From eb60ceac0ebd11e9896ed1ca84fcad71a4b23601 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Fri, 2 Feb 2007 09:18:22 -0500
Subject: Btrfs: Add backing store, memory management

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/disk-io.h | 21 +++++++++++++++++++++
 1 file changed, 21 insertions(+)
 create mode 100644 fs/btrfs/disk-io.h

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
new file mode 100644
index 000000000000..ee95fa05c4a3
--- /dev/null
+++ b/fs/btrfs/disk-io.h
@@ -0,0 +1,21 @@
+#ifndef __DISKIO__
+#define __DISKIO__
+
+struct tree_buffer {
+	u64 blocknr;
+	int count;
+	union {
+		struct node node;
+		struct leaf leaf;
+	};
+};
+
+struct tree_buffer *read_tree_block(struct ctree_root *root, u64 blocknr);
+int write_tree_block(struct ctree_root *root, struct tree_buffer *buf);
+struct ctree_root *open_ctree(char *filename);
+int close_ctree(struct ctree_root *root);
+void tree_block_release(struct ctree_root *root, struct tree_buffer *buf);
+struct tree_buffer *alloc_free_block(struct ctree_root *root);
+int update_root_block(struct ctree_root *root);
+
+#endif
-- 
cgit v1.2.3


From d97e63b69ef21c02b67e20e41d9968b0e503572e Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Tue, 20 Feb 2007 16:40:44 -0500
Subject: Btrfs: early extent mapping support

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/Makefile  |   4 +-
 fs/btrfs/ctree.c   | 152 +++++++++++++++++++++++++++++++++++++++++++++++++++--
 fs/btrfs/ctree.h   |  34 ++++++++++--
 fs/btrfs/disk-io.c |  76 +++++++++++++++++++--------
 fs/btrfs/disk-io.h |   3 ++
 5 files changed, 236 insertions(+), 33 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 63360212a075..df065dd2dce7 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -4,8 +4,8 @@ CFLAGS= -g -Wall
 .c.o:
 	$(CC) $(CFLAGS) -c $<
 
-ctree: ctree.o disk-io.h ctree.h disk-io.o radix-tree.o radix-tree.h
-	gcc $(CFLAGS) -o ctree ctree.o disk-io.o radix-tree.o
+ctree: ctree.o disk-io.h ctree.h disk-io.o radix-tree.o radix-tree.h mkfs.o
+	gcc $(CFLAGS) -o ctree ctree.o disk-io.o radix-tree.o mkfs.o
 
 clean:
 	rm ctree *.o
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 6b64f49a0279..2177744dedd3 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -5,6 +5,8 @@
 #include "ctree.h"
 #include "disk-io.h"
 
+static int refill_alloc_extent(struct ctree_root *root);
+
 static inline void init_path(struct ctree_path *p)
 {
 	memset(p, 0, sizeof(*p));
@@ -29,7 +31,7 @@ static inline unsigned int leaf_data_end(struct leaf *leaf)
 {
 	unsigned int nr = leaf->header.nritems;
 	if (nr == 0)
-		return ARRAY_SIZE(leaf->data);
+		return sizeof(leaf->data);
 	return leaf->items[nr-1].offset;
 }
 
@@ -421,7 +423,7 @@ int insert_ptr(struct ctree_root *root,
 	 * due to splitting.  Once we've done all the splitting required
 	 * do the inserts based on the data in the bal array.
 	 */
-	memset(bal, 0, ARRAY_SIZE(bal));
+	memset(bal, 0, sizeof(bal));
 	while(t && t->node.header.nritems == NODEPTRS_PER_BLOCK) {
 		c = &t->node;
 		if (push_node_left(root, path,
@@ -756,6 +758,7 @@ int insert_item(struct ctree_root *root, struct key *key,
 	if (leaf_free_space(leaf) < 0)
 		BUG();
 	release_path(root, &path);
+	refill_alloc_extent(root);
 	return 0;
 }
 
@@ -884,6 +887,135 @@ int del_item(struct ctree_root *root, struct ctree_path *path)
 	return 0;
 }
 
+int next_leaf(struct ctree_root *root, struct ctree_path *path)
+{
+	int slot;
+	int level = 1;
+	u64 blocknr;
+	struct tree_buffer *c;
+	struct tree_buffer *next;
+
+	while(level < MAX_LEVEL) {
+		if (!path->nodes[level])
+			return -1;
+		slot = path->slots[level] + 1;
+		c = path->nodes[level];
+		if (slot >= c->node.header.nritems) {
+			level++;
+			continue;
+		}
+		blocknr = c->node.blockptrs[slot];
+		next = read_tree_block(root, blocknr);
+		break;
+	}
+	path->slots[level] = slot;
+	while(1) {
+		level--;
+		c = path->nodes[level];
+		tree_block_release(root, c);
+		path->nodes[level] = next;
+		path->slots[level] = 0;
+		if (!level)
+			break;
+		next = read_tree_block(root, next->node.blockptrs[0]);
+	}
+	return 0;
+}
+
+int alloc_extent(struct ctree_root *root, u64 num_blocks, u64 search_start,
+		 u64 search_end, u64 owner, struct key *ins)
+{
+	struct ctree_path path;
+	struct key *key;
+	int ret;
+	u64 hole_size = 0;
+	int slot = 0;
+	u64 last_block;
+	int start_found = 0;
+	struct leaf *l;
+	struct extent_item extent_item;
+
+	init_path(&path);
+	ins->objectid = search_start;
+	ins->offset = 0;
+	ins->flags = 0;
+
+	ret = search_slot(root, ins, &path);
+	while (1) {
+		l = &path.nodes[0]->leaf;
+		slot = path.slots[0];
+		if (!l) {
+			// FIXME allocate root
+		}
+		if (slot >= l->header.nritems) {
+			ret = next_leaf(root, &path);
+			if (ret == 0)
+				continue;
+			if (!start_found) {
+				ins->objectid = search_start;
+				ins->offset = num_blocks;
+				hole_size = search_end - search_start;
+				goto insert;
+			}
+			ins->objectid = last_block;
+			ins->offset = num_blocks;
+			hole_size = search_end - last_block;
+			goto insert;
+		}
+		key = &l->items[slot].key;
+		if (start_found) {
+			hole_size = key->objectid - last_block;
+			if (hole_size > num_blocks) {
+				ins->objectid = last_block;
+				ins->offset = num_blocks;
+				goto insert;
+			}
+		} else
+			start_found = 1;
+		last_block = key->objectid + key->offset;
+		path.slots[0]++;
+		printf("last block is not %lu\n", last_block);
+	}
+	// FIXME -ENOSPC
+insert:
+	extent_item.refs = 1;
+	extent_item.owner = owner;
+	ret = insert_item(root, ins, &extent_item, sizeof(extent_item));
+	return ret;
+}
+
+static int refill_alloc_extent(struct ctree_root *root)
+{
+	struct alloc_extent *ae = root->alloc_extent;
+	struct key key;
+	int ret;
+	int min_blocks = MAX_LEVEL * 2;
+
+	printf("refill alloc root %p, numused %lu total %lu\n", root, ae->num_used, ae->num_blocks);
+	if (ae->num_blocks > ae->num_used && ae->num_blocks - ae->num_used >
+	    min_blocks)
+		return 0;
+	ae = root->reserve_extent;
+	if (ae->num_blocks > ae->num_used) {
+		if (root->alloc_extent->num_blocks == 0) {
+			/* we should swap reserve/alloc_extent when alloc
+			 * fills up
+			 */
+			BUG();
+		}
+		if (ae->num_blocks - ae->num_used < min_blocks)
+			BUG();
+		return 0;
+	}
+	// FIXME, this recurses
+	ret = alloc_extent(root->extent_root,
+			   min_blocks * 2, 0, (unsigned long)-1, 0, &key);
+	ae->blocknr = key.objectid;
+	ae->num_blocks = key.offset;
+	ae->num_used = 0;
+	return ret;
+}
+
 void print_leaf(struct leaf *l)
 {
 	int i;
@@ -948,8 +1080,8 @@ void print_tree(struct ctree_root *root, struct tree_buffer *t)
 
 /* for testing only */
 int next_key(int i, int max_key) {
-	return rand() % max_key;
-	// return i;
+	// return rand() % max_key;
+	return i;
 }
 
 int main() {
@@ -960,7 +1092,7 @@ int main() {
 	int i;
 	int num;
 	int ret;
-	int run_size = 25000;
+	int run_size = 256;
 	int max_key = 100000000;
 	int tree_size = 0;
 	struct ctree_path path;
@@ -980,10 +1112,20 @@ int main() {
 		ins.objectid = num;
 		ins.offset = 0;
 		ins.flags = 0;
+		printf("insert %d\n", i);
 		ret = insert_item(root, &ins, buf, strlen(buf));
 		if (!ret)
 			tree_size++;
+		printf("done insert %d\n", i);
 	}
+	printf("root used: %lu\n", root->alloc_extent->num_used);
+	printf("root tree\n");
+	print_tree(root, root->node);
+	printf("map tree\n");
+	printf("map used: %lu\n", root->extent_root->alloc_extent->num_used);
+	print_tree(root->extent_root, root->extent_root->node);
+	exit(1);
+
 	close_ctree(root);
 	root = open_ctree("dbfile");
 	printf("starting search\n");
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 586bf1866042..b737925be314 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1,7 +1,7 @@
 #ifndef __CTREE__
 #define __CTREE__
 
-#define CTREE_BLOCKSIZE 4096
+#define CTREE_BLOCKSIZE 256
 
 struct key {
 	u64 objectid;
@@ -22,18 +22,41 @@ struct header {
 #define NODEPTRS_PER_BLOCK ((CTREE_BLOCKSIZE - sizeof(struct header)) / \
 			    (sizeof(struct key) + sizeof(u64)))
 
-#define LEVEL_BITS 3
-#define MAX_LEVEL (1 << LEVEL_BITS)
+#define MAX_LEVEL 8
 #define node_level(f) ((f) & (MAX_LEVEL-1))
 #define is_leaf(f) (node_level(f) == 0)
 
 struct tree_buffer;
+
+struct alloc_extent {
+	u64 blocknr;
+	u64 num_blocks;
+	u64 num_used;
+} __attribute__ ((__packed__));
+
 struct ctree_root {
 	struct tree_buffer *node;
+	struct ctree_root *extent_root;
+	struct alloc_extent *alloc_extent;
+	struct alloc_extent *reserve_extent;
 	int fp;
 	struct radix_tree_root cache_radix;
+	struct alloc_extent ai1;
+	struct alloc_extent ai2;
 };
 
+struct ctree_root_info {
+	u64 fsid[2]; /* FS specific uuid */
+	u64 blocknr; /* blocknr of this block */
+	u64 objectid; /* inode number of this root */
+	u64 tree_root; /* the tree root */
+	u32 csum;
+	u32 ham;
+	struct alloc_extent alloc_extent;
+	struct alloc_extent reserve_extent;
+	u64 snapuuid[2]; /* root specific uuid */
+} __attribute__ ((__packed__));
+
 struct item {
 	struct key key;
 	u16 offset;
@@ -55,6 +78,11 @@ struct node {
 	u64 blockptrs[NODEPTRS_PER_BLOCK];
 } __attribute__ ((__packed__));
 
+struct extent_item {
+	u32 refs;
+	u64 owner;
+} __attribute__ ((__packed__));
+
 struct ctree_path {
 	struct tree_buffer *nodes[MAX_LEVEL];
 	int slots[MAX_LEVEL];
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 8d51a07051d9..653f18aab330 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -12,25 +12,27 @@
 
 static int allocated_blocks = 0;
 
-struct ctree_header {
-	u64 root_block;
-} __attribute__ ((__packed__));
-
 static int get_free_block(struct ctree_root *root, u64 *block)
 {
 	struct stat st;
 	int ret;
 
+	if (root->alloc_extent->num_used >= root->alloc_extent->num_blocks)
+		return -1;
+
+	*block = root->alloc_extent->blocknr + root->alloc_extent->num_used;
+	root->alloc_extent->num_used += 1;
+	if (root->alloc_extent->num_used >= root->alloc_extent->num_blocks) {
+		struct alloc_extent *ae = root->alloc_extent;
+		root->alloc_extent = root->reserve_extent;
+		root->reserve_extent = ae;
+		ae->num_blocks = 0;
+	}
 	st.st_size = 0;
 	ret = fstat(root->fp, &st);
-	if (st.st_size > sizeof(struct ctree_header)) {
-		*block = (st.st_size -
-			sizeof(struct ctree_header)) / CTREE_BLOCKSIZE;
-	} else {
-		*block = 0;
-	}
-	ret = ftruncate(root->fp, sizeof(struct ctree_header) + (*block + 1) *
-			CTREE_BLOCKSIZE);
+	if (st.st_size < (*block + 1) * CTREE_BLOCKSIZE)
+		ret = ftruncate(root->fp,
+				(*block + 1) * CTREE_BLOCKSIZE);
 	return ret;
 }
 
@@ -72,7 +74,7 @@ struct tree_buffer *alloc_free_block(struct ctree_root *root)
 
 struct tree_buffer *read_tree_block(struct ctree_root *root, u64 blocknr)
 {
-	loff_t offset = blocknr * CTREE_BLOCKSIZE + sizeof(struct ctree_header);
+	loff_t offset = blocknr * CTREE_BLOCKSIZE;
 	struct tree_buffer *buf;
 	int ret;
 
@@ -101,7 +103,7 @@ struct tree_buffer *read_tree_block(struct ctree_root *root, u64 blocknr)
 int write_tree_block(struct ctree_root *root, struct tree_buffer *buf)
 {
 	u64 blocknr = buf->blocknr;
-	loff_t offset = blocknr * CTREE_BLOCKSIZE + sizeof(struct ctree_header);
+	loff_t offset = blocknr * CTREE_BLOCKSIZE;
 	int ret;
 
 	if (buf->blocknr != buf->node.header.blocknr)
@@ -114,11 +116,32 @@ int write_tree_block(struct ctree_root *root, struct tree_buffer *buf)
 	return 0;
 }
 
+struct ctree_super_block {
+	struct ctree_root_info root_info;
+	struct ctree_root_info extent_info;
+} __attribute__ ((__packed__));
+
+static int __setup_root(struct ctree_root *root, struct ctree_root *extent_root,
+			struct ctree_root_info *info, int fp)
+{
+	root->fp = fp;
+	root->node = read_tree_block(root, info->tree_root);
+	root->extent_root = extent_root;
+	memcpy(&root->ai1, &info->alloc_extent, sizeof(info->alloc_extent));
+	memcpy(&root->ai2, &info->reserve_extent, sizeof(info->reserve_extent));
+	root->alloc_extent = &root->ai1;
+	root->reserve_extent = &root->ai2;
+	INIT_RADIX_TREE(&root->cache_radix, GFP_KERNEL);
+	printf("setup done reading root %p, used %lu\n", root, root->alloc_extent->num_used);
+	return 0;
+}
+
 struct ctree_root *open_ctree(char *filename)
 {
 	struct ctree_root *root = malloc(sizeof(struct ctree_root));
+	struct ctree_root *extent_root = malloc(sizeof(struct ctree_root));
+	struct ctree_super_block super;
 	int fp;
-	u64 root_block;
 	int ret;
 
 	fp = open(filename, O_CREAT | O_RDWR);
@@ -126,14 +149,20 @@ struct ctree_root *open_ctree(char *filename)
 		free(root);
 		return NULL;
 	}
-	root->fp = fp;
-	INIT_RADIX_TREE(&root->cache_radix, GFP_KERNEL);
-	ret = pread(fp, &root_block, sizeof(u64), 0);
-	if (ret == sizeof(u64)) {
-		printf("reading root node at block %lu\n", root_block);
-		root->node = read_tree_block(root, root_block);
-	} else
-		root->node = NULL;
+	ret = pread(fp, &super, sizeof(struct ctree_super_block),
+		     CTREE_SUPER_INFO_OFFSET(CTREE_BLOCKSIZE));
+	if (ret == 0) {
+		ret = mkfs(fp);
+		if (ret)
+			return NULL;
+		ret = pread(fp, &super, sizeof(struct ctree_super_block),
+			     CTREE_SUPER_INFO_OFFSET(CTREE_BLOCKSIZE));
+		if (ret != sizeof(struct ctree_super_block))
+			return NULL;
+	}
+	BUG_ON(ret < 0);
+	__setup_root(root, extent_root, &super.root_info, fp);
+	__setup_root(extent_root, extent_root, &super.extent_info, fp);
 	return root;
 }
 
@@ -160,6 +189,7 @@ int update_root_block(struct ctree_root *root)
 
 void tree_block_release(struct ctree_root *root, struct tree_buffer *buf)
 {
+	return;
 	buf->count--;
 	if (buf->count == 0) {
 		if (!radix_tree_lookup(&root->cache_radix, buf->blocknr))
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index ee95fa05c4a3..41acb65eee19 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -17,5 +17,8 @@ int close_ctree(struct ctree_root *root);
 void tree_block_release(struct ctree_root *root, struct tree_buffer *buf);
 struct tree_buffer *alloc_free_block(struct ctree_root *root);
 int update_root_block(struct ctree_root *root);
+int mkfs(int fd);
+
+#define CTREE_SUPER_INFO_OFFSET(bs) (16 * (bs))
 
 #endif
-- 
cgit v1.2.3


From cfaa72952fa7b44aa5d967cbc266110900552aef Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Wed, 21 Feb 2007 17:04:57 -0500
Subject: Btrfs: extent fixes

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.c   | 69 +++++++++++++++++++++++++++++++----------------
 fs/btrfs/ctree.h   |  5 ++++
 fs/btrfs/disk-io.c | 79 +++++++++++++++++++++++++++++++-----------------------
 fs/btrfs/disk-io.h |  4 +--
 fs/btrfs/mkfs.c    |  7 +++--
 5 files changed, 103 insertions(+), 61 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 2177744dedd3..2891b582e26f 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -354,6 +354,7 @@ int __insert_ptr(struct ctree_root *root,
 		c->header.nritems = 2;
 		c->header.flags = node_level(level);
 		c->header.blocknr = t->blocknr;
+		c->header.parentid = root->node->node.header.parentid;
 		lower = &path->nodes[level-1]->node;
 		if (is_leaf(lower->header.flags))
 			lower_key = &((struct leaf *)lower)->items[0].key;
@@ -363,7 +364,7 @@ int __insert_ptr(struct ctree_root *root,
 		memcpy(c->keys + 1, key, sizeof(struct key));
 		c->blockptrs[0] = path->nodes[level-1]->blocknr;
 		c->blockptrs[1] = blocknr;
-		/* the path has an extra ref to root->node */
+		/* the super has an extra ref to root->node */
 		tree_block_release(root, root->node);
 		root->node = t;
 		t->count++;
@@ -439,6 +440,7 @@ int insert_ptr(struct ctree_root *root,
 		b = &b_buffer->node;
 		b->header.flags = c->header.flags;
 		b->header.blocknr = b_buffer->blocknr;
+		b->header.parentid = root->node->node.header.parentid;
 		mid = (c->header.nritems + 1) / 2;
 		memcpy(b->keys, c->keys + mid,
 			(c->header.nritems - mid) * sizeof(struct key));
@@ -642,6 +644,7 @@ int split_leaf(struct ctree_root *root, struct ctree_path *path, int data_size)
 	right->header.nritems = nritems - mid;
 	right->header.blocknr = right_buffer->blocknr;
 	right->header.flags = node_level(0);
+	right->header.parentid = root->node->node.header.parentid;
 	data_copy_size = l->items[mid].offset + l->items[mid].size -
 			 leaf_data_end(l);
 	memcpy(right->items, l->items + mid,
@@ -689,8 +692,12 @@ int insert_item(struct ctree_root *root, struct key *key,
 	unsigned int data_end;
 	struct ctree_path path;
 
+	refill_alloc_extent(root);
+
 	/* create a root if there isn't one */
 	if (!root->node) {
+		BUG();
+#if 0
 		struct tree_buffer *t;
 		t = alloc_free_block(root);
 		BUG_ON(!t);
@@ -699,6 +706,7 @@ int insert_item(struct ctree_root *root, struct key *key,
 		t->node.header.blocknr = t->blocknr;
 		root->node = t;
 		write_tree_block(root, t);
+#endif
 	}
 	init_path(&path);
 	ret = search_slot(root, key, &path);
@@ -758,7 +766,6 @@ int insert_item(struct ctree_root *root, struct key *key,
 	if (leaf_free_space(leaf) < 0)
 		BUG();
 	release_path(root, &path);
-	refill_alloc_extent(root);
 	return 0;
 }
 
@@ -893,7 +900,7 @@ int next_leaf(struct ctree_root *root, struct ctree_path *path)
 	int level = 1;
 	u64 blocknr;
 	struct tree_buffer *c;
-	struct tree_buffer *next;
+	struct tree_buffer *next = NULL;
 
 	while(level < MAX_LEVEL) {
 		if (!path->nodes[level])
@@ -905,6 +912,8 @@ int next_leaf(struct ctree_root *root, struct ctree_path *path)
 			continue;
 		}
 		blocknr = c->node.blockptrs[slot];
+		if (next)
+			tree_block_release(root, next);
 		next = read_tree_block(root, blocknr);
 		break;
 	}
@@ -922,7 +931,7 @@ int next_leaf(struct ctree_root *root, struct ctree_path *path)
 	return 0;
 }
 
-int alloc_extent(struct ctree_root *root, u64 num_blocks, u64 search_start,
+int alloc_extent(struct ctree_root *orig_root, u64 num_blocks, u64 search_start,
 		 u64 search_end, u64 owner, struct key *ins)
 {
 	struct ctree_path path;
@@ -934,6 +943,7 @@ int alloc_extent(struct ctree_root *root, u64 num_blocks, u64 search_start,
 	int start_found = 0;
 	struct leaf *l;
 	struct extent_item extent_item;
+	struct ctree_root * root = orig_root->extent_root;
 
 	init_path(&path);
 	ins->objectid = search_start;
@@ -974,13 +984,18 @@ int alloc_extent(struct ctree_root *root, u64 num_blocks, u64 search_start,
 			start_found = 1;
 		last_block = key->objectid + key->offset;
 		path.slots[0]++;
-		printf("last block is not %lu\n", last_block);
 	}
 	// FIXME -ENOSPC
 insert:
+	release_path(root, &path);
 	extent_item.refs = 1;
 	extent_item.owner = owner;
-	ret = insert_item(root, ins, &extent_item, sizeof(extent_item));
+	if (root == orig_root && root->reserve_extent->num_blocks == 0) {
+		root->reserve_extent->blocknr = ins->objectid;
+		root->reserve_extent->num_blocks = ins->offset;
+		root->reserve_extent->num_used = 0;
+	}
+	ret = insert_item(root->extent_root, ins, &extent_item, sizeof(extent_item));
 	return ret;
 }
 
@@ -991,7 +1006,6 @@ static int refill_alloc_extent(struct ctree_root *root)
 	int ret;
 	int min_blocks = MAX_LEVEL * 2;
 
-	printf("refill alloc root %p, numused %lu total %lu\n", root, ae->num_used, ae->num_blocks);
 	if (ae->num_blocks > ae->num_used && ae->num_blocks - ae->num_used >
 	    min_blocks)
 		return 0;
@@ -1007,9 +1021,9 @@ static int refill_alloc_extent(struct ctree_root *root)
 			BUG();
 		return 0;
 	}
-	// FIXME, this recurses
-	ret = alloc_extent(root->extent_root,
-			   min_blocks * 2, 0, (unsigned long)-1, 0, &key);
+	ret = alloc_extent(root,
+			   min_blocks * 2, 0, (unsigned long)-1,
+			   root->node->node.header.parentid, &key);
 	ae->blocknr = key.objectid;
 	ae->num_blocks = key.offset;
 	ae->num_used = 0;
@@ -1021,6 +1035,7 @@ void print_leaf(struct leaf *l)
 	int i;
 	int nr = l->header.nritems;
 	struct item *item;
+	struct extent_item *ei;
 	printf("leaf %lu total ptrs %d free space %d\n", l->header.blocknr, nr,
 	       leaf_free_space(l));
 	fflush(stdout);
@@ -1032,6 +1047,8 @@ void print_leaf(struct leaf *l)
 			item->offset, item->size);
 		fflush(stdout);
 		printf("\t\titem data %.*s\n", item->size, l->data+item->offset);
+		ei = (struct extent_item *)(l->data + item->offset);
+		printf("\t\textent data %u %lu\n", ei->refs, ei->owner);
 		fflush(stdout);
 	}
 }
@@ -1080,8 +1097,8 @@ void print_tree(struct ctree_root *root, struct tree_buffer *t)
 
 /* for testing only */
 int next_key(int i, int max_key) {
-	// return rand() % max_key;
-	return i;
+	return rand() % max_key;
+	// return i;
 }
 
 int main() {
@@ -1092,15 +1109,20 @@ int main() {
 	int i;
 	int num;
 	int ret;
-	int run_size = 256;
+	int run_size = 10000;
 	int max_key = 100000000;
 	int tree_size = 0;
 	struct ctree_path path;
+	struct ctree_super_block super;
 
 	radix_tree_init();
 
 
-	root = open_ctree("dbfile");
+	root = open_ctree("dbfile", &super);
+	printf("root tree\n");
+	print_tree(root, root->node);
+	printf("map tree\n");
+	print_tree(root->extent_root, root->extent_root->node);
 
 	srand(55);
 	for (i = 0; i < run_size; i++) {
@@ -1112,22 +1134,20 @@ int main() {
 		ins.objectid = num;
 		ins.offset = 0;
 		ins.flags = 0;
-		printf("insert %d\n", i);
 		ret = insert_item(root, &ins, buf, strlen(buf));
 		if (!ret)
 			tree_size++;
-		printf("done insert %d\n", i);
 	}
 	printf("root used: %lu\n", root->alloc_extent->num_used);
 	printf("root tree\n");
-	print_tree(root, root->node);
+	// print_tree(root, root->node);
 	printf("map tree\n");
 	printf("map used: %lu\n", root->extent_root->alloc_extent->num_used);
-	print_tree(root->extent_root, root->extent_root->node);
-	exit(1);
-
+	// print_tree(root->extent_root, root->extent_root->node);
+	write_ctree_super(root, &super);
 	close_ctree(root);
-	root = open_ctree("dbfile");
+
+	root = open_ctree("dbfile", &super);
 	printf("starting search\n");
 	srand(55);
 	for (i = 0; i < run_size; i++) {
@@ -1142,8 +1162,9 @@ int main() {
 		}
 		release_path(root, &path);
 	}
+	write_ctree_super(root, &super);
 	close_ctree(root);
-	root = open_ctree("dbfile");
+	root = open_ctree("dbfile", &super);
 	printf("node %p level %d total ptrs %d free spc %lu\n", root->node,
 	        node_level(root->node->node.header.flags),
 		root->node->node.header.nritems,
@@ -1174,8 +1195,9 @@ int main() {
 		if (!ret)
 			tree_size++;
 	}
+	write_ctree_super(root, &super);
 	close_ctree(root);
-	root = open_ctree("dbfile");
+	root = open_ctree("dbfile", &super);
 	printf("starting search2\n");
 	srand(128);
 	for (i = 0; i < run_size; i++) {
@@ -1221,6 +1243,7 @@ int main() {
 		}
 		release_path(root, &path);
 	}
+	write_ctree_super(root, &super);
 	close_ctree(root);
 	printf("tree size is now %d\n", tree_size);
 	return 0;
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index b737925be314..78407d32db78 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -57,6 +57,11 @@ struct ctree_root_info {
 	u64 snapuuid[2]; /* root specific uuid */
 } __attribute__ ((__packed__));
 
+struct ctree_super_block {
+	struct ctree_root_info root_info;
+	struct ctree_root_info extent_info;
+} __attribute__ ((__packed__));
+
 struct item {
 	struct key key;
 	u16 offset;
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 653f18aab330..2f71ccf1c3cb 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -15,7 +15,7 @@ static int allocated_blocks = 0;
 static int get_free_block(struct ctree_root *root, u64 *block)
 {
 	struct stat st;
-	int ret;
+	int ret = 0;
 
 	if (root->alloc_extent->num_used >= root->alloc_extent->num_blocks)
 		return -1;
@@ -30,9 +30,14 @@ static int get_free_block(struct ctree_root *root, u64 *block)
 	}
 	st.st_size = 0;
 	ret = fstat(root->fp, &st);
-	if (st.st_size < (*block + 1) * CTREE_BLOCKSIZE)
+	if (st.st_size < (*block + 1) * CTREE_BLOCKSIZE) {
 		ret = ftruncate(root->fp,
 				(*block + 1) * CTREE_BLOCKSIZE);
+		if (ret) {
+			perror("ftruncate");
+			exit(1);
+		}
+	}
 	return ret;
 }
 
@@ -81,11 +86,7 @@ struct tree_buffer *read_tree_block(struct ctree_root *root, u64 blocknr)
 	buf = radix_tree_lookup(&root->cache_radix, blocknr);
 	if (buf) {
 		buf->count++;
-		if (buf->blocknr != blocknr)
-			BUG();
-		if (buf->blocknr != buf->node.header.blocknr)
-			BUG();
-		return buf;
+		goto test;
 	}
 	buf = alloc_tree_block(root, blocknr);
 	if (!buf)
@@ -95,8 +96,11 @@ struct tree_buffer *read_tree_block(struct ctree_root *root, u64 blocknr)
 		free(buf);
 		return NULL;
 	}
+test:
 	if (buf->blocknr != buf->node.header.blocknr)
 		BUG();
+	if (root->node && buf->node.header.parentid != root->node->node.header.parentid)
+		BUG();
 	return buf;
 }
 
@@ -111,36 +115,30 @@ int write_tree_block(struct ctree_root *root, struct tree_buffer *buf)
 	ret = pwrite(root->fp, &buf->node, CTREE_BLOCKSIZE, offset);
 	if (ret != CTREE_BLOCKSIZE)
 		return ret;
-	if (buf == root->node)
-		return update_root_block(root);
 	return 0;
 }
 
-struct ctree_super_block {
-	struct ctree_root_info root_info;
-	struct ctree_root_info extent_info;
-} __attribute__ ((__packed__));
-
 static int __setup_root(struct ctree_root *root, struct ctree_root *extent_root,
 			struct ctree_root_info *info, int fp)
 {
+	INIT_RADIX_TREE(&root->cache_radix, GFP_KERNEL);
 	root->fp = fp;
+	root->node = NULL;
 	root->node = read_tree_block(root, info->tree_root);
 	root->extent_root = extent_root;
 	memcpy(&root->ai1, &info->alloc_extent, sizeof(info->alloc_extent));
 	memcpy(&root->ai2, &info->reserve_extent, sizeof(info->reserve_extent));
 	root->alloc_extent = &root->ai1;
 	root->reserve_extent = &root->ai2;
-	INIT_RADIX_TREE(&root->cache_radix, GFP_KERNEL);
-	printf("setup done reading root %p, used %lu\n", root, root->alloc_extent->num_used);
+	printf("setup done reading root %p, used %lu available %lu\n", root, root->alloc_extent->num_used, root->alloc_extent->num_blocks);
+	printf("setup done reading root %p, reserve used %lu available %lu\n", root, root->reserve_extent->num_used, root->reserve_extent->num_blocks);
 	return 0;
 }
 
-struct ctree_root *open_ctree(char *filename)
+struct ctree_root *open_ctree(char *filename, struct ctree_super_block *super)
 {
 	struct ctree_root *root = malloc(sizeof(struct ctree_root));
 	struct ctree_root *extent_root = malloc(sizeof(struct ctree_root));
-	struct ctree_super_block super;
 	int fp;
 	int ret;
 
@@ -149,48 +147,61 @@ struct ctree_root *open_ctree(char *filename)
 		free(root);
 		return NULL;
 	}
-	ret = pread(fp, &super, sizeof(struct ctree_super_block),
+	ret = pread(fp, super, sizeof(struct ctree_super_block),
 		     CTREE_SUPER_INFO_OFFSET(CTREE_BLOCKSIZE));
 	if (ret == 0) {
 		ret = mkfs(fp);
 		if (ret)
 			return NULL;
-		ret = pread(fp, &super, sizeof(struct ctree_super_block),
+		ret = pread(fp, super, sizeof(struct ctree_super_block),
 			     CTREE_SUPER_INFO_OFFSET(CTREE_BLOCKSIZE));
 		if (ret != sizeof(struct ctree_super_block))
 			return NULL;
 	}
 	BUG_ON(ret < 0);
-	__setup_root(root, extent_root, &super.root_info, fp);
-	__setup_root(extent_root, extent_root, &super.extent_info, fp);
+	__setup_root(root, extent_root, &super->root_info, fp);
+	__setup_root(extent_root, extent_root, &super->extent_info, fp);
 	return root;
 }
 
-int close_ctree(struct ctree_root *root)
+static int __update_root(struct ctree_root *root, struct ctree_root_info *info)
 {
-	close(root->fp);
-	if (root->node)
-		tree_block_release(root, root->node);
-	free(root);
-	printf("on close %d blocks are allocated\n", allocated_blocks);
+	info->tree_root = root->node->blocknr;
+	memcpy(&info->alloc_extent, root->alloc_extent, sizeof(struct alloc_extent));
+	memcpy(&info->reserve_extent, root->reserve_extent, sizeof(struct alloc_extent));
 	return 0;
 }
 
-int update_root_block(struct ctree_root *root)
+int write_ctree_super(struct ctree_root *root, struct ctree_super_block *s)
 {
 	int ret;
-	u64 root_block = root->node->blocknr;
-
-	ret = pwrite(root->fp, &root_block, sizeof(u64), 0);
-	if (ret != sizeof(u64))
+	__update_root(root, &s->root_info);
+	__update_root(root->extent_root, &s->extent_info);
+	ret = pwrite(root->fp, s, sizeof(*s), CTREE_SUPER_INFO_OFFSET(CTREE_BLOCKSIZE));
+	if (ret != sizeof(*s)) {
+		fprintf(stderr, "failed to write new super block err %d\n", ret);
 		return ret;
+	}
+	return 0;
+}
+
+int close_ctree(struct ctree_root *root)
+{
+	close(root->fp);
+	if (root->node)
+		tree_block_release(root, root->node);
+	if (root->extent_root->node)
+		tree_block_release(root->extent_root, root->extent_root->node);
+	free(root);
+	printf("on close %d blocks are allocated\n", allocated_blocks);
 	return 0;
 }
 
 void tree_block_release(struct ctree_root *root, struct tree_buffer *buf)
 {
-	return;
 	buf->count--;
+	if (buf->count < 0)
+		BUG();
 	if (buf->count == 0) {
 		if (!radix_tree_lookup(&root->cache_radix, buf->blocknr))
 			BUG();
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 41acb65eee19..e288fe883b9a 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -12,11 +12,11 @@ struct tree_buffer {
 
 struct tree_buffer *read_tree_block(struct ctree_root *root, u64 blocknr);
 int write_tree_block(struct ctree_root *root, struct tree_buffer *buf);
-struct ctree_root *open_ctree(char *filename);
+struct ctree_root *open_ctree(char *filename, struct ctree_super_block *s);
 int close_ctree(struct ctree_root *root);
 void tree_block_release(struct ctree_root *root, struct tree_buffer *buf);
 struct tree_buffer *alloc_free_block(struct ctree_root *root);
-int update_root_block(struct ctree_root *root);
+int write_ctree_super(struct ctree_root *root, struct ctree_super_block *s);
 int mkfs(int fd);
 
 #define CTREE_SUPER_INFO_OFFSET(bs) (16 * (bs))
diff --git a/fs/btrfs/mkfs.c b/fs/btrfs/mkfs.c
index bab98c6d6fa7..584aba44e801 100644
--- a/fs/btrfs/mkfs.c
+++ b/fs/btrfs/mkfs.c
@@ -18,12 +18,13 @@ int mkfs(int fd)
 	struct extent_item extent_item;
 	int ret;
 
+	/* setup the super block area */
 	memset(info, 0, sizeof(info));
 	info[0].blocknr = 16;
 	info[0].objectid = 1;
 	info[0].tree_root = 17;
 	info[0].alloc_extent.blocknr = 0;
-	info[0].alloc_extent.num_blocks = 20;
+	info[0].alloc_extent.num_blocks = 64;
 	/* 0-17 are used (inclusive) */
 	info[0].alloc_extent.num_used = 18;
 
@@ -31,12 +32,14 @@ int mkfs(int fd)
 	info[1].objectid = 2;
 	info[1].tree_root = 64;
 	info[1].alloc_extent.blocknr = 64;
-	info[1].alloc_extent.num_blocks = 8;
+	info[1].alloc_extent.num_blocks = 64;
 	info[1].alloc_extent.num_used = 1;
 	ret = pwrite(fd, info, sizeof(info),
 		     CTREE_SUPER_INFO_OFFSET(CTREE_BLOCKSIZE));
 	if (ret != sizeof(info))
 		return -1;
+
+	/* create leaves for the tree root and extent root */
 	memset(&empty_leaf, 0, sizeof(empty_leaf));
 	empty_leaf.header.parentid = 1;
 	empty_leaf.header.blocknr = 17;
-- 
cgit v1.2.3


From 9a8dd1502de6aa683ae46cf0397e9b6e636416fb Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Fri, 23 Feb 2007 08:38:36 -0500
Subject: Btrfs: Block sized tree extents and extent deletion

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.c   | 245 +++++++++++++++++++++++++++++++++++++++++++----------
 fs/btrfs/ctree.h   |  13 +--
 fs/btrfs/disk-io.c |  90 +++++++-------------
 fs/btrfs/disk-io.h |   2 +-
 fs/btrfs/mkfs.c    |  37 ++++----
 5 files changed, 252 insertions(+), 135 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 1b4e82d8074d..f0abcf1f3939 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -8,9 +8,12 @@
 #define SEARCH_READ 0
 #define SEARCH_WRITE 1
 
-static int refill_alloc_extent(struct ctree_root *root);
+#define CTREE_EXTENT_PENDING 0
+
 int split_node(struct ctree_root *root, struct ctree_path *path, int level);
 int split_leaf(struct ctree_root *root, struct ctree_path *path, int data_size);
+struct tree_buffer *alloc_free_block(struct ctree_root *root);
+int free_extent(struct ctree_root *root, u64 blocknr, u64 num_blocks);
 
 static inline void init_path(struct ctree_path *p)
 {
@@ -682,8 +685,6 @@ int insert_item(struct ctree_root *root, struct key *key,
 	unsigned int data_end;
 	struct ctree_path path;
 
-	refill_alloc_extent(root);
-
 	/* create a root if there isn't one */
 	if (!root->node)
 		BUG();
@@ -756,6 +757,7 @@ int del_ptr(struct ctree_root *root, struct ctree_path *path, int level)
 	struct tree_buffer *t;
 	struct node *node;
 	int nritems;
+	u64 blocknr;
 
 	while(1) {
 		t = path->nodes[level];
@@ -774,6 +776,7 @@ int del_ptr(struct ctree_root *root, struct ctree_path *path, int level)
 		}
 		node->header.nritems--;
 		write_tree_block(root, t);
+		blocknr = t->blocknr;
 		if (node->header.nritems != 0) {
 			int tslot;
 			if (slot == 0)
@@ -799,6 +802,7 @@ int del_ptr(struct ctree_root *root, struct ctree_path *path, int level)
 			break;
 		}
 		level++;
+		free_extent(root, blocknr, 1);
 		if (!path->nodes[level])
 			BUG();
 	}
@@ -841,8 +845,10 @@ int del_item(struct ctree_root *root, struct ctree_path *path)
 		if (leaf_buf == root->node) {
 			leaf->header.flags = node_level(0);
 			write_tree_block(root, leaf_buf);
-		} else
+		} else {
 			del_ptr(root, path, 1);
+			free_extent(root, leaf_buf->blocknr, 1);
+		}
 	} else {
 		if (slot == 0)
 			fixup_low_keys(root, path, &leaf->items[0].key, 1);
@@ -867,6 +873,72 @@ int del_item(struct ctree_root *root, struct ctree_path *path)
 	return 0;
 }
 
+static int del_pending_extents(struct ctree_root *extent_root)
+{
+	int ret;
+	struct key key;
+	struct tree_buffer *gang[4];
+	int i;
+	struct ctree_path path;
+
+	while(1) {
+		ret = radix_tree_gang_lookup_tag(&extent_root->cache_radix,
+						 (void **)gang, 0, ARRAY_SIZE(gang),
+						 CTREE_EXTENT_PENDING);
+		if (!ret)
+			break;
+		for (i = 0; i < ret; i++) {
+			key.objectid = gang[i]->blocknr;
+			key.flags = 0;
+			key.offset = 1;
+			init_path(&path);
+			ret = search_slot(extent_root, &key, &path, 0);
+			if (ret) {
+				BUG();
+				// FIXME undo it and return sane
+				return ret;
+			}
+			ret = del_item(extent_root, &path);
+			if (ret) {
+				BUG();
+				return ret;
+			}
+			release_path(extent_root, &path);
+			radix_tree_tag_clear(&extent_root->cache_radix, gang[i]->blocknr,
+						CTREE_EXTENT_PENDING);
+			tree_block_release(extent_root, gang[i]);
+		}
+	}
+	return 0;
+}
+
+int free_extent(struct ctree_root *root, u64 blocknr, u64 num_blocks)
+{
+	struct ctree_path path;
+	struct key key;
+	struct ctree_root *extent_root = root->extent_root;
+	struct tree_buffer *t;
+	int pending_ret;
+	int ret;
+
+	key.objectid = blocknr;
+	key.flags = 0;
+	key.offset = num_blocks;
+	if (root == extent_root) {
+		t = read_tree_block(root, key.objectid);
+		radix_tree_tag_set(&root->cache_radix, key.objectid, CTREE_EXTENT_PENDING);
+		return 0;
+	}
+	init_path(&path);
+	ret = search_slot(extent_root, &key, &path, 0);
+	if (ret)
+		BUG();
+	ret = del_item(extent_root, &path);
+	release_path(extent_root, &path);
+	pending_ret = del_pending_extents(root->extent_root);
+	return ret ? ret : pending_ret;
+}
+
 int next_leaf(struct ctree_root *root, struct ctree_path *path)
 {
 	int slot;
@@ -904,8 +976,8 @@ int next_leaf(struct ctree_root *root, struct ctree_path *path)
 	return 0;
 }
 
-int alloc_extent(struct ctree_root *orig_root, u64 num_blocks, u64 search_start,
-		 u64 search_end, u64 owner, struct key *ins)
+int find_free_extent(struct ctree_root *orig_root, u64 num_blocks, u64 search_start,
+			 u64 search_end, struct key *ins)
 {
 	struct ctree_path path;
 	struct key *key;
@@ -915,15 +987,13 @@ int alloc_extent(struct ctree_root *orig_root, u64 num_blocks, u64 search_start,
 	u64 last_block;
 	int start_found = 0;
 	struct leaf *l;
-	struct extent_item extent_item;
 	struct ctree_root * root = orig_root->extent_root;
 
 	init_path(&path);
 	ins->objectid = search_start;
 	ins->offset = 0;
 	ins->flags = 0;
-
-	ret = search_slot(root, ins, &path, sizeof(struct extent_item));
+	ret = search_slot(root, ins, &path, 0);
 	while (1) {
 		l = &path.nodes[0]->leaf;
 		slot = path.slots[0];
@@ -938,6 +1008,7 @@ int alloc_extent(struct ctree_root *orig_root, u64 num_blocks, u64 search_start,
 				ins->objectid = search_start;
 				ins->offset = num_blocks;
 				hole_size = search_end - search_start;
+				start_found = 1;
 				goto insert;
 			}
 			ins->objectid = last_block;
@@ -956,51 +1027,119 @@ int alloc_extent(struct ctree_root *orig_root, u64 num_blocks, u64 search_start,
 		} else
 			start_found = 1;
 		last_block = key->objectid + key->offset;
+insert_failed:
 		path.slots[0]++;
 	}
 	// FIXME -ENOSPC
 insert:
+	if (orig_root->extent_root == orig_root) {
+		BUG_ON(num_blocks != 1);
+		if ((root->current_insert.objectid <= ins->objectid &&
+		    root->current_insert.objectid + root->current_insert.offset >
+		    ins->objectid) ||
+		   (root->current_insert.objectid > ins->objectid &&
+		    root->current_insert.objectid <= ins->objectid + ins->offset) ||
+		   radix_tree_tag_get(&root->cache_radix, ins->objectid,
+				      CTREE_EXTENT_PENDING)) {
+			last_block = ins->objectid + 1;
+			search_start = last_block;
+			goto insert_failed;
+		}
+	}
 	release_path(root, &path);
+	if (ins->offset != 1)
+		BUG();
+	return 0;
+}
+
+static int insert_pending_extents(struct ctree_root *extent_root)
+{
+	int ret;
+	struct key key;
+	struct extent_item item;
+	struct tree_buffer *gang[4];
+	int i;
+
+	// FIXME -ENOSPC
+	item.refs = 1;
+	item.owner = extent_root->node->node.header.parentid;
+	while(1) {
+		ret = radix_tree_gang_lookup_tag(&extent_root->cache_radix,
+						 (void **)gang, 0, ARRAY_SIZE(gang),
+						 CTREE_EXTENT_PENDING);
+		if (!ret)
+			break;
+		for (i = 0; i < ret; i++) {
+			key.objectid = gang[i]->blocknr;
+			key.flags = 0;
+			key.offset = 1;
+			ret = insert_item(extent_root, &key, &item, sizeof(item));
+			if (ret) {
+				BUG();
+				// FIXME undo it and return sane
+				return ret;
+			}
+			radix_tree_tag_clear(&extent_root->cache_radix, gang[i]->blocknr,
+						CTREE_EXTENT_PENDING);
+			tree_block_release(extent_root, gang[i]);
+		}
+	}
+	return 0;
+}
+
+int alloc_extent(struct ctree_root *root, u64 num_blocks, u64 search_start,
+			 u64 search_end, u64 owner, struct key *ins, struct tree_buffer **buf)
+{
+	int ret;
+	int pending_ret;
+	struct extent_item extent_item;
+
 	extent_item.refs = 1;
 	extent_item.owner = owner;
-	if (root == orig_root && root->reserve_extent->num_blocks == 0) {
-		root->reserve_extent->blocknr = ins->objectid;
-		root->reserve_extent->num_blocks = ins->offset;
-		root->reserve_extent->num_used = 0;
+
+	ret = find_free_extent(root, num_blocks, search_start, search_end, ins);
+	if (ret)
+		return ret;
+
+	if (root != root->extent_root) {
+		memcpy(&root->extent_root->current_insert, ins, sizeof(*ins));
+		ret = insert_item(root->extent_root, ins, &extent_item, sizeof(extent_item));
+		memset(&root->extent_root->current_insert, 0, sizeof(struct key));
+		pending_ret = insert_pending_extents(root->extent_root);
+		if (ret)
+			return ret;
+		if (pending_ret)
+			return pending_ret;
+		*buf = find_tree_block(root, ins->objectid);
+		return 0;
 	}
-	ret = insert_item(root->extent_root, ins, &extent_item, sizeof(extent_item));
-	return ret;
+	/* we're allocating an extent for the extent tree, don't recurse */
+	BUG_ON(ins->offset != 1);
+	*buf = find_tree_block(root, ins->objectid);
+	BUG_ON(!*buf);
+	radix_tree_tag_set(&root->cache_radix, ins->objectid, CTREE_EXTENT_PENDING);
+	(*buf)->count++;
+	return 0;
+
 }
 
-static int refill_alloc_extent(struct ctree_root *root)
+struct tree_buffer *alloc_free_block(struct ctree_root *root)
 {
-	struct alloc_extent *ae = root->alloc_extent;
-	struct key key;
+	struct key ins;
 	int ret;
-	int min_blocks = MAX_LEVEL * 2;
+	struct tree_buffer *buf = NULL;
 
-	if (ae->num_blocks > ae->num_used && ae->num_blocks - ae->num_used >
-	    min_blocks)
-		return 0;
-	ae = root->reserve_extent;
-	if (ae->num_blocks > ae->num_used) {
-		if (root->alloc_extent->num_blocks == 0) {
-			/* we should swap reserve/alloc_extent when alloc
-			 * fills up
-			 */
-			BUG();
-		}
-		if (ae->num_blocks - ae->num_used < min_blocks)
-			BUG();
-		return 0;
+	ret = alloc_extent(root, 1, 0, (unsigned long)-1, root->node->node.header.parentid,
+			   &ins, &buf);
+
+	if (ret) {
+		BUG();
+		return NULL;
 	}
-	ret = alloc_extent(root,
-			   min_blocks * 2, 0, (unsigned long)-1,
-			   root->node->node.header.parentid, &key);
-	ae->blocknr = key.objectid;
-	ae->num_blocks = key.offset;
-	ae->num_used = 0;
-	return ret;
+	if (root != root->extent_root)
+		BUG_ON(radix_tree_tag_get(&root->extent_root->cache_radix, buf->blocknr,
+					  CTREE_EXTENT_PENDING));
+	return buf;
 }
 
 void print_leaf(struct leaf *l)
@@ -1096,6 +1235,7 @@ int main() {
 	print_tree(root, root->node);
 	printf("map tree\n");
 	print_tree(root->extent_root, root->extent_root->node);
+	fflush(stdout);
 
 	srand(55);
 	for (i = 0; i < run_size; i++) {
@@ -1111,12 +1251,6 @@ int main() {
 		if (!ret)
 			tree_size++;
 	}
-	printf("root used: %lu\n", root->alloc_extent->num_used);
-	printf("root tree\n");
-	// print_tree(root, root->node);
-	printf("map tree\n");
-	printf("map used: %lu\n", root->extent_root->alloc_extent->num_used);
-	// print_tree(root->extent_root, root->extent_root->node);
 	write_ctree_super(root, &super);
 	close_ctree(root);
 
@@ -1167,12 +1301,27 @@ int main() {
 		ret = insert_item(root, &ins, buf, strlen(buf));
 		if (!ret)
 			tree_size++;
+		if (i >= 5) {
+			struct key ugh;
+			ugh.objectid = 5;
+			ugh.flags = 0;
+			ugh.offset = 0;
+			init_path(&path);
+			ret = search_slot(root, &ugh, &path, 0);
+			if (ret) {
+				print_tree(root, root->node);
+				printf("unable to find 5 %d\n", num);
+				exit(1);
+			}
+			release_path(root, &path);
+
+		}
 	}
 	write_ctree_super(root, &super);
 	close_ctree(root);
 	root = open_ctree("dbfile", &super);
-	printf("starting search2\n");
 	srand(128);
+	printf("starting search2\n");
 	for (i = 0; i < run_size; i++) {
 		num = next_key(i, max_key);
 		ins.objectid = num;
@@ -1219,5 +1368,7 @@ int main() {
 	write_ctree_super(root, &super);
 	close_ctree(root);
 	printf("tree size is now %d\n", tree_size);
+	printf("map tree\n");
+	print_tree(root->extent_root, root->extent_root->node);
 	return 0;
 }
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 78407d32db78..8c32c0e9267d 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -28,21 +28,12 @@ struct header {
 
 struct tree_buffer;
 
-struct alloc_extent {
-	u64 blocknr;
-	u64 num_blocks;
-	u64 num_used;
-} __attribute__ ((__packed__));
-
 struct ctree_root {
 	struct tree_buffer *node;
 	struct ctree_root *extent_root;
-	struct alloc_extent *alloc_extent;
-	struct alloc_extent *reserve_extent;
+	struct key current_insert;
 	int fp;
 	struct radix_tree_root cache_radix;
-	struct alloc_extent ai1;
-	struct alloc_extent ai2;
 };
 
 struct ctree_root_info {
@@ -52,8 +43,6 @@ struct ctree_root_info {
 	u64 tree_root; /* the tree root */
 	u32 csum;
 	u32 ham;
-	struct alloc_extent alloc_extent;
-	struct alloc_extent reserve_extent;
 	u64 snapuuid[2]; /* root specific uuid */
 } __attribute__ ((__packed__));
 
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index a696a4278ac5..14955e440773 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -12,33 +12,13 @@
 
 static int allocated_blocks = 0;
 
-static int get_free_block(struct ctree_root *root, u64 *block)
+static int check_tree_block(struct ctree_root *root, struct tree_buffer *buf)
 {
-	struct stat st;
-	int ret = 0;
-
-	if (root->alloc_extent->num_used >= root->alloc_extent->num_blocks)
-		return -1;
-
-	*block = root->alloc_extent->blocknr + root->alloc_extent->num_used;
-	root->alloc_extent->num_used += 1;
-	if (root->alloc_extent->num_used >= root->alloc_extent->num_blocks) {
-		struct alloc_extent *ae = root->alloc_extent;
-		root->alloc_extent = root->reserve_extent;
-		root->reserve_extent = ae;
-		ae->num_blocks = 0;
-	}
-	st.st_size = 0;
-	ret = fstat(root->fp, &st);
-	if (st.st_size < (*block + 1) * CTREE_BLOCKSIZE) {
-		ret = ftruncate(root->fp,
-				(*block + 1) * CTREE_BLOCKSIZE);
-		if (ret) {
-			perror("ftruncate");
-			exit(1);
-		}
-	}
-	return ret;
+	if (buf->blocknr != buf->node.header.blocknr)
+		BUG();
+	if (root->node && buf->node.header.parentid != root->node->node.header.parentid)
+		BUG();
+	return 0;
 }
 
 struct tree_buffer *alloc_tree_block(struct ctree_root *root, u64 blocknr)
@@ -61,22 +41,23 @@ struct tree_buffer *alloc_tree_block(struct ctree_root *root, u64 blocknr)
 	return buf;
 }
 
-struct tree_buffer *alloc_free_block(struct ctree_root *root)
+struct tree_buffer *find_tree_block(struct ctree_root *root, u64 blocknr)
 {
-	u64 free_block;
-	int ret;
-	struct tree_buffer * buf;
-	ret = get_free_block(root, &free_block);
-	if (ret) {
-		BUG();
-		return NULL;
+	struct tree_buffer *buf;
+	buf = radix_tree_lookup(&root->cache_radix, blocknr);
+	if (buf) {
+		buf->count++;
+	} else {
+		buf = alloc_tree_block(root, blocknr);
+		if (!buf) {
+			BUG();
+			return NULL;
+		}
 	}
-	buf = alloc_tree_block(root, free_block);
-	if (!buf)
-		BUG();
 	return buf;
 }
 
+
 struct tree_buffer *read_tree_block(struct ctree_root *root, u64 blocknr)
 {
 	loff_t offset = blocknr * CTREE_BLOCKSIZE;
@@ -86,20 +67,17 @@ struct tree_buffer *read_tree_block(struct ctree_root *root, u64 blocknr)
 	buf = radix_tree_lookup(&root->cache_radix, blocknr);
 	if (buf) {
 		buf->count++;
-		goto test;
-	}
-	buf = alloc_tree_block(root, blocknr);
-	if (!buf)
-		return NULL;
-	ret = pread(root->fp, &buf->node, CTREE_BLOCKSIZE, offset);
-	if (ret != CTREE_BLOCKSIZE) {
-		free(buf);
-		return NULL;
+	} else {
+		buf = alloc_tree_block(root, blocknr);
+		if (!buf)
+			return NULL;
+		ret = pread(root->fp, &buf->node, CTREE_BLOCKSIZE, offset);
+		if (ret != CTREE_BLOCKSIZE) {
+			free(buf);
+			return NULL;
+		}
 	}
-test:
-	if (buf->blocknr != buf->node.header.blocknr)
-		BUG();
-	if (root->node && buf->node.header.parentid != root->node->node.header.parentid)
+	if (check_tree_block(root, buf))
 		BUG();
 	return buf;
 }
@@ -121,17 +99,10 @@ int write_tree_block(struct ctree_root *root, struct tree_buffer *buf)
 static int __setup_root(struct ctree_root *root, struct ctree_root *extent_root,
 			struct ctree_root_info *info, int fp)
 {
-	INIT_RADIX_TREE(&root->cache_radix, GFP_KERNEL);
 	root->fp = fp;
 	root->node = NULL;
 	root->node = read_tree_block(root, info->tree_root);
 	root->extent_root = extent_root;
-	memcpy(&root->ai1, &info->alloc_extent, sizeof(info->alloc_extent));
-	memcpy(&root->ai2, &info->reserve_extent, sizeof(info->reserve_extent));
-	root->alloc_extent = &root->ai1;
-	root->reserve_extent = &root->ai2;
-	printf("setup done reading root %p, used %lu available %lu\n", root, root->alloc_extent->num_used, root->alloc_extent->num_blocks);
-	printf("setup done reading root %p, reserve used %lu available %lu\n", root, root->reserve_extent->num_used, root->reserve_extent->num_blocks);
 	return 0;
 }
 
@@ -147,6 +118,8 @@ struct ctree_root *open_ctree(char *filename, struct ctree_super_block *super)
 		free(root);
 		return NULL;
 	}
+	INIT_RADIX_TREE(&root->cache_radix, GFP_KERNEL);
+	INIT_RADIX_TREE(&extent_root->cache_radix, GFP_KERNEL);
 	ret = pread(fp, super, sizeof(struct ctree_super_block),
 		     CTREE_SUPER_INFO_OFFSET(CTREE_BLOCKSIZE));
 	if (ret == 0 || super->root_info.tree_root == 0) {
@@ -168,8 +141,6 @@ struct ctree_root *open_ctree(char *filename, struct ctree_super_block *super)
 static int __update_root(struct ctree_root *root, struct ctree_root_info *info)
 {
 	info->tree_root = root->node->blocknr;
-	memcpy(&info->alloc_extent, root->alloc_extent, sizeof(struct alloc_extent));
-	memcpy(&info->reserve_extent, root->reserve_extent, sizeof(struct alloc_extent));
 	return 0;
 }
 
@@ -201,6 +172,7 @@ int close_ctree(struct ctree_root *root)
 void tree_block_release(struct ctree_root *root, struct tree_buffer *buf)
 {
 	buf->count--;
+	write_tree_block(root, buf);
 	if (buf->count < 0)
 		BUG();
 	if (buf->count == 0) {
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index e288fe883b9a..2729b757ddc1 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -11,11 +11,11 @@ struct tree_buffer {
 };
 
 struct tree_buffer *read_tree_block(struct ctree_root *root, u64 blocknr);
+struct tree_buffer *find_tree_block(struct ctree_root *root, u64 blocknr);
 int write_tree_block(struct ctree_root *root, struct tree_buffer *buf);
 struct ctree_root *open_ctree(char *filename, struct ctree_super_block *s);
 int close_ctree(struct ctree_root *root);
 void tree_block_release(struct ctree_root *root, struct tree_buffer *buf);
-struct tree_buffer *alloc_free_block(struct ctree_root *root);
 int write_ctree_super(struct ctree_root *root, struct ctree_super_block *s);
 int mkfs(int fd);
 
diff --git a/fs/btrfs/mkfs.c b/fs/btrfs/mkfs.c
index 584aba44e801..fd4e5dea7e1b 100644
--- a/fs/btrfs/mkfs.c
+++ b/fs/btrfs/mkfs.c
@@ -23,17 +23,10 @@ int mkfs(int fd)
 	info[0].blocknr = 16;
 	info[0].objectid = 1;
 	info[0].tree_root = 17;
-	info[0].alloc_extent.blocknr = 0;
-	info[0].alloc_extent.num_blocks = 64;
-	/* 0-17 are used (inclusive) */
-	info[0].alloc_extent.num_used = 18;
 
 	info[1].blocknr = 16;
 	info[1].objectid = 2;
-	info[1].tree_root = 64;
-	info[1].alloc_extent.blocknr = 64;
-	info[1].alloc_extent.num_blocks = 64;
-	info[1].alloc_extent.num_used = 1;
+	info[1].tree_root = 18;
 	ret = pwrite(fd, info, sizeof(info),
 		     CTREE_SUPER_INFO_OFFSET(CTREE_BLOCKSIZE));
 	if (ret != sizeof(info))
@@ -48,24 +41,36 @@ int mkfs(int fd)
 		return -1;
 
 	empty_leaf.header.parentid = 2;
-	empty_leaf.header.blocknr = 64;
-	empty_leaf.header.nritems = 2;
+	empty_leaf.header.blocknr = 18;
+	empty_leaf.header.nritems = 3;
+
+	/* item1, reserve blocks 0-16 */
 	item.key.objectid = 0;
-	item.key.offset = 64;
+	item.key.offset = 17;
 	item.key.flags = 0;
 	item.offset = LEAF_DATA_SIZE - sizeof(struct extent_item);
 	item.size = sizeof(struct extent_item);
 	extent_item.refs = 1;
-	extent_item.owner = 1;
+	extent_item.owner = 0;
 	memcpy(empty_leaf.items, &item, sizeof(item));
 	memcpy(empty_leaf.data + item.offset, &extent_item, item.size);
-	item.key.objectid = 64;
-	item.key.offset = 64;
+
+	/* item2, give block 17 to the root */
+	item.key.objectid = 17;
+	item.key.offset = 1;
 	item.offset = LEAF_DATA_SIZE - sizeof(struct extent_item) * 2;
-	extent_item.owner = 2;
+	extent_item.owner = 1;
 	memcpy(empty_leaf.items + 1, &item, sizeof(item));
 	memcpy(empty_leaf.data + item.offset, &extent_item, item.size);
-	ret = pwrite(fd, &empty_leaf, sizeof(empty_leaf), 64 * CTREE_BLOCKSIZE);
+
+	/* item3, give block 18 for the extent root */
+	item.key.objectid = 18;
+	item.key.offset = 1;
+	item.offset = LEAF_DATA_SIZE - sizeof(struct extent_item) * 3;
+	extent_item.owner = 2;
+	memcpy(empty_leaf.items + 2, &item, sizeof(item));
+	memcpy(empty_leaf.data + item.offset, &extent_item, item.size);
+	ret = pwrite(fd, &empty_leaf, sizeof(empty_leaf), 18 * CTREE_BLOCKSIZE);
 	if (ret != sizeof(empty_leaf))
 		return -1;
 	return 0;
-- 
cgit v1.2.3


From ed2ff2cba766dfe7976a0113f667c9a0a50dff02 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Thu, 1 Mar 2007 18:59:40 -0500
Subject: Btrfs: pretend page cache & commit code

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.c   |  66 +++++----
 fs/btrfs/ctree.h   |   5 +
 fs/btrfs/disk-io.c |  91 +++++++++++-
 fs/btrfs/disk-io.h |   6 +
 fs/btrfs/list.h    | 418 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 5 files changed, 558 insertions(+), 28 deletions(-)
 create mode 100644 fs/btrfs/list.h

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index afa5bc5c7c1a..57fa505fb9f6 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -250,6 +250,7 @@ static int balance_level(struct ctree_root *root, struct ctree_path *path,
 		tree_block_release(root, mid_buf);
 		/* once for the root ptr */
 		tree_block_release(root, mid_buf);
+		clean_tree_block(root, mid_buf);
 		return free_extent(root, blocknr, 1);
 	}
 	parent = &parent_buf->node;
@@ -280,6 +281,7 @@ static int balance_level(struct ctree_root *root, struct ctree_path *path,
 		if (right->header.nritems == 0) {
 			u64 blocknr = right_buf->blocknr;
 			tree_block_release(root, right_buf);
+			clean_tree_block(root, right_buf);
 			right_buf = NULL;
 			right = NULL;
 			wret = del_ptr(root, path, level + 1, pslot + 1);
@@ -291,7 +293,7 @@ static int balance_level(struct ctree_root *root, struct ctree_path *path,
 		} else {
 			memcpy(parent->keys + pslot + 1, right->keys,
 				sizeof(struct key));
-			wret = write_tree_block(root, parent_buf);
+			wret = dirty_tree_block(root, parent_buf);
 			if (wret)
 				ret = wret;
 		}
@@ -316,6 +318,7 @@ static int balance_level(struct ctree_root *root, struct ctree_path *path,
 		/* we've managed to empty the middle node, drop it */
 		u64 blocknr = mid_buf->blocknr;
 		tree_block_release(root, mid_buf);
+		clean_tree_block(root, mid_buf);
 		mid_buf = NULL;
 		mid = NULL;
 		wret = del_ptr(root, path, level + 1, pslot);
@@ -327,7 +330,7 @@ static int balance_level(struct ctree_root *root, struct ctree_path *path,
 	} else {
 		/* update the parent key to reflect our changes */
 		memcpy(parent->keys + pslot, mid->keys, sizeof(struct key));
-		wret = write_tree_block(root, parent_buf);
+		wret = dirty_tree_block(root, parent_buf);
 		if (wret)
 			ret = wret;
 	}
@@ -458,7 +461,7 @@ static int fixup_low_keys(struct ctree_root *root,
 			break;
 		t = &path->nodes[i]->node;
 		memcpy(t->keys + tslot, key, sizeof(*key));
-		wret = write_tree_block(root, path->nodes[i]);
+		wret = dirty_tree_block(root, path->nodes[i]);
 		if (wret)
 			ret = wret;
 		if (tslot != 0)
@@ -508,11 +511,11 @@ static int push_node_left(struct ctree_root *root, struct tree_buffer *dst_buf,
 	src->header.nritems -= push_items;
 	dst->header.nritems += push_items;
 
-	wret = write_tree_block(root, src_buf);
+	wret = dirty_tree_block(root, src_buf);
 	if (wret < 0)
 		ret = wret;
 
-	wret = write_tree_block(root, dst_buf);
+	wret = dirty_tree_block(root, dst_buf);
 	if (wret < 0)
 		ret = wret;
 	return ret;
@@ -566,11 +569,11 @@ static int balance_node_right(struct ctree_root *root,
 	src->header.nritems -= push_items;
 	dst->header.nritems += push_items;
 
-	wret = write_tree_block(root, src_buf);
+	wret = dirty_tree_block(root, src_buf);
 	if (wret < 0)
 		ret = wret;
 
-	wret = write_tree_block(root, dst_buf);
+	wret = dirty_tree_block(root, dst_buf);
 	if (wret < 0)
 		ret = wret;
 	return ret;
@@ -612,7 +615,7 @@ static int insert_new_root(struct ctree_root *root,
 	tree_block_release(root, root->node);
 	root->node = t;
 	t->count++;
-	write_tree_block(root, t);
+	dirty_tree_block(root, t);
 	path->nodes[level] = t;
 	path->slots[level] = 0;
 	return 0;
@@ -652,7 +655,7 @@ static int insert_ptr(struct ctree_root *root,
 	lower->header.nritems++;
 	if (lower->keys[1].objectid == 0)
 			BUG();
-	write_tree_block(root, path->nodes[level]);
+	dirty_tree_block(root, path->nodes[level]);
 	return 0;
 }
 
@@ -698,10 +701,10 @@ static int split_node(struct ctree_root *root, struct ctree_path *path,
 	c->header.nritems = mid;
 	ret = 0;
 
-	wret = write_tree_block(root, t);
+	wret = dirty_tree_block(root, t);
 	if (wret)
 		ret = wret;
-	wret = write_tree_block(root, split_buffer);
+	wret = dirty_tree_block(root, split_buffer);
 	if (wret)
 		ret = wret;
 	wret = insert_ptr(root, path, split->keys, split_buffer->blocknr,
@@ -815,11 +818,11 @@ static int push_leaf_right(struct ctree_root *root, struct ctree_path *path,
 	}
 	left->header.nritems -= push_items;
 
-	write_tree_block(root, left_buf);
-	write_tree_block(root, right_buf);
+	dirty_tree_block(root, left_buf);
+	dirty_tree_block(root, right_buf);
 	memcpy(upper->node.keys + slot + 1,
 		&right->items[0].key, sizeof(struct key));
-	write_tree_block(root, upper);
+	dirty_tree_block(root, upper);
 	/* then fixup the leaf pointer in the path */
 	if (path->slots[0] >= left->header.nritems) {
 		path->slots[0] -= left->header.nritems;
@@ -909,10 +912,10 @@ static int push_leaf_left(struct ctree_root *root, struct ctree_path *path,
 		push_space = right->items[i].offset;
 	}
 
-	wret = write_tree_block(root, t);
+	wret = dirty_tree_block(root, t);
 	if (wret)
 		ret = wret;
-	wret = write_tree_block(root, right_buf);
+	wret = dirty_tree_block(root, right_buf);
 	if (wret)
 		ret = wret;
 
@@ -1019,10 +1022,10 @@ static int split_leaf(struct ctree_root *root, struct ctree_path *path,
 			  right_buffer->blocknr, path->slots[1] + 1, 1);
 	if (wret)
 		ret = wret;
-	wret = write_tree_block(root, right_buffer);
+	wret = dirty_tree_block(root, right_buffer);
 	if (wret)
 		ret = wret;
-	wret = write_tree_block(root, l_buf);
+	wret = dirty_tree_block(root, l_buf);
 	if (wret)
 		ret = wret;
 
@@ -1062,12 +1065,14 @@ int insert_item(struct ctree_root *root, struct key *key,
 	ret = search_slot(root, key, &path, data_size);
 	if (ret == 0) {
 		release_path(root, &path);
-		return -EEXIST;
-	}
-	if (ret < 0) {
-		release_path(root, &path);
+		ret = -EEXIST;
+		wret = commit_transaction(root);
+		if (wret)
+			ret = wret;
 		return ret;
 	}
+	if (ret < 0)
+		goto out;
 
 	slot_orig = path.slots[0];
 	leaf_buf = path.nodes[0];
@@ -1113,14 +1118,18 @@ int insert_item(struct ctree_root *root, struct key *key,
 	if (slot == 0)
 		ret = fixup_low_keys(root, &path, key, 1);
 
-	wret = write_tree_block(root, leaf_buf);
+	wret = dirty_tree_block(root, leaf_buf);
 	if (wret)
 		ret = wret;
 
 	if (leaf_free_space(leaf) < 0)
 		BUG();
 	check_leaf(&path, 0);
+out:
 	release_path(root, &path);
+	wret = commit_transaction(root);
+	if (wret)
+		ret = wret;
 	return ret;
 }
 
@@ -1160,7 +1169,7 @@ static int del_ptr(struct ctree_root *root, struct ctree_path *path, int level,
 		if (wret)
 			ret = wret;
 	}
-	wret = write_tree_block(root, parent);
+	wret = dirty_tree_block(root, parent);
 	if (wret)
 		ret = wret;
 	return ret;
@@ -1203,8 +1212,9 @@ int del_item(struct ctree_root *root, struct ctree_path *path)
 	if (leaf->header.nritems == 0) {
 		if (leaf_buf == root->node) {
 			leaf->header.flags = node_level(0);
-			write_tree_block(root, leaf_buf);
+			dirty_tree_block(root, leaf_buf);
 		} else {
+			clean_tree_block(root, leaf_buf);
 			wret = del_ptr(root, path, 1, path->slots[1]);
 			if (wret)
 				ret = wret;
@@ -1220,7 +1230,7 @@ int del_item(struct ctree_root *root, struct ctree_path *path)
 			if (wret)
 				ret = wret;
 		}
-		wret = write_tree_block(root, leaf_buf);
+		wret = dirty_tree_block(root, leaf_buf);
 		if (wret)
 			ret = wret;
 
@@ -1242,6 +1252,7 @@ int del_item(struct ctree_root *root, struct ctree_path *path)
 			}
 			if (leaf->header.nritems == 0) {
 				u64 blocknr = leaf_buf->blocknr;
+				clean_tree_block(root, leaf_buf);
 				wret = del_ptr(root, path, 1, slot);
 				if (wret)
 					ret = wret;
@@ -1254,6 +1265,9 @@ int del_item(struct ctree_root *root, struct ctree_path *path)
 			}
 		}
 	}
+	wret = commit_transaction(root);
+	if (wret)
+		ret = wret;
 	return ret;
 }
 
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 18daccd84535..6b4dabd47698 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1,6 +1,8 @@
 #ifndef __CTREE__
 #define __CTREE__
 
+#include "list.h"
+
 #define CTREE_BLOCKSIZE 1024
 
 /*
@@ -53,6 +55,9 @@ struct ctree_root {
 	struct key current_insert;
 	int fp;
 	struct radix_tree_root cache_radix;
+	struct list_head trans;
+	struct list_head cache;
+	int cache_size;
 };
 
 /*
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index c42dc72706bf..656ace6147a8 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -11,6 +11,8 @@
 #include "disk-io.h"
 
 static int allocated_blocks = 0;
+int cache_size = 0;
+int cache_max = 10000;
 
 static int check_tree_block(struct ctree_root *root, struct tree_buffer *buf)
 {
@@ -21,6 +23,25 @@ static int check_tree_block(struct ctree_root *root, struct tree_buffer *buf)
 	return 0;
 }
 
+static int free_some_buffers(struct ctree_root *root)
+{
+	struct list_head *node, *next;
+	struct tree_buffer *b;
+	if (root->cache_size < cache_max)
+		return 0;
+	list_for_each_safe(node, next, &root->cache) {
+		b = list_entry(node, struct tree_buffer, cache);
+		if (b->count == 1) {
+			BUG_ON(!list_empty(&b->dirty));
+			list_del_init(&b->cache);
+			tree_block_release(root, b);
+			if (root->cache_size < cache_max)
+				return 0;
+		}
+	}
+	return 0;
+}
+
 struct tree_buffer *alloc_tree_block(struct ctree_root *root, u64 blocknr)
 {
 	struct tree_buffer *buf;
@@ -30,10 +51,14 @@ struct tree_buffer *alloc_tree_block(struct ctree_root *root, u64 blocknr)
 		return buf;
 	allocated_blocks++;
 	buf->blocknr = blocknr;
-	buf->count = 1;
+	buf->count = 2;
+	INIT_LIST_HEAD(&buf->dirty);
+	free_some_buffers(root);
 	radix_tree_preload(GFP_KERNEL);
 	ret = radix_tree_insert(&root->cache_radix, blocknr, buf);
 	radix_tree_preload_end();
+	list_add_tail(&buf->cache, &root->cache);
+	root->cache_size++;
 	if (ret) {
 		free(buf);
 		return NULL;
@@ -57,7 +82,6 @@ struct tree_buffer *find_tree_block(struct ctree_root *root, u64 blocknr)
 	return buf;
 }
 
-
 struct tree_buffer *read_tree_block(struct ctree_root *root, u64 blocknr)
 {
 	loff_t offset = blocknr * CTREE_BLOCKSIZE;
@@ -82,6 +106,24 @@ struct tree_buffer *read_tree_block(struct ctree_root *root, u64 blocknr)
 	return buf;
 }
 
+int dirty_tree_block(struct ctree_root *root, struct tree_buffer *buf)
+{
+	if (!list_empty(&buf->dirty))
+		return 0;
+	list_add_tail(&buf->dirty, &root->trans);
+	buf->count++;
+	return 0;
+}
+
+int clean_tree_block(struct ctree_root *root, struct tree_buffer *buf)
+{
+	if (!list_empty(&buf->dirty)) {
+		list_del_init(&buf->dirty);
+		tree_block_release(root, buf);
+	}
+	return 0;
+}
+
 int write_tree_block(struct ctree_root *root, struct tree_buffer *buf)
 {
 	u64 blocknr = buf->blocknr;
@@ -96,9 +138,37 @@ int write_tree_block(struct ctree_root *root, struct tree_buffer *buf)
 	return 0;
 }
 
+static int __commit_transaction(struct ctree_root *root)
+{
+	struct tree_buffer *b;
+	int ret = 0;
+	int wret;
+	while(!list_empty(&root->trans)) {
+		b = list_entry(root->trans.next, struct tree_buffer, dirty);
+		list_del_init(&b->dirty);
+		wret = write_tree_block(root, b);
+		if (wret)
+			ret = wret;
+		tree_block_release(root, b);
+	}
+	return ret;
+}
+
+int commit_transaction(struct ctree_root *root)
+{
+	int ret;
+	ret = __commit_transaction(root);
+	if (!ret && root != root->extent_root)
+		ret = __commit_transaction(root->extent_root);
+	BUG_ON(ret);
+	return ret;
+}
+
 static int __setup_root(struct ctree_root *root, struct ctree_root *extent_root,
 			struct ctree_root_info *info, int fp)
 {
+	INIT_LIST_HEAD(&root->trans);
+	INIT_LIST_HEAD(&root->cache);
 	root->fp = fp;
 	root->node = NULL;
 	root->node = read_tree_block(root, info->tree_root);
@@ -157,8 +227,23 @@ int write_ctree_super(struct ctree_root *root, struct ctree_super_block *s)
 	return 0;
 }
 
+static int drop_cache(struct ctree_root *root)
+{
+	while(!list_empty(&root->cache)) {
+		struct tree_buffer *b = list_entry(root->cache.next,
+						   struct tree_buffer, cache);
+		list_del_init(&b->cache);
+		tree_block_release(root, b);
+	}
+	return 0;
+}
 int close_ctree(struct ctree_root *root)
 {
+	drop_cache(root->extent_root);
+	drop_cache(root);
+	BUG_ON(!list_empty(&root->trans));
+	BUG_ON(!list_empty(&root->extent_root->trans));
+
 	close(root->fp);
 	if (root->node)
 		tree_block_release(root, root->node);
@@ -182,6 +267,8 @@ void tree_block_release(struct ctree_root *root, struct tree_buffer *buf)
 		free(buf);
 		BUG_ON(allocated_blocks == 0);
 		allocated_blocks--;
+		BUG_ON(root->cache_size == 0);
+		root->cache_size--;
 	}
 }
 
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 2729b757ddc1..b5dee2fae4da 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -1,5 +1,6 @@
 #ifndef __DISKIO__
 #define __DISKIO__
+#include "list.h"
 
 struct tree_buffer {
 	u64 blocknr;
@@ -8,11 +9,16 @@ struct tree_buffer {
 		struct node node;
 		struct leaf leaf;
 	};
+	struct list_head dirty;
+	struct list_head cache;
 };
 
 struct tree_buffer *read_tree_block(struct ctree_root *root, u64 blocknr);
 struct tree_buffer *find_tree_block(struct ctree_root *root, u64 blocknr);
 int write_tree_block(struct ctree_root *root, struct tree_buffer *buf);
+int dirty_tree_block(struct ctree_root *root, struct tree_buffer *buf);
+int clean_tree_block(struct ctree_root *root, struct tree_buffer *buf);
+int commit_transaction(struct ctree_root *root);
 struct ctree_root *open_ctree(char *filename, struct ctree_super_block *s);
 int close_ctree(struct ctree_root *root);
 void tree_block_release(struct ctree_root *root, struct tree_buffer *buf);
diff --git a/fs/btrfs/list.h b/fs/btrfs/list.h
new file mode 100644
index 000000000000..1aafafb13370
--- /dev/null
+++ b/fs/btrfs/list.h
@@ -0,0 +1,418 @@
+#ifndef _LINUX_LIST_H
+#define _LINUX_LIST_H
+
+#define LIST_POISON1  ((void *) 0x00100100)
+#define LIST_POISON2  ((void *) 0x00200200)
+
+/*
+ * Simple doubly linked list implementation.
+ *
+ * Some of the internal functions ("__xxx") are useful when
+ * manipulating whole lists rather than single entries, as
+ * sometimes we already know the next/prev entries and we can
+ * generate better code by using them directly rather than
+ * using the generic single-entry routines.
+ */
+
+struct list_head {
+	struct list_head *next, *prev;
+};
+
+#define LIST_HEAD_INIT(name) { &(name), &(name) }
+
+#define LIST_HEAD(name) \
+	struct list_head name = LIST_HEAD_INIT(name)
+
+static inline void INIT_LIST_HEAD(struct list_head *list)
+{
+	list->next = list;
+	list->prev = list;
+}
+
+/*
+ * Insert a new entry between two known consecutive entries.
+ *
+ * This is only for internal list manipulation where we know
+ * the prev/next entries already!
+ */
+#ifndef CONFIG_DEBUG_LIST
+static inline void __list_add(struct list_head *new,
+			      struct list_head *prev,
+			      struct list_head *next)
+{
+	next->prev = new;
+	new->next = next;
+	new->prev = prev;
+	prev->next = new;
+}
+#else
+extern void __list_add(struct list_head *new,
+			      struct list_head *prev,
+			      struct list_head *next);
+#endif
+
+/**
+ * list_add - add a new entry
+ * @new: new entry to be added
+ * @head: list head to add it after
+ *
+ * Insert a new entry after the specified head.
+ * This is good for implementing stacks.
+ */
+#ifndef CONFIG_DEBUG_LIST
+static inline void list_add(struct list_head *new, struct list_head *head)
+{
+	__list_add(new, head, head->next);
+}
+#else
+extern void list_add(struct list_head *new, struct list_head *head);
+#endif
+
+
+/**
+ * list_add_tail - add a new entry
+ * @new: new entry to be added
+ * @head: list head to add it before
+ *
+ * Insert a new entry before the specified head.
+ * This is useful for implementing queues.
+ */
+static inline void list_add_tail(struct list_head *new, struct list_head *head)
+{
+	__list_add(new, head->prev, head);
+}
+
+/*
+ * Delete a list entry by making the prev/next entries
+ * point to each other.
+ *
+ * This is only for internal list manipulation where we know
+ * the prev/next entries already!
+ */
+static inline void __list_del(struct list_head * prev, struct list_head * next)
+{
+	next->prev = prev;
+	prev->next = next;
+}
+
+/**
+ * list_del - deletes entry from list.
+ * @entry: the element to delete from the list.
+ * Note: list_empty on entry does not return true after this, the entry is
+ * in an undefined state.
+ */
+#ifndef CONFIG_DEBUG_LIST
+static inline void list_del(struct list_head *entry)
+{
+	__list_del(entry->prev, entry->next);
+	entry->next = LIST_POISON1;
+	entry->prev = LIST_POISON2;
+}
+#else
+extern void list_del(struct list_head *entry);
+#endif
+
+/**
+ * list_replace - replace old entry by new one
+ * @old : the element to be replaced
+ * @new : the new element to insert
+ * Note: if 'old' was empty, it will be overwritten.
+ */
+static inline void list_replace(struct list_head *old,
+				struct list_head *new)
+{
+	new->next = old->next;
+	new->next->prev = new;
+	new->prev = old->prev;
+	new->prev->next = new;
+}
+
+static inline void list_replace_init(struct list_head *old,
+					struct list_head *new)
+{
+	list_replace(old, new);
+	INIT_LIST_HEAD(old);
+}
+/**
+ * list_del_init - deletes entry from list and reinitialize it.
+ * @entry: the element to delete from the list.
+ */
+static inline void list_del_init(struct list_head *entry)
+{
+	__list_del(entry->prev, entry->next);
+	INIT_LIST_HEAD(entry);
+}
+
+/**
+ * list_move - delete from one list and add as another's head
+ * @list: the entry to move
+ * @head: the head that will precede our entry
+ */
+static inline void list_move(struct list_head *list, struct list_head *head)
+{
+        __list_del(list->prev, list->next);
+        list_add(list, head);
+}
+
+/**
+ * list_move_tail - delete from one list and add as another's tail
+ * @list: the entry to move
+ * @head: the head that will follow our entry
+ */
+static inline void list_move_tail(struct list_head *list,
+				  struct list_head *head)
+{
+        __list_del(list->prev, list->next);
+        list_add_tail(list, head);
+}
+
+/**
+ * list_is_last - tests whether @list is the last entry in list @head
+ * @list: the entry to test
+ * @head: the head of the list
+ */
+static inline int list_is_last(const struct list_head *list,
+				const struct list_head *head)
+{
+	return list->next == head;
+}
+
+/**
+ * list_empty - tests whether a list is empty
+ * @head: the list to test.
+ */
+static inline int list_empty(const struct list_head *head)
+{
+	return head->next == head;
+}
+
+/**
+ * list_empty_careful - tests whether a list is empty and not being modified
+ * @head: the list to test
+ *
+ * Description:
+ * tests whether a list is empty _and_ checks that no other CPU might be
+ * in the process of modifying either member (next or prev)
+ *
+ * NOTE: using list_empty_careful() without synchronization
+ * can only be safe if the only activity that can happen
+ * to the list entry is list_del_init(). Eg. it cannot be used
+ * if another CPU could re-list_add() it.
+ */
+static inline int list_empty_careful(const struct list_head *head)
+{
+	struct list_head *next = head->next;
+	return (next == head) && (next == head->prev);
+}
+
+static inline void __list_splice(struct list_head *list,
+				 struct list_head *head)
+{
+	struct list_head *first = list->next;
+	struct list_head *last = list->prev;
+	struct list_head *at = head->next;
+
+	first->prev = head;
+	head->next = first;
+
+	last->next = at;
+	at->prev = last;
+}
+
+/**
+ * list_splice - join two lists
+ * @list: the new list to add.
+ * @head: the place to add it in the first list.
+ */
+static inline void list_splice(struct list_head *list, struct list_head *head)
+{
+	if (!list_empty(list))
+		__list_splice(list, head);
+}
+
+/**
+ * list_splice_init - join two lists and reinitialise the emptied list.
+ * @list: the new list to add.
+ * @head: the place to add it in the first list.
+ *
+ * The list at @list is reinitialised
+ */
+static inline void list_splice_init(struct list_head *list,
+				    struct list_head *head)
+{
+	if (!list_empty(list)) {
+		__list_splice(list, head);
+		INIT_LIST_HEAD(list);
+	}
+}
+
+/**
+ * list_entry - get the struct for this entry
+ * @ptr:	the &struct list_head pointer.
+ * @type:	the type of the struct this is embedded in.
+ * @member:	the name of the list_struct within the struct.
+ */
+#define list_entry(ptr, type, member) \
+	container_of(ptr, type, member)
+
+/**
+ * list_for_each	-	iterate over a list
+ * @pos:	the &struct list_head to use as a loop cursor.
+ * @head:	the head for your list.
+ */
+#define list_for_each(pos, head) \
+	for (pos = (head)->next; prefetch(pos->next), pos != (head); \
+        	pos = pos->next)
+
+/**
+ * __list_for_each	-	iterate over a list
+ * @pos:	the &struct list_head to use as a loop cursor.
+ * @head:	the head for your list.
+ *
+ * This variant differs from list_for_each() in that it's the
+ * simplest possible list iteration code, no prefetching is done.
+ * Use this for code that knows the list to be very short (empty
+ * or 1 entry) most of the time.
+ */
+#define __list_for_each(pos, head) \
+	for (pos = (head)->next; pos != (head); pos = pos->next)
+
+/**
+ * list_for_each_prev	-	iterate over a list backwards
+ * @pos:	the &struct list_head to use as a loop cursor.
+ * @head:	the head for your list.
+ */
+#define list_for_each_prev(pos, head) \
+	for (pos = (head)->prev; prefetch(pos->prev), pos != (head); \
+        	pos = pos->prev)
+
+/**
+ * list_for_each_safe - iterate over a list safe against removal of list entry
+ * @pos:	the &struct list_head to use as a loop cursor.
+ * @n:		another &struct list_head to use as temporary storage
+ * @head:	the head for your list.
+ */
+#define list_for_each_safe(pos, n, head) \
+	for (pos = (head)->next, n = pos->next; pos != (head); \
+		pos = n, n = pos->next)
+
+/**
+ * list_for_each_entry	-	iterate over list of given type
+ * @pos:	the type * to use as a loop cursor.
+ * @head:	the head for your list.
+ * @member:	the name of the list_struct within the struct.
+ */
+#define list_for_each_entry(pos, head, member)				\
+	for (pos = list_entry((head)->next, typeof(*pos), member);	\
+	     prefetch(pos->member.next), &pos->member != (head); 	\
+	     pos = list_entry(pos->member.next, typeof(*pos), member))
+
+/**
+ * list_for_each_entry_reverse - iterate backwards over list of given type.
+ * @pos:	the type * to use as a loop cursor.
+ * @head:	the head for your list.
+ * @member:	the name of the list_struct within the struct.
+ */
+#define list_for_each_entry_reverse(pos, head, member)			\
+	for (pos = list_entry((head)->prev, typeof(*pos), member);	\
+	     prefetch(pos->member.prev), &pos->member != (head); 	\
+	     pos = list_entry(pos->member.prev, typeof(*pos), member))
+
+/**
+ * list_prepare_entry - prepare a pos entry for use in list_for_each_entry_continue
+ * @pos:	the type * to use as a start point
+ * @head:	the head of the list
+ * @member:	the name of the list_struct within the struct.
+ *
+ * Prepares a pos entry for use as a start point in list_for_each_entry_continue.
+ */
+#define list_prepare_entry(pos, head, member) \
+	((pos) ? : list_entry(head, typeof(*pos), member))
+
+/**
+ * list_for_each_entry_continue - continue iteration over list of given type
+ * @pos:	the type * to use as a loop cursor.
+ * @head:	the head for your list.
+ * @member:	the name of the list_struct within the struct.
+ *
+ * Continue to iterate over list of given type, continuing after
+ * the current position.
+ */
+#define list_for_each_entry_continue(pos, head, member) 		\
+	for (pos = list_entry(pos->member.next, typeof(*pos), member);	\
+	     prefetch(pos->member.next), &pos->member != (head);	\
+	     pos = list_entry(pos->member.next, typeof(*pos), member))
+
+/**
+ * list_for_each_entry_from - iterate over list of given type from the current point
+ * @pos:	the type * to use as a loop cursor.
+ * @head:	the head for your list.
+ * @member:	the name of the list_struct within the struct.
+ *
+ * Iterate over list of given type, continuing from current position.
+ */
+#define list_for_each_entry_from(pos, head, member) 			\
+	for (; prefetch(pos->member.next), &pos->member != (head);	\
+	     pos = list_entry(pos->member.next, typeof(*pos), member))
+
+/**
+ * list_for_each_entry_safe - iterate over list of given type safe against removal of list entry
+ * @pos:	the type * to use as a loop cursor.
+ * @n:		another type * to use as temporary storage
+ * @head:	the head for your list.
+ * @member:	the name of the list_struct within the struct.
+ */
+#define list_for_each_entry_safe(pos, n, head, member)			\
+	for (pos = list_entry((head)->next, typeof(*pos), member),	\
+		n = list_entry(pos->member.next, typeof(*pos), member);	\
+	     &pos->member != (head); 					\
+	     pos = n, n = list_entry(n->member.next, typeof(*n), member))
+
+/**
+ * list_for_each_entry_safe_continue
+ * @pos:	the type * to use as a loop cursor.
+ * @n:		another type * to use as temporary storage
+ * @head:	the head for your list.
+ * @member:	the name of the list_struct within the struct.
+ *
+ * Iterate over list of given type, continuing after current point,
+ * safe against removal of list entry.
+ */
+#define list_for_each_entry_safe_continue(pos, n, head, member) 		\
+	for (pos = list_entry(pos->member.next, typeof(*pos), member), 		\
+		n = list_entry(pos->member.next, typeof(*pos), member);		\
+	     &pos->member != (head);						\
+	     pos = n, n = list_entry(n->member.next, typeof(*n), member))
+
+/**
+ * list_for_each_entry_safe_from
+ * @pos:	the type * to use as a loop cursor.
+ * @n:		another type * to use as temporary storage
+ * @head:	the head for your list.
+ * @member:	the name of the list_struct within the struct.
+ *
+ * Iterate over list of given type from current point, safe against
+ * removal of list entry.
+ */
+#define list_for_each_entry_safe_from(pos, n, head, member) 			\
+	for (n = list_entry(pos->member.next, typeof(*pos), member);		\
+	     &pos->member != (head);						\
+	     pos = n, n = list_entry(n->member.next, typeof(*n), member))
+
+/**
+ * list_for_each_entry_safe_reverse
+ * @pos:	the type * to use as a loop cursor.
+ * @n:		another type * to use as temporary storage
+ * @head:	the head for your list.
+ * @member:	the name of the list_struct within the struct.
+ *
+ * Iterate backwards over list of given type, safe against removal
+ * of list entry.
+ */
+#define list_for_each_entry_safe_reverse(pos, n, head, member)		\
+	for (pos = list_entry((head)->prev, typeof(*pos), member),	\
+		n = list_entry(pos->member.prev, typeof(*pos), member);	\
+	     &pos->member != (head); 					\
+	     pos = n, n = list_entry(n->member.prev, typeof(*n), member))
+
+#endif
-- 
cgit v1.2.3


From a28ec19775d62d673b034082128aca95780d3737 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Tue, 6 Mar 2007 20:08:01 -0500
Subject: Btrfs: Fixup reference counting on cows

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/Makefile      |   2 +-
 fs/btrfs/ctree.c       |   9 +-
 fs/btrfs/ctree.h       |   4 +
 fs/btrfs/disk-io.c     |  31 +++++-
 fs/btrfs/disk-io.h     |   4 +-
 fs/btrfs/extent-tree.c | 291 ++++++++++++++++++++++++++++++++++---------------
 fs/btrfs/print-tree.c  |   6 +-
 fs/btrfs/quick-test.c  |  26 +++--
 fs/btrfs/random-test.c |  21 ++--
 9 files changed, 267 insertions(+), 127 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 756b1cd4cf4d..ae7f4c00c39c 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -1,7 +1,7 @@
 
 CC=gcc
 CFLAGS = -g -Wall
-headers = radix-tree.h ctree.h disk-io.h kerncompat.h print-tree.h
+headers = radix-tree.h ctree.h disk-io.h kerncompat.h print-tree.h list.h
 objects = ctree.o disk-io.o radix-tree.o mkfs.o extent-tree.o print-tree.o
 
 # if you don't have sparse installed, use ls instead
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 1118986d7569..72816381d203 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -50,16 +50,17 @@ int btrfs_cow_block(struct ctree_root *root,
 	memcpy(&cow->node, &buf->node, sizeof(buf->node));
 	cow->node.header.blocknr = cow->blocknr;
 	*cow_ret = cow;
+	btrfs_inc_ref(root, buf);
 	if (buf == root->node) {
 		root->node = cow;
 		cow->count++;
+		if (buf != root->commit_root)
+			free_extent(root, buf->blocknr, 1);
 		tree_block_release(root, buf);
 	} else {
 		parent->node.blockptrs[parent_slot] = cow->blocknr;
 		BUG_ON(list_empty(&parent->dirty));
-	}
-	if (0 && root != root->extent_root && !is_leaf(cow->node.header.flags)) {
-		btrfs_inc_ref(root, cow);
+		free_extent(root, buf->blocknr, 1);
 	}
 	tree_block_release(root, buf);
 	return 0;
@@ -1018,7 +1019,6 @@ static int split_leaf(struct ctree_root *root, struct ctree_path *path,
 	slot = path->slots[0];
 	nritems = l->header.nritems;
 	mid = (nritems + 1)/ 2;
-
 	right_buffer = alloc_free_block(root);
 	BUG_ON(!right_buffer);
 	BUG_ON(mid == nritems);
@@ -1170,7 +1170,6 @@ static int del_ptr(struct ctree_root *root, struct ctree_path *path, int level,
 
 	node = &parent->node;
 	nritems = node->header.nritems;
-
 	if (slot != nritems -1) {
 		memmove(node->keys + slot, node->keys + slot + 1,
 			sizeof(struct key) * (nritems - slot - 1));
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 9fe8ba6e25c1..4a7bc4e6e747 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -51,10 +51,12 @@ struct tree_buffer;
  */
 struct ctree_root {
 	struct tree_buffer *node;
+	struct tree_buffer *commit_root;
 	struct ctree_root *extent_root;
 	struct key current_insert;
 	int fp;
 	struct radix_tree_root cache_radix;
+	struct radix_tree_root pinned_radix;
 	struct list_head trans;
 	struct list_head cache;
 	int cache_size;
@@ -151,4 +153,6 @@ int del_item(struct ctree_root *root, struct ctree_path *path);
 int insert_item(struct ctree_root *root, struct key *key, void *data, int data_size);
 int next_leaf(struct ctree_root *root, struct ctree_path *path);
 int leaf_free_space(struct leaf *leaf);
+int btrfs_drop_snapshot(struct ctree_root *root, struct tree_buffer *snap);
+int btrfs_finish_extent_commit(struct ctree_root *root);
 #endif
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 0e1c31e682fb..2fe31c3508c1 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -153,13 +153,24 @@ static int __commit_transaction(struct ctree_root *root)
 	return ret;
 }
 
-int commit_transaction(struct ctree_root *root)
+int commit_transaction(struct ctree_root *root, struct ctree_super_block *s)
 {
-	int ret;
+	int ret = 0;
+
 	ret = __commit_transaction(root);
 	if (!ret && root != root->extent_root)
 		ret = __commit_transaction(root->extent_root);
 	BUG_ON(ret);
+	if (root->commit_root != root->node) {
+		struct tree_buffer *snap = root->commit_root;
+		root->commit_root = root->node;
+		root->node->count++;
+		ret = btrfs_drop_snapshot(root, snap);
+		BUG_ON(ret);
+		tree_block_release(root, snap);
+	}
+        write_ctree_super(root, s);
+	btrfs_finish_extent_commit(root);
 	return ret;
 }
 
@@ -168,10 +179,13 @@ static int __setup_root(struct ctree_root *root, struct ctree_root *extent_root,
 {
 	INIT_LIST_HEAD(&root->trans);
 	INIT_LIST_HEAD(&root->cache);
+	root->cache_size = 0;
 	root->fp = fp;
 	root->node = NULL;
-	root->node = read_tree_block(root, info->tree_root);
 	root->extent_root = extent_root;
+	root->commit_root = NULL;
+	root->node = read_tree_block(root, info->tree_root);
+	memset(&root->current_insert, 0, sizeof(root->current_insert));
 	return 0;
 }
 
@@ -188,6 +202,8 @@ struct ctree_root *open_ctree(char *filename, struct ctree_super_block *super)
 		return NULL;
 	}
 	INIT_RADIX_TREE(&root->cache_radix, GFP_KERNEL);
+	INIT_RADIX_TREE(&root->pinned_radix, GFP_KERNEL);
+	INIT_RADIX_TREE(&extent_root->pinned_radix, GFP_KERNEL);
 	INIT_RADIX_TREE(&extent_root->cache_radix, GFP_KERNEL);
 	ret = pread(fp, super, sizeof(struct ctree_super_block),
 		     CTREE_SUPER_INFO_OFFSET(CTREE_BLOCKSIZE));
@@ -204,6 +220,8 @@ struct ctree_root *open_ctree(char *filename, struct ctree_super_block *super)
 	BUG_ON(ret < 0);
 	__setup_root(root, extent_root, &super->root_info, fp);
 	__setup_root(extent_root, extent_root, &super->extent_info, fp);
+	root->commit_root = root->node;
+	root->node->count++;
 	return root;
 }
 
@@ -236,9 +254,11 @@ static int drop_cache(struct ctree_root *root)
 	}
 	return 0;
 }
-int close_ctree(struct ctree_root *root)
+int close_ctree(struct ctree_root *root, struct ctree_super_block *s)
 {
-	commit_transaction(root);
+	commit_transaction(root, s);
+	__commit_transaction(root->extent_root);
+	write_ctree_super(root, s);
 	drop_cache(root->extent_root);
 	drop_cache(root);
 	BUG_ON(!list_empty(&root->trans));
@@ -249,6 +269,7 @@ int close_ctree(struct ctree_root *root)
 		tree_block_release(root, root->node);
 	if (root->extent_root->node)
 		tree_block_release(root->extent_root, root->extent_root->node);
+	tree_block_release(root, root->commit_root);
 	free(root);
 	printf("on close %d blocks are allocated\n", allocated_blocks);
 	return 0;
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index b5dee2fae4da..1c0af7c56c21 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -18,9 +18,9 @@ struct tree_buffer *find_tree_block(struct ctree_root *root, u64 blocknr);
 int write_tree_block(struct ctree_root *root, struct tree_buffer *buf);
 int dirty_tree_block(struct ctree_root *root, struct tree_buffer *buf);
 int clean_tree_block(struct ctree_root *root, struct tree_buffer *buf);
-int commit_transaction(struct ctree_root *root);
+int commit_transaction(struct ctree_root *root, struct ctree_super_block *s);
 struct ctree_root *open_ctree(char *filename, struct ctree_super_block *s);
-int close_ctree(struct ctree_root *root);
+int close_ctree(struct ctree_root *root, struct ctree_super_block *s);
 void tree_block_release(struct ctree_root *root, struct tree_buffer *buf);
 int write_ctree_super(struct ctree_root *root, struct ctree_super_block *s);
 int mkfs(int fd);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 25d9cd169209..0723b7f3f0c3 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -13,7 +13,8 @@
  * other allocations are done.  The pending tag is also used in the same
  * manner for deletes.
  */
-#define CTREE_EXTENT_PENDING 0
+#define CTREE_EXTENT_PENDING_ADD 0
+#define CTREE_EXTENT_PENDING_DEL 1
 
 static int inc_block_ref(struct ctree_root *root, u64 blocknr)
 {
@@ -27,20 +28,51 @@ static int inc_block_ref(struct ctree_root *root, u64 blocknr)
 	key.flags = 0;
 	key.offset = 1;
 	ret = search_slot(root->extent_root, &key, &path, 0, 1);
+	if (ret != 0)
+		BUG();
 	BUG_ON(ret != 0);
 	l = &path.nodes[0]->leaf;
 	item = (struct extent_item *)(l->data +
 				      l->items[path.slots[0]].offset);
 	item->refs++;
+
 	BUG_ON(list_empty(&path.nodes[0]->dirty));
 	release_path(root->extent_root, &path);
 	return 0;
 }
 
+static int lookup_block_ref(struct ctree_root *root, u64 blocknr, int *refs)
+{
+	struct ctree_path path;
+	int ret;
+	struct key key;
+	struct leaf *l;
+	struct extent_item *item;
+	init_path(&path);
+	key.objectid = blocknr;
+	key.flags = 0;
+	key.offset = 1;
+	ret = search_slot(root->extent_root, &key, &path, 0, 0);
+	if (ret != 0)
+		BUG();
+	l = &path.nodes[0]->leaf;
+	item = (struct extent_item *)(l->data +
+				      l->items[path.slots[0]].offset);
+	*refs = item->refs;
+	release_path(root->extent_root, &path);
+	return 0;
+}
+
 int btrfs_inc_ref(struct ctree_root *root, struct tree_buffer *buf)
 {
 	u64 blocknr;
 	int i;
+
+	if (root == root->extent_root)
+		return 0;
+	if (is_leaf(buf->node.header.flags))
+		return 0;
+
 	for (i = 0; i < buf->node.header.nritems; i++) {
 		blocknr = buf->node.blockptrs[i];
 		inc_block_ref(root, blocknr);
@@ -48,85 +80,187 @@ int btrfs_inc_ref(struct ctree_root *root, struct tree_buffer *buf)
 	return 0;
 }
 
+int btrfs_finish_extent_commit(struct ctree_root *root)
+{
+	struct ctree_root *extent_root = root->extent_root;
+	unsigned long gang[8];
+	int ret;
+	int i;
+
+	while(1) {
+		ret = radix_tree_gang_lookup(&extent_root->pinned_radix,
+						 (void **)gang, 0,
+						 ARRAY_SIZE(gang));
+		if (!ret)
+			break;
+		for (i = 0; i < ret; i++)
+			radix_tree_delete(&extent_root->pinned_radix, gang[i]);
+	}
+	return 0;
+}
+
 /*
- * find all the blocks marked as pending in the radix tree and remove
- * them from the extent map
+ * remove an extent from the root, returns 0 on success
  */
-static int del_pending_extents(struct ctree_root *extent_root)
+int __free_extent(struct ctree_root *root, u64 blocknr, u64 num_blocks)
+{
+	struct ctree_path path;
+	struct key key;
+	struct ctree_root *extent_root = root->extent_root;
+	int ret;
+	struct item *item;
+	struct extent_item *ei;
+	key.objectid = blocknr;
+	key.flags = 0;
+	key.offset = num_blocks;
+
+	init_path(&path);
+	ret = search_slot(extent_root, &key, &path, -1, 1);
+	if (ret) {
+		printf("failed to find %Lu\n", key.objectid);
+		print_tree(extent_root, extent_root->node);
+		printf("failed to find %Lu\n", key.objectid);
+		BUG();
+	}
+	item = path.nodes[0]->leaf.items + path.slots[0];
+	ei = (struct extent_item *)(path.nodes[0]->leaf.data + item->offset);
+	BUG_ON(ei->refs == 0);
+	ei->refs--;
+	if (ei->refs == 0) {
+		if (root == extent_root) {
+			int err;
+			radix_tree_preload(GFP_KERNEL);
+			err = radix_tree_insert(&extent_root->pinned_radix,
+					  blocknr, (void *)blocknr);
+			BUG_ON(err);
+			radix_tree_preload_end();
+		}
+		ret = del_item(extent_root, &path);
+		if (ret)
+			BUG();
+	}
+	release_path(extent_root, &path);
+	return ret;
+}
+
+/*
+ * insert all of the pending extents reserved during the original
+ * allocation.  (CTREE_EXTENT_PENDING).  Returns zero if it all worked out
+ */
+static int insert_pending_extents(struct ctree_root *extent_root)
 {
 	int ret;
 	struct key key;
+	struct extent_item item;
 	struct tree_buffer *gang[4];
 	int i;
-	struct ctree_path path;
 
+	// FIXME -ENOSPC
+	item.owner = extent_root->node->node.header.parentid;
+	item.refs = 1;
 	while(1) {
 		ret = radix_tree_gang_lookup_tag(&extent_root->cache_radix,
 						 (void **)gang, 0,
 						 ARRAY_SIZE(gang),
-						 CTREE_EXTENT_PENDING);
+						 CTREE_EXTENT_PENDING_ADD);
 		if (!ret)
 			break;
 		for (i = 0; i < ret; i++) {
 			key.objectid = gang[i]->blocknr;
 			key.flags = 0;
 			key.offset = 1;
-			init_path(&path);
-			ret = search_slot(extent_root, &key, &path, -1, 1);
+			ret = insert_item(extent_root, &key, &item,
+					  sizeof(item));
 			if (ret) {
+				printf("%Lu already in tree\n", key.objectid);
 				print_tree(extent_root, extent_root->node);
-				printf("unable to find %Lu\n", key.objectid);
 				BUG();
 				// FIXME undo it and return sane
 				return ret;
 			}
-			ret = del_item(extent_root, &path);
-			if (ret) {
-				BUG();
-				return ret;
-			}
-			release_path(extent_root, &path);
+			radix_tree_tag_clear(&extent_root->cache_radix,
+					     gang[i]->blocknr,
+					     CTREE_EXTENT_PENDING_ADD);
+			tree_block_release(extent_root, gang[i]);
+		}
+	}
+	return 0;
+}
+
+/*
+ * find all the blocks marked as pending in the radix tree and remove
+ * them from the extent map
+ */
+static int del_pending_extents(struct ctree_root *extent_root)
+{
+	int ret;
+	struct tree_buffer *gang[4];
+	int i;
+
+	while(1) {
+		ret = radix_tree_gang_lookup_tag(&extent_root->cache_radix,
+						 (void **)gang, 0,
+						 ARRAY_SIZE(gang),
+						 CTREE_EXTENT_PENDING_DEL);
+		if (!ret)
+			break;
+		for (i = 0; i < ret; i++) {
+			ret = __free_extent(extent_root, gang[i]->blocknr, 1);
 			radix_tree_tag_clear(&extent_root->cache_radix,
 						gang[i]->blocknr,
-						CTREE_EXTENT_PENDING);
+						CTREE_EXTENT_PENDING_DEL);
 			tree_block_release(extent_root, gang[i]);
 		}
 	}
 	return 0;
 }
 
+static int run_pending(struct ctree_root *extent_root)
+{
+	while(radix_tree_tagged(&extent_root->cache_radix,
+			        CTREE_EXTENT_PENDING_DEL) ||
+	      radix_tree_tagged(&extent_root->cache_radix,
+				CTREE_EXTENT_PENDING_ADD)) {
+		insert_pending_extents(extent_root);
+		del_pending_extents(extent_root);
+	}
+	return 0;
+}
+
+
 /*
  * remove an extent from the root, returns 0 on success
  */
 int free_extent(struct ctree_root *root, u64 blocknr, u64 num_blocks)
 {
-	struct ctree_path path;
 	struct key key;
 	struct ctree_root *extent_root = root->extent_root;
 	struct tree_buffer *t;
 	int pending_ret;
 	int ret;
-	key.objectid = blocknr;
-	key.flags = 0;
-	key.offset = num_blocks;
+
 	if (root == extent_root) {
-		t = read_tree_block(root, key.objectid);
-		radix_tree_tag_set(&root->cache_radix, key.objectid,
-				   CTREE_EXTENT_PENDING);
+		t = find_tree_block(root, blocknr);
+		if (radix_tree_tag_get(&root->cache_radix, blocknr,
+				      CTREE_EXTENT_PENDING_ADD)) {
+			radix_tree_tag_clear(&root->cache_radix,
+					     blocknr,
+					     CTREE_EXTENT_PENDING_ADD);
+			/* once for us */
+			tree_block_release(root, t);
+			/* once for the pending add */
+			tree_block_release(root, t);
+		} else {
+			radix_tree_tag_set(&root->cache_radix, blocknr,
+				   CTREE_EXTENT_PENDING_DEL);
+		}
 		return 0;
 	}
-	init_path(&path);
-	ret = search_slot(extent_root, &key, &path, -1, 1);
-	if (ret) {
-		print_tree(extent_root, extent_root->node);
-		printf("failed to find %Lu\n", key.objectid);
-		BUG();
-	}
-	ret = del_item(extent_root, &path);
-	if (ret)
-		BUG();
-	release_path(extent_root, &path);
-	pending_ret = del_pending_extents(root->extent_root);
+	key.objectid = blocknr;
+	key.flags = 0;
+	key.offset = num_blocks;
+	ret = __free_extent(root, blocknr, num_blocks);
+	pending_ret = run_pending(root->extent_root);
 	return ret ? ret : pending_ret;
 }
 
@@ -203,7 +337,7 @@ check_pending:
 	 */
 	release_path(root, &path);
 	BUG_ON(ins->objectid < search_start);
-	if (orig_root->extent_root == orig_root) {
+	if (1 || orig_root->extent_root == orig_root) {
 		BUG_ON(num_blocks != 1);
 		if ((root->current_insert.objectid <= ins->objectid &&
 		    root->current_insert.objectid +
@@ -211,8 +345,9 @@ check_pending:
 		   (root->current_insert.objectid > ins->objectid &&
 		    root->current_insert.objectid <= ins->objectid +
 		    ins->offset) ||
+		   radix_tree_lookup(&root->pinned_radix, ins->objectid) ||
 		   radix_tree_tag_get(&root->cache_radix, ins->objectid,
-				      CTREE_EXTENT_PENDING)) {
+				      CTREE_EXTENT_PENDING_ADD)) {
 			search_start = ins->objectid + 1;
 			goto check_failed;
 		}
@@ -225,51 +360,6 @@ error:
 	return ret;
 }
 
-/*
- * insert all of the pending extents reserved during the original
- * allocation.  (CTREE_EXTENT_PENDING).  Returns zero if it all worked out
- */
-static int insert_pending_extents(struct ctree_root *extent_root)
-{
-	int ret;
-	struct key key;
-	struct extent_item item;
-	struct tree_buffer *gang[4];
-	int i;
-
-	// FIXME -ENOSPC
-	item.refs = 1;
-	item.owner = extent_root->node->node.header.parentid;
-	while(1) {
-		ret = radix_tree_gang_lookup_tag(&extent_root->cache_radix,
-						 (void **)gang, 0,
-						 ARRAY_SIZE(gang),
-						 CTREE_EXTENT_PENDING);
-		if (!ret)
-			break;
-		for (i = 0; i < ret; i++) {
-			key.objectid = gang[i]->blocknr;
-			key.flags = 0;
-			key.offset = 1;
-			ret = insert_item(extent_root, &key, &item,
-					  sizeof(item));
-			if (ret) {
-				printf("%Lu already in tree\n", key.objectid);
-				print_tree(extent_root, extent_root->node);
-				BUG();
-				// FIXME undo it and return sane
-				return ret;
-			}
-			radix_tree_tag_clear(&extent_root->cache_radix,
-					     gang[i]->blocknr,
-					     CTREE_EXTENT_PENDING);
-			printf("%Lu is not pending\n", gang[i]->blocknr);
-			tree_block_release(extent_root, gang[i]);
-		}
-	}
-	return 0;
-}
-
 /*
  * finds a free extent and does all the dirty work required for allocation
  * returns the key for the extent through ins, and a tree buffer for
@@ -296,7 +386,7 @@ int alloc_extent(struct ctree_root *root, u64 num_blocks, u64 search_start,
 				  sizeof(extent_item));
 		memset(&root->extent_root->current_insert, 0,
 		       sizeof(struct key));
-		pending_ret = insert_pending_extents(root->extent_root);
+		pending_ret = run_pending(root->extent_root);
 		if (ret)
 			return ret;
 		if (pending_ret)
@@ -309,9 +399,8 @@ int alloc_extent(struct ctree_root *root, u64 num_blocks, u64 search_start,
 	BUG_ON(ins->offset != 1);
 	*buf = find_tree_block(root, ins->objectid);
 	BUG_ON(!*buf);
-	printf("%Lu is pending\n", ins->objectid);
 	radix_tree_tag_set(&root->cache_radix, ins->objectid,
-			   CTREE_EXTENT_PENDING);
+			   CTREE_EXTENT_PENDING_ADD);
 	(*buf)->count++;
 	dirty_tree_block(root, *buf);
 	return 0;
@@ -331,13 +420,41 @@ struct tree_buffer *alloc_free_block(struct ctree_root *root)
 	ret = alloc_extent(root, 1, 0, (unsigned long)-1,
 			   root->node->node.header.parentid,
 			   &ins, &buf);
-
 	if (ret) {
 		BUG();
 		return NULL;
 	}
 	if (root != root->extent_root)
 		BUG_ON(radix_tree_tag_get(&root->extent_root->cache_radix,
-					  buf->blocknr, CTREE_EXTENT_PENDING));
+					  buf->blocknr,
+					  CTREE_EXTENT_PENDING_ADD));
 	return buf;
 }
+
+int btrfs_drop_snapshot(struct ctree_root *root, struct tree_buffer *snap)
+{
+	int ret;
+	int level;
+	int refs;
+	u64 blocknr = snap->blocknr;
+
+	level = node_level(snap->node.header.flags);
+	ret = lookup_block_ref(root, snap->blocknr, &refs);
+	BUG_ON(ret);
+	if (refs == 1 && level != 0) {
+		struct node *n = &snap->node;
+		struct tree_buffer *b;
+		int i;
+		for (i = 0; i < n->header.nritems; i++) {
+			b = read_tree_block(root, n->blockptrs[i]);
+			/* FIXME, don't recurse here */
+			ret = btrfs_drop_snapshot(root, b);
+			BUG_ON(ret);
+			tree_block_release(root, b);
+		}
+	}
+	ret = free_extent(root, blocknr, 1);
+	BUG_ON(ret);
+	return 0;
+}
+
diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c
index dda08f32c154..e32a959dd3e5 100644
--- a/fs/btrfs/print-tree.c
+++ b/fs/btrfs/print-tree.c
@@ -21,9 +21,11 @@ void print_leaf(struct leaf *l)
 			item->key.objectid, item->key.flags, item->key.offset,
 			item->offset, item->size);
 		fflush(stdout);
-		printf("\t\titem data %.*s\n", item->size, l->data+item->offset);
+		printf("\t\titem data %.*s\n", item->size,
+			l->data+item->offset);
 		ei = (struct extent_item *)(l->data + item->offset);
-		printf("\t\textent data %u %Lu\n", ei->refs, ei->owner);
+		printf("\t\textent data refs %u owner %Lu\n", ei->refs,
+			ei->owner);
 		fflush(stdout);
 	}
 }
diff --git a/fs/btrfs/quick-test.c b/fs/btrfs/quick-test.c
index 8255f79ceca5..6400c7100a6a 100644
--- a/fs/btrfs/quick-test.c
+++ b/fs/btrfs/quick-test.c
@@ -19,7 +19,7 @@ int main(int ac, char **av) {
 	int i;
 	int num;
 	int ret;
-	int run_size = 1024;
+	int run_size = 100000;
 	int max_key =  100000000;
 	int tree_size = 0;
 	struct ctree_path path;
@@ -44,9 +44,9 @@ int main(int ac, char **av) {
 		if (!ret)
 			tree_size++;
 		free(buf);
+
 	}
-	write_ctree_super(root, &super);
-	close_ctree(root);
+	close_ctree(root, &super);
 
 	root = open_ctree("dbfile", &super);
 	printf("starting search\n");
@@ -65,8 +65,7 @@ int main(int ac, char **av) {
 		}
 		release_path(root, &path);
 	}
-	write_ctree_super(root, &super);
-	close_ctree(root);
+	close_ctree(root, &super);
 	root = open_ctree("dbfile", &super);
 	printf("node %p level %d total ptrs %d free spc %lu\n", root->node,
 	        node_level(root->node->node.header.flags),
@@ -90,8 +89,7 @@ int main(int ac, char **av) {
 		}
 		release_path(root, &path);
 	}
-	write_ctree_super(root, &super);
-	close_ctree(root);
+	close_ctree(root, &super);
 	root = open_ctree("dbfile", &super);
 	srand(128);
 	for (i = 0; i < run_size; i++) {
@@ -106,8 +104,7 @@ int main(int ac, char **av) {
 			tree_size++;
 		free(buf);
 	}
-	write_ctree_super(root, &super);
-	close_ctree(root);
+	close_ctree(root, &super);
 	root = open_ctree("dbfile", &super);
 	srand(128);
 	printf("starting search2\n");
@@ -156,10 +153,17 @@ int main(int ac, char **av) {
 		}
 		release_path(root, &path);
 	}
+	/*
+	printf("previous tree:\n");
+	print_tree(root, root->commit_root);
+	printf("map before commit\n");
+	print_tree(root->extent_root, root->extent_root->node);
+	*/
+	commit_transaction(root, &super);
 	printf("tree size is now %d\n", tree_size);
+	printf("root %p commit root %p\n", root->node, root->commit_root);
 	printf("map tree\n");
 	print_tree(root->extent_root, root->extent_root->node);
-	write_ctree_super(root, &super);
-	close_ctree(root);
+	close_ctree(root, &super);
 	return 0;
 }
diff --git a/fs/btrfs/random-test.c b/fs/btrfs/random-test.c
index dcc852ad6737..7b37b6bae105 100644
--- a/fs/btrfs/random-test.c
+++ b/fs/btrfs/random-test.c
@@ -8,6 +8,7 @@
 #include "print-tree.h"
 
 int keep_running = 1;
+struct ctree_super_block super;
 
 static int setup_key(struct radix_tree_root *root, struct key *key, int exists)
 {
@@ -59,11 +60,6 @@ error:
 	return -1;
 }
 
-static int run_commit(struct ctree_root *root, struct radix_tree_root *radix)
-{
-	return commit_transaction(root);
-}
-
 static int insert_dup(struct ctree_root *root, struct radix_tree_root *radix)
 {
 	struct ctree_path path;
@@ -210,7 +206,7 @@ static int fill_tree(struct ctree_root *root, struct radix_tree_root *radix,
 			goto out;
 		}
 		if (i % 1000 == 0) {
-			ret = commit_transaction(root);
+			ret = commit_transaction(root, &super);
 			if (ret) {
 				fprintf(stderr, "fill commit failed\n");
 				return ret;
@@ -229,7 +225,7 @@ out:
 static int bulk_op(struct ctree_root *root, struct radix_tree_root *radix)
 {
 	int ret;
-	int nr = rand() % 20000;
+	int nr = rand() % 5000;
 	static int run_nr = 0;
 
 	/* do the bulk op much less frequently */
@@ -247,7 +243,7 @@ static int bulk_op(struct ctree_root *root, struct radix_tree_root *radix)
 
 int (*ops[])(struct ctree_root *root, struct radix_tree_root *radix) =
 	{ ins_one, insert_dup, del_one, lookup_item,
-	  lookup_enoent, bulk_op, run_commit };
+	  lookup_enoent, bulk_op };
 
 static int fill_radix(struct ctree_root *root, struct radix_tree_root *radix)
 {
@@ -314,7 +310,6 @@ int print_usage(void)
 int main(int ac, char **av)
 {
 	RADIX_TREE(radix, GFP_KERNEL);
-	struct ctree_super_block super;
 	struct ctree_root *root;
 	int i;
 	int ret;
@@ -365,8 +360,7 @@ int main(int ac, char **av)
 			printf("open & close, root level %d nritems %d\n",
 				node_level(root->node->node.header.flags),
 				root->node->node.header.nritems);
-			write_ctree_super(root, &super);
-			close_ctree(root);
+			close_ctree(root, &super);
 			root = open_ctree("dbfile", &super);
 		}
 		while(count--) {
@@ -380,7 +374,7 @@ int main(int ac, char **av)
 				err = ret;
 				goto out;
 			}
-			if (ops[op] == bulk_op || ops[op] == run_commit)
+			if (ops[op] == bulk_op)
 				break;
 			if (keep_running == 0) {
 				err = 0;
@@ -389,8 +383,7 @@ int main(int ac, char **av)
 		}
 	}
 out:
-	write_ctree_super(root, &super);
-	close_ctree(root);
+	close_ctree(root, &super);
 	return err;
 }
 
-- 
cgit v1.2.3


From 234b63a091e1df6bd4261dd7b3a7490074830628 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Tue, 13 Mar 2007 10:46:10 -0400
Subject: rename funcs and structs to btrfs

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.c       | 317 +++++++++++++++++++++++++------------------------
 fs/btrfs/ctree.h       |  95 +++++++--------
 fs/btrfs/debug-tree.c  |   8 +-
 fs/btrfs/disk-io.c     | 106 +++++++++--------
 fs/btrfs/disk-io.h     |  29 ++---
 fs/btrfs/extent-tree.c | 147 +++++++++++------------
 fs/btrfs/mkfs.c        |  26 ++--
 fs/btrfs/print-tree.c  |  23 ++--
 fs/btrfs/print-tree.h  |   4 +-
 fs/btrfs/quick-test.c  |  54 ++++-----
 fs/btrfs/random-test.c |  92 +++++++-------
 11 files changed, 454 insertions(+), 447 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 762810731001..c8529b394d5f 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -6,47 +6,47 @@
 #include "disk-io.h"
 #include "print-tree.h"
 
-static int split_node(struct ctree_root *root, struct ctree_path *path,
+static int split_node(struct btrfs_root *root, struct btrfs_path *path,
 		      int level);
-static int split_leaf(struct ctree_root *root, struct ctree_path *path,
+static int split_leaf(struct btrfs_root *root, struct btrfs_path *path,
 		      int data_size);
-static int push_node_left(struct ctree_root *root, struct tree_buffer *dst,
-			  struct tree_buffer *src);
-static int balance_node_right(struct ctree_root *root,
-			      struct tree_buffer *dst_buf,
-			      struct tree_buffer *src_buf);
-static int del_ptr(struct ctree_root *root, struct ctree_path *path, int level,
+static int push_node_left(struct btrfs_root *root, struct btrfs_buffer *dst,
+			  struct btrfs_buffer *src);
+static int balance_node_right(struct btrfs_root *root,
+			      struct btrfs_buffer *dst_buf,
+			      struct btrfs_buffer *src_buf);
+static int del_ptr(struct btrfs_root *root, struct btrfs_path *path, int level,
 		   int slot);
 
-inline void init_path(struct ctree_path *p)
+inline void btrfs_init_path(struct btrfs_path *p)
 {
 	memset(p, 0, sizeof(*p));
 }
 
-void release_path(struct ctree_root *root, struct ctree_path *p)
+void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p)
 {
 	int i;
-	for (i = 0; i < MAX_LEVEL; i++) {
+	for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
 		if (!p->nodes[i])
 			break;
-		tree_block_release(root, p->nodes[i]);
+		btrfs_block_release(root, p->nodes[i]);
 	}
 	memset(p, 0, sizeof(*p));
 }
 
-int btrfs_cow_block(struct ctree_root *root,
-		    struct tree_buffer *buf,
-		    struct tree_buffer *parent,
+int btrfs_cow_block(struct btrfs_root *root,
+		    struct btrfs_buffer *buf,
+		    struct btrfs_buffer *parent,
 		    int parent_slot,
-		    struct tree_buffer **cow_ret)
+		    struct btrfs_buffer **cow_ret)
 {
-	struct tree_buffer *cow;
+	struct btrfs_buffer *cow;
 
 	if (!list_empty(&buf->dirty)) {
 		*cow_ret = buf;
 		return 0;
 	}
-	cow = alloc_free_block(root);
+	cow = btrfs_alloc_free_block(root);
 	memcpy(&cow->node, &buf->node, sizeof(buf->node));
 	btrfs_set_header_blocknr(&cow->node.header, cow->blocknr);
 	*cow_ret = cow;
@@ -55,15 +55,15 @@ int btrfs_cow_block(struct ctree_root *root,
 		root->node = cow;
 		cow->count++;
 		if (buf != root->commit_root)
-			free_extent(root, buf->blocknr, 1);
-		tree_block_release(root, buf);
+			btrfs_free_extent(root, buf->blocknr, 1);
+		btrfs_block_release(root, buf);
 	} else {
 		btrfs_set_node_blockptr(&parent->node, parent_slot,
 					cow->blocknr);
 		BUG_ON(list_empty(&parent->dirty));
-		free_extent(root, buf->blocknr, 1);
+		btrfs_free_extent(root, buf->blocknr, 1);
 	}
-	tree_block_release(root, buf);
+	btrfs_block_release(root, buf);
 	return 0;
 }
 
@@ -72,7 +72,7 @@ int btrfs_cow_block(struct ctree_root *root,
  * this returns the address of the start of the last item,
  * which is the stop of the leaf data stack
  */
-static inline unsigned int leaf_data_end(struct leaf *leaf)
+static inline unsigned int leaf_data_end(struct btrfs_leaf *leaf)
 {
 	u32 nr = btrfs_header_nritems(&leaf->header);
 	if (nr == 0)
@@ -85,7 +85,7 @@ static inline unsigned int leaf_data_end(struct leaf *leaf)
  * the start of the leaf data.  IOW, how much room
  * the leaf has left for both items and data
  */
-int leaf_free_space(struct leaf *leaf)
+int btrfs_leaf_free_space(struct btrfs_leaf *leaf)
 {
 	int data_end = leaf_data_end(leaf);
 	int nritems = btrfs_header_nritems(&leaf->header);
@@ -117,11 +117,11 @@ int comp_keys(struct btrfs_disk_key *disk, struct btrfs_key *k2)
 	return 0;
 }
 
-int check_node(struct ctree_path *path, int level)
+int check_node(struct btrfs_path *path, int level)
 {
 	int i;
-	struct node *parent = NULL;
-	struct node *node = &path->nodes[level]->node;
+	struct btrfs_node *parent = NULL;
+	struct btrfs_node *node = &path->nodes[level]->node;
 	int parent_slot;
 	u32 nritems = btrfs_header_nritems(&node->header);
 
@@ -146,18 +146,18 @@ int check_node(struct ctree_path *path, int level)
 	return 0;
 }
 
-int check_leaf(struct ctree_path *path, int level)
+int check_leaf(struct btrfs_path *path, int level)
 {
 	int i;
-	struct leaf *leaf = &path->nodes[level]->leaf;
-	struct node *parent = NULL;
+	struct btrfs_leaf *leaf = &path->nodes[level]->leaf;
+	struct btrfs_node *parent = NULL;
 	int parent_slot;
 	u32 nritems = btrfs_header_nritems(&leaf->header);
 
 	if (path->nodes[level + 1])
 		parent = &path->nodes[level + 1]->node;
 	parent_slot = path->slots[level + 1];
-	BUG_ON(leaf_free_space(leaf) < 0);
+	BUG_ON(btrfs_leaf_free_space(leaf) < 0);
 
 	if (nritems == 0)
 		return 0;
@@ -186,7 +186,7 @@ int check_leaf(struct ctree_path *path, int level)
 	return 0;
 }
 
-int check_block(struct ctree_path *path, int level)
+int check_block(struct btrfs_path *path, int level)
 {
 	if (level == 0)
 		return check_leaf(path, level);
@@ -233,10 +233,10 @@ int generic_bin_search(char *p, int item_size, struct btrfs_key *key,
  * simple bin_search frontend that does the right thing for
  * leaves vs nodes
  */
-int bin_search(struct node *c, struct btrfs_key *key, int *slot)
+int bin_search(struct btrfs_node *c, struct btrfs_key *key, int *slot)
 {
 	if (btrfs_is_leaf(c)) {
-		struct leaf *l = (struct leaf *)c;
+		struct btrfs_leaf *l = (struct btrfs_leaf *)c;
 		return generic_bin_search((void *)l->items,
 					  sizeof(struct btrfs_item),
 					  key, btrfs_header_nritems(&c->header),
@@ -250,11 +250,11 @@ int bin_search(struct node *c, struct btrfs_key *key, int *slot)
 	return -1;
 }
 
-struct tree_buffer *read_node_slot(struct ctree_root *root,
-				   struct tree_buffer *parent_buf,
+struct btrfs_buffer *read_node_slot(struct btrfs_root *root,
+				   struct btrfs_buffer *parent_buf,
 				   int slot)
 {
-	struct node *node = &parent_buf->node;
+	struct btrfs_node *node = &parent_buf->node;
 	if (slot < 0)
 		return NULL;
 	if (slot >= btrfs_header_nritems(&node->header))
@@ -262,17 +262,17 @@ struct tree_buffer *read_node_slot(struct ctree_root *root,
 	return read_tree_block(root, btrfs_node_blockptr(node, slot));
 }
 
-static int balance_level(struct ctree_root *root, struct ctree_path *path,
+static int balance_level(struct btrfs_root *root, struct btrfs_path *path,
 			int level)
 {
-	struct tree_buffer *right_buf;
-	struct tree_buffer *mid_buf;
-	struct tree_buffer *left_buf;
-	struct tree_buffer *parent_buf = NULL;
-	struct node *right = NULL;
-	struct node *mid;
-	struct node *left = NULL;
-	struct node *parent = NULL;
+	struct btrfs_buffer *right_buf;
+	struct btrfs_buffer *mid_buf;
+	struct btrfs_buffer *left_buf;
+	struct btrfs_buffer *parent_buf = NULL;
+	struct btrfs_node *right = NULL;
+	struct btrfs_node *mid;
+	struct btrfs_node *left = NULL;
+	struct btrfs_node *parent = NULL;
 	int ret = 0;
 	int wret;
 	int pslot;
@@ -286,12 +286,12 @@ static int balance_level(struct ctree_root *root, struct ctree_path *path,
 	mid = &mid_buf->node;
 	orig_ptr = btrfs_node_blockptr(mid, orig_slot);
 
-	if (level < MAX_LEVEL - 1)
+	if (level < BTRFS_MAX_LEVEL - 1)
 		parent_buf = path->nodes[level + 1];
 	pslot = path->slots[level + 1];
 
 	if (!parent_buf) {
-		struct tree_buffer *child;
+		struct btrfs_buffer *child;
 		u64 blocknr = mid_buf->blocknr;
 
 		if (btrfs_header_nritems(&mid->header) != 1)
@@ -303,11 +303,11 @@ static int balance_level(struct ctree_root *root, struct ctree_path *path,
 		root->node = child;
 		path->nodes[level] = NULL;
 		/* once for the path */
-		tree_block_release(root, mid_buf);
+		btrfs_block_release(root, mid_buf);
 		/* once for the root ptr */
-		tree_block_release(root, mid_buf);
+		btrfs_block_release(root, mid_buf);
 		clean_tree_block(root, mid_buf);
-		return free_extent(root, blocknr, 1);
+		return btrfs_free_extent(root, blocknr, 1);
 	}
 	parent = &parent_buf->node;
 
@@ -340,14 +340,14 @@ static int balance_level(struct ctree_root *root, struct ctree_path *path,
 			ret = wret;
 		if (btrfs_header_nritems(&right->header) == 0) {
 			u64 blocknr = right_buf->blocknr;
-			tree_block_release(root, right_buf);
+			btrfs_block_release(root, right_buf);
 			clean_tree_block(root, right_buf);
 			right_buf = NULL;
 			right = NULL;
 			wret = del_ptr(root, path, level + 1, pslot + 1);
 			if (wret)
 				ret = wret;
-			wret = free_extent(root, blocknr, 1);
+			wret = btrfs_free_extent(root, blocknr, 1);
 			if (wret)
 				ret = wret;
 		} else {
@@ -375,14 +375,14 @@ static int balance_level(struct ctree_root *root, struct ctree_path *path,
 	if (btrfs_header_nritems(&mid->header) == 0) {
 		/* we've managed to empty the middle node, drop it */
 		u64 blocknr = mid_buf->blocknr;
-		tree_block_release(root, mid_buf);
+		btrfs_block_release(root, mid_buf);
 		clean_tree_block(root, mid_buf);
 		mid_buf = NULL;
 		mid = NULL;
 		wret = del_ptr(root, path, level + 1, pslot);
 		if (wret)
 			ret = wret;
-		wret = free_extent(root, blocknr, 1);
+		wret = btrfs_free_extent(root, blocknr, 1);
 		if (wret)
 			ret = wret;
 	} else {
@@ -400,7 +400,7 @@ static int balance_level(struct ctree_root *root, struct ctree_path *path,
 			path->slots[level + 1] -= 1;
 			path->slots[level] = orig_slot;
 			if (mid_buf)
-				tree_block_release(root, mid_buf);
+				btrfs_block_release(root, mid_buf);
 		} else {
 			orig_slot -= btrfs_header_nritems(&left->header);
 			path->slots[level] = orig_slot;
@@ -413,9 +413,9 @@ static int balance_level(struct ctree_root *root, struct ctree_path *path,
 		BUG();
 
 	if (right_buf)
-		tree_block_release(root, right_buf);
+		btrfs_block_release(root, right_buf);
 	if (left_buf)
-		tree_block_release(root, left_buf);
+		btrfs_block_release(root, left_buf);
 	return ret;
 }
 
@@ -432,12 +432,12 @@ static int balance_level(struct ctree_root *root, struct ctree_path *path,
  * tree.  if ins_len < 0, nodes will be merged as we walk down the tree (if
  * possible)
  */
-int search_slot(struct ctree_root *root, struct btrfs_key *key,
-		struct ctree_path *p, int ins_len, int cow)
+int btrfs_search_slot(struct btrfs_root *root, struct btrfs_key *key,
+		struct btrfs_path *p, int ins_len, int cow)
 {
-	struct tree_buffer *b;
-	struct tree_buffer *cow_buf;
-	struct node *c;
+	struct btrfs_buffer *b;
+	struct btrfs_buffer *cow_buf;
+	struct btrfs_node *c;
 	int slot;
 	int ret;
 	int level;
@@ -486,9 +486,9 @@ again:
 			}
 			b = read_tree_block(root, btrfs_node_blockptr(c, slot));
 		} else {
-			struct leaf *l = (struct leaf *)c;
+			struct btrfs_leaf *l = (struct btrfs_leaf *)c;
 			p->slots[level] = slot;
-			if (ins_len > 0 && leaf_free_space(l) <
+			if (ins_len > 0 && btrfs_leaf_free_space(l) <
 			    sizeof(struct btrfs_item) + ins_len) {
 				int sret = split_leaf(root, p, ins_len);
 				BUG_ON(sret > 0);
@@ -513,14 +513,14 @@ again:
  * If this fails to write a tree block, it returns -1, but continues
  * fixing up the blocks in ram so the tree is consistent.
  */
-static int fixup_low_keys(struct ctree_root *root,
-			   struct ctree_path *path, struct btrfs_disk_key *key,
+static int fixup_low_keys(struct btrfs_root *root,
+			   struct btrfs_path *path, struct btrfs_disk_key *key,
 			   int level)
 {
 	int i;
 	int ret = 0;
-	for (i = level; i < MAX_LEVEL; i++) {
-		struct node *t;
+	for (i = level; i < BTRFS_MAX_LEVEL; i++) {
+		struct btrfs_node *t;
 		int tslot = path->slots[i];
 		if (!path->nodes[i])
 			break;
@@ -540,11 +540,11 @@ static int fixup_low_keys(struct ctree_root *root,
  * returns 0 if some ptrs were pushed left, < 0 if there was some horrible
  * error, and > 0 if there was no room in the left hand block.
  */
-static int push_node_left(struct ctree_root *root, struct tree_buffer *dst_buf,
-			  struct tree_buffer *src_buf)
+static int push_node_left(struct btrfs_root *root, struct btrfs_buffer *dst_buf,
+			  struct btrfs_buffer *src_buf)
 {
-	struct node *src = &src_buf->node;
-	struct node *dst = &dst_buf->node;
+	struct btrfs_node *src = &src_buf->node;
+	struct btrfs_node *dst = &dst_buf->node;
 	int push_items = 0;
 	int src_nritems;
 	int dst_nritems;
@@ -587,12 +587,12 @@ static int push_node_left(struct ctree_root *root, struct tree_buffer *dst_buf,
  *
  * this will  only push up to 1/2 the contents of the left node over
  */
-static int balance_node_right(struct ctree_root *root,
-			      struct tree_buffer *dst_buf,
-			      struct tree_buffer *src_buf)
+static int balance_node_right(struct btrfs_root *root,
+			      struct btrfs_buffer *dst_buf,
+			      struct btrfs_buffer *src_buf)
 {
-	struct node *src = &src_buf->node;
-	struct node *dst = &dst_buf->node;
+	struct btrfs_node *src = &src_buf->node;
+	struct btrfs_node *dst = &dst_buf->node;
 	int push_items = 0;
 	int max_push;
 	int src_nritems;
@@ -637,18 +637,18 @@ static int balance_node_right(struct ctree_root *root,
  *
  * returns zero on success or < 0 on failure.
  */
-static int insert_new_root(struct ctree_root *root,
-			   struct ctree_path *path, int level)
+static int insert_new_root(struct btrfs_root *root,
+			   struct btrfs_path *path, int level)
 {
-	struct tree_buffer *t;
-	struct node *lower;
-	struct node *c;
+	struct btrfs_buffer *t;
+	struct btrfs_node *lower;
+	struct btrfs_node *c;
 	struct btrfs_disk_key *lower_key;
 
 	BUG_ON(path->nodes[level]);
 	BUG_ON(path->nodes[level-1] != root->node);
 
-	t = alloc_free_block(root);
+	t = btrfs_alloc_free_block(root);
 	c = &t->node;
 	memset(c, 0, sizeof(c));
 	btrfs_set_header_nritems(&c->header, 1);
@@ -658,13 +658,13 @@ static int insert_new_root(struct ctree_root *root,
 	                       btrfs_header_parentid(&root->node->node.header));
 	lower = &path->nodes[level-1]->node;
 	if (btrfs_is_leaf(lower))
-		lower_key = &((struct leaf *)lower)->items[0].key;
+		lower_key = &((struct btrfs_leaf *)lower)->items[0].key;
 	else
 		lower_key = lower->keys;
 	memcpy(c->keys, lower_key, sizeof(struct btrfs_disk_key));
 	btrfs_set_node_blockptr(c, 0, path->nodes[level - 1]->blocknr);
 	/* the super has an extra ref to root->node */
-	tree_block_release(root, root->node);
+	btrfs_block_release(root, root->node);
 	root->node = t;
 	t->count++;
 	path->nodes[level] = t;
@@ -681,11 +681,11 @@ static int insert_new_root(struct ctree_root *root,
  *
  * returns zero on success and < 0 on any error
  */
-static int insert_ptr(struct ctree_root *root,
-		struct ctree_path *path, struct btrfs_disk_key *key,
+static int insert_ptr(struct btrfs_root *root,
+		struct btrfs_path *path, struct btrfs_disk_key *key,
 		u64 blocknr, int slot, int level)
 {
-	struct node *lower;
+	struct btrfs_node *lower;
 	int nritems;
 
 	BUG_ON(!path->nodes[level]);
@@ -719,13 +719,13 @@ static int insert_ptr(struct ctree_root *root,
  *
  * returns 0 on success and < 0 on failure
  */
-static int split_node(struct ctree_root *root, struct ctree_path *path,
+static int split_node(struct btrfs_root *root, struct btrfs_path *path,
 		      int level)
 {
-	struct tree_buffer *t;
-	struct node *c;
-	struct tree_buffer *split_buffer;
-	struct node *split;
+	struct btrfs_buffer *t;
+	struct btrfs_node *c;
+	struct btrfs_buffer *split_buffer;
+	struct btrfs_node *split;
 	int mid;
 	int ret;
 	int wret;
@@ -740,7 +740,7 @@ static int split_node(struct ctree_root *root, struct ctree_path *path,
 			return ret;
 	}
 	c_nritems = btrfs_header_nritems(&c->header);
-	split_buffer = alloc_free_block(root);
+	split_buffer = btrfs_alloc_free_block(root);
 	split = &split_buffer->node;
 	btrfs_set_header_flags(&split->header, btrfs_header_flags(&c->header));
 	btrfs_set_header_blocknr(&split->header, split_buffer->blocknr);
@@ -763,11 +763,11 @@ static int split_node(struct ctree_root *root, struct ctree_path *path,
 
 	if (path->slots[level] >= mid) {
 		path->slots[level] -= mid;
-		tree_block_release(root, t);
+		btrfs_block_release(root, t);
 		path->nodes[level] = split_buffer;
 		path->slots[level + 1] += 1;
 	} else {
-		tree_block_release(root, split_buffer);
+		btrfs_block_release(root, split_buffer);
 	}
 	return ret;
 }
@@ -777,7 +777,7 @@ static int split_node(struct ctree_root *root, struct ctree_path *path,
  * and nr indicate which items in the leaf to check.  This totals up the
  * space used both by the item structs and the item data
  */
-static int leaf_space_used(struct leaf *l, int start, int nr)
+static int leaf_space_used(struct btrfs_leaf *l, int start, int nr)
 {
 	int data_len;
 	int end = start + nr - 1;
@@ -797,14 +797,14 @@ static int leaf_space_used(struct leaf *l, int start, int nr)
  * returns 1 if the push failed because the other node didn't have enough
  * room, 0 if everything worked out and < 0 if there were major errors.
  */
-static int push_leaf_right(struct ctree_root *root, struct ctree_path *path,
+static int push_leaf_right(struct btrfs_root *root, struct btrfs_path *path,
 			   int data_size)
 {
-	struct tree_buffer *left_buf = path->nodes[0];
-	struct leaf *left = &left_buf->leaf;
-	struct leaf *right;
-	struct tree_buffer *right_buf;
-	struct tree_buffer *upper;
+	struct btrfs_buffer *left_buf = path->nodes[0];
+	struct btrfs_leaf *left = &left_buf->leaf;
+	struct btrfs_leaf *right;
+	struct btrfs_buffer *right_buf;
+	struct btrfs_buffer *upper;
 	int slot;
 	int i;
 	int free_space;
@@ -825,17 +825,17 @@ static int push_leaf_right(struct ctree_root *root, struct ctree_path *path,
 	right_buf = read_tree_block(root, btrfs_node_blockptr(&upper->node,
 							      slot + 1));
 	right = &right_buf->leaf;
-	free_space = leaf_free_space(right);
+	free_space = btrfs_leaf_free_space(right);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
-		tree_block_release(root, right_buf);
+		btrfs_block_release(root, right_buf);
 		return 1;
 	}
 	/* cow and double check */
 	btrfs_cow_block(root, right_buf, upper, slot + 1, &right_buf);
 	right = &right_buf->leaf;
-	free_space = leaf_free_space(right);
+	free_space = btrfs_leaf_free_space(right);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
-		tree_block_release(root, right_buf);
+		btrfs_block_release(root, right_buf);
 		return 1;
 	}
 
@@ -851,7 +851,7 @@ static int push_leaf_right(struct ctree_root *root, struct ctree_path *path,
 		push_space += btrfs_item_size(item) + sizeof(*item);
 	}
 	if (push_items == 0) {
-		tree_block_release(root, right_buf);
+		btrfs_block_release(root, right_buf);
 		return 1;
 	}
 	right_nritems = btrfs_header_nritems(&right->header);
@@ -893,11 +893,11 @@ static int push_leaf_right(struct ctree_root *root, struct ctree_path *path,
 	/* then fixup the leaf pointer in the path */
 	if (path->slots[0] >= left_nritems) {
 		path->slots[0] -= left_nritems;
-		tree_block_release(root, path->nodes[0]);
+		btrfs_block_release(root, path->nodes[0]);
 		path->nodes[0] = right_buf;
 		path->slots[1] += 1;
 	} else {
-		tree_block_release(root, right_buf);
+		btrfs_block_release(root, right_buf);
 	}
 	return 0;
 }
@@ -905,13 +905,13 @@ static int push_leaf_right(struct ctree_root *root, struct ctree_path *path,
  * push some data in the path leaf to the left, trying to free up at
  * least data_size bytes.  returns zero if the push worked, nonzero otherwise
  */
-static int push_leaf_left(struct ctree_root *root, struct ctree_path *path,
+static int push_leaf_left(struct btrfs_root *root, struct btrfs_path *path,
 			  int data_size)
 {
-	struct tree_buffer *right_buf = path->nodes[0];
-	struct leaf *right = &right_buf->leaf;
-	struct tree_buffer *t;
-	struct leaf *left;
+	struct btrfs_buffer *right_buf = path->nodes[0];
+	struct btrfs_leaf *right = &right_buf->leaf;
+	struct btrfs_buffer *t;
+	struct btrfs_leaf *left;
 	int slot;
 	int i;
 	int free_space;
@@ -932,18 +932,18 @@ static int push_leaf_left(struct ctree_root *root, struct ctree_path *path,
 	t = read_tree_block(root, btrfs_node_blockptr(&path->nodes[1]->node,
 						      slot - 1));
 	left = &t->leaf;
-	free_space = leaf_free_space(left);
+	free_space = btrfs_leaf_free_space(left);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
-		tree_block_release(root, t);
+		btrfs_block_release(root, t);
 		return 1;
 	}
 
 	/* cow and double check */
 	btrfs_cow_block(root, t, path->nodes[1], slot - 1, &t);
 	left = &t->leaf;
-	free_space = leaf_free_space(left);
+	free_space = btrfs_leaf_free_space(left);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
-		tree_block_release(root, t);
+		btrfs_block_release(root, t);
 		return 1;
 	}
 
@@ -958,7 +958,7 @@ static int push_leaf_left(struct ctree_root *root, struct ctree_path *path,
 		push_space += btrfs_item_size(item) + sizeof(*item);
 	}
 	if (push_items == 0) {
-		tree_block_release(root, t);
+		btrfs_block_release(root, t);
 		return 1;
 	}
 	/* push data from right to left */
@@ -1009,11 +1009,11 @@ static int push_leaf_left(struct ctree_root *root, struct ctree_path *path,
 	/* then fixup the leaf pointer in the path */
 	if (path->slots[0] < push_items) {
 		path->slots[0] += old_left_nritems;
-		tree_block_release(root, path->nodes[0]);
+		btrfs_block_release(root, path->nodes[0]);
 		path->nodes[0] = t;
 		path->slots[1] -= 1;
 	} else {
-		tree_block_release(root, t);
+		btrfs_block_release(root, t);
 		path->slots[0] -= push_items;
 	}
 	BUG_ON(path->slots[0] < 0);
@@ -1026,16 +1026,16 @@ static int push_leaf_left(struct ctree_root *root, struct ctree_path *path,
  *
  * returns 0 if all went well and < 0 on failure.
  */
-static int split_leaf(struct ctree_root *root, struct ctree_path *path,
+static int split_leaf(struct btrfs_root *root, struct btrfs_path *path,
 		      int data_size)
 {
-	struct tree_buffer *l_buf;
-	struct leaf *l;
+	struct btrfs_buffer *l_buf;
+	struct btrfs_leaf *l;
 	u32 nritems;
 	int mid;
 	int slot;
-	struct leaf *right;
-	struct tree_buffer *right_buffer;
+	struct btrfs_leaf *right;
+	struct btrfs_buffer *right_buffer;
 	int space_needed = data_size + sizeof(struct btrfs_item);
 	int data_copy_size;
 	int rt_data_off;
@@ -1047,7 +1047,7 @@ static int split_leaf(struct ctree_root *root, struct ctree_path *path,
 	l = &l_buf->leaf;
 
 	/* did the pushes work? */
-	if (leaf_free_space(l) >= sizeof(struct btrfs_item) + data_size)
+	if (btrfs_leaf_free_space(l) >= sizeof(struct btrfs_item) + data_size)
 		return 0;
 
 	if (!path->nodes[1]) {
@@ -1058,7 +1058,7 @@ static int split_leaf(struct ctree_root *root, struct ctree_path *path,
 	slot = path->slots[0];
 	nritems = btrfs_header_nritems(&l->header);
 	mid = (nritems + 1)/ 2;
-	right_buffer = alloc_free_block(root);
+	right_buffer = btrfs_alloc_free_block(root);
 	BUG_ON(!right_buffer);
 	BUG_ON(mid == nritems);
 	right = &right_buffer->leaf;
@@ -1101,12 +1101,12 @@ static int split_leaf(struct ctree_root *root, struct ctree_path *path,
 	BUG_ON(list_empty(&l_buf->dirty));
 	BUG_ON(path->slots[0] != slot);
 	if (mid <= slot) {
-		tree_block_release(root, path->nodes[0]);
+		btrfs_block_release(root, path->nodes[0]);
 		path->nodes[0] = right_buffer;
 		path->slots[0] -= mid;
 		path->slots[1] += 1;
 	} else
-		tree_block_release(root, right_buffer);
+		btrfs_block_release(root, right_buffer);
 	BUG_ON(path->slots[0] < 0);
 	return ret;
 }
@@ -1115,17 +1115,17 @@ static int split_leaf(struct ctree_root *root, struct ctree_path *path,
  * Given a key and some data, insert an item into the tree.
  * This does all the path init required, making room in the tree if needed.
  */
-int insert_item(struct ctree_root *root, struct btrfs_key *cpu_key,
+int btrfs_insert_item(struct btrfs_root *root, struct btrfs_key *cpu_key,
 			  void *data, int data_size)
 {
 	int ret = 0;
 	int slot;
 	int slot_orig;
-	struct leaf *leaf;
-	struct tree_buffer *leaf_buf;
+	struct btrfs_leaf *leaf;
+	struct btrfs_buffer *leaf_buf;
 	u32 nritems;
 	unsigned int data_end;
-	struct ctree_path path;
+	struct btrfs_path path;
 	struct btrfs_disk_key disk_key;
 
 	btrfs_cpu_key_to_disk(&disk_key, cpu_key);
@@ -1133,10 +1133,10 @@ int insert_item(struct ctree_root *root, struct btrfs_key *cpu_key,
 	/* create a root if there isn't one */
 	if (!root->node)
 		BUG();
-	init_path(&path);
-	ret = search_slot(root, cpu_key, &path, data_size, 1);
+	btrfs_init_path(&path);
+	ret = btrfs_search_slot(root, cpu_key, &path, data_size, 1);
 	if (ret == 0) {
-		release_path(root, &path);
+		btrfs_release_path(root, &path);
 		return -EEXIST;
 	}
 	if (ret < 0)
@@ -1149,7 +1149,8 @@ int insert_item(struct ctree_root *root, struct btrfs_key *cpu_key,
 	nritems = btrfs_header_nritems(&leaf->header);
 	data_end = leaf_data_end(leaf);
 
-	if (leaf_free_space(leaf) <  sizeof(struct btrfs_item) + data_size)
+	if (btrfs_leaf_free_space(leaf) <
+	    sizeof(struct btrfs_item) + data_size)
 		BUG();
 
 	slot = path.slots[0];
@@ -1190,11 +1191,11 @@ int insert_item(struct ctree_root *root, struct btrfs_key *cpu_key,
 		ret = fixup_low_keys(root, &path, &disk_key, 1);
 
 	BUG_ON(list_empty(&leaf_buf->dirty));
-	if (leaf_free_space(leaf) < 0)
+	if (btrfs_leaf_free_space(leaf) < 0)
 		BUG();
 	check_leaf(&path, 0);
 out:
-	release_path(root, &path);
+	btrfs_release_path(root, &path);
 	return ret;
 }
 
@@ -1205,11 +1206,11 @@ out:
  * continuing all the way the root if required.  The root is converted into
  * a leaf if all the nodes are emptied.
  */
-static int del_ptr(struct ctree_root *root, struct ctree_path *path, int level,
+static int del_ptr(struct btrfs_root *root, struct btrfs_path *path, int level,
 		   int slot)
 {
-	struct node *node;
-	struct tree_buffer *parent = path->nodes[level];
+	struct btrfs_node *node;
+	struct btrfs_buffer *parent = path->nodes[level];
 	u32 nritems;
 	int ret = 0;
 	int wret;
@@ -1242,11 +1243,11 @@ static int del_ptr(struct ctree_root *root, struct ctree_path *path, int level,
  * delete the item at the leaf level in path.  If that empties
  * the leaf, remove it from the tree
  */
-int del_item(struct ctree_root *root, struct ctree_path *path)
+int btrfs_del_item(struct btrfs_root *root, struct btrfs_path *path)
 {
 	int slot;
-	struct leaf *leaf;
-	struct tree_buffer *leaf_buf;
+	struct btrfs_leaf *leaf;
+	struct btrfs_buffer *leaf_buf;
 	int doff;
 	int dsize;
 	int ret = 0;
@@ -1286,7 +1287,7 @@ int del_item(struct ctree_root *root, struct ctree_path *path)
 			wret = del_ptr(root, path, 1, path->slots[1]);
 			if (wret)
 				ret = wret;
-			wret = free_extent(root, leaf_buf->blocknr, 1);
+			wret = btrfs_free_extent(root, leaf_buf->blocknr, 1);
 			if (wret)
 				ret = wret;
 		}
@@ -1323,12 +1324,12 @@ int del_item(struct ctree_root *root, struct ctree_path *path)
 				wret = del_ptr(root, path, 1, slot);
 				if (wret)
 					ret = wret;
-				tree_block_release(root, leaf_buf);
-				wret = free_extent(root, blocknr, 1);
+				btrfs_block_release(root, leaf_buf);
+				wret = btrfs_free_extent(root, blocknr, 1);
 				if (wret)
 					ret = wret;
 			} else {
-				tree_block_release(root, leaf_buf);
+				btrfs_block_release(root, leaf_buf);
 			}
 		}
 	}
@@ -1340,15 +1341,15 @@ int del_item(struct ctree_root *root, struct ctree_path *path)
  * returns 0 if it found something or 1 if there are no greater leaves.
  * returns < 0 on io errors.
  */
-int next_leaf(struct ctree_root *root, struct ctree_path *path)
+int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 {
 	int slot;
 	int level = 1;
 	u64 blocknr;
-	struct tree_buffer *c;
-	struct tree_buffer *next = NULL;
+	struct btrfs_buffer *c;
+	struct btrfs_buffer *next = NULL;
 
-	while(level < MAX_LEVEL) {
+	while(level < BTRFS_MAX_LEVEL) {
 		if (!path->nodes[level])
 			return 1;
 		slot = path->slots[level] + 1;
@@ -1359,7 +1360,7 @@ int next_leaf(struct ctree_root *root, struct ctree_path *path)
 		}
 		blocknr = btrfs_node_blockptr(&c->node, slot);
 		if (next)
-			tree_block_release(root, next);
+			btrfs_block_release(root, next);
 		next = read_tree_block(root, blocknr);
 		break;
 	}
@@ -1367,7 +1368,7 @@ int next_leaf(struct ctree_root *root, struct ctree_path *path)
 	while(1) {
 		level--;
 		c = path->nodes[level];
-		tree_block_release(root, c);
+		btrfs_block_release(root, c);
 		path->nodes[level] = next;
 		path->slots[level] = 0;
 		if (!level)
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 90cc2c9bd79e..5dfcc90812c1 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1,10 +1,10 @@
-#ifndef __CTREE__
-#define __CTREE__
+#ifndef __BTRFS__
+#define __BTRFS__
 
 #include "list.h"
 #include "kerncompat.h"
 
-#define CTREE_BLOCKSIZE 1024
+#define BTRFS_BLOCKSIZE 1024
 
 /*
  * the key defines the order in the tree, and so it also defines (optimal)
@@ -46,21 +46,21 @@ struct btrfs_header {
 	/* generation flags to be added */
 } __attribute__ ((__packed__));
 
-#define MAX_LEVEL 8
-#define NODEPTRS_PER_BLOCK ((CTREE_BLOCKSIZE - sizeof(struct btrfs_header)) / \
+#define BTRFS_MAX_LEVEL 8
+#define NODEPTRS_PER_BLOCK ((BTRFS_BLOCKSIZE - sizeof(struct btrfs_header)) / \
 			    (sizeof(struct btrfs_disk_key) + sizeof(u64)))
 
-struct tree_buffer;
+struct btrfs_buffer;
 
 /*
  * in ram representation of the tree.  extent_root is used for all allocations
  * and for the extent tree extent_root root.  current_insert is used
  * only for the extent tree.
  */
-struct ctree_root {
-	struct tree_buffer *node;
-	struct tree_buffer *commit_root;
-	struct ctree_root *extent_root;
+struct btrfs_root {
+	struct btrfs_buffer *node;
+	struct btrfs_buffer *commit_root;
+	struct btrfs_root *extent_root;
 	struct btrfs_key current_insert;
 	struct btrfs_key last_insert;
 	int fp;
@@ -74,7 +74,7 @@ struct ctree_root {
 /*
  * describes a tree on disk
  */
-struct ctree_root_info {
+struct btrfs_root_info {
 	u64 fsid[2]; /* FS specific uuid */
 	u64 blocknr; /* blocknr of this block */
 	u64 objectid; /* inode number of this root */
@@ -88,9 +88,9 @@ struct ctree_root_info {
  * the super block basically lists the main trees of the FS
  * it currently lacks any block count etc etc
  */
-struct ctree_super_block {
-	struct ctree_root_info root_info;
-	struct ctree_root_info extent_info;
+struct btrfs_super_block {
+	struct btrfs_root_info root_info;
+	struct btrfs_root_info extent_info;
 } __attribute__ ((__packed__));
 
 /*
@@ -111,13 +111,13 @@ struct btrfs_item {
  * The data is separate from the items to get the keys closer together
  * during searches.
  */
-#define LEAF_DATA_SIZE (CTREE_BLOCKSIZE - sizeof(struct btrfs_header))
-struct leaf {
+#define LEAF_DATA_SIZE (BTRFS_BLOCKSIZE - sizeof(struct btrfs_header))
+struct btrfs_leaf {
 	struct btrfs_header header;
 	union {
 		struct btrfs_item items[LEAF_DATA_SIZE/
 				        sizeof(struct btrfs_item)];
-		u8 data[CTREE_BLOCKSIZE-sizeof(struct btrfs_header)];
+		u8 data[BTRFS_BLOCKSIZE - sizeof(struct btrfs_header)];
 	};
 } __attribute__ ((__packed__));
 
@@ -125,7 +125,7 @@ struct leaf {
  * all non-leaf blocks are nodes, they hold only keys and pointers to
  * other blocks
  */
-struct node {
+struct btrfs_node {
 	struct btrfs_header header;
 	struct btrfs_disk_key keys[NODEPTRS_PER_BLOCK];
 	__le64 blockptrs[NODEPTRS_PER_BLOCK];
@@ -135,50 +135,51 @@ struct node {
  * items in the extent btree are used to record the objectid of the
  * owner of the block and the number of references
  */
-struct extent_item {
+struct btrfs_extent_item {
 	__le32 refs;
 	__le64 owner;
 } __attribute__ ((__packed__));
 
 /*
- * ctree_paths remember the path taken from the root down to the leaf.
- * level 0 is always the leaf, and nodes[1...MAX_LEVEL] will point
+ * btrfs_paths remember the path taken from the root down to the leaf.
+ * level 0 is always the leaf, and nodes[1...BTRFS_MAX_LEVEL] will point
  * to any other levels that are present.
  *
  * The slots array records the index of the item or block pointer
  * used while walking the tree.
  */
-struct ctree_path {
-	struct tree_buffer *nodes[MAX_LEVEL];
-	int slots[MAX_LEVEL];
+struct btrfs_path {
+	struct btrfs_buffer *nodes[BTRFS_MAX_LEVEL];
+	int slots[BTRFS_MAX_LEVEL];
 };
 
-static inline u64 btrfs_extent_owner(struct extent_item *ei)
+static inline u64 btrfs_extent_owner(struct btrfs_extent_item *ei)
 {
 	return le64_to_cpu(ei->owner);
 }
 
-static inline void btrfs_set_extent_owner(struct extent_item *ei, u64 val)
+static inline void btrfs_set_extent_owner(struct btrfs_extent_item *ei, u64 val)
 {
 	ei->owner = cpu_to_le64(val);
 }
 
-static inline u32 btrfs_extent_refs(struct extent_item *ei)
+static inline u32 btrfs_extent_refs(struct btrfs_extent_item *ei)
 {
 	return le32_to_cpu(ei->refs);
 }
 
-static inline void btrfs_set_extent_refs(struct extent_item *ei, u32 val)
+static inline void btrfs_set_extent_refs(struct btrfs_extent_item *ei, u32 val)
 {
 	ei->refs = cpu_to_le32(val);
 }
 
-static inline u64 btrfs_node_blockptr(struct node *n, int nr)
+static inline u64 btrfs_node_blockptr(struct btrfs_node *n, int nr)
 {
 	return le64_to_cpu(n->blockptrs[nr]);
 }
 
-static inline void btrfs_set_node_blockptr(struct node *n, int nr, u64 val)
+static inline void btrfs_set_node_blockptr(struct btrfs_node *n, int nr,
+					   u64 val)
 {
 	n->blockptrs[nr] = cpu_to_le64(val);
 }
@@ -300,34 +301,34 @@ static inline void btrfs_set_header_flags(struct btrfs_header *h, u16 val)
 
 static inline int btrfs_header_level(struct btrfs_header *h)
 {
-	return btrfs_header_flags(h) & (MAX_LEVEL - 1);
+	return btrfs_header_flags(h) & (BTRFS_MAX_LEVEL - 1);
 }
 
 static inline void btrfs_set_header_level(struct btrfs_header *h, int level)
 {
 	u16 flags;
-	BUG_ON(level > MAX_LEVEL);
-	flags = btrfs_header_flags(h) & ~(MAX_LEVEL - 1);
+	BUG_ON(level > BTRFS_MAX_LEVEL);
+	flags = btrfs_header_flags(h) & ~(BTRFS_MAX_LEVEL - 1);
 	btrfs_set_header_flags(h, flags | level);
 }
 
-static inline int btrfs_is_leaf(struct node *n)
+static inline int btrfs_is_leaf(struct btrfs_node *n)
 {
 	return (btrfs_header_level(&n->header) == 0);
 }
 
-struct tree_buffer *alloc_free_block(struct ctree_root *root);
-int btrfs_inc_ref(struct ctree_root *root, struct tree_buffer *buf);
-int free_extent(struct ctree_root *root, u64 blocknr, u64 num_blocks);
-int search_slot(struct ctree_root *root, struct btrfs_key *key,
-		struct ctree_path *p, int ins_len, int cow);
-void release_path(struct ctree_root *root, struct ctree_path *p);
-void init_path(struct ctree_path *p);
-int del_item(struct ctree_root *root, struct ctree_path *path);
-int insert_item(struct ctree_root *root, struct btrfs_key *key,
+struct btrfs_buffer *btrfs_alloc_free_block(struct btrfs_root *root);
+int btrfs_inc_ref(struct btrfs_root *root, struct btrfs_buffer *buf);
+int btrfs_free_extent(struct btrfs_root *root, u64 blocknr, u64 num_blocks);
+int btrfs_search_slot(struct btrfs_root *root, struct btrfs_key *key,
+		struct btrfs_path *p, int ins_len, int cow);
+void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p);
+void btrfs_init_path(struct btrfs_path *p);
+int btrfs_del_item(struct btrfs_root *root, struct btrfs_path *path);
+int btrfs_insert_item(struct btrfs_root *root, struct btrfs_key *key,
 		void *data, int data_size);
-int next_leaf(struct ctree_root *root, struct ctree_path *path);
-int leaf_free_space(struct leaf *leaf);
-int btrfs_drop_snapshot(struct ctree_root *root, struct tree_buffer *snap);
-int btrfs_finish_extent_commit(struct ctree_root *root);
+int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path);
+int btrfs_leaf_free_space(struct btrfs_leaf *leaf);
+int btrfs_drop_snapshot(struct btrfs_root *root, struct btrfs_buffer *snap);
+int btrfs_finish_extent_commit(struct btrfs_root *root);
 #endif
diff --git a/fs/btrfs/debug-tree.c b/fs/btrfs/debug-tree.c
index 21f607d8b050..6da0a7aa0f47 100644
--- a/fs/btrfs/debug-tree.c
+++ b/fs/btrfs/debug-tree.c
@@ -7,13 +7,13 @@
 #include "print-tree.h"
 
 int main(int ac, char **av) {
-	struct ctree_super_block super;
-	struct ctree_root *root;
+	struct btrfs_super_block super;
+	struct btrfs_root *root;
 	radix_tree_init();
 	root = open_ctree("dbfile", &super);
 	printf("root tree\n");
-	print_tree(root, root->node);
+	btrfs_print_tree(root, root->node);
 	printf("map tree\n");
-	print_tree(root->extent_root, root->extent_root->node);
+	btrfs_print_tree(root->extent_root, root->extent_root->node);
 	return 0;
 }
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 065e888d2c0b..c34c0c60935f 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -13,7 +13,7 @@
 static int allocated_blocks = 0;
 int cache_max = 10000;
 
-static int check_tree_block(struct ctree_root *root, struct tree_buffer *buf)
+static int check_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf)
 {
 	if (buf->blocknr != btrfs_header_blocknr(&buf->node.header))
 		BUG();
@@ -23,18 +23,18 @@ static int check_tree_block(struct ctree_root *root, struct tree_buffer *buf)
 	return 0;
 }
 
-static int free_some_buffers(struct ctree_root *root)
+static int free_some_buffers(struct btrfs_root *root)
 {
 	struct list_head *node, *next;
-	struct tree_buffer *b;
+	struct btrfs_buffer *b;
 	if (root->cache_size < cache_max)
 		return 0;
 	list_for_each_safe(node, next, &root->cache) {
-		b = list_entry(node, struct tree_buffer, cache);
+		b = list_entry(node, struct btrfs_buffer, cache);
 		if (b->count == 1) {
 			BUG_ON(!list_empty(&b->dirty));
 			list_del_init(&b->cache);
-			tree_block_release(root, b);
+			btrfs_block_release(root, b);
 			if (root->cache_size < cache_max)
 				break;
 		}
@@ -42,11 +42,11 @@ static int free_some_buffers(struct ctree_root *root)
 	return 0;
 }
 
-struct tree_buffer *alloc_tree_block(struct ctree_root *root, u64 blocknr)
+struct btrfs_buffer *alloc_tree_block(struct btrfs_root *root, u64 blocknr)
 {
-	struct tree_buffer *buf;
+	struct btrfs_buffer *buf;
 	int ret;
-	buf = malloc(sizeof(struct tree_buffer));
+	buf = malloc(sizeof(struct btrfs_buffer));
 	if (!buf)
 		return buf;
 	allocated_blocks++;
@@ -66,9 +66,9 @@ struct tree_buffer *alloc_tree_block(struct ctree_root *root, u64 blocknr)
 	return buf;
 }
 
-struct tree_buffer *find_tree_block(struct ctree_root *root, u64 blocknr)
+struct btrfs_buffer *find_tree_block(struct btrfs_root *root, u64 blocknr)
 {
-	struct tree_buffer *buf;
+	struct btrfs_buffer *buf;
 	buf = radix_tree_lookup(&root->cache_radix, blocknr);
 	if (buf) {
 		buf->count++;
@@ -82,10 +82,10 @@ struct tree_buffer *find_tree_block(struct ctree_root *root, u64 blocknr)
 	return buf;
 }
 
-struct tree_buffer *read_tree_block(struct ctree_root *root, u64 blocknr)
+struct btrfs_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr)
 {
-	loff_t offset = blocknr * CTREE_BLOCKSIZE;
-	struct tree_buffer *buf;
+	loff_t offset = blocknr * BTRFS_BLOCKSIZE;
+	struct btrfs_buffer *buf;
 	int ret;
 
 	buf = radix_tree_lookup(&root->cache_radix, blocknr);
@@ -95,8 +95,8 @@ struct tree_buffer *read_tree_block(struct ctree_root *root, u64 blocknr)
 		buf = alloc_tree_block(root, blocknr);
 		if (!buf)
 			return NULL;
-		ret = pread(root->fp, &buf->node, CTREE_BLOCKSIZE, offset);
-		if (ret != CTREE_BLOCKSIZE) {
+		ret = pread(root->fp, &buf->node, BTRFS_BLOCKSIZE, offset);
+		if (ret != BTRFS_BLOCKSIZE) {
 			free(buf);
 			return NULL;
 		}
@@ -106,7 +106,7 @@ struct tree_buffer *read_tree_block(struct ctree_root *root, u64 blocknr)
 	return buf;
 }
 
-int dirty_tree_block(struct ctree_root *root, struct tree_buffer *buf)
+int dirty_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf)
 {
 	if (!list_empty(&buf->dirty))
 		return 0;
@@ -115,46 +115,47 @@ int dirty_tree_block(struct ctree_root *root, struct tree_buffer *buf)
 	return 0;
 }
 
-int clean_tree_block(struct ctree_root *root, struct tree_buffer *buf)
+int clean_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf)
 {
 	if (!list_empty(&buf->dirty)) {
 		list_del_init(&buf->dirty);
-		tree_block_release(root, buf);
+		btrfs_block_release(root, buf);
 	}
 	return 0;
 }
 
-int write_tree_block(struct ctree_root *root, struct tree_buffer *buf)
+int write_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf)
 {
 	u64 blocknr = buf->blocknr;
-	loff_t offset = blocknr * CTREE_BLOCKSIZE;
+	loff_t offset = blocknr * BTRFS_BLOCKSIZE;
 	int ret;
 
 	if (buf->blocknr != btrfs_header_blocknr(&buf->node.header))
 		BUG();
-	ret = pwrite(root->fp, &buf->node, CTREE_BLOCKSIZE, offset);
-	if (ret != CTREE_BLOCKSIZE)
+	ret = pwrite(root->fp, &buf->node, BTRFS_BLOCKSIZE, offset);
+	if (ret != BTRFS_BLOCKSIZE)
 		return ret;
 	return 0;
 }
 
-static int __commit_transaction(struct ctree_root *root)
+static int __commit_transaction(struct btrfs_root *root)
 {
-	struct tree_buffer *b;
+	struct btrfs_buffer *b;
 	int ret = 0;
 	int wret;
 	while(!list_empty(&root->trans)) {
-		b = list_entry(root->trans.next, struct tree_buffer, dirty);
+		b = list_entry(root->trans.next, struct btrfs_buffer, dirty);
 		list_del_init(&b->dirty);
 		wret = write_tree_block(root, b);
 		if (wret)
 			ret = wret;
-		tree_block_release(root, b);
+		btrfs_block_release(root, b);
 	}
 	return ret;
 }
 
-int commit_transaction(struct ctree_root *root, struct ctree_super_block *s)
+int btrfs_commit_transaction(struct btrfs_root *root,
+			     struct btrfs_super_block *s)
 {
 	int ret = 0;
 
@@ -163,20 +164,20 @@ int commit_transaction(struct ctree_root *root, struct ctree_super_block *s)
 		ret = __commit_transaction(root->extent_root);
 	BUG_ON(ret);
 	if (root->commit_root != root->node) {
-		struct tree_buffer *snap = root->commit_root;
+		struct btrfs_buffer *snap = root->commit_root;
 		root->commit_root = root->node;
 		root->node->count++;
 		ret = btrfs_drop_snapshot(root, snap);
 		BUG_ON(ret);
-		// tree_block_release(root, snap);
+		// btrfs_block_release(root, snap);
 	}
         write_ctree_super(root, s);
 	btrfs_finish_extent_commit(root);
 	return ret;
 }
 
-static int __setup_root(struct ctree_root *root, struct ctree_root *extent_root,
-			struct ctree_root_info *info, int fp)
+static int __setup_root(struct btrfs_root *root, struct btrfs_root *extent_root,
+			struct btrfs_root_info *info, int fp)
 {
 	INIT_LIST_HEAD(&root->trans);
 	INIT_LIST_HEAD(&root->cache);
@@ -191,10 +192,10 @@ static int __setup_root(struct ctree_root *root, struct ctree_root *extent_root,
 	return 0;
 }
 
-struct ctree_root *open_ctree(char *filename, struct ctree_super_block *super)
+struct btrfs_root *open_ctree(char *filename, struct btrfs_super_block *super)
 {
-	struct ctree_root *root = malloc(sizeof(struct ctree_root));
-	struct ctree_root *extent_root = malloc(sizeof(struct ctree_root));
+	struct btrfs_root *root = malloc(sizeof(struct btrfs_root));
+	struct btrfs_root *extent_root = malloc(sizeof(struct btrfs_root));
 	int fp;
 	int ret;
 
@@ -207,16 +208,16 @@ struct ctree_root *open_ctree(char *filename, struct ctree_super_block *super)
 	INIT_RADIX_TREE(&root->pinned_radix, GFP_KERNEL);
 	INIT_RADIX_TREE(&extent_root->pinned_radix, GFP_KERNEL);
 	INIT_RADIX_TREE(&extent_root->cache_radix, GFP_KERNEL);
-	ret = pread(fp, super, sizeof(struct ctree_super_block),
-		     CTREE_SUPER_INFO_OFFSET(CTREE_BLOCKSIZE));
+	ret = pread(fp, super, sizeof(struct btrfs_super_block),
+		     BTRFS_SUPER_INFO_OFFSET(BTRFS_BLOCKSIZE));
 	if (ret == 0 || super->root_info.tree_root == 0) {
 		printf("making new FS!\n");
 		ret = mkfs(fp);
 		if (ret)
 			return NULL;
-		ret = pread(fp, super, sizeof(struct ctree_super_block),
-			     CTREE_SUPER_INFO_OFFSET(CTREE_BLOCKSIZE));
-		if (ret != sizeof(struct ctree_super_block))
+		ret = pread(fp, super, sizeof(struct btrfs_super_block),
+			     BTRFS_SUPER_INFO_OFFSET(BTRFS_BLOCKSIZE));
+		if (ret != sizeof(struct btrfs_super_block))
 			return NULL;
 	}
 	BUG_ON(ret < 0);
@@ -227,18 +228,19 @@ struct ctree_root *open_ctree(char *filename, struct ctree_super_block *super)
 	return root;
 }
 
-static int __update_root(struct ctree_root *root, struct ctree_root_info *info)
+static int __update_root(struct btrfs_root *root, struct btrfs_root_info *info)
 {
 	info->tree_root = root->node->blocknr;
 	return 0;
 }
 
-int write_ctree_super(struct ctree_root *root, struct ctree_super_block *s)
+int write_ctree_super(struct btrfs_root *root, struct btrfs_super_block *s)
 {
 	int ret;
 	__update_root(root, &s->root_info);
 	__update_root(root->extent_root, &s->extent_info);
-	ret = pwrite(root->fp, s, sizeof(*s), CTREE_SUPER_INFO_OFFSET(CTREE_BLOCKSIZE));
+	ret = pwrite(root->fp, s, sizeof(*s),
+		     BTRFS_SUPER_INFO_OFFSET(BTRFS_BLOCKSIZE));
 	if (ret != sizeof(*s)) {
 		fprintf(stderr, "failed to write new super block err %d\n", ret);
 		return ret;
@@ -246,19 +248,19 @@ int write_ctree_super(struct ctree_root *root, struct ctree_super_block *s)
 	return 0;
 }
 
-static int drop_cache(struct ctree_root *root)
+static int drop_cache(struct btrfs_root *root)
 {
 	while(!list_empty(&root->cache)) {
-		struct tree_buffer *b = list_entry(root->cache.next,
-						   struct tree_buffer, cache);
+		struct btrfs_buffer *b = list_entry(root->cache.next,
+						   struct btrfs_buffer, cache);
 		list_del_init(&b->cache);
-		tree_block_release(root, b);
+		btrfs_block_release(root, b);
 	}
 	return 0;
 }
-int close_ctree(struct ctree_root *root, struct ctree_super_block *s)
+int close_ctree(struct btrfs_root *root, struct btrfs_super_block *s)
 {
-	commit_transaction(root, s);
+	btrfs_commit_transaction(root, s);
 	__commit_transaction(root->extent_root);
 	write_ctree_super(root, s);
 	drop_cache(root->extent_root);
@@ -268,16 +270,16 @@ int close_ctree(struct ctree_root *root, struct ctree_super_block *s)
 
 	close(root->fp);
 	if (root->node)
-		tree_block_release(root, root->node);
+		btrfs_block_release(root, root->node);
 	if (root->extent_root->node)
-		tree_block_release(root->extent_root, root->extent_root->node);
-	tree_block_release(root, root->commit_root);
+		btrfs_block_release(root->extent_root, root->extent_root->node);
+	btrfs_block_release(root, root->commit_root);
 	free(root);
 	printf("on close %d blocks are allocated\n", allocated_blocks);
 	return 0;
 }
 
-void tree_block_release(struct ctree_root *root, struct tree_buffer *buf)
+void btrfs_block_release(struct btrfs_root *root, struct btrfs_buffer *buf)
 {
 	buf->count--;
 	if (buf->count < 0)
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 1c0af7c56c21..b391335864b1 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -2,29 +2,30 @@
 #define __DISKIO__
 #include "list.h"
 
-struct tree_buffer {
+struct btrfs_buffer {
 	u64 blocknr;
 	int count;
 	union {
-		struct node node;
-		struct leaf leaf;
+		struct btrfs_node node;
+		struct btrfs_leaf leaf;
 	};
 	struct list_head dirty;
 	struct list_head cache;
 };
 
-struct tree_buffer *read_tree_block(struct ctree_root *root, u64 blocknr);
-struct tree_buffer *find_tree_block(struct ctree_root *root, u64 blocknr);
-int write_tree_block(struct ctree_root *root, struct tree_buffer *buf);
-int dirty_tree_block(struct ctree_root *root, struct tree_buffer *buf);
-int clean_tree_block(struct ctree_root *root, struct tree_buffer *buf);
-int commit_transaction(struct ctree_root *root, struct ctree_super_block *s);
-struct ctree_root *open_ctree(char *filename, struct ctree_super_block *s);
-int close_ctree(struct ctree_root *root, struct ctree_super_block *s);
-void tree_block_release(struct ctree_root *root, struct tree_buffer *buf);
-int write_ctree_super(struct ctree_root *root, struct ctree_super_block *s);
+struct btrfs_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr);
+struct btrfs_buffer *find_tree_block(struct btrfs_root *root, u64 blocknr);
+int write_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf);
+int dirty_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf);
+int clean_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf);
+int btrfs_commit_transaction(struct btrfs_root *root,
+			     struct btrfs_super_block *s);
+struct btrfs_root *open_ctree(char *filename, struct btrfs_super_block *s);
+int close_ctree(struct btrfs_root *root, struct btrfs_super_block *s);
+void btrfs_block_release(struct btrfs_root *root, struct btrfs_buffer *buf);
+int write_ctree_super(struct btrfs_root *root, struct btrfs_super_block *s);
 int mkfs(int fd);
 
-#define CTREE_SUPER_INFO_OFFSET(bs) (16 * (bs))
+#define BTRFS_SUPER_INFO_OFFSET(bs) (16 * (bs))
 
 #endif
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 1676a6595cce..4a4f2d810d9f 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -6,11 +6,11 @@
 #include "disk-io.h"
 #include "print-tree.h"
 
-static int find_free_extent(struct ctree_root *orig_root, u64 num_blocks,
+static int find_free_extent(struct btrfs_root *orig_root, u64 num_blocks,
 			    u64 search_start, u64 search_end,
 			    struct btrfs_key *ins);
-static int finish_current_insert(struct ctree_root *extent_root);
-static int run_pending(struct ctree_root *extent_root);
+static int finish_current_insert(struct btrfs_root *extent_root);
+static int run_pending(struct btrfs_root *extent_root);
 
 /*
  * pending extents are blocks that we're trying to allocate in the extent
@@ -21,62 +21,63 @@ static int run_pending(struct ctree_root *extent_root);
  */
 #define CTREE_EXTENT_PENDING_DEL 0
 
-static int inc_block_ref(struct ctree_root *root, u64 blocknr)
+static int inc_block_ref(struct btrfs_root *root, u64 blocknr)
 {
-	struct ctree_path path;
+	struct btrfs_path path;
 	int ret;
 	struct btrfs_key key;
-	struct leaf *l;
-	struct extent_item *item;
+	struct btrfs_leaf *l;
+	struct btrfs_extent_item *item;
 	struct btrfs_key ins;
 	u32 refs;
 
 	find_free_extent(root->extent_root, 0, 0, (u64)-1, &ins);
-	init_path(&path);
+	btrfs_init_path(&path);
 	key.objectid = blocknr;
 	key.flags = 0;
 	key.offset = 1;
-	ret = search_slot(root->extent_root, &key, &path, 0, 1);
+	ret = btrfs_search_slot(root->extent_root, &key, &path, 0, 1);
 	if (ret != 0)
 		BUG();
 	BUG_ON(ret != 0);
 	l = &path.nodes[0]->leaf;
-	item = (struct extent_item *)(l->data + btrfs_item_offset(l->items +
-								path.slots[0]));
+	item = (struct btrfs_extent_item *)(l->data +
+					    btrfs_item_offset(l->items +
+							      path.slots[0]));
 	refs = btrfs_extent_refs(item);
 	btrfs_set_extent_refs(item, refs + 1);
 
 	BUG_ON(list_empty(&path.nodes[0]->dirty));
-	release_path(root->extent_root, &path);
+	btrfs_release_path(root->extent_root, &path);
 	finish_current_insert(root->extent_root);
 	run_pending(root->extent_root);
 	return 0;
 }
 
-static int lookup_block_ref(struct ctree_root *root, u64 blocknr, u32 *refs)
+static int lookup_block_ref(struct btrfs_root *root, u64 blocknr, u32 *refs)
 {
-	struct ctree_path path;
+	struct btrfs_path path;
 	int ret;
 	struct btrfs_key key;
-	struct leaf *l;
-	struct extent_item *item;
-	init_path(&path);
+	struct btrfs_leaf *l;
+	struct btrfs_extent_item *item;
+	btrfs_init_path(&path);
 	key.objectid = blocknr;
 	key.flags = 0;
 	key.offset = 1;
-	ret = search_slot(root->extent_root, &key, &path, 0, 0);
+	ret = btrfs_search_slot(root->extent_root, &key, &path, 0, 0);
 	if (ret != 0)
 		BUG();
 	l = &path.nodes[0]->leaf;
-	item = (struct extent_item *)(l->data +
+	item = (struct btrfs_extent_item *)(l->data +
 				      btrfs_item_offset(l->items +
 							path.slots[0]));
 	*refs = btrfs_extent_refs(item);
-	release_path(root->extent_root, &path);
+	btrfs_release_path(root->extent_root, &path);
 	return 0;
 }
 
-int btrfs_inc_ref(struct ctree_root *root, struct tree_buffer *buf)
+int btrfs_inc_ref(struct btrfs_root *root, struct btrfs_buffer *buf)
 {
 	u64 blocknr;
 	int i;
@@ -93,9 +94,9 @@ int btrfs_inc_ref(struct ctree_root *root, struct tree_buffer *buf)
 	return 0;
 }
 
-int btrfs_finish_extent_commit(struct ctree_root *root)
+int btrfs_finish_extent_commit(struct btrfs_root *root)
 {
-	struct ctree_root *extent_root = root->extent_root;
+	struct btrfs_root *extent_root = root->extent_root;
 	unsigned long gang[8];
 	int ret;
 	int i;
@@ -115,10 +116,10 @@ int btrfs_finish_extent_commit(struct ctree_root *root)
 	return 0;
 }
 
-static int finish_current_insert(struct ctree_root *extent_root)
+static int finish_current_insert(struct btrfs_root *extent_root)
 {
 	struct btrfs_key ins;
-	struct extent_item extent_item;
+	struct btrfs_extent_item extent_item;
 	int i;
 	int ret;
 
@@ -130,7 +131,7 @@ static int finish_current_insert(struct ctree_root *extent_root)
 
 	for (i = 0; i < extent_root->current_insert.flags; i++) {
 		ins.objectid = extent_root->current_insert.objectid + i;
-		ret = insert_item(extent_root, &ins, &extent_item,
+		ret = btrfs_insert_item(extent_root, &ins, &extent_item,
 				  sizeof(extent_item));
 		BUG_ON(ret);
 	}
@@ -141,14 +142,14 @@ static int finish_current_insert(struct ctree_root *extent_root)
 /*
  * remove an extent from the root, returns 0 on success
  */
-int __free_extent(struct ctree_root *root, u64 blocknr, u64 num_blocks)
+static int __free_extent(struct btrfs_root *root, u64 blocknr, u64 num_blocks)
 {
-	struct ctree_path path;
+	struct btrfs_path path;
 	struct btrfs_key key;
-	struct ctree_root *extent_root = root->extent_root;
+	struct btrfs_root *extent_root = root->extent_root;
 	int ret;
 	struct btrfs_item *item;
-	struct extent_item *ei;
+	struct btrfs_extent_item *ei;
 	struct btrfs_key ins;
 	u32 refs;
 
@@ -157,16 +158,16 @@ int __free_extent(struct ctree_root *root, u64 blocknr, u64 num_blocks)
 	key.offset = num_blocks;
 
 	find_free_extent(root, 0, 0, (u64)-1, &ins);
-	init_path(&path);
-	ret = search_slot(extent_root, &key, &path, -1, 1);
+	btrfs_init_path(&path);
+	ret = btrfs_search_slot(extent_root, &key, &path, -1, 1);
 	if (ret) {
 		printf("failed to find %Lu\n", key.objectid);
-		print_tree(extent_root, extent_root->node);
+		btrfs_print_tree(extent_root, extent_root->node);
 		printf("failed to find %Lu\n", key.objectid);
 		BUG();
 	}
 	item = path.nodes[0]->leaf.items + path.slots[0];
-	ei = (struct extent_item *)(path.nodes[0]->leaf.data +
+	ei = (struct btrfs_extent_item *)(path.nodes[0]->leaf.data +
 				    btrfs_item_offset(item));
 	BUG_ON(ei->refs == 0);
 	refs = btrfs_extent_refs(ei) - 1;
@@ -180,14 +181,14 @@ int __free_extent(struct ctree_root *root, u64 blocknr, u64 num_blocks)
 			BUG_ON(err);
 			radix_tree_preload_end();
 		}
-		ret = del_item(extent_root, &path);
+		ret = btrfs_del_item(extent_root, &path);
 		if (root != extent_root &&
 		    extent_root->last_insert.objectid < blocknr)
 			extent_root->last_insert.objectid = blocknr;
 		if (ret)
 			BUG();
 	}
-	release_path(extent_root, &path);
+	btrfs_release_path(extent_root, &path);
 	finish_current_insert(extent_root);
 	return ret;
 }
@@ -196,10 +197,10 @@ int __free_extent(struct ctree_root *root, u64 blocknr, u64 num_blocks)
  * find all the blocks marked as pending in the radix tree and remove
  * them from the extent map
  */
-static int del_pending_extents(struct ctree_root *extent_root)
+static int del_pending_extents(struct btrfs_root *extent_root)
 {
 	int ret;
-	struct tree_buffer *gang[4];
+	struct btrfs_buffer *gang[4];
 	int i;
 
 	while(1) {
@@ -214,13 +215,13 @@ static int del_pending_extents(struct ctree_root *extent_root)
 			radix_tree_tag_clear(&extent_root->cache_radix,
 						gang[i]->blocknr,
 						CTREE_EXTENT_PENDING_DEL);
-			tree_block_release(extent_root, gang[i]);
+			btrfs_block_release(extent_root, gang[i]);
 		}
 	}
 	return 0;
 }
 
-static int run_pending(struct ctree_root *extent_root)
+static int run_pending(struct btrfs_root *extent_root)
 {
 	while(radix_tree_tagged(&extent_root->cache_radix,
 			        CTREE_EXTENT_PENDING_DEL))
@@ -232,11 +233,11 @@ static int run_pending(struct ctree_root *extent_root)
 /*
  * remove an extent from the root, returns 0 on success
  */
-int free_extent(struct ctree_root *root, u64 blocknr, u64 num_blocks)
+int btrfs_free_extent(struct btrfs_root *root, u64 blocknr, u64 num_blocks)
 {
 	struct btrfs_key key;
-	struct ctree_root *extent_root = root->extent_root;
-	struct tree_buffer *t;
+	struct btrfs_root *extent_root = root->extent_root;
+	struct btrfs_buffer *t;
 	int pending_ret;
 	int ret;
 
@@ -262,11 +263,11 @@ int free_extent(struct ctree_root *root, u64 blocknr, u64 num_blocks)
  * ins->offset == number of blocks
  * Any available blocks before search_start are skipped.
  */
-static int find_free_extent(struct ctree_root *orig_root, u64 num_blocks,
+static int find_free_extent(struct btrfs_root *orig_root, u64 num_blocks,
 			    u64 search_start, u64 search_end,
 			    struct btrfs_key *ins)
 {
-	struct ctree_path path;
+	struct btrfs_path path;
 	struct btrfs_key key;
 	int ret;
 	u64 hole_size = 0;
@@ -274,20 +275,20 @@ static int find_free_extent(struct ctree_root *orig_root, u64 num_blocks,
 	u64 last_block;
 	u64 test_block;
 	int start_found;
-	struct leaf *l;
-	struct ctree_root * root = orig_root->extent_root;
+	struct btrfs_leaf *l;
+	struct btrfs_root * root = orig_root->extent_root;
 	int total_needed = num_blocks;
 
 	total_needed += (btrfs_header_level(&root->node->node.header) + 1) * 3;
 	if (root->last_insert.objectid > search_start)
 		search_start = root->last_insert.objectid;
 check_failed:
-	init_path(&path);
+	btrfs_init_path(&path);
 	ins->objectid = search_start;
 	ins->offset = 0;
 	ins->flags = 0;
 	start_found = 0;
-	ret = search_slot(root, ins, &path, 0, 0);
+	ret = btrfs_search_slot(root, ins, &path, 0, 0);
 	if (ret < 0)
 		goto error;
 
@@ -298,7 +299,7 @@ check_failed:
 		l = &path.nodes[0]->leaf;
 		slot = path.slots[0];
 		if (slot >= btrfs_header_nritems(&l->header)) {
-			ret = next_leaf(root, &path);
+			ret = btrfs_next_leaf(root, &path);
 			if (ret == 0)
 				continue;
 			if (ret < 0)
@@ -336,7 +337,7 @@ check_pending:
 	/* we have to make sure we didn't find an extent that has already
 	 * been allocated by the map tree or the original allocation
 	 */
-	release_path(root, &path);
+	btrfs_release_path(root, &path);
 	BUG_ON(ins->objectid < search_start);
 	for (test_block = ins->objectid;
 	     test_block < ins->objectid + total_needed; test_block++) {
@@ -353,7 +354,7 @@ check_pending:
 	ins->offset = num_blocks;
 	return 0;
 error:
-	release_path(root, &path);
+	btrfs_release_path(root, &path);
 	return ret;
 }
 
@@ -364,13 +365,13 @@ error:
  *
  * returns 0 if everything worked, non-zero otherwise.
  */
-int alloc_extent(struct ctree_root *root, u64 num_blocks, u64 search_start,
+int alloc_extent(struct btrfs_root *root, u64 num_blocks, u64 search_start,
 			 u64 search_end, u64 owner, struct btrfs_key *ins)
 {
 	int ret;
 	int pending_ret;
-	struct ctree_root *extent_root = root->extent_root;
-	struct extent_item extent_item;
+	struct btrfs_root *extent_root = root->extent_root;
+	struct btrfs_extent_item extent_item;
 
 	btrfs_set_extent_refs(&extent_item, 1);
 	btrfs_set_extent_owner(&extent_item, owner);
@@ -390,7 +391,7 @@ int alloc_extent(struct ctree_root *root, u64 num_blocks, u64 search_start,
 	if (ret)
 		return ret;
 
-	ret = insert_item(extent_root, ins, &extent_item,
+	ret = btrfs_insert_item(extent_root, ins, &extent_item,
 			  sizeof(extent_item));
 
 	finish_current_insert(extent_root);
@@ -406,11 +407,11 @@ int alloc_extent(struct ctree_root *root, u64 num_blocks, u64 search_start,
  * helper function to allocate a block for a given tree
  * returns the tree buffer or NULL.
  */
-struct tree_buffer *alloc_free_block(struct ctree_root *root)
+struct btrfs_buffer *btrfs_alloc_free_block(struct btrfs_root *root)
 {
 	struct btrfs_key ins;
 	int ret;
-	struct tree_buffer *buf;
+	struct btrfs_buffer *buf;
 
 	ret = alloc_extent(root, 1, 0, (unsigned long)-1,
 			   btrfs_header_parentid(&root->node->node.header),
@@ -424,10 +425,10 @@ struct tree_buffer *alloc_free_block(struct ctree_root *root)
 	return buf;
 }
 
-int walk_down_tree(struct ctree_root *root, struct ctree_path *path, int *level)
+int walk_down_tree(struct btrfs_root *root, struct btrfs_path *path, int *level)
 {
-	struct tree_buffer *next;
-	struct tree_buffer *cur;
+	struct btrfs_buffer *next;
+	struct btrfs_buffer *cur;
 	u64 blocknr;
 	int ret;
 	u32 refs;
@@ -445,33 +446,33 @@ int walk_down_tree(struct ctree_root *root, struct ctree_path *path, int *level)
 		ret = lookup_block_ref(root, blocknr, &refs);
 		if (refs != 1 || *level == 1) {
 			path->slots[*level]++;
-			ret = free_extent(root, blocknr, 1);
+			ret = btrfs_free_extent(root, blocknr, 1);
 			BUG_ON(ret);
 			continue;
 		}
 		BUG_ON(ret);
 		next = read_tree_block(root, blocknr);
 		if (path->nodes[*level-1])
-			tree_block_release(root, path->nodes[*level-1]);
+			btrfs_block_release(root, path->nodes[*level-1]);
 		path->nodes[*level-1] = next;
 		*level = btrfs_header_level(&next->node.header);
 		path->slots[*level] = 0;
 	}
 out:
-	ret = free_extent(root, path->nodes[*level]->blocknr, 1);
-	tree_block_release(root, path->nodes[*level]);
+	ret = btrfs_free_extent(root, path->nodes[*level]->blocknr, 1);
+	btrfs_block_release(root, path->nodes[*level]);
 	path->nodes[*level] = NULL;
 	*level += 1;
 	BUG_ON(ret);
 	return 0;
 }
 
-int walk_up_tree(struct ctree_root *root, struct ctree_path *path, int *level)
+int walk_up_tree(struct btrfs_root *root, struct btrfs_path *path, int *level)
 {
 	int i;
 	int slot;
 	int ret;
-	for(i = *level; i < MAX_LEVEL - 1 && path->nodes[i]; i++) {
+	for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
 		slot = path->slots[i];
 		if (slot <
 		    btrfs_header_nritems(&path->nodes[i]->node.header)- 1) {
@@ -479,9 +480,9 @@ int walk_up_tree(struct ctree_root *root, struct ctree_path *path, int *level)
 			*level = i;
 			return 0;
 		} else {
-			ret = free_extent(root,
+			ret = btrfs_free_extent(root,
 					  path->nodes[*level]->blocknr, 1);
-			tree_block_release(root, path->nodes[*level]);
+			btrfs_block_release(root, path->nodes[*level]);
 			path->nodes[*level] = NULL;
 			*level = i + 1;
 			BUG_ON(ret);
@@ -490,15 +491,15 @@ int walk_up_tree(struct ctree_root *root, struct ctree_path *path, int *level)
 	return 1;
 }
 
-int btrfs_drop_snapshot(struct ctree_root *root, struct tree_buffer *snap)
+int btrfs_drop_snapshot(struct btrfs_root *root, struct btrfs_buffer *snap)
 {
 	int ret;
 	int level;
-	struct ctree_path path;
+	struct btrfs_path path;
 	int i;
 	int orig_level;
 
-	init_path(&path);
+	btrfs_init_path(&path);
 
 	level = btrfs_header_level(&snap->node.header);
 	orig_level = level;
@@ -514,7 +515,7 @@ int btrfs_drop_snapshot(struct ctree_root *root, struct tree_buffer *snap)
 	}
 	for (i = 0; i <= orig_level; i++) {
 		if (path.nodes[i]) {
-			tree_block_release(root, path.nodes[i]);
+			btrfs_block_release(root, path.nodes[i]);
 		}
 	}
 
diff --git a/fs/btrfs/mkfs.c b/fs/btrfs/mkfs.c
index e0d3bc1d0259..fc1923320de3 100644
--- a/fs/btrfs/mkfs.c
+++ b/fs/btrfs/mkfs.c
@@ -12,10 +12,10 @@
 
 int mkfs(int fd)
 {
-	struct ctree_root_info info[2];
-	struct leaf empty_leaf;
+	struct btrfs_root_info info[2];
+	struct btrfs_leaf empty_leaf;
 	struct btrfs_item item;
-	struct extent_item extent_item;
+	struct btrfs_extent_item extent_item;
 	int ret;
 
 	/* setup the super block area */
@@ -28,7 +28,7 @@ int mkfs(int fd)
 	info[1].objectid = 2;
 	info[1].tree_root = 18;
 	ret = pwrite(fd, info, sizeof(info),
-		     CTREE_SUPER_INFO_OFFSET(CTREE_BLOCKSIZE));
+		     BTRFS_SUPER_INFO_OFFSET(BTRFS_BLOCKSIZE));
 	if (ret != sizeof(info))
 		return -1;
 
@@ -36,7 +36,7 @@ int mkfs(int fd)
 	memset(&empty_leaf, 0, sizeof(empty_leaf));
 	btrfs_set_header_parentid(&empty_leaf.header, 1);
 	btrfs_set_header_blocknr(&empty_leaf.header, 17);
-	ret = pwrite(fd, &empty_leaf, sizeof(empty_leaf), 17 * CTREE_BLOCKSIZE);
+	ret = pwrite(fd, &empty_leaf, sizeof(empty_leaf), 17 * BTRFS_BLOCKSIZE);
 	if (ret != sizeof(empty_leaf))
 		return -1;
 
@@ -48,9 +48,9 @@ int mkfs(int fd)
 	btrfs_set_key_objectid(&item.key, 0);
 	btrfs_set_key_offset(&item.key, 17);
 	btrfs_set_key_flags(&item.key, 0);
-	btrfs_set_item_offset(&item,
-			      LEAF_DATA_SIZE - sizeof(struct extent_item));
-	btrfs_set_item_size(&item, sizeof(struct extent_item));
+	btrfs_set_item_offset(&item, LEAF_DATA_SIZE -
+			      sizeof(struct btrfs_extent_item));
+	btrfs_set_item_size(&item, sizeof(struct btrfs_extent_item));
 	btrfs_set_extent_refs(&extent_item, 1);
 	btrfs_set_extent_owner(&extent_item, 0);
 	memcpy(empty_leaf.items, &item, sizeof(item));
@@ -60,8 +60,8 @@ int mkfs(int fd)
 	/* item2, give block 17 to the root */
 	btrfs_set_key_objectid(&item.key, 17);
 	btrfs_set_key_offset(&item.key, 1);
-	btrfs_set_item_offset(&item,
-			      LEAF_DATA_SIZE - sizeof(struct extent_item) * 2);
+	btrfs_set_item_offset(&item, LEAF_DATA_SIZE -
+			      sizeof(struct btrfs_extent_item) * 2);
 	btrfs_set_extent_owner(&extent_item, 1);
 	memcpy(empty_leaf.items + 1, &item, sizeof(item));
 	memcpy(empty_leaf.data + btrfs_item_offset(&item), &extent_item,
@@ -70,13 +70,13 @@ int mkfs(int fd)
 	/* item3, give block 18 for the extent root */
 	btrfs_set_key_objectid(&item.key, 18);
 	btrfs_set_key_offset(&item.key, 1);
-	btrfs_set_item_offset(&item,
-			      LEAF_DATA_SIZE - sizeof(struct extent_item) * 3);
+	btrfs_set_item_offset(&item, LEAF_DATA_SIZE -
+			      sizeof(struct btrfs_extent_item) * 3);
 	btrfs_set_extent_owner(&extent_item, 2);
 	memcpy(empty_leaf.items + 2, &item, sizeof(item));
 	memcpy(empty_leaf.data + btrfs_item_offset(&item), &extent_item,
 		btrfs_item_size(&item));
-	ret = pwrite(fd, &empty_leaf, sizeof(empty_leaf), 18 * CTREE_BLOCKSIZE);
+	ret = pwrite(fd, &empty_leaf, sizeof(empty_leaf), 18 * BTRFS_BLOCKSIZE);
 	if (ret != sizeof(empty_leaf))
 		return -1;
 	return 0;
diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c
index 101278e1139a..c95c85640aa9 100644
--- a/fs/btrfs/print-tree.c
+++ b/fs/btrfs/print-tree.c
@@ -5,14 +5,14 @@
 #include "ctree.h"
 #include "disk-io.h"
 
-void print_leaf(struct leaf *l)
+void btrfs_print_leaf(struct btrfs_leaf *l)
 {
 	int i;
 	u32 nr = btrfs_header_nritems(&l->header);
 	struct btrfs_item *item;
-	struct extent_item *ei;
+	struct btrfs_extent_item *ei;
 	printf("leaf %Lu total ptrs %d free space %d\n",
-		btrfs_header_blocknr(&l->header), nr, leaf_free_space(l));
+		btrfs_header_blocknr(&l->header), nr, btrfs_leaf_free_space(l));
 	fflush(stdout);
 	for (i = 0 ; i < nr ; i++) {
 		item = l->items + i;
@@ -26,24 +26,25 @@ void print_leaf(struct leaf *l)
 		fflush(stdout);
 		printf("\t\titem data %.*s\n", btrfs_item_size(item),
 			l->data + btrfs_item_offset(item));
-		ei = (struct extent_item *)(l->data + btrfs_item_offset(item));
+		ei = (struct btrfs_extent_item *)(l->data +
+						  btrfs_item_offset(item));
 		printf("\t\textent data refs %u owner %Lu\n", ei->refs,
 			ei->owner);
 		fflush(stdout);
 	}
 }
-void print_tree(struct ctree_root *root, struct tree_buffer *t)
+void btrfs_print_tree(struct btrfs_root *root, struct btrfs_buffer *t)
 {
 	int i;
 	u32 nr;
-	struct node *c;
+	struct btrfs_node *c;
 
 	if (!t)
 		return;
 	c = &t->node;
 	nr = btrfs_header_nritems(&c->header);
 	if (btrfs_is_leaf(c)) {
-		print_leaf((struct leaf *)c);
+		btrfs_print_leaf((struct btrfs_leaf *)c);
 		return;
 	}
 	printf("node %Lu level %d total ptrs %d free spc %u\n", t->blocknr,
@@ -58,17 +59,17 @@ void print_tree(struct ctree_root *root, struct tree_buffer *t)
 		fflush(stdout);
 	}
 	for (i = 0; i < nr; i++) {
-		struct tree_buffer *next_buf = read_tree_block(root,
+		struct btrfs_buffer *next_buf = read_tree_block(root,
 						btrfs_node_blockptr(c, i));
-		struct node *next = &next_buf->node;
+		struct btrfs_node *next = &next_buf->node;
 		if (btrfs_is_leaf(next) &&
 		    btrfs_header_level(&c->header) != 1)
 			BUG();
 		if (btrfs_header_level(&next->header) !=
 			btrfs_header_level(&c->header) - 1)
 			BUG();
-		print_tree(root, next_buf);
-		tree_block_release(root, next_buf);
+		btrfs_print_tree(root, next_buf);
+		btrfs_block_release(root, next_buf);
 	}
 
 }
diff --git a/fs/btrfs/print-tree.h b/fs/btrfs/print-tree.h
index 3c1e9a3e0260..e8d0b847c023 100644
--- a/fs/btrfs/print-tree.h
+++ b/fs/btrfs/print-tree.h
@@ -1,3 +1,3 @@
 
-void print_leaf(struct leaf *l);
-void print_tree(struct ctree_root *root, struct tree_buffer *t);
+void btrfs_print_leaf(struct btrfs_leaf *l);
+void btrfs_print_tree(struct btrfs_root *root, struct btrfs_buffer *t);
diff --git a/fs/btrfs/quick-test.c b/fs/btrfs/quick-test.c
index ab3bda53a2f6..66bdc57905d7 100644
--- a/fs/btrfs/quick-test.c
+++ b/fs/btrfs/quick-test.c
@@ -22,9 +22,9 @@ int main(int ac, char **av) {
 	int run_size = 100000;
 	int max_key =  100000000;
 	int tree_size = 0;
-	struct ctree_path path;
-	struct ctree_super_block super;
-	struct ctree_root *root;
+	struct btrfs_path path;
+	struct btrfs_super_block super;
+	struct btrfs_root *root;
 
 	radix_tree_init();
 
@@ -40,12 +40,12 @@ int main(int ac, char **av) {
 		ins.objectid = num;
 		ins.offset = 0;
 		ins.flags = 0;
-		ret = insert_item(root, &ins, buf, strlen(buf));
+		ret = btrfs_insert_item(root, &ins, buf, strlen(buf));
 		if (!ret)
 			tree_size++;
 		free(buf);
 		if (i == run_size - 5) {
-			commit_transaction(root, &super);
+			btrfs_commit_transaction(root, &super);
 		}
 
 	}
@@ -57,16 +57,16 @@ int main(int ac, char **av) {
 	for (i = 0; i < run_size; i++) {
 		num = next_key(i, max_key);
 		ins.objectid = num;
-		init_path(&path);
+		btrfs_init_path(&path);
 		if (i % 10000 == 0)
 			fprintf(stderr, "search %d:%d\n", num, i);
-		ret = search_slot(root, &ins, &path, 0, 0);
+		ret = btrfs_search_slot(root, &ins, &path, 0, 0);
 		if (ret) {
-			print_tree(root, root->node);
+			btrfs_print_tree(root, root->node);
 			printf("unable to find %d\n", num);
 			exit(1);
 		}
-		release_path(root, &path);
+		btrfs_release_path(root, &path);
 	}
 	close_ctree(root, &super);
 	root = open_ctree("dbfile", &super);
@@ -81,17 +81,17 @@ int main(int ac, char **av) {
 	for (i = 0 ; i < run_size/4; i++) {
 		num = next_key(i, max_key);
 		ins.objectid = num;
-		init_path(&path);
-		ret = search_slot(root, &ins, &path, -1, 1);
+		btrfs_init_path(&path);
+		ret = btrfs_search_slot(root, &ins, &path, -1, 1);
 		if (!ret) {
 			if (i % 10000 == 0)
 				fprintf(stderr, "del %d:%d\n", num, i);
-			ret = del_item(root, &path);
+			ret = btrfs_del_item(root, &path);
 			if (ret != 0)
 				BUG();
 			tree_size--;
 		}
-		release_path(root, &path);
+		btrfs_release_path(root, &path);
 	}
 	close_ctree(root, &super);
 	root = open_ctree("dbfile", &super);
@@ -103,7 +103,7 @@ int main(int ac, char **av) {
 		ins.objectid = num;
 		if (i % 10000 == 0)
 			fprintf(stderr, "insert %d:%d\n", num, i);
-		ret = insert_item(root, &ins, buf, strlen(buf));
+		ret = btrfs_insert_item(root, &ins, buf, strlen(buf));
 		if (!ret)
 			tree_size++;
 		free(buf);
@@ -115,25 +115,25 @@ int main(int ac, char **av) {
 	for (i = 0; i < run_size; i++) {
 		num = next_key(i, max_key);
 		ins.objectid = num;
-		init_path(&path);
+		btrfs_init_path(&path);
 		if (i % 10000 == 0)
 			fprintf(stderr, "search %d:%d\n", num, i);
-		ret = search_slot(root, &ins, &path, 0, 0);
+		ret = btrfs_search_slot(root, &ins, &path, 0, 0);
 		if (ret) {
-			print_tree(root, root->node);
+			btrfs_print_tree(root, root->node);
 			printf("unable to find %d\n", num);
 			exit(1);
 		}
-		release_path(root, &path);
+		btrfs_release_path(root, &path);
 	}
 	printf("starting big long delete run\n");
 	while(root->node &&
 	      btrfs_header_nritems(&root->node->node.header) > 0) {
-		struct leaf *leaf;
+		struct btrfs_leaf *leaf;
 		int slot;
 		ins.objectid = (u64)-1;
-		init_path(&path);
-		ret = search_slot(root, &ins, &path, -1, 1);
+		btrfs_init_path(&path);
+		ret = btrfs_search_slot(root, &ins, &path, -1, 1);
 		if (ret == 0)
 			BUG();
 
@@ -149,26 +149,26 @@ int main(int ac, char **av) {
 			btrfs_disk_key_to_cpu(&last, &leaf->items[slot].key);
 			if (tree_size % 10000 == 0)
 				printf("big del %d:%d\n", tree_size, i);
-			ret = del_item(root, &path);
+			ret = btrfs_del_item(root, &path);
 			if (ret != 0) {
 				printf("del_item returned %d\n", ret);
 				BUG();
 			}
 			tree_size--;
 		}
-		release_path(root, &path);
+		btrfs_release_path(root, &path);
 	}
 	/*
 	printf("previous tree:\n");
-	print_tree(root, root->commit_root);
+	btrfs_print_tree(root, root->commit_root);
 	printf("map before commit\n");
-	print_tree(root->extent_root, root->extent_root->node);
+	btrfs_print_tree(root->extent_root, root->extent_root->node);
 	*/
-	commit_transaction(root, &super);
+	btrfs_commit_transaction(root, &super);
 	printf("tree size is now %d\n", tree_size);
 	printf("root %p commit root %p\n", root->node, root->commit_root);
 	printf("map tree\n");
-	print_tree(root->extent_root, root->extent_root->node);
+	btrfs_print_tree(root->extent_root, root->extent_root->node);
 	close_ctree(root, &super);
 	return 0;
 }
diff --git a/fs/btrfs/random-test.c b/fs/btrfs/random-test.c
index e767528bc521..a8ef0478991b 100644
--- a/fs/btrfs/random-test.c
+++ b/fs/btrfs/random-test.c
@@ -8,7 +8,7 @@
 #include "print-tree.h"
 
 int keep_running = 1;
-struct ctree_super_block super;
+struct btrfs_super_block super;
 
 static int setup_key(struct radix_tree_root *root, struct btrfs_key *key,
 		     int exists)
@@ -36,17 +36,17 @@ again:
 	return 0;
 }
 
-static int ins_one(struct ctree_root *root, struct radix_tree_root *radix)
+static int ins_one(struct btrfs_root *root, struct radix_tree_root *radix)
 {
-	struct ctree_path path;
+	struct btrfs_path path;
 	struct btrfs_key key;
 	int ret;
 	char buf[128];
 	unsigned long oid;
-	init_path(&path);
+	btrfs_init_path(&path);
 	ret = setup_key(radix, &key, 0);
 	sprintf(buf, "str-%Lu\n", key.objectid);
-	ret = insert_item(root, &key, buf, strlen(buf));
+	ret = btrfs_insert_item(root, &key, buf, strlen(buf));
 	if (ret)
 		goto error;
 	oid = (unsigned long)key.objectid;
@@ -61,18 +61,18 @@ error:
 	return -1;
 }
 
-static int insert_dup(struct ctree_root *root, struct radix_tree_root *radix)
+static int insert_dup(struct btrfs_root *root, struct radix_tree_root *radix)
 {
-	struct ctree_path path;
+	struct btrfs_path path;
 	struct btrfs_key key;
 	int ret;
 	char buf[128];
-	init_path(&path);
+	btrfs_init_path(&path);
 	ret = setup_key(radix, &key, 1);
 	if (ret < 0)
 		return 0;
 	sprintf(buf, "str-%Lu\n", key.objectid);
-	ret = insert_item(root, &key, buf, strlen(buf));
+	ret = btrfs_insert_item(root, &key, buf, strlen(buf));
 	if (ret != -EEXIST) {
 		printf("insert on %Lu gave us %d\n", key.objectid, ret);
 		return 1;
@@ -80,21 +80,21 @@ static int insert_dup(struct ctree_root *root, struct radix_tree_root *radix)
 	return 0;
 }
 
-static int del_one(struct ctree_root *root, struct radix_tree_root *radix)
+static int del_one(struct btrfs_root *root, struct radix_tree_root *radix)
 {
-	struct ctree_path path;
+	struct btrfs_path path;
 	struct btrfs_key key;
 	int ret;
 	unsigned long *ptr;
-	init_path(&path);
+	btrfs_init_path(&path);
 	ret = setup_key(radix, &key, 1);
 	if (ret < 0)
 		return 0;
-	ret = search_slot(root, &key, &path, -1, 1);
+	ret = btrfs_search_slot(root, &key, &path, -1, 1);
 	if (ret)
 		goto error;
-	ret = del_item(root, &path);
-	release_path(root, &path);
+	ret = btrfs_del_item(root, &path);
+	btrfs_release_path(root, &path);
 	if (ret != 0)
 		goto error;
 	ptr = radix_tree_delete(radix, key.objectid);
@@ -106,17 +106,17 @@ error:
 	return -1;
 }
 
-static int lookup_item(struct ctree_root *root, struct radix_tree_root *radix)
+static int lookup_item(struct btrfs_root *root, struct radix_tree_root *radix)
 {
-	struct ctree_path path;
+	struct btrfs_path path;
 	struct btrfs_key key;
 	int ret;
-	init_path(&path);
+	btrfs_init_path(&path);
 	ret = setup_key(radix, &key, 1);
 	if (ret < 0)
 		return 0;
-	ret = search_slot(root, &key, &path, 0, 1);
-	release_path(root, &path);
+	ret = btrfs_search_slot(root, &key, &path, 0, 1);
+	btrfs_release_path(root, &path);
 	if (ret)
 		goto error;
 	return 0;
@@ -125,17 +125,17 @@ error:
 	return -1;
 }
 
-static int lookup_enoent(struct ctree_root *root, struct radix_tree_root *radix)
+static int lookup_enoent(struct btrfs_root *root, struct radix_tree_root *radix)
 {
-	struct ctree_path path;
+	struct btrfs_path path;
 	struct btrfs_key key;
 	int ret;
-	init_path(&path);
+	btrfs_init_path(&path);
 	ret = setup_key(radix, &key, 0);
 	if (ret < 0)
 		return ret;
-	ret = search_slot(root, &key, &path, 0, 0);
-	release_path(root, &path);
+	ret = btrfs_search_slot(root, &key, &path, 0, 0);
+	btrfs_release_path(root, &path);
 	if (ret <= 0)
 		goto error;
 	return 0;
@@ -144,10 +144,10 @@ error:
 	return -1;
 }
 
-static int empty_tree(struct ctree_root *root, struct radix_tree_root *radix,
+static int empty_tree(struct btrfs_root *root, struct radix_tree_root *radix,
 		      int nr)
 {
-	struct ctree_path path;
+	struct btrfs_path path;
 	struct btrfs_key key;
 	unsigned long found = 0;
 	int ret;
@@ -159,22 +159,22 @@ static int empty_tree(struct ctree_root *root, struct radix_tree_root *radix,
 	key.flags = 0;
 	key.objectid = (unsigned long)-1;
 	while(nr-- >= 0) {
-		init_path(&path);
-		ret = search_slot(root, &key, &path, -1, 1);
+		btrfs_init_path(&path);
+		ret = btrfs_search_slot(root, &key, &path, -1, 1);
 		if (ret < 0) {
-			release_path(root, &path);
+			btrfs_release_path(root, &path);
 			return ret;
 		}
 		if (ret != 0) {
 			if (path.slots[0] == 0) {
-				release_path(root, &path);
+				btrfs_release_path(root, &path);
 				break;
 			}
 			path.slots[0] -= 1;
 		}
 		slot = path.slots[0];
 		found=btrfs_key_objectid(&path.nodes[0]->leaf.items[slot].key);
-		ret = del_item(root, &path);
+		ret = btrfs_del_item(root, &path);
 		count++;
 		if (ret) {
 			fprintf(stderr,
@@ -182,7 +182,7 @@ static int empty_tree(struct ctree_root *root, struct radix_tree_root *radix,
 				found);
 			return -1;
 		}
-		release_path(root, &path);
+		btrfs_release_path(root, &path);
 		ptr = radix_tree_delete(radix, found);
 		if (!ptr)
 			goto error;
@@ -195,7 +195,7 @@ error:
 	return -1;
 }
 
-static int fill_tree(struct ctree_root *root, struct radix_tree_root *radix,
+static int fill_tree(struct btrfs_root *root, struct radix_tree_root *radix,
 		     int count)
 {
 	int i;
@@ -207,7 +207,7 @@ static int fill_tree(struct ctree_root *root, struct radix_tree_root *radix,
 			goto out;
 		}
 		if (i % 1000 == 0) {
-			ret = commit_transaction(root, &super);
+			ret = btrfs_commit_transaction(root, &super);
 			if (ret) {
 				fprintf(stderr, "fill commit failed\n");
 				return ret;
@@ -223,7 +223,7 @@ out:
 	return ret;
 }
 
-static int bulk_op(struct ctree_root *root, struct radix_tree_root *radix)
+static int bulk_op(struct btrfs_root *root, struct radix_tree_root *radix)
 {
 	int ret;
 	int nr = rand() % 5000;
@@ -242,13 +242,13 @@ static int bulk_op(struct ctree_root *root, struct radix_tree_root *radix)
 }
 
 
-int (*ops[])(struct ctree_root *root, struct radix_tree_root *radix) =
+int (*ops[])(struct btrfs_root *root, struct radix_tree_root *radix) =
 	{ ins_one, insert_dup, del_one, lookup_item,
 	  lookup_enoent, bulk_op };
 
-static int fill_radix(struct ctree_root *root, struct radix_tree_root *radix)
+static int fill_radix(struct btrfs_root *root, struct radix_tree_root *radix)
 {
-	struct ctree_path path;
+	struct btrfs_path path;
 	struct btrfs_key key;
 	unsigned long found;
 	int ret;
@@ -259,16 +259,16 @@ static int fill_radix(struct ctree_root *root, struct radix_tree_root *radix)
 	key.flags = 0;
 	key.objectid = (unsigned long)-1;
 	while(1) {
-		init_path(&path);
-		ret = search_slot(root, &key, &path, 0, 0);
+		btrfs_init_path(&path);
+		ret = btrfs_search_slot(root, &key, &path, 0, 0);
 		if (ret < 0) {
-			release_path(root, &path);
+			btrfs_release_path(root, &path);
 			return ret;
 		}
 		slot = path.slots[0];
 		if (ret != 0) {
 			if (slot == 0) {
-				release_path(root, &path);
+				btrfs_release_path(root, &path);
 				break;
 			}
 			slot -= 1;
@@ -287,7 +287,7 @@ static int fill_radix(struct ctree_root *root, struct radix_tree_root *radix)
 
 			radix_tree_preload_end();
 		}
-		release_path(root, &path);
+		btrfs_release_path(root, &path);
 		key.objectid = found - 1;
 		if (key.objectid > found)
 			break;
@@ -312,7 +312,7 @@ int print_usage(void)
 int main(int ac, char **av)
 {
 	RADIX_TREE(radix, GFP_KERNEL);
-	struct ctree_root *root;
+	struct btrfs_root *root;
 	int i;
 	int ret;
 	int count;
@@ -370,7 +370,7 @@ int main(int ac, char **av)
 			if (ret) {
 				fprintf(stderr, "op %d failed %d:%d\n",
 					op, i, iterations);
-				print_tree(root, root->node);
+				btrfs_print_tree(root, root->node);
 				fprintf(stderr, "op %d failed %d:%d\n",
 					op, i, iterations);
 				err = ret;
-- 
cgit v1.2.3


From 3768f3689fc76ecea17414936dff7a02746a4355 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Tue, 13 Mar 2007 16:47:54 -0400
Subject: Btrfs: Change the super to point to a tree of trees to enable
 persistent snapshots

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/Makefile      |   3 +-
 fs/btrfs/TODO          |   1 +
 fs/btrfs/ctree.h       | 123 ++++++++++++++++++++++++++++++++++++++------
 fs/btrfs/debug-tree.c  |   4 +-
 fs/btrfs/disk-io.c     | 135 +++++++++++++++++++++++++++++++++++++++----------
 fs/btrfs/disk-io.h     |   3 +-
 fs/btrfs/extent-tree.c |  15 +++---
 fs/btrfs/mkfs.c        | 128 +++++++++++++++++++++++++++++++++++++++++++---
 fs/btrfs/print-tree.c  |  10 ++--
 fs/btrfs/root-tree.c   |  88 ++++++++++++++++++++++++++++++++
 10 files changed, 445 insertions(+), 65 deletions(-)
 create mode 100644 fs/btrfs/root-tree.c

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index f8532200e9ad..13ed6313ac85 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -2,7 +2,8 @@
 CC=gcc
 CFLAGS = -g -Wall
 headers = radix-tree.h ctree.h disk-io.h kerncompat.h print-tree.h list.h
-objects = ctree.o disk-io.o radix-tree.o mkfs.o extent-tree.o print-tree.o
+objects = ctree.o disk-io.o radix-tree.o mkfs.o extent-tree.o print-tree.o \
+	  root-tree.o
 
 # if you don't have sparse installed, use ls instead
 CHECKFLAGS=-D__linux__ -Dlinux -D__STDC__ -Dunix -D__unix__ -Wbitwise \
diff --git a/fs/btrfs/TODO b/fs/btrfs/TODO
index ad29267b8738..ff90e5eaf431 100644
--- a/fs/btrfs/TODO
+++ b/fs/btrfs/TODO
@@ -5,6 +5,7 @@
 * make a real mkfs and superblock
 * Do checksumming
 * Define FS objects in terms of different item types
+* add inode tree
 * Add block mapping tree (simple dm layer)
 * Add simple tree locking (semaphore per tree)
 * Make allocator smarter
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 5dfcc90812c1..0968899fb7f9 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -4,8 +4,13 @@
 #include "list.h"
 #include "kerncompat.h"
 
+#define BTRFS_MAGIC "_BtRfS_M"
 #define BTRFS_BLOCKSIZE 1024
 
+#define BTRFS_ROOT_TREE_OBJECTID 1
+#define BTRFS_EXTENT_TREE_OBJECTID 2
+#define BTRFS_FS_TREE_OBJECTID 3
+
 /*
  * the key defines the order in the tree, and so it also defines (optimal)
  * block layout.  objectid corresonds to the inode number.  The flags
@@ -36,7 +41,7 @@ struct btrfs_key {
  * every tree block (leaf or node) starts with this header.
  */
 struct btrfs_header {
-	__le64 fsid[2]; /* FS specific uuid */
+	u8 fsid[16]; /* FS specific uuid */
 	__le64 blocknr; /* which block this node is supposed to live in */
 	__le64 parentid; /* objectid of the tree root */
 	__le32 csum;
@@ -52,6 +57,14 @@ struct btrfs_header {
 
 struct btrfs_buffer;
 
+struct btrfs_root_item {
+	__le64 blocknr;
+	__le32 flags;
+	__le64 block_limit;
+	__le64 blocks_used;
+	__le32 refs;
+};
+
 /*
  * in ram representation of the tree.  extent_root is used for all allocations
  * and for the extent tree extent_root root.  current_insert is used
@@ -61,6 +74,7 @@ struct btrfs_root {
 	struct btrfs_buffer *node;
 	struct btrfs_buffer *commit_root;
 	struct btrfs_root *extent_root;
+	struct btrfs_root *tree_root;
 	struct btrfs_key current_insert;
 	struct btrfs_key last_insert;
 	int fp;
@@ -69,28 +83,25 @@ struct btrfs_root {
 	struct list_head trans;
 	struct list_head cache;
 	int cache_size;
+	int ref_cows;
+	struct btrfs_root_item root_item;
+	struct btrfs_key root_key;
 };
 
-/*
- * describes a tree on disk
- */
-struct btrfs_root_info {
-	u64 fsid[2]; /* FS specific uuid */
-	u64 blocknr; /* blocknr of this block */
-	u64 objectid; /* inode number of this root */
-	u64 tree_root; /* the tree root block */
-	u32 csum;
-	u32 ham;
-	u64 snapuuid[2]; /* root specific uuid */
-} __attribute__ ((__packed__));
-
 /*
  * the super block basically lists the main trees of the FS
  * it currently lacks any block count etc etc
  */
 struct btrfs_super_block {
-	struct btrfs_root_info root_info;
-	struct btrfs_root_info extent_info;
+	u8 fsid[16];    /* FS specific uuid */
+	__le64 blocknr; /* this block number */
+	__le32 csum;
+	__le64 magic;
+	__le16 blocksize;
+	__le64 generation;
+	__le64 root;
+	__le64 total_blocks;
+	__le64 blocks_used;
 } __attribute__ ((__packed__));
 
 /*
@@ -317,6 +328,79 @@ static inline int btrfs_is_leaf(struct btrfs_node *n)
 	return (btrfs_header_level(&n->header) == 0);
 }
 
+static inline u64 btrfs_root_blocknr(struct btrfs_root_item *item)
+{
+	return le64_to_cpu(item->blocknr);
+}
+
+static inline void btrfs_set_root_blocknr(struct btrfs_root_item *item, u64 val)
+{
+	item->blocknr = cpu_to_le64(val);
+}
+
+static inline u32 btrfs_root_refs(struct btrfs_root_item *item)
+{
+	return le32_to_cpu(item->refs);
+}
+
+static inline void btrfs_set_root_refs(struct btrfs_root_item *item, u32 val)
+{
+	item->refs = cpu_to_le32(val);
+}
+
+static inline u64 btrfs_super_blocknr(struct btrfs_super_block *s)
+{
+	return le64_to_cpu(s->blocknr);
+}
+
+static inline void btrfs_set_super_blocknr(struct btrfs_super_block *s, u64 val)
+{
+	s->blocknr = cpu_to_le64(val);
+}
+
+static inline u64 btrfs_super_root(struct btrfs_super_block *s)
+{
+	return le64_to_cpu(s->root);
+}
+
+static inline void btrfs_set_super_root(struct btrfs_super_block *s, u64 val)
+{
+	s->root = cpu_to_le64(val);
+}
+
+static inline u64 btrfs_super_total_blocks(struct btrfs_super_block *s)
+{
+	return le64_to_cpu(s->total_blocks);
+}
+
+static inline void btrfs_set_super_total_blocks(struct btrfs_super_block *s,
+						u64 val)
+{
+	s->total_blocks = cpu_to_le64(val);
+}
+
+static inline u64 btrfs_super_blocks_used(struct btrfs_super_block *s)
+{
+	return le64_to_cpu(s->blocks_used);
+}
+
+static inline void btrfs_set_super_blocks_used(struct btrfs_super_block *s,
+						u64 val)
+{
+	s->blocks_used = cpu_to_le64(val);
+}
+
+static inline u16 btrfs_super_blocksize(struct btrfs_super_block *s)
+{
+	return le16_to_cpu(s->blocksize);
+}
+
+static inline void btrfs_set_super_blocksize(struct btrfs_super_block *s,
+						u16 val)
+{
+	s->blocksize = cpu_to_le16(val);
+}
+
 struct btrfs_buffer *btrfs_alloc_free_block(struct btrfs_root *root);
 int btrfs_inc_ref(struct btrfs_root *root, struct btrfs_buffer *buf);
 int btrfs_free_extent(struct btrfs_root *root, u64 blocknr, u64 num_blocks);
@@ -331,4 +415,11 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path);
 int btrfs_leaf_free_space(struct btrfs_leaf *leaf);
 int btrfs_drop_snapshot(struct btrfs_root *root, struct btrfs_buffer *snap);
 int btrfs_finish_extent_commit(struct btrfs_root *root);
+int btrfs_del_root(struct btrfs_root *root, struct btrfs_key *key);
+int btrfs_insert_root(struct btrfs_root *root, struct btrfs_key *key,
+		      struct btrfs_root_item *item);
+int btrfs_update_root(struct btrfs_root *root, struct btrfs_key *key,
+		      struct btrfs_root_item *item);
+int btrfs_find_last_root(struct btrfs_root *root, u64 objectid,
+			struct btrfs_root_item *item, struct btrfs_key *key);
 #endif
diff --git a/fs/btrfs/debug-tree.c b/fs/btrfs/debug-tree.c
index 6da0a7aa0f47..de45fb4dfddd 100644
--- a/fs/btrfs/debug-tree.c
+++ b/fs/btrfs/debug-tree.c
@@ -11,9 +11,11 @@ int main(int ac, char **av) {
 	struct btrfs_root *root;
 	radix_tree_init();
 	root = open_ctree("dbfile", &super);
-	printf("root tree\n");
+	printf("fs tree\n");
 	btrfs_print_tree(root, root->node);
 	printf("map tree\n");
 	btrfs_print_tree(root->extent_root, root->extent_root->node);
+	printf("root tree\n");
+	btrfs_print_tree(root->tree_root, root->tree_root->node);
 	return 0;
 }
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index c34c0c60935f..3d4bf6833f2a 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -154,41 +154,96 @@ static int __commit_transaction(struct btrfs_root *root)
 	return ret;
 }
 
+static int commit_extent_and_tree_roots(struct btrfs_root *tree_root,
+					struct btrfs_root *extent_root)
+{
+	int ret;
+	u64 old_extent_block;
+
+	while(1) {
+		old_extent_block = btrfs_root_blocknr(&extent_root->root_item);
+		if (old_extent_block == extent_root->node->blocknr)
+			break;
+		btrfs_set_root_blocknr(&extent_root->root_item,
+				       extent_root->node->blocknr);
+		ret = btrfs_update_root(tree_root,
+					&extent_root->root_key,
+					&extent_root->root_item);
+		BUG_ON(ret);
+	}
+	__commit_transaction(extent_root);
+	__commit_transaction(tree_root);
+	return 0;
+}
+
 int btrfs_commit_transaction(struct btrfs_root *root,
 			     struct btrfs_super_block *s)
 {
 	int ret = 0;
+	struct btrfs_buffer *snap = root->commit_root;
+	struct btrfs_key snap_key;
 
 	ret = __commit_transaction(root);
-	if (!ret && root != root->extent_root)
-		ret = __commit_transaction(root->extent_root);
 	BUG_ON(ret);
-	if (root->commit_root != root->node) {
-		struct btrfs_buffer *snap = root->commit_root;
-		root->commit_root = root->node;
-		root->node->count++;
-		ret = btrfs_drop_snapshot(root, snap);
-		BUG_ON(ret);
-		// btrfs_block_release(root, snap);
-	}
+
+	if (root->commit_root == root->node)
+		return 0;
+
+	memcpy(&snap_key, &root->root_key, sizeof(snap_key));
+	root->root_key.offset++;
+
+	btrfs_set_root_blocknr(&root->root_item, root->node->blocknr);
+	ret = btrfs_insert_root(root->tree_root, &root->root_key,
+				&root->root_item);
+	BUG_ON(ret);
+
+	ret = commit_extent_and_tree_roots(root->tree_root, root->extent_root);
+	BUG_ON(ret);
+
         write_ctree_super(root, s);
-	btrfs_finish_extent_commit(root);
+	btrfs_finish_extent_commit(root->extent_root);
+	btrfs_finish_extent_commit(root->tree_root);
+
+	root->commit_root = root->node;
+	root->node->count++;
+	ret = btrfs_drop_snapshot(root, snap);
+	BUG_ON(ret);
+
+	ret = btrfs_del_root(root->tree_root, &snap_key);
+	BUG_ON(ret);
+
 	return ret;
 }
 
-static int __setup_root(struct btrfs_root *root, struct btrfs_root *extent_root,
-			struct btrfs_root_info *info, int fp)
+static int __setup_root(struct btrfs_root *root, u64 objectid, int fp)
 {
 	INIT_LIST_HEAD(&root->trans);
 	INIT_LIST_HEAD(&root->cache);
 	root->cache_size = 0;
 	root->fp = fp;
 	root->node = NULL;
-	root->extent_root = extent_root;
 	root->commit_root = NULL;
-	root->node = read_tree_block(root, info->tree_root);
 	memset(&root->current_insert, 0, sizeof(root->current_insert));
 	memset(&root->last_insert, 0, sizeof(root->last_insert));
+	memset(&root->root_key, 0, sizeof(root->root_key));
+	memset(&root->root_item, 0, sizeof(root->root_item));
+	return 0;
+}
+
+static int find_and_setup_root(struct btrfs_root *tree_root, u64 objectid,
+			struct btrfs_root *root, int fp)
+{
+	int ret;
+
+	__setup_root(root, objectid, fp);
+	ret = btrfs_find_last_root(tree_root, objectid,
+				   &root->root_item, &root->root_key);
+	BUG_ON(ret);
+
+	root->node = read_tree_block(root,
+				     btrfs_root_blocknr(&root->root_item));
+	root->ref_cows = 0;
+	BUG_ON(!root->node);
 	return 0;
 }
 
@@ -196,9 +251,19 @@ struct btrfs_root *open_ctree(char *filename, struct btrfs_super_block *super)
 {
 	struct btrfs_root *root = malloc(sizeof(struct btrfs_root));
 	struct btrfs_root *extent_root = malloc(sizeof(struct btrfs_root));
+	struct btrfs_root *tree_root = malloc(sizeof(struct btrfs_root));
 	int fp;
 	int ret;
 
+	root->extent_root = extent_root;
+	root->tree_root = tree_root;
+
+	extent_root->extent_root = extent_root;
+	extent_root->tree_root = tree_root;
+
+	tree_root->extent_root = extent_root;
+	tree_root->tree_root = tree_root;
+
 	fp = open(filename, O_CREAT | O_RDWR, 0600);
 	if (fp < 0) {
 		free(root);
@@ -208,11 +273,14 @@ struct btrfs_root *open_ctree(char *filename, struct btrfs_super_block *super)
 	INIT_RADIX_TREE(&root->pinned_radix, GFP_KERNEL);
 	INIT_RADIX_TREE(&extent_root->pinned_radix, GFP_KERNEL);
 	INIT_RADIX_TREE(&extent_root->cache_radix, GFP_KERNEL);
+	INIT_RADIX_TREE(&tree_root->pinned_radix, GFP_KERNEL);
+	INIT_RADIX_TREE(&tree_root->cache_radix, GFP_KERNEL);
+
 	ret = pread(fp, super, sizeof(struct btrfs_super_block),
 		     BTRFS_SUPER_INFO_OFFSET(BTRFS_BLOCKSIZE));
-	if (ret == 0 || super->root_info.tree_root == 0) {
+	if (ret == 0 || btrfs_super_root(super) == 0) {
 		printf("making new FS!\n");
-		ret = mkfs(fp);
+		ret = mkfs(fp, 0, BTRFS_BLOCKSIZE);
 		if (ret)
 			return NULL;
 		ret = pread(fp, super, sizeof(struct btrfs_super_block),
@@ -221,24 +289,29 @@ struct btrfs_root *open_ctree(char *filename, struct btrfs_super_block *super)
 			return NULL;
 	}
 	BUG_ON(ret < 0);
-	__setup_root(root, extent_root, &super->root_info, fp);
-	__setup_root(extent_root, extent_root, &super->extent_info, fp);
+
+	__setup_root(tree_root, BTRFS_ROOT_TREE_OBJECTID, fp);
+	tree_root->node = read_tree_block(tree_root, btrfs_super_root(super));
+	BUG_ON(!tree_root->node);
+
+	ret = find_and_setup_root(tree_root, BTRFS_EXTENT_TREE_OBJECTID,
+				  extent_root, fp);
+	BUG_ON(ret);
+
+	ret = find_and_setup_root(tree_root, BTRFS_FS_TREE_OBJECTID,
+				  root, fp);
+	BUG_ON(ret);
+
 	root->commit_root = root->node;
 	root->node->count++;
+	root->ref_cows = 1;
 	return root;
 }
 
-static int __update_root(struct btrfs_root *root, struct btrfs_root_info *info)
-{
-	info->tree_root = root->node->blocknr;
-	return 0;
-}
-
 int write_ctree_super(struct btrfs_root *root, struct btrfs_super_block *s)
 {
 	int ret;
-	__update_root(root, &s->root_info);
-	__update_root(root->extent_root, &s->extent_info);
+	btrfs_set_super_root(s, root->tree_root->node->blocknr);
 	ret = pwrite(root->fp, s, sizeof(*s),
 		     BTRFS_SUPER_INFO_OFFSET(BTRFS_BLOCKSIZE));
 	if (ret != sizeof(*s)) {
@@ -260,19 +333,25 @@ static int drop_cache(struct btrfs_root *root)
 }
 int close_ctree(struct btrfs_root *root, struct btrfs_super_block *s)
 {
+	int ret;
 	btrfs_commit_transaction(root, s);
-	__commit_transaction(root->extent_root);
+	ret = commit_extent_and_tree_roots(root->tree_root, root->extent_root);
+	BUG_ON(ret);
 	write_ctree_super(root, s);
 	drop_cache(root->extent_root);
+	drop_cache(root->tree_root);
 	drop_cache(root);
 	BUG_ON(!list_empty(&root->trans));
 	BUG_ON(!list_empty(&root->extent_root->trans));
+	BUG_ON(!list_empty(&root->tree_root->trans));
 
 	close(root->fp);
 	if (root->node)
 		btrfs_block_release(root, root->node);
 	if (root->extent_root->node)
 		btrfs_block_release(root->extent_root, root->extent_root->node);
+	if (root->tree_root->node)
+		btrfs_block_release(root->tree_root, root->tree_root->node);
 	btrfs_block_release(root, root->commit_root);
 	free(root);
 	printf("on close %d blocks are allocated\n", allocated_blocks);
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index b391335864b1..c22a61f9233f 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -24,7 +24,8 @@ struct btrfs_root *open_ctree(char *filename, struct btrfs_super_block *s);
 int close_ctree(struct btrfs_root *root, struct btrfs_super_block *s);
 void btrfs_block_release(struct btrfs_root *root, struct btrfs_buffer *buf);
 int write_ctree_super(struct btrfs_root *root, struct btrfs_super_block *s);
-int mkfs(int fd);
+int mkfs(int fd, u64 num_blocks, u16 blocksize);
+
 
 #define BTRFS_SUPER_INFO_OFFSET(bs) (16 * (bs))
 
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index a6969538bca2..21f39b404697 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -82,7 +82,7 @@ int btrfs_inc_ref(struct btrfs_root *root, struct btrfs_buffer *buf)
 	u64 blocknr;
 	int i;
 
-	if (root == root->extent_root)
+	if (!root->ref_cows)
 		return 0;
 	if (btrfs_is_leaf(&buf->node))
 		return 0;
@@ -96,23 +96,22 @@ int btrfs_inc_ref(struct btrfs_root *root, struct btrfs_buffer *buf)
 
 int btrfs_finish_extent_commit(struct btrfs_root *root)
 {
-	struct btrfs_root *extent_root = root->extent_root;
 	unsigned long gang[8];
 	int ret;
 	int i;
 
 	while(1) {
-		ret = radix_tree_gang_lookup(&extent_root->pinned_radix,
+		ret = radix_tree_gang_lookup(&root->pinned_radix,
 						 (void **)gang, 0,
 						 ARRAY_SIZE(gang));
 		if (!ret)
 			break;
 		for (i = 0; i < ret; i++) {
-			radix_tree_delete(&extent_root->pinned_radix, gang[i]);
+			radix_tree_delete(&root->pinned_radix, gang[i]);
 		}
 	}
-	extent_root->last_insert.objectid = 0;
-	extent_root->last_insert.offset = 0;
+	root->last_insert.objectid = 0;
+	root->last_insert.offset = 0;
 	return 0;
 }
 
@@ -173,7 +172,7 @@ static int __free_extent(struct btrfs_root *root, u64 blocknr, u64 num_blocks)
 	refs = btrfs_extent_refs(ei) - 1;
 	btrfs_set_extent_refs(ei, refs);
 	if (refs == 0) {
-		if (root == extent_root) {
+		if (!root->ref_cows) {
 			int err;
 			radix_tree_preload(GFP_KERNEL);
 			err = radix_tree_insert(&extent_root->pinned_radix,
@@ -513,7 +512,7 @@ static int walk_up_tree(struct btrfs_root *root, struct btrfs_path *path,
  */
 int btrfs_drop_snapshot(struct btrfs_root *root, struct btrfs_buffer *snap)
 {
-	int ret = 0;;
+	int ret = 0;
 	int wret;
 	int level;
 	struct btrfs_path path;
diff --git a/fs/btrfs/mkfs.c b/fs/btrfs/mkfs.c
index fc1923320de3..dd14ed4fea6c 100644
--- a/fs/btrfs/mkfs.c
+++ b/fs/btrfs/mkfs.c
@@ -10,6 +10,120 @@
 #include "ctree.h"
 #include "disk-io.h"
 
+int mkfs(int fd, u64 num_blocks, u16 blocksize)
+{
+	struct btrfs_super_block super;
+	struct btrfs_leaf empty_leaf;
+	struct btrfs_root_item root_item;
+	struct btrfs_item item;
+	struct btrfs_extent_item extent_item;
+	char *block;
+	int ret;
+	u16 itemoff;
+
+	btrfs_set_super_blocknr(&super, 16);
+	btrfs_set_super_root(&super, 17);
+	strcpy((char *)(&super.magic), BTRFS_MAGIC);
+	btrfs_set_super_blocksize(&super, blocksize);
+	btrfs_set_super_total_blocks(&super, num_blocks);
+	btrfs_set_super_blocks_used(&super, 0);
+
+	block = malloc(blocksize);
+	memset(block, 0, blocksize);
+	BUG_ON(sizeof(super) > blocksize);
+	memcpy(block, &super, sizeof(super));
+	ret = pwrite(fd, block, blocksize, BTRFS_SUPER_INFO_OFFSET(blocksize));
+	BUG_ON(ret != blocksize);
+
+	/* create the tree of root objects */
+	memset(&empty_leaf, 0, sizeof(empty_leaf));
+	btrfs_set_header_parentid(&empty_leaf.header, BTRFS_ROOT_TREE_OBJECTID);
+	btrfs_set_header_blocknr(&empty_leaf.header, 17);
+	btrfs_set_header_nritems(&empty_leaf.header, 2);
+
+	/* create the items for the root tree */
+	btrfs_set_root_blocknr(&root_item, 18);
+	btrfs_set_root_refs(&root_item, 1);
+	itemoff = LEAF_DATA_SIZE - sizeof(root_item);
+	btrfs_set_item_offset(&item, itemoff);
+	btrfs_set_item_size(&item, sizeof(root_item));
+	btrfs_set_key_objectid(&item.key, BTRFS_EXTENT_TREE_OBJECTID);
+	btrfs_set_key_offset(&item.key, 0);
+	btrfs_set_key_flags(&item.key, 0);
+	memcpy(empty_leaf.items, &item, sizeof(item));
+	memcpy(empty_leaf.data + itemoff, &root_item, sizeof(root_item));
+
+	btrfs_set_root_blocknr(&root_item, 19);
+	itemoff = itemoff - sizeof(root_item);
+	btrfs_set_item_offset(&item, itemoff);
+	btrfs_set_key_objectid(&item.key, BTRFS_FS_TREE_OBJECTID);
+	memcpy(empty_leaf.items + 1, &item, sizeof(item));
+	memcpy(empty_leaf.data + itemoff, &root_item, sizeof(root_item));
+	ret = pwrite(fd, &empty_leaf, blocksize, 17 * blocksize);
+
+	/* create the items for the extent tree */
+	btrfs_set_header_parentid(&empty_leaf.header,
+				  BTRFS_EXTENT_TREE_OBJECTID);
+	btrfs_set_header_blocknr(&empty_leaf.header, 18);
+	btrfs_set_header_nritems(&empty_leaf.header, 4);
+
+	/* item1, reserve blocks 0-16 */
+	btrfs_set_key_objectid(&item.key, 0);
+	btrfs_set_key_offset(&item.key, 17);
+	btrfs_set_key_flags(&item.key, 0);
+	itemoff = LEAF_DATA_SIZE - sizeof(struct btrfs_extent_item);
+	btrfs_set_item_offset(&item, itemoff);
+	btrfs_set_item_size(&item, sizeof(struct btrfs_extent_item));
+	btrfs_set_extent_refs(&extent_item, 1);
+	btrfs_set_extent_owner(&extent_item, 0);
+	memcpy(empty_leaf.items, &item, sizeof(item));
+	memcpy(empty_leaf.data + btrfs_item_offset(&item), &extent_item,
+		btrfs_item_size(&item));
+
+	/* item2, give block 17 to the root */
+	btrfs_set_key_objectid(&item.key, 17);
+	btrfs_set_key_offset(&item.key, 1);
+	itemoff = itemoff - sizeof(struct btrfs_extent_item);
+	btrfs_set_item_offset(&item, itemoff);
+	btrfs_set_extent_owner(&extent_item, BTRFS_ROOT_TREE_OBJECTID);
+	memcpy(empty_leaf.items + 1, &item, sizeof(item));
+	memcpy(empty_leaf.data + btrfs_item_offset(&item), &extent_item,
+		btrfs_item_size(&item));
+
+	/* item3, give block 18 to the extent root */
+	btrfs_set_key_objectid(&item.key, 18);
+	btrfs_set_key_offset(&item.key, 1);
+	itemoff = itemoff - sizeof(struct btrfs_extent_item);
+	btrfs_set_item_offset(&item, itemoff);
+	btrfs_set_extent_owner(&extent_item, BTRFS_EXTENT_TREE_OBJECTID);
+	memcpy(empty_leaf.items + 2, &item, sizeof(item));
+	memcpy(empty_leaf.data + btrfs_item_offset(&item), &extent_item,
+		btrfs_item_size(&item));
+
+	/* item4, give block 19 to the FS root */
+	btrfs_set_key_objectid(&item.key, 19);
+	btrfs_set_key_offset(&item.key, 1);
+	itemoff = itemoff - sizeof(struct btrfs_extent_item);
+	btrfs_set_item_offset(&item, itemoff);
+	btrfs_set_extent_owner(&extent_item, BTRFS_FS_TREE_OBJECTID);
+	memcpy(empty_leaf.items + 3, &item, sizeof(item));
+	memcpy(empty_leaf.data + btrfs_item_offset(&item), &extent_item,
+		btrfs_item_size(&item));
+	ret = pwrite(fd, &empty_leaf, blocksize, 18 * blocksize);
+	if (ret != sizeof(empty_leaf))
+		return -1;
+
+	/* finally create the FS root */
+	btrfs_set_header_parentid(&empty_leaf.header, BTRFS_FS_TREE_OBJECTID);
+	btrfs_set_header_blocknr(&empty_leaf.header, 19);
+	btrfs_set_header_nritems(&empty_leaf.header, 0);
+	ret = pwrite(fd, &empty_leaf, blocksize, 19 * blocksize);
+	if (ret != sizeof(empty_leaf))
+		return -1;
+	return 0;
+}
+
+#if 0
 int mkfs(int fd)
 {
 	struct btrfs_root_info info[2];
@@ -20,13 +134,14 @@ int mkfs(int fd)
 
 	/* setup the super block area */
 	memset(info, 0, sizeof(info));
-	info[0].blocknr = 16;
-	info[0].objectid = 1;
-	info[0].tree_root = 17;
+	btrfs_set_root_blocknr(info, 16);
+	btrfs_set_root_objectid(info, 1);
+	btrfs_set_root_tree_root(info, 17);
+
+	btrfs_set_root_blocknr(info + 1, 16);
+	btrfs_set_root_objectid(info + 1, 2);
+	btrfs_set_root_tree_root(info + 1, 18);
 
-	info[1].blocknr = 16;
-	info[1].objectid = 2;
-	info[1].tree_root = 18;
 	ret = pwrite(fd, info, sizeof(info),
 		     BTRFS_SUPER_INFO_OFFSET(BTRFS_BLOCKSIZE));
 	if (ret != sizeof(info))
@@ -81,3 +196,4 @@ int mkfs(int fd)
 		return -1;
 	return 0;
 }
+#endif
diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c
index c95c85640aa9..64829b6b90f9 100644
--- a/fs/btrfs/print-tree.c
+++ b/fs/btrfs/print-tree.c
@@ -11,6 +11,7 @@ void btrfs_print_leaf(struct btrfs_leaf *l)
 	u32 nr = btrfs_header_nritems(&l->header);
 	struct btrfs_item *item;
 	struct btrfs_extent_item *ei;
+	struct btrfs_root_item *ri;
 	printf("leaf %Lu total ptrs %d free space %d\n",
 		btrfs_header_blocknr(&l->header), nr, btrfs_leaf_free_space(l));
 	fflush(stdout);
@@ -23,13 +24,15 @@ void btrfs_print_leaf(struct btrfs_leaf *l)
 			btrfs_key_offset(&item->key),
 			btrfs_item_offset(item),
 			btrfs_item_size(item));
-		fflush(stdout);
 		printf("\t\titem data %.*s\n", btrfs_item_size(item),
 			l->data + btrfs_item_offset(item));
 		ei = (struct btrfs_extent_item *)(l->data +
 						  btrfs_item_offset(item));
-		printf("\t\textent data refs %u owner %Lu\n", ei->refs,
-			ei->owner);
+		printf("\t\textent data refs %u owner %Lu\n",
+			btrfs_extent_refs(ei), btrfs_extent_owner(ei));
+		ri = (struct btrfs_root_item *)ei;
+		printf("\t\troot data blocknr %Lu refs %u\n",
+			btrfs_root_blocknr(ri), btrfs_root_refs(ri));
 		fflush(stdout);
 	}
 }
@@ -71,6 +74,5 @@ void btrfs_print_tree(struct btrfs_root *root, struct btrfs_buffer *t)
 		btrfs_print_tree(root, next_buf);
 		btrfs_block_release(root, next_buf);
 	}
-
 }
 
diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c
new file mode 100644
index 000000000000..0ab90cfea98f
--- /dev/null
+++ b/fs/btrfs/root-tree.c
@@ -0,0 +1,88 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include "kerncompat.h"
+#include "radix-tree.h"
+#include "ctree.h"
+#include "disk-io.h"
+#include "print-tree.h"
+
+int btrfs_find_last_root(struct btrfs_root *root, u64 objectid,
+			struct btrfs_root_item *item, struct btrfs_key *key)
+{
+	struct btrfs_path path;
+	struct btrfs_key search_key;
+	struct btrfs_leaf *l;
+	int ret;
+	int slot;
+
+	search_key.objectid = objectid;
+	search_key.flags = (u32)-1;
+	search_key.offset = (u32)-1;
+
+	btrfs_init_path(&path);
+	ret = btrfs_search_slot(root, &search_key, &path, 0, 0);
+	if (ret < 0)
+		goto out;
+	BUG_ON(ret == 0);
+	l = &path.nodes[0]->leaf;
+	BUG_ON(path.slots[0] == 0);
+	slot = path.slots[0] - 1;
+	if (btrfs_key_objectid(&l->items[slot].key) != objectid) {
+		ret = 1;
+		goto out;
+	}
+	memcpy(item, l->data + btrfs_item_offset(l->items + slot),
+		sizeof(*item));
+	btrfs_disk_key_to_cpu(key, &l->items[slot].key);
+	btrfs_release_path(root, &path);
+	ret = 0;
+out:
+	return ret;
+}
+
+int btrfs_update_root(struct btrfs_root *root, struct btrfs_key *key,
+		      struct btrfs_root_item *item)
+{
+	struct btrfs_path path;
+	struct btrfs_leaf *l;
+	int ret;
+	int slot;
+
+	btrfs_init_path(&path);
+	ret = btrfs_search_slot(root, key, &path, 0, 1);
+	if (ret < 0)
+		goto out;
+	BUG_ON(ret != 0);
+	l = &path.nodes[0]->leaf;
+	slot = path.slots[0];
+	memcpy(l->data + btrfs_item_offset(l->items + slot), item,
+		sizeof(*item));
+out:
+	btrfs_release_path(root, &path);
+	return ret;
+}
+
+int btrfs_insert_root(struct btrfs_root *root, struct btrfs_key *key,
+		      struct btrfs_root_item *item)
+{
+	int ret;
+	ret = btrfs_insert_item(root, key, item, sizeof(*item));
+	BUG_ON(ret);
+	return ret;
+}
+
+int btrfs_del_root(struct btrfs_root *root, struct btrfs_key *key)
+{
+	struct btrfs_path path;
+	int ret;
+
+	btrfs_init_path(&path);
+	ret = btrfs_search_slot(root, key, &path, -1, 1);
+	if (ret < 0)
+		goto out;
+	BUG_ON(ret != 0);
+	ret = btrfs_del_item(root, &path);
+out:
+	btrfs_release_path(root, &path);
+	return ret;
+}
-- 
cgit v1.2.3


From 123abc88c9087b9c5605566ee3491aaef17fd837 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Wed, 14 Mar 2007 14:14:43 -0400
Subject: Btrfs: variable block size support

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.c       | 225 ++++++++++++++++++++++++-------------------------
 fs/btrfs/ctree.h       |  62 ++++++++------
 fs/btrfs/disk-io.c     |  42 ++++-----
 fs/btrfs/disk-io.h     |   9 +-
 fs/btrfs/extent-tree.c |   6 +-
 fs/btrfs/mkfs.c        | 171 +++++++++++--------------------------
 fs/btrfs/print-tree.c  |  15 ++--
 fs/btrfs/print-tree.h  |   2 +-
 fs/btrfs/quick-test.c  |   2 +-
 fs/btrfs/root-tree.c   |   4 +-
 10 files changed, 238 insertions(+), 300 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index dc620186e501..ece8de7f38ef 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -47,7 +47,7 @@ static int btrfs_cow_block(struct btrfs_root *root,
 		return 0;
 	}
 	cow = btrfs_alloc_free_block(root);
-	memcpy(&cow->node, &buf->node, sizeof(buf->node));
+	memcpy(&cow->node, &buf->node, root->blocksize);
 	btrfs_set_header_blocknr(&cow->node.header, cow->blocknr);
 	*cow_ret = cow;
 	btrfs_inc_ref(root, buf);
@@ -72,11 +72,12 @@ static int btrfs_cow_block(struct btrfs_root *root,
  * this returns the address of the start of the last item,
  * which is the stop of the leaf data stack
  */
-static inline unsigned int leaf_data_end(struct btrfs_leaf *leaf)
+static inline unsigned int leaf_data_end(struct btrfs_root *root,
+					 struct btrfs_leaf *leaf)
 {
 	u32 nr = btrfs_header_nritems(&leaf->header);
 	if (nr == 0)
-		return sizeof(leaf->data);
+		return BTRFS_LEAF_DATA_SIZE(root);
 	return btrfs_item_offset(leaf->items + nr - 1);
 }
 
@@ -85,12 +86,12 @@ static inline unsigned int leaf_data_end(struct btrfs_leaf *leaf)
  * the start of the leaf data.  IOW, how much room
  * the leaf has left for both items and data
  */
-int btrfs_leaf_free_space(struct btrfs_leaf *leaf)
+int btrfs_leaf_free_space(struct btrfs_root *root, struct btrfs_leaf *leaf)
 {
-	int data_end = leaf_data_end(leaf);
+	int data_end = leaf_data_end(root, leaf);
 	int nritems = btrfs_header_nritems(&leaf->header);
 	char *items_end = (char *)(leaf->items + nritems + 1);
-	return (char *)(leaf->data + data_end) - (char *)items_end;
+	return (char *)(btrfs_leaf_data(leaf) + data_end) - (char *)items_end;
 }
 
 /*
@@ -117,7 +118,8 @@ static int comp_keys(struct btrfs_disk_key *disk, struct btrfs_key *k2)
 	return 0;
 }
 
-static int check_node(struct btrfs_path *path, int level)
+static int check_node(struct btrfs_root *root, struct btrfs_path *path,
+		      int level)
 {
 	int i;
 	struct btrfs_node *parent = NULL;
@@ -131,22 +133,23 @@ static int check_node(struct btrfs_path *path, int level)
 	BUG_ON(nritems == 0);
 	if (parent) {
 		struct btrfs_disk_key *parent_key;
-		parent_key = &parent->keys[parent_slot];
-		BUG_ON(memcmp(parent_key, node->keys,
+		parent_key = &parent->ptrs[parent_slot].key;
+		BUG_ON(memcmp(parent_key, &node->ptrs[0].key,
 			      sizeof(struct btrfs_disk_key)));
 		BUG_ON(btrfs_node_blockptr(parent, parent_slot) !=
 		       btrfs_header_blocknr(&node->header));
 	}
-	BUG_ON(nritems > NODEPTRS_PER_BLOCK);
+	BUG_ON(nritems > BTRFS_NODEPTRS_PER_BLOCK(root));
 	for (i = 0; nritems > 1 && i < nritems - 2; i++) {
 		struct btrfs_key cpukey;
-		btrfs_disk_key_to_cpu(&cpukey, &node->keys[i + 1]);
-		BUG_ON(comp_keys(&node->keys[i], &cpukey) >= 0);
+		btrfs_disk_key_to_cpu(&cpukey, &node->ptrs[i + 1].key);
+		BUG_ON(comp_keys(&node->ptrs[i].key, &cpukey) >= 0);
 	}
 	return 0;
 }
 
-static int check_leaf(struct btrfs_path *path, int level)
+static int check_leaf(struct btrfs_root *root, struct btrfs_path *path,
+		      int level)
 {
 	int i;
 	struct btrfs_leaf *leaf = &path->nodes[level]->leaf;
@@ -157,14 +160,14 @@ static int check_leaf(struct btrfs_path *path, int level)
 	if (path->nodes[level + 1])
 		parent = &path->nodes[level + 1]->node;
 	parent_slot = path->slots[level + 1];
-	BUG_ON(btrfs_leaf_free_space(leaf) < 0);
+	BUG_ON(btrfs_leaf_free_space(root, leaf) < 0);
 
 	if (nritems == 0)
 		return 0;
 
 	if (parent) {
 		struct btrfs_disk_key *parent_key;
-		parent_key = &parent->keys[parent_slot];
+		parent_key = &parent->ptrs[parent_slot].key;
 		BUG_ON(memcmp(parent_key, &leaf->items[0].key,
 		       sizeof(struct btrfs_disk_key)));
 		BUG_ON(btrfs_node_blockptr(parent, parent_slot) !=
@@ -180,17 +183,18 @@ static int check_leaf(struct btrfs_path *path, int level)
 		if (i == 0) {
 			BUG_ON(btrfs_item_offset(leaf->items + i) +
 			       btrfs_item_size(leaf->items + i) !=
-			       LEAF_DATA_SIZE);
+			       BTRFS_LEAF_DATA_SIZE(root));
 		}
 	}
 	return 0;
 }
 
-static int check_block(struct btrfs_path *path, int level)
+static int check_block(struct btrfs_root *root, struct btrfs_path *path,
+			int level)
 {
 	if (level == 0)
-		return check_leaf(path, level);
-	return check_node(path, level);
+		return check_leaf(root, path, level);
+	return check_node(root, path, level);
 }
 
 /*
@@ -242,8 +246,8 @@ static int bin_search(struct btrfs_node *c, struct btrfs_key *key, int *slot)
 					  key, btrfs_header_nritems(&c->header),
 					  slot);
 	} else {
-		return generic_bin_search((void *)c->keys,
-					  sizeof(struct btrfs_disk_key),
+		return generic_bin_search((void *)c->ptrs,
+					  sizeof(struct btrfs_key_ptr),
 					  key, btrfs_header_nritems(&c->header),
 					  slot);
 	}
@@ -311,7 +315,8 @@ static int balance_level(struct btrfs_root *root, struct btrfs_path *path,
 	}
 	parent = &parent_buf->node;
 
-	if (btrfs_header_nritems(&mid->header) > NODEPTRS_PER_BLOCK / 4)
+	if (btrfs_header_nritems(&mid->header) >
+	    BTRFS_NODEPTRS_PER_BLOCK(root) / 4)
 		return 0;
 
 	left_buf = read_node_slot(root, parent_buf, pslot - 1);
@@ -351,7 +356,8 @@ static int balance_level(struct btrfs_root *root, struct btrfs_path *path,
 			if (wret)
 				ret = wret;
 		} else {
-			memcpy(parent->keys + pslot + 1, right->keys,
+			memcpy(&parent->ptrs[pslot + 1].key,
+				&right->ptrs[0].key,
 				sizeof(struct btrfs_disk_key));
 			BUG_ON(list_empty(&parent_buf->dirty));
 		}
@@ -387,7 +393,7 @@ static int balance_level(struct btrfs_root *root, struct btrfs_path *path,
 			ret = wret;
 	} else {
 		/* update the parent key to reflect our changes */
-		memcpy(parent->keys + pslot, mid->keys,
+		memcpy(&parent->ptrs[pslot].key, &mid->ptrs[0].key,
 		       sizeof(struct btrfs_disk_key));
 		BUG_ON(list_empty(&parent_buf->dirty));
 	}
@@ -407,7 +413,7 @@ static int balance_level(struct btrfs_root *root, struct btrfs_path *path,
 		}
 	}
 	/* double check we haven't messed things up */
-	check_block(path, level);
+	check_block(root, path, level);
 	if (orig_ptr != btrfs_node_blockptr(&path->nodes[level]->node,
 					    path->slots[level]))
 		BUG();
@@ -456,7 +462,7 @@ again:
 		BUG_ON(!cow && ins_len);
 		c = &b->node;
 		p->nodes[level] = b;
-		ret = check_block(p, level);
+		ret = check_block(root, p, level);
 		if (ret)
 			return -1;
 		ret = bin_search(c, key, &slot);
@@ -465,7 +471,7 @@ again:
 				slot -= 1;
 			p->slots[level] = slot;
 			if (ins_len > 0 && btrfs_header_nritems(&c->header) ==
-			    NODEPTRS_PER_BLOCK) {
+			    BTRFS_NODEPTRS_PER_BLOCK(root)) {
 				int sret = split_node(root, p, level);
 				BUG_ON(sret > 0);
 				if (sret)
@@ -488,7 +494,7 @@ again:
 		} else {
 			struct btrfs_leaf *l = (struct btrfs_leaf *)c;
 			p->slots[level] = slot;
-			if (ins_len > 0 && btrfs_leaf_free_space(l) <
+			if (ins_len > 0 && btrfs_leaf_free_space(root, l) <
 			    sizeof(struct btrfs_item) + ins_len) {
 				int sret = split_leaf(root, p, ins_len);
 				BUG_ON(sret > 0);
@@ -525,7 +531,7 @@ static int fixup_low_keys(struct btrfs_root *root,
 		if (!path->nodes[i])
 			break;
 		t = &path->nodes[i]->node;
-		memcpy(t->keys + tslot, key, sizeof(*key));
+		memcpy(&t->ptrs[tslot].key, key, sizeof(*key));
 		BUG_ON(list_empty(&path->nodes[i]->dirty));
 		if (tslot != 0)
 			break;
@@ -552,7 +558,7 @@ static int push_node_left(struct btrfs_root *root, struct btrfs_buffer *dst_buf,
 
 	src_nritems = btrfs_header_nritems(&src->header);
 	dst_nritems = btrfs_header_nritems(&dst->header);
-	push_items = NODEPTRS_PER_BLOCK - dst_nritems;
+	push_items = BTRFS_NODEPTRS_PER_BLOCK(root) - dst_nritems;
 	if (push_items <= 0) {
 		return 1;
 	}
@@ -560,16 +566,12 @@ static int push_node_left(struct btrfs_root *root, struct btrfs_buffer *dst_buf,
 	if (src_nritems < push_items)
 		push_items = src_nritems;
 
-	memcpy(dst->keys + dst_nritems, src->keys,
-		push_items * sizeof(struct btrfs_disk_key));
-	memcpy(dst->blockptrs + dst_nritems, src->blockptrs,
-		push_items * sizeof(u64));
+	memcpy(dst->ptrs + dst_nritems, src->ptrs,
+		push_items * sizeof(struct btrfs_key_ptr));
 	if (push_items < src_nritems) {
-		memmove(src->keys, src->keys + push_items,
+		memmove(src->ptrs, src->ptrs + push_items,
 			(src_nritems - push_items) *
-			sizeof(struct btrfs_disk_key));
-		memmove(src->blockptrs, src->blockptrs + push_items,
-			(src_nritems - push_items) * sizeof(u64));
+			sizeof(struct btrfs_key_ptr));
 	}
 	btrfs_set_header_nritems(&src->header, src_nritems - push_items);
 	btrfs_set_header_nritems(&dst->header, dst_nritems + push_items);
@@ -601,7 +603,7 @@ static int balance_node_right(struct btrfs_root *root,
 
 	src_nritems = btrfs_header_nritems(&src->header);
 	dst_nritems = btrfs_header_nritems(&dst->header);
-	push_items = NODEPTRS_PER_BLOCK - dst_nritems;
+	push_items = BTRFS_NODEPTRS_PER_BLOCK(root) - dst_nritems;
 	if (push_items <= 0) {
 		return 1;
 	}
@@ -613,14 +615,10 @@ static int balance_node_right(struct btrfs_root *root,
 	if (max_push < push_items)
 		push_items = max_push;
 
-	memmove(dst->keys + push_items, dst->keys,
-		dst_nritems * sizeof(struct btrfs_disk_key));
-	memmove(dst->blockptrs + push_items, dst->blockptrs,
-		dst_nritems * sizeof(u64));
-	memcpy(dst->keys, src->keys + src_nritems - push_items,
-		push_items * sizeof(struct btrfs_disk_key));
-	memcpy(dst->blockptrs, src->blockptrs + src_nritems - push_items,
-		push_items * sizeof(u64));
+	memmove(dst->ptrs + push_items, dst->ptrs,
+		dst_nritems * sizeof(struct btrfs_key_ptr));
+	memcpy(dst->ptrs, src->ptrs + src_nritems - push_items,
+		push_items * sizeof(struct btrfs_key_ptr));
 
 	btrfs_set_header_nritems(&src->header, src_nritems - push_items);
 	btrfs_set_header_nritems(&dst->header, dst_nritems + push_items);
@@ -650,7 +648,7 @@ static int insert_new_root(struct btrfs_root *root,
 
 	t = btrfs_alloc_free_block(root);
 	c = &t->node;
-	memset(c, 0, sizeof(c));
+	memset(c, 0, root->blocksize);
 	btrfs_set_header_nritems(&c->header, 1);
 	btrfs_set_header_level(&c->header, level);
 	btrfs_set_header_blocknr(&c->header, t->blocknr);
@@ -660,8 +658,8 @@ static int insert_new_root(struct btrfs_root *root,
 	if (btrfs_is_leaf(lower))
 		lower_key = &((struct btrfs_leaf *)lower)->items[0].key;
 	else
-		lower_key = lower->keys;
-	memcpy(c->keys, lower_key, sizeof(struct btrfs_disk_key));
+		lower_key = &lower->ptrs[0].key;
+	memcpy(&c->ptrs[0].key, lower_key, sizeof(struct btrfs_disk_key));
 	btrfs_set_node_blockptr(c, 0, path->nodes[level - 1]->blocknr);
 	/* the super has an extra ref to root->node */
 	btrfs_block_release(root, root->node);
@@ -693,19 +691,15 @@ static int insert_ptr(struct btrfs_root *root,
 	nritems = btrfs_header_nritems(&lower->header);
 	if (slot > nritems)
 		BUG();
-	if (nritems == NODEPTRS_PER_BLOCK)
+	if (nritems == BTRFS_NODEPTRS_PER_BLOCK(root))
 		BUG();
 	if (slot != nritems) {
-		memmove(lower->keys + slot + 1, lower->keys + slot,
-			(nritems - slot) * sizeof(struct btrfs_disk_key));
-		memmove(lower->blockptrs + slot + 1, lower->blockptrs + slot,
-			(nritems - slot) * sizeof(u64));
+		memmove(lower->ptrs + slot + 1, lower->ptrs + slot,
+			(nritems - slot) * sizeof(struct btrfs_key_ptr));
 	}
-	memcpy(lower->keys + slot, key, sizeof(struct btrfs_disk_key));
+	memcpy(&lower->ptrs[slot].key, key, sizeof(struct btrfs_disk_key));
 	btrfs_set_node_blockptr(lower, slot, blocknr);
 	btrfs_set_header_nritems(&lower->header, nritems + 1);
-	if (lower->keys[1].objectid == 0)
-			BUG();
 	BUG_ON(list_empty(&path->nodes[level]->dirty));
 	return 0;
 }
@@ -747,17 +741,16 @@ static int split_node(struct btrfs_root *root, struct btrfs_path *path,
 	btrfs_set_header_parentid(&split->header,
 	                       btrfs_header_parentid(&root->node->node.header));
 	mid = (c_nritems + 1) / 2;
-	memcpy(split->keys, c->keys + mid,
-		(c_nritems - mid) * sizeof(struct btrfs_disk_key));
-	memcpy(split->blockptrs, c->blockptrs + mid,
-		(c_nritems - mid) * sizeof(u64));
+	memcpy(split->ptrs, c->ptrs + mid,
+		(c_nritems - mid) * sizeof(struct btrfs_key_ptr));
 	btrfs_set_header_nritems(&split->header, c_nritems - mid);
 	btrfs_set_header_nritems(&c->header, mid);
 	ret = 0;
 
 	BUG_ON(list_empty(&t->dirty));
-	wret = insert_ptr(root, path, split->keys, split_buffer->blocknr,
-			  path->slots[level + 1] + 1, level + 1);
+	wret = insert_ptr(root, path, &split->ptrs[0].key,
+			  split_buffer->blocknr, path->slots[level + 1] + 1,
+			  level + 1);
 	if (wret)
 		ret = wret;
 
@@ -825,7 +818,7 @@ static int push_leaf_right(struct btrfs_root *root, struct btrfs_path *path,
 	right_buf = read_tree_block(root, btrfs_node_blockptr(&upper->node,
 							      slot + 1));
 	right = &right_buf->leaf;
-	free_space = btrfs_leaf_free_space(right);
+	free_space = btrfs_leaf_free_space(root, right);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
 		btrfs_block_release(root, right_buf);
 		return 1;
@@ -833,7 +826,7 @@ static int push_leaf_right(struct btrfs_root *root, struct btrfs_path *path,
 	/* cow and double check */
 	btrfs_cow_block(root, right_buf, upper, slot + 1, &right_buf);
 	right = &right_buf->leaf;
-	free_space = btrfs_leaf_free_space(right);
+	free_space = btrfs_leaf_free_space(root, right);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
 		btrfs_block_release(root, right_buf);
 		return 1;
@@ -857,15 +850,14 @@ static int push_leaf_right(struct btrfs_root *root, struct btrfs_path *path,
 	right_nritems = btrfs_header_nritems(&right->header);
 	/* push left to right */
 	push_space = btrfs_item_end(left->items + left_nritems - push_items);
-	push_space -= leaf_data_end(left);
+	push_space -= leaf_data_end(root, left);
 	/* make room in the right data area */
-	memmove(right->data + leaf_data_end(right) - push_space,
-		right->data + leaf_data_end(right),
-		LEAF_DATA_SIZE - leaf_data_end(right));
+	memmove(btrfs_leaf_data(right) + leaf_data_end(root, right) -
+		push_space, btrfs_leaf_data(right) + leaf_data_end(root, right),
+		BTRFS_LEAF_DATA_SIZE(root) - leaf_data_end(root, right));
 	/* copy from the left data area */
-	memcpy(right->data + LEAF_DATA_SIZE - push_space,
-		left->data + leaf_data_end(left),
-		push_space);
+	memcpy(btrfs_leaf_data(right) + BTRFS_LEAF_DATA_SIZE(root) - push_space,
+		btrfs_leaf_data(left) + leaf_data_end(root, left), push_space);
 	memmove(right->items + push_items, right->items,
 		right_nritems * sizeof(struct btrfs_item));
 	/* copy the items from left to right */
@@ -875,7 +867,7 @@ static int push_leaf_right(struct btrfs_root *root, struct btrfs_path *path,
 	/* update the item pointers */
 	right_nritems += push_items;
 	btrfs_set_header_nritems(&right->header, right_nritems);
-	push_space = LEAF_DATA_SIZE;
+	push_space = BTRFS_LEAF_DATA_SIZE(root);
 	for (i = 0; i < right_nritems; i++) {
 		btrfs_set_item_offset(right->items + i, push_space -
 				      btrfs_item_size(right->items + i));
@@ -886,7 +878,7 @@ static int push_leaf_right(struct btrfs_root *root, struct btrfs_path *path,
 
 	BUG_ON(list_empty(&left_buf->dirty));
 	BUG_ON(list_empty(&right_buf->dirty));
-	memcpy(upper->node.keys + slot + 1,
+	memcpy(&upper->node.ptrs[slot + 1].key,
 		&right->items[0].key, sizeof(struct btrfs_disk_key));
 	BUG_ON(list_empty(&upper->dirty));
 
@@ -932,7 +924,7 @@ static int push_leaf_left(struct btrfs_root *root, struct btrfs_path *path,
 	t = read_tree_block(root, btrfs_node_blockptr(&path->nodes[1]->node,
 						      slot - 1));
 	left = &t->leaf;
-	free_space = btrfs_leaf_free_space(left);
+	free_space = btrfs_leaf_free_space(root, left);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
 		btrfs_block_release(root, t);
 		return 1;
@@ -941,7 +933,7 @@ static int push_leaf_left(struct btrfs_root *root, struct btrfs_path *path,
 	/* cow and double check */
 	btrfs_cow_block(root, t, path->nodes[1], slot - 1, &t);
 	left = &t->leaf;
-	free_space = btrfs_leaf_free_space(left);
+	free_space = btrfs_leaf_free_space(root, left);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
 		btrfs_block_release(root, t);
 		return 1;
@@ -964,17 +956,19 @@ static int push_leaf_left(struct btrfs_root *root, struct btrfs_path *path,
 	/* push data from right to left */
 	memcpy(left->items + btrfs_header_nritems(&left->header),
 		right->items, push_items * sizeof(struct btrfs_item));
-	push_space = LEAF_DATA_SIZE -
+	push_space = BTRFS_LEAF_DATA_SIZE(root) -
 		     btrfs_item_offset(right->items + push_items -1);
-	memcpy(left->data + leaf_data_end(left) - push_space,
-		right->data + btrfs_item_offset(right->items + push_items - 1),
+	memcpy(btrfs_leaf_data(left) + leaf_data_end(root, left) - push_space,
+		btrfs_leaf_data(right) +
+		btrfs_item_offset(right->items + push_items - 1),
 		push_space);
 	old_left_nritems = btrfs_header_nritems(&left->header);
 	BUG_ON(old_left_nritems < 0);
 
 	for (i = old_left_nritems; i < old_left_nritems + push_items; i++) {
-		u16 ioff = btrfs_item_offset(left->items + i);
-		btrfs_set_item_offset(left->items + i, ioff - (LEAF_DATA_SIZE -
+		u32 ioff = btrfs_item_offset(left->items + i);
+		btrfs_set_item_offset(left->items + i, ioff -
+				     (BTRFS_LEAF_DATA_SIZE(root) -
 				      btrfs_item_offset(left->items +
 						        old_left_nritems - 1)));
 	}
@@ -982,16 +976,17 @@ static int push_leaf_left(struct btrfs_root *root, struct btrfs_path *path,
 
 	/* fixup right node */
 	push_space = btrfs_item_offset(right->items + push_items - 1) -
-		     leaf_data_end(right);
-	memmove(right->data + LEAF_DATA_SIZE - push_space, right->data +
-		leaf_data_end(right), push_space);
+		     leaf_data_end(root, right);
+	memmove(btrfs_leaf_data(right) + BTRFS_LEAF_DATA_SIZE(root) -
+		push_space, btrfs_leaf_data(right) +
+		leaf_data_end(root, right), push_space);
 	memmove(right->items, right->items + push_items,
 		(btrfs_header_nritems(&right->header) - push_items) *
 		sizeof(struct btrfs_item));
 	btrfs_set_header_nritems(&right->header,
 				 btrfs_header_nritems(&right->header) -
 				 push_items);
-	push_space = LEAF_DATA_SIZE;
+	push_space = BTRFS_LEAF_DATA_SIZE(root);
 
 	for (i = 0; i < btrfs_header_nritems(&right->header); i++) {
 		btrfs_set_item_offset(right->items + i, push_space -
@@ -1051,12 +1046,12 @@ static int split_leaf(struct btrfs_root *root, struct btrfs_path *path,
 		if (wret < 0)
 			return wret;
 	}
-
 	l_buf = path->nodes[0];
 	l = &l_buf->leaf;
 
 	/* did the pushes work? */
-	if (btrfs_leaf_free_space(l) >= sizeof(struct btrfs_item) + data_size)
+	if (btrfs_leaf_free_space(root, l) >=
+	    sizeof(struct btrfs_item) + data_size)
 		return 0;
 
 	if (!path->nodes[1]) {
@@ -1071,16 +1066,16 @@ static int split_leaf(struct btrfs_root *root, struct btrfs_path *path,
 	BUG_ON(!right_buffer);
 	BUG_ON(mid == nritems);
 	right = &right_buffer->leaf;
-	memset(right, 0, sizeof(*right));
+	memset(&right->header, 0, sizeof(right->header));
 	if (mid <= slot) {
 		/* FIXME, just alloc a new leaf here */
 		if (leaf_space_used(l, mid, nritems - mid) + space_needed >
-			LEAF_DATA_SIZE)
+			BTRFS_LEAF_DATA_SIZE(root))
 			BUG();
 	} else {
 		/* FIXME, just alloc a new leaf here */
 		if (leaf_space_used(l, 0, mid + 1) + space_needed >
-			LEAF_DATA_SIZE)
+			BTRFS_LEAF_DATA_SIZE(root))
 			BUG();
 	}
 	btrfs_set_header_nritems(&right->header, nritems - mid);
@@ -1088,15 +1083,18 @@ static int split_leaf(struct btrfs_root *root, struct btrfs_path *path,
 	btrfs_set_header_level(&right->header, 0);
 	btrfs_set_header_parentid(&right->header,
 	                       btrfs_header_parentid(&root->node->node.header));
-	data_copy_size = btrfs_item_end(l->items + mid) - leaf_data_end(l);
+	data_copy_size = btrfs_item_end(l->items + mid) -
+			 leaf_data_end(root, l);
 	memcpy(right->items, l->items + mid,
 	       (nritems - mid) * sizeof(struct btrfs_item));
-	memcpy(right->data + LEAF_DATA_SIZE - data_copy_size,
-	       l->data + leaf_data_end(l), data_copy_size);
-	rt_data_off = LEAF_DATA_SIZE - btrfs_item_end(l->items + mid);
+	memcpy(btrfs_leaf_data(right) + BTRFS_LEAF_DATA_SIZE(root) -
+		data_copy_size, btrfs_leaf_data(l) +
+		leaf_data_end(root, l), data_copy_size);
+	rt_data_off = BTRFS_LEAF_DATA_SIZE(root) -
+		      btrfs_item_end(l->items + mid);
 
 	for (i = 0; i < btrfs_header_nritems(&right->header); i++) {
-		u16 ioff = btrfs_item_offset(right->items + i);
+		u32 ioff = btrfs_item_offset(right->items + i);
 		btrfs_set_item_offset(right->items + i, ioff + rt_data_off);
 	}
 
@@ -1156,9 +1154,9 @@ int btrfs_insert_item(struct btrfs_root *root, struct btrfs_key *cpu_key,
 	leaf = &leaf_buf->leaf;
 
 	nritems = btrfs_header_nritems(&leaf->header);
-	data_end = leaf_data_end(leaf);
+	data_end = leaf_data_end(root, leaf);
 
-	if (btrfs_leaf_free_space(leaf) <
+	if (btrfs_leaf_free_space(root, leaf) <
 	    sizeof(struct btrfs_item) + data_size)
 		BUG();
 
@@ -1173,7 +1171,7 @@ int btrfs_insert_item(struct btrfs_root *root, struct btrfs_key *cpu_key,
 		 */
 		/* first correct the data pointers */
 		for (i = slot; i < nritems; i++) {
-			u16 ioff = btrfs_item_offset(leaf->items + i);
+			u32 ioff = btrfs_item_offset(leaf->items + i);
 			btrfs_set_item_offset(leaf->items + i,
 					      ioff - data_size);
 		}
@@ -1183,7 +1181,8 @@ int btrfs_insert_item(struct btrfs_root *root, struct btrfs_key *cpu_key,
 		        (nritems - slot) * sizeof(struct btrfs_item));
 
 		/* shift the data */
-		memmove(leaf->data + data_end - data_size, leaf->data +
+		memmove(btrfs_leaf_data(leaf) + data_end - data_size,
+			btrfs_leaf_data(leaf) +
 		        data_end, old_data - data_end);
 		data_end = old_data;
 	}
@@ -1192,7 +1191,7 @@ int btrfs_insert_item(struct btrfs_root *root, struct btrfs_key *cpu_key,
 		sizeof(struct btrfs_disk_key));
 	btrfs_set_item_offset(leaf->items + slot, data_end - data_size);
 	btrfs_set_item_size(leaf->items + slot, data_size);
-	memcpy(leaf->data + data_end - data_size, data, data_size);
+	memcpy(btrfs_leaf_data(leaf) + data_end - data_size, data, data_size);
 	btrfs_set_header_nritems(&leaf->header, nritems + 1);
 
 	ret = 0;
@@ -1200,9 +1199,9 @@ int btrfs_insert_item(struct btrfs_root *root, struct btrfs_key *cpu_key,
 		ret = fixup_low_keys(root, &path, &disk_key, 1);
 
 	BUG_ON(list_empty(&leaf_buf->dirty));
-	if (btrfs_leaf_free_space(leaf) < 0)
+	if (btrfs_leaf_free_space(root, leaf) < 0)
 		BUG();
-	check_leaf(&path, 0);
+	check_leaf(root, &path, 0);
 out:
 	btrfs_release_path(root, &path);
 	return ret;
@@ -1227,11 +1226,8 @@ static int del_ptr(struct btrfs_root *root, struct btrfs_path *path, int level,
 	node = &parent->node;
 	nritems = btrfs_header_nritems(&node->header);
 	if (slot != nritems -1) {
-		memmove(node->keys + slot, node->keys + slot + 1,
-			sizeof(struct btrfs_disk_key) * (nritems - slot - 1));
-		memmove(node->blockptrs + slot,
-			node->blockptrs + slot + 1,
-			sizeof(u64) * (nritems - slot - 1));
+		memmove(node->ptrs + slot, node->ptrs + slot + 1,
+			sizeof(struct btrfs_key_ptr) * (nritems - slot - 1));
 	}
 	nritems--;
 	btrfs_set_header_nritems(&node->header, nritems);
@@ -1240,7 +1236,8 @@ static int del_ptr(struct btrfs_root *root, struct btrfs_path *path, int level,
 		/* just turn the root into a leaf and break */
 		btrfs_set_header_level(&root->node->node.header, 0);
 	} else if (slot == 0) {
-		wret = fixup_low_keys(root, path, node->keys, level + 1);
+		wret = fixup_low_keys(root, path, &node->ptrs[0].key,
+				      level + 1);
 		if (wret)
 			ret = wret;
 	}
@@ -1272,12 +1269,12 @@ int btrfs_del_item(struct btrfs_root *root, struct btrfs_path *path)
 
 	if (slot != nritems - 1) {
 		int i;
-		int data_end = leaf_data_end(leaf);
-		memmove(leaf->data + data_end + dsize,
-			leaf->data + data_end,
+		int data_end = leaf_data_end(root, leaf);
+		memmove(btrfs_leaf_data(leaf) + data_end + dsize,
+			btrfs_leaf_data(leaf) + data_end,
 			doff - data_end);
 		for (i = slot + 1; i < nritems; i++) {
-			u16 ioff = btrfs_item_offset(leaf->items + i);
+			u32 ioff = btrfs_item_offset(leaf->items + i);
 			btrfs_set_item_offset(leaf->items + i, ioff + dsize);
 		}
 		memmove(leaf->items + slot, leaf->items + slot + 1,
@@ -1311,7 +1308,7 @@ int btrfs_del_item(struct btrfs_root *root, struct btrfs_path *path)
 		BUG_ON(list_empty(&leaf_buf->dirty));
 
 		/* delete the leaf if it is mostly empty */
-		if (used < LEAF_DATA_SIZE / 3) {
+		if (used < BTRFS_LEAF_DATA_SIZE(root) / 3) {
 			/* push_leaf_left fixes the path.
 			 * make sure the path still points to our leaf
 			 * for possible call to del_ptr below
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 05c7707263f5..c61ad0f69be9 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -5,7 +5,6 @@
 #include "kerncompat.h"
 
 #define BTRFS_MAGIC "_BtRfS_M"
-#define BTRFS_BLOCKSIZE 1024
 
 #define BTRFS_ROOT_TREE_OBJECTID 1
 #define BTRFS_EXTENT_TREE_OBJECTID 2
@@ -52,8 +51,11 @@ struct btrfs_header {
 } __attribute__ ((__packed__));
 
 #define BTRFS_MAX_LEVEL 8
-#define NODEPTRS_PER_BLOCK ((BTRFS_BLOCKSIZE - sizeof(struct btrfs_header)) / \
-			    (sizeof(struct btrfs_disk_key) + sizeof(u64)))
+#define BTRFS_NODEPTRS_PER_BLOCK(r) (((r)->blocksize - \
+			        sizeof(struct btrfs_header)) / \
+			       (sizeof(struct btrfs_disk_key) + sizeof(u64)))
+#define __BTRFS_LEAF_DATA_SIZE(bs) ((bs) - sizeof(struct btrfs_header))
+#define BTRFS_LEAF_DATA_SIZE(r) (__BTRFS_LEAF_DATA_SIZE(r->blocksize))
 
 struct btrfs_buffer;
 
@@ -86,6 +88,7 @@ struct btrfs_root {
 	int ref_cows;
 	struct btrfs_root_item root_item;
 	struct btrfs_key root_key;
+	u32 blocksize;
 };
 
 /*
@@ -97,7 +100,7 @@ struct btrfs_super_block {
 	__le64 blocknr; /* this block number */
 	__le32 csum;
 	__le64 magic;
-	__le16 blocksize;
+	__le32 blocksize;
 	__le64 generation;
 	__le64 root;
 	__le64 total_blocks;
@@ -111,7 +114,7 @@ struct btrfs_super_block {
  */
 struct btrfs_item {
 	struct btrfs_disk_key key;
-	__le16 offset;
+	__le32 offset;
 	__le16 size;
 } __attribute__ ((__packed__));
 
@@ -122,24 +125,23 @@ struct btrfs_item {
  * The data is separate from the items to get the keys closer together
  * during searches.
  */
-#define LEAF_DATA_SIZE (BTRFS_BLOCKSIZE - sizeof(struct btrfs_header))
 struct btrfs_leaf {
 	struct btrfs_header header;
-	union {
-		struct btrfs_item items[LEAF_DATA_SIZE/
-				        sizeof(struct btrfs_item)];
-		u8 data[BTRFS_BLOCKSIZE - sizeof(struct btrfs_header)];
-	};
+	struct btrfs_item items[];
 } __attribute__ ((__packed__));
 
 /*
  * all non-leaf blocks are nodes, they hold only keys and pointers to
  * other blocks
  */
+struct btrfs_key_ptr {
+	struct btrfs_disk_key key;
+	__le64 blockptr;
+} __attribute__ ((__packed__));
+
 struct btrfs_node {
 	struct btrfs_header header;
-	struct btrfs_disk_key keys[NODEPTRS_PER_BLOCK];
-	__le64 blockptrs[NODEPTRS_PER_BLOCK];
+	struct btrfs_key_ptr ptrs[];
 } __attribute__ ((__packed__));
 
 /*
@@ -186,28 +188,28 @@ static inline void btrfs_set_extent_refs(struct btrfs_extent_item *ei, u32 val)
 
 static inline u64 btrfs_node_blockptr(struct btrfs_node *n, int nr)
 {
-	return le64_to_cpu(n->blockptrs[nr]);
+	return le64_to_cpu(n->ptrs[nr].blockptr);
 }
 
 static inline void btrfs_set_node_blockptr(struct btrfs_node *n, int nr,
 					   u64 val)
 {
-	n->blockptrs[nr] = cpu_to_le64(val);
+	n->ptrs[nr].blockptr = cpu_to_le64(val);
 }
 
-static inline u16 btrfs_item_offset(struct btrfs_item *item)
+static inline u32 btrfs_item_offset(struct btrfs_item *item)
 {
-	return le16_to_cpu(item->offset);
+	return le32_to_cpu(item->offset);
 }
 
-static inline void btrfs_set_item_offset(struct btrfs_item *item, u16 val)
+static inline void btrfs_set_item_offset(struct btrfs_item *item, u32 val)
 {
-	item->offset = cpu_to_le16(val);
+	item->offset = cpu_to_le32(val);
 }
 
-static inline u16 btrfs_item_end(struct btrfs_item *item)
+static inline u32 btrfs_item_end(struct btrfs_item *item)
 {
-	return le16_to_cpu(item->offset) + le16_to_cpu(item->size);
+	return le32_to_cpu(item->offset) + le16_to_cpu(item->size);
 }
 
 static inline u16 btrfs_item_size(struct btrfs_item *item)
@@ -390,20 +392,26 @@ static inline void btrfs_set_super_blocks_used(struct btrfs_super_block *s,
 	s->blocks_used = cpu_to_le64(val);
 }
 
-static inline u16 btrfs_super_blocksize(struct btrfs_super_block *s)
+static inline u32 btrfs_super_blocksize(struct btrfs_super_block *s)
 {
-	return le16_to_cpu(s->blocksize);
+	return le32_to_cpu(s->blocksize);
 }
 
 static inline void btrfs_set_super_blocksize(struct btrfs_super_block *s,
-						u16 val)
+						u32 val)
+{
+	s->blocksize = cpu_to_le32(val);
+}
+
+static inline u8 *btrfs_leaf_data(struct btrfs_leaf *l)
 {
-	s->blocksize = cpu_to_le16(val);
+	return (u8 *)l->items;
 }
 
 /* helper function to cast into the data area of the leaf. */
 #define btrfs_item_ptr(leaf, slot, type) \
-	((type *)((leaf)->data + btrfs_item_offset((leaf)->items + (slot))))
+	((type *)(btrfs_leaf_data(leaf) + \
+	btrfs_item_offset((leaf)->items + (slot))))
 
 struct btrfs_buffer *btrfs_alloc_free_block(struct btrfs_root *root);
 int btrfs_inc_ref(struct btrfs_root *root, struct btrfs_buffer *buf);
@@ -416,7 +424,7 @@ int btrfs_del_item(struct btrfs_root *root, struct btrfs_path *path);
 int btrfs_insert_item(struct btrfs_root *root, struct btrfs_key *key,
 		void *data, int data_size);
 int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path);
-int btrfs_leaf_free_space(struct btrfs_leaf *leaf);
+int btrfs_leaf_free_space(struct btrfs_root *root, struct btrfs_leaf *leaf);
 int btrfs_drop_snapshot(struct btrfs_root *root, struct btrfs_buffer *snap);
 int btrfs_finish_extent_commit(struct btrfs_root *root);
 int btrfs_del_root(struct btrfs_root *root, struct btrfs_key *key);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 3d4bf6833f2a..8d9457b5aef5 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -46,7 +46,8 @@ struct btrfs_buffer *alloc_tree_block(struct btrfs_root *root, u64 blocknr)
 {
 	struct btrfs_buffer *buf;
 	int ret;
-	buf = malloc(sizeof(struct btrfs_buffer));
+
+	buf = malloc(sizeof(struct btrfs_buffer) + root->blocksize);
 	if (!buf)
 		return buf;
 	allocated_blocks++;
@@ -84,7 +85,7 @@ struct btrfs_buffer *find_tree_block(struct btrfs_root *root, u64 blocknr)
 
 struct btrfs_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr)
 {
-	loff_t offset = blocknr * BTRFS_BLOCKSIZE;
+	loff_t offset = blocknr * root->blocksize;
 	struct btrfs_buffer *buf;
 	int ret;
 
@@ -95,8 +96,8 @@ struct btrfs_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr)
 		buf = alloc_tree_block(root, blocknr);
 		if (!buf)
 			return NULL;
-		ret = pread(root->fp, &buf->node, BTRFS_BLOCKSIZE, offset);
-		if (ret != BTRFS_BLOCKSIZE) {
+		ret = pread(root->fp, &buf->node, root->blocksize, offset);
+		if (ret != root->blocksize) {
 			free(buf);
 			return NULL;
 		}
@@ -127,13 +128,13 @@ int clean_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf)
 int write_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf)
 {
 	u64 blocknr = buf->blocknr;
-	loff_t offset = blocknr * BTRFS_BLOCKSIZE;
+	loff_t offset = blocknr * root->blocksize;
 	int ret;
 
 	if (buf->blocknr != btrfs_header_blocknr(&buf->node.header))
 		BUG();
-	ret = pwrite(root->fp, &buf->node, BTRFS_BLOCKSIZE, offset);
-	if (ret != BTRFS_BLOCKSIZE)
+	ret = pwrite(root->fp, &buf->node, root->blocksize, offset);
+	if (ret != root->blocksize)
 		return ret;
 	return 0;
 }
@@ -215,7 +216,8 @@ int btrfs_commit_transaction(struct btrfs_root *root,
 	return ret;
 }
 
-static int __setup_root(struct btrfs_root *root, u64 objectid, int fp)
+static int __setup_root(struct btrfs_super_block *super,
+			struct btrfs_root *root, u64 objectid, int fp)
 {
 	INIT_LIST_HEAD(&root->trans);
 	INIT_LIST_HEAD(&root->cache);
@@ -223,6 +225,8 @@ static int __setup_root(struct btrfs_root *root, u64 objectid, int fp)
 	root->fp = fp;
 	root->node = NULL;
 	root->commit_root = NULL;
+	root->blocksize = btrfs_super_blocksize(super);
+	root->ref_cows = 0;
 	memset(&root->current_insert, 0, sizeof(root->current_insert));
 	memset(&root->last_insert, 0, sizeof(root->last_insert));
 	memset(&root->root_key, 0, sizeof(root->root_key));
@@ -230,19 +234,19 @@ static int __setup_root(struct btrfs_root *root, u64 objectid, int fp)
 	return 0;
 }
 
-static int find_and_setup_root(struct btrfs_root *tree_root, u64 objectid,
-			struct btrfs_root *root, int fp)
+static int find_and_setup_root(struct btrfs_super_block *super,
+			       struct btrfs_root *tree_root, u64 objectid,
+			       struct btrfs_root *root, int fp)
 {
 	int ret;
 
-	__setup_root(root, objectid, fp);
+	__setup_root(super, root, objectid, fp);
 	ret = btrfs_find_last_root(tree_root, objectid,
 				   &root->root_item, &root->root_key);
 	BUG_ON(ret);
 
 	root->node = read_tree_block(root,
 				     btrfs_root_blocknr(&root->root_item));
-	root->ref_cows = 0;
 	BUG_ON(!root->node);
 	return 0;
 }
@@ -277,28 +281,28 @@ struct btrfs_root *open_ctree(char *filename, struct btrfs_super_block *super)
 	INIT_RADIX_TREE(&tree_root->cache_radix, GFP_KERNEL);
 
 	ret = pread(fp, super, sizeof(struct btrfs_super_block),
-		     BTRFS_SUPER_INFO_OFFSET(BTRFS_BLOCKSIZE));
+		     BTRFS_SUPER_INFO_OFFSET);
 	if (ret == 0 || btrfs_super_root(super) == 0) {
 		printf("making new FS!\n");
-		ret = mkfs(fp, 0, BTRFS_BLOCKSIZE);
+		ret = mkfs(fp, 0, 1024);
 		if (ret)
 			return NULL;
 		ret = pread(fp, super, sizeof(struct btrfs_super_block),
-			     BTRFS_SUPER_INFO_OFFSET(BTRFS_BLOCKSIZE));
+			     BTRFS_SUPER_INFO_OFFSET);
 		if (ret != sizeof(struct btrfs_super_block))
 			return NULL;
 	}
 	BUG_ON(ret < 0);
 
-	__setup_root(tree_root, BTRFS_ROOT_TREE_OBJECTID, fp);
+	__setup_root(super, tree_root, BTRFS_ROOT_TREE_OBJECTID, fp);
 	tree_root->node = read_tree_block(tree_root, btrfs_super_root(super));
 	BUG_ON(!tree_root->node);
 
-	ret = find_and_setup_root(tree_root, BTRFS_EXTENT_TREE_OBJECTID,
+	ret = find_and_setup_root(super, tree_root, BTRFS_EXTENT_TREE_OBJECTID,
 				  extent_root, fp);
 	BUG_ON(ret);
 
-	ret = find_and_setup_root(tree_root, BTRFS_FS_TREE_OBJECTID,
+	ret = find_and_setup_root(super, tree_root, BTRFS_FS_TREE_OBJECTID,
 				  root, fp);
 	BUG_ON(ret);
 
@@ -313,7 +317,7 @@ int write_ctree_super(struct btrfs_root *root, struct btrfs_super_block *s)
 	int ret;
 	btrfs_set_super_root(s, root->tree_root->node->blocknr);
 	ret = pwrite(root->fp, s, sizeof(*s),
-		     BTRFS_SUPER_INFO_OFFSET(BTRFS_BLOCKSIZE));
+		     BTRFS_SUPER_INFO_OFFSET);
 	if (ret != sizeof(*s)) {
 		fprintf(stderr, "failed to write new super block err %d\n", ret);
 		return ret;
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index c22a61f9233f..5771bb90acb2 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -5,12 +5,12 @@
 struct btrfs_buffer {
 	u64 blocknr;
 	int count;
+	struct list_head dirty;
+	struct list_head cache;
 	union {
 		struct btrfs_node node;
 		struct btrfs_leaf leaf;
 	};
-	struct list_head dirty;
-	struct list_head cache;
 };
 
 struct btrfs_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr);
@@ -24,9 +24,8 @@ struct btrfs_root *open_ctree(char *filename, struct btrfs_super_block *s);
 int close_ctree(struct btrfs_root *root, struct btrfs_super_block *s);
 void btrfs_block_release(struct btrfs_root *root, struct btrfs_buffer *buf);
 int write_ctree_super(struct btrfs_root *root, struct btrfs_super_block *s);
-int mkfs(int fd, u64 num_blocks, u16 blocksize);
-
+int mkfs(int fd, u64 num_blocks, u32 blocksize);
 
-#define BTRFS_SUPER_INFO_OFFSET(bs) (16 * (bs))
+#define BTRFS_SUPER_INFO_OFFSET (16 * 1024)
 
 #endif
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 3767744f659e..d4f1ec328399 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -143,7 +143,6 @@ static int __free_extent(struct btrfs_root *root, u64 blocknr, u64 num_blocks)
 	struct btrfs_key key;
 	struct btrfs_root *extent_root = root->extent_root;
 	int ret;
-	struct btrfs_item *item;
 	struct btrfs_extent_item *ei;
 	struct btrfs_key ins;
 	u32 refs;
@@ -161,9 +160,8 @@ static int __free_extent(struct btrfs_root *root, u64 blocknr, u64 num_blocks)
 		printf("failed to find %Lu\n", key.objectid);
 		BUG();
 	}
-	item = path.nodes[0]->leaf.items + path.slots[0];
-	ei = (struct btrfs_extent_item *)(path.nodes[0]->leaf.data +
-				    btrfs_item_offset(item));
+	ei = btrfs_item_ptr(&path.nodes[0]->leaf, path.slots[0],
+			    struct btrfs_extent_item);
 	BUG_ON(ei->refs == 0);
 	refs = btrfs_extent_refs(ei) - 1;
 	btrfs_set_extent_refs(ei, refs);
diff --git a/fs/btrfs/mkfs.c b/fs/btrfs/mkfs.c
index dd14ed4fea6c..317d20ce759d 100644
--- a/fs/btrfs/mkfs.c
+++ b/fs/btrfs/mkfs.c
@@ -10,19 +10,20 @@
 #include "ctree.h"
 #include "disk-io.h"
 
-int mkfs(int fd, u64 num_blocks, u16 blocksize)
+int mkfs(int fd, u64 num_blocks, u32 blocksize)
 {
 	struct btrfs_super_block super;
-	struct btrfs_leaf empty_leaf;
+	struct btrfs_leaf *empty_leaf;
 	struct btrfs_root_item root_item;
 	struct btrfs_item item;
 	struct btrfs_extent_item extent_item;
 	char *block;
 	int ret;
-	u16 itemoff;
+	u32 itemoff;
+	u32 start_block = BTRFS_SUPER_INFO_OFFSET / blocksize;
 
-	btrfs_set_super_blocknr(&super, 16);
-	btrfs_set_super_root(&super, 17);
+	btrfs_set_super_blocknr(&super, start_block);
+	btrfs_set_super_root(&super, start_block + 1);
 	strcpy((char *)(&super.magic), BTRFS_MAGIC);
 	btrfs_set_super_blocksize(&super, blocksize);
 	btrfs_set_super_total_blocks(&super, num_blocks);
@@ -32,168 +33,98 @@ int mkfs(int fd, u64 num_blocks, u16 blocksize)
 	memset(block, 0, blocksize);
 	BUG_ON(sizeof(super) > blocksize);
 	memcpy(block, &super, sizeof(super));
-	ret = pwrite(fd, block, blocksize, BTRFS_SUPER_INFO_OFFSET(blocksize));
+	ret = pwrite(fd, block, blocksize, BTRFS_SUPER_INFO_OFFSET);
 	BUG_ON(ret != blocksize);
 
 	/* create the tree of root objects */
-	memset(&empty_leaf, 0, sizeof(empty_leaf));
-	btrfs_set_header_parentid(&empty_leaf.header, BTRFS_ROOT_TREE_OBJECTID);
-	btrfs_set_header_blocknr(&empty_leaf.header, 17);
-	btrfs_set_header_nritems(&empty_leaf.header, 2);
+	empty_leaf = malloc(blocksize);
+	memset(empty_leaf, 0, blocksize);
+	btrfs_set_header_parentid(&empty_leaf->header,
+				  BTRFS_ROOT_TREE_OBJECTID);
+	btrfs_set_header_blocknr(&empty_leaf->header, start_block + 1);
+	btrfs_set_header_nritems(&empty_leaf->header, 2);
 
 	/* create the items for the root tree */
-	btrfs_set_root_blocknr(&root_item, 18);
+	btrfs_set_root_blocknr(&root_item, start_block + 2);
 	btrfs_set_root_refs(&root_item, 1);
-	itemoff = LEAF_DATA_SIZE - sizeof(root_item);
+	itemoff = __BTRFS_LEAF_DATA_SIZE(blocksize) - sizeof(root_item);
 	btrfs_set_item_offset(&item, itemoff);
 	btrfs_set_item_size(&item, sizeof(root_item));
 	btrfs_set_key_objectid(&item.key, BTRFS_EXTENT_TREE_OBJECTID);
 	btrfs_set_key_offset(&item.key, 0);
 	btrfs_set_key_flags(&item.key, 0);
-	memcpy(empty_leaf.items, &item, sizeof(item));
-	memcpy(empty_leaf.data + itemoff, &root_item, sizeof(root_item));
+	memcpy(empty_leaf->items, &item, sizeof(item));
+	memcpy(btrfs_leaf_data(empty_leaf) + itemoff,
+		&root_item, sizeof(root_item));
 
-	btrfs_set_root_blocknr(&root_item, 19);
+	btrfs_set_root_blocknr(&root_item, start_block + 3);
 	itemoff = itemoff - sizeof(root_item);
 	btrfs_set_item_offset(&item, itemoff);
 	btrfs_set_key_objectid(&item.key, BTRFS_FS_TREE_OBJECTID);
-	memcpy(empty_leaf.items + 1, &item, sizeof(item));
-	memcpy(empty_leaf.data + itemoff, &root_item, sizeof(root_item));
-	ret = pwrite(fd, &empty_leaf, blocksize, 17 * blocksize);
+	memcpy(empty_leaf->items + 1, &item, sizeof(item));
+	memcpy(btrfs_leaf_data(empty_leaf) + itemoff,
+		&root_item, sizeof(root_item));
+	ret = pwrite(fd, empty_leaf, blocksize, (start_block + 1) * blocksize);
 
 	/* create the items for the extent tree */
-	btrfs_set_header_parentid(&empty_leaf.header,
+	btrfs_set_header_parentid(&empty_leaf->header,
 				  BTRFS_EXTENT_TREE_OBJECTID);
-	btrfs_set_header_blocknr(&empty_leaf.header, 18);
-	btrfs_set_header_nritems(&empty_leaf.header, 4);
+	btrfs_set_header_blocknr(&empty_leaf->header, start_block + 2);
+	btrfs_set_header_nritems(&empty_leaf->header, 4);
 
 	/* item1, reserve blocks 0-16 */
 	btrfs_set_key_objectid(&item.key, 0);
-	btrfs_set_key_offset(&item.key, 17);
+	btrfs_set_key_offset(&item.key, start_block + 1);
 	btrfs_set_key_flags(&item.key, 0);
-	itemoff = LEAF_DATA_SIZE - sizeof(struct btrfs_extent_item);
+	itemoff = __BTRFS_LEAF_DATA_SIZE(blocksize) -
+			sizeof(struct btrfs_extent_item);
 	btrfs_set_item_offset(&item, itemoff);
 	btrfs_set_item_size(&item, sizeof(struct btrfs_extent_item));
 	btrfs_set_extent_refs(&extent_item, 1);
 	btrfs_set_extent_owner(&extent_item, 0);
-	memcpy(empty_leaf.items, &item, sizeof(item));
-	memcpy(empty_leaf.data + btrfs_item_offset(&item), &extent_item,
-		btrfs_item_size(&item));
+	memcpy(empty_leaf->items, &item, sizeof(item));
+	memcpy(btrfs_leaf_data(empty_leaf) + btrfs_item_offset(&item),
+		&extent_item, btrfs_item_size(&item));
 
 	/* item2, give block 17 to the root */
-	btrfs_set_key_objectid(&item.key, 17);
+	btrfs_set_key_objectid(&item.key, start_block + 1);
 	btrfs_set_key_offset(&item.key, 1);
 	itemoff = itemoff - sizeof(struct btrfs_extent_item);
 	btrfs_set_item_offset(&item, itemoff);
 	btrfs_set_extent_owner(&extent_item, BTRFS_ROOT_TREE_OBJECTID);
-	memcpy(empty_leaf.items + 1, &item, sizeof(item));
-	memcpy(empty_leaf.data + btrfs_item_offset(&item), &extent_item,
-		btrfs_item_size(&item));
+	memcpy(empty_leaf->items + 1, &item, sizeof(item));
+	memcpy(btrfs_leaf_data(empty_leaf) + btrfs_item_offset(&item),
+		&extent_item, btrfs_item_size(&item));
 
 	/* item3, give block 18 to the extent root */
-	btrfs_set_key_objectid(&item.key, 18);
+	btrfs_set_key_objectid(&item.key, start_block + 2);
 	btrfs_set_key_offset(&item.key, 1);
 	itemoff = itemoff - sizeof(struct btrfs_extent_item);
 	btrfs_set_item_offset(&item, itemoff);
 	btrfs_set_extent_owner(&extent_item, BTRFS_EXTENT_TREE_OBJECTID);
-	memcpy(empty_leaf.items + 2, &item, sizeof(item));
-	memcpy(empty_leaf.data + btrfs_item_offset(&item), &extent_item,
-		btrfs_item_size(&item));
+	memcpy(empty_leaf->items + 2, &item, sizeof(item));
+	memcpy(btrfs_leaf_data(empty_leaf) + btrfs_item_offset(&item),
+		&extent_item, btrfs_item_size(&item));
 
 	/* item4, give block 19 to the FS root */
-	btrfs_set_key_objectid(&item.key, 19);
+	btrfs_set_key_objectid(&item.key, start_block + 3);
 	btrfs_set_key_offset(&item.key, 1);
 	itemoff = itemoff - sizeof(struct btrfs_extent_item);
 	btrfs_set_item_offset(&item, itemoff);
 	btrfs_set_extent_owner(&extent_item, BTRFS_FS_TREE_OBJECTID);
-	memcpy(empty_leaf.items + 3, &item, sizeof(item));
-	memcpy(empty_leaf.data + btrfs_item_offset(&item), &extent_item,
-		btrfs_item_size(&item));
-	ret = pwrite(fd, &empty_leaf, blocksize, 18 * blocksize);
-	if (ret != sizeof(empty_leaf))
+	memcpy(empty_leaf->items + 3, &item, sizeof(item));
+	memcpy(btrfs_leaf_data(empty_leaf) + btrfs_item_offset(&item),
+		&extent_item, btrfs_item_size(&item));
+	ret = pwrite(fd, empty_leaf, blocksize, (start_block + 2) * blocksize);
+	if (ret != blocksize)
 		return -1;
 
 	/* finally create the FS root */
-	btrfs_set_header_parentid(&empty_leaf.header, BTRFS_FS_TREE_OBJECTID);
-	btrfs_set_header_blocknr(&empty_leaf.header, 19);
-	btrfs_set_header_nritems(&empty_leaf.header, 0);
-	ret = pwrite(fd, &empty_leaf, blocksize, 19 * blocksize);
-	if (ret != sizeof(empty_leaf))
+	btrfs_set_header_parentid(&empty_leaf->header, BTRFS_FS_TREE_OBJECTID);
+	btrfs_set_header_blocknr(&empty_leaf->header, start_block + 3);
+	btrfs_set_header_nritems(&empty_leaf->header, 0);
+	ret = pwrite(fd, empty_leaf, blocksize, (start_block + 3) * blocksize);
+	if (ret != blocksize)
 		return -1;
 	return 0;
 }
-
-#if 0
-int mkfs(int fd)
-{
-	struct btrfs_root_info info[2];
-	struct btrfs_leaf empty_leaf;
-	struct btrfs_item item;
-	struct btrfs_extent_item extent_item;
-	int ret;
-
-	/* setup the super block area */
-	memset(info, 0, sizeof(info));
-	btrfs_set_root_blocknr(info, 16);
-	btrfs_set_root_objectid(info, 1);
-	btrfs_set_root_tree_root(info, 17);
-
-	btrfs_set_root_blocknr(info + 1, 16);
-	btrfs_set_root_objectid(info + 1, 2);
-	btrfs_set_root_tree_root(info + 1, 18);
-
-	ret = pwrite(fd, info, sizeof(info),
-		     BTRFS_SUPER_INFO_OFFSET(BTRFS_BLOCKSIZE));
-	if (ret != sizeof(info))
-		return -1;
-
-	/* create leaves for the tree root and extent root */
-	memset(&empty_leaf, 0, sizeof(empty_leaf));
-	btrfs_set_header_parentid(&empty_leaf.header, 1);
-	btrfs_set_header_blocknr(&empty_leaf.header, 17);
-	ret = pwrite(fd, &empty_leaf, sizeof(empty_leaf), 17 * BTRFS_BLOCKSIZE);
-	if (ret != sizeof(empty_leaf))
-		return -1;
-
-	btrfs_set_header_parentid(&empty_leaf.header, 2);
-	btrfs_set_header_blocknr(&empty_leaf.header, 18);
-	btrfs_set_header_nritems(&empty_leaf.header, 3);
-
-	/* item1, reserve blocks 0-16 */
-	btrfs_set_key_objectid(&item.key, 0);
-	btrfs_set_key_offset(&item.key, 17);
-	btrfs_set_key_flags(&item.key, 0);
-	btrfs_set_item_offset(&item, LEAF_DATA_SIZE -
-			      sizeof(struct btrfs_extent_item));
-	btrfs_set_item_size(&item, sizeof(struct btrfs_extent_item));
-	btrfs_set_extent_refs(&extent_item, 1);
-	btrfs_set_extent_owner(&extent_item, 0);
-	memcpy(empty_leaf.items, &item, sizeof(item));
-	memcpy(empty_leaf.data + btrfs_item_offset(&item), &extent_item,
-		btrfs_item_size(&item));
-
-	/* item2, give block 17 to the root */
-	btrfs_set_key_objectid(&item.key, 17);
-	btrfs_set_key_offset(&item.key, 1);
-	btrfs_set_item_offset(&item, LEAF_DATA_SIZE -
-			      sizeof(struct btrfs_extent_item) * 2);
-	btrfs_set_extent_owner(&extent_item, 1);
-	memcpy(empty_leaf.items + 1, &item, sizeof(item));
-	memcpy(empty_leaf.data + btrfs_item_offset(&item), &extent_item,
-		btrfs_item_size(&item));
-
-	/* item3, give block 18 for the extent root */
-	btrfs_set_key_objectid(&item.key, 18);
-	btrfs_set_key_offset(&item.key, 1);
-	btrfs_set_item_offset(&item, LEAF_DATA_SIZE -
-			      sizeof(struct btrfs_extent_item) * 3);
-	btrfs_set_extent_owner(&extent_item, 2);
-	memcpy(empty_leaf.items + 2, &item, sizeof(item));
-	memcpy(empty_leaf.data + btrfs_item_offset(&item), &extent_item,
-		btrfs_item_size(&item));
-	ret = pwrite(fd, &empty_leaf, sizeof(empty_leaf), 18 * BTRFS_BLOCKSIZE);
-	if (ret != sizeof(empty_leaf))
-		return -1;
-	return 0;
-}
-#endif
diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c
index e769f36cf051..7a189eaa589f 100644
--- a/fs/btrfs/print-tree.c
+++ b/fs/btrfs/print-tree.c
@@ -5,7 +5,7 @@
 #include "ctree.h"
 #include "disk-io.h"
 
-void btrfs_print_leaf(struct btrfs_leaf *l)
+void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l)
 {
 	int i;
 	u32 nr = btrfs_header_nritems(&l->header);
@@ -13,7 +13,8 @@ void btrfs_print_leaf(struct btrfs_leaf *l)
 	struct btrfs_extent_item *ei;
 	struct btrfs_root_item *ri;
 	printf("leaf %Lu total ptrs %d free space %d\n",
-		btrfs_header_blocknr(&l->header), nr, btrfs_leaf_free_space(l));
+		btrfs_header_blocknr(&l->header), nr,
+		btrfs_leaf_free_space(root, l));
 	fflush(stdout);
 	for (i = 0 ; i < nr ; i++) {
 		item = l->items + i;
@@ -25,7 +26,7 @@ void btrfs_print_leaf(struct btrfs_leaf *l)
 			btrfs_item_offset(item),
 			btrfs_item_size(item));
 		printf("\t\titem data %.*s\n", btrfs_item_size(item),
-			l->data + btrfs_item_offset(item));
+			btrfs_leaf_data(l) + btrfs_item_offset(item));
 		ei = btrfs_item_ptr(l, i, struct btrfs_extent_item);
 		printf("\t\textent data refs %u owner %Lu\n",
 			btrfs_extent_refs(ei), btrfs_extent_owner(ei));
@@ -46,18 +47,18 @@ void btrfs_print_tree(struct btrfs_root *root, struct btrfs_buffer *t)
 	c = &t->node;
 	nr = btrfs_header_nritems(&c->header);
 	if (btrfs_is_leaf(c)) {
-		btrfs_print_leaf((struct btrfs_leaf *)c);
+		btrfs_print_leaf(root, (struct btrfs_leaf *)c);
 		return;
 	}
 	printf("node %Lu level %d total ptrs %d free spc %u\n", t->blocknr,
 	        btrfs_header_level(&c->header), nr,
-		(u32)NODEPTRS_PER_BLOCK - nr);
+		(u32)BTRFS_NODEPTRS_PER_BLOCK(root) - nr);
 	fflush(stdout);
 	for (i = 0; i < nr; i++) {
 		printf("\tkey %d (%Lu %u %Lu) block %Lu\n",
 		       i,
-		       c->keys[i].objectid, c->keys[i].flags, c->keys[i].offset,
-		       btrfs_node_blockptr(c, i));
+		       c->ptrs[i].key.objectid, c->ptrs[i].key.flags,
+		       c->ptrs[i].key.offset, btrfs_node_blockptr(c, i));
 		fflush(stdout);
 	}
 	for (i = 0; i < nr; i++) {
diff --git a/fs/btrfs/print-tree.h b/fs/btrfs/print-tree.h
index e8d0b847c023..9236abda2a74 100644
--- a/fs/btrfs/print-tree.h
+++ b/fs/btrfs/print-tree.h
@@ -1,3 +1,3 @@
 
-void btrfs_print_leaf(struct btrfs_leaf *l);
+void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l);
 void btrfs_print_tree(struct btrfs_root *root, struct btrfs_buffer *t);
diff --git a/fs/btrfs/quick-test.c b/fs/btrfs/quick-test.c
index 66bdc57905d7..d9287cad35c9 100644
--- a/fs/btrfs/quick-test.c
+++ b/fs/btrfs/quick-test.c
@@ -73,7 +73,7 @@ int main(int ac, char **av) {
 	printf("node %p level %d total ptrs %d free spc %lu\n", root->node,
 	        btrfs_header_level(&root->node->node.header),
 		btrfs_header_nritems(&root->node->node.header),
-		NODEPTRS_PER_BLOCK -
+		BTRFS_NODEPTRS_PER_BLOCK(root) -
 		btrfs_header_nritems(&root->node->node.header));
 	printf("all searches good, deleting some items\n");
 	i = 0;
diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c
index 0ab90cfea98f..03aa7c2c6336 100644
--- a/fs/btrfs/root-tree.c
+++ b/fs/btrfs/root-tree.c
@@ -31,7 +31,7 @@ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid,
 		ret = 1;
 		goto out;
 	}
-	memcpy(item, l->data + btrfs_item_offset(l->items + slot),
+	memcpy(item, btrfs_item_ptr(l, slot, struct btrfs_root_item),
 		sizeof(*item));
 	btrfs_disk_key_to_cpu(key, &l->items[slot].key);
 	btrfs_release_path(root, &path);
@@ -55,7 +55,7 @@ int btrfs_update_root(struct btrfs_root *root, struct btrfs_key *key,
 	BUG_ON(ret != 0);
 	l = &path.nodes[0]->leaf;
 	slot = path.slots[0];
-	memcpy(l->data + btrfs_item_offset(l->items + slot), item,
+	memcpy(btrfs_item_ptr(l, slot, struct btrfs_root_item), item,
 		sizeof(*item));
 out:
 	btrfs_release_path(root, &path);
-- 
cgit v1.2.3


From e089f05c18ab36ed5fa7e2319052e03ab800d518 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Fri, 16 Mar 2007 16:20:31 -0400
Subject: Btrfs: transaction handles everywhere

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/Makefile      |   4 +-
 fs/btrfs/TODO          |   2 +
 fs/btrfs/ctree.c       | 204 +++++++++++++++++++++++++------------------------
 fs/btrfs/ctree.h       |  63 +++++++++------
 fs/btrfs/debug-tree.c  |   1 +
 fs/btrfs/dir-item.c    |  20 +++--
 fs/btrfs/dir-test.c    |  70 ++++++++++-------
 fs/btrfs/disk-io.c     |  60 +++++++++------
 fs/btrfs/disk-io.h     |  16 ++--
 fs/btrfs/extent-tree.c | 135 +++++++++++++++++---------------
 fs/btrfs/inode-item.c  |  15 ++--
 fs/btrfs/print-tree.h  |   4 +-
 fs/btrfs/quick-test.c  |  23 +++---
 fs/btrfs/random-test.c |  62 ++++++++-------
 fs/btrfs/root-tree.c   |  23 +++---
 fs/btrfs/transaction.h |  27 +++++++
 16 files changed, 426 insertions(+), 303 deletions(-)
 create mode 100644 fs/btrfs/transaction.h

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 36964f54bb47..76bcd87d8575 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -1,5 +1,5 @@
 CC=gcc
-CFLAGS = -g -Wall
+CFLAGS = -g -Wall -Werror
 headers = radix-tree.h ctree.h disk-io.h kerncompat.h print-tree.h list.h
 objects = ctree.o disk-io.o radix-tree.o mkfs.o extent-tree.o print-tree.o \
 	  root-tree.o dir-item.o hash.o file-item.o inode-item.o
@@ -14,7 +14,7 @@ check=sparse $(CHECKFLAGS)
 	$(check) $<
 	$(CC) $(CFLAGS) -c $<
 
-all: tester debug-tree quick-test dir-test
+all: tester debug-tree quick-test dir-test tags
 
 debug-tree: $(objects) debug-tree.o
 	gcc $(CFLAGS) -o debug-tree $(objects) debug-tree.o
diff --git a/fs/btrfs/TODO b/fs/btrfs/TODO
index 2ae4b3aae1e5..e9894999322e 100644
--- a/fs/btrfs/TODO
+++ b/fs/btrfs/TODO
@@ -8,6 +8,8 @@
 * Add block mapping tree (simple dm layer)
 * Add simple tree locking (semaphore per tree)
 * Make allocator smarter
+* Only pin blocks allocated in this transaction
+* Add transaction handles
 * Do actual block accounting
 * Check compat and incompat flags on the inode
 * Port into the kernel
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 13128b5ed65c..17a3ff2f1828 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -6,17 +6,18 @@
 #include "disk-io.h"
 #include "print-tree.h"
 
-static int split_node(struct btrfs_root *root, struct btrfs_path *path,
-		      int level);
-static int split_leaf(struct btrfs_root *root, struct btrfs_path *path,
-		      int data_size);
-static int push_node_left(struct btrfs_root *root, struct btrfs_buffer *dst,
-			  struct btrfs_buffer *src);
-static int balance_node_right(struct btrfs_root *root,
-			      struct btrfs_buffer *dst_buf,
+static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, struct btrfs_path *path, int level);
+static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, struct btrfs_path *path, int data_size);
+static int push_node_left(struct btrfs_trans_handle *trans, struct btrfs_root
+			  *root, struct btrfs_buffer *dst, struct btrfs_buffer
+			  *src);
+static int balance_node_right(struct btrfs_trans_handle *trans, struct
+			      btrfs_root *root, struct btrfs_buffer *dst_buf,
 			      struct btrfs_buffer *src_buf);
-static int del_ptr(struct btrfs_root *root, struct btrfs_path *path, int level,
-		   int slot);
+static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		   struct btrfs_path *path, int level, int slot);
 
 inline void btrfs_init_path(struct btrfs_path *p)
 {
@@ -34,11 +35,10 @@ void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p)
 	memset(p, 0, sizeof(*p));
 }
 
-static int btrfs_cow_block(struct btrfs_root *root,
-		    struct btrfs_buffer *buf,
-		    struct btrfs_buffer *parent,
-		    int parent_slot,
-		    struct btrfs_buffer **cow_ret)
+static int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root
+			   *root, struct btrfs_buffer *buf, struct btrfs_buffer
+			   *parent, int parent_slot, struct btrfs_buffer
+			   **cow_ret)
 {
 	struct btrfs_buffer *cow;
 
@@ -46,22 +46,22 @@ static int btrfs_cow_block(struct btrfs_root *root,
 		*cow_ret = buf;
 		return 0;
 	}
-	cow = btrfs_alloc_free_block(root);
+	cow = btrfs_alloc_free_block(trans, root);
 	memcpy(&cow->node, &buf->node, root->blocksize);
 	btrfs_set_header_blocknr(&cow->node.header, cow->blocknr);
 	*cow_ret = cow;
-	btrfs_inc_ref(root, buf);
+	btrfs_inc_ref(trans, root, buf);
 	if (buf == root->node) {
 		root->node = cow;
 		cow->count++;
 		if (buf != root->commit_root)
-			btrfs_free_extent(root, buf->blocknr, 1, 1);
+			btrfs_free_extent(trans, root, buf->blocknr, 1, 1);
 		btrfs_block_release(root, buf);
 	} else {
 		btrfs_set_node_blockptr(&parent->node, parent_slot,
 					cow->blocknr);
 		BUG_ON(list_empty(&parent->dirty));
-		btrfs_free_extent(root, buf->blocknr, 1, 1);
+		btrfs_free_extent(trans, root, buf->blocknr, 1, 1);
 	}
 	btrfs_block_release(root, buf);
 	return 0;
@@ -266,8 +266,8 @@ static struct btrfs_buffer *read_node_slot(struct btrfs_root *root,
 	return read_tree_block(root, btrfs_node_blockptr(node, slot));
 }
 
-static int balance_level(struct btrfs_root *root, struct btrfs_path *path,
-			int level)
+static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
+			 *root, struct btrfs_path *path, int level)
 {
 	struct btrfs_buffer *right_buf;
 	struct btrfs_buffer *mid_buf;
@@ -310,8 +310,8 @@ static int balance_level(struct btrfs_root *root, struct btrfs_path *path,
 		btrfs_block_release(root, mid_buf);
 		/* once for the root ptr */
 		btrfs_block_release(root, mid_buf);
-		clean_tree_block(root, mid_buf);
-		return btrfs_free_extent(root, blocknr, 1, 1);
+		clean_tree_block(trans, root, mid_buf);
+		return btrfs_free_extent(trans, root, blocknr, 1, 1);
 	}
 	parent = &parent_buf->node;
 
@@ -324,11 +324,11 @@ static int balance_level(struct btrfs_root *root, struct btrfs_path *path,
 
 	/* first, try to make some room in the middle buffer */
 	if (left_buf) {
-		btrfs_cow_block(root, left_buf, parent_buf,
-				pslot - 1, &left_buf);
+		btrfs_cow_block(trans, root, left_buf, parent_buf, pslot - 1,
+				&left_buf);
 		left = &left_buf->node;
 		orig_slot += btrfs_header_nritems(&left->header);
-		wret = push_node_left(root, left_buf, mid_buf);
+		wret = push_node_left(trans, root, left_buf, mid_buf);
 		if (wret < 0)
 			ret = wret;
 	}
@@ -337,22 +337,23 @@ static int balance_level(struct btrfs_root *root, struct btrfs_path *path,
 	 * then try to empty the right most buffer into the middle
 	 */
 	if (right_buf) {
-		btrfs_cow_block(root, right_buf, parent_buf,
-				pslot + 1, &right_buf);
+		btrfs_cow_block(trans, root, right_buf, parent_buf, pslot + 1,
+				&right_buf);
 		right = &right_buf->node;
-		wret = push_node_left(root, mid_buf, right_buf);
+		wret = push_node_left(trans, root, mid_buf, right_buf);
 		if (wret < 0)
 			ret = wret;
 		if (btrfs_header_nritems(&right->header) == 0) {
 			u64 blocknr = right_buf->blocknr;
 			btrfs_block_release(root, right_buf);
-			clean_tree_block(root, right_buf);
+			clean_tree_block(trans, root, right_buf);
 			right_buf = NULL;
 			right = NULL;
-			wret = del_ptr(root, path, level + 1, pslot + 1);
+			wret = del_ptr(trans, root, path, level + 1, pslot +
+				       1);
 			if (wret)
 				ret = wret;
-			wret = btrfs_free_extent(root, blocknr, 1, 1);
+			wret = btrfs_free_extent(trans, root, blocknr, 1, 1);
 			if (wret)
 				ret = wret;
 		} else {
@@ -373,7 +374,7 @@ static int balance_level(struct btrfs_root *root, struct btrfs_path *path,
 		 * right
 		 */
 		BUG_ON(!left_buf);
-		wret = balance_node_right(root, mid_buf, left_buf);
+		wret = balance_node_right(trans, root, mid_buf, left_buf);
 		if (wret < 0)
 			ret = wret;
 		BUG_ON(wret == 1);
@@ -382,13 +383,13 @@ static int balance_level(struct btrfs_root *root, struct btrfs_path *path,
 		/* we've managed to empty the middle node, drop it */
 		u64 blocknr = mid_buf->blocknr;
 		btrfs_block_release(root, mid_buf);
-		clean_tree_block(root, mid_buf);
+		clean_tree_block(trans, root, mid_buf);
 		mid_buf = NULL;
 		mid = NULL;
-		wret = del_ptr(root, path, level + 1, pslot);
+		wret = del_ptr(trans, root, path, level + 1, pslot);
 		if (wret)
 			ret = wret;
-		wret = btrfs_free_extent(root, blocknr, 1, 1);
+		wret = btrfs_free_extent(trans, root, blocknr, 1, 1);
 		if (wret)
 			ret = wret;
 	} else {
@@ -438,8 +439,9 @@ static int balance_level(struct btrfs_root *root, struct btrfs_path *path,
  * tree.  if ins_len < 0, nodes will be merged as we walk down the tree (if
  * possible)
  */
-int btrfs_search_slot(struct btrfs_root *root, struct btrfs_key *key,
-		struct btrfs_path *p, int ins_len, int cow)
+int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, struct btrfs_key *key, struct btrfs_path *p, int
+		      ins_len, int cow)
 {
 	struct btrfs_buffer *b;
 	struct btrfs_buffer *cow_buf;
@@ -455,8 +457,9 @@ again:
 		level = btrfs_header_level(&b->node.header);
 		if (cow) {
 			int wret;
-			wret = btrfs_cow_block(root, b, p->nodes[level + 1],
-					       p->slots[level + 1], &cow_buf);
+			wret = btrfs_cow_block(trans, root, b, p->nodes[level +
+					       1], p->slots[level + 1],
+					       &cow_buf);
 			b = cow_buf;
 		}
 		BUG_ON(!cow && ins_len);
@@ -472,7 +475,7 @@ again:
 			p->slots[level] = slot;
 			if (ins_len > 0 && btrfs_header_nritems(&c->header) ==
 			    BTRFS_NODEPTRS_PER_BLOCK(root)) {
-				int sret = split_node(root, p, level);
+				int sret = split_node(trans, root, p, level);
 				BUG_ON(sret > 0);
 				if (sret)
 					return sret;
@@ -480,7 +483,8 @@ again:
 				c = &b->node;
 				slot = p->slots[level];
 			} else if (ins_len < 0) {
-				int sret = balance_level(root, p, level);
+				int sret = balance_level(trans, root, p,
+							 level);
 				if (sret)
 					return sret;
 				b = p->nodes[level];
@@ -496,7 +500,7 @@ again:
 			p->slots[level] = slot;
 			if (ins_len > 0 && btrfs_leaf_free_space(root, l) <
 			    sizeof(struct btrfs_item) + ins_len) {
-				int sret = split_leaf(root, p, ins_len);
+				int sret = split_leaf(trans, root, p, ins_len);
 				BUG_ON(sret > 0);
 				if (sret)
 					return sret;
@@ -519,9 +523,9 @@ again:
  * If this fails to write a tree block, it returns -1, but continues
  * fixing up the blocks in ram so the tree is consistent.
  */
-static int fixup_low_keys(struct btrfs_root *root,
-			   struct btrfs_path *path, struct btrfs_disk_key *key,
-			   int level)
+static int fixup_low_keys(struct btrfs_trans_handle *trans, struct btrfs_root
+			  *root, struct btrfs_path *path, struct btrfs_disk_key
+			  *key, int level)
 {
 	int i;
 	int ret = 0;
@@ -546,8 +550,9 @@ static int fixup_low_keys(struct btrfs_root *root,
  * returns 0 if some ptrs were pushed left, < 0 if there was some horrible
  * error, and > 0 if there was no room in the left hand block.
  */
-static int push_node_left(struct btrfs_root *root, struct btrfs_buffer *dst_buf,
-			  struct btrfs_buffer *src_buf)
+static int push_node_left(struct btrfs_trans_handle *trans, struct btrfs_root
+			  *root, struct btrfs_buffer *dst_buf, struct
+			  btrfs_buffer *src_buf)
 {
 	struct btrfs_node *src = &src_buf->node;
 	struct btrfs_node *dst = &dst_buf->node;
@@ -589,8 +594,8 @@ static int push_node_left(struct btrfs_root *root, struct btrfs_buffer *dst_buf,
  *
  * this will  only push up to 1/2 the contents of the left node over
  */
-static int balance_node_right(struct btrfs_root *root,
-			      struct btrfs_buffer *dst_buf,
+static int balance_node_right(struct btrfs_trans_handle *trans, struct
+			      btrfs_root *root, struct btrfs_buffer *dst_buf,
 			      struct btrfs_buffer *src_buf)
 {
 	struct btrfs_node *src = &src_buf->node;
@@ -635,8 +640,8 @@ static int balance_node_right(struct btrfs_root *root,
  *
  * returns zero on success or < 0 on failure.
  */
-static int insert_new_root(struct btrfs_root *root,
-			   struct btrfs_path *path, int level)
+static int insert_new_root(struct btrfs_trans_handle *trans, struct btrfs_root
+			   *root, struct btrfs_path *path, int level)
 {
 	struct btrfs_buffer *t;
 	struct btrfs_node *lower;
@@ -646,7 +651,7 @@ static int insert_new_root(struct btrfs_root *root,
 	BUG_ON(path->nodes[level]);
 	BUG_ON(path->nodes[level-1] != root->node);
 
-	t = btrfs_alloc_free_block(root);
+	t = btrfs_alloc_free_block(trans, root);
 	c = &t->node;
 	memset(c, 0, root->blocksize);
 	btrfs_set_header_nritems(&c->header, 1);
@@ -679,9 +684,9 @@ static int insert_new_root(struct btrfs_root *root,
  *
  * returns zero on success and < 0 on any error
  */
-static int insert_ptr(struct btrfs_root *root,
-		struct btrfs_path *path, struct btrfs_disk_key *key,
-		u64 blocknr, int slot, int level)
+static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, struct btrfs_path *path, struct btrfs_disk_key
+		      *key, u64 blocknr, int slot, int level)
 {
 	struct btrfs_node *lower;
 	int nritems;
@@ -713,8 +718,8 @@ static int insert_ptr(struct btrfs_root *root,
  *
  * returns 0 on success and < 0 on failure
  */
-static int split_node(struct btrfs_root *root, struct btrfs_path *path,
-		      int level)
+static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, struct btrfs_path *path, int level)
 {
 	struct btrfs_buffer *t;
 	struct btrfs_node *c;
@@ -729,12 +734,12 @@ static int split_node(struct btrfs_root *root, struct btrfs_path *path,
 	c = &t->node;
 	if (t == root->node) {
 		/* trying to split the root, lets make a new one */
-		ret = insert_new_root(root, path, level + 1);
+		ret = insert_new_root(trans, root, path, level + 1);
 		if (ret)
 			return ret;
 	}
 	c_nritems = btrfs_header_nritems(&c->header);
-	split_buffer = btrfs_alloc_free_block(root);
+	split_buffer = btrfs_alloc_free_block(trans, root);
 	split = &split_buffer->node;
 	btrfs_set_header_flags(&split->header, btrfs_header_flags(&c->header));
 	btrfs_set_header_blocknr(&split->header, split_buffer->blocknr);
@@ -748,7 +753,7 @@ static int split_node(struct btrfs_root *root, struct btrfs_path *path,
 	ret = 0;
 
 	BUG_ON(list_empty(&t->dirty));
-	wret = insert_ptr(root, path, &split->ptrs[0].key,
+	wret = insert_ptr(trans, root, path, &split->ptrs[0].key,
 			  split_buffer->blocknr, path->slots[level + 1] + 1,
 			  level + 1);
 	if (wret)
@@ -790,8 +795,8 @@ static int leaf_space_used(struct btrfs_leaf *l, int start, int nr)
  * returns 1 if the push failed because the other node didn't have enough
  * room, 0 if everything worked out and < 0 if there were major errors.
  */
-static int push_leaf_right(struct btrfs_root *root, struct btrfs_path *path,
-			   int data_size)
+static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
+			   *root, struct btrfs_path *path, int data_size)
 {
 	struct btrfs_buffer *left_buf = path->nodes[0];
 	struct btrfs_leaf *left = &left_buf->leaf;
@@ -824,7 +829,7 @@ static int push_leaf_right(struct btrfs_root *root, struct btrfs_path *path,
 		return 1;
 	}
 	/* cow and double check */
-	btrfs_cow_block(root, right_buf, upper, slot + 1, &right_buf);
+	btrfs_cow_block(trans, root, right_buf, upper, slot + 1, &right_buf);
 	right = &right_buf->leaf;
 	free_space = btrfs_leaf_free_space(root, right);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
@@ -897,8 +902,8 @@ static int push_leaf_right(struct btrfs_root *root, struct btrfs_path *path,
  * push some data in the path leaf to the left, trying to free up at
  * least data_size bytes.  returns zero if the push worked, nonzero otherwise
  */
-static int push_leaf_left(struct btrfs_root *root, struct btrfs_path *path,
-			  int data_size)
+static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
+			  *root, struct btrfs_path *path, int data_size)
 {
 	struct btrfs_buffer *right_buf = path->nodes[0];
 	struct btrfs_leaf *right = &right_buf->leaf;
@@ -931,7 +936,7 @@ static int push_leaf_left(struct btrfs_root *root, struct btrfs_path *path,
 	}
 
 	/* cow and double check */
-	btrfs_cow_block(root, t, path->nodes[1], slot - 1, &t);
+	btrfs_cow_block(trans, root, t, path->nodes[1], slot - 1, &t);
 	left = &t->leaf;
 	free_space = btrfs_leaf_free_space(root, left);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
@@ -997,7 +1002,7 @@ static int push_leaf_left(struct btrfs_root *root, struct btrfs_path *path,
 	BUG_ON(list_empty(&t->dirty));
 	BUG_ON(list_empty(&right_buf->dirty));
 
-	wret = fixup_low_keys(root, path, &right->items[0].key, 1);
+	wret = fixup_low_keys(trans, root, path, &right->items[0].key, 1);
 	if (wret)
 		ret = wret;
 
@@ -1021,8 +1026,8 @@ static int push_leaf_left(struct btrfs_root *root, struct btrfs_path *path,
  *
  * returns 0 if all went well and < 0 on failure.
  */
-static int split_leaf(struct btrfs_root *root, struct btrfs_path *path,
-		      int data_size)
+static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, struct btrfs_path *path, int data_size)
 {
 	struct btrfs_buffer *l_buf;
 	struct btrfs_leaf *l;
@@ -1038,11 +1043,11 @@ static int split_leaf(struct btrfs_root *root, struct btrfs_path *path,
 	int ret;
 	int wret;
 
-	wret = push_leaf_left(root, path, data_size);
+	wret = push_leaf_left(trans, root, path, data_size);
 	if (wret < 0)
 		return wret;
 	if (wret) {
-		wret = push_leaf_right(root, path, data_size);
+		wret = push_leaf_right(trans, root, path, data_size);
 		if (wret < 0)
 			return wret;
 	}
@@ -1055,14 +1060,14 @@ static int split_leaf(struct btrfs_root *root, struct btrfs_path *path,
 		return 0;
 
 	if (!path->nodes[1]) {
-		ret = insert_new_root(root, path, 1);
+		ret = insert_new_root(trans, root, path, 1);
 		if (ret)
 			return ret;
 	}
 	slot = path->slots[0];
 	nritems = btrfs_header_nritems(&l->header);
 	mid = (nritems + 1)/ 2;
-	right_buffer = btrfs_alloc_free_block(root);
+	right_buffer = btrfs_alloc_free_block(trans, root);
 	BUG_ON(!right_buffer);
 	BUG_ON(mid == nritems);
 	right = &right_buffer->leaf;
@@ -1100,7 +1105,7 @@ static int split_leaf(struct btrfs_root *root, struct btrfs_path *path,
 
 	btrfs_set_header_nritems(&l->header, mid);
 	ret = 0;
-	wret = insert_ptr(root, path, &right->items[0].key,
+	wret = insert_ptr(trans, root, path, &right->items[0].key,
 			  right_buffer->blocknr, path->slots[1] + 1, 1);
 	if (wret)
 		ret = wret;
@@ -1122,8 +1127,9 @@ static int split_leaf(struct btrfs_root *root, struct btrfs_path *path,
  * Given a key and some data, insert an item into the tree.
  * This does all the path init required, making room in the tree if needed.
  */
-int btrfs_insert_empty_item(struct btrfs_root *root, struct btrfs_path *path,
-			    struct btrfs_key *cpu_key, u32 data_size)
+int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, struct btrfs_root
+			    *root, struct btrfs_path *path, struct btrfs_key
+			    *cpu_key, u32 data_size)
 {
 	int ret = 0;
 	int slot;
@@ -1139,7 +1145,7 @@ int btrfs_insert_empty_item(struct btrfs_root *root, struct btrfs_path *path,
 	/* create a root if there isn't one */
 	if (!root->node)
 		BUG();
-	ret = btrfs_search_slot(root, cpu_key, path, data_size, 1);
+	ret = btrfs_search_slot(trans, root, cpu_key, path, data_size, 1);
 	if (ret == 0) {
 		btrfs_release_path(root, path);
 		return -EEXIST;
@@ -1193,7 +1199,7 @@ int btrfs_insert_empty_item(struct btrfs_root *root, struct btrfs_path *path,
 
 	ret = 0;
 	if (slot == 0)
-		ret = fixup_low_keys(root, path, &disk_key, 1);
+		ret = fixup_low_keys(trans, root, path, &disk_key, 1);
 
 	BUG_ON(list_empty(&leaf_buf->dirty));
 	if (btrfs_leaf_free_space(root, leaf) < 0)
@@ -1207,15 +1213,16 @@ out:
  * Given a key and some data, insert an item into the tree.
  * This does all the path init required, making room in the tree if needed.
  */
-int btrfs_insert_item(struct btrfs_root *root, struct btrfs_key *cpu_key,
-			  void *data, u32 data_size)
+int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, struct btrfs_key *cpu_key, void *data, u32
+		      data_size)
 {
 	int ret = 0;
 	struct btrfs_path path;
 	u8 *ptr;
 
 	btrfs_init_path(&path);
-	ret = btrfs_insert_empty_item(root, &path, cpu_key, data_size);
+	ret = btrfs_insert_empty_item(trans, root, &path, cpu_key, data_size);
 	if (!ret) {
 		ptr = btrfs_item_ptr(&path.nodes[0]->leaf, path.slots[0], u8);
 		memcpy(ptr, data, data_size);
@@ -1231,8 +1238,8 @@ int btrfs_insert_item(struct btrfs_root *root, struct btrfs_key *cpu_key,
  * continuing all the way the root if required.  The root is converted into
  * a leaf if all the nodes are emptied.
  */
-static int del_ptr(struct btrfs_root *root, struct btrfs_path *path, int level,
-		   int slot)
+static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		   struct btrfs_path *path, int level, int slot)
 {
 	struct btrfs_node *node;
 	struct btrfs_buffer *parent = path->nodes[level];
@@ -1253,7 +1260,7 @@ static int del_ptr(struct btrfs_root *root, struct btrfs_path *path, int level,
 		/* just turn the root into a leaf and break */
 		btrfs_set_header_level(&root->node->node.header, 0);
 	} else if (slot == 0) {
-		wret = fixup_low_keys(root, path, &node->ptrs[0].key,
+		wret = fixup_low_keys(trans, root, path, &node->ptrs[0].key,
 				      level + 1);
 		if (wret)
 			ret = wret;
@@ -1266,7 +1273,8 @@ static int del_ptr(struct btrfs_root *root, struct btrfs_path *path, int level,
  * delete the item at the leaf level in path.  If that empties
  * the leaf, remove it from the tree
  */
-int btrfs_del_item(struct btrfs_root *root, struct btrfs_path *path)
+int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		   struct btrfs_path *path)
 {
 	int slot;
 	struct btrfs_leaf *leaf;
@@ -1306,19 +1314,20 @@ int btrfs_del_item(struct btrfs_root *root, struct btrfs_path *path)
 			btrfs_set_header_level(&leaf->header, 0);
 			BUG_ON(list_empty(&leaf_buf->dirty));
 		} else {
-			clean_tree_block(root, leaf_buf);
-			wret = del_ptr(root, path, 1, path->slots[1]);
+			clean_tree_block(trans, root, leaf_buf);
+			wret = del_ptr(trans, root, path, 1, path->slots[1]);
 			if (wret)
 				ret = wret;
-			wret = btrfs_free_extent(root, leaf_buf->blocknr, 1, 1);
+			wret = btrfs_free_extent(trans, root,
+						 leaf_buf->blocknr, 1, 1);
 			if (wret)
 				ret = wret;
 		}
 	} else {
 		int used = leaf_space_used(leaf, 0, nritems);
 		if (slot == 0) {
-			wret = fixup_low_keys(root, path,
-						   &leaf->items[0].key, 1);
+			wret = fixup_low_keys(trans, root, path,
+					      &leaf->items[0].key, 1);
 			if (wret)
 				ret = wret;
 		}
@@ -1332,23 +1341,24 @@ int btrfs_del_item(struct btrfs_root *root, struct btrfs_path *path)
 			 */
 			slot = path->slots[1];
 			leaf_buf->count++;
-			wret = push_leaf_left(root, path, 1);
+			wret = push_leaf_left(trans, root, path, 1);
 			if (wret < 0)
 				ret = wret;
 			if (path->nodes[0] == leaf_buf &&
 			    btrfs_header_nritems(&leaf->header)) {
-				wret = push_leaf_right(root, path, 1);
+				wret = push_leaf_right(trans, root, path, 1);
 				if (wret < 0)
 					ret = wret;
 			}
 			if (btrfs_header_nritems(&leaf->header) == 0) {
 				u64 blocknr = leaf_buf->blocknr;
-				clean_tree_block(root, leaf_buf);
-				wret = del_ptr(root, path, 1, slot);
+				clean_tree_block(trans, root, leaf_buf);
+				wret = del_ptr(trans, root, path, 1, slot);
 				if (wret)
 					ret = wret;
 				btrfs_block_release(root, leaf_buf);
-				wret = btrfs_free_extent(root, blocknr, 1, 1);
+				wret = btrfs_free_extent(trans, root, blocknr,
+							 1, 1);
 				if (wret)
 					ret = wret;
 			} else {
@@ -1401,5 +1411,3 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 	}
 	return 0;
 }
-
-
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index d15a2ed95076..68f0af39777f 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -4,6 +4,8 @@
 #include "list.h"
 #include "kerncompat.h"
 
+struct btrfs_trans_handle;
+
 #define BTRFS_MAGIC "_BtRfS_M"
 
 #define BTRFS_ROOT_TREE_OBJECTID 1
@@ -200,6 +202,7 @@ struct btrfs_root {
 	struct btrfs_root_item root_item;
 	struct btrfs_key root_key;
 	u32 blocksize;
+	struct btrfs_trans_handle *running_transaction;
 };
 
 /* the lower bits in the key flags defines the item type */
@@ -656,34 +659,46 @@ static inline u8 *btrfs_leaf_data(struct btrfs_leaf *l)
 	((type *)(btrfs_leaf_data(leaf) + \
 	btrfs_item_offset((leaf)->items + (slot))))
 
-struct btrfs_buffer *btrfs_alloc_free_block(struct btrfs_root *root);
-int btrfs_inc_ref(struct btrfs_root *root, struct btrfs_buffer *buf);
-int btrfs_free_extent(struct btrfs_root *root, u64 blocknr, u64 num_blocks,
-		      int pin);
-int btrfs_search_slot(struct btrfs_root *root, struct btrfs_key *key,
-		struct btrfs_path *p, int ins_len, int cow);
+struct btrfs_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
+					    struct btrfs_root *root);
+int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		  struct btrfs_buffer *buf);
+int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, u64 blocknr, u64 num_blocks, int pin);
+int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, struct btrfs_key *key, struct btrfs_path *p, int
+		      ins_len, int cow);
 void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p);
 void btrfs_init_path(struct btrfs_path *p);
-int btrfs_del_item(struct btrfs_root *root, struct btrfs_path *path);
-int btrfs_insert_item(struct btrfs_root *root, struct btrfs_key *key,
-		void *data, u32 data_size);
-int btrfs_insert_empty_item(struct btrfs_root *root, struct btrfs_path *path,
-			    struct btrfs_key *cpu_key, u32 data_size);
+int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		   struct btrfs_path *path);
+int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, struct btrfs_key *key, void *data, u32 data_size);
+int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, struct btrfs_root
+			    *root, struct btrfs_path *path, struct btrfs_key
+			    *cpu_key, u32 data_size);
 int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path);
 int btrfs_leaf_free_space(struct btrfs_root *root, struct btrfs_leaf *leaf);
-int btrfs_drop_snapshot(struct btrfs_root *root, struct btrfs_buffer *snap);
-int btrfs_finish_extent_commit(struct btrfs_root *root);
-int btrfs_del_root(struct btrfs_root *root, struct btrfs_key *key);
-int btrfs_insert_root(struct btrfs_root *root, struct btrfs_key *key,
-		      struct btrfs_root_item *item);
-int btrfs_update_root(struct btrfs_root *root, struct btrfs_key *key,
-		      struct btrfs_root_item *item);
-int btrfs_find_last_root(struct btrfs_root *root, u64 objectid,
-			struct btrfs_root_item *item, struct btrfs_key *key);
-int btrfs_insert_dir_item(struct btrfs_root *root, char *name, int name_len,
-			  u64 dir, u64 objectid, u8 type);
-int btrfs_lookup_dir_item(struct btrfs_root *root, struct btrfs_path *path,
-			  u64 dir, char *name, int name_len, int mod);
+int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
+			*root, struct btrfs_buffer *snap);
+int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct
+			       btrfs_root *root);
+int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		   struct btrfs_key *key);
+int btrfs_insert_root(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, struct btrfs_key *key, struct btrfs_root_item
+		      *item);
+int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, struct btrfs_key *key, struct btrfs_root_item
+		      *item);
+int btrfs_find_last_root(struct btrfs_root *root, u64 objectid, struct
+			 btrfs_root_item *item, struct btrfs_key *key);
+int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
+			  *root, char *name, int name_len, u64 dir, u64
+			  objectid, u8 type);
+int btrfs_lookup_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
+			  *root, struct btrfs_path *path, u64 dir, char *name,
+			  int name_len, int mod);
 int btrfs_match_dir_item_name(struct btrfs_root *root, struct btrfs_path *path,
 			      char *name, int name_len);
 #endif
diff --git a/fs/btrfs/debug-tree.c b/fs/btrfs/debug-tree.c
index de45fb4dfddd..91dea7a0a476 100644
--- a/fs/btrfs/debug-tree.c
+++ b/fs/btrfs/debug-tree.c
@@ -5,6 +5,7 @@
 #include "ctree.h"
 #include "disk-io.h"
 #include "print-tree.h"
+#include "transaction.h"
 
 int main(int ac, char **av) {
 	struct btrfs_super_block super;
diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c
index a42a67b99753..949c4e526798 100644
--- a/fs/btrfs/dir-item.c
+++ b/fs/btrfs/dir-item.c
@@ -5,9 +5,11 @@
 #include "ctree.h"
 #include "disk-io.h"
 #include "hash.h"
+#include "transaction.h"
 
-int btrfs_insert_dir_item(struct btrfs_root *root, char *name, int name_len,
-			  u64 dir, u64 objectid, u8 type)
+int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
+			  *root, char *name, int name_len, u64 dir, u64
+			  objectid, u8 type)
 {
 	int ret = 0;
 	struct btrfs_path path;
@@ -23,7 +25,7 @@ int btrfs_insert_dir_item(struct btrfs_root *root, char *name, int name_len,
 	BUG_ON(ret);
 	btrfs_init_path(&path);
 	data_size = sizeof(*dir_item) + name_len;
-	ret = btrfs_insert_empty_item(root, &path, &key, data_size);
+	ret = btrfs_insert_empty_item(trans, root, &path, &key, data_size);
 	if (ret)
 		goto out;
 
@@ -40,8 +42,9 @@ out:
 	return ret;
 }
 
-int btrfs_lookup_dir_item(struct btrfs_root *root, struct btrfs_path *path,
-			  u64 dir, char *name, int name_len, int mod)
+int btrfs_lookup_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
+			  *root, struct btrfs_path *path, u64 dir, char *name,
+			  int name_len, int mod)
 {
 	int ret;
 	struct btrfs_key key;
@@ -53,12 +56,13 @@ int btrfs_lookup_dir_item(struct btrfs_root *root, struct btrfs_path *path,
 	btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY);
 	ret = btrfs_name_hash(name, name_len, &key.offset);
 	BUG_ON(ret);
-	ret = btrfs_search_slot(root, &key, path, ins_len, cow);
+	ret = btrfs_search_slot(trans, root, &key, path, ins_len, cow);
 	return ret;
 }
 
-int btrfs_match_dir_item_name(struct btrfs_root *root, struct btrfs_path *path,
-			      char *name, int name_len)
+int btrfs_match_dir_item_name(struct btrfs_root *root,
+			      struct btrfs_path *path, char
+			      *name, int name_len)
 {
 	struct btrfs_dir_item *dir_item;
 	char *name_ptr;
diff --git a/fs/btrfs/dir-test.c b/fs/btrfs/dir-test.c
index f73aa7623398..e908c0c588cc 100644
--- a/fs/btrfs/dir-test.c
+++ b/fs/btrfs/dir-test.c
@@ -8,6 +8,7 @@
 #include "disk-io.h"
 #include "print-tree.h"
 #include "hash.h"
+#include "transaction.h"
 
 int keep_running = 1;
 struct btrfs_super_block super;
@@ -38,7 +39,8 @@ again:
 	return 0;
 }
 
-static int ins_one(struct btrfs_root *root, struct radix_tree_root *radix)
+static int ins_one(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		   struct radix_tree_root *radix)
 {
 	int ret;
 	char buf[128];
@@ -48,8 +50,8 @@ static int ins_one(struct btrfs_root *root, struct radix_tree_root *radix)
 	find_num(radix, &oid, 0);
 	sprintf(buf, "str-%lu", oid);
 
-	ret = btrfs_insert_dir_item(root, buf, strlen(buf), dir_oid, file_oid,
-				    1);
+	ret = btrfs_insert_dir_item(trans, root, buf, strlen(buf), dir_oid,
+				    file_oid, 1);
 	if (ret)
 		goto error;
 
@@ -68,7 +70,8 @@ error:
 	 * check
 	 */
 	btrfs_init_path(&path);
-	ret = btrfs_lookup_dir_item(root, &path, dir_oid, buf, strlen(buf), 0);
+	ret = btrfs_lookup_dir_item(trans, root, &path, dir_oid, buf,
+				    strlen(buf), 0);
 	if (ret)
 		goto fatal_release;
 	if (!btrfs_match_dir_item_name(root, &path, buf, strlen(buf))) {
@@ -96,7 +99,8 @@ fatal:
 	return -1;
 }
 
-static int insert_dup(struct btrfs_root *root, struct radix_tree_root *radix)
+static int insert_dup(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, struct radix_tree_root *radix)
 {
 	int ret;
 	char buf[128];
@@ -107,8 +111,8 @@ static int insert_dup(struct btrfs_root *root, struct radix_tree_root *radix)
 		return 0;
 	sprintf(buf, "str-%lu", oid);
 
-	ret = btrfs_insert_dir_item(root, buf, strlen(buf), dir_oid, file_oid,
-				    1);
+	ret = btrfs_insert_dir_item(trans, root, buf, strlen(buf), dir_oid,
+				    file_oid, 1);
 	if (ret != -EEXIST) {
 		printf("insert on %s gave us %d\n", buf, ret);
 		return 1;
@@ -116,7 +120,8 @@ static int insert_dup(struct btrfs_root *root, struct radix_tree_root *radix)
 	return 0;
 }
 
-static int del_one(struct btrfs_root *root, struct radix_tree_root *radix)
+static int del_one(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		   struct radix_tree_root *radix)
 {
 	int ret;
 	char buf[128];
@@ -129,10 +134,11 @@ static int del_one(struct btrfs_root *root, struct radix_tree_root *radix)
 		return 0;
 	sprintf(buf, "str-%lu", oid);
 	btrfs_init_path(&path);
-	ret = btrfs_lookup_dir_item(root, &path, dir_oid, buf, strlen(buf), -1);
+	ret = btrfs_lookup_dir_item(trans, root, &path, dir_oid, buf,
+				    strlen(buf), -1);
 	if (ret)
 		goto out_release;
-	ret = btrfs_del_item(root, &path);
+	ret = btrfs_del_item(trans, root, &path);
 	if (ret)
 		goto out_release;
 	btrfs_release_path(root, &path);
@@ -149,7 +155,8 @@ out:
 	return -1;
 }
 
-static int lookup_item(struct btrfs_root *root, struct radix_tree_root *radix)
+static int lookup_item(struct btrfs_trans_handle *trans, struct btrfs_root
+		       *root, struct radix_tree_root *radix)
 {
 	struct btrfs_path path;
 	char buf[128];
@@ -161,7 +168,8 @@ static int lookup_item(struct btrfs_root *root, struct radix_tree_root *radix)
 		return 0;
 	sprintf(buf, "str-%lu", oid);
 	btrfs_init_path(&path);
-	ret = btrfs_lookup_dir_item(root, &path, dir_oid, buf, strlen(buf), 0);
+	ret = btrfs_lookup_dir_item(trans, root, &path, dir_oid, buf,
+				    strlen(buf), 0);
 	btrfs_release_path(root, &path);
 	if (ret) {
 		printf("unable to find key %lu\n", oid);
@@ -170,7 +178,8 @@ static int lookup_item(struct btrfs_root *root, struct radix_tree_root *radix)
 	return 0;
 }
 
-static int lookup_enoent(struct btrfs_root *root, struct radix_tree_root *radix)
+static int lookup_enoent(struct btrfs_trans_handle *trans, struct btrfs_root
+			 *root, struct radix_tree_root *radix)
 {
 	struct btrfs_path path;
 	char buf[128];
@@ -182,7 +191,8 @@ static int lookup_enoent(struct btrfs_root *root, struct radix_tree_root *radix)
 		return 0;
 	sprintf(buf, "str-%lu", oid);
 	btrfs_init_path(&path);
-	ret = btrfs_lookup_dir_item(root, &path, dir_oid, buf, strlen(buf), 0);
+	ret = btrfs_lookup_dir_item(trans, root, &path, dir_oid, buf,
+				    strlen(buf), 0);
 	btrfs_release_path(root, &path);
 	if (!ret) {
 		printf("able to find key that should not exist %lu\n", oid);
@@ -191,8 +201,8 @@ static int lookup_enoent(struct btrfs_root *root, struct radix_tree_root *radix)
 	return 0;
 }
 
-static int empty_tree(struct btrfs_root *root, struct radix_tree_root *radix,
-		      int nr)
+static int empty_tree(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, struct radix_tree_root *radix, int nr)
 {
 	struct btrfs_path path;
 	struct btrfs_key key;
@@ -211,7 +221,7 @@ static int empty_tree(struct btrfs_root *root, struct radix_tree_root *radix,
 	key.objectid = dir_oid;
 	while(nr-- >= 0) {
 		btrfs_init_path(&path);
-		ret = btrfs_search_slot(root, &key, &path, -1, 1);
+		ret = btrfs_search_slot(trans, root, &key, &path, -1, 1);
 		if (ret < 0) {
 			btrfs_release_path(root, &path);
 			return ret;
@@ -231,7 +241,7 @@ static int empty_tree(struct btrfs_root *root, struct radix_tree_root *radix,
 		BUG_ON(found_len > 128);
 		buf[found_len] = '\0';
 		found = atoi(buf + 4);
-		ret = btrfs_del_item(root, &path);
+		ret = btrfs_del_item(trans, root, &path);
 		count++;
 		if (ret) {
 			fprintf(stderr,
@@ -252,19 +262,19 @@ error:
 	return -1;
 }
 
-static int fill_tree(struct btrfs_root *root, struct radix_tree_root *radix,
-		     int count)
+static int fill_tree(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		     struct radix_tree_root *radix, int count)
 {
 	int i;
 	int ret = 0;
 	for (i = 0; i < count; i++) {
-		ret = ins_one(root, radix);
+		ret = ins_one(trans, root, radix);
 		if (ret) {
 			fprintf(stderr, "fill failed\n");
 			goto out;
 		}
 		if (i % 1000 == 0) {
-			ret = btrfs_commit_transaction(root, &super);
+			ret = btrfs_commit_transaction(trans, root, &super);
 			if (ret) {
 				fprintf(stderr, "fill commit failed\n");
 				return ret;
@@ -280,7 +290,8 @@ out:
 	return ret;
 }
 
-static int bulk_op(struct btrfs_root *root, struct radix_tree_root *radix)
+static int bulk_op(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		   struct radix_tree_root *radix)
 {
 	int ret;
 	int nr = rand() % 5000;
@@ -289,17 +300,18 @@ static int bulk_op(struct btrfs_root *root, struct radix_tree_root *radix)
 	/* do the bulk op much less frequently */
 	if (run_nr++ % 100)
 		return 0;
-	ret = empty_tree(root, radix, nr);
+	ret = empty_tree(trans, root, radix, nr);
 	if (ret)
 		return ret;
-	ret = fill_tree(root, radix, nr);
+	ret = fill_tree(trans, root, radix, nr);
 	if (ret)
 		return ret;
 	return 0;
 }
 
 
-int (*ops[])(struct btrfs_root *root, struct radix_tree_root *radix) =
+int (*ops[])(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct
+	     radix_tree_root *radix) =
 	{ ins_one, insert_dup, del_one, lookup_item,
 	  lookup_enoent, bulk_op };
 
@@ -330,11 +342,13 @@ int main(int ac, char **av)
 	int init_fill_count = 800000;
 	int err = 0;
 	int initial_only = 0;
+	struct btrfs_trans_handle *trans;
 	radix_tree_init();
 
 	printf("removing old tree\n");
 	unlink("dbfile");
 	root = open_ctree("dbfile", &super);
+	trans = btrfs_start_transaction(root, 1);
 
 	signal(SIGTERM, sigstopper);
 	signal(SIGINT, sigstopper);
@@ -353,7 +367,7 @@ int main(int ac, char **av)
 		}
 	}
 	printf("initial fill\n");
-	ret = fill_tree(root, &radix, init_fill_count);
+	ret = fill_tree(trans, root, &radix, init_fill_count);
 	printf("starting run\n");
 	if (ret) {
 		err = ret;
@@ -377,7 +391,7 @@ int main(int ac, char **av)
 			root = open_ctree("dbfile", &super);
 		}
 		while(count--) {
-			ret = ops[op](root, &radix);
+			ret = ops[op](trans, root, &radix);
 			if (ret) {
 				fprintf(stderr, "op %d failed %d:%d\n",
 					op, i, iterations);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 8d9457b5aef5..25ce07908ee3 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -9,6 +9,7 @@
 #include "radix-tree.h"
 #include "ctree.h"
 #include "disk-io.h"
+#include "transaction.h"
 
 static int allocated_blocks = 0;
 int cache_max = 10000;
@@ -107,7 +108,8 @@ struct btrfs_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr)
 	return buf;
 }
 
-int dirty_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf)
+int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		     struct btrfs_buffer *buf)
 {
 	if (!list_empty(&buf->dirty))
 		return 0;
@@ -116,7 +118,8 @@ int dirty_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf)
 	return 0;
 }
 
-int clean_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf)
+int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		     struct btrfs_buffer *buf)
 {
 	if (!list_empty(&buf->dirty)) {
 		list_del_init(&buf->dirty);
@@ -125,7 +128,8 @@ int clean_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf)
 	return 0;
 }
 
-int write_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf)
+int write_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		     struct btrfs_buffer *buf)
 {
 	u64 blocknr = buf->blocknr;
 	loff_t offset = blocknr * root->blocksize;
@@ -139,7 +143,8 @@ int write_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf)
 	return 0;
 }
 
-static int __commit_transaction(struct btrfs_root *root)
+static int __commit_transaction(struct btrfs_trans_handle *trans, struct
+				btrfs_root *root)
 {
 	struct btrfs_buffer *b;
 	int ret = 0;
@@ -147,7 +152,7 @@ static int __commit_transaction(struct btrfs_root *root)
 	while(!list_empty(&root->trans)) {
 		b = list_entry(root->trans.next, struct btrfs_buffer, dirty);
 		list_del_init(&b->dirty);
-		wret = write_tree_block(root, b);
+		wret = write_tree_block(trans, root, b);
 		if (wret)
 			ret = wret;
 		btrfs_block_release(root, b);
@@ -155,8 +160,9 @@ static int __commit_transaction(struct btrfs_root *root)
 	return ret;
 }
 
-static int commit_extent_and_tree_roots(struct btrfs_root *tree_root,
-					struct btrfs_root *extent_root)
+static int commit_extent_and_tree_roots(struct btrfs_trans_handle *trans,
+					struct btrfs_root *tree_root, struct
+					btrfs_root *extent_root)
 {
 	int ret;
 	u64 old_extent_block;
@@ -167,24 +173,24 @@ static int commit_extent_and_tree_roots(struct btrfs_root *tree_root,
 			break;
 		btrfs_set_root_blocknr(&extent_root->root_item,
 				       extent_root->node->blocknr);
-		ret = btrfs_update_root(tree_root,
+		ret = btrfs_update_root(trans, tree_root,
 					&extent_root->root_key,
 					&extent_root->root_item);
 		BUG_ON(ret);
 	}
-	__commit_transaction(extent_root);
-	__commit_transaction(tree_root);
+	__commit_transaction(trans, extent_root);
+	__commit_transaction(trans, tree_root);
 	return 0;
 }
 
-int btrfs_commit_transaction(struct btrfs_root *root,
-			     struct btrfs_super_block *s)
+int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct
+			     btrfs_root *root, struct btrfs_super_block *s)
 {
 	int ret = 0;
 	struct btrfs_buffer *snap = root->commit_root;
 	struct btrfs_key snap_key;
 
-	ret = __commit_transaction(root);
+	ret = __commit_transaction(trans, root);
 	BUG_ON(ret);
 
 	if (root->commit_root == root->node)
@@ -194,23 +200,24 @@ int btrfs_commit_transaction(struct btrfs_root *root,
 	root->root_key.offset++;
 
 	btrfs_set_root_blocknr(&root->root_item, root->node->blocknr);
-	ret = btrfs_insert_root(root->tree_root, &root->root_key,
+	ret = btrfs_insert_root(trans, root->tree_root, &root->root_key,
 				&root->root_item);
 	BUG_ON(ret);
 
-	ret = commit_extent_and_tree_roots(root->tree_root, root->extent_root);
+	ret = commit_extent_and_tree_roots(trans, root->tree_root,
+					   root->extent_root);
 	BUG_ON(ret);
 
-        write_ctree_super(root, s);
-	btrfs_finish_extent_commit(root->extent_root);
-	btrfs_finish_extent_commit(root->tree_root);
+	write_ctree_super(trans, root, s);
+	btrfs_finish_extent_commit(trans, root->extent_root);
+	btrfs_finish_extent_commit(trans, root->tree_root);
 
 	root->commit_root = root->node;
 	root->node->count++;
-	ret = btrfs_drop_snapshot(root, snap);
+	ret = btrfs_drop_snapshot(trans, root, snap);
 	BUG_ON(ret);
 
-	ret = btrfs_del_root(root->tree_root, &snap_key);
+	ret = btrfs_del_root(trans, root->tree_root, &snap_key);
 	BUG_ON(ret);
 
 	return ret;
@@ -312,7 +319,8 @@ struct btrfs_root *open_ctree(char *filename, struct btrfs_super_block *super)
 	return root;
 }
 
-int write_ctree_super(struct btrfs_root *root, struct btrfs_super_block *s)
+int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, struct btrfs_super_block *s)
 {
 	int ret;
 	btrfs_set_super_root(s, root->tree_root->node->blocknr);
@@ -338,10 +346,14 @@ static int drop_cache(struct btrfs_root *root)
 int close_ctree(struct btrfs_root *root, struct btrfs_super_block *s)
 {
 	int ret;
-	btrfs_commit_transaction(root, s);
-	ret = commit_extent_and_tree_roots(root->tree_root, root->extent_root);
+	struct btrfs_trans_handle *trans;
+
+	trans = root->running_transaction;
+	btrfs_commit_transaction(trans, root, s);
+	ret = commit_extent_and_tree_roots(trans, root->tree_root,
+					   root->extent_root);
 	BUG_ON(ret);
-	write_ctree_super(root, s);
+	write_ctree_super(trans, root, s);
 	drop_cache(root->extent_root);
 	drop_cache(root->tree_root);
 	drop_cache(root);
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 5771bb90acb2..24a9e77c8311 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -15,15 +15,19 @@ struct btrfs_buffer {
 
 struct btrfs_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr);
 struct btrfs_buffer *find_tree_block(struct btrfs_root *root, u64 blocknr);
-int write_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf);
-int dirty_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf);
-int clean_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf);
-int btrfs_commit_transaction(struct btrfs_root *root,
-			     struct btrfs_super_block *s);
+int write_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		     struct btrfs_buffer *buf);
+int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		     struct btrfs_buffer *buf);
+int clean_tree_block(struct btrfs_trans_handle *trans,
+		     struct btrfs_root *root, struct btrfs_buffer *buf);
+int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct btrfs_root
+			     *root, struct btrfs_super_block *s);
 struct btrfs_root *open_ctree(char *filename, struct btrfs_super_block *s);
 int close_ctree(struct btrfs_root *root, struct btrfs_super_block *s);
 void btrfs_block_release(struct btrfs_root *root, struct btrfs_buffer *buf);
-int write_ctree_super(struct btrfs_root *root, struct btrfs_super_block *s);
+int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		      struct btrfs_super_block *s);
 int mkfs(int fd, u64 num_blocks, u32 blocksize);
 
 #define BTRFS_SUPER_INFO_OFFSET (16 * 1024)
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 4a40282b45f7..c29b92d440e0 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -5,12 +5,15 @@
 #include "ctree.h"
 #include "disk-io.h"
 #include "print-tree.h"
+#include "transaction.h"
 
-static int find_free_extent(struct btrfs_root *orig_root, u64 num_blocks,
-			    u64 search_start, u64 search_end,
-			    struct btrfs_key *ins);
-static int finish_current_insert(struct btrfs_root *extent_root);
-static int run_pending(struct btrfs_root *extent_root);
+static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
+			    *orig_root, u64 num_blocks, u64 search_start, u64
+			    search_end, struct btrfs_key *ins);
+static int finish_current_insert(struct btrfs_trans_handle *trans, struct
+				 btrfs_root *extent_root);
+static int run_pending(struct btrfs_trans_handle *trans, struct btrfs_root
+		       *extent_root);
 
 /*
  * pending extents are blocks that we're trying to allocate in the extent
@@ -21,7 +24,8 @@ static int run_pending(struct btrfs_root *extent_root);
  */
 #define CTREE_EXTENT_PENDING_DEL 0
 
-static int inc_block_ref(struct btrfs_root *root, u64 blocknr)
+static int inc_block_ref(struct btrfs_trans_handle *trans, struct btrfs_root
+			 *root, u64 blocknr)
 {
 	struct btrfs_path path;
 	int ret;
@@ -31,13 +35,13 @@ static int inc_block_ref(struct btrfs_root *root, u64 blocknr)
 	struct btrfs_key ins;
 	u32 refs;
 
-	find_free_extent(root->extent_root, 0, 0, (u64)-1, &ins);
+	find_free_extent(trans, root->extent_root, 0, 0, (u64)-1, &ins);
 	btrfs_init_path(&path);
 	key.objectid = blocknr;
 	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 	key.offset = 1;
-	ret = btrfs_search_slot(root->extent_root, &key, &path, 0, 1);
+	ret = btrfs_search_slot(trans, root->extent_root, &key, &path, 0, 1);
 	if (ret != 0)
 		BUG();
 	BUG_ON(ret != 0);
@@ -48,12 +52,13 @@ static int inc_block_ref(struct btrfs_root *root, u64 blocknr)
 
 	BUG_ON(list_empty(&path.nodes[0]->dirty));
 	btrfs_release_path(root->extent_root, &path);
-	finish_current_insert(root->extent_root);
-	run_pending(root->extent_root);
+	finish_current_insert(trans, root->extent_root);
+	run_pending(trans, root->extent_root);
 	return 0;
 }
 
-static int lookup_block_ref(struct btrfs_root *root, u64 blocknr, u32 *refs)
+static int lookup_block_ref(struct btrfs_trans_handle *trans, struct btrfs_root
+			    *root, u64 blocknr, u32 *refs)
 {
 	struct btrfs_path path;
 	int ret;
@@ -65,7 +70,7 @@ static int lookup_block_ref(struct btrfs_root *root, u64 blocknr, u32 *refs)
 	key.offset = 1;
 	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
-	ret = btrfs_search_slot(root->extent_root, &key, &path, 0, 0);
+	ret = btrfs_search_slot(trans, root->extent_root, &key, &path, 0, 0);
 	if (ret != 0)
 		BUG();
 	l = &path.nodes[0]->leaf;
@@ -75,7 +80,8 @@ static int lookup_block_ref(struct btrfs_root *root, u64 blocknr, u32 *refs)
 	return 0;
 }
 
-int btrfs_inc_ref(struct btrfs_root *root, struct btrfs_buffer *buf)
+int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		  struct btrfs_buffer *buf)
 {
 	u64 blocknr;
 	int i;
@@ -87,12 +93,13 @@ int btrfs_inc_ref(struct btrfs_root *root, struct btrfs_buffer *buf)
 
 	for (i = 0; i < btrfs_header_nritems(&buf->node.header); i++) {
 		blocknr = btrfs_node_blockptr(&buf->node, i);
-		inc_block_ref(root, blocknr);
+		inc_block_ref(trans, root, blocknr);
 	}
 	return 0;
 }
 
-int btrfs_finish_extent_commit(struct btrfs_root *root)
+int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct
+			       btrfs_root *root)
 {
 	unsigned long gang[8];
 	u64 first = 0;
@@ -116,7 +123,8 @@ int btrfs_finish_extent_commit(struct btrfs_root *root)
 	return 0;
 }
 
-static int finish_current_insert(struct btrfs_root *extent_root)
+static int finish_current_insert(struct btrfs_trans_handle *trans, struct
+				 btrfs_root *extent_root)
 {
 	struct btrfs_key ins;
 	struct btrfs_extent_item extent_item;
@@ -132,8 +140,8 @@ static int finish_current_insert(struct btrfs_root *extent_root)
 
 	for (i = 0; i < extent_root->current_insert.flags; i++) {
 		ins.objectid = extent_root->current_insert.objectid + i;
-		ret = btrfs_insert_item(extent_root, &ins, &extent_item,
-				  sizeof(extent_item));
+		ret = btrfs_insert_item(trans, extent_root, &ins, &extent_item,
+					sizeof(extent_item));
 		BUG_ON(ret);
 	}
 	extent_root->current_insert.offset = 0;
@@ -143,8 +151,8 @@ static int finish_current_insert(struct btrfs_root *extent_root)
 /*
  * remove an extent from the root, returns 0 on success
  */
-static int __free_extent(struct btrfs_root *root, u64 blocknr, u64 num_blocks,
-			 int pin)
+static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
+			 *root, u64 blocknr, u64 num_blocks, int pin)
 {
 	struct btrfs_path path;
 	struct btrfs_key key;
@@ -160,9 +168,9 @@ static int __free_extent(struct btrfs_root *root, u64 blocknr, u64 num_blocks,
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 	key.offset = num_blocks;
 
-	find_free_extent(root, 0, 0, (u64)-1, &ins);
+	find_free_extent(trans, root, 0, 0, (u64)-1, &ins);
 	btrfs_init_path(&path);
-	ret = btrfs_search_slot(extent_root, &key, &path, -1, 1);
+	ret = btrfs_search_slot(trans, extent_root, &key, &path, -1, 1);
 	if (ret) {
 		printf("failed to find %Lu\n", key.objectid);
 		btrfs_print_tree(extent_root, extent_root->node);
@@ -183,14 +191,14 @@ static int __free_extent(struct btrfs_root *root, u64 blocknr, u64 num_blocks,
 			BUG_ON(err);
 			radix_tree_preload_end();
 		}
-		ret = btrfs_del_item(extent_root, &path);
+		ret = btrfs_del_item(trans, extent_root, &path);
 		if (!pin && extent_root->last_insert.objectid > blocknr)
 			extent_root->last_insert.objectid = blocknr;
 		if (ret)
 			BUG();
 	}
 	btrfs_release_path(extent_root, &path);
-	finish_current_insert(extent_root);
+	finish_current_insert(trans, extent_root);
 	return ret;
 }
 
@@ -198,7 +206,8 @@ static int __free_extent(struct btrfs_root *root, u64 blocknr, u64 num_blocks,
  * find all the blocks marked as pending in the radix tree and remove
  * them from the extent map
  */
-static int del_pending_extents(struct btrfs_root *extent_root)
+static int del_pending_extents(struct btrfs_trans_handle *trans, struct
+			       btrfs_root *extent_root)
 {
 	int ret;
 	struct btrfs_buffer *gang[4];
@@ -212,7 +221,7 @@ static int del_pending_extents(struct btrfs_root *extent_root)
 		if (!ret)
 			break;
 		for (i = 0; i < ret; i++) {
-			ret = __free_extent(extent_root,
+			ret = __free_extent(trans, extent_root,
 					    gang[i]->blocknr, 1, 1);
 			radix_tree_tag_clear(&extent_root->cache_radix,
 						gang[i]->blocknr,
@@ -223,11 +232,12 @@ static int del_pending_extents(struct btrfs_root *extent_root)
 	return 0;
 }
 
-static int run_pending(struct btrfs_root *extent_root)
+static int run_pending(struct btrfs_trans_handle *trans, struct btrfs_root
+		       *extent_root)
 {
 	while(radix_tree_tagged(&extent_root->cache_radix,
 			        CTREE_EXTENT_PENDING_DEL))
-		del_pending_extents(extent_root);
+		del_pending_extents(trans, extent_root);
 	return 0;
 }
 
@@ -235,8 +245,8 @@ static int run_pending(struct btrfs_root *extent_root)
 /*
  * remove an extent from the root, returns 0 on success
  */
-int btrfs_free_extent(struct btrfs_root *root, u64 blocknr, u64 num_blocks,
-		      int pin)
+int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, u64 blocknr, u64 num_blocks, int pin)
 {
 	struct btrfs_root *extent_root = root->extent_root;
 	struct btrfs_buffer *t;
@@ -249,8 +259,8 @@ int btrfs_free_extent(struct btrfs_root *root, u64 blocknr, u64 num_blocks,
 				   CTREE_EXTENT_PENDING_DEL);
 		return 0;
 	}
-	ret = __free_extent(root, blocknr, num_blocks, pin);
-	pending_ret = run_pending(root->extent_root);
+	ret = __free_extent(trans, root, blocknr, num_blocks, pin);
+	pending_ret = run_pending(trans, root->extent_root);
 	return ret ? ret : pending_ret;
 }
 
@@ -262,9 +272,9 @@ int btrfs_free_extent(struct btrfs_root *root, u64 blocknr, u64 num_blocks,
  * ins->offset == number of blocks
  * Any available blocks before search_start are skipped.
  */
-static int find_free_extent(struct btrfs_root *orig_root, u64 num_blocks,
-			    u64 search_start, u64 search_end,
-			    struct btrfs_key *ins)
+static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
+			    *orig_root, u64 num_blocks, u64 search_start, u64
+			    search_end, struct btrfs_key *ins)
 {
 	struct btrfs_path path;
 	struct btrfs_key key;
@@ -290,7 +300,7 @@ check_failed:
 	ins->objectid = search_start;
 	ins->offset = 0;
 	start_found = 0;
-	ret = btrfs_search_slot(root, ins, &path, 0, 0);
+	ret = btrfs_search_slot(trans, root, ins, &path, 0, 0);
 	if (ret < 0)
 		goto error;
 
@@ -367,9 +377,9 @@ error:
  *
  * returns 0 if everything worked, non-zero otherwise.
  */
-static int alloc_extent(struct btrfs_root *root, u64 num_blocks,
-			u64 search_start, u64 search_end, u64 owner,
-			struct btrfs_key *ins)
+static int alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_root
+			*root, u64 num_blocks, u64 search_start, u64
+			search_end, u64 owner, struct btrfs_key *ins)
 {
 	int ret;
 	int pending_ret;
@@ -389,16 +399,16 @@ static int alloc_extent(struct btrfs_root *root, u64 num_blocks,
 				extent_root->current_insert.flags++;
 		return 0;
 	}
-	ret = find_free_extent(root, num_blocks, search_start,
+	ret = find_free_extent(trans, root, num_blocks, search_start,
 			       search_end, ins);
 	if (ret)
 		return ret;
 
-	ret = btrfs_insert_item(extent_root, ins, &extent_item,
-			  sizeof(extent_item));
+	ret = btrfs_insert_item(trans, extent_root, ins, &extent_item,
+				sizeof(extent_item));
 
-	finish_current_insert(extent_root);
-	pending_ret = run_pending(extent_root);
+	finish_current_insert(trans, extent_root);
+	pending_ret = run_pending(trans, extent_root);
 	if (ret)
 		return ret;
 	if (pending_ret)
@@ -410,13 +420,14 @@ static int alloc_extent(struct btrfs_root *root, u64 num_blocks,
  * helper function to allocate a block for a given tree
  * returns the tree buffer or NULL.
  */
-struct btrfs_buffer *btrfs_alloc_free_block(struct btrfs_root *root)
+struct btrfs_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
+					    struct btrfs_root *root)
 {
 	struct btrfs_key ins;
 	int ret;
 	struct btrfs_buffer *buf;
 
-	ret = alloc_extent(root, 1, 0, (unsigned long)-1,
+	ret = alloc_extent(trans, root, 1, 0, (unsigned long)-1,
 			   btrfs_header_parentid(&root->node->node.header),
 			   &ins);
 	if (ret) {
@@ -424,7 +435,7 @@ struct btrfs_buffer *btrfs_alloc_free_block(struct btrfs_root *root)
 		return NULL;
 	}
 	buf = find_tree_block(root, ins.objectid);
-	dirty_tree_block(root, buf);
+	dirty_tree_block(trans, root, buf);
 	return buf;
 }
 
@@ -432,8 +443,8 @@ struct btrfs_buffer *btrfs_alloc_free_block(struct btrfs_root *root)
  * helper function for drop_snapshot, this walks down the tree dropping ref
  * counts as it goes.
  */
-static int walk_down_tree(struct btrfs_root *root,
-			  struct btrfs_path *path, int *level)
+static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
+			  *root, struct btrfs_path *path, int *level)
 {
 	struct btrfs_buffer *next;
 	struct btrfs_buffer *cur;
@@ -441,7 +452,8 @@ static int walk_down_tree(struct btrfs_root *root,
 	int ret;
 	u32 refs;
 
-	ret = lookup_block_ref(root, path->nodes[*level]->blocknr, &refs);
+	ret = lookup_block_ref(trans, root, path->nodes[*level]->blocknr,
+			       &refs);
 	BUG_ON(ret);
 	if (refs > 1)
 		goto out;
@@ -454,10 +466,10 @@ static int walk_down_tree(struct btrfs_root *root,
 		    btrfs_header_nritems(&cur->node.header))
 			break;
 		blocknr = btrfs_node_blockptr(&cur->node, path->slots[*level]);
-		ret = lookup_block_ref(root, blocknr, &refs);
+		ret = lookup_block_ref(trans, root, blocknr, &refs);
 		if (refs != 1 || *level == 1) {
 			path->slots[*level]++;
-			ret = btrfs_free_extent(root, blocknr, 1, 1);
+			ret = btrfs_free_extent(trans, root, blocknr, 1, 1);
 			BUG_ON(ret);
 			continue;
 		}
@@ -470,7 +482,8 @@ static int walk_down_tree(struct btrfs_root *root,
 		path->slots[*level] = 0;
 	}
 out:
-	ret = btrfs_free_extent(root, path->nodes[*level]->blocknr, 1, 1);
+	ret = btrfs_free_extent(trans, root, path->nodes[*level]->blocknr, 1,
+				1);
 	btrfs_block_release(root, path->nodes[*level]);
 	path->nodes[*level] = NULL;
 	*level += 1;
@@ -483,8 +496,8 @@ out:
  * to find the first node higher up where we haven't yet gone through
  * all the slots
  */
-static int walk_up_tree(struct btrfs_root *root, struct btrfs_path *path,
-			int *level)
+static int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root
+			*root, struct btrfs_path *path, int *level)
 {
 	int i;
 	int slot;
@@ -497,8 +510,9 @@ static int walk_up_tree(struct btrfs_root *root, struct btrfs_path *path,
 			*level = i;
 			return 0;
 		} else {
-			ret = btrfs_free_extent(root,
-					  path->nodes[*level]->blocknr, 1, 1);
+			ret = btrfs_free_extent(trans, root,
+						path->nodes[*level]->blocknr,
+						1, 1);
 			btrfs_block_release(root, path->nodes[*level]);
 			path->nodes[*level] = NULL;
 			*level = i + 1;
@@ -513,7 +527,8 @@ static int walk_up_tree(struct btrfs_root *root, struct btrfs_path *path,
  * the tree freeing any blocks that have a ref count of zero after being
  * decremented.
  */
-int btrfs_drop_snapshot(struct btrfs_root *root, struct btrfs_buffer *snap)
+int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
+			*root, struct btrfs_buffer *snap)
 {
 	int ret = 0;
 	int wret;
@@ -529,13 +544,13 @@ int btrfs_drop_snapshot(struct btrfs_root *root, struct btrfs_buffer *snap)
 	path.nodes[level] = snap;
 	path.slots[level] = 0;
 	while(1) {
-		wret = walk_down_tree(root, &path, &level);
+		wret = walk_down_tree(trans, root, &path, &level);
 		if (wret > 0)
 			break;
 		if (wret < 0)
 			ret = wret;
 
-		wret = walk_up_tree(root, &path, &level);
+		wret = walk_up_tree(trans, root, &path, &level);
 		if (wret > 0)
 			break;
 		if (wret < 0)
diff --git a/fs/btrfs/inode-item.c b/fs/btrfs/inode-item.c
index 4908f9afdf86..7caeb11e8756 100644
--- a/fs/btrfs/inode-item.c
+++ b/fs/btrfs/inode-item.c
@@ -4,9 +4,11 @@
 #include "radix-tree.h"
 #include "ctree.h"
 #include "disk-io.h"
+#include "transaction.h"
 
-int btrfs_insert_inode(struct btrfs_root *root, u64 objectid,
-		       struct btrfs_inode_item *inode_item)
+int btrfs_insert_inode(struct btrfs_trans_handle *trans, struct btrfs_root
+		       *root, u64 objectid, struct btrfs_inode_item
+		       *inode_item)
 {
 	struct btrfs_path path;
 	struct btrfs_key key;
@@ -17,13 +19,14 @@ int btrfs_insert_inode(struct btrfs_root *root, u64 objectid,
 	key.offset = 0;
 
 	btrfs_init_path(&path);
-	ret = btrfs_insert_item(root, &key, inode_item, sizeof(*inode_item));
+	ret = btrfs_insert_item(trans, root, &key, inode_item,
+				sizeof(*inode_item));
 	btrfs_release_path(root, &path);
 	return ret;
 }
 
-int btrfs_lookup_inode(struct btrfs_root *root, struct btrfs_path *path,
-			u64 objectid, int mod)
+int btrfs_lookup_inode(struct btrfs_trans_handle *trans, struct btrfs_root
+		       *root, struct btrfs_path *path, u64 objectid, int mod)
 {
 	struct btrfs_key key;
 	int ins_len = mod < 0 ? -1 : 0;
@@ -33,5 +36,5 @@ int btrfs_lookup_inode(struct btrfs_root *root, struct btrfs_path *path,
 	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
 	key.offset = 0;
-	return btrfs_search_slot(root, &key, path, ins_len, cow);
+	return btrfs_search_slot(trans, root, &key, path, ins_len, cow);
 }
diff --git a/fs/btrfs/print-tree.h b/fs/btrfs/print-tree.h
index 9236abda2a74..0882ca904eca 100644
--- a/fs/btrfs/print-tree.h
+++ b/fs/btrfs/print-tree.h
@@ -1,3 +1,5 @@
-
+#ifndef __PRINT_TREE_
+#define __PRINT_TREE_
 void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l);
 void btrfs_print_tree(struct btrfs_root *root, struct btrfs_buffer *t);
+#endif
diff --git a/fs/btrfs/quick-test.c b/fs/btrfs/quick-test.c
index 3a2571617121..d676577185d5 100644
--- a/fs/btrfs/quick-test.c
+++ b/fs/btrfs/quick-test.c
@@ -5,6 +5,7 @@
 #include "ctree.h"
 #include "disk-io.h"
 #include "print-tree.h"
+#include "transaction.h"
 
 /* for testing only */
 int next_key(int i, int max_key) {
@@ -25,10 +26,12 @@ int main(int ac, char **av) {
 	struct btrfs_path path;
 	struct btrfs_super_block super;
 	struct btrfs_root *root;
+	struct btrfs_trans_handle *trans;
 
 	radix_tree_init();
 
 	root = open_ctree("dbfile", &super);
+	trans = btrfs_start_transaction(root, 1);
 	srand(55);
 	ins.flags = 0;
 	btrfs_set_key_type(&ins, BTRFS_STRING_ITEM_KEY);
@@ -41,12 +44,12 @@ int main(int ac, char **av) {
 			fprintf(stderr, "insert %d:%d\n", num, i);
 		ins.objectid = num;
 		ins.offset = 0;
-		ret = btrfs_insert_item(root, &ins, buf, strlen(buf));
+		ret = btrfs_insert_item(trans, root, &ins, buf, strlen(buf));
 		if (!ret)
 			tree_size++;
 		free(buf);
 		if (i == run_size - 5) {
-			btrfs_commit_transaction(root, &super);
+			btrfs_commit_transaction(trans, root, &super);
 		}
 
 	}
@@ -61,7 +64,7 @@ int main(int ac, char **av) {
 		btrfs_init_path(&path);
 		if (i % 10000 == 0)
 			fprintf(stderr, "search %d:%d\n", num, i);
-		ret = btrfs_search_slot(root, &ins, &path, 0, 0);
+		ret = btrfs_search_slot(trans, root, &ins, &path, 0, 0);
 		if (ret) {
 			btrfs_print_tree(root, root->node);
 			printf("unable to find %d\n", num);
@@ -83,11 +86,11 @@ int main(int ac, char **av) {
 		num = next_key(i, max_key);
 		ins.objectid = num;
 		btrfs_init_path(&path);
-		ret = btrfs_search_slot(root, &ins, &path, -1, 1);
+		ret = btrfs_search_slot(trans, root, &ins, &path, -1, 1);
 		if (!ret) {
 			if (i % 10000 == 0)
 				fprintf(stderr, "del %d:%d\n", num, i);
-			ret = btrfs_del_item(root, &path);
+			ret = btrfs_del_item(trans, root, &path);
 			if (ret != 0)
 				BUG();
 			tree_size--;
@@ -104,7 +107,7 @@ int main(int ac, char **av) {
 		ins.objectid = num;
 		if (i % 10000 == 0)
 			fprintf(stderr, "insert %d:%d\n", num, i);
-		ret = btrfs_insert_item(root, &ins, buf, strlen(buf));
+		ret = btrfs_insert_item(trans, root, &ins, buf, strlen(buf));
 		if (!ret)
 			tree_size++;
 		free(buf);
@@ -119,7 +122,7 @@ int main(int ac, char **av) {
 		btrfs_init_path(&path);
 		if (i % 10000 == 0)
 			fprintf(stderr, "search %d:%d\n", num, i);
-		ret = btrfs_search_slot(root, &ins, &path, 0, 0);
+		ret = btrfs_search_slot(trans, root, &ins, &path, 0, 0);
 		if (ret) {
 			btrfs_print_tree(root, root->node);
 			printf("unable to find %d\n", num);
@@ -134,7 +137,7 @@ int main(int ac, char **av) {
 		int slot;
 		ins.objectid = (u64)-1;
 		btrfs_init_path(&path);
-		ret = btrfs_search_slot(root, &ins, &path, -1, 1);
+		ret = btrfs_search_slot(trans, root, &ins, &path, -1, 1);
 		if (ret == 0)
 			BUG();
 
@@ -150,7 +153,7 @@ int main(int ac, char **av) {
 			btrfs_disk_key_to_cpu(&last, &leaf->items[slot].key);
 			if (tree_size % 10000 == 0)
 				printf("big del %d:%d\n", tree_size, i);
-			ret = btrfs_del_item(root, &path);
+			ret = btrfs_del_item(trans, root, &path);
 			if (ret != 0) {
 				printf("del_item returned %d\n", ret);
 				BUG();
@@ -165,7 +168,7 @@ int main(int ac, char **av) {
 	printf("map before commit\n");
 	btrfs_print_tree(root->extent_root, root->extent_root->node);
 	*/
-	btrfs_commit_transaction(root, &super);
+	btrfs_commit_transaction(trans, root, &super);
 	printf("tree size is now %d\n", tree_size);
 	printf("root %p commit root %p\n", root->node, root->commit_root);
 	printf("map tree\n");
diff --git a/fs/btrfs/random-test.c b/fs/btrfs/random-test.c
index f05135f13f9a..3a38ae7a886d 100644
--- a/fs/btrfs/random-test.c
+++ b/fs/btrfs/random-test.c
@@ -6,6 +6,7 @@
 #include "ctree.h"
 #include "disk-io.h"
 #include "print-tree.h"
+#include "transaction.h"
 
 int keep_running = 1;
 struct btrfs_super_block super;
@@ -37,7 +38,8 @@ again:
 	return 0;
 }
 
-static int ins_one(struct btrfs_root *root, struct radix_tree_root *radix)
+static int ins_one(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		   struct radix_tree_root *radix)
 {
 	struct btrfs_path path;
 	struct btrfs_key key;
@@ -47,7 +49,7 @@ static int ins_one(struct btrfs_root *root, struct radix_tree_root *radix)
 	btrfs_init_path(&path);
 	ret = setup_key(radix, &key, 0);
 	sprintf(buf, "str-%Lu\n", key.objectid);
-	ret = btrfs_insert_item(root, &key, buf, strlen(buf));
+	ret = btrfs_insert_item(trans, root, &key, buf, strlen(buf));
 	if (ret)
 		goto error;
 	oid = (unsigned long)key.objectid;
@@ -62,7 +64,8 @@ error:
 	return -1;
 }
 
-static int insert_dup(struct btrfs_root *root, struct radix_tree_root *radix)
+static int insert_dup(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, struct radix_tree_root *radix)
 {
 	struct btrfs_path path;
 	struct btrfs_key key;
@@ -73,7 +76,7 @@ static int insert_dup(struct btrfs_root *root, struct radix_tree_root *radix)
 	if (ret < 0)
 		return 0;
 	sprintf(buf, "str-%Lu\n", key.objectid);
-	ret = btrfs_insert_item(root, &key, buf, strlen(buf));
+	ret = btrfs_insert_item(trans, root, &key, buf, strlen(buf));
 	if (ret != -EEXIST) {
 		printf("insert on %Lu gave us %d\n", key.objectid, ret);
 		return 1;
@@ -81,7 +84,8 @@ static int insert_dup(struct btrfs_root *root, struct radix_tree_root *radix)
 	return 0;
 }
 
-static int del_one(struct btrfs_root *root, struct radix_tree_root *radix)
+static int del_one(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		   struct radix_tree_root *radix)
 {
 	struct btrfs_path path;
 	struct btrfs_key key;
@@ -91,10 +95,10 @@ static int del_one(struct btrfs_root *root, struct radix_tree_root *radix)
 	ret = setup_key(radix, &key, 1);
 	if (ret < 0)
 		return 0;
-	ret = btrfs_search_slot(root, &key, &path, -1, 1);
+	ret = btrfs_search_slot(trans, root, &key, &path, -1, 1);
 	if (ret)
 		goto error;
-	ret = btrfs_del_item(root, &path);
+	ret = btrfs_del_item(trans, root, &path);
 	btrfs_release_path(root, &path);
 	if (ret != 0)
 		goto error;
@@ -107,7 +111,8 @@ error:
 	return -1;
 }
 
-static int lookup_item(struct btrfs_root *root, struct radix_tree_root *radix)
+static int lookup_item(struct btrfs_trans_handle *trans, struct btrfs_root
+		       *root, struct radix_tree_root *radix)
 {
 	struct btrfs_path path;
 	struct btrfs_key key;
@@ -116,7 +121,7 @@ static int lookup_item(struct btrfs_root *root, struct radix_tree_root *radix)
 	ret = setup_key(radix, &key, 1);
 	if (ret < 0)
 		return 0;
-	ret = btrfs_search_slot(root, &key, &path, 0, 1);
+	ret = btrfs_search_slot(trans, root, &key, &path, 0, 1);
 	btrfs_release_path(root, &path);
 	if (ret)
 		goto error;
@@ -126,7 +131,8 @@ error:
 	return -1;
 }
 
-static int lookup_enoent(struct btrfs_root *root, struct radix_tree_root *radix)
+static int lookup_enoent(struct btrfs_trans_handle *trans, struct btrfs_root
+			 *root, struct radix_tree_root *radix)
 {
 	struct btrfs_path path;
 	struct btrfs_key key;
@@ -135,7 +141,7 @@ static int lookup_enoent(struct btrfs_root *root, struct radix_tree_root *radix)
 	ret = setup_key(radix, &key, 0);
 	if (ret < 0)
 		return ret;
-	ret = btrfs_search_slot(root, &key, &path, 0, 0);
+	ret = btrfs_search_slot(trans, root, &key, &path, 0, 0);
 	btrfs_release_path(root, &path);
 	if (ret <= 0)
 		goto error;
@@ -145,8 +151,8 @@ error:
 	return -1;
 }
 
-static int empty_tree(struct btrfs_root *root, struct radix_tree_root *radix,
-		      int nr)
+static int empty_tree(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, struct radix_tree_root *radix, int nr)
 {
 	struct btrfs_path path;
 	struct btrfs_key key;
@@ -162,7 +168,7 @@ static int empty_tree(struct btrfs_root *root, struct radix_tree_root *radix,
 	key.objectid = (unsigned long)-1;
 	while(nr-- >= 0) {
 		btrfs_init_path(&path);
-		ret = btrfs_search_slot(root, &key, &path, -1, 1);
+		ret = btrfs_search_slot(trans, root, &key, &path, -1, 1);
 		if (ret < 0) {
 			btrfs_release_path(root, &path);
 			return ret;
@@ -177,7 +183,7 @@ static int empty_tree(struct btrfs_root *root, struct radix_tree_root *radix,
 		slot = path.slots[0];
 		found = btrfs_disk_key_objectid(
 					&path.nodes[0]->leaf.items[slot].key);
-		ret = btrfs_del_item(root, &path);
+		ret = btrfs_del_item(trans, root, &path);
 		count++;
 		if (ret) {
 			fprintf(stderr,
@@ -198,19 +204,19 @@ error:
 	return -1;
 }
 
-static int fill_tree(struct btrfs_root *root, struct radix_tree_root *radix,
-		     int count)
+static int fill_tree(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		     struct radix_tree_root *radix, int count)
 {
 	int i;
 	int ret = 0;
 	for (i = 0; i < count; i++) {
-		ret = ins_one(root, radix);
+		ret = ins_one(trans, root, radix);
 		if (ret) {
 			fprintf(stderr, "fill failed\n");
 			goto out;
 		}
 		if (i % 1000 == 0) {
-			ret = btrfs_commit_transaction(root, &super);
+			ret = btrfs_commit_transaction(trans, root, &super);
 			if (ret) {
 				fprintf(stderr, "fill commit failed\n");
 				return ret;
@@ -226,7 +232,8 @@ out:
 	return ret;
 }
 
-static int bulk_op(struct btrfs_root *root, struct radix_tree_root *radix)
+static int bulk_op(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		   struct radix_tree_root *radix)
 {
 	int ret;
 	int nr = rand() % 5000;
@@ -235,17 +242,18 @@ static int bulk_op(struct btrfs_root *root, struct radix_tree_root *radix)
 	/* do the bulk op much less frequently */
 	if (run_nr++ % 100)
 		return 0;
-	ret = empty_tree(root, radix, nr);
+	ret = empty_tree(trans, root, radix, nr);
 	if (ret)
 		return ret;
-	ret = fill_tree(root, radix, nr);
+	ret = fill_tree(trans, root, radix, nr);
 	if (ret)
 		return ret;
 	return 0;
 }
 
 
-int (*ops[])(struct btrfs_root *root, struct radix_tree_root *radix) =
+int (*ops[])(struct btrfs_trans_handle *,
+	     struct btrfs_root *root, struct radix_tree_root *radix) =
 	{ ins_one, insert_dup, del_one, lookup_item,
 	  lookup_enoent, bulk_op };
 
@@ -264,7 +272,7 @@ static int fill_radix(struct btrfs_root *root, struct radix_tree_root *radix)
 	key.objectid = (unsigned long)-1;
 	while(1) {
 		btrfs_init_path(&path);
-		ret = btrfs_search_slot(root, &key, &path, 0, 0);
+		ret = btrfs_search_slot(NULL, root, &key, &path, 0, 0);
 		if (ret < 0) {
 			btrfs_release_path(root, &path);
 			return ret;
@@ -325,6 +333,7 @@ int main(int ac, char **av)
 	int init_fill_count = 800000;
 	int err = 0;
 	int initial_only = 0;
+	struct btrfs_trans_handle *trans;
 	radix_tree_init();
 	root = open_ctree("dbfile", &super);
 	fill_radix(root, &radix);
@@ -346,7 +355,8 @@ int main(int ac, char **av)
 		}
 	}
 	printf("initial fill\n");
-	ret = fill_tree(root, &radix, init_fill_count);
+	trans = btrfs_start_transaction(root, 1);
+	ret = fill_tree(trans, root, &radix, init_fill_count);
 	printf("starting run\n");
 	if (ret) {
 		err = ret;
@@ -370,7 +380,7 @@ int main(int ac, char **av)
 			root = open_ctree("dbfile", &super);
 		}
 		while(count--) {
-			ret = ops[op](root, &radix);
+			ret = ops[op](trans, root, &radix);
 			if (ret) {
 				fprintf(stderr, "op %d failed %d:%d\n",
 					op, i, iterations);
diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c
index a24b47274037..9cccecc0f431 100644
--- a/fs/btrfs/root-tree.c
+++ b/fs/btrfs/root-tree.c
@@ -20,7 +20,7 @@ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid,
 	search_key.offset = (u32)-1;
 
 	btrfs_init_path(&path);
-	ret = btrfs_search_slot(root, &search_key, &path, 0, 0);
+	ret = btrfs_search_slot(NULL, root, &search_key, &path, 0, 0);
 	if (ret < 0)
 		goto out;
 	BUG_ON(ret == 0);
@@ -40,8 +40,9 @@ out:
 	return ret;
 }
 
-int btrfs_update_root(struct btrfs_root *root, struct btrfs_key *key,
-		      struct btrfs_root_item *item)
+int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, struct btrfs_key *key, struct btrfs_root_item
+		      *item)
 {
 	struct btrfs_path path;
 	struct btrfs_leaf *l;
@@ -49,7 +50,7 @@ int btrfs_update_root(struct btrfs_root *root, struct btrfs_key *key,
 	int slot;
 
 	btrfs_init_path(&path);
-	ret = btrfs_search_slot(root, key, &path, 0, 1);
+	ret = btrfs_search_slot(trans, root, key, &path, 0, 1);
 	if (ret < 0)
 		goto out;
 	BUG_ON(ret != 0);
@@ -62,26 +63,28 @@ out:
 	return ret;
 }
 
-int btrfs_insert_root(struct btrfs_root *root, struct btrfs_key *key,
-		      struct btrfs_root_item *item)
+int btrfs_insert_root(struct btrfs_trans_handle *trans, struct btrfs_root
+		      *root, struct btrfs_key *key, struct btrfs_root_item
+		      *item)
 {
 	int ret;
-	ret = btrfs_insert_item(root, key, item, sizeof(*item));
+	ret = btrfs_insert_item(trans, root, key, item, sizeof(*item));
 	BUG_ON(ret);
 	return ret;
 }
 
-int btrfs_del_root(struct btrfs_root *root, struct btrfs_key *key)
+int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		   struct btrfs_key *key)
 {
 	struct btrfs_path path;
 	int ret;
 
 	btrfs_init_path(&path);
-	ret = btrfs_search_slot(root, key, &path, -1, 1);
+	ret = btrfs_search_slot(trans, root, key, &path, -1, 1);
 	if (ret < 0)
 		goto out;
 	BUG_ON(ret != 0);
-	ret = btrfs_del_item(root, &path);
+	ret = btrfs_del_item(trans, root, &path);
 out:
 	btrfs_release_path(root, &path);
 	return ret;
diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h
new file mode 100644
index 000000000000..3adb6e69fe43
--- /dev/null
+++ b/fs/btrfs/transaction.h
@@ -0,0 +1,27 @@
+#ifndef __TRANSACTION__
+#define __TRANSACTION__
+
+struct btrfs_trans_handle {
+	u64 transid;
+	unsigned long blocks_reserved;
+	unsigned long blocks_used;
+};
+
+static inline struct btrfs_trans_handle *
+btrfs_start_transaction(struct btrfs_root *root, int num_blocks)
+{
+	struct btrfs_trans_handle *h = malloc(sizeof(*h));
+	h->transid = root->root_key.offset;
+	h->blocks_reserved = num_blocks;
+	h->blocks_used = 0;
+	return h;
+}
+
+static inline void btrfs_free_transaction(struct btrfs_root *root,
+					  struct btrfs_trans_handle *handle)
+{
+	memset(handle, 0, sizeof(*handle));
+	free(handle);
+}
+
+#endif
-- 
cgit v1.2.3


From 2e635a278354a1a7951e16cfea4c247d6d0e7c99 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Wed, 21 Mar 2007 11:12:56 -0400
Subject: Btrfs: initial move to kernel module land

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/Makefile      |  50 +--
 fs/btrfs/ctree.c       |   6 +-
 fs/btrfs/ctree.h       |  15 +-
 fs/btrfs/debug-tree.c  |  38 ---
 fs/btrfs/dir-item.c    |  12 +-
 fs/btrfs/dir-test.c    | 494 -----------------------------
 fs/btrfs/disk-io.c     |  17 +-
 fs/btrfs/disk-io.h     |   1 +
 fs/btrfs/extent-tree.c |  10 +-
 fs/btrfs/file-item.c   |   6 +-
 fs/btrfs/hash.c        |   1 -
 fs/btrfs/hasher.c      |  23 --
 fs/btrfs/inode-item.c  |   5 +-
 fs/btrfs/inode-map.c   |   5 +-
 fs/btrfs/kerncompat.h  |  96 ------
 fs/btrfs/list.h        | 418 -------------------------
 fs/btrfs/mkfs.c        | 255 ---------------
 fs/btrfs/print-tree.c  |  30 +-
 fs/btrfs/quick-test.c  | 179 -----------
 fs/btrfs/radix-tree.c  | 836 -------------------------------------------------
 fs/btrfs/radix-tree.h  |  73 -----
 fs/btrfs/random-test.c | 405 ------------------------
 fs/btrfs/root-tree.c   |   5 +-
 fs/btrfs/super.c       | 205 ++++++++++++
 24 files changed, 274 insertions(+), 2911 deletions(-)
 delete mode 100644 fs/btrfs/debug-tree.c
 delete mode 100644 fs/btrfs/dir-test.c
 delete mode 100644 fs/btrfs/hasher.c
 delete mode 100644 fs/btrfs/kerncompat.h
 delete mode 100644 fs/btrfs/list.h
 delete mode 100644 fs/btrfs/mkfs.c
 delete mode 100644 fs/btrfs/quick-test.c
 delete mode 100644 fs/btrfs/radix-tree.c
 delete mode 100644 fs/btrfs/radix-tree.h
 delete mode 100644 fs/btrfs/random-test.c
 create mode 100644 fs/btrfs/super.c

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 0720169b6d66..99e45a54ebd6 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -1,40 +1,20 @@
-CC=gcc
-CFLAGS = -g -Wall -Werror
-headers = radix-tree.h ctree.h disk-io.h kerncompat.h print-tree.h list.h \
-	  transaction.h
-objects = ctree.o disk-io.o radix-tree.o extent-tree.o print-tree.o \
-	  root-tree.o dir-item.o hash.o file-item.o inode-item.o \
-	  inode-map.o \
+ifneq ($(KERNELRELEASE),)
+# kbuild part of makefile
 
-# if you don't have sparse installed, use ls instead
-CHECKFLAGS=-D__linux__ -Dlinux -D__STDC__ -Dunix -D__unix__ -Wbitwise \
-		-Wcontext -Wcast-truncate -Wuninitialized -Wshadow -Wundef
-check=sparse $(CHECKFLAGS)
-#check=ls
+obj-m  := btrfs.o
+btrfs-y := super.o
 
-.c.o:
-	$(check) $<
-	$(CC) $(CFLAGS) -c $<
+#btrfs-y := ctree.o disk-io.o radix-tree.o extent-tree.o print-tree.o \
+#	  root-tree.o dir-item.o hash.o file-item.o inode-item.o \
+#	  inode-map.o \
 
-all: tester debug-tree quick-test dir-test tags mkfs.btrfs
-
-mkfs.btrfs: $(objects) mkfs.o
-	gcc $(CFLAGS) -o mkfs.btrfs $(objects) mkfs.o
-
-debug-tree: $(objects) debug-tree.o
-	gcc $(CFLAGS) -o debug-tree $(objects) debug-tree.o
-
-tester: $(objects) random-test.o
-	gcc $(CFLAGS) -o tester $(objects) random-test.o
-
-dir-test: $(objects) dir-test.o
-	gcc $(CFLAGS) -o dir-test $(objects) dir-test.o
-quick-test: $(objects) quick-test.o
-	gcc $(CFLAGS) -o quick-test $(objects) quick-test.o
-
-$(objects): $(headers)
-
-clean :
-	rm debug-tree tester *.o
+else
 
+# Normal Makefile
 
+KERNELDIR := /lib/modules/`uname -r`/build
+all::
+	$(MAKE) -C $(KERNELDIR) M=`pwd` modules
+clean::
+	rm *.o btrfs.ko
+endif
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 32922643b5bc..9fbd07c37fde 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -1,10 +1,6 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include "kerncompat.h"
-#include "radix-tree.h"
+#include <linux/module.h>
 #include "ctree.h"
 #include "disk-io.h"
-#include "print-tree.h"
 
 static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_path *path, int level);
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 1a4d1d6fa401..ae8518cb94bf 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1,9 +1,6 @@
 #ifndef __BTRFS__
 #define __BTRFS__
 
-#include "list.h"
-#include "kerncompat.h"
-
 struct btrfs_trans_handle;
 
 #define BTRFS_MAGIC "_BtRfS_M"
@@ -75,6 +72,7 @@ struct btrfs_super_block {
 	__le64 root;
 	__le64 total_blocks;
 	__le64 blocks_used;
+	__le64 root_dir_objectid;
 } __attribute__ ((__packed__));
 
 /*
@@ -693,6 +691,17 @@ static inline void btrfs_set_super_blocksize(struct btrfs_super_block *s,
 	s->blocksize = cpu_to_le32(val);
 }
 
+static inline u64 btrfs_super_root_dir(struct btrfs_super_block *s)
+{
+	return le64_to_cpu(s->root_dir_objectid);
+}
+
+static inline void btrfs_set_super_root_dir(struct btrfs_super_block *s, u64
+					    val)
+{
+	s->root_dir_objectid = cpu_to_le64(val);
+}
+
 static inline u8 *btrfs_leaf_data(struct btrfs_leaf *l)
 {
 	return (u8 *)l->items;
diff --git a/fs/btrfs/debug-tree.c b/fs/btrfs/debug-tree.c
deleted file mode 100644
index fd07969600c6..000000000000
--- a/fs/btrfs/debug-tree.c
+++ /dev/null
@@ -1,38 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include "kerncompat.h"
-#include "radix-tree.h"
-#include "ctree.h"
-#include "disk-io.h"
-#include "print-tree.h"
-#include "transaction.h"
-
-int main(int ac, char **av) {
-	struct btrfs_super_block super;
-	struct btrfs_root *root;
-
-	if (ac != 2) {
-		fprintf(stderr, "usage: %s device\n", av[0]);
-		exit(1);
-	}
-	radix_tree_init();
-	root = open_ctree(av[1], &super);
-	if (!root) {
-		fprintf(stderr, "unable to open %s\n", av[1]);
-		exit(1);
-	}
-	printf("fs tree\n");
-	btrfs_print_tree(root, root->node);
-	printf("map tree\n");
-	btrfs_print_tree(root->fs_info->extent_root,
-			 root->fs_info->extent_root->node);
-	printf("inode tree\n");
-	btrfs_print_tree(root->fs_info->inode_root,
-			 root->fs_info->inode_root->node);
-	printf("root tree\n");
-	btrfs_print_tree(root->fs_info->tree_root,
-			 root->fs_info->tree_root->node);
-	printf("total blocks %Lu\n", btrfs_super_total_blocks(&super));
-	printf("blocks used %Lu\n", btrfs_super_blocks_used(&super));
-	return 0;
-}
diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c
index 949c4e526798..4d8083d92fa0 100644
--- a/fs/btrfs/dir-item.c
+++ b/fs/btrfs/dir-item.c
@@ -1,7 +1,4 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include "kerncompat.h"
-#include "radix-tree.h"
+#include <linux/module.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "hash.h"
@@ -21,7 +18,12 @@ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
 	key.objectid = dir;
 	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY);
-	ret = btrfs_name_hash(name, name_len, &key.offset);
+	if (name_len == 1 && *name == '.')
+		key.offset = 1;
+	else if (name_len == 2 && name[0] == '.' && name[1] == '.')
+		key.offset = 2;
+	else
+		ret = btrfs_name_hash(name, name_len, &key.offset);
 	BUG_ON(ret);
 	btrfs_init_path(&path);
 	data_size = sizeof(*dir_item) + name_len;
diff --git a/fs/btrfs/dir-test.c b/fs/btrfs/dir-test.c
deleted file mode 100644
index b673982a1f3c..000000000000
--- a/fs/btrfs/dir-test.c
+++ /dev/null
@@ -1,494 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <signal.h>
-#include <unistd.h>
-#include "kerncompat.h"
-#include "radix-tree.h"
-#include "ctree.h"
-#include "disk-io.h"
-#include "print-tree.h"
-#include "hash.h"
-#include "transaction.h"
-
-int keep_running = 1;
-struct btrfs_super_block super;
-static u64 dir_oid = 44556;
-static u64 file_oid = 33778;
-
-static int find_num(struct radix_tree_root *root, unsigned long *num_ret,
-		     int exists)
-{
-	unsigned long num = rand();
-	unsigned long res[2];
-	int ret;
-
-again:
-	ret = radix_tree_gang_lookup(root, (void **)res, num, 2);
-	if (exists) {
-		if (ret == 0)
-			return -1;
-		num = res[0];
-	} else if (ret != 0 && num == res[0]) {
-		num++;
-		if (ret > 1 && num == res[1]) {
-			num++;
-			goto again;
-		}
-	}
-	*num_ret = num;
-	return 0;
-}
-
-static void initial_inode_init(struct btrfs_root *root,
-			       struct btrfs_inode_item *inode_item)
-{
-	memset(inode_item, 0, sizeof(*inode_item));
-	btrfs_set_inode_generation(inode_item, root->fs_info->generation);
-}
-
-static int ins_one(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		   struct radix_tree_root *radix)
-{
-	int ret;
-	char buf[128];
-	unsigned long oid;
-	u64 objectid;
-	struct btrfs_path path;
-	struct btrfs_key inode_map;
-	struct btrfs_inode_item inode_item;
-
-	find_num(radix, &oid, 0);
-	sprintf(buf, "str-%lu", oid);
-
-	ret = btrfs_find_free_objectid(trans, root, dir_oid + 1, &objectid);
-	if (ret)
-		goto error;
-
-	inode_map.objectid = objectid;
-	inode_map.flags = 0;
-	inode_map.offset = 0;
-
-	ret = btrfs_insert_inode_map(trans, root, objectid, &inode_map);
-	if (ret)
-		goto error;
-
-	initial_inode_init(root, &inode_item);
-	ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
-	if (ret)
-		goto error;
-	ret = btrfs_insert_dir_item(trans, root, buf, strlen(buf), dir_oid,
-				    objectid, 1);
-	if (ret)
-		goto error;
-
-	radix_tree_preload(GFP_KERNEL);
-	ret = radix_tree_insert(radix, oid, (void *)oid);
-	radix_tree_preload_end();
-	if (ret)
-		goto error;
-	return ret;
-error:
-	if (ret != -EEXIST)
-		goto fatal;
-
-	/*
-	 * if we got an EEXIST, it may be due to hash collision, double
-	 * check
-	 */
-	btrfs_init_path(&path);
-	ret = btrfs_lookup_dir_item(trans, root, &path, dir_oid, buf,
-				    strlen(buf), 0);
-	if (ret)
-		goto fatal_release;
-	if (!btrfs_match_dir_item_name(root, &path, buf, strlen(buf))) {
-		struct btrfs_dir_item *di;
-		char *found;
-		u32 found_len;
-		u64 myhash;
-		u64 foundhash;
-
-		di = btrfs_item_ptr(&path.nodes[0]->leaf, path.slots[0],
-				    struct btrfs_dir_item);
-		found = (char *)(di + 1);
-		found_len = btrfs_dir_name_len(di);
-		btrfs_name_hash(buf, strlen(buf), &myhash);
-		btrfs_name_hash(found, found_len, &foundhash);
-		if (myhash != foundhash)
-			goto fatal_release;
-		btrfs_release_path(root, &path);
-		return 0;
-	}
-fatal_release:
-	btrfs_release_path(root, &path);
-fatal:
-	printf("failed to insert %lu ret %d\n", oid, ret);
-	return -1;
-}
-
-static int insert_dup(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, struct radix_tree_root *radix)
-{
-	int ret;
-	char buf[128];
-	unsigned long oid;
-
-	ret = find_num(radix, &oid, 1);
-	if (ret < 0)
-		return 0;
-	sprintf(buf, "str-%lu", oid);
-
-	ret = btrfs_insert_dir_item(trans, root, buf, strlen(buf), dir_oid,
-				    file_oid, 1);
-	if (ret != -EEXIST) {
-		printf("insert on %s gave us %d\n", buf, ret);
-		return 1;
-	}
-	return 0;
-}
-
-static int del_dir_item(struct btrfs_trans_handle *trans,
-			struct btrfs_root *root,
-			struct radix_tree_root *radix,
-			unsigned long radix_index,
-			struct btrfs_path *path)
-{
-	int ret;
-	unsigned long *ptr;
-	u64 file_objectid;
-	struct btrfs_dir_item *di;
-
-	/* find the inode number of the file */
-	di = btrfs_item_ptr(&path->nodes[0]->leaf, path->slots[0],
-			    struct btrfs_dir_item);
-	file_objectid = btrfs_dir_objectid(di);
-
-	/* delete the directory item */
-	ret = btrfs_del_item(trans, root, path);
-	if (ret)
-		goto out_release;
-	btrfs_release_path(root, path);
-
-	/* delete the inode */
-	btrfs_init_path(path);
-	ret = btrfs_lookup_inode(trans, root, path, file_objectid, -1);
-	if (ret)
-		goto out_release;
-	ret = btrfs_del_item(trans, root, path);
-	if (ret)
-		goto out_release;
-	btrfs_release_path(root, path);
-
-	/* delete the inode mapping */
-	btrfs_init_path(path);
-	ret = btrfs_lookup_inode_map(trans, root, path, file_objectid, -1);
-	if (ret)
-		goto out_release;
-	ret = btrfs_del_item(trans, root->fs_info->inode_root, path);
-	if (ret)
-		goto out_release;
-
-	if (root->fs_info->last_inode_alloc > file_objectid)
-		root->fs_info->last_inode_alloc = file_objectid;
-	btrfs_release_path(root, path);
-	ptr = radix_tree_delete(radix, radix_index);
-	if (!ptr) {
-		ret = -5555;
-		goto out;
-	}
-	return 0;
-out_release:
-	btrfs_release_path(root, path);
-out:
-	printf("failed to delete %lu %d\n", radix_index, ret);
-	return -1;
-}
-
-static int del_one(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		   struct radix_tree_root *radix)
-{
-	int ret;
-	char buf[128];
-	unsigned long oid;
-	struct btrfs_path path;
-
-	ret = find_num(radix, &oid, 1);
-	if (ret < 0)
-		return 0;
-	sprintf(buf, "str-%lu", oid);
-	btrfs_init_path(&path);
-	ret = btrfs_lookup_dir_item(trans, root, &path, dir_oid, buf,
-				    strlen(buf), -1);
-	if (ret)
-		goto out_release;
-
-	ret = del_dir_item(trans, root, radix, oid, &path);
-	if (ret)
-		goto out_release;
-	return ret;
-out_release:
-	btrfs_release_path(root, &path);
-	printf("failed to delete %lu %d\n", oid, ret);
-	return -1;
-}
-
-static int lookup_item(struct btrfs_trans_handle *trans, struct btrfs_root
-		       *root, struct radix_tree_root *radix)
-{
-	struct btrfs_path path;
-	char buf[128];
-	int ret;
-	unsigned long oid;
-	u64 objectid;
-	struct btrfs_dir_item *di;
-
-	ret = find_num(radix, &oid, 1);
-	if (ret < 0)
-		return 0;
-	sprintf(buf, "str-%lu", oid);
-	btrfs_init_path(&path);
-	ret = btrfs_lookup_dir_item(trans, root, &path, dir_oid, buf,
-				    strlen(buf), 0);
-	if (!ret) {
-		di = btrfs_item_ptr(&path.nodes[0]->leaf, path.slots[0],
-				    struct btrfs_dir_item);
-		objectid = btrfs_dir_objectid(di);
-		btrfs_release_path(root, &path);
-		btrfs_init_path(&path);
-		ret = btrfs_lookup_inode_map(trans, root, &path, objectid, 0);
-	}
-	btrfs_release_path(root, &path);
-	if (ret) {
-		printf("unable to find key %lu\n", oid);
-		return -1;
-	}
-	return 0;
-}
-
-static int lookup_enoent(struct btrfs_trans_handle *trans, struct btrfs_root
-			 *root, struct radix_tree_root *radix)
-{
-	struct btrfs_path path;
-	char buf[128];
-	int ret;
-	unsigned long oid;
-
-	ret = find_num(radix, &oid, 0);
-	if (ret < 0)
-		return 0;
-	sprintf(buf, "str-%lu", oid);
-	btrfs_init_path(&path);
-	ret = btrfs_lookup_dir_item(trans, root, &path, dir_oid, buf,
-				    strlen(buf), 0);
-	btrfs_release_path(root, &path);
-	if (!ret) {
-		printf("able to find key that should not exist %lu\n", oid);
-		return -1;
-	}
-	return 0;
-}
-
-static int empty_tree(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, struct radix_tree_root *radix, int nr)
-{
-	struct btrfs_path path;
-	struct btrfs_key key;
-	unsigned long found = 0;
-	u32 found_len;
-	int ret;
-	int slot;
-	int count = 0;
-	char buf[128];
-	struct btrfs_dir_item *di;
-
-	key.offset = (u64)-1;
-	key.flags = 0;
-	btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY);
-	key.objectid = dir_oid;
-	while(nr-- >= 0) {
-		btrfs_init_path(&path);
-		ret = btrfs_search_slot(trans, root, &key, &path, -1, 1);
-		if (ret < 0) {
-			btrfs_release_path(root, &path);
-			return ret;
-		}
-		if (ret != 0) {
-			if (path.slots[0] == 0) {
-				btrfs_release_path(root, &path);
-				break;
-			}
-			path.slots[0] -= 1;
-		}
-		slot = path.slots[0];
-		di = btrfs_item_ptr(&path.nodes[0]->leaf, slot,
-				    struct btrfs_dir_item);
-		found_len = btrfs_dir_name_len(di);
-		memcpy(buf, (char *)(di + 1), found_len);
-		BUG_ON(found_len > 128);
-		buf[found_len] = '\0';
-		found = atoi(buf + 4);
-		ret = del_dir_item(trans, root, radix, found, &path);
-		count++;
-		if (ret) {
-			fprintf(stderr,
-				"failed to remove %lu from tree\n",
-				found);
-			return -1;
-		}
-		if (!keep_running)
-			break;
-	}
-	return 0;
-	fprintf(stderr, "failed to delete from the radix %lu\n", found);
-	return -1;
-}
-
-static int fill_tree(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		     struct radix_tree_root *radix, int count)
-{
-	int i;
-	int ret = 0;
-	for (i = 0; i < count; i++) {
-		ret = ins_one(trans, root, radix);
-		if (ret) {
-			fprintf(stderr, "fill failed\n");
-			goto out;
-		}
-		if (i % 1000 == 0) {
-			ret = btrfs_commit_transaction(trans, root, &super);
-			if (ret) {
-				fprintf(stderr, "fill commit failed\n");
-				return ret;
-			}
-		}
-		if (i && i % 10000 == 0) {
-			printf("bigfill %d\n", i);
-		}
-		if (!keep_running)
-			break;
-	}
-out:
-	return ret;
-}
-
-static int bulk_op(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		   struct radix_tree_root *radix)
-{
-	int ret;
-	int nr = rand() % 5000;
-	static int run_nr = 0;
-
-	/* do the bulk op much less frequently */
-	if (run_nr++ % 100)
-		return 0;
-	ret = empty_tree(trans, root, radix, nr);
-	if (ret)
-		return ret;
-	ret = fill_tree(trans, root, radix, nr);
-	if (ret)
-		return ret;
-	return 0;
-}
-
-
-int (*ops[])(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct
-	     radix_tree_root *radix) =
-	{ ins_one, insert_dup, del_one, lookup_item,
-	  lookup_enoent, bulk_op };
-
-void sigstopper(int ignored)
-{
-	keep_running = 0;
-	fprintf(stderr, "caught exit signal, stopping\n");
-}
-
-int print_usage(void)
-{
-	printf("usage: tester [-ih] [-c count] [-f count]\n");
-	printf("\t -c count -- iteration count after filling\n");
-	printf("\t -f count -- run this many random inserts before starting\n");
-	printf("\t -i       -- only do initial fill\n");
-	printf("\t -h       -- this help text\n");
-	exit(1);
-}
-int main(int ac, char **av)
-{
-	RADIX_TREE(radix, GFP_KERNEL);
-	struct btrfs_root *root;
-	int i;
-	int ret;
-	int count;
-	int op;
-	int iterations = 20000;
-	int init_fill_count = 800000;
-	int err = 0;
-	int initial_only = 0;
-	struct btrfs_trans_handle *trans;
-	radix_tree_init();
-
-	root = open_ctree("dbfile", &super);
-	trans = btrfs_start_transaction(root, 1);
-
-	signal(SIGTERM, sigstopper);
-	signal(SIGINT, sigstopper);
-
-	for (i = 1 ; i < ac ; i++) {
-		if (strcmp(av[i], "-i") == 0) {
-			initial_only = 1;
-		} else if (strcmp(av[i], "-c") == 0) {
-			iterations = atoi(av[i+1]);
-			i++;
-		} else if (strcmp(av[i], "-f") == 0) {
-			init_fill_count = atoi(av[i+1]);
-			i++;
-		} else {
-			print_usage();
-		}
-	}
-	printf("initial fill\n");
-	ret = fill_tree(trans, root, &radix, init_fill_count);
-	printf("starting run\n");
-	if (ret) {
-		err = ret;
-		goto out;
-	}
-	if (initial_only == 1) {
-		goto out;
-	}
-	for (i = 0; i < iterations; i++) {
-		op = rand() % ARRAY_SIZE(ops);
-		count = rand() % 128;
-		if (i % 2000 == 0) {
-			printf("%d\n", i);
-			fflush(stdout);
-		}
-		if (i && i % 5000 == 0) {
-			printf("open & close, root level %d nritems %d\n",
-				btrfs_header_level(&root->node->node.header),
-				btrfs_header_nritems(&root->node->node.header));
-			close_ctree(root, &super);
-			root = open_ctree("dbfile", &super);
-		}
-		while(count--) {
-			ret = ops[op](trans, root, &radix);
-			if (ret) {
-				fprintf(stderr, "op %d failed %d:%d\n",
-					op, i, iterations);
-				btrfs_print_tree(root, root->node);
-				fprintf(stderr, "op %d failed %d:%d\n",
-					op, i, iterations);
-				err = ret;
-				goto out;
-			}
-			if (ops[op] == bulk_op)
-				break;
-			if (keep_running == 0) {
-				err = 0;
-				goto out;
-			}
-		}
-	}
-out:
-	close_ctree(root, &super);
-	return err;
-}
-
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 0322c55162cb..05637f9fd7c7 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -266,20 +266,25 @@ static int find_and_setup_root(struct btrfs_super_block *super,
 }
 
 struct btrfs_root *open_ctree(char *filename, struct btrfs_super_block *super)
+{
+	int fp;
+
+	fp = open(filename, O_CREAT | O_RDWR, 0600);
+	if (fp < 0) {
+		return NULL;
+	}
+	return open_ctree_fd(fp, super);
+}
+
+struct btrfs_root *open_ctree_fd(int fp, struct btrfs_super_block *super)
 {
 	struct btrfs_root *root = malloc(sizeof(struct btrfs_root));
 	struct btrfs_root *extent_root = malloc(sizeof(struct btrfs_root));
 	struct btrfs_root *tree_root = malloc(sizeof(struct btrfs_root));
 	struct btrfs_root *inode_root = malloc(sizeof(struct btrfs_root));
 	struct btrfs_fs_info *fs_info = malloc(sizeof(*fs_info));
-	int fp;
 	int ret;
 
-	fp = open(filename, O_CREAT | O_RDWR, 0600);
-	if (fp < 0) {
-		free(root);
-		return NULL;
-	}
 	INIT_RADIX_TREE(&fs_info->cache_radix, GFP_KERNEL);
 	INIT_RADIX_TREE(&fs_info->pinned_radix, GFP_KERNEL);
 	INIT_LIST_HEAD(&fs_info->trans);
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 24a9e77c8311..d888cf5c350b 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -24,6 +24,7 @@ int clean_tree_block(struct btrfs_trans_handle *trans,
 int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct btrfs_root
 			     *root, struct btrfs_super_block *s);
 struct btrfs_root *open_ctree(char *filename, struct btrfs_super_block *s);
+struct btrfs_root *open_ctree_fd(int fp, struct btrfs_super_block *super);
 int close_ctree(struct btrfs_root *root, struct btrfs_super_block *s);
 void btrfs_block_release(struct btrfs_root *root, struct btrfs_buffer *buf);
 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root *root,
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 9bc4ad38876d..53a7550b5c1e 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1,7 +1,5 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include "kerncompat.h"
-#include "radix-tree.h"
+#include <linux/module.h>
+#include <linux/radix-tree.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "print-tree.h"
@@ -183,9 +181,9 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_init_path(&path);
 	ret = btrfs_search_slot(trans, extent_root, &key, &path, -1, 1);
 	if (ret) {
-		printf("failed to find %Lu\n", key.objectid);
+		printk("failed to find %Lu\n", key.objectid);
 		btrfs_print_tree(extent_root, extent_root->node);
-		printf("failed to find %Lu\n", key.objectid);
+		printk("failed to find %Lu\n", key.objectid);
 		BUG();
 	}
 	ei = btrfs_item_ptr(&path.nodes[0]->leaf, path.slots[0],
diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c
index 24cfd6d85240..8e1e5b4ccfaf 100644
--- a/fs/btrfs/file-item.c
+++ b/fs/btrfs/file-item.c
@@ -1,9 +1,5 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include "kerncompat.h"
-#include "radix-tree.h"
+#include <linux/module.h>
 #include "ctree.h"
-#include "disk-io.h"
 #include "transaction.h"
 
 int btrfs_create_file(struct btrfs_trans_handle *trans,
diff --git a/fs/btrfs/hash.c b/fs/btrfs/hash.c
index 92c37eece1f2..6c2a71a46c7d 100644
--- a/fs/btrfs/hash.c
+++ b/fs/btrfs/hash.c
@@ -10,7 +10,6 @@
  * License.
  */
 
-#include "kerncompat.h"
 #define DELTA 0x9E3779B9
 
 static void TEA_transform(__u32 buf[2], __u32 const in[])
diff --git a/fs/btrfs/hasher.c b/fs/btrfs/hasher.c
deleted file mode 100644
index 96702da4329c..000000000000
--- a/fs/btrfs/hasher.c
+++ /dev/null
@@ -1,23 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include "kerncompat.h"
-#include "hash.h"
-
-int main() {
-	u64 result;
-	int ret;
-	char line[255];
-	char *p;
-	while(1) {
-		p = fgets(line, 255, stdin);
-		if (!p)
-			break;
-		if (strlen(line) == 0)
-			continue;
-		ret = btrfs_name_hash(line, strlen(line), &result);
-		BUG_ON(ret);
-		printf("hash returns %Lu\n", result);
-	}
-	return 0;
-}
diff --git a/fs/btrfs/inode-item.c b/fs/btrfs/inode-item.c
index 7caeb11e8756..8d8c26a6c1a0 100644
--- a/fs/btrfs/inode-item.c
+++ b/fs/btrfs/inode-item.c
@@ -1,7 +1,4 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include "kerncompat.h"
-#include "radix-tree.h"
+#include <linux/module.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c
index f412b3392135..c7fda3bf7b21 100644
--- a/fs/btrfs/inode-map.c
+++ b/fs/btrfs/inode-map.c
@@ -1,7 +1,4 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include "kerncompat.h"
-#include "radix-tree.h"
+#include <linux/module.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
diff --git a/fs/btrfs/kerncompat.h b/fs/btrfs/kerncompat.h
deleted file mode 100644
index 105d3f584089..000000000000
--- a/fs/btrfs/kerncompat.h
+++ /dev/null
@@ -1,96 +0,0 @@
-#ifndef __KERNCOMPAT
-#define __KERNCOMPAT
-#define gfp_t int
-#define get_cpu_var(p) (p)
-#define __get_cpu_var(p) (p)
-#define BITS_PER_LONG 64
-#define __GFP_BITS_SHIFT 20
-#define __GFP_BITS_MASK ((int)((1 << __GFP_BITS_SHIFT) - 1))
-#define GFP_KERNEL 0
-#define __read_mostly
-#define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
-#define PAGE_SHIFT 12
-#define ULONG_MAX       (~0UL)
-#define BUG() abort()
-#ifdef __CHECKER__
-#define __force    __attribute__((force))
-#define __bitwise__ __attribute__((bitwise))
-#else
-#define __force
-#define __bitwise__
-#endif
-
-typedef unsigned int u32;
-typedef u32 __u32;
-typedef unsigned long long u64;
-typedef unsigned char u8;
-typedef unsigned short u16;
-
-typedef unsigned long pgoff_t;
-
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-
-struct vma_shared { int prio_tree_node; };
-struct vm_area_struct {
-	unsigned long vm_pgoff;
-	unsigned long vm_start;
-	unsigned long vm_end;
-	struct vma_shared shared;
-};
-
-struct page {
-	unsigned long index;
-};
-
-static inline void preempt_enable(void) { do {; } while(0);}
-static inline void preempt_disable(void) { do {; } while(0);}
-
-static inline void __set_bit(int bit, unsigned long *map) {
-	unsigned long *p = map + bit / BITS_PER_LONG;
-	bit = bit & (BITS_PER_LONG -1);
-	*p |= 1UL << bit;
-}
-
-static inline int test_bit(int bit, unsigned long *map) {
-	unsigned long *p = map + bit / BITS_PER_LONG;
-	bit = bit & (BITS_PER_LONG -1);
-	return *p & (1UL << bit) ? 1 : 0;
-}
-
-static inline void __clear_bit(int bit, unsigned long *map) {
-	unsigned long *p = map + bit / BITS_PER_LONG;
-	bit = bit & (BITS_PER_LONG -1);
-	*p &= ~(1UL << bit);
-}
-#define BUG_ON(c) do { if (c) abort(); } while (0)
-
-#define container_of(ptr, type, member) ({                      \
-        const typeof( ((type *)0)->member ) *__mptr = (ptr);    \
-	        (type *)( (char *)__mptr - __builtin_offsetof(type,member) );})
-
-#define ENOMEM 5
-#define EEXIST 6
-
-#define __CHECK_ENDIAN__
-#ifdef __CHECK_ENDIAN__
-#define __bitwise __bitwise__
-#else
-#define __bitwise
-#endif
-
-typedef u16 __bitwise __le16;
-typedef u16 __bitwise __be16;
-typedef u32 __bitwise __le32;
-typedef u32 __bitwise __be32;
-typedef u64 __bitwise __le64;
-typedef u64 __bitwise __be64;
-
-#define cpu_to_le64(x) ((__force __le64)(u64)(x))
-#define le64_to_cpu(x) ((__force u64)(__le64)(x))
-#define cpu_to_le32(x) ((__force __le32)(u32)(x))
-#define le32_to_cpu(x) ((__force u32)(__le32)(x))
-#define cpu_to_le16(x) ((__force __le16)(u16)(x))
-#define le16_to_cpu(x) ((__force u16)(__le16)(x))
-#endif
diff --git a/fs/btrfs/list.h b/fs/btrfs/list.h
deleted file mode 100644
index 1aafafb13370..000000000000
--- a/fs/btrfs/list.h
+++ /dev/null
@@ -1,418 +0,0 @@
-#ifndef _LINUX_LIST_H
-#define _LINUX_LIST_H
-
-#define LIST_POISON1  ((void *) 0x00100100)
-#define LIST_POISON2  ((void *) 0x00200200)
-
-/*
- * Simple doubly linked list implementation.
- *
- * Some of the internal functions ("__xxx") are useful when
- * manipulating whole lists rather than single entries, as
- * sometimes we already know the next/prev entries and we can
- * generate better code by using them directly rather than
- * using the generic single-entry routines.
- */
-
-struct list_head {
-	struct list_head *next, *prev;
-};
-
-#define LIST_HEAD_INIT(name) { &(name), &(name) }
-
-#define LIST_HEAD(name) \
-	struct list_head name = LIST_HEAD_INIT(name)
-
-static inline void INIT_LIST_HEAD(struct list_head *list)
-{
-	list->next = list;
-	list->prev = list;
-}
-
-/*
- * Insert a new entry between two known consecutive entries.
- *
- * This is only for internal list manipulation where we know
- * the prev/next entries already!
- */
-#ifndef CONFIG_DEBUG_LIST
-static inline void __list_add(struct list_head *new,
-			      struct list_head *prev,
-			      struct list_head *next)
-{
-	next->prev = new;
-	new->next = next;
-	new->prev = prev;
-	prev->next = new;
-}
-#else
-extern void __list_add(struct list_head *new,
-			      struct list_head *prev,
-			      struct list_head *next);
-#endif
-
-/**
- * list_add - add a new entry
- * @new: new entry to be added
- * @head: list head to add it after
- *
- * Insert a new entry after the specified head.
- * This is good for implementing stacks.
- */
-#ifndef CONFIG_DEBUG_LIST
-static inline void list_add(struct list_head *new, struct list_head *head)
-{
-	__list_add(new, head, head->next);
-}
-#else
-extern void list_add(struct list_head *new, struct list_head *head);
-#endif
-
-
-/**
- * list_add_tail - add a new entry
- * @new: new entry to be added
- * @head: list head to add it before
- *
- * Insert a new entry before the specified head.
- * This is useful for implementing queues.
- */
-static inline void list_add_tail(struct list_head *new, struct list_head *head)
-{
-	__list_add(new, head->prev, head);
-}
-
-/*
- * Delete a list entry by making the prev/next entries
- * point to each other.
- *
- * This is only for internal list manipulation where we know
- * the prev/next entries already!
- */
-static inline void __list_del(struct list_head * prev, struct list_head * next)
-{
-	next->prev = prev;
-	prev->next = next;
-}
-
-/**
- * list_del - deletes entry from list.
- * @entry: the element to delete from the list.
- * Note: list_empty on entry does not return true after this, the entry is
- * in an undefined state.
- */
-#ifndef CONFIG_DEBUG_LIST
-static inline void list_del(struct list_head *entry)
-{
-	__list_del(entry->prev, entry->next);
-	entry->next = LIST_POISON1;
-	entry->prev = LIST_POISON2;
-}
-#else
-extern void list_del(struct list_head *entry);
-#endif
-
-/**
- * list_replace - replace old entry by new one
- * @old : the element to be replaced
- * @new : the new element to insert
- * Note: if 'old' was empty, it will be overwritten.
- */
-static inline void list_replace(struct list_head *old,
-				struct list_head *new)
-{
-	new->next = old->next;
-	new->next->prev = new;
-	new->prev = old->prev;
-	new->prev->next = new;
-}
-
-static inline void list_replace_init(struct list_head *old,
-					struct list_head *new)
-{
-	list_replace(old, new);
-	INIT_LIST_HEAD(old);
-}
-/**
- * list_del_init - deletes entry from list and reinitialize it.
- * @entry: the element to delete from the list.
- */
-static inline void list_del_init(struct list_head *entry)
-{
-	__list_del(entry->prev, entry->next);
-	INIT_LIST_HEAD(entry);
-}
-
-/**
- * list_move - delete from one list and add as another's head
- * @list: the entry to move
- * @head: the head that will precede our entry
- */
-static inline void list_move(struct list_head *list, struct list_head *head)
-{
-        __list_del(list->prev, list->next);
-        list_add(list, head);
-}
-
-/**
- * list_move_tail - delete from one list and add as another's tail
- * @list: the entry to move
- * @head: the head that will follow our entry
- */
-static inline void list_move_tail(struct list_head *list,
-				  struct list_head *head)
-{
-        __list_del(list->prev, list->next);
-        list_add_tail(list, head);
-}
-
-/**
- * list_is_last - tests whether @list is the last entry in list @head
- * @list: the entry to test
- * @head: the head of the list
- */
-static inline int list_is_last(const struct list_head *list,
-				const struct list_head *head)
-{
-	return list->next == head;
-}
-
-/**
- * list_empty - tests whether a list is empty
- * @head: the list to test.
- */
-static inline int list_empty(const struct list_head *head)
-{
-	return head->next == head;
-}
-
-/**
- * list_empty_careful - tests whether a list is empty and not being modified
- * @head: the list to test
- *
- * Description:
- * tests whether a list is empty _and_ checks that no other CPU might be
- * in the process of modifying either member (next or prev)
- *
- * NOTE: using list_empty_careful() without synchronization
- * can only be safe if the only activity that can happen
- * to the list entry is list_del_init(). Eg. it cannot be used
- * if another CPU could re-list_add() it.
- */
-static inline int list_empty_careful(const struct list_head *head)
-{
-	struct list_head *next = head->next;
-	return (next == head) && (next == head->prev);
-}
-
-static inline void __list_splice(struct list_head *list,
-				 struct list_head *head)
-{
-	struct list_head *first = list->next;
-	struct list_head *last = list->prev;
-	struct list_head *at = head->next;
-
-	first->prev = head;
-	head->next = first;
-
-	last->next = at;
-	at->prev = last;
-}
-
-/**
- * list_splice - join two lists
- * @list: the new list to add.
- * @head: the place to add it in the first list.
- */
-static inline void list_splice(struct list_head *list, struct list_head *head)
-{
-	if (!list_empty(list))
-		__list_splice(list, head);
-}
-
-/**
- * list_splice_init - join two lists and reinitialise the emptied list.
- * @list: the new list to add.
- * @head: the place to add it in the first list.
- *
- * The list at @list is reinitialised
- */
-static inline void list_splice_init(struct list_head *list,
-				    struct list_head *head)
-{
-	if (!list_empty(list)) {
-		__list_splice(list, head);
-		INIT_LIST_HEAD(list);
-	}
-}
-
-/**
- * list_entry - get the struct for this entry
- * @ptr:	the &struct list_head pointer.
- * @type:	the type of the struct this is embedded in.
- * @member:	the name of the list_struct within the struct.
- */
-#define list_entry(ptr, type, member) \
-	container_of(ptr, type, member)
-
-/**
- * list_for_each	-	iterate over a list
- * @pos:	the &struct list_head to use as a loop cursor.
- * @head:	the head for your list.
- */
-#define list_for_each(pos, head) \
-	for (pos = (head)->next; prefetch(pos->next), pos != (head); \
-        	pos = pos->next)
-
-/**
- * __list_for_each	-	iterate over a list
- * @pos:	the &struct list_head to use as a loop cursor.
- * @head:	the head for your list.
- *
- * This variant differs from list_for_each() in that it's the
- * simplest possible list iteration code, no prefetching is done.
- * Use this for code that knows the list to be very short (empty
- * or 1 entry) most of the time.
- */
-#define __list_for_each(pos, head) \
-	for (pos = (head)->next; pos != (head); pos = pos->next)
-
-/**
- * list_for_each_prev	-	iterate over a list backwards
- * @pos:	the &struct list_head to use as a loop cursor.
- * @head:	the head for your list.
- */
-#define list_for_each_prev(pos, head) \
-	for (pos = (head)->prev; prefetch(pos->prev), pos != (head); \
-        	pos = pos->prev)
-
-/**
- * list_for_each_safe - iterate over a list safe against removal of list entry
- * @pos:	the &struct list_head to use as a loop cursor.
- * @n:		another &struct list_head to use as temporary storage
- * @head:	the head for your list.
- */
-#define list_for_each_safe(pos, n, head) \
-	for (pos = (head)->next, n = pos->next; pos != (head); \
-		pos = n, n = pos->next)
-
-/**
- * list_for_each_entry	-	iterate over list of given type
- * @pos:	the type * to use as a loop cursor.
- * @head:	the head for your list.
- * @member:	the name of the list_struct within the struct.
- */
-#define list_for_each_entry(pos, head, member)				\
-	for (pos = list_entry((head)->next, typeof(*pos), member);	\
-	     prefetch(pos->member.next), &pos->member != (head); 	\
-	     pos = list_entry(pos->member.next, typeof(*pos), member))
-
-/**
- * list_for_each_entry_reverse - iterate backwards over list of given type.
- * @pos:	the type * to use as a loop cursor.
- * @head:	the head for your list.
- * @member:	the name of the list_struct within the struct.
- */
-#define list_for_each_entry_reverse(pos, head, member)			\
-	for (pos = list_entry((head)->prev, typeof(*pos), member);	\
-	     prefetch(pos->member.prev), &pos->member != (head); 	\
-	     pos = list_entry(pos->member.prev, typeof(*pos), member))
-
-/**
- * list_prepare_entry - prepare a pos entry for use in list_for_each_entry_continue
- * @pos:	the type * to use as a start point
- * @head:	the head of the list
- * @member:	the name of the list_struct within the struct.
- *
- * Prepares a pos entry for use as a start point in list_for_each_entry_continue.
- */
-#define list_prepare_entry(pos, head, member) \
-	((pos) ? : list_entry(head, typeof(*pos), member))
-
-/**
- * list_for_each_entry_continue - continue iteration over list of given type
- * @pos:	the type * to use as a loop cursor.
- * @head:	the head for your list.
- * @member:	the name of the list_struct within the struct.
- *
- * Continue to iterate over list of given type, continuing after
- * the current position.
- */
-#define list_for_each_entry_continue(pos, head, member) 		\
-	for (pos = list_entry(pos->member.next, typeof(*pos), member);	\
-	     prefetch(pos->member.next), &pos->member != (head);	\
-	     pos = list_entry(pos->member.next, typeof(*pos), member))
-
-/**
- * list_for_each_entry_from - iterate over list of given type from the current point
- * @pos:	the type * to use as a loop cursor.
- * @head:	the head for your list.
- * @member:	the name of the list_struct within the struct.
- *
- * Iterate over list of given type, continuing from current position.
- */
-#define list_for_each_entry_from(pos, head, member) 			\
-	for (; prefetch(pos->member.next), &pos->member != (head);	\
-	     pos = list_entry(pos->member.next, typeof(*pos), member))
-
-/**
- * list_for_each_entry_safe - iterate over list of given type safe against removal of list entry
- * @pos:	the type * to use as a loop cursor.
- * @n:		another type * to use as temporary storage
- * @head:	the head for your list.
- * @member:	the name of the list_struct within the struct.
- */
-#define list_for_each_entry_safe(pos, n, head, member)			\
-	for (pos = list_entry((head)->next, typeof(*pos), member),	\
-		n = list_entry(pos->member.next, typeof(*pos), member);	\
-	     &pos->member != (head); 					\
-	     pos = n, n = list_entry(n->member.next, typeof(*n), member))
-
-/**
- * list_for_each_entry_safe_continue
- * @pos:	the type * to use as a loop cursor.
- * @n:		another type * to use as temporary storage
- * @head:	the head for your list.
- * @member:	the name of the list_struct within the struct.
- *
- * Iterate over list of given type, continuing after current point,
- * safe against removal of list entry.
- */
-#define list_for_each_entry_safe_continue(pos, n, head, member) 		\
-	for (pos = list_entry(pos->member.next, typeof(*pos), member), 		\
-		n = list_entry(pos->member.next, typeof(*pos), member);		\
-	     &pos->member != (head);						\
-	     pos = n, n = list_entry(n->member.next, typeof(*n), member))
-
-/**
- * list_for_each_entry_safe_from
- * @pos:	the type * to use as a loop cursor.
- * @n:		another type * to use as temporary storage
- * @head:	the head for your list.
- * @member:	the name of the list_struct within the struct.
- *
- * Iterate over list of given type from current point, safe against
- * removal of list entry.
- */
-#define list_for_each_entry_safe_from(pos, n, head, member) 			\
-	for (n = list_entry(pos->member.next, typeof(*pos), member);		\
-	     &pos->member != (head);						\
-	     pos = n, n = list_entry(n->member.next, typeof(*n), member))
-
-/**
- * list_for_each_entry_safe_reverse
- * @pos:	the type * to use as a loop cursor.
- * @n:		another type * to use as temporary storage
- * @head:	the head for your list.
- * @member:	the name of the list_struct within the struct.
- *
- * Iterate backwards over list of given type, safe against removal
- * of list entry.
- */
-#define list_for_each_entry_safe_reverse(pos, n, head, member)		\
-	for (pos = list_entry((head)->prev, typeof(*pos), member),	\
-		n = list_entry(pos->member.prev, typeof(*pos), member);	\
-	     &pos->member != (head); 					\
-	     pos = n, n = list_entry(n->member.prev, typeof(*n), member))
-
-#endif
diff --git a/fs/btrfs/mkfs.c b/fs/btrfs/mkfs.c
deleted file mode 100644
index f7efc8a5fb1a..000000000000
--- a/fs/btrfs/mkfs.c
+++ /dev/null
@@ -1,255 +0,0 @@
-#define _XOPEN_SOURCE 500
-#ifndef __CHECKER__
-#include <sys/ioctl.h>
-#include <sys/mount.h>
-#endif
-#include <stdio.h>
-#include <stdlib.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <unistd.h>
-#include "kerncompat.h"
-#include "radix-tree.h"
-#include "ctree.h"
-#include "disk-io.h"
-
-#ifdef __CHECKER__
-#define BLKGETSIZE64 0
-static inline int ioctl(int fd, int define, u64 *size) { return 0; }
-#endif
-
-#if 0
-#if defined(__linux__) && defined(_IOR) && !defined(BLKGETSIZE64)
-#   define BLKGETSIZE64 _IOR(0x12, 114, __u64)
-#endif
-#endif
-
-int mkfs(int fd, u64 num_blocks, u32 blocksize)
-{
-	struct btrfs_super_block super;
-	struct btrfs_leaf *empty_leaf;
-	struct btrfs_root_item root_item;
-	struct btrfs_item item;
-	struct btrfs_extent_item extent_item;
-	char *block;
-	int ret;
-	u32 itemoff;
-	u32 start_block = BTRFS_SUPER_INFO_OFFSET / blocksize;
-
-	btrfs_set_super_blocknr(&super, start_block);
-	btrfs_set_super_root(&super, start_block + 1);
-	strcpy((char *)(&super.magic), BTRFS_MAGIC);
-	btrfs_set_super_blocksize(&super, blocksize);
-	btrfs_set_super_total_blocks(&super, num_blocks);
-	btrfs_set_super_blocks_used(&super, start_block + 5);
-
-	block = malloc(blocksize);
-	memset(block, 0, blocksize);
-	BUG_ON(sizeof(super) > blocksize);
-	memcpy(block, &super, sizeof(super));
-	ret = pwrite(fd, block, blocksize, BTRFS_SUPER_INFO_OFFSET);
-	BUG_ON(ret != blocksize);
-
-	/* create the tree of root objects */
-	empty_leaf = malloc(blocksize);
-	memset(empty_leaf, 0, blocksize);
-	btrfs_set_header_parentid(&empty_leaf->header,
-				  BTRFS_ROOT_TREE_OBJECTID);
-	btrfs_set_header_blocknr(&empty_leaf->header, start_block + 1);
-	btrfs_set_header_nritems(&empty_leaf->header, 3);
-
-	/* create the items for the root tree */
-	btrfs_set_root_blocknr(&root_item, start_block + 2);
-	btrfs_set_root_refs(&root_item, 1);
-	itemoff = __BTRFS_LEAF_DATA_SIZE(blocksize) - sizeof(root_item);
-	btrfs_set_item_offset(&item, itemoff);
-	btrfs_set_item_size(&item, sizeof(root_item));
-	btrfs_set_disk_key_objectid(&item.key, BTRFS_EXTENT_TREE_OBJECTID);
-	btrfs_set_disk_key_offset(&item.key, 0);
-	btrfs_set_disk_key_flags(&item.key, 0);
-	btrfs_set_disk_key_type(&item.key, BTRFS_ROOT_ITEM_KEY);
-	memcpy(empty_leaf->items, &item, sizeof(item));
-	memcpy(btrfs_leaf_data(empty_leaf) + itemoff,
-		&root_item, sizeof(root_item));
-
-	btrfs_set_root_blocknr(&root_item, start_block + 3);
-	itemoff = itemoff - sizeof(root_item);
-	btrfs_set_item_offset(&item, itemoff);
-	btrfs_set_disk_key_objectid(&item.key, BTRFS_INODE_MAP_OBJECTID);
-	memcpy(empty_leaf->items + 1, &item, sizeof(item));
-	memcpy(btrfs_leaf_data(empty_leaf) + itemoff,
-		&root_item, sizeof(root_item));
-
-	btrfs_set_root_blocknr(&root_item, start_block + 4);
-	itemoff = itemoff - sizeof(root_item);
-	btrfs_set_item_offset(&item, itemoff);
-	btrfs_set_disk_key_objectid(&item.key, BTRFS_FS_TREE_OBJECTID);
-	memcpy(empty_leaf->items + 2, &item, sizeof(item));
-	memcpy(btrfs_leaf_data(empty_leaf) + itemoff,
-		&root_item, sizeof(root_item));
-	ret = pwrite(fd, empty_leaf, blocksize, (start_block + 1) * blocksize);
-
-	/* create the items for the extent tree */
-	btrfs_set_header_parentid(&empty_leaf->header,
-				  BTRFS_EXTENT_TREE_OBJECTID);
-	btrfs_set_header_blocknr(&empty_leaf->header, start_block + 2);
-	btrfs_set_header_nritems(&empty_leaf->header, 5);
-
-	/* item1, reserve blocks 0-16 */
-	btrfs_set_disk_key_objectid(&item.key, 0);
-	btrfs_set_disk_key_offset(&item.key, start_block + 1);
-	btrfs_set_disk_key_flags(&item.key, 0);
-	btrfs_set_disk_key_type(&item.key, BTRFS_EXTENT_ITEM_KEY);
-	itemoff = __BTRFS_LEAF_DATA_SIZE(blocksize) -
-			sizeof(struct btrfs_extent_item);
-	btrfs_set_item_offset(&item, itemoff);
-	btrfs_set_item_size(&item, sizeof(struct btrfs_extent_item));
-	btrfs_set_extent_refs(&extent_item, 1);
-	btrfs_set_extent_owner(&extent_item, 0);
-	memcpy(empty_leaf->items, &item, sizeof(item));
-	memcpy(btrfs_leaf_data(empty_leaf) + btrfs_item_offset(&item),
-		&extent_item, btrfs_item_size(&item));
-
-	/* item2, give block 17 to the root */
-	btrfs_set_disk_key_objectid(&item.key, start_block + 1);
-	btrfs_set_disk_key_offset(&item.key, 1);
-	itemoff = itemoff - sizeof(struct btrfs_extent_item);
-	btrfs_set_item_offset(&item, itemoff);
-	btrfs_set_extent_owner(&extent_item, BTRFS_ROOT_TREE_OBJECTID);
-	memcpy(empty_leaf->items + 1, &item, sizeof(item));
-	memcpy(btrfs_leaf_data(empty_leaf) + btrfs_item_offset(&item),
-		&extent_item, btrfs_item_size(&item));
-
-	/* item3, give block 18 to the extent root */
-	btrfs_set_disk_key_objectid(&item.key, start_block + 2);
-	btrfs_set_disk_key_offset(&item.key, 1);
-	itemoff = itemoff - sizeof(struct btrfs_extent_item);
-	btrfs_set_item_offset(&item, itemoff);
-	btrfs_set_extent_owner(&extent_item, BTRFS_EXTENT_TREE_OBJECTID);
-	memcpy(empty_leaf->items + 2, &item, sizeof(item));
-	memcpy(btrfs_leaf_data(empty_leaf) + btrfs_item_offset(&item),
-		&extent_item, btrfs_item_size(&item));
-
-	/* item4, give block 19 to the inode map */
-	btrfs_set_disk_key_objectid(&item.key, start_block + 3);
-	btrfs_set_disk_key_offset(&item.key, 1);
-	itemoff = itemoff - sizeof(struct btrfs_extent_item);
-	btrfs_set_item_offset(&item, itemoff);
-	btrfs_set_extent_owner(&extent_item, BTRFS_INODE_MAP_OBJECTID);
-	memcpy(empty_leaf->items + 3, &item, sizeof(item));
-	memcpy(btrfs_leaf_data(empty_leaf) + btrfs_item_offset(&item),
-		&extent_item, btrfs_item_size(&item));
-	ret = pwrite(fd, empty_leaf, blocksize, (start_block + 2) * blocksize);
-	if (ret != blocksize)
-		return -1;
-
-	/* item5, give block 20 to the FS root */
-	btrfs_set_disk_key_objectid(&item.key, start_block + 4);
-	btrfs_set_disk_key_offset(&item.key, 1);
-	itemoff = itemoff - sizeof(struct btrfs_extent_item);
-	btrfs_set_item_offset(&item, itemoff);
-	btrfs_set_extent_owner(&extent_item, BTRFS_FS_TREE_OBJECTID);
-	memcpy(empty_leaf->items + 4, &item, sizeof(item));
-	memcpy(btrfs_leaf_data(empty_leaf) + btrfs_item_offset(&item),
-		&extent_item, btrfs_item_size(&item));
-	ret = pwrite(fd, empty_leaf, blocksize, (start_block + 2) * blocksize);
-	if (ret != blocksize)
-		return -1;
-
-	/* create the inode map */
-	btrfs_set_header_parentid(&empty_leaf->header,
-				  BTRFS_INODE_MAP_OBJECTID);
-	btrfs_set_header_blocknr(&empty_leaf->header, start_block + 3);
-	btrfs_set_header_nritems(&empty_leaf->header, 0);
-	ret = pwrite(fd, empty_leaf, blocksize, (start_block + 3) * blocksize);
-	if (ret != blocksize)
-		return -1;
-
-	/* finally create the FS root */
-	btrfs_set_header_parentid(&empty_leaf->header, BTRFS_FS_TREE_OBJECTID);
-	btrfs_set_header_blocknr(&empty_leaf->header, start_block + 4);
-	btrfs_set_header_nritems(&empty_leaf->header, 0);
-	ret = pwrite(fd, empty_leaf, blocksize, (start_block + 4) * blocksize);
-	if (ret != blocksize)
-		return -1;
-	return 0;
-}
-
-u64 device_size(int fd, struct stat *st)
-{
-	u64 size;
-	if (S_ISREG(st->st_mode)) {
-		return st->st_size;
-	}
-	if (!S_ISBLK(st->st_mode)) {
-		return 0;
-	}
-	if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
-		return size;
-	}
-	return 0;
-}
-
-int main(int ac, char **av)
-{
-	char *file;
-	u64 block_count = 0;
-	int fd;
-	struct stat st;
-	int ret;
-	int i;
-	char *buf = malloc(4096);
-	if (ac >= 2) {
-		file = av[1];
-		if (ac == 3) {
-			block_count = atoi(av[2]);
-			if (!block_count) {
-				fprintf(stderr, "error finding block count\n");
-				exit(1);
-			}
-		}
-	} else {
-		fprintf(stderr, "usage: mkfs.btrfs file [block count]\n");
-		exit(1);
-	}
-	fd = open(file, O_RDWR);
-	if (fd < 0) {
-		fprintf(stderr, "unable to open %s\n", file);
-		exit(1);
-	}
-	ret = fstat(fd, &st);
-	if (ret < 0) {
-		fprintf(stderr, "unable to stat %s\n", file);
-		exit(1);
-	}
-	if (block_count == 0) {
-		block_count = device_size(fd, &st);
-		if (block_count == 0) {
-			fprintf(stderr, "unable to find %s size\n", file);
-			exit(1);
-		}
-	}
-	block_count /= 4096;
-	if (block_count < 256) {
-		fprintf(stderr, "device %s is too small\n", file);
-		exit(1);
-	}
-	memset(buf, 0, 4096);
-	for(i = 0; i < 6; i++) {
-		ret = write(fd, buf, 4096);
-		if (ret != 4096) {
-			fprintf(stderr, "unable to zero fill device\n");
-			exit(1);
-		}
-	}
-	ret = mkfs(fd, block_count, 4096);
-	if (ret) {
-		fprintf(stderr, "error during mkfs %d\n", ret);
-		exit(1);
-	}
-	printf("fs created on %s blocksize %d blocks %Lu\n",
-	       file, 4096, block_count);
-	return 0;
-}
-
diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c
index f53b99da12f3..aa2d3fac8804 100644
--- a/fs/btrfs/print-tree.c
+++ b/fs/btrfs/print-tree.c
@@ -1,7 +1,4 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include "kerncompat.h"
-#include "radix-tree.h"
+#include <linux/module.h>
 #include "ctree.h"
 #include "disk-io.h"
 
@@ -17,14 +14,14 @@ void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l)
 	struct btrfs_inode_item *ii;
 	u32 type;
 
-	printf("leaf %Lu total ptrs %d free space %d\n",
+	printk("leaf %Lu total ptrs %d free space %d\n",
 		btrfs_header_blocknr(&l->header), nr,
 		btrfs_leaf_free_space(root, l));
 	fflush(stdout);
 	for (i = 0 ; i < nr ; i++) {
 		item = l->items + i;
 		type = btrfs_disk_key_type(&item->key);
-		printf("\titem %d key (%Lu %u %Lu) itemoff %d itemsize %d\n",
+		printk("\titem %d key (%Lu %u %Lu) itemoff %d itemsize %d\n",
 			i,
 			btrfs_disk_key_objectid(&item->key),
 			btrfs_disk_key_flags(&item->key),
@@ -34,38 +31,39 @@ void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l)
 		switch (type) {
 		case BTRFS_INODE_ITEM_KEY:
 			ii = btrfs_item_ptr(l, i, struct btrfs_inode_item);
-			printf("\t\tinode generation %Lu size %Lu\n",
+			printk("\t\tinode generation %Lu size %Lu mode %o\n",
 			       btrfs_inode_generation(ii),
-			       btrfs_inode_size(ii));
+			       btrfs_inode_size(ii),
+			       btrfs_inode_mode(ii));
 			break;
 		case BTRFS_DIR_ITEM_KEY:
 			di = btrfs_item_ptr(l, i, struct btrfs_dir_item);
-			printf("\t\tdir oid %Lu flags %u type %u\n",
+			printk("\t\tdir oid %Lu flags %u type %u\n",
 				btrfs_dir_objectid(di),
 				btrfs_dir_flags(di),
 				btrfs_dir_type(di));
-			printf("\t\tname %.*s\n",
+			printk("\t\tname %.*s\n",
 			       btrfs_dir_name_len(di),(char *)(di + 1));
 			break;
 		case BTRFS_ROOT_ITEM_KEY:
 			ri = btrfs_item_ptr(l, i, struct btrfs_root_item);
-			printf("\t\troot data blocknr %Lu refs %u\n",
+			printk("\t\troot data blocknr %Lu refs %u\n",
 				btrfs_root_blocknr(ri), btrfs_root_refs(ri));
 			break;
 		case BTRFS_EXTENT_ITEM_KEY:
 			ei = btrfs_item_ptr(l, i, struct btrfs_extent_item);
-			printf("\t\textent data refs %u owner %Lu\n",
+			printk("\t\textent data refs %u owner %Lu\n",
 				btrfs_extent_refs(ei), btrfs_extent_owner(ei));
 			break;
 		case BTRFS_INODE_MAP_ITEM_KEY:
 			mi = btrfs_item_ptr(l, i, struct btrfs_inode_map_item);
-			printf("\t\tinode map key %Lu %u %Lu\n",
+			printk("\t\tinode map key %Lu %u %Lu\n",
 			       btrfs_disk_key_objectid(&mi->key),
 			       btrfs_disk_key_flags(&mi->key),
 			       btrfs_disk_key_offset(&mi->key));
 			break;
 		case BTRFS_STRING_ITEM_KEY:
-			printf("\t\titem data %.*s\n", btrfs_item_size(item),
+			printk("\t\titem data %.*s\n", btrfs_item_size(item),
 				btrfs_leaf_data(l) + btrfs_item_offset(item));
 			break;
 		};
@@ -86,12 +84,12 @@ void btrfs_print_tree(struct btrfs_root *root, struct btrfs_buffer *t)
 		btrfs_print_leaf(root, (struct btrfs_leaf *)c);
 		return;
 	}
-	printf("node %Lu level %d total ptrs %d free spc %u\n", t->blocknr,
+	printk("node %Lu level %d total ptrs %d free spc %u\n", t->blocknr,
 	        btrfs_header_level(&c->header), nr,
 		(u32)BTRFS_NODEPTRS_PER_BLOCK(root) - nr);
 	fflush(stdout);
 	for (i = 0; i < nr; i++) {
-		printf("\tkey %d (%Lu %u %Lu) block %Lu\n",
+		printk("\tkey %d (%Lu %u %Lu) block %Lu\n",
 		       i,
 		       c->ptrs[i].key.objectid,
 		       c->ptrs[i].key.flags,
diff --git a/fs/btrfs/quick-test.c b/fs/btrfs/quick-test.c
deleted file mode 100644
index ef12cf888783..000000000000
--- a/fs/btrfs/quick-test.c
+++ /dev/null
@@ -1,179 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include "kerncompat.h"
-#include "radix-tree.h"
-#include "ctree.h"
-#include "disk-io.h"
-#include "print-tree.h"
-#include "transaction.h"
-
-/* for testing only */
-int next_key(int i, int max_key) {
-	return rand() % max_key;
-	// return i;
-}
-
-int main(int ac, char **av) {
-	struct btrfs_key ins;
-	struct btrfs_key last = { (u64)-1, 0, 0};
-	char *buf;
-	int i;
-	int num;
-	int ret;
-	int run_size = 100000;
-	int max_key =  100000000;
-	int tree_size = 0;
-	struct btrfs_path path;
-	struct btrfs_super_block super;
-	struct btrfs_root *root;
-	struct btrfs_trans_handle *trans;
-
-	radix_tree_init();
-
-	root = open_ctree("dbfile", &super);
-	trans = btrfs_start_transaction(root, 1);
-	srand(55);
-	ins.flags = 0;
-	btrfs_set_key_type(&ins, BTRFS_STRING_ITEM_KEY);
-	for (i = 0; i < run_size; i++) {
-		buf = malloc(64);
-		num = next_key(i, max_key);
-		// num = i;
-		sprintf(buf, "string-%d", num);
-		if (i % 10000 == 0)
-			fprintf(stderr, "insert %d:%d\n", num, i);
-		ins.objectid = num;
-		ins.offset = 0;
-		ret = btrfs_insert_item(trans, root, &ins, buf, strlen(buf));
-		if (!ret)
-			tree_size++;
-		free(buf);
-		if (i == run_size - 5) {
-			btrfs_commit_transaction(trans, root, &super);
-		}
-
-	}
-	close_ctree(root, &super);
-
-	root = open_ctree("dbfile", &super);
-	printf("starting search\n");
-	srand(55);
-	for (i = 0; i < run_size; i++) {
-		num = next_key(i, max_key);
-		ins.objectid = num;
-		btrfs_init_path(&path);
-		if (i % 10000 == 0)
-			fprintf(stderr, "search %d:%d\n", num, i);
-		ret = btrfs_search_slot(trans, root, &ins, &path, 0, 0);
-		if (ret) {
-			btrfs_print_tree(root, root->node);
-			printf("unable to find %d\n", num);
-			exit(1);
-		}
-		btrfs_release_path(root, &path);
-	}
-	close_ctree(root, &super);
-	root = open_ctree("dbfile", &super);
-	printf("node %p level %d total ptrs %d free spc %lu\n", root->node,
-	        btrfs_header_level(&root->node->node.header),
-		btrfs_header_nritems(&root->node->node.header),
-		BTRFS_NODEPTRS_PER_BLOCK(root) -
-		btrfs_header_nritems(&root->node->node.header));
-	printf("all searches good, deleting some items\n");
-	i = 0;
-	srand(55);
-	for (i = 0 ; i < run_size/4; i++) {
-		num = next_key(i, max_key);
-		ins.objectid = num;
-		btrfs_init_path(&path);
-		ret = btrfs_search_slot(trans, root, &ins, &path, -1, 1);
-		if (!ret) {
-			if (i % 10000 == 0)
-				fprintf(stderr, "del %d:%d\n", num, i);
-			ret = btrfs_del_item(trans, root, &path);
-			if (ret != 0)
-				BUG();
-			tree_size--;
-		}
-		btrfs_release_path(root, &path);
-	}
-	close_ctree(root, &super);
-	root = open_ctree("dbfile", &super);
-	srand(128);
-	for (i = 0; i < run_size; i++) {
-		buf = malloc(64);
-		num = next_key(i, max_key);
-		sprintf(buf, "string-%d", num);
-		ins.objectid = num;
-		if (i % 10000 == 0)
-			fprintf(stderr, "insert %d:%d\n", num, i);
-		ret = btrfs_insert_item(trans, root, &ins, buf, strlen(buf));
-		if (!ret)
-			tree_size++;
-		free(buf);
-	}
-	close_ctree(root, &super);
-	root = open_ctree("dbfile", &super);
-	srand(128);
-	printf("starting search2\n");
-	for (i = 0; i < run_size; i++) {
-		num = next_key(i, max_key);
-		ins.objectid = num;
-		btrfs_init_path(&path);
-		if (i % 10000 == 0)
-			fprintf(stderr, "search %d:%d\n", num, i);
-		ret = btrfs_search_slot(trans, root, &ins, &path, 0, 0);
-		if (ret) {
-			btrfs_print_tree(root, root->node);
-			printf("unable to find %d\n", num);
-			exit(1);
-		}
-		btrfs_release_path(root, &path);
-	}
-	printf("starting big long delete run\n");
-	while(root->node &&
-	      btrfs_header_nritems(&root->node->node.header) > 0) {
-		struct btrfs_leaf *leaf;
-		int slot;
-		ins.objectid = (u64)-1;
-		btrfs_init_path(&path);
-		ret = btrfs_search_slot(trans, root, &ins, &path, -1, 1);
-		if (ret == 0)
-			BUG();
-
-		leaf = &path.nodes[0]->leaf;
-		slot = path.slots[0];
-		if (slot != btrfs_header_nritems(&leaf->header))
-			BUG();
-		while(path.slots[0] > 0) {
-			path.slots[0] -= 1;
-			slot = path.slots[0];
-			leaf = &path.nodes[0]->leaf;
-
-			btrfs_disk_key_to_cpu(&last, &leaf->items[slot].key);
-			if (tree_size % 10000 == 0)
-				printf("big del %d:%d\n", tree_size, i);
-			ret = btrfs_del_item(trans, root, &path);
-			if (ret != 0) {
-				printf("del_item returned %d\n", ret);
-				BUG();
-			}
-			tree_size--;
-		}
-		btrfs_release_path(root, &path);
-	}
-	/*
-	printf("previous tree:\n");
-	btrfs_print_tree(root, root->commit_root);
-	printf("map before commit\n");
-	btrfs_print_tree(root->extent_root, root->extent_root->node);
-	*/
-	btrfs_commit_transaction(trans, root, &super);
-	printf("tree size is now %d\n", tree_size);
-	printf("root %p commit root %p\n", root->node, root->commit_root);
-	printf("map tree\n");
-	btrfs_print_tree(root->fs_info->extent_root,
-			 root->fs_info->extent_root->node);
-	close_ctree(root, &super);
-	return 0;
-}
diff --git a/fs/btrfs/radix-tree.c b/fs/btrfs/radix-tree.c
deleted file mode 100644
index baa25ca1c2ac..000000000000
--- a/fs/btrfs/radix-tree.c
+++ /dev/null
@@ -1,836 +0,0 @@
-/*
- * Copyright (C) 2001 Momchil Velikov
- * Portions Copyright (C) 2001 Christoph Hellwig
- * Copyright (C) 2005 SGI, Christoph Lameter <clameter@sgi.com>
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License as
- * published by the Free Software Foundation; either version 2, or (at
- * your option) any later version.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write to the Free Software
- * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
- */
-
-#include "kerncompat.h"
-#include "radix-tree.h"
-#ifdef __KERNEL__
-#define RADIX_TREE_MAP_SHIFT	(CONFIG_BASE_SMALL ? 4 : 6)
-#else
-#define RADIX_TREE_MAP_SHIFT	3	/* For more stressful testing */
-#endif
-
-#define RADIX_TREE_MAP_SIZE	(1UL << RADIX_TREE_MAP_SHIFT)
-#define RADIX_TREE_MAP_MASK	(RADIX_TREE_MAP_SIZE-1)
-
-#define RADIX_TREE_TAG_LONGS	\
-	((RADIX_TREE_MAP_SIZE + BITS_PER_LONG - 1) / BITS_PER_LONG)
-
-struct radix_tree_node {
-	unsigned int	count;
-	void		*slots[RADIX_TREE_MAP_SIZE];
-	unsigned long	tags[RADIX_TREE_MAX_TAGS][RADIX_TREE_TAG_LONGS];
-};
-
-struct radix_tree_path {
-	struct radix_tree_node *node;
-	int offset;
-};
-
-#define RADIX_TREE_INDEX_BITS  (8 /* CHAR_BIT */ * sizeof(unsigned long))
-#define RADIX_TREE_MAX_PATH (RADIX_TREE_INDEX_BITS/RADIX_TREE_MAP_SHIFT + 2)
-
-static unsigned long height_to_maxindex[RADIX_TREE_MAX_PATH] __read_mostly;
-
-/*
- * Per-cpu pool of preloaded nodes
- */
-struct radix_tree_preload {
-	int nr;
-	struct radix_tree_node *nodes[RADIX_TREE_MAX_PATH];
-};
-struct radix_tree_preload radix_tree_preloads = { 0, };
-
-static inline gfp_t root_gfp_mask(struct radix_tree_root *root)
-{
-	return root->gfp_mask & __GFP_BITS_MASK;
-}
-
-static int internal_nodes = 0;
-/*
- * This assumes that the caller has performed appropriate preallocation, and
- * that the caller has pinned this thread of control to the current CPU.
- */
-static struct radix_tree_node *
-radix_tree_node_alloc(struct radix_tree_root *root)
-{
-	struct radix_tree_node *ret;
-	ret = malloc(sizeof(struct radix_tree_node));
-	if (ret) {
-		memset(ret, 0, sizeof(struct radix_tree_node));
-		internal_nodes++;
-	}
-	return ret;
-}
-
-static inline void
-radix_tree_node_free(struct radix_tree_node *node)
-{
-	internal_nodes--;
-	free(node);
-}
-
-/*
- * Load up this CPU's radix_tree_node buffer with sufficient objects to
- * ensure that the addition of a single element in the tree cannot fail.  On
- * success, return zero, with preemption disabled.  On error, return -ENOMEM
- * with preemption not disabled.
- */
-int radix_tree_preload(gfp_t gfp_mask)
-{
-	struct radix_tree_preload *rtp;
-	struct radix_tree_node *node;
-	int ret = -ENOMEM;
-
-	preempt_disable();
-	rtp = &__get_cpu_var(radix_tree_preloads);
-	while (rtp->nr < ARRAY_SIZE(rtp->nodes)) {
-		preempt_enable();
-		node = radix_tree_node_alloc(NULL);
-		if (node == NULL)
-			goto out;
-		preempt_disable();
-		rtp = &__get_cpu_var(radix_tree_preloads);
-		if (rtp->nr < ARRAY_SIZE(rtp->nodes))
-			rtp->nodes[rtp->nr++] = node;
-		else
-			radix_tree_node_free(node);
-	}
-	ret = 0;
-out:
-	return ret;
-}
-
-static inline void tag_set(struct radix_tree_node *node, unsigned int tag,
-		int offset)
-{
-	__set_bit(offset, node->tags[tag]);
-}
-
-static inline void tag_clear(struct radix_tree_node *node, unsigned int tag,
-		int offset)
-{
-	__clear_bit(offset, node->tags[tag]);
-}
-
-static inline int tag_get(struct radix_tree_node *node, unsigned int tag,
-		int offset)
-{
-	return test_bit(offset, node->tags[tag]);
-}
-
-static inline void root_tag_set(struct radix_tree_root *root, unsigned int tag)
-{
-	root->gfp_mask |= (__force gfp_t)(1 << (tag + __GFP_BITS_SHIFT));
-}
-
-
-static inline void root_tag_clear(struct radix_tree_root *root, unsigned int tag)
-{
-	root->gfp_mask &= (__force gfp_t)~(1 << (tag + __GFP_BITS_SHIFT));
-}
-
-static inline void root_tag_clear_all(struct radix_tree_root *root)
-{
-	root->gfp_mask &= __GFP_BITS_MASK;
-}
-
-static inline int root_tag_get(struct radix_tree_root *root, unsigned int tag)
-{
-	return (__force unsigned)root->gfp_mask & (1 << (tag + __GFP_BITS_SHIFT));
-}
-
-/*
- * Returns 1 if any slot in the node has this tag set.
- * Otherwise returns 0.
- */
-static inline int any_tag_set(struct radix_tree_node *node, unsigned int tag)
-{
-	int idx;
-	for (idx = 0; idx < RADIX_TREE_TAG_LONGS; idx++) {
-		if (node->tags[tag][idx])
-			return 1;
-	}
-	return 0;
-}
-
-/*
- *	Return the maximum key which can be store into a
- *	radix tree with height HEIGHT.
- */
-static inline unsigned long radix_tree_maxindex(unsigned int height)
-{
-	return height_to_maxindex[height];
-}
-
-/*
- *	Extend a radix tree so it can store key @index.
- */
-static int radix_tree_extend(struct radix_tree_root *root, unsigned long index)
-{
-	struct radix_tree_node *node;
-	unsigned int height;
-	int tag;
-
-	/* Figure out what the height should be.  */
-	height = root->height + 1;
-	while (index > radix_tree_maxindex(height))
-		height++;
-
-	if (root->rnode == NULL) {
-		root->height = height;
-		goto out;
-	}
-
-	do {
-		if (!(node = radix_tree_node_alloc(root)))
-			return -ENOMEM;
-
-		/* Increase the height.  */
-		node->slots[0] = root->rnode;
-
-		/* Propagate the aggregated tag info into the new root */
-		for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++) {
-			if (root_tag_get(root, tag))
-				tag_set(node, tag, 0);
-		}
-
-		node->count = 1;
-		root->rnode = node;
-		root->height++;
-	} while (height > root->height);
-out:
-	return 0;
-}
-
-/**
- *	radix_tree_insert    -    insert into a radix tree
- *	@root:		radix tree root
- *	@index:		index key
- *	@item:		item to insert
- *
- *	Insert an item into the radix tree at position @index.
- */
-int radix_tree_insert(struct radix_tree_root *root,
-			unsigned long index, void *item)
-{
-	struct radix_tree_node *node = NULL, *slot;
-	unsigned int height, shift;
-	int offset;
-	int error;
-
-	/* Make sure the tree is high enough.  */
-	if (index > radix_tree_maxindex(root->height)) {
-		error = radix_tree_extend(root, index);
-		if (error)
-			return error;
-	}
-
-	slot = root->rnode;
-	height = root->height;
-	shift = (height-1) * RADIX_TREE_MAP_SHIFT;
-
-	offset = 0;			/* uninitialised var warning */
-	while (height > 0) {
-		if (slot == NULL) {
-			/* Have to add a child node.  */
-			if (!(slot = radix_tree_node_alloc(root)))
-				return -ENOMEM;
-			if (node) {
-				node->slots[offset] = slot;
-				node->count++;
-			} else
-				root->rnode = slot;
-		}
-
-		/* Go a level down */
-		offset = (index >> shift) & RADIX_TREE_MAP_MASK;
-		node = slot;
-		slot = node->slots[offset];
-		shift -= RADIX_TREE_MAP_SHIFT;
-		height--;
-	}
-
-	if (slot != NULL)
-		return -EEXIST;
-
-	if (node) {
-		node->count++;
-		node->slots[offset] = item;
-		BUG_ON(tag_get(node, 0, offset));
-		BUG_ON(tag_get(node, 1, offset));
-	} else {
-		root->rnode = item;
-		BUG_ON(root_tag_get(root, 0));
-		BUG_ON(root_tag_get(root, 1));
-	}
-
-	return 0;
-}
-
-static inline void **__lookup_slot(struct radix_tree_root *root,
-				   unsigned long index)
-{
-	unsigned int height, shift;
-	struct radix_tree_node **slot;
-
-	height = root->height;
-
-	if (index > radix_tree_maxindex(height))
-		return NULL;
-
-	if (height == 0 && root->rnode)
-		return (void **)&root->rnode;
-
-	shift = (height-1) * RADIX_TREE_MAP_SHIFT;
-	slot = &root->rnode;
-
-	while (height > 0) {
-		if (*slot == NULL)
-			return NULL;
-
-		slot = (struct radix_tree_node **)
-			((*slot)->slots +
-				((index >> shift) & RADIX_TREE_MAP_MASK));
-		shift -= RADIX_TREE_MAP_SHIFT;
-		height--;
-	}
-
-	return (void **)slot;
-}
-
-/**
- *	radix_tree_lookup_slot    -    lookup a slot in a radix tree
- *	@root:		radix tree root
- *	@index:		index key
- *
- *	Lookup the slot corresponding to the position @index in the radix tree
- *	@root. This is useful for update-if-exists operations.
- */
-void **radix_tree_lookup_slot(struct radix_tree_root *root, unsigned long index)
-{
-	return __lookup_slot(root, index);
-}
-
-/**
- *	radix_tree_lookup    -    perform lookup operation on a radix tree
- *	@root:		radix tree root
- *	@index:		index key
- *
- *	Lookup the item at the position @index in the radix tree @root.
- */
-void *radix_tree_lookup(struct radix_tree_root *root, unsigned long index)
-{
-	void **slot;
-
-	slot = __lookup_slot(root, index);
-	return slot != NULL ? *slot : NULL;
-}
-
-/**
- *	radix_tree_tag_set - set a tag on a radix tree node
- *	@root:		radix tree root
- *	@index:		index key
- *	@tag: 		tag index
- *
- *	Set the search tag (which must be < RADIX_TREE_MAX_TAGS)
- *	corresponding to @index in the radix tree.  From
- *	the root all the way down to the leaf node.
- *
- *	Returns the address of the tagged item.   Setting a tag on a not-present
- *	item is a bug.
- */
-void *radix_tree_tag_set(struct radix_tree_root *root,
-			unsigned long index, unsigned int tag)
-{
-	unsigned int height, shift;
-	struct radix_tree_node *slot;
-
-	height = root->height;
-	BUG_ON(index > radix_tree_maxindex(height));
-
-	slot = root->rnode;
-	shift = (height - 1) * RADIX_TREE_MAP_SHIFT;
-
-	while (height > 0) {
-		int offset;
-
-		offset = (index >> shift) & RADIX_TREE_MAP_MASK;
-		if (!tag_get(slot, tag, offset))
-			tag_set(slot, tag, offset);
-		slot = slot->slots[offset];
-		BUG_ON(slot == NULL);
-		shift -= RADIX_TREE_MAP_SHIFT;
-		height--;
-	}
-
-	/* set the root's tag bit */
-	if (slot && !root_tag_get(root, tag))
-		root_tag_set(root, tag);
-
-	return slot;
-}
-
-/**
- *	radix_tree_tag_clear - clear a tag on a radix tree node
- *	@root:		radix tree root
- *	@index:		index key
- *	@tag: 		tag index
- *
- *	Clear the search tag (which must be < RADIX_TREE_MAX_TAGS)
- *	corresponding to @index in the radix tree.  If
- *	this causes the leaf node to have no tags set then clear the tag in the
- *	next-to-leaf node, etc.
- *
- *	Returns the address of the tagged item on success, else NULL.  ie:
- *	has the same return value and semantics as radix_tree_lookup().
- */
-void *radix_tree_tag_clear(struct radix_tree_root *root,
-			unsigned long index, unsigned int tag)
-{
-	struct radix_tree_path path[RADIX_TREE_MAX_PATH], *pathp = path;
-	struct radix_tree_node *slot = NULL;
-	unsigned int height, shift;
-
-	height = root->height;
-	if (index > radix_tree_maxindex(height))
-		goto out;
-
-	shift = (height - 1) * RADIX_TREE_MAP_SHIFT;
-	pathp->node = NULL;
-	slot = root->rnode;
-
-	while (height > 0) {
-		int offset;
-
-		if (slot == NULL)
-			goto out;
-
-		offset = (index >> shift) & RADIX_TREE_MAP_MASK;
-		pathp[1].offset = offset;
-		pathp[1].node = slot;
-		slot = slot->slots[offset];
-		pathp++;
-		shift -= RADIX_TREE_MAP_SHIFT;
-		height--;
-	}
-
-	if (slot == NULL)
-		goto out;
-
-	while (pathp->node) {
-		if (!tag_get(pathp->node, tag, pathp->offset))
-			goto out;
-		tag_clear(pathp->node, tag, pathp->offset);
-		if (any_tag_set(pathp->node, tag))
-			goto out;
-		pathp--;
-	}
-
-	/* clear the root's tag bit */
-	if (root_tag_get(root, tag))
-		root_tag_clear(root, tag);
-
-out:
-	return slot;
-}
-
-#ifndef __KERNEL__	/* Only the test harness uses this at present */
-/**
- * radix_tree_tag_get - get a tag on a radix tree node
- * @root:		radix tree root
- * @index:		index key
- * @tag: 		tag index (< RADIX_TREE_MAX_TAGS)
- *
- * Return values:
- *
- *  0: tag not present or not set
- *  1: tag set
- */
-int radix_tree_tag_get(struct radix_tree_root *root,
-			unsigned long index, unsigned int tag)
-{
-	unsigned int height, shift;
-	struct radix_tree_node *slot;
-	int saw_unset_tag = 0;
-
-	height = root->height;
-	if (index > radix_tree_maxindex(height))
-		return 0;
-
-	/* check the root's tag bit */
-	if (!root_tag_get(root, tag))
-		return 0;
-
-	if (height == 0)
-		return 1;
-
-	shift = (height - 1) * RADIX_TREE_MAP_SHIFT;
-	slot = root->rnode;
-
-	for ( ; ; ) {
-		int offset;
-
-		if (slot == NULL)
-			return 0;
-
-		offset = (index >> shift) & RADIX_TREE_MAP_MASK;
-
-		/*
-		 * This is just a debug check.  Later, we can bale as soon as
-		 * we see an unset tag.
-		 */
-		if (!tag_get(slot, tag, offset))
-			saw_unset_tag = 1;
-		if (height == 1) {
-			int ret = tag_get(slot, tag, offset);
-
-			BUG_ON(ret && saw_unset_tag);
-			return !!ret;
-		}
-		slot = slot->slots[offset];
-		shift -= RADIX_TREE_MAP_SHIFT;
-		height--;
-	}
-}
-#endif
-
-static unsigned int
-__lookup(struct radix_tree_root *root, void **results, unsigned long index,
-	unsigned int max_items, unsigned long *next_index)
-{
-	unsigned int nr_found = 0;
-	unsigned int shift, height;
-	struct radix_tree_node *slot;
-	unsigned long i;
-
-	height = root->height;
-	if (height == 0) {
-		if (root->rnode && index == 0)
-			results[nr_found++] = root->rnode;
-		goto out;
-	}
-
-	shift = (height-1) * RADIX_TREE_MAP_SHIFT;
-	slot = root->rnode;
-
-	for ( ; height > 1; height--) {
-
-		for (i = (index >> shift) & RADIX_TREE_MAP_MASK ;
-				i < RADIX_TREE_MAP_SIZE; i++) {
-			if (slot->slots[i] != NULL)
-				break;
-			index &= ~((1UL << shift) - 1);
-			index += 1UL << shift;
-			if (index == 0)
-				goto out;	/* 32-bit wraparound */
-		}
-		if (i == RADIX_TREE_MAP_SIZE)
-			goto out;
-
-		shift -= RADIX_TREE_MAP_SHIFT;
-		slot = slot->slots[i];
-	}
-
-	/* Bottom level: grab some items */
-	for (i = index & RADIX_TREE_MAP_MASK; i < RADIX_TREE_MAP_SIZE; i++) {
-		index++;
-		if (slot->slots[i]) {
-			results[nr_found++] = slot->slots[i];
-			if (nr_found == max_items)
-				goto out;
-		}
-	}
-out:
-	*next_index = index;
-	return nr_found;
-}
-
-/**
- *	radix_tree_gang_lookup - perform multiple lookup on a radix tree
- *	@root:		radix tree root
- *	@results:	where the results of the lookup are placed
- *	@first_index:	start the lookup from this key
- *	@max_items:	place up to this many items at *results
- *
- *	Performs an index-ascending scan of the tree for present items.  Places
- *	them at *@results and returns the number of items which were placed at
- *	*@results.
- *
- *	The implementation is naive.
- */
-unsigned int
-radix_tree_gang_lookup(struct radix_tree_root *root, void **results,
-			unsigned long first_index, unsigned int max_items)
-{
-	const unsigned long max_index = radix_tree_maxindex(root->height);
-	unsigned long cur_index = first_index;
-	unsigned int ret = 0;
-
-	while (ret < max_items) {
-		unsigned int nr_found;
-		unsigned long next_index;	/* Index of next search */
-
-		if (cur_index > max_index)
-			break;
-		nr_found = __lookup(root, results + ret, cur_index,
-					max_items - ret, &next_index);
-		ret += nr_found;
-		if (next_index == 0)
-			break;
-		cur_index = next_index;
-	}
-	return ret;
-}
-
-/*
- * FIXME: the two tag_get()s here should use find_next_bit() instead of
- * open-coding the search.
- */
-static unsigned int
-__lookup_tag(struct radix_tree_root *root, void **results, unsigned long index,
-	unsigned int max_items, unsigned long *next_index, unsigned int tag)
-{
-	unsigned int nr_found = 0;
-	unsigned int shift;
-	unsigned int height = root->height;
-	struct radix_tree_node *slot;
-
-	if (height == 0) {
-		if (root->rnode && index == 0)
-			results[nr_found++] = root->rnode;
-		goto out;
-	}
-
-	shift = (height - 1) * RADIX_TREE_MAP_SHIFT;
-	slot = root->rnode;
-
-	do {
-		unsigned long i = (index >> shift) & RADIX_TREE_MAP_MASK;
-
-		for ( ; i < RADIX_TREE_MAP_SIZE; i++) {
-			if (tag_get(slot, tag, i)) {
-				BUG_ON(slot->slots[i] == NULL);
-				break;
-			}
-			index &= ~((1UL << shift) - 1);
-			index += 1UL << shift;
-			if (index == 0)
-				goto out;	/* 32-bit wraparound */
-		}
-		if (i == RADIX_TREE_MAP_SIZE)
-			goto out;
-		height--;
-		if (height == 0) {	/* Bottom level: grab some items */
-			unsigned long j = index & RADIX_TREE_MAP_MASK;
-
-			for ( ; j < RADIX_TREE_MAP_SIZE; j++) {
-				index++;
-				if (tag_get(slot, tag, j)) {
-					BUG_ON(slot->slots[j] == NULL);
-					results[nr_found++] = slot->slots[j];
-					if (nr_found == max_items)
-						goto out;
-				}
-			}
-		}
-		shift -= RADIX_TREE_MAP_SHIFT;
-		slot = slot->slots[i];
-	} while (height > 0);
-out:
-	*next_index = index;
-	return nr_found;
-}
-
-/**
- *	radix_tree_gang_lookup_tag - perform multiple lookup on a radix tree
- *	                             based on a tag
- *	@root:		radix tree root
- *	@results:	where the results of the lookup are placed
- *	@first_index:	start the lookup from this key
- *	@max_items:	place up to this many items at *results
- *	@tag:		the tag index (< RADIX_TREE_MAX_TAGS)
- *
- *	Performs an index-ascending scan of the tree for present items which
- *	have the tag indexed by @tag set.  Places the items at *@results and
- *	returns the number of items which were placed at *@results.
- */
-unsigned int
-radix_tree_gang_lookup_tag(struct radix_tree_root *root, void **results,
-		unsigned long first_index, unsigned int max_items,
-		unsigned int tag)
-{
-	const unsigned long max_index = radix_tree_maxindex(root->height);
-	unsigned long cur_index = first_index;
-	unsigned int ret = 0;
-
-	/* check the root's tag bit */
-	if (!root_tag_get(root, tag))
-		return 0;
-
-	while (ret < max_items) {
-		unsigned int nr_found;
-		unsigned long next_index;	/* Index of next search */
-
-		if (cur_index > max_index)
-			break;
-		nr_found = __lookup_tag(root, results + ret, cur_index,
-					max_items - ret, &next_index, tag);
-		ret += nr_found;
-		if (next_index == 0)
-			break;
-		cur_index = next_index;
-	}
-	return ret;
-}
-
-/**
- *	radix_tree_shrink    -    shrink height of a radix tree to minimal
- *	@root		radix tree root
- */
-static inline void radix_tree_shrink(struct radix_tree_root *root)
-{
-	/* try to shrink tree height */
-	while (root->height > 0 &&
-			root->rnode->count == 1 &&
-			root->rnode->slots[0]) {
-		struct radix_tree_node *to_free = root->rnode;
-
-		root->rnode = to_free->slots[0];
-		root->height--;
-		/* must only free zeroed nodes into the slab */
-		tag_clear(to_free, 0, 0);
-		tag_clear(to_free, 1, 0);
-		to_free->slots[0] = NULL;
-		to_free->count = 0;
-		radix_tree_node_free(to_free);
-	}
-}
-
-/**
- *	radix_tree_delete    -    delete an item from a radix tree
- *	@root:		radix tree root
- *	@index:		index key
- *
- *	Remove the item at @index from the radix tree rooted at @root.
- *
- *	Returns the address of the deleted item, or NULL if it was not present.
- */
-void *radix_tree_delete(struct radix_tree_root *root, unsigned long index)
-{
-	struct radix_tree_path path[RADIX_TREE_MAX_PATH], *pathp = path;
-	struct radix_tree_node *slot = NULL;
-	unsigned int height, shift;
-	int tag;
-	int offset;
-
-	height = root->height;
-	if (index > radix_tree_maxindex(height))
-		goto out;
-
-	slot = root->rnode;
-	if (height == 0 && root->rnode) {
-		root_tag_clear_all(root);
-		root->rnode = NULL;
-		goto out;
-	}
-
-	shift = (height - 1) * RADIX_TREE_MAP_SHIFT;
-	pathp->node = NULL;
-
-	do {
-		if (slot == NULL)
-			goto out;
-
-		pathp++;
-		offset = (index >> shift) & RADIX_TREE_MAP_MASK;
-		pathp->offset = offset;
-		pathp->node = slot;
-		slot = slot->slots[offset];
-		shift -= RADIX_TREE_MAP_SHIFT;
-		height--;
-	} while (height > 0);
-
-	if (slot == NULL)
-		goto out;
-
-	/*
-	 * Clear all tags associated with the just-deleted item
-	 */
-	for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++) {
-		if (tag_get(pathp->node, tag, pathp->offset))
-			radix_tree_tag_clear(root, index, tag);
-	}
-
-	/* Now free the nodes we do not need anymore */
-	while (pathp->node) {
-		pathp->node->slots[pathp->offset] = NULL;
-		pathp->node->count--;
-
-		if (pathp->node->count) {
-			if (pathp->node == root->rnode)
-				radix_tree_shrink(root);
-			goto out;
-		}
-
-		/* Node with zero slots in use so free it */
-		radix_tree_node_free(pathp->node);
-
-		pathp--;
-	}
-	root_tag_clear_all(root);
-	root->height = 0;
-	root->rnode = NULL;
-
-out:
-	return slot;
-}
-
-/**
- *	radix_tree_tagged - test whether any items in the tree are tagged
- *	@root:		radix tree root
- *	@tag:		tag to test
- */
-int radix_tree_tagged(struct radix_tree_root *root, unsigned int tag)
-{
-	return root_tag_get(root, tag);
-}
-
-static unsigned long __maxindex(unsigned int height)
-{
-	unsigned int tmp = height * RADIX_TREE_MAP_SHIFT;
-	unsigned long index = (~0UL >> (RADIX_TREE_INDEX_BITS - tmp - 1)) >> 1;
-
-	if (tmp >= RADIX_TREE_INDEX_BITS)
-		index = ~0UL;
-	return index;
-}
-
-static void radix_tree_init_maxindex(void)
-{
-	unsigned int i;
-
-	for (i = 0; i < ARRAY_SIZE(height_to_maxindex); i++)
-		height_to_maxindex[i] = __maxindex(i);
-}
-
-void radix_tree_init(void)
-{
-	radix_tree_init_maxindex();
-}
diff --git a/fs/btrfs/radix-tree.h b/fs/btrfs/radix-tree.h
deleted file mode 100644
index c3ce88137f77..000000000000
--- a/fs/btrfs/radix-tree.h
+++ /dev/null
@@ -1,73 +0,0 @@
-/*
- * Copyright (C) 2001 Momchil Velikov
- * Portions Copyright (C) 2001 Christoph Hellwig
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License as
- * published by the Free Software Foundation; either version 2, or (at
- * your option) any later version.
- * 
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- * 
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write to the Free Software
- * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
- */
-#ifndef _LINUX_RADIX_TREE_H
-#define _LINUX_RADIX_TREE_H
-
-#define RADIX_TREE_MAX_TAGS 2
-
-/* root tags are stored in gfp_mask, shifted by __GFP_BITS_SHIFT */
-struct radix_tree_root {
-	unsigned int		height;
-	gfp_t			gfp_mask;
-	struct radix_tree_node	*rnode;
-};
-
-#define RADIX_TREE_INIT(mask)	{					\
-	.height = 0,							\
-	.gfp_mask = (mask),						\
-	.rnode = NULL,							\
-}
-
-#define RADIX_TREE(name, mask) \
-	struct radix_tree_root name = RADIX_TREE_INIT(mask)
-
-#define INIT_RADIX_TREE(root, mask)					\
-do {									\
-	(root)->height = 0;						\
-	(root)->gfp_mask = (mask);					\
-	(root)->rnode = NULL;						\
-} while (0)
-
-int radix_tree_insert(struct radix_tree_root *, unsigned long, void *);
-void *radix_tree_lookup(struct radix_tree_root *, unsigned long);
-void **radix_tree_lookup_slot(struct radix_tree_root *, unsigned long);
-void *radix_tree_delete(struct radix_tree_root *, unsigned long);
-unsigned int
-radix_tree_gang_lookup(struct radix_tree_root *root, void **results,
-			unsigned long first_index, unsigned int max_items);
-int radix_tree_preload(gfp_t gfp_mask);
-void radix_tree_init(void);
-void *radix_tree_tag_set(struct radix_tree_root *root,
-			unsigned long index, unsigned int tag);
-void *radix_tree_tag_clear(struct radix_tree_root *root,
-			unsigned long index, unsigned int tag);
-int radix_tree_tag_get(struct radix_tree_root *root,
-			unsigned long index, unsigned int tag);
-unsigned int
-radix_tree_gang_lookup_tag(struct radix_tree_root *root, void **results,
-		unsigned long first_index, unsigned int max_items,
-		unsigned int tag);
-int radix_tree_tagged(struct radix_tree_root *root, unsigned int tag);
-
-static inline void radix_tree_preload_end(void)
-{
-	preempt_enable();
-}
-
-#endif /* _LINUX_RADIX_TREE_H */
diff --git a/fs/btrfs/random-test.c b/fs/btrfs/random-test.c
deleted file mode 100644
index 3a38ae7a886d..000000000000
--- a/fs/btrfs/random-test.c
+++ /dev/null
@@ -1,405 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <signal.h>
-#include "kerncompat.h"
-#include "radix-tree.h"
-#include "ctree.h"
-#include "disk-io.h"
-#include "print-tree.h"
-#include "transaction.h"
-
-int keep_running = 1;
-struct btrfs_super_block super;
-
-static int setup_key(struct radix_tree_root *root, struct btrfs_key *key,
-		     int exists)
-{
-	int num = rand();
-	unsigned long res[2];
-	int ret;
-
-	key->flags = 0;
-	btrfs_set_key_type(key, BTRFS_STRING_ITEM_KEY);
-	key->offset = 0;
-again:
-	ret = radix_tree_gang_lookup(root, (void **)res, num, 2);
-	if (exists) {
-		if (ret == 0)
-			return -1;
-		num = res[0];
-	} else if (ret != 0 && num == res[0]) {
-		num++;
-		if (ret > 1 && num == res[1]) {
-			num++;
-			goto again;
-		}
-	}
-	key->objectid = num;
-	return 0;
-}
-
-static int ins_one(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		   struct radix_tree_root *radix)
-{
-	struct btrfs_path path;
-	struct btrfs_key key;
-	int ret;
-	char buf[128];
-	unsigned long oid;
-	btrfs_init_path(&path);
-	ret = setup_key(radix, &key, 0);
-	sprintf(buf, "str-%Lu\n", key.objectid);
-	ret = btrfs_insert_item(trans, root, &key, buf, strlen(buf));
-	if (ret)
-		goto error;
-	oid = (unsigned long)key.objectid;
-	radix_tree_preload(GFP_KERNEL);
-	ret = radix_tree_insert(radix, oid, (void *)oid);
-	radix_tree_preload_end();
-	if (ret)
-		goto error;
-	return ret;
-error:
-	printf("failed to insert %Lu\n", key.objectid);
-	return -1;
-}
-
-static int insert_dup(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, struct radix_tree_root *radix)
-{
-	struct btrfs_path path;
-	struct btrfs_key key;
-	int ret;
-	char buf[128];
-	btrfs_init_path(&path);
-	ret = setup_key(radix, &key, 1);
-	if (ret < 0)
-		return 0;
-	sprintf(buf, "str-%Lu\n", key.objectid);
-	ret = btrfs_insert_item(trans, root, &key, buf, strlen(buf));
-	if (ret != -EEXIST) {
-		printf("insert on %Lu gave us %d\n", key.objectid, ret);
-		return 1;
-	}
-	return 0;
-}
-
-static int del_one(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		   struct radix_tree_root *radix)
-{
-	struct btrfs_path path;
-	struct btrfs_key key;
-	int ret;
-	unsigned long *ptr;
-	btrfs_init_path(&path);
-	ret = setup_key(radix, &key, 1);
-	if (ret < 0)
-		return 0;
-	ret = btrfs_search_slot(trans, root, &key, &path, -1, 1);
-	if (ret)
-		goto error;
-	ret = btrfs_del_item(trans, root, &path);
-	btrfs_release_path(root, &path);
-	if (ret != 0)
-		goto error;
-	ptr = radix_tree_delete(radix, key.objectid);
-	if (!ptr)
-		goto error;
-	return 0;
-error:
-	printf("failed to delete %Lu\n", key.objectid);
-	return -1;
-}
-
-static int lookup_item(struct btrfs_trans_handle *trans, struct btrfs_root
-		       *root, struct radix_tree_root *radix)
-{
-	struct btrfs_path path;
-	struct btrfs_key key;
-	int ret;
-	btrfs_init_path(&path);
-	ret = setup_key(radix, &key, 1);
-	if (ret < 0)
-		return 0;
-	ret = btrfs_search_slot(trans, root, &key, &path, 0, 1);
-	btrfs_release_path(root, &path);
-	if (ret)
-		goto error;
-	return 0;
-error:
-	printf("unable to find key %Lu\n", key.objectid);
-	return -1;
-}
-
-static int lookup_enoent(struct btrfs_trans_handle *trans, struct btrfs_root
-			 *root, struct radix_tree_root *radix)
-{
-	struct btrfs_path path;
-	struct btrfs_key key;
-	int ret;
-	btrfs_init_path(&path);
-	ret = setup_key(radix, &key, 0);
-	if (ret < 0)
-		return ret;
-	ret = btrfs_search_slot(trans, root, &key, &path, 0, 0);
-	btrfs_release_path(root, &path);
-	if (ret <= 0)
-		goto error;
-	return 0;
-error:
-	printf("able to find key that should not exist %Lu\n", key.objectid);
-	return -1;
-}
-
-static int empty_tree(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, struct radix_tree_root *radix, int nr)
-{
-	struct btrfs_path path;
-	struct btrfs_key key;
-	unsigned long found = 0;
-	int ret;
-	int slot;
-	int *ptr;
-	int count = 0;
-
-	key.offset = 0;
-	key.flags = 0;
-	btrfs_set_key_type(&key, BTRFS_STRING_ITEM_KEY);
-	key.objectid = (unsigned long)-1;
-	while(nr-- >= 0) {
-		btrfs_init_path(&path);
-		ret = btrfs_search_slot(trans, root, &key, &path, -1, 1);
-		if (ret < 0) {
-			btrfs_release_path(root, &path);
-			return ret;
-		}
-		if (ret != 0) {
-			if (path.slots[0] == 0) {
-				btrfs_release_path(root, &path);
-				break;
-			}
-			path.slots[0] -= 1;
-		}
-		slot = path.slots[0];
-		found = btrfs_disk_key_objectid(
-					&path.nodes[0]->leaf.items[slot].key);
-		ret = btrfs_del_item(trans, root, &path);
-		count++;
-		if (ret) {
-			fprintf(stderr,
-				"failed to remove %lu from tree\n",
-				found);
-			return -1;
-		}
-		btrfs_release_path(root, &path);
-		ptr = radix_tree_delete(radix, found);
-		if (!ptr)
-			goto error;
-		if (!keep_running)
-			break;
-	}
-	return 0;
-error:
-	fprintf(stderr, "failed to delete from the radix %lu\n", found);
-	return -1;
-}
-
-static int fill_tree(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		     struct radix_tree_root *radix, int count)
-{
-	int i;
-	int ret = 0;
-	for (i = 0; i < count; i++) {
-		ret = ins_one(trans, root, radix);
-		if (ret) {
-			fprintf(stderr, "fill failed\n");
-			goto out;
-		}
-		if (i % 1000 == 0) {
-			ret = btrfs_commit_transaction(trans, root, &super);
-			if (ret) {
-				fprintf(stderr, "fill commit failed\n");
-				return ret;
-			}
-		}
-		if (i && i % 10000 == 0) {
-			printf("bigfill %d\n", i);
-		}
-		if (!keep_running)
-			break;
-	}
-out:
-	return ret;
-}
-
-static int bulk_op(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		   struct radix_tree_root *radix)
-{
-	int ret;
-	int nr = rand() % 5000;
-	static int run_nr = 0;
-
-	/* do the bulk op much less frequently */
-	if (run_nr++ % 100)
-		return 0;
-	ret = empty_tree(trans, root, radix, nr);
-	if (ret)
-		return ret;
-	ret = fill_tree(trans, root, radix, nr);
-	if (ret)
-		return ret;
-	return 0;
-}
-
-
-int (*ops[])(struct btrfs_trans_handle *,
-	     struct btrfs_root *root, struct radix_tree_root *radix) =
-	{ ins_one, insert_dup, del_one, lookup_item,
-	  lookup_enoent, bulk_op };
-
-static int fill_radix(struct btrfs_root *root, struct radix_tree_root *radix)
-{
-	struct btrfs_path path;
-	struct btrfs_key key;
-	unsigned long found;
-	int ret;
-	int slot;
-	int i;
-
-	key.offset = 0;
-	key.flags = 0;
-	btrfs_set_key_type(&key, BTRFS_STRING_ITEM_KEY);
-	key.objectid = (unsigned long)-1;
-	while(1) {
-		btrfs_init_path(&path);
-		ret = btrfs_search_slot(NULL, root, &key, &path, 0, 0);
-		if (ret < 0) {
-			btrfs_release_path(root, &path);
-			return ret;
-		}
-		slot = path.slots[0];
-		if (ret != 0) {
-			if (slot == 0) {
-				btrfs_release_path(root, &path);
-				break;
-			}
-			slot -= 1;
-		}
-		for (i = slot; i >= 0; i--) {
-			found = btrfs_disk_key_objectid(&path.nodes[0]->
-							leaf.items[i].key);
-			radix_tree_preload(GFP_KERNEL);
-			ret = radix_tree_insert(radix, found, (void *)found);
-			if (ret) {
-				fprintf(stderr,
-					"failed to insert %lu into radix\n",
-					found);
-				exit(1);
-			}
-
-			radix_tree_preload_end();
-		}
-		btrfs_release_path(root, &path);
-		key.objectid = found - 1;
-		if (key.objectid > found)
-			break;
-	}
-	return 0;
-}
-void sigstopper(int ignored)
-{
-	keep_running = 0;
-	fprintf(stderr, "caught exit signal, stopping\n");
-}
-
-int print_usage(void)
-{
-	printf("usage: tester [-ih] [-c count] [-f count]\n");
-	printf("\t -c count -- iteration count after filling\n");
-	printf("\t -f count -- run this many random inserts before starting\n");
-	printf("\t -i       -- only do initial fill\n");
-	printf("\t -h       -- this help text\n");
-	exit(1);
-}
-int main(int ac, char **av)
-{
-	RADIX_TREE(radix, GFP_KERNEL);
-	struct btrfs_root *root;
-	int i;
-	int ret;
-	int count;
-	int op;
-	int iterations = 20000;
-	int init_fill_count = 800000;
-	int err = 0;
-	int initial_only = 0;
-	struct btrfs_trans_handle *trans;
-	radix_tree_init();
-	root = open_ctree("dbfile", &super);
-	fill_radix(root, &radix);
-
-	signal(SIGTERM, sigstopper);
-	signal(SIGINT, sigstopper);
-
-	for (i = 1 ; i < ac ; i++) {
-		if (strcmp(av[i], "-i") == 0) {
-			initial_only = 1;
-		} else if (strcmp(av[i], "-c") == 0) {
-			iterations = atoi(av[i+1]);
-			i++;
-		} else if (strcmp(av[i], "-f") == 0) {
-			init_fill_count = atoi(av[i+1]);
-			i++;
-		} else {
-			print_usage();
-		}
-	}
-	printf("initial fill\n");
-	trans = btrfs_start_transaction(root, 1);
-	ret = fill_tree(trans, root, &radix, init_fill_count);
-	printf("starting run\n");
-	if (ret) {
-		err = ret;
-		goto out;
-	}
-	if (initial_only == 1) {
-		goto out;
-	}
-	for (i = 0; i < iterations; i++) {
-		op = rand() % ARRAY_SIZE(ops);
-		count = rand() % 128;
-		if (i % 2000 == 0) {
-			printf("%d\n", i);
-			fflush(stdout);
-		}
-		if (i && i % 5000 == 0) {
-			printf("open & close, root level %d nritems %d\n",
-				btrfs_header_level(&root->node->node.header),
-				btrfs_header_nritems(&root->node->node.header));
-			close_ctree(root, &super);
-			root = open_ctree("dbfile", &super);
-		}
-		while(count--) {
-			ret = ops[op](trans, root, &radix);
-			if (ret) {
-				fprintf(stderr, "op %d failed %d:%d\n",
-					op, i, iterations);
-				btrfs_print_tree(root, root->node);
-				fprintf(stderr, "op %d failed %d:%d\n",
-					op, i, iterations);
-				err = ret;
-				goto out;
-			}
-			if (ops[op] == bulk_op)
-				break;
-			if (keep_running == 0) {
-				err = 0;
-				goto out;
-			}
-		}
-	}
-out:
-	close_ctree(root, &super);
-	return err;
-}
-
diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c
index 9cccecc0f431..52c83be4b307 100644
--- a/fs/btrfs/root-tree.c
+++ b/fs/btrfs/root-tree.c
@@ -1,7 +1,4 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include "kerncompat.h"
-#include "radix-tree.h"
+#include <linux/module.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "print-tree.h"
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
new file mode 100644
index 000000000000..4ae76044aeab
--- /dev/null
+++ b/fs/btrfs/super.c
@@ -0,0 +1,205 @@
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/pagemap.h>
+#include <linux/highmem.h>
+#include <linux/time.h>
+#include <linux/init.h>
+#include <linux/string.h>
+#include <linux/smp_lock.h>
+#include <linux/backing-dev.h>
+#include "ctree.h"
+
+#define BTRFS_SUPER_MAGIC 0x9123682E
+#if 0
+/* some random number */
+
+static struct super_operations ramfs_ops;
+static struct inode_operations ramfs_dir_inode_operations;
+
+static struct backing_dev_info ramfs_backing_dev_info = {
+	.ra_pages	= 0,	/* No readahead */
+	.capabilities	= BDI_CAP_NO_ACCT_DIRTY | BDI_CAP_NO_WRITEBACK |
+			  BDI_CAP_MAP_DIRECT | BDI_CAP_MAP_COPY |
+			  BDI_CAP_READ_MAP | BDI_CAP_WRITE_MAP | BDI_CAP_EXEC_MAP,
+};
+
+struct inode *ramfs_get_inode(struct super_block *sb, int mode, dev_t dev)
+{
+	struct inode * inode = new_inode(sb);
+
+	if (inode) {
+		inode->i_mode = mode;
+		inode->i_uid = current->fsuid;
+		inode->i_gid = current->fsgid;
+		inode->i_blocks = 0;
+		inode->i_mapping->a_ops = &ramfs_aops;
+		inode->i_mapping->backing_dev_info = &ramfs_backing_dev_info;
+		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
+		switch (mode & S_IFMT) {
+		default:
+			init_special_inode(inode, mode, dev);
+			break;
+		case S_IFREG:
+			inode->i_op = &ramfs_file_inode_operations;
+			inode->i_fop = &ramfs_file_operations;
+			break;
+		case S_IFDIR:
+			inode->i_op = &ramfs_dir_inode_operations;
+			inode->i_fop = &simple_dir_operations;
+
+			/* directory inodes start off with i_nlink == 2 (for "." entry) */
+			inc_nlink(inode);
+			break;
+		case S_IFLNK:
+			inode->i_op = &page_symlink_inode_operations;
+			break;
+		}
+	}
+	return inode;
+}
+
+/*
+ * File creation. Allocate an inode, and we're done..
+ */
+/* SMP-safe */
+static int
+ramfs_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev)
+{
+	struct inode * inode = ramfs_get_inode(dir->i_sb, mode, dev);
+	int error = -ENOSPC;
+
+	if (inode) {
+		if (dir->i_mode & S_ISGID) {
+			inode->i_gid = dir->i_gid;
+			if (S_ISDIR(mode))
+				inode->i_mode |= S_ISGID;
+		}
+		d_instantiate(dentry, inode);
+		dget(dentry);	/* Extra count - pin the dentry in core */
+		error = 0;
+		dir->i_mtime = dir->i_ctime = CURRENT_TIME;
+	}
+	return error;
+}
+
+static int ramfs_mkdir(struct inode * dir, struct dentry * dentry, int mode)
+{
+	int retval = ramfs_mknod(dir, dentry, mode | S_IFDIR, 0);
+	if (!retval)
+		inc_nlink(dir);
+	return retval;
+}
+
+static int ramfs_create(struct inode *dir, struct dentry *dentry, int mode, struct nameidata *nd)
+{
+	return ramfs_mknod(dir, dentry, mode | S_IFREG, 0);
+}
+
+static int ramfs_symlink(struct inode * dir, struct dentry *dentry, const char * symname)
+{
+	struct inode *inode;
+	int error = -ENOSPC;
+
+	inode = ramfs_get_inode(dir->i_sb, S_IFLNK|S_IRWXUGO, 0);
+	if (inode) {
+		int l = strlen(symname)+1;
+		error = page_symlink(inode, symname, l);
+		if (!error) {
+			if (dir->i_mode & S_ISGID)
+				inode->i_gid = dir->i_gid;
+			d_instantiate(dentry, inode);
+			dget(dentry);
+			dir->i_mtime = dir->i_ctime = CURRENT_TIME;
+		} else
+			iput(inode);
+	}
+	return error;
+}
+
+static struct inode_operations ramfs_dir_inode_operations = {
+	.create		= ramfs_create,
+	.lookup		= simple_lookup,
+	.link		= simple_link,
+	.unlink		= simple_unlink,
+	.symlink	= ramfs_symlink,
+	.mkdir		= ramfs_mkdir,
+	.rmdir		= simple_rmdir,
+	.mknod		= ramfs_mknod,
+	.rename		= simple_rename,
+};
+#endif
+
+struct inode *btrfs_get_inode(struct super_block *sb, int mode, dev_t dev)
+{
+	struct inode * inode = new_inode(sb);
+
+	if (inode) {
+		inode->i_mode = mode;
+		inode->i_uid = current->fsuid;
+		inode->i_gid = current->fsgid;
+		inode->i_blocks = 0;
+		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
+	}
+	return inode;
+}
+
+static struct super_operations btrfs_ops = {
+	.statfs		= simple_statfs,
+	.drop_inode	= generic_delete_inode,
+};
+
+static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
+{
+	struct inode * inode;
+	struct dentry * root;
+
+	sb->s_maxbytes = MAX_LFS_FILESIZE;
+	sb->s_blocksize = PAGE_CACHE_SIZE;
+	sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
+	sb->s_magic = BTRFS_SUPER_MAGIC;
+	sb->s_op = &btrfs_ops;
+	sb->s_time_gran = 1;
+	inode = btrfs_get_inode(sb, S_IFDIR | 0755, 0);
+	if (!inode)
+		return -ENOMEM;
+
+	root = d_alloc_root(inode);
+	if (!root) {
+		iput(inode);
+		return -ENOMEM;
+	}
+	sb->s_root = root;
+	return 0;
+}
+
+static int btrfs_get_sb(struct file_system_type *fs_type,
+	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
+{
+	return get_sb_bdev(fs_type, flags, dev_name, data,
+			   btrfs_fill_super, mnt);
+}
+
+static struct file_system_type btrfs_fs_type = {
+	.owner		= THIS_MODULE,
+	.name		= "btrfs",
+	.get_sb		= btrfs_get_sb,
+	.kill_sb	= kill_block_super,
+	.fs_flags	= FS_REQUIRES_DEV,
+};
+
+static int __init init_btrfs_fs(void)
+{
+	printk("btrfs loaded!\n");
+	return register_filesystem(&btrfs_fs_type);
+}
+
+static void __exit exit_btrfs_fs(void)
+{
+	unregister_filesystem(&btrfs_fs_type);
+	printk("btrfs unloaded\n");
+}
+
+module_init(init_btrfs_fs)
+module_exit(exit_btrfs_fs)
+
+MODULE_LICENSE("GPL");
-- 
cgit v1.2.3


From e20d96d64f9cf9288ffecc9ad4714e91c3b97ca8 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Thu, 22 Mar 2007 12:13:20 -0400
Subject: Mountable btrfs, with readdir

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/Makefile      |   5 +-
 fs/btrfs/ctree.c       | 273 ++++++++++++++++++++++++-----------------------
 fs/btrfs/ctree.h       |  62 ++++++++---
 fs/btrfs/dir-item.c    |  17 ++-
 fs/btrfs/disk-io.c     | 281 +++++++++++++++----------------------------------
 fs/btrfs/disk-io.h     |  51 +++++----
 fs/btrfs/extent-tree.c | 144 +++++++++++++------------
 fs/btrfs/hash.c        |   9 ++
 fs/btrfs/inode-map.c   |   8 +-
 fs/btrfs/print-tree.c  |  20 ++--
 fs/btrfs/print-tree.h  |   2 +-
 fs/btrfs/root-tree.c   |   4 +-
 fs/btrfs/super.c       | 261 +++++++++++++++++++++++++++++++++++++++++----
 fs/btrfs/transaction.h |   4 +-
 14 files changed, 651 insertions(+), 490 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 99e45a54ebd6..1a0fb7422393 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -2,7 +2,8 @@ ifneq ($(KERNELRELEASE),)
 # kbuild part of makefile
 
 obj-m  := btrfs.o
-btrfs-y := super.o
+btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
+	   hash.o file-item.o inode-item.o inode-map.o disk-io.o
 
 #btrfs-y := ctree.o disk-io.o radix-tree.o extent-tree.o print-tree.o \
 #	  root-tree.o dir-item.o hash.o file-item.o inode-item.o \
@@ -14,7 +15,7 @@ else
 
 KERNELDIR := /lib/modules/`uname -r`/build
 all::
-	$(MAKE) -C $(KERNELDIR) M=`pwd` modules
+	$(MAKE) C=1 -C $(KERNELDIR) M=`pwd` modules
 clean::
 	rm *.o btrfs.ko
 endif
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 9fbd07c37fde..e690e2bb47d2 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -7,11 +7,11 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_path *path, int data_size);
 static int push_node_left(struct btrfs_trans_handle *trans, struct btrfs_root
-			  *root, struct btrfs_buffer *dst, struct btrfs_buffer
+			  *root, struct buffer_head *dst, struct buffer_head
 			  *src);
 static int balance_node_right(struct btrfs_trans_handle *trans, struct
-			      btrfs_root *root, struct btrfs_buffer *dst_buf,
-			      struct btrfs_buffer *src_buf);
+			      btrfs_root *root, struct buffer_head *dst_buf,
+			      struct buffer_head *src_buf);
 static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		   struct btrfs_path *path, int level, int slot);
 
@@ -32,32 +32,34 @@ void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p)
 }
 
 static int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root
-			   *root, struct btrfs_buffer *buf, struct btrfs_buffer
-			   *parent, int parent_slot, struct btrfs_buffer
+			   *root, struct buffer_head *buf, struct buffer_head
+			   *parent, int parent_slot, struct buffer_head
 			   **cow_ret)
 {
-	struct btrfs_buffer *cow;
+	struct buffer_head *cow;
+	struct btrfs_node *cow_node;
 
-	if (!list_empty(&buf->dirty)) {
+	if (!buffer_dirty(buf)) {
 		*cow_ret = buf;
 		return 0;
 	}
 	cow = btrfs_alloc_free_block(trans, root);
-	memcpy(&cow->node, &buf->node, root->blocksize);
-	btrfs_set_header_blocknr(&cow->node.header, cow->blocknr);
+	cow_node = btrfs_buffer_node(cow);
+	memcpy(cow_node, btrfs_buffer_node(buf), root->blocksize);
+	btrfs_set_header_blocknr(&cow_node->header, cow->b_blocknr);
 	*cow_ret = cow;
 	btrfs_inc_ref(trans, root, buf);
 	if (buf == root->node) {
 		root->node = cow;
-		cow->count++;
+		get_bh(cow);
 		if (buf != root->commit_root)
-			btrfs_free_extent(trans, root, buf->blocknr, 1, 1);
+			btrfs_free_extent(trans, root, buf->b_blocknr, 1, 1);
 		btrfs_block_release(root, buf);
 	} else {
-		btrfs_set_node_blockptr(&parent->node, parent_slot,
-					cow->blocknr);
-		BUG_ON(list_empty(&parent->dirty));
-		btrfs_free_extent(trans, root, buf->blocknr, 1, 1);
+		btrfs_set_node_blockptr(btrfs_buffer_node(parent), parent_slot,
+					cow->b_blocknr);
+		BUG_ON(!buffer_dirty(parent));
+		btrfs_free_extent(trans, root, buf->b_blocknr, 1, 1);
 	}
 	btrfs_block_release(root, buf);
 	return 0;
@@ -119,12 +121,12 @@ static int check_node(struct btrfs_root *root, struct btrfs_path *path,
 {
 	int i;
 	struct btrfs_node *parent = NULL;
-	struct btrfs_node *node = &path->nodes[level]->node;
+	struct btrfs_node *node = btrfs_buffer_node(path->nodes[level]);
 	int parent_slot;
 	u32 nritems = btrfs_header_nritems(&node->header);
 
 	if (path->nodes[level + 1])
-		parent = &path->nodes[level + 1]->node;
+		parent = btrfs_buffer_node(path->nodes[level + 1]);
 	parent_slot = path->slots[level + 1];
 	BUG_ON(nritems == 0);
 	if (parent) {
@@ -148,13 +150,13 @@ static int check_leaf(struct btrfs_root *root, struct btrfs_path *path,
 		      int level)
 {
 	int i;
-	struct btrfs_leaf *leaf = &path->nodes[level]->leaf;
+	struct btrfs_leaf *leaf = btrfs_buffer_leaf(path->nodes[level]);
 	struct btrfs_node *parent = NULL;
 	int parent_slot;
 	u32 nritems = btrfs_header_nritems(&leaf->header);
 
 	if (path->nodes[level + 1])
-		parent = &path->nodes[level + 1]->node;
+		parent = btrfs_buffer_node(path->nodes[level + 1]);
 	parent_slot = path->slots[level + 1];
 	BUG_ON(btrfs_leaf_free_space(root, leaf) < 0);
 
@@ -250,11 +252,11 @@ static int bin_search(struct btrfs_node *c, struct btrfs_key *key, int *slot)
 	return -1;
 }
 
-static struct btrfs_buffer *read_node_slot(struct btrfs_root *root,
-				   struct btrfs_buffer *parent_buf,
+static struct buffer_head *read_node_slot(struct btrfs_root *root,
+				   struct buffer_head *parent_buf,
 				   int slot)
 {
-	struct btrfs_node *node = &parent_buf->node;
+	struct btrfs_node *node = btrfs_buffer_node(parent_buf);
 	if (slot < 0)
 		return NULL;
 	if (slot >= btrfs_header_nritems(&node->header))
@@ -265,10 +267,10 @@ static struct btrfs_buffer *read_node_slot(struct btrfs_root *root,
 static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 			 *root, struct btrfs_path *path, int level)
 {
-	struct btrfs_buffer *right_buf;
-	struct btrfs_buffer *mid_buf;
-	struct btrfs_buffer *left_buf;
-	struct btrfs_buffer *parent_buf = NULL;
+	struct buffer_head *right_buf;
+	struct buffer_head *mid_buf;
+	struct buffer_head *left_buf;
+	struct buffer_head *parent_buf = NULL;
 	struct btrfs_node *right = NULL;
 	struct btrfs_node *mid;
 	struct btrfs_node *left = NULL;
@@ -283,7 +285,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 		return 0;
 
 	mid_buf = path->nodes[level];
-	mid = &mid_buf->node;
+	mid = btrfs_buffer_node(mid_buf);
 	orig_ptr = btrfs_node_blockptr(mid, orig_slot);
 
 	if (level < BTRFS_MAX_LEVEL - 1)
@@ -295,8 +297,8 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 	 * by promoting the node below to a root
 	 */
 	if (!parent_buf) {
-		struct btrfs_buffer *child;
-		u64 blocknr = mid_buf->blocknr;
+		struct buffer_head *child;
+		u64 blocknr = mid_buf->b_blocknr;
 
 		if (btrfs_header_nritems(&mid->header) != 1)
 			return 0;
@@ -313,7 +315,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 		clean_tree_block(trans, root, mid_buf);
 		return btrfs_free_extent(trans, root, blocknr, 1, 1);
 	}
-	parent = &parent_buf->node;
+	parent = btrfs_buffer_node(parent_buf);
 
 	if (btrfs_header_nritems(&mid->header) >
 	    BTRFS_NODEPTRS_PER_BLOCK(root) / 4)
@@ -326,7 +328,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 	if (left_buf) {
 		btrfs_cow_block(trans, root, left_buf, parent_buf, pslot - 1,
 				&left_buf);
-		left = &left_buf->node;
+		left = btrfs_buffer_node(left_buf);
 		orig_slot += btrfs_header_nritems(&left->header);
 		wret = push_node_left(trans, root, left_buf, mid_buf);
 		if (wret < 0)
@@ -339,12 +341,12 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 	if (right_buf) {
 		btrfs_cow_block(trans, root, right_buf, parent_buf, pslot + 1,
 				&right_buf);
-		right = &right_buf->node;
+		right = btrfs_buffer_node(right_buf);
 		wret = push_node_left(trans, root, mid_buf, right_buf);
 		if (wret < 0)
 			ret = wret;
 		if (btrfs_header_nritems(&right->header) == 0) {
-			u64 blocknr = right_buf->blocknr;
+			u64 blocknr = right_buf->b_blocknr;
 			btrfs_block_release(root, right_buf);
 			clean_tree_block(trans, root, right_buf);
 			right_buf = NULL;
@@ -360,7 +362,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 			memcpy(&parent->ptrs[pslot + 1].key,
 				&right->ptrs[0].key,
 				sizeof(struct btrfs_disk_key));
-			BUG_ON(list_empty(&parent_buf->dirty));
+			BUG_ON(!buffer_dirty(parent_buf));
 		}
 	}
 	if (btrfs_header_nritems(&mid->header) == 1) {
@@ -381,7 +383,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 	}
 	if (btrfs_header_nritems(&mid->header) == 0) {
 		/* we've managed to empty the middle node, drop it */
-		u64 blocknr = mid_buf->blocknr;
+		u64 blocknr = mid_buf->b_blocknr;
 		btrfs_block_release(root, mid_buf);
 		clean_tree_block(trans, root, mid_buf);
 		mid_buf = NULL;
@@ -396,13 +398,13 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 		/* update the parent key to reflect our changes */
 		memcpy(&parent->ptrs[pslot].key, &mid->ptrs[0].key,
 		       sizeof(struct btrfs_disk_key));
-		BUG_ON(list_empty(&parent_buf->dirty));
+		BUG_ON(!buffer_dirty(parent_buf));
 	}
 
 	/* update the path */
 	if (left_buf) {
 		if (btrfs_header_nritems(&left->header) > orig_slot) {
-			left_buf->count++; // released below
+			get_bh(left_buf);
 			path->nodes[level] = left_buf;
 			path->slots[level + 1] -= 1;
 			path->slots[level] = orig_slot;
@@ -415,8 +417,9 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 	}
 	/* double check we haven't messed things up */
 	check_block(root, path, level);
-	if (orig_ptr != btrfs_node_blockptr(&path->nodes[level]->node,
-					    path->slots[level]))
+	if (orig_ptr !=
+	    btrfs_node_blockptr(btrfs_buffer_node(path->nodes[level]),
+				path->slots[level]))
 		BUG();
 
 	if (right_buf)
@@ -443,8 +446,8 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_key *key, struct btrfs_path *p, int
 		      ins_len, int cow)
 {
-	struct btrfs_buffer *b;
-	struct btrfs_buffer *cow_buf;
+	struct buffer_head *b;
+	struct buffer_head *cow_buf;
 	struct btrfs_node *c;
 	int slot;
 	int ret;
@@ -452,18 +455,20 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
 
 again:
 	b = root->node;
-	b->count++;
+	get_bh(b);
 	while (b) {
-		level = btrfs_header_level(&b->node.header);
+		c = btrfs_buffer_node(b);
+		level = btrfs_header_level(&c->header);
 		if (cow) {
 			int wret;
-			wret = btrfs_cow_block(trans, root, b, p->nodes[level +
-					       1], p->slots[level + 1],
+			wret = btrfs_cow_block(trans, root, b,
+					       p->nodes[level + 1],
+					       p->slots[level + 1],
 					       &cow_buf);
 			b = cow_buf;
 		}
 		BUG_ON(!cow && ins_len);
-		c = &b->node;
+		c = btrfs_buffer_node(b);
 		p->nodes[level] = b;
 		ret = check_block(root, p, level);
 		if (ret)
@@ -480,7 +485,7 @@ again:
 				if (sret)
 					return sret;
 				b = p->nodes[level];
-				c = &b->node;
+				c = btrfs_buffer_node(b);
 				slot = p->slots[level];
 			} else if (ins_len < 0) {
 				int sret = balance_level(trans, root, p,
@@ -490,7 +495,7 @@ again:
 				b = p->nodes[level];
 				if (!b)
 					goto again;
-				c = &b->node;
+				c = btrfs_buffer_node(b);
 				slot = p->slots[level];
 				BUG_ON(btrfs_header_nritems(&c->header) == 1);
 			}
@@ -505,11 +510,9 @@ again:
 				if (sret)
 					return sret;
 			}
-			BUG_ON(root->node->count == 1);
 			return ret;
 		}
 	}
-	BUG_ON(root->node->count == 1);
 	return 1;
 }
 
@@ -534,9 +537,9 @@ static int fixup_low_keys(struct btrfs_trans_handle *trans, struct btrfs_root
 		int tslot = path->slots[i];
 		if (!path->nodes[i])
 			break;
-		t = &path->nodes[i]->node;
+		t = btrfs_buffer_node(path->nodes[i]);
 		memcpy(&t->ptrs[tslot].key, key, sizeof(*key));
-		BUG_ON(list_empty(&path->nodes[i]->dirty));
+		BUG_ON(!buffer_dirty(path->nodes[i]));
 		if (tslot != 0)
 			break;
 	}
@@ -551,11 +554,11 @@ static int fixup_low_keys(struct btrfs_trans_handle *trans, struct btrfs_root
  * error, and > 0 if there was no room in the left hand block.
  */
 static int push_node_left(struct btrfs_trans_handle *trans, struct btrfs_root
-			  *root, struct btrfs_buffer *dst_buf, struct
-			  btrfs_buffer *src_buf)
+			  *root, struct buffer_head *dst_buf, struct
+			  buffer_head *src_buf)
 {
-	struct btrfs_node *src = &src_buf->node;
-	struct btrfs_node *dst = &dst_buf->node;
+	struct btrfs_node *src = btrfs_buffer_node(src_buf);
+	struct btrfs_node *dst = btrfs_buffer_node(dst_buf);
 	int push_items = 0;
 	int src_nritems;
 	int dst_nritems;
@@ -580,8 +583,8 @@ static int push_node_left(struct btrfs_trans_handle *trans, struct btrfs_root
 	}
 	btrfs_set_header_nritems(&src->header, src_nritems - push_items);
 	btrfs_set_header_nritems(&dst->header, dst_nritems + push_items);
-	BUG_ON(list_empty(&src_buf->dirty));
-	BUG_ON(list_empty(&dst_buf->dirty));
+	BUG_ON(!buffer_dirty(src_buf));
+	BUG_ON(!buffer_dirty(dst_buf));
 	return ret;
 }
 
@@ -595,11 +598,11 @@ static int push_node_left(struct btrfs_trans_handle *trans, struct btrfs_root
  * this will  only push up to 1/2 the contents of the left node over
  */
 static int balance_node_right(struct btrfs_trans_handle *trans, struct
-			      btrfs_root *root, struct btrfs_buffer *dst_buf,
-			      struct btrfs_buffer *src_buf)
+			      btrfs_root *root, struct buffer_head *dst_buf,
+			      struct buffer_head *src_buf)
 {
-	struct btrfs_node *src = &src_buf->node;
-	struct btrfs_node *dst = &dst_buf->node;
+	struct btrfs_node *src = btrfs_buffer_node(src_buf);
+	struct btrfs_node *dst = btrfs_buffer_node(dst_buf);
 	int push_items = 0;
 	int max_push;
 	int src_nritems;
@@ -628,8 +631,8 @@ static int balance_node_right(struct btrfs_trans_handle *trans, struct
 	btrfs_set_header_nritems(&src->header, src_nritems - push_items);
 	btrfs_set_header_nritems(&dst->header, dst_nritems + push_items);
 
-	BUG_ON(list_empty(&src_buf->dirty));
-	BUG_ON(list_empty(&dst_buf->dirty));
+	BUG_ON(!buffer_dirty(src_buf));
+	BUG_ON(!buffer_dirty(dst_buf));
 	return ret;
 }
 
@@ -643,7 +646,7 @@ static int balance_node_right(struct btrfs_trans_handle *trans, struct
 static int insert_new_root(struct btrfs_trans_handle *trans, struct btrfs_root
 			   *root, struct btrfs_path *path, int level)
 {
-	struct btrfs_buffer *t;
+	struct buffer_head *t;
 	struct btrfs_node *lower;
 	struct btrfs_node *c;
 	struct btrfs_disk_key *lower_key;
@@ -652,24 +655,24 @@ static int insert_new_root(struct btrfs_trans_handle *trans, struct btrfs_root
 	BUG_ON(path->nodes[level-1] != root->node);
 
 	t = btrfs_alloc_free_block(trans, root);
-	c = &t->node;
+	c = btrfs_buffer_node(t);
 	memset(c, 0, root->blocksize);
 	btrfs_set_header_nritems(&c->header, 1);
 	btrfs_set_header_level(&c->header, level);
-	btrfs_set_header_blocknr(&c->header, t->blocknr);
+	btrfs_set_header_blocknr(&c->header, t->b_blocknr);
 	btrfs_set_header_parentid(&c->header,
-	                       btrfs_header_parentid(&root->node->node.header));
-	lower = &path->nodes[level-1]->node;
+	      btrfs_header_parentid(btrfs_buffer_header(root->node)));
+	lower = btrfs_buffer_node(path->nodes[level-1]);
 	if (btrfs_is_leaf(lower))
 		lower_key = &((struct btrfs_leaf *)lower)->items[0].key;
 	else
 		lower_key = &lower->ptrs[0].key;
 	memcpy(&c->ptrs[0].key, lower_key, sizeof(struct btrfs_disk_key));
-	btrfs_set_node_blockptr(c, 0, path->nodes[level - 1]->blocknr);
+	btrfs_set_node_blockptr(c, 0, path->nodes[level - 1]->b_blocknr);
 	/* the super has an extra ref to root->node */
 	btrfs_block_release(root, root->node);
 	root->node = t;
-	t->count++;
+	get_bh(t);
 	path->nodes[level] = t;
 	path->slots[level] = 0;
 	return 0;
@@ -692,7 +695,7 @@ static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root
 	int nritems;
 
 	BUG_ON(!path->nodes[level]);
-	lower = &path->nodes[level]->node;
+	lower = btrfs_buffer_node(path->nodes[level]);
 	nritems = btrfs_header_nritems(&lower->header);
 	if (slot > nritems)
 		BUG();
@@ -705,7 +708,7 @@ static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root
 	memcpy(&lower->ptrs[slot].key, key, sizeof(struct btrfs_disk_key));
 	btrfs_set_node_blockptr(lower, slot, blocknr);
 	btrfs_set_header_nritems(&lower->header, nritems + 1);
-	BUG_ON(list_empty(&path->nodes[level]->dirty));
+	BUG_ON(!buffer_dirty(path->nodes[level]));
 	return 0;
 }
 
@@ -721,9 +724,9 @@ static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root
 static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_path *path, int level)
 {
-	struct btrfs_buffer *t;
+	struct buffer_head *t;
 	struct btrfs_node *c;
-	struct btrfs_buffer *split_buffer;
+	struct buffer_head *split_buffer;
 	struct btrfs_node *split;
 	int mid;
 	int ret;
@@ -731,7 +734,7 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 	u32 c_nritems;
 
 	t = path->nodes[level];
-	c = &t->node;
+	c = btrfs_buffer_node(t);
 	if (t == root->node) {
 		/* trying to split the root, lets make a new one */
 		ret = insert_new_root(trans, root, path, level + 1);
@@ -740,11 +743,11 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 	}
 	c_nritems = btrfs_header_nritems(&c->header);
 	split_buffer = btrfs_alloc_free_block(trans, root);
-	split = &split_buffer->node;
+	split = btrfs_buffer_node(split_buffer);
 	btrfs_set_header_flags(&split->header, btrfs_header_flags(&c->header));
-	btrfs_set_header_blocknr(&split->header, split_buffer->blocknr);
+	btrfs_set_header_blocknr(&split->header, split_buffer->b_blocknr);
 	btrfs_set_header_parentid(&split->header,
-	                       btrfs_header_parentid(&root->node->node.header));
+	      btrfs_header_parentid(btrfs_buffer_header(root->node)));
 	mid = (c_nritems + 1) / 2;
 	memcpy(split->ptrs, c->ptrs + mid,
 		(c_nritems - mid) * sizeof(struct btrfs_key_ptr));
@@ -752,9 +755,9 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_set_header_nritems(&c->header, mid);
 	ret = 0;
 
-	BUG_ON(list_empty(&t->dirty));
+	BUG_ON(!buffer_dirty(t));
 	wret = insert_ptr(trans, root, path, &split->ptrs[0].key,
-			  split_buffer->blocknr, path->slots[level + 1] + 1,
+			  split_buffer->b_blocknr, path->slots[level + 1] + 1,
 			  level + 1);
 	if (wret)
 		ret = wret;
@@ -798,11 +801,12 @@ static int leaf_space_used(struct btrfs_leaf *l, int start, int nr)
 static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 			   *root, struct btrfs_path *path, int data_size)
 {
-	struct btrfs_buffer *left_buf = path->nodes[0];
-	struct btrfs_leaf *left = &left_buf->leaf;
+	struct buffer_head *left_buf = path->nodes[0];
+	struct btrfs_leaf *left = btrfs_buffer_leaf(left_buf);
 	struct btrfs_leaf *right;
-	struct btrfs_buffer *right_buf;
-	struct btrfs_buffer *upper;
+	struct buffer_head *right_buf;
+	struct buffer_head *upper;
+	struct btrfs_node *upper_node;
 	int slot;
 	int i;
 	int free_space;
@@ -817,12 +821,13 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 		return 1;
 	}
 	upper = path->nodes[1];
-	if (slot >= btrfs_header_nritems(&upper->node.header) - 1) {
+	upper_node = btrfs_buffer_node(upper);
+	if (slot >= btrfs_header_nritems(&upper_node->header) - 1) {
 		return 1;
 	}
-	right_buf = read_tree_block(root, btrfs_node_blockptr(&upper->node,
-							      slot + 1));
-	right = &right_buf->leaf;
+	right_buf = read_tree_block(root,
+		    btrfs_node_blockptr(btrfs_buffer_node(upper), slot + 1));
+	right = btrfs_buffer_leaf(right_buf);
 	free_space = btrfs_leaf_free_space(root, right);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
 		btrfs_block_release(root, right_buf);
@@ -830,7 +835,7 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 	}
 	/* cow and double check */
 	btrfs_cow_block(trans, root, right_buf, upper, slot + 1, &right_buf);
-	right = &right_buf->leaf;
+	right = btrfs_buffer_leaf(right_buf);
 	free_space = btrfs_leaf_free_space(root, right);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
 		btrfs_block_release(root, right_buf);
@@ -881,11 +886,11 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 	left_nritems -= push_items;
 	btrfs_set_header_nritems(&left->header, left_nritems);
 
-	BUG_ON(list_empty(&left_buf->dirty));
-	BUG_ON(list_empty(&right_buf->dirty));
-	memcpy(&upper->node.ptrs[slot + 1].key,
+	BUG_ON(!buffer_dirty(left_buf));
+	BUG_ON(!buffer_dirty(right_buf));
+	memcpy(&upper_node->ptrs[slot + 1].key,
 		&right->items[0].key, sizeof(struct btrfs_disk_key));
-	BUG_ON(list_empty(&upper->dirty));
+	BUG_ON(!buffer_dirty(upper));
 
 	/* then fixup the leaf pointer in the path */
 	if (path->slots[0] >= left_nritems) {
@@ -905,9 +910,9 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 			  *root, struct btrfs_path *path, int data_size)
 {
-	struct btrfs_buffer *right_buf = path->nodes[0];
-	struct btrfs_leaf *right = &right_buf->leaf;
-	struct btrfs_buffer *t;
+	struct buffer_head *right_buf = path->nodes[0];
+	struct btrfs_leaf *right = btrfs_buffer_leaf(right_buf);
+	struct buffer_head *t;
 	struct btrfs_leaf *left;
 	int slot;
 	int i;
@@ -926,9 +931,9 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 	if (!path->nodes[1]) {
 		return 1;
 	}
-	t = read_tree_block(root, btrfs_node_blockptr(&path->nodes[1]->node,
-						      slot - 1));
-	left = &t->leaf;
+	t = read_tree_block(root,
+	    btrfs_node_blockptr(btrfs_buffer_node(path->nodes[1]), slot - 1));
+	left = btrfs_buffer_leaf(t);
 	free_space = btrfs_leaf_free_space(root, left);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
 		btrfs_block_release(root, t);
@@ -937,7 +942,7 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	/* cow and double check */
 	btrfs_cow_block(trans, root, t, path->nodes[1], slot - 1, &t);
-	left = &t->leaf;
+	left = btrfs_buffer_leaf(t);
 	free_space = btrfs_leaf_free_space(root, left);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
 		btrfs_block_release(root, t);
@@ -999,8 +1004,8 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 		push_space = btrfs_item_offset(right->items + i);
 	}
 
-	BUG_ON(list_empty(&t->dirty));
-	BUG_ON(list_empty(&right_buf->dirty));
+	BUG_ON(!buffer_dirty(t));
+	BUG_ON(!buffer_dirty(right_buf));
 
 	wret = fixup_low_keys(trans, root, path, &right->items[0].key, 1);
 	if (wret)
@@ -1029,13 +1034,13 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_path *path, int data_size)
 {
-	struct btrfs_buffer *l_buf;
+	struct buffer_head *l_buf;
 	struct btrfs_leaf *l;
 	u32 nritems;
 	int mid;
 	int slot;
 	struct btrfs_leaf *right;
-	struct btrfs_buffer *right_buffer;
+	struct buffer_head *right_buffer;
 	int space_needed = data_size + sizeof(struct btrfs_item);
 	int data_copy_size;
 	int rt_data_off;
@@ -1053,7 +1058,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 			return wret;
 	}
 	l_buf = path->nodes[0];
-	l = &l_buf->leaf;
+	l = btrfs_buffer_leaf(l_buf);
 
 	/* did the pushes work? */
 	if (btrfs_leaf_free_space(root, l) >=
@@ -1071,7 +1076,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 	right_buffer = btrfs_alloc_free_block(trans, root);
 	BUG_ON(!right_buffer);
 	BUG_ON(mid == nritems);
-	right = &right_buffer->leaf;
+	right = btrfs_buffer_leaf(right_buffer);
 	memset(&right->header, 0, sizeof(right->header));
 	if (mid <= slot) {
 		/* FIXME, just alloc a new leaf here */
@@ -1085,10 +1090,10 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 			BUG();
 	}
 	btrfs_set_header_nritems(&right->header, nritems - mid);
-	btrfs_set_header_blocknr(&right->header, right_buffer->blocknr);
+	btrfs_set_header_blocknr(&right->header, right_buffer->b_blocknr);
 	btrfs_set_header_level(&right->header, 0);
 	btrfs_set_header_parentid(&right->header,
-	                       btrfs_header_parentid(&root->node->node.header));
+	      btrfs_header_parentid(btrfs_buffer_header(root->node)));
 	data_copy_size = btrfs_item_end(l->items + mid) -
 			 leaf_data_end(root, l);
 	memcpy(right->items, l->items + mid,
@@ -1107,11 +1112,11 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_set_header_nritems(&l->header, mid);
 	ret = 0;
 	wret = insert_ptr(trans, root, path, &right->items[0].key,
-			  right_buffer->blocknr, path->slots[1] + 1, 1);
+			  right_buffer->b_blocknr, path->slots[1] + 1, 1);
 	if (wret)
 		ret = wret;
-	BUG_ON(list_empty(&right_buffer->dirty));
-	BUG_ON(list_empty(&l_buf->dirty));
+	BUG_ON(!buffer_dirty(right_buffer));
+	BUG_ON(!buffer_dirty(l_buf));
 	BUG_ON(path->slots[0] != slot);
 	if (mid <= slot) {
 		btrfs_block_release(root, path->nodes[0]);
@@ -1136,7 +1141,7 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, struct btrfs_root
 	int slot;
 	int slot_orig;
 	struct btrfs_leaf *leaf;
-	struct btrfs_buffer *leaf_buf;
+	struct buffer_head *leaf_buf;
 	u32 nritems;
 	unsigned int data_end;
 	struct btrfs_disk_key disk_key;
@@ -1156,7 +1161,7 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	slot_orig = path->slots[0];
 	leaf_buf = path->nodes[0];
-	leaf = &leaf_buf->leaf;
+	leaf = btrfs_buffer_leaf(leaf_buf);
 
 	nritems = btrfs_header_nritems(&leaf->header);
 	data_end = leaf_data_end(root, leaf);
@@ -1202,7 +1207,7 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, struct btrfs_root
 	if (slot == 0)
 		ret = fixup_low_keys(trans, root, path, &disk_key, 1);
 
-	BUG_ON(list_empty(&leaf_buf->dirty));
+	BUG_ON(!buffer_dirty(leaf_buf));
 	if (btrfs_leaf_free_space(root, leaf) < 0)
 		BUG();
 	check_leaf(root, path, 0);
@@ -1225,7 +1230,8 @@ int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_init_path(&path);
 	ret = btrfs_insert_empty_item(trans, root, &path, cpu_key, data_size);
 	if (!ret) {
-		ptr = btrfs_item_ptr(&path.nodes[0]->leaf, path.slots[0], u8);
+		ptr = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]),
+				     path.slots[0], u8);
 		memcpy(ptr, data, data_size);
 	}
 	btrfs_release_path(root, &path);
@@ -1243,12 +1249,12 @@ static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		   struct btrfs_path *path, int level, int slot)
 {
 	struct btrfs_node *node;
-	struct btrfs_buffer *parent = path->nodes[level];
+	struct buffer_head *parent = path->nodes[level];
 	u32 nritems;
 	int ret = 0;
 	int wret;
 
-	node = &parent->node;
+	node = btrfs_buffer_node(parent);
 	nritems = btrfs_header_nritems(&node->header);
 	if (slot != nritems -1) {
 		memmove(node->ptrs + slot, node->ptrs + slot + 1,
@@ -1257,16 +1263,17 @@ static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	nritems--;
 	btrfs_set_header_nritems(&node->header, nritems);
 	if (nritems == 0 && parent == root->node) {
-		BUG_ON(btrfs_header_level(&root->node->node.header) != 1);
+		struct btrfs_header *header = btrfs_buffer_header(root->node);
+		BUG_ON(btrfs_header_level(header) != 1);
 		/* just turn the root into a leaf and break */
-		btrfs_set_header_level(&root->node->node.header, 0);
+		btrfs_set_header_level(header, 0);
 	} else if (slot == 0) {
 		wret = fixup_low_keys(trans, root, path, &node->ptrs[0].key,
 				      level + 1);
 		if (wret)
 			ret = wret;
 	}
-	BUG_ON(list_empty(&parent->dirty));
+	BUG_ON(!buffer_dirty(parent));
 	return ret;
 }
 
@@ -1279,7 +1286,7 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 {
 	int slot;
 	struct btrfs_leaf *leaf;
-	struct btrfs_buffer *leaf_buf;
+	struct buffer_head *leaf_buf;
 	int doff;
 	int dsize;
 	int ret = 0;
@@ -1287,7 +1294,7 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	u32 nritems;
 
 	leaf_buf = path->nodes[0];
-	leaf = &leaf_buf->leaf;
+	leaf = btrfs_buffer_leaf(leaf_buf);
 	slot = path->slots[0];
 	doff = btrfs_item_offset(leaf->items + slot);
 	dsize = btrfs_item_size(leaf->items + slot);
@@ -1313,14 +1320,13 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	if (nritems == 0) {
 		if (leaf_buf == root->node) {
 			btrfs_set_header_level(&leaf->header, 0);
-			BUG_ON(list_empty(&leaf_buf->dirty));
 		} else {
 			clean_tree_block(trans, root, leaf_buf);
 			wret = del_ptr(trans, root, path, 1, path->slots[1]);
 			if (wret)
 				ret = wret;
 			wret = btrfs_free_extent(trans, root,
-						 leaf_buf->blocknr, 1, 1);
+						 leaf_buf->b_blocknr, 1, 1);
 			if (wret)
 				ret = wret;
 		}
@@ -1332,7 +1338,6 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 			if (wret)
 				ret = wret;
 		}
-		BUG_ON(list_empty(&leaf_buf->dirty));
 
 		/* delete the leaf if it is mostly empty */
 		if (used < BTRFS_LEAF_DATA_SIZE(root) / 3) {
@@ -1341,7 +1346,7 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 			 * for possible call to del_ptr below
 			 */
 			slot = path->slots[1];
-			leaf_buf->count++;
+			get_bh(leaf_buf);
 			wret = push_leaf_left(trans, root, path, 1);
 			if (wret < 0)
 				ret = wret;
@@ -1352,7 +1357,7 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 					ret = wret;
 			}
 			if (btrfs_header_nritems(&leaf->header) == 0) {
-				u64 blocknr = leaf_buf->blocknr;
+				u64 blocknr = leaf_buf->b_blocknr;
 				clean_tree_block(trans, root, leaf_buf);
 				wret = del_ptr(trans, root, path, 1, slot);
 				if (wret)
@@ -1380,19 +1385,21 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 	int slot;
 	int level = 1;
 	u64 blocknr;
-	struct btrfs_buffer *c;
-	struct btrfs_buffer *next = NULL;
+	struct buffer_head *c;
+	struct btrfs_node *c_node;
+	struct buffer_head *next = NULL;
 
 	while(level < BTRFS_MAX_LEVEL) {
 		if (!path->nodes[level])
 			return 1;
 		slot = path->slots[level] + 1;
 		c = path->nodes[level];
-		if (slot >= btrfs_header_nritems(&c->node.header)) {
+		c_node = btrfs_buffer_node(c);
+		if (slot >= btrfs_header_nritems(&c_node->header)) {
 			level++;
 			continue;
 		}
-		blocknr = btrfs_node_blockptr(&c->node, slot);
+		blocknr = btrfs_node_blockptr(c_node, slot);
 		if (next)
 			btrfs_block_release(root, next);
 		next = read_tree_block(root, blocknr);
@@ -1408,7 +1415,7 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 		if (!level)
 			break;
 		next = read_tree_block(root,
-				       btrfs_node_blockptr(&next->node, 0));
+		       btrfs_node_blockptr(btrfs_buffer_node(next), 0));
 	}
 	return 0;
 }
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index ae8518cb94bf..7748eecd9304 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1,6 +1,9 @@
 #ifndef __BTRFS__
 #define __BTRFS__
 
+#include <linux/radix-tree.h>
+#include <linux/fs.h>
+
 struct btrfs_trans_handle;
 
 #define BTRFS_MAGIC "_BtRfS_M"
@@ -10,6 +13,12 @@ struct btrfs_trans_handle;
 #define BTRFS_INODE_MAP_OBJECTID 3
 #define BTRFS_FS_TREE_OBJECTID 4
 
+/*
+ * we can actually store much bigger names, but lets not confuse the rest
+ * of linux
+ */
+#define BTRFS_NAME_LEN 255
+
 /*
  * the key defines the order in the tree, and so it also defines (optimal)
  * block layout.  objectid corresonds to the inode number.  The flags
@@ -57,7 +66,7 @@ struct btrfs_header {
 #define __BTRFS_LEAF_DATA_SIZE(bs) ((bs) - sizeof(struct btrfs_header))
 #define BTRFS_LEAF_DATA_SIZE(r) (__BTRFS_LEAF_DATA_SIZE(r->blocksize))
 
-struct btrfs_buffer;
+struct buffer_head;
 /*
  * the super block basically lists the main trees of the FS
  * it currently lacks any block count etc etc
@@ -120,7 +129,7 @@ struct btrfs_node {
  * used while walking the tree.
  */
 struct btrfs_path {
-	struct btrfs_buffer *nodes[BTRFS_MAX_LEVEL];
+	struct buffer_head *nodes[BTRFS_MAX_LEVEL];
 	int slots[BTRFS_MAX_LEVEL];
 };
 
@@ -211,17 +220,14 @@ struct btrfs_fs_info {
 	struct btrfs_root *inode_root;
 	struct btrfs_key current_insert;
 	struct btrfs_key last_insert;
-	struct radix_tree_root cache_radix;
 	struct radix_tree_root pinned_radix;
-	struct list_head trans;
-	struct list_head cache;
 	u64 last_inode_alloc;
 	u64 last_inode_alloc_dirid;
 	u64 generation;
-	int cache_size;
-	int fp;
 	struct btrfs_trans_handle *running_transaction;
 	struct btrfs_super_block *disk_super;
+	struct buffer_head *sb_buffer;
+	struct super_block *sb;
 };
 
 /*
@@ -230,8 +236,8 @@ struct btrfs_fs_info {
  * only for the extent tree.
  */
 struct btrfs_root {
-	struct btrfs_buffer *node;
-	struct btrfs_buffer *commit_root;
+	struct buffer_head *node;
+	struct buffer_head *commit_root;
 	struct btrfs_root_item root_item;
 	struct btrfs_key root_key;
 	struct btrfs_fs_info *fs_info;
@@ -389,6 +395,29 @@ static inline void btrfs_set_inode_compat_flags(struct btrfs_inode_item *i,
 	i->compat_flags = cpu_to_le16(val);
 }
 
+static inline u32 btrfs_timespec_sec(struct btrfs_inode_timespec *ts)
+{
+	return le32_to_cpu(ts->sec);
+}
+
+static inline void btrfs_set_timespec_sec(struct btrfs_inode_timespec *ts,
+					  u32 val)
+{
+	ts->sec = cpu_to_le32(val);
+}
+
+static inline u32 btrfs_timespec_nsec(struct btrfs_inode_timespec *ts)
+{
+	return le32_to_cpu(ts->nsec);
+}
+
+static inline void btrfs_set_timespec_nsec(struct btrfs_inode_timespec *ts,
+					  u32 val)
+{
+	ts->nsec = cpu_to_le32(val);
+}
+
+
 
 static inline u64 btrfs_extent_owner(struct btrfs_extent_item *ei)
 {
@@ -757,15 +786,20 @@ static inline void btrfs_set_file_extent_num_blocks(struct
 	e->num_blocks = cpu_to_le64(val);
 }
 
+static inline struct btrfs_root *btrfs_sb(struct super_block *sb)
+{
+	return sb->s_fs_info;
+}
+
 /* helper function to cast into the data area of the leaf. */
 #define btrfs_item_ptr(leaf, slot, type) \
 	((type *)(btrfs_leaf_data(leaf) + \
 	btrfs_item_offset((leaf)->items + (slot))))
 
-struct btrfs_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
+struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 					    struct btrfs_root *root);
 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		  struct btrfs_buffer *buf);
+		  struct buffer_head *buf);
 int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, u64 blocknr, u64 num_blocks, int pin);
 int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
@@ -783,7 +817,7 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, struct btrfs_root
 int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path);
 int btrfs_leaf_free_space(struct btrfs_root *root, struct btrfs_leaf *leaf);
 int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
-			*root, struct btrfs_buffer *snap);
+			*root, struct buffer_head *snap);
 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct
 			       btrfs_root *root);
 int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
@@ -800,8 +834,8 @@ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
 			  *root, char *name, int name_len, u64 dir, u64
 			  objectid, u8 type);
 int btrfs_lookup_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
-			  *root, struct btrfs_path *path, u64 dir, char *name,
-			  int name_len, int mod);
+			  *root, struct btrfs_path *path, u64 dir,
+			  const char *name, int name_len, int mod);
 int btrfs_match_dir_item_name(struct btrfs_root *root, struct btrfs_path *path,
 			      char *name, int name_len);
 int btrfs_find_free_objectid(struct btrfs_trans_handle *trans,
diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c
index 4d8083d92fa0..75d6e373e98d 100644
--- a/fs/btrfs/dir-item.c
+++ b/fs/btrfs/dir-item.c
@@ -18,12 +18,7 @@ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
 	key.objectid = dir;
 	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY);
-	if (name_len == 1 && *name == '.')
-		key.offset = 1;
-	else if (name_len == 2 && name[0] == '.' && name[1] == '.')
-		key.offset = 2;
-	else
-		ret = btrfs_name_hash(name, name_len, &key.offset);
+	ret = btrfs_name_hash(name, name_len, &key.offset);
 	BUG_ON(ret);
 	btrfs_init_path(&path);
 	data_size = sizeof(*dir_item) + name_len;
@@ -31,7 +26,8 @@ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
 	if (ret)
 		goto out;
 
-	dir_item = btrfs_item_ptr(&path.nodes[0]->leaf, path.slots[0],
+	dir_item = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]),
+				  path.slots[0],
 				  struct btrfs_dir_item);
 	btrfs_set_dir_objectid(dir_item, objectid);
 	btrfs_set_dir_type(dir_item, type);
@@ -45,8 +41,8 @@ out:
 }
 
 int btrfs_lookup_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
-			  *root, struct btrfs_path *path, u64 dir, char *name,
-			  int name_len, int mod)
+			  *root, struct btrfs_path *path, u64 dir,
+			  const char *name, int name_len, int mod)
 {
 	int ret;
 	struct btrfs_key key;
@@ -69,7 +65,8 @@ int btrfs_match_dir_item_name(struct btrfs_root *root,
 	struct btrfs_dir_item *dir_item;
 	char *name_ptr;
 
-	dir_item = btrfs_item_ptr(&path->nodes[0]->leaf, path->slots[0],
+	dir_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
+				  path->slots[0],
 				  struct btrfs_dir_item);
 	if (btrfs_dir_name_len(dir_item) != name_len)
 		return 0;
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 05637f9fd7c7..df2061a735cd 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1,165 +1,67 @@
-#define _XOPEN_SOURCE 500
-#include <stdio.h>
-#include <stdlib.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <unistd.h>
-#include "kerncompat.h"
-#include "radix-tree.h"
+#include <linux/module.h>
+#include <linux/fs.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
 
-static int allocated_blocks = 0;
-int cache_max = 10000;
-
-static int check_tree_block(struct btrfs_root *root, struct btrfs_buffer *buf)
+static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
 {
-	if (buf->blocknr != btrfs_header_blocknr(&buf->node.header))
+	struct btrfs_node *node = btrfs_buffer_node(buf);
+	if (buf->b_blocknr != btrfs_header_blocknr(&node->header))
 		BUG();
-	if (root->node && btrfs_header_parentid(&buf->node.header) !=
-	    btrfs_header_parentid(&root->node->node.header))
+	if (root->node && btrfs_header_parentid(&node->header) !=
+	    btrfs_header_parentid(btrfs_buffer_header(root->node)))
 		BUG();
 	return 0;
 }
 
-static int free_some_buffers(struct btrfs_root *root)
+struct buffer_head *alloc_tree_block(struct btrfs_root *root, u64 blocknr)
 {
-	struct list_head *node, *next;
-	struct btrfs_buffer *b;
-	if (root->fs_info->cache_size < cache_max)
-		return 0;
-	list_for_each_safe(node, next, &root->fs_info->cache) {
-		b = list_entry(node, struct btrfs_buffer, cache);
-		if (b->count == 1) {
-			BUG_ON(!list_empty(&b->dirty));
-			list_del_init(&b->cache);
-			btrfs_block_release(root, b);
-			if (root->fs_info->cache_size < cache_max)
-				break;
-		}
-	}
-	return 0;
+	return sb_getblk(root->fs_info->sb, blocknr);
 }
 
-struct btrfs_buffer *alloc_tree_block(struct btrfs_root *root, u64 blocknr)
+struct buffer_head *find_tree_block(struct btrfs_root *root, u64 blocknr)
 {
-	struct btrfs_buffer *buf;
-	int ret;
-
-	buf = malloc(sizeof(struct btrfs_buffer) + root->blocksize);
-	if (!buf)
-		return buf;
-	allocated_blocks++;
-	buf->blocknr = blocknr;
-	buf->count = 2;
-	INIT_LIST_HEAD(&buf->dirty);
-	free_some_buffers(root);
-	radix_tree_preload(GFP_KERNEL);
-	ret = radix_tree_insert(&root->fs_info->cache_radix, blocknr, buf);
-	radix_tree_preload_end();
-	list_add_tail(&buf->cache, &root->fs_info->cache);
-	root->fs_info->cache_size++;
-	if (ret) {
-		free(buf);
-		return NULL;
-	}
-	return buf;
+	return sb_getblk(root->fs_info->sb, blocknr);
 }
 
-struct btrfs_buffer *find_tree_block(struct btrfs_root *root, u64 blocknr)
+struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr)
 {
-	struct btrfs_buffer *buf;
-	buf = radix_tree_lookup(&root->fs_info->cache_radix, blocknr);
-	if (buf) {
-		buf->count++;
-	} else {
-		buf = alloc_tree_block(root, blocknr);
-		if (!buf) {
-			BUG();
-			return NULL;
-		}
-	}
-	return buf;
-}
-
-struct btrfs_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr)
-{
-	loff_t offset = blocknr * root->blocksize;
-	struct btrfs_buffer *buf;
-	int ret;
+	struct buffer_head *buf = sb_bread(root->fs_info->sb, blocknr);
 
-	buf = radix_tree_lookup(&root->fs_info->cache_radix, blocknr);
-	if (buf) {
-		buf->count++;
-	} else {
-		buf = alloc_tree_block(root, blocknr);
-		if (!buf)
-			return NULL;
-		ret = pread(root->fs_info->fp, &buf->node, root->blocksize,
-			    offset);
-		if (ret != root->blocksize) {
-			free(buf);
-			return NULL;
-		}
-	}
+	if (!buf)
+		return buf;
 	if (check_tree_block(root, buf))
 		BUG();
 	return buf;
 }
 
 int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		     struct btrfs_buffer *buf)
+		     struct buffer_head *buf)
 {
-	if (!list_empty(&buf->dirty))
-		return 0;
-	list_add_tail(&buf->dirty, &root->fs_info->trans);
-	buf->count++;
+	mark_buffer_dirty(buf);
 	return 0;
 }
 
 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		     struct btrfs_buffer *buf)
+		     struct buffer_head *buf)
 {
-	if (!list_empty(&buf->dirty)) {
-		list_del_init(&buf->dirty);
-		btrfs_block_release(root, buf);
-	}
+	clear_buffer_dirty(buf);
 	return 0;
 }
 
 int write_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		     struct btrfs_buffer *buf)
+		     struct buffer_head *buf)
 {
-	u64 blocknr = buf->blocknr;
-	loff_t offset = blocknr * root->blocksize;
-	int ret;
-
-	if (buf->blocknr != btrfs_header_blocknr(&buf->node.header))
-		BUG();
-	ret = pwrite(root->fs_info->fp, &buf->node, root->blocksize, offset);
-	if (ret != root->blocksize)
-		return ret;
+	mark_buffer_dirty(buf);
 	return 0;
 }
 
 static int __commit_transaction(struct btrfs_trans_handle *trans, struct
 				btrfs_root *root)
 {
-	struct btrfs_buffer *b;
-	int ret = 0;
-	int wret;
-	while(!list_empty(&root->fs_info->trans)) {
-		b = list_entry(root->fs_info->trans.next, struct btrfs_buffer,
-			       dirty);
-		list_del_init(&b->dirty);
-		wret = write_tree_block(trans, root, b);
-		if (wret)
-			ret = wret;
-		btrfs_block_release(root, b);
-	}
-	return ret;
+	filemap_write_and_wait(root->fs_info->sb->s_bdev->bd_inode->i_mapping);
+	return 0;
 }
 
 static int commit_tree_roots(struct btrfs_trans_handle *trans,
@@ -172,17 +74,17 @@ static int commit_tree_roots(struct btrfs_trans_handle *trans,
 	struct btrfs_root *inode_root = fs_info->inode_root;
 
 	btrfs_set_root_blocknr(&inode_root->root_item,
-			       inode_root->node->blocknr);
+			       inode_root->node->b_blocknr);
 	ret = btrfs_update_root(trans, tree_root,
 				&inode_root->root_key,
 				&inode_root->root_item);
 	BUG_ON(ret);
 	while(1) {
 		old_extent_block = btrfs_root_blocknr(&extent_root->root_item);
-		if (old_extent_block == extent_root->node->blocknr)
+		if (old_extent_block == extent_root->node->b_blocknr)
 			break;
 		btrfs_set_root_blocknr(&extent_root->root_item,
-				       extent_root->node->blocknr);
+				       extent_root->node->b_blocknr);
 		ret = btrfs_update_root(trans, tree_root,
 					&extent_root->root_key,
 					&extent_root->root_item);
@@ -195,7 +97,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct
 			     btrfs_root *root, struct btrfs_super_block *s)
 {
 	int ret = 0;
-	struct btrfs_buffer *snap = root->commit_root;
+	struct buffer_head *snap = root->commit_root;
 	struct btrfs_key snap_key;
 
 	if (root->commit_root == root->node)
@@ -204,7 +106,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct
 	memcpy(&snap_key, &root->root_key, sizeof(snap_key));
 	root->root_key.offset++;
 
-	btrfs_set_root_blocknr(&root->root_item, root->node->blocknr);
+	btrfs_set_root_blocknr(&root->root_item, root->node->b_blocknr);
 	ret = btrfs_insert_root(trans, root->fs_info->tree_root,
 				&root->root_key, &root->root_item);
 	BUG_ON(ret);
@@ -220,7 +122,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct
 	btrfs_finish_extent_commit(trans, root->fs_info->tree_root);
 
 	root->commit_root = root->node;
-	root->node->count++;
+	get_bh(root->node);
 	ret = btrfs_drop_snapshot(trans, root, snap);
 	BUG_ON(ret);
 
@@ -234,7 +136,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct
 static int __setup_root(struct btrfs_super_block *super,
 			struct btrfs_root *root,
 			struct btrfs_fs_info *fs_info,
-			u64 objectid, int fp)
+			u64 objectid)
 {
 	root->node = NULL;
 	root->commit_root = NULL;
@@ -250,11 +152,11 @@ static int find_and_setup_root(struct btrfs_super_block *super,
 			       struct btrfs_root *tree_root,
 			       struct btrfs_fs_info *fs_info,
 			       u64 objectid,
-			       struct btrfs_root *root, int fp)
+			       struct btrfs_root *root)
 {
 	int ret;
 
-	__setup_root(super, root, fs_info, objectid, fp);
+	__setup_root(super, root, fs_info, objectid);
 	ret = btrfs_find_last_root(tree_root, objectid,
 				   &root->root_item, &root->root_key);
 	BUG_ON(ret);
@@ -265,32 +167,26 @@ static int find_and_setup_root(struct btrfs_super_block *super,
 	return 0;
 }
 
-struct btrfs_root *open_ctree(char *filename, struct btrfs_super_block *super)
+struct btrfs_root *open_ctree(struct super_block *sb,
+			      struct buffer_head *sb_buffer,
+			      struct btrfs_super_block *disk_super)
 {
-	int fp;
-
-	fp = open(filename, O_CREAT | O_RDWR, 0600);
-	if (fp < 0) {
-		return NULL;
-	}
-	return open_ctree_fd(fp, super);
-}
-
-struct btrfs_root *open_ctree_fd(int fp, struct btrfs_super_block *super)
-{
-	struct btrfs_root *root = malloc(sizeof(struct btrfs_root));
-	struct btrfs_root *extent_root = malloc(sizeof(struct btrfs_root));
-	struct btrfs_root *tree_root = malloc(sizeof(struct btrfs_root));
-	struct btrfs_root *inode_root = malloc(sizeof(struct btrfs_root));
-	struct btrfs_fs_info *fs_info = malloc(sizeof(*fs_info));
+	struct btrfs_root *root = kmalloc(sizeof(struct btrfs_root),
+					  GFP_NOFS);
+	struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
+						 GFP_NOFS);
+	struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
+					       GFP_NOFS);
+	struct btrfs_root *inode_root = kmalloc(sizeof(struct btrfs_root),
+						GFP_NOFS);
+	struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
+						GFP_NOFS);
 	int ret;
 
-	INIT_RADIX_TREE(&fs_info->cache_radix, GFP_KERNEL);
+	/* FIXME: don't be stupid */
+	if (!btrfs_super_root(disk_super))
+		return NULL;
 	INIT_RADIX_TREE(&fs_info->pinned_radix, GFP_KERNEL);
-	INIT_LIST_HEAD(&fs_info->trans);
-	INIT_LIST_HEAD(&fs_info->cache);
-	fs_info->cache_size = 0;
-	fs_info->fp = fp;
 	fs_info->running_transaction = NULL;
 	fs_info->fs_root = root;
 	fs_info->tree_root = tree_root;
@@ -298,36 +194,31 @@ struct btrfs_root *open_ctree_fd(int fp, struct btrfs_super_block *super)
 	fs_info->inode_root = inode_root;
 	fs_info->last_inode_alloc = 0;
 	fs_info->last_inode_alloc_dirid = 0;
-	fs_info->disk_super = super;
+	fs_info->disk_super = disk_super;
+	fs_info->sb_buffer = sb_buffer;
+	fs_info->sb = sb;
 	memset(&fs_info->current_insert, 0, sizeof(fs_info->current_insert));
 	memset(&fs_info->last_insert, 0, sizeof(fs_info->last_insert));
 
-	ret = pread(fp, super, sizeof(struct btrfs_super_block),
-		     BTRFS_SUPER_INFO_OFFSET);
-	if (ret == 0 || btrfs_super_root(super) == 0) {
-		BUG();
-		return NULL;
-	}
-	BUG_ON(ret < 0);
-
-	__setup_root(super, tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID, fp);
-	tree_root->node = read_tree_block(tree_root, btrfs_super_root(super));
+	__setup_root(disk_super, tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID);
+	tree_root->node = read_tree_block(tree_root,
+					  btrfs_super_root(disk_super));
 	BUG_ON(!tree_root->node);
 
-	ret = find_and_setup_root(super, tree_root, fs_info,
-				  BTRFS_EXTENT_TREE_OBJECTID, extent_root, fp);
+	ret = find_and_setup_root(disk_super, tree_root, fs_info,
+				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
 	BUG_ON(ret);
 
-	ret = find_and_setup_root(super, tree_root, fs_info,
-				  BTRFS_INODE_MAP_OBJECTID, inode_root, fp);
+	ret = find_and_setup_root(disk_super, tree_root, fs_info,
+				  BTRFS_INODE_MAP_OBJECTID, inode_root);
 	BUG_ON(ret);
 
-	ret = find_and_setup_root(super, tree_root, fs_info,
-				  BTRFS_FS_TREE_OBJECTID, root, fp);
+	ret = find_and_setup_root(disk_super, tree_root, fs_info,
+				  BTRFS_FS_TREE_OBJECTID, root);
 	BUG_ON(ret);
 
 	root->commit_root = root->node;
-	root->node->count++;
+	get_bh(root->node);
 	root->ref_cows = 1;
 	root->fs_info->generation = root->root_key.offset + 1;
 	return root;
@@ -336,8 +227,11 @@ struct btrfs_root *open_ctree_fd(int fp, struct btrfs_super_block *super)
 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_super_block *s)
 {
+	return 0;
+#if 0
 	int ret;
-	btrfs_set_super_root(s, root->fs_info->tree_root->node->blocknr);
+	btrfs_set_super_root(s, root->fs_info->tree_root->node->b_blocknr);
+
 	ret = pwrite(root->fs_info->fp, s, sizeof(*s),
 		     BTRFS_SUPER_INFO_OFFSET);
 	if (ret != sizeof(*s)) {
@@ -345,35 +239,38 @@ int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
 		return ret;
 	}
 	return 0;
+#endif
 }
 
 static int drop_cache(struct btrfs_root *root)
 {
+	return 0;
+#if 0
 	while(!list_empty(&root->fs_info->cache)) {
-		struct btrfs_buffer *b = list_entry(root->fs_info->cache.next,
-						    struct btrfs_buffer,
+		struct buffer_head *b = list_entry(root->fs_info->cache.next,
+						    struct buffer_head,
 						    cache);
 		list_del_init(&b->cache);
 		btrfs_block_release(root, b);
 	}
 	return 0;
+#endif
 }
-int close_ctree(struct btrfs_root *root, struct btrfs_super_block *s)
+
+int close_ctree(struct btrfs_root *root)
 {
 	int ret;
 	struct btrfs_trans_handle *trans;
 
 	trans = root->fs_info->running_transaction;
-	btrfs_commit_transaction(trans, root, s);
+	btrfs_commit_transaction(trans, root, root->fs_info->disk_super);
 	ret = commit_tree_roots(trans, root->fs_info);
 	BUG_ON(ret);
 	ret = __commit_transaction(trans, root);
 	BUG_ON(ret);
-	write_ctree_super(trans, root, s);
+	write_ctree_super(trans, root, root->fs_info->disk_super);
 	drop_cache(root);
-	BUG_ON(!list_empty(&root->fs_info->trans));
 
-	close(root->fs_info->fp);
 	if (root->node)
 		btrfs_block_release(root, root->node);
 	if (root->fs_info->extent_root->node)
@@ -386,29 +283,17 @@ int close_ctree(struct btrfs_root *root, struct btrfs_super_block *s)
 		btrfs_block_release(root->fs_info->tree_root,
 				    root->fs_info->tree_root->node);
 	btrfs_block_release(root, root->commit_root);
-	free(root);
-	printf("on close %d blocks are allocated\n", allocated_blocks);
+	btrfs_block_release(root, root->fs_info->sb_buffer);
+	kfree(root->fs_info->extent_root);
+	kfree(root->fs_info->inode_root);
+	kfree(root->fs_info->tree_root);
+	kfree(root->fs_info);
+	kfree(root);
 	return 0;
 }
 
-void btrfs_block_release(struct btrfs_root *root, struct btrfs_buffer *buf)
+void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf)
 {
-	buf->count--;
-	if (buf->count < 0)
-		BUG();
-	if (buf->count == 0) {
-		BUG_ON(!list_empty(&buf->cache));
-		BUG_ON(!list_empty(&buf->dirty));
-		if (!radix_tree_lookup(&root->fs_info->cache_radix,
-				       buf->blocknr))
-			BUG();
-		radix_tree_delete(&root->fs_info->cache_radix, buf->blocknr);
-		memset(buf, 0, sizeof(*buf));
-		free(buf);
-		BUG_ON(allocated_blocks == 0);
-		allocated_blocks--;
-		BUG_ON(root->fs_info->cache_size == 0);
-		root->fs_info->cache_size--;
-	}
+	brelse(buf);
 }
 
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index d888cf5c350b..7f4bb729b734 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -1,36 +1,41 @@
 #ifndef __DISKIO__
 #define __DISKIO__
-#include "list.h"
 
-struct btrfs_buffer {
-	u64 blocknr;
-	int count;
-	struct list_head dirty;
-	struct list_head cache;
-	union {
-		struct btrfs_node node;
-		struct btrfs_leaf leaf;
-	};
-};
+#include <linux/buffer_head.h>
 
-struct btrfs_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr);
-struct btrfs_buffer *find_tree_block(struct btrfs_root *root, u64 blocknr);
+#define BTRFS_SUPER_INFO_OFFSET (16 * 1024)
+
+static inline struct btrfs_node *btrfs_buffer_node(struct buffer_head *bh)
+{
+	return (struct btrfs_node *)bh->b_data;
+}
+
+static inline struct btrfs_leaf *btrfs_buffer_leaf(struct buffer_head *bh)
+{
+	return (struct btrfs_leaf *)bh->b_data;
+}
+
+static inline struct btrfs_header *btrfs_buffer_header(struct buffer_head *bh)
+{
+	return &((struct btrfs_node *)bh->b_data)->header;
+}
+
+struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr);
+struct buffer_head *find_tree_block(struct btrfs_root *root, u64 blocknr);
 int write_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		     struct btrfs_buffer *buf);
+		     struct buffer_head *buf);
 int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		     struct btrfs_buffer *buf);
+		     struct buffer_head *buf);
 int clean_tree_block(struct btrfs_trans_handle *trans,
-		     struct btrfs_root *root, struct btrfs_buffer *buf);
+		     struct btrfs_root *root, struct buffer_head *buf);
 int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct btrfs_root
 			     *root, struct btrfs_super_block *s);
-struct btrfs_root *open_ctree(char *filename, struct btrfs_super_block *s);
-struct btrfs_root *open_ctree_fd(int fp, struct btrfs_super_block *super);
-int close_ctree(struct btrfs_root *root, struct btrfs_super_block *s);
-void btrfs_block_release(struct btrfs_root *root, struct btrfs_buffer *buf);
+struct btrfs_root *open_ctree(struct super_block *sb,
+			      struct buffer_head *sb_buffer,
+			      struct btrfs_super_block *disk_super);
+int close_ctree(struct btrfs_root *root);
+void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf);
 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		      struct btrfs_super_block *s);
 int mkfs(int fd, u64 num_blocks, u32 blocksize);
-
-#define BTRFS_SUPER_INFO_OFFSET (16 * 1024)
-
 #endif
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 53a7550b5c1e..e3af2c035687 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -10,9 +10,8 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 			    search_end, struct btrfs_key *ins);
 static int finish_current_insert(struct btrfs_trans_handle *trans, struct
 				 btrfs_root *extent_root);
-static int run_pending(struct btrfs_trans_handle *trans, struct btrfs_root
-		       *extent_root);
-
+static int del_pending_extents(struct btrfs_trans_handle *trans, struct
+			       btrfs_root *extent_root);
 /*
  * pending extents are blocks that we're trying to allocate in the extent
  * map while trying to grow the map because of other allocations.  To avoid
@@ -21,6 +20,7 @@ static int run_pending(struct btrfs_trans_handle *trans, struct btrfs_root
  * manner for deletes.
  */
 #define CTREE_EXTENT_PENDING_DEL 0
+#define CTREE_EXTENT_PINNED 1
 
 static int inc_block_ref(struct btrfs_trans_handle *trans, struct btrfs_root
 			 *root, u64 blocknr)
@@ -45,15 +45,14 @@ static int inc_block_ref(struct btrfs_trans_handle *trans, struct btrfs_root
 	if (ret != 0)
 		BUG();
 	BUG_ON(ret != 0);
-	l = &path.nodes[0]->leaf;
+	l = btrfs_buffer_leaf(path.nodes[0]);
 	item = btrfs_item_ptr(l, path.slots[0], struct btrfs_extent_item);
 	refs = btrfs_extent_refs(item);
 	btrfs_set_extent_refs(item, refs + 1);
 
-	BUG_ON(list_empty(&path.nodes[0]->dirty));
 	btrfs_release_path(root->fs_info->extent_root, &path);
 	finish_current_insert(trans, root->fs_info->extent_root);
-	run_pending(trans, root->fs_info->extent_root);
+	del_pending_extents(trans, root->fs_info->extent_root);
 	return 0;
 }
 
@@ -74,7 +73,7 @@ static int lookup_block_ref(struct btrfs_trans_handle *trans, struct btrfs_root
 				0, 0);
 	if (ret != 0)
 		BUG();
-	l = &path.nodes[0]->leaf;
+	l = btrfs_buffer_leaf(path.nodes[0]);
 	item = btrfs_item_ptr(l, path.slots[0], struct btrfs_extent_item);
 	*refs = btrfs_extent_refs(item);
 	btrfs_release_path(root->fs_info->extent_root, &path);
@@ -82,18 +81,20 @@ static int lookup_block_ref(struct btrfs_trans_handle *trans, struct btrfs_root
 }
 
 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		  struct btrfs_buffer *buf)
+		  struct buffer_head *buf)
 {
 	u64 blocknr;
+	struct btrfs_node *buf_node;
 	int i;
 
 	if (!root->ref_cows)
 		return 0;
-	if (btrfs_is_leaf(&buf->node))
+	buf_node = btrfs_buffer_node(buf);
+	if (btrfs_is_leaf(buf_node))
 		return 0;
 
-	for (i = 0; i < btrfs_header_nritems(&buf->node.header); i++) {
-		blocknr = btrfs_node_blockptr(&buf->node, i);
+	for (i = 0; i < btrfs_header_nritems(&buf_node->header); i++) {
+		blocknr = btrfs_node_blockptr(buf_node, i);
 		inc_block_ref(trans, root, blocknr);
 	}
 	return 0;
@@ -108,9 +109,10 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct
 	int i;
 
 	while(1) {
-		ret = radix_tree_gang_lookup(&root->fs_info->pinned_radix,
+		ret = radix_tree_gang_lookup_tag(&root->fs_info->pinned_radix,
 					     (void **)gang, 0,
-					     ARRAY_SIZE(gang));
+					     ARRAY_SIZE(gang),
+					     CTREE_EXTENT_PINNED);
 		if (!ret)
 			break;
 		if (!first)
@@ -137,7 +139,7 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct
 
 	btrfs_set_extent_refs(&extent_item, 1);
 	btrfs_set_extent_owner(&extent_item,
-		btrfs_header_parentid(&extent_root->node->node.header));
+		btrfs_header_parentid(btrfs_buffer_header(extent_root->node)));
 	ins.offset = 1;
 	ins.flags = 0;
 	btrfs_set_key_type(&ins, BTRFS_EXTENT_ITEM_KEY);
@@ -156,11 +158,24 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct
 	return 0;
 }
 
+static int pin_down_block(struct btrfs_root *root, u64 blocknr, int tag)
+{
+	int err;
+	err = radix_tree_insert(&root->fs_info->pinned_radix,
+				blocknr, (void *)blocknr);
+	BUG_ON(err);
+	if (err)
+		return err;
+	radix_tree_tag_set(&root->fs_info->pinned_radix, blocknr,
+			   tag);
+	return 0;
+}
+
 /*
  * remove an extent from the root, returns 0 on success
  */
 static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
-			 *root, u64 blocknr, u64 num_blocks, int pin)
+			 *root, u64 blocknr, u64 num_blocks)
 {
 	struct btrfs_path path;
 	struct btrfs_key key;
@@ -171,7 +186,6 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	struct btrfs_key ins;
 	u32 refs;
 
-	BUG_ON(pin && num_blocks != 1);
 	key.objectid = blocknr;
 	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
@@ -186,26 +200,18 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 		printk("failed to find %Lu\n", key.objectid);
 		BUG();
 	}
-	ei = btrfs_item_ptr(&path.nodes[0]->leaf, path.slots[0],
+	ei = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]), path.slots[0],
 			    struct btrfs_extent_item);
 	BUG_ON(ei->refs == 0);
 	refs = btrfs_extent_refs(ei) - 1;
 	btrfs_set_extent_refs(ei, refs);
 	if (refs == 0) {
 		u64 super_blocks_used;
-		if (pin) {
-			int err;
-			radix_tree_preload(GFP_KERNEL);
-			err = radix_tree_insert(&info->pinned_radix,
-						blocknr, (void *)blocknr);
-			BUG_ON(err);
-			radix_tree_preload_end();
-		}
 		super_blocks_used = btrfs_super_blocks_used(info->disk_super);
 		btrfs_set_super_blocks_used(info->disk_super,
 					    super_blocks_used - num_blocks);
 		ret = btrfs_del_item(trans, extent_root, &path);
-		if (!pin && extent_root->fs_info->last_insert.objectid >
+		if (extent_root->fs_info->last_insert.objectid >
 		    blocknr)
 			extent_root->fs_info->last_insert.objectid = blocknr;
 		if (ret)
@@ -224,39 +230,32 @@ static int del_pending_extents(struct btrfs_trans_handle *trans, struct
 			       btrfs_root *extent_root)
 {
 	int ret;
-	struct btrfs_buffer *gang[4];
+	int wret;
+	int err = 0;
+	unsigned long gang[4];
 	int i;
+	struct radix_tree_root *radix = &extent_root->fs_info->pinned_radix;
 
 	while(1) {
 		ret = radix_tree_gang_lookup_tag(
-					&extent_root->fs_info->cache_radix,
+					&extent_root->fs_info->pinned_radix,
 					(void **)gang, 0,
 					ARRAY_SIZE(gang),
 					CTREE_EXTENT_PENDING_DEL);
 		if (!ret)
 			break;
 		for (i = 0; i < ret; i++) {
-			ret = __free_extent(trans, extent_root,
-					    gang[i]->blocknr, 1, 1);
-			radix_tree_tag_clear(&extent_root->fs_info->cache_radix,
-					     gang[i]->blocknr,
+			radix_tree_tag_set(radix, gang[i], CTREE_EXTENT_PINNED);
+			radix_tree_tag_clear(radix, gang[i],
 					     CTREE_EXTENT_PENDING_DEL);
-			btrfs_block_release(extent_root, gang[i]);
+			wret = __free_extent(trans, extent_root, gang[i], 1);
+			if (wret)
+				err = wret;
 		}
 	}
-	return 0;
+	return err;
 }
 
-static int run_pending(struct btrfs_trans_handle *trans, struct btrfs_root
-		       *extent_root)
-{
-	while(radix_tree_tagged(&extent_root->fs_info->cache_radix,
-				CTREE_EXTENT_PENDING_DEL))
-		del_pending_extents(trans, extent_root);
-	return 0;
-}
-
-
 /*
  * remove an extent from the root, returns 0 on success
  */
@@ -264,18 +263,21 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, u64 blocknr, u64 num_blocks, int pin)
 {
 	struct btrfs_root *extent_root = root->fs_info->extent_root;
-	struct btrfs_buffer *t;
+	struct buffer_head *t;
 	int pending_ret;
 	int ret;
 
 	if (root == extent_root) {
 		t = find_tree_block(root, blocknr);
-		radix_tree_tag_set(&root->fs_info->cache_radix, blocknr,
-				   CTREE_EXTENT_PENDING_DEL);
+		pin_down_block(root, blocknr, CTREE_EXTENT_PENDING_DEL);
 		return 0;
 	}
-	ret = __free_extent(trans, root, blocknr, num_blocks, pin);
-	pending_ret = run_pending(trans, root->fs_info->extent_root);
+	if (pin) {
+		ret = pin_down_block(root, blocknr, CTREE_EXTENT_PINNED);
+		BUG_ON(ret);
+	}
+	ret = __free_extent(trans, root, blocknr, num_blocks);
+	pending_ret = del_pending_extents(trans, root->fs_info->extent_root);
 	return ret ? ret : pending_ret;
 }
 
@@ -296,14 +298,16 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	int ret;
 	u64 hole_size = 0;
 	int slot = 0;
-	u64 last_block;
+	u64 last_block = 0;
 	u64 test_block;
 	int start_found;
 	struct btrfs_leaf *l;
 	struct btrfs_root * root = orig_root->fs_info->extent_root;
 	int total_needed = num_blocks;
+	int level;
 
-	total_needed += (btrfs_header_level(&root->node->node.header) + 1) * 3;
+	level = btrfs_header_level(btrfs_buffer_header(root->node));
+	total_needed += (level + 1) * 3;
 	if (root->fs_info->last_insert.objectid > search_start)
 		search_start = root->fs_info->last_insert.objectid;
 
@@ -323,7 +327,7 @@ check_failed:
 		path.slots[0]--;
 
 	while (1) {
-		l = &path.nodes[0]->leaf;
+		l = btrfs_buffer_leaf(path.nodes[0]);
 		slot = path.slots[0];
 		if (slot >= btrfs_header_nritems(&l->header)) {
 			ret = btrfs_next_leaf(root, &path);
@@ -429,7 +433,7 @@ static int alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 				sizeof(extent_item));
 
 	finish_current_insert(trans, extent_root);
-	pending_ret = run_pending(trans, extent_root);
+	pending_ret = del_pending_extents(trans, extent_root);
 	if (ret)
 		return ret;
 	if (pending_ret)
@@ -441,16 +445,15 @@ static int alloc_extent(struct btrfs_trans_handle *trans, struct btrfs_root
  * helper function to allocate a block for a given tree
  * returns the tree buffer or NULL.
  */
-struct btrfs_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
+struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 					    struct btrfs_root *root)
 {
 	struct btrfs_key ins;
 	int ret;
-	struct btrfs_buffer *buf;
+	struct buffer_head *buf;
 
 	ret = alloc_extent(trans, root, 1, 0, (unsigned long)-1,
-			   btrfs_header_parentid(&root->node->node.header),
-			   &ins);
+		btrfs_header_parentid(btrfs_buffer_header(root->node)), &ins);
 	if (ret) {
 		BUG();
 		return NULL;
@@ -467,13 +470,13 @@ struct btrfs_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 			  *root, struct btrfs_path *path, int *level)
 {
-	struct btrfs_buffer *next;
-	struct btrfs_buffer *cur;
+	struct buffer_head *next;
+	struct buffer_head *cur;
 	u64 blocknr;
 	int ret;
 	u32 refs;
 
-	ret = lookup_block_ref(trans, root, path->nodes[*level]->blocknr,
+	ret = lookup_block_ref(trans, root, path->nodes[*level]->b_blocknr,
 			       &refs);
 	BUG_ON(ret);
 	if (refs > 1)
@@ -484,9 +487,10 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 	while(*level > 0) {
 		cur = path->nodes[*level];
 		if (path->slots[*level] >=
-		    btrfs_header_nritems(&cur->node.header))
+		    btrfs_header_nritems(btrfs_buffer_header(cur)))
 			break;
-		blocknr = btrfs_node_blockptr(&cur->node, path->slots[*level]);
+		blocknr = btrfs_node_blockptr(btrfs_buffer_node(cur),
+					      path->slots[*level]);
 		ret = lookup_block_ref(trans, root, blocknr, &refs);
 		if (refs != 1 || *level == 1) {
 			path->slots[*level]++;
@@ -499,12 +503,12 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 		if (path->nodes[*level-1])
 			btrfs_block_release(root, path->nodes[*level-1]);
 		path->nodes[*level-1] = next;
-		*level = btrfs_header_level(&next->node.header);
+		*level = btrfs_header_level(btrfs_buffer_header(next));
 		path->slots[*level] = 0;
 	}
 out:
-	ret = btrfs_free_extent(trans, root, path->nodes[*level]->blocknr, 1,
-				1);
+	ret = btrfs_free_extent(trans, root, path->nodes[*level]->b_blocknr,
+				1, 1);
 	btrfs_block_release(root, path->nodes[*level]);
 	path->nodes[*level] = NULL;
 	*level += 1;
@@ -525,14 +529,14 @@ static int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 	int ret;
 	for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
 		slot = path->slots[i];
-		if (slot <
-		    btrfs_header_nritems(&path->nodes[i]->node.header)- 1) {
+		if (slot < btrfs_header_nritems(
+		    btrfs_buffer_header(path->nodes[i])) - 1) {
 			path->slots[i]++;
 			*level = i;
 			return 0;
 		} else {
 			ret = btrfs_free_extent(trans, root,
-						path->nodes[*level]->blocknr,
+						path->nodes[*level]->b_blocknr,
 						1, 1);
 			btrfs_block_release(root, path->nodes[*level]);
 			path->nodes[*level] = NULL;
@@ -549,7 +553,7 @@ static int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root
  * decremented.
  */
 int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
-			*root, struct btrfs_buffer *snap)
+			*root, struct buffer_head *snap)
 {
 	int ret = 0;
 	int wret;
@@ -560,7 +564,7 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	btrfs_init_path(&path);
 
-	level = btrfs_header_level(&snap->node.header);
+	level = btrfs_header_level(btrfs_buffer_header(snap));
 	orig_level = level;
 	path.nodes[level] = snap;
 	path.slots[level] = 0;
diff --git a/fs/btrfs/hash.c b/fs/btrfs/hash.c
index 6c2a71a46c7d..22519b8e0cf2 100644
--- a/fs/btrfs/hash.c
+++ b/fs/btrfs/hash.c
@@ -10,6 +10,7 @@
  * License.
  */
 
+#include <linux/types.h>
 #define DELTA 0x9E3779B9
 
 static void TEA_transform(__u32 buf[2], __u32 const in[])
@@ -63,6 +64,14 @@ int btrfs_name_hash(const char *name, int len, u64 *hash_result)
 	const char	*p;
 	__u32		in[8], buf[2];
 
+	if (len == 1 && *name == '.') {
+		*hash_result = 1;
+		return 0;
+	} else if (len == 2 && name[0] == '.' && name[1] == '.') {
+		*hash_result = 2;
+		return 0;
+	}
+
 	/* Initialize the default seed for the hash checksum functions */
 	buf[0] = 0x67452301;
 	buf[1] = 0xefcdab89;
diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c
index c7fda3bf7b21..c45aec258bd5 100644
--- a/fs/btrfs/inode-map.c
+++ b/fs/btrfs/inode-map.c
@@ -15,7 +15,7 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans,
 	int ret;
 	u64 hole_size = 0;
 	int slot = 0;
-	u64 last_ino;
+	u64 last_ino = 0;
 	int start_found;
 	struct btrfs_leaf *l;
 	struct btrfs_root *root = fs_root->fs_info->inode_root;
@@ -40,7 +40,7 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans,
 		path.slots[0]--;
 
 	while (1) {
-		l = &path.nodes[0]->leaf;
+		l = btrfs_buffer_leaf(path.nodes[0]);
 		slot = path.slots[0];
 		if (slot >= btrfs_header_nritems(&l->header)) {
 			ret = btrfs_next_leaf(root, &path);
@@ -105,8 +105,8 @@ int btrfs_insert_inode_map(struct btrfs_trans_handle *trans,
 	if (ret)
 		goto out;
 
-	inode_item = btrfs_item_ptr(&path.nodes[0]->leaf, path.slots[0],
-				  struct btrfs_inode_map_item);
+	inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]),
+				    path.slots[0], struct btrfs_inode_map_item);
 	btrfs_cpu_key_to_disk(&inode_item->key, location);
 out:
 	btrfs_release_path(inode_root, &path);
diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c
index aa2d3fac8804..c8ee938c1251 100644
--- a/fs/btrfs/print-tree.c
+++ b/fs/btrfs/print-tree.c
@@ -17,7 +17,6 @@ void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l)
 	printk("leaf %Lu total ptrs %d free space %d\n",
 		btrfs_header_blocknr(&l->header), nr,
 		btrfs_leaf_free_space(root, l));
-	fflush(stdout);
 	for (i = 0 ; i < nr ; i++) {
 		item = l->items + i;
 		type = btrfs_disk_key_type(&item->key);
@@ -67,10 +66,10 @@ void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l)
 				btrfs_leaf_data(l) + btrfs_item_offset(item));
 			break;
 		};
-		fflush(stdout);
 	}
 }
-void btrfs_print_tree(struct btrfs_root *root, struct btrfs_buffer *t)
+
+void btrfs_print_tree(struct btrfs_root *root, struct buffer_head *t)
 {
 	int i;
 	u32 nr;
@@ -78,16 +77,16 @@ void btrfs_print_tree(struct btrfs_root *root, struct btrfs_buffer *t)
 
 	if (!t)
 		return;
-	c = &t->node;
+	c = btrfs_buffer_node(t);
 	nr = btrfs_header_nritems(&c->header);
 	if (btrfs_is_leaf(c)) {
 		btrfs_print_leaf(root, (struct btrfs_leaf *)c);
 		return;
 	}
-	printk("node %Lu level %d total ptrs %d free spc %u\n", t->blocknr,
-	        btrfs_header_level(&c->header), nr,
-		(u32)BTRFS_NODEPTRS_PER_BLOCK(root) - nr);
-	fflush(stdout);
+	printk("node %Lu level %d total ptrs %d free spc %u\n",
+	       btrfs_header_blocknr(&c->header),
+	       btrfs_header_level(&c->header), nr,
+	       (u32)BTRFS_NODEPTRS_PER_BLOCK(root) - nr);
 	for (i = 0; i < nr; i++) {
 		printk("\tkey %d (%Lu %u %Lu) block %Lu\n",
 		       i,
@@ -95,12 +94,11 @@ void btrfs_print_tree(struct btrfs_root *root, struct btrfs_buffer *t)
 		       c->ptrs[i].key.flags,
 		       c->ptrs[i].key.offset,
 		       btrfs_node_blockptr(c, i));
-		fflush(stdout);
 	}
 	for (i = 0; i < nr; i++) {
-		struct btrfs_buffer *next_buf = read_tree_block(root,
+		struct buffer_head *next_buf = read_tree_block(root,
 						btrfs_node_blockptr(c, i));
-		struct btrfs_node *next = &next_buf->node;
+		struct btrfs_node *next = btrfs_buffer_node(next_buf);
 		if (btrfs_is_leaf(next) &&
 		    btrfs_header_level(&c->header) != 1)
 			BUG();
diff --git a/fs/btrfs/print-tree.h b/fs/btrfs/print-tree.h
index 0882ca904eca..396041a05cfa 100644
--- a/fs/btrfs/print-tree.h
+++ b/fs/btrfs/print-tree.h
@@ -1,5 +1,5 @@
 #ifndef __PRINT_TREE_
 #define __PRINT_TREE_
 void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l);
-void btrfs_print_tree(struct btrfs_root *root, struct btrfs_buffer *t);
+void btrfs_print_tree(struct btrfs_root *root, struct buffer_head *t);
 #endif
diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c
index 52c83be4b307..a4554c007ef7 100644
--- a/fs/btrfs/root-tree.c
+++ b/fs/btrfs/root-tree.c
@@ -21,7 +21,7 @@ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid,
 	if (ret < 0)
 		goto out;
 	BUG_ON(ret == 0);
-	l = &path.nodes[0]->leaf;
+	l = btrfs_buffer_leaf(path.nodes[0]);
 	BUG_ON(path.slots[0] == 0);
 	slot = path.slots[0] - 1;
 	if (btrfs_disk_key_objectid(&l->items[slot].key) != objectid) {
@@ -51,7 +51,7 @@ int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root
 	if (ret < 0)
 		goto out;
 	BUG_ON(ret != 0);
-	l = &path.nodes[0]->leaf;
+	l = btrfs_buffer_leaf(path.nodes[0]);
 	slot = path.slots[0];
 	memcpy(btrfs_item_ptr(l, slot, struct btrfs_root_item), item,
 		sizeof(*item));
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 4ae76044aeab..ccc056aad692 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -1,4 +1,5 @@
 #include <linux/module.h>
+#include <linux/buffer_head.h>
 #include <linux/fs.h>
 #include <linux/pagemap.h>
 #include <linux/highmem.h>
@@ -8,13 +9,18 @@
 #include <linux/smp_lock.h>
 #include <linux/backing-dev.h>
 #include "ctree.h"
+#include "disk-io.h"
 
 #define BTRFS_SUPER_MAGIC 0x9123682E
+
+static struct inode_operations btrfs_dir_inode_operations;
+static struct super_operations btrfs_super_ops;
+static struct file_operations btrfs_dir_file_operations;
+
 #if 0
 /* some random number */
 
 static struct super_operations ramfs_ops;
-static struct inode_operations ramfs_dir_inode_operations;
 
 static struct backing_dev_info ramfs_backing_dev_info = {
 	.ra_pages	= 0,	/* No readahead */
@@ -129,46 +135,243 @@ static struct inode_operations ramfs_dir_inode_operations = {
 };
 #endif
 
-struct inode *btrfs_get_inode(struct super_block *sb, int mode, dev_t dev)
+static void btrfs_read_locked_inode(struct inode *inode)
 {
-	struct inode * inode = new_inode(sb);
+	struct btrfs_path path;
+	struct btrfs_inode_item *inode_item;
+	struct btrfs_root *root = btrfs_sb(inode->i_sb);
+	int ret;
+printk("read locked inode %lu\n", inode->i_ino);
+	btrfs_init_path(&path);
+	ret = btrfs_lookup_inode(NULL, root, &path, inode->i_ino, 0);
+	if (ret) {
+		make_bad_inode(inode);
+		return;
+	}
+	inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]),
+				  path.slots[0],
+				  struct btrfs_inode_item);
 
-	if (inode) {
-		inode->i_mode = mode;
-		inode->i_uid = current->fsuid;
-		inode->i_gid = current->fsgid;
-		inode->i_blocks = 0;
-		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
+printk("found locked inode %lu\n", inode->i_ino);
+	inode->i_mode = btrfs_inode_mode(inode_item);
+	inode->i_nlink = btrfs_inode_nlink(inode_item);
+	inode->i_uid = btrfs_inode_uid(inode_item);
+	inode->i_gid = btrfs_inode_gid(inode_item);
+	inode->i_size = btrfs_inode_size(inode_item);
+	inode->i_atime.tv_sec = btrfs_timespec_sec(&inode_item->atime);
+	inode->i_atime.tv_nsec = btrfs_timespec_nsec(&inode_item->atime);
+	inode->i_mtime.tv_sec = btrfs_timespec_sec(&inode_item->mtime);
+	inode->i_mtime.tv_nsec = btrfs_timespec_nsec(&inode_item->mtime);
+	inode->i_ctime.tv_sec = btrfs_timespec_sec(&inode_item->ctime);
+	inode->i_ctime.tv_nsec = btrfs_timespec_nsec(&inode_item->ctime);
+	inode->i_blocks = btrfs_inode_nblocks(inode_item);
+	inode->i_generation = btrfs_inode_generation(inode_item);
+printk("about to release\n");
+	btrfs_release_path(root, &path);
+	switch (inode->i_mode & S_IFMT) {
+#if 0
+	default:
+		init_special_inode(inode, inode->i_mode,
+				   btrfs_inode_rdev(inode_item));
+		break;
+#endif
+	case S_IFREG:
+printk("inode %lu now a file\n", inode->i_ino);
+		break;
+	case S_IFDIR:
+printk("inode %lu now a directory\n", inode->i_ino);
+		inode->i_op = &btrfs_dir_inode_operations;
+		inode->i_fop = &btrfs_dir_file_operations;
+		break;
+	case S_IFLNK:
+printk("inode %lu now a link\n", inode->i_ino);
+		// inode->i_op = &page_symlink_inode_operations;
+		break;
 	}
-	return inode;
+printk("returning!\n");
+	return;
 }
 
-static struct super_operations btrfs_ops = {
-	.statfs		= simple_statfs,
-	.drop_inode	= generic_delete_inode,
-};
+static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
+			      ino_t *ino)
+{
+	const char *name = dentry->d_name.name;
+	int namelen = dentry->d_name.len;
+	struct btrfs_dir_item *di;
+	struct btrfs_path path;
+	struct btrfs_root *root = btrfs_sb(dir->i_sb);
+	int ret;
+
+	btrfs_init_path(&path);
+	ret = btrfs_lookup_dir_item(NULL, root, &path, dir->i_ino, name,
+				    namelen, 0);
+	if (ret) {
+		*ino = 0;
+		goto out;
+	}
+	di = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]), path.slots[0],
+			    struct btrfs_dir_item);
+	*ino = btrfs_dir_objectid(di);
+out:
+	btrfs_release_path(root, &path);
+	return ret;
+}
+
+static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
+				   struct nameidata *nd)
+{
+	struct inode * inode;
+	ino_t ino;
+	int ret;
+
+	if (dentry->d_name.len > BTRFS_NAME_LEN)
+		return ERR_PTR(-ENAMETOOLONG);
+
+	ret = btrfs_inode_by_name(dir, dentry, &ino);
+	if (ret < 0)
+		return ERR_PTR(ret);
+	inode = NULL;
+	if (ino) {
+printk("lookup on %.*s returns %lu\n", dentry->d_name.len, dentry->d_name.name, ino);
+		inode = iget(dir->i_sb, ino);
+		if (!inode)
+			return ERR_PTR(-EACCES);
+	}
+	return d_splice_alias(inode, dentry);
+}
+
+static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
+{
+	struct inode *inode = filp->f_path.dentry->d_inode;
+	struct btrfs_root *root = btrfs_sb(inode->i_sb);
+	struct btrfs_item *item;
+	struct btrfs_dir_item *di;
+	struct btrfs_key key;
+	struct btrfs_path path;
+	int ret;
+	u32 nritems;
+	struct btrfs_leaf *leaf;
+	int slot;
+	int advance;
+	unsigned char d_type = DT_UNKNOWN;
+	int over;
+
+	key.objectid = inode->i_ino;
+printk("readdir on dir %Lu pos %Lu\n", key.objectid, filp->f_pos);
+	key.flags = 0;
+	btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY);
+	key.offset = filp->f_pos;
+	btrfs_init_path(&path);
+	ret = btrfs_search_slot(NULL, root, &key, &path, 0, 0);
+	if (ret < 0) {
+		goto err;
+	}
+printk("first ret %d\n", ret);
+	advance = filp->f_pos > 0 && ret != 0;
+	while(1) {
+		leaf = btrfs_buffer_leaf(path.nodes[0]);
+		nritems = btrfs_header_nritems(&leaf->header);
+		slot = path.slots[0];
+printk("leaf %Lu nritems %lu slot %d\n", path.nodes[0]->b_blocknr, nritems, slot);
+		if (advance) {
+printk("advancing!\n");
+			if (slot == nritems -1) {
+				ret = btrfs_next_leaf(root, &path);
+				if (ret)
+					break;
+				leaf = btrfs_buffer_leaf(path.nodes[0]);
+				nritems = btrfs_header_nritems(&leaf->header);
+				slot = path.slots[0];
+printk("2leaf %Lu nritems %lu slot %d\n", path.nodes[0]->b_blocknr, nritems, slot);
+			} else {
+				slot++;
+				path.slots[0]++;
+			}
+		}
+		advance = 1;
+		item = leaf->items + slot;
+printk("item key %Lu %u %Lu\n", btrfs_disk_key_objectid(&item->key),
+       btrfs_disk_key_flags(&item->key), btrfs_disk_key_offset(&item->key));
+		if (btrfs_disk_key_objectid(&item->key) != key.objectid)
+			break;
+		if (btrfs_disk_key_type(&item->key) != BTRFS_DIR_ITEM_KEY)
+			continue;
+		di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
+printk("filldir name %.*s, objectid %Lu\n", btrfs_dir_name_len(di),
+       (const char *)(di + 1), btrfs_dir_objectid(di));
+		over = filldir(dirent, (const char *)(di + 1),
+			       btrfs_dir_name_len(di),
+			       btrfs_disk_key_offset(&item->key),
+			       btrfs_dir_objectid(di), d_type);
+		if (over)
+			break;
+		filp->f_pos = btrfs_disk_key_offset(&item->key) + 1;
+	}
+printk("filldir all done\n");
+	ret = 0;
+err:
+	btrfs_release_path(root, &path);
+	return ret;
+}
+
+static void btrfs_put_super (struct super_block * sb)
+{
+	struct btrfs_root *root = btrfs_sb(sb);
+	int ret;
+
+	ret = close_ctree(root);
+	if (ret) {
+		printk("close ctree returns %d\n", ret);
+	}
+	sb->s_fs_info = NULL;
+}
 
 static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
 {
 	struct inode * inode;
-	struct dentry * root;
+	struct dentry * root_dentry;
+	struct btrfs_super_block *disk_super;
+	struct buffer_head *bh;
+	struct btrfs_root *root;
 
 	sb->s_maxbytes = MAX_LFS_FILESIZE;
 	sb->s_blocksize = PAGE_CACHE_SIZE;
 	sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
 	sb->s_magic = BTRFS_SUPER_MAGIC;
-	sb->s_op = &btrfs_ops;
+	sb->s_op = &btrfs_super_ops;
 	sb->s_time_gran = 1;
-	inode = btrfs_get_inode(sb, S_IFDIR | 0755, 0);
+
+	bh = sb_bread(sb, BTRFS_SUPER_INFO_OFFSET / sb->s_blocksize);
+	if (!bh) {
+		printk("btrfs: unable to read on disk super\n");
+		return -EIO;
+	}
+	disk_super = (struct btrfs_super_block *)bh->b_data;
+	root = open_ctree(sb, bh, disk_super);
+	sb->s_fs_info = root;
+	if (!root) {
+		printk("btrfs: open_ctree failed\n");
+		return -EIO;
+	}
+	printk("read in super total blocks %Lu root %Lu\n",
+	       btrfs_super_total_blocks(disk_super),
+	       btrfs_super_root_dir(disk_super));
+
+	inode = iget_locked(sb, btrfs_super_root_dir(disk_super));
 	if (!inode)
 		return -ENOMEM;
+	if (inode->i_state & I_NEW) {
+		btrfs_read_locked_inode(inode);
+		unlock_new_inode(inode);
+	}
 
-	root = d_alloc_root(inode);
-	if (!root) {
+	root_dentry = d_alloc_root(inode);
+	if (!root_dentry) {
 		iput(inode);
 		return -ENOMEM;
 	}
-	sb->s_root = root;
+	sb->s_root = root_dentry;
+
 	return 0;
 }
 
@@ -187,6 +390,24 @@ static struct file_system_type btrfs_fs_type = {
 	.fs_flags	= FS_REQUIRES_DEV,
 };
 
+static struct super_operations btrfs_super_ops = {
+	.statfs		= simple_statfs,
+	.drop_inode	= generic_delete_inode,
+	.put_super	= btrfs_put_super,
+	.read_inode	= btrfs_read_locked_inode,
+};
+
+static struct inode_operations btrfs_dir_inode_operations = {
+	.lookup		= btrfs_lookup,
+};
+
+static struct file_operations btrfs_dir_file_operations = {
+	.llseek		= generic_file_llseek,
+	.read		= generic_read_dir,
+	.readdir	= btrfs_readdir,
+};
+
+
 static int __init init_btrfs_fs(void)
 {
 	printk("btrfs loaded!\n");
diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h
index 3adb6e69fe43..9ab27b7491c5 100644
--- a/fs/btrfs/transaction.h
+++ b/fs/btrfs/transaction.h
@@ -10,7 +10,7 @@ struct btrfs_trans_handle {
 static inline struct btrfs_trans_handle *
 btrfs_start_transaction(struct btrfs_root *root, int num_blocks)
 {
-	struct btrfs_trans_handle *h = malloc(sizeof(*h));
+	struct btrfs_trans_handle *h = kmalloc(sizeof(*h), GFP_NOFS);
 	h->transid = root->root_key.offset;
 	h->blocks_reserved = num_blocks;
 	h->blocks_used = 0;
@@ -21,7 +21,7 @@ static inline void btrfs_free_transaction(struct btrfs_root *root,
 					  struct btrfs_trans_handle *handle)
 {
 	memset(handle, 0, sizeof(*handle));
-	free(handle);
+	kfree(handle);
 }
 
 #endif
-- 
cgit v1.2.3


From 79154b1b5bcf87903db7ff16a30b360b78d6fe4f Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Thu, 22 Mar 2007 15:59:16 -0400
Subject: Btrfs: transaction rework

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/Makefile      |   2 +-
 fs/btrfs/ctree.h       |   4 +-
 fs/btrfs/disk-io.c     | 115 +++-------------------------
 fs/btrfs/disk-io.h     |   8 +-
 fs/btrfs/super.c       |  18 -----
 fs/btrfs/transaction.c | 199 +++++++++++++++++++++++++++++++++++++++++++++++++
 fs/btrfs/transaction.h |  35 +++++----
 7 files changed, 235 insertions(+), 146 deletions(-)
 create mode 100644 fs/btrfs/transaction.c

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 1a0fb7422393..be7d74cdca04 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -3,7 +3,7 @@ ifneq ($(KERNELRELEASE),)
 
 obj-m  := btrfs.o
 btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
-	   hash.o file-item.o inode-item.o inode-map.o disk-io.o
+	   hash.o file-item.o inode-item.o inode-map.o disk-io.o transaction.o
 
 #btrfs-y := ctree.o disk-io.o radix-tree.o extent-tree.o print-tree.o \
 #	  root-tree.o dir-item.o hash.o file-item.o inode-item.o \
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 7748eecd9304..4c19a3f12afd 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -5,6 +5,7 @@
 #include <linux/fs.h>
 
 struct btrfs_trans_handle;
+struct btrfs_transaction;
 
 #define BTRFS_MAGIC "_BtRfS_M"
 
@@ -224,10 +225,11 @@ struct btrfs_fs_info {
 	u64 last_inode_alloc;
 	u64 last_inode_alloc_dirid;
 	u64 generation;
-	struct btrfs_trans_handle *running_transaction;
+	struct btrfs_transaction *running_transaction;
 	struct btrfs_super_block *disk_super;
 	struct buffer_head *sb_buffer;
 	struct super_block *sb;
+	struct mutex trans_mutex;
 };
 
 /*
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index df2061a735cd..9cacca0c525c 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -50,89 +50,6 @@ int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	return 0;
 }
 
-int write_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		     struct buffer_head *buf)
-{
-	mark_buffer_dirty(buf);
-	return 0;
-}
-
-static int __commit_transaction(struct btrfs_trans_handle *trans, struct
-				btrfs_root *root)
-{
-	filemap_write_and_wait(root->fs_info->sb->s_bdev->bd_inode->i_mapping);
-	return 0;
-}
-
-static int commit_tree_roots(struct btrfs_trans_handle *trans,
-			     struct btrfs_fs_info *fs_info)
-{
-	int ret;
-	u64 old_extent_block;
-	struct btrfs_root *tree_root = fs_info->tree_root;
-	struct btrfs_root *extent_root = fs_info->extent_root;
-	struct btrfs_root *inode_root = fs_info->inode_root;
-
-	btrfs_set_root_blocknr(&inode_root->root_item,
-			       inode_root->node->b_blocknr);
-	ret = btrfs_update_root(trans, tree_root,
-				&inode_root->root_key,
-				&inode_root->root_item);
-	BUG_ON(ret);
-	while(1) {
-		old_extent_block = btrfs_root_blocknr(&extent_root->root_item);
-		if (old_extent_block == extent_root->node->b_blocknr)
-			break;
-		btrfs_set_root_blocknr(&extent_root->root_item,
-				       extent_root->node->b_blocknr);
-		ret = btrfs_update_root(trans, tree_root,
-					&extent_root->root_key,
-					&extent_root->root_item);
-		BUG_ON(ret);
-	}
-	return 0;
-}
-
-int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct
-			     btrfs_root *root, struct btrfs_super_block *s)
-{
-	int ret = 0;
-	struct buffer_head *snap = root->commit_root;
-	struct btrfs_key snap_key;
-
-	if (root->commit_root == root->node)
-		return 0;
-
-	memcpy(&snap_key, &root->root_key, sizeof(snap_key));
-	root->root_key.offset++;
-
-	btrfs_set_root_blocknr(&root->root_item, root->node->b_blocknr);
-	ret = btrfs_insert_root(trans, root->fs_info->tree_root,
-				&root->root_key, &root->root_item);
-	BUG_ON(ret);
-
-	ret = commit_tree_roots(trans, root->fs_info);
-	BUG_ON(ret);
-
-	ret = __commit_transaction(trans, root);
-	BUG_ON(ret);
-
-	write_ctree_super(trans, root, s);
-	btrfs_finish_extent_commit(trans, root->fs_info->extent_root);
-	btrfs_finish_extent_commit(trans, root->fs_info->tree_root);
-
-	root->commit_root = root->node;
-	get_bh(root->node);
-	ret = btrfs_drop_snapshot(trans, root, snap);
-	BUG_ON(ret);
-
-	ret = btrfs_del_root(trans, root->fs_info->tree_root, &snap_key);
-	BUG_ON(ret);
-	root->fs_info->generation = root->root_key.offset + 1;
-
-	return ret;
-}
-
 static int __setup_root(struct btrfs_super_block *super,
 			struct btrfs_root *root,
 			struct btrfs_fs_info *fs_info,
@@ -197,6 +114,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	fs_info->disk_super = disk_super;
 	fs_info->sb_buffer = sb_buffer;
 	fs_info->sb = sb;
+	mutex_init(&fs_info->trans_mutex);
 	memset(&fs_info->current_insert, 0, sizeof(fs_info->current_insert));
 	memset(&fs_info->last_insert, 0, sizeof(fs_info->last_insert));
 
@@ -225,7 +143,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 }
 
 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, struct btrfs_super_block *s)
+		      *root)
 {
 	return 0;
 #if 0
@@ -242,34 +160,19 @@ int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
 #endif
 }
 
-static int drop_cache(struct btrfs_root *root)
-{
-	return 0;
-#if 0
-	while(!list_empty(&root->fs_info->cache)) {
-		struct buffer_head *b = list_entry(root->fs_info->cache.next,
-						    struct buffer_head,
-						    cache);
-		list_del_init(&b->cache);
-		btrfs_block_release(root, b);
-	}
-	return 0;
-#endif
-}
-
 int close_ctree(struct btrfs_root *root)
 {
 	int ret;
 	struct btrfs_trans_handle *trans;
 
-	trans = root->fs_info->running_transaction;
-	btrfs_commit_transaction(trans, root, root->fs_info->disk_super);
-	ret = commit_tree_roots(trans, root->fs_info);
-	BUG_ON(ret);
-	ret = __commit_transaction(trans, root);
+	trans = btrfs_start_transaction(root, 1);
+	btrfs_commit_transaction(trans, root);
+	/* run commit again to  drop the original snapshot */
+	trans = btrfs_start_transaction(root, 1);
+	btrfs_commit_transaction(trans, root);
+	ret = btrfs_write_and_wait_transaction(NULL, root);
 	BUG_ON(ret);
-	write_ctree_super(trans, root, root->fs_info->disk_super);
-	drop_cache(root);
+	write_ctree_super(NULL, root);
 
 	if (root->node)
 		btrfs_block_release(root, root->node);
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 7f4bb729b734..099f7eea0ec7 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -28,14 +28,14 @@ int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		     struct buffer_head *buf);
 int clean_tree_block(struct btrfs_trans_handle *trans,
 		     struct btrfs_root *root, struct buffer_head *buf);
-int btrfs_commit_transaction(struct btrfs_trans_handle *trans, struct btrfs_root
-			     *root, struct btrfs_super_block *s);
+int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root);
 struct btrfs_root *open_ctree(struct super_block *sb,
 			      struct buffer_head *sb_buffer,
 			      struct btrfs_super_block *disk_super);
 int close_ctree(struct btrfs_root *root);
 void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf);
-int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		      struct btrfs_super_block *s);
+int write_ctree_super(struct btrfs_trans_handle *trans,
+		      struct btrfs_root *root);
 int mkfs(int fd, u64 num_blocks, u32 blocksize);
 #endif
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index ccc056aad692..620430825840 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -141,7 +141,6 @@ static void btrfs_read_locked_inode(struct inode *inode)
 	struct btrfs_inode_item *inode_item;
 	struct btrfs_root *root = btrfs_sb(inode->i_sb);
 	int ret;
-printk("read locked inode %lu\n", inode->i_ino);
 	btrfs_init_path(&path);
 	ret = btrfs_lookup_inode(NULL, root, &path, inode->i_ino, 0);
 	if (ret) {
@@ -152,7 +151,6 @@ printk("read locked inode %lu\n", inode->i_ino);
 				  path.slots[0],
 				  struct btrfs_inode_item);
 
-printk("found locked inode %lu\n", inode->i_ino);
 	inode->i_mode = btrfs_inode_mode(inode_item);
 	inode->i_nlink = btrfs_inode_nlink(inode_item);
 	inode->i_uid = btrfs_inode_uid(inode_item);
@@ -166,7 +164,6 @@ printk("found locked inode %lu\n", inode->i_ino);
 	inode->i_ctime.tv_nsec = btrfs_timespec_nsec(&inode_item->ctime);
 	inode->i_blocks = btrfs_inode_nblocks(inode_item);
 	inode->i_generation = btrfs_inode_generation(inode_item);
-printk("about to release\n");
 	btrfs_release_path(root, &path);
 	switch (inode->i_mode & S_IFMT) {
 #if 0
@@ -176,19 +173,15 @@ printk("about to release\n");
 		break;
 #endif
 	case S_IFREG:
-printk("inode %lu now a file\n", inode->i_ino);
 		break;
 	case S_IFDIR:
-printk("inode %lu now a directory\n", inode->i_ino);
 		inode->i_op = &btrfs_dir_inode_operations;
 		inode->i_fop = &btrfs_dir_file_operations;
 		break;
 	case S_IFLNK:
-printk("inode %lu now a link\n", inode->i_ino);
 		// inode->i_op = &page_symlink_inode_operations;
 		break;
 	}
-printk("returning!\n");
 	return;
 }
 
@@ -232,7 +225,6 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
 		return ERR_PTR(ret);
 	inode = NULL;
 	if (ino) {
-printk("lookup on %.*s returns %lu\n", dentry->d_name.len, dentry->d_name.name, ino);
 		inode = iget(dir->i_sb, ino);
 		if (!inode)
 			return ERR_PTR(-EACCES);
@@ -257,7 +249,6 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
 	int over;
 
 	key.objectid = inode->i_ino;
-printk("readdir on dir %Lu pos %Lu\n", key.objectid, filp->f_pos);
 	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY);
 	key.offset = filp->f_pos;
@@ -266,15 +257,12 @@ printk("readdir on dir %Lu pos %Lu\n", key.objectid, filp->f_pos);
 	if (ret < 0) {
 		goto err;
 	}
-printk("first ret %d\n", ret);
 	advance = filp->f_pos > 0 && ret != 0;
 	while(1) {
 		leaf = btrfs_buffer_leaf(path.nodes[0]);
 		nritems = btrfs_header_nritems(&leaf->header);
 		slot = path.slots[0];
-printk("leaf %Lu nritems %lu slot %d\n", path.nodes[0]->b_blocknr, nritems, slot);
 		if (advance) {
-printk("advancing!\n");
 			if (slot == nritems -1) {
 				ret = btrfs_next_leaf(root, &path);
 				if (ret)
@@ -282,7 +270,6 @@ printk("advancing!\n");
 				leaf = btrfs_buffer_leaf(path.nodes[0]);
 				nritems = btrfs_header_nritems(&leaf->header);
 				slot = path.slots[0];
-printk("2leaf %Lu nritems %lu slot %d\n", path.nodes[0]->b_blocknr, nritems, slot);
 			} else {
 				slot++;
 				path.slots[0]++;
@@ -290,15 +277,11 @@ printk("2leaf %Lu nritems %lu slot %d\n", path.nodes[0]->b_blocknr, nritems, slo
 		}
 		advance = 1;
 		item = leaf->items + slot;
-printk("item key %Lu %u %Lu\n", btrfs_disk_key_objectid(&item->key),
-       btrfs_disk_key_flags(&item->key), btrfs_disk_key_offset(&item->key));
 		if (btrfs_disk_key_objectid(&item->key) != key.objectid)
 			break;
 		if (btrfs_disk_key_type(&item->key) != BTRFS_DIR_ITEM_KEY)
 			continue;
 		di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
-printk("filldir name %.*s, objectid %Lu\n", btrfs_dir_name_len(di),
-       (const char *)(di + 1), btrfs_dir_objectid(di));
 		over = filldir(dirent, (const char *)(di + 1),
 			       btrfs_dir_name_len(di),
 			       btrfs_disk_key_offset(&item->key),
@@ -307,7 +290,6 @@ printk("filldir name %.*s, objectid %Lu\n", btrfs_dir_name_len(di),
 			break;
 		filp->f_pos = btrfs_disk_key_offset(&item->key) + 1;
 	}
-printk("filldir all done\n");
 	ret = 0;
 err:
 	btrfs_release_path(root, &path);
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
new file mode 100644
index 000000000000..8dc1c170f10f
--- /dev/null
+++ b/fs/btrfs/transaction.c
@@ -0,0 +1,199 @@
+#include <linux/module.h>
+#include <linux/fs.h>
+#include "ctree.h"
+#include "disk-io.h"
+#include "transaction.h"
+
+
+static void put_transaction(struct btrfs_transaction *transaction)
+{
+	transaction->use_count--;
+	if (transaction->use_count == 0)
+		kfree(transaction);
+}
+
+static int join_transaction(struct btrfs_root *root)
+{
+	struct btrfs_transaction *cur_trans;
+	cur_trans = root->fs_info->running_transaction;
+	if (!cur_trans) {
+		cur_trans = kmalloc(sizeof(*cur_trans), GFP_NOFS);
+		BUG_ON(!cur_trans);
+		root->fs_info->running_transaction = cur_trans;
+		cur_trans->num_writers = 0;
+		cur_trans->transid = root->root_key.offset + 1;
+		init_waitqueue_head(&cur_trans->writer_wait);
+		init_waitqueue_head(&cur_trans->commit_wait);
+		cur_trans->in_commit = 0;
+		cur_trans->use_count = 0;
+		cur_trans->commit_done = 0;
+	}
+	cur_trans->num_writers++;
+	return 0;
+}
+
+struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
+						   int num_blocks)
+{
+	struct btrfs_trans_handle *h = kmalloc(sizeof(*h), GFP_NOFS);
+	int ret;
+
+	mutex_lock(&root->fs_info->trans_mutex);
+	ret = join_transaction(root);
+	BUG_ON(ret);
+	h->transid = root->fs_info->running_transaction->transid;
+	h->transaction = root->fs_info->running_transaction;
+	h->blocks_reserved = num_blocks;
+	h->blocks_used = 0;
+	root->fs_info->running_transaction->use_count++;
+	mutex_unlock(&root->fs_info->trans_mutex);
+	return h;
+}
+
+int btrfs_end_transaction(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root)
+{
+	struct btrfs_transaction *cur_trans;
+	mutex_lock(&root->fs_info->trans_mutex);
+	cur_trans = root->fs_info->running_transaction;
+	WARN_ON(cur_trans->num_writers <= 1);
+	if (waitqueue_active(&cur_trans->writer_wait))
+		wake_up(&cur_trans->writer_wait);
+	cur_trans->num_writers--;
+	put_transaction(cur_trans);
+	mutex_unlock(&root->fs_info->trans_mutex);
+	kfree(trans);
+	return 0;
+}
+
+
+int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
+				     struct btrfs_root *root)
+{
+	filemap_write_and_wait(root->fs_info->sb->s_bdev->bd_inode->i_mapping);
+	return 0;
+}
+
+int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
+			    struct btrfs_root *root)
+{
+	int ret;
+	u64 old_extent_block;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	struct btrfs_root *tree_root = fs_info->tree_root;
+	struct btrfs_root *extent_root = fs_info->extent_root;
+	struct btrfs_root *inode_root = fs_info->inode_root;
+
+	btrfs_set_root_blocknr(&inode_root->root_item,
+			       inode_root->node->b_blocknr);
+	ret = btrfs_update_root(trans, tree_root,
+				&inode_root->root_key,
+				&inode_root->root_item);
+	BUG_ON(ret);
+	while(1) {
+		old_extent_block = btrfs_root_blocknr(&extent_root->root_item);
+		if (old_extent_block == extent_root->node->b_blocknr)
+			break;
+		btrfs_set_root_blocknr(&extent_root->root_item,
+				       extent_root->node->b_blocknr);
+		ret = btrfs_update_root(trans, tree_root,
+					&extent_root->root_key,
+					&extent_root->root_item);
+		BUG_ON(ret);
+	}
+	return 0;
+}
+
+static int wait_for_commit(struct btrfs_root *root,
+			   struct btrfs_transaction *commit)
+{
+	DEFINE_WAIT(wait);
+	commit->use_count++;
+	while(!commit->commit_done) {
+		prepare_to_wait(&commit->commit_wait, &wait,
+				TASK_UNINTERRUPTIBLE);
+		if (commit->commit_done)
+			break;
+		mutex_unlock(&root->fs_info->trans_mutex);
+		schedule();
+		mutex_lock(&root->fs_info->trans_mutex);
+	}
+	finish_wait(&commit->commit_wait, &wait);
+	return 0;
+}
+
+int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root)
+{
+	int ret = 0;
+	struct buffer_head *snap = root->commit_root;
+	struct btrfs_key snap_key;
+	struct btrfs_transaction *cur_trans;
+	DEFINE_WAIT(wait);
+
+	mutex_lock(&root->fs_info->trans_mutex);
+	if (trans->transaction->in_commit) {
+		cur_trans = trans->transaction;
+		trans->transaction->use_count++;
+		btrfs_end_transaction(trans, root);
+		ret = wait_for_commit(root, cur_trans);
+		BUG_ON(ret);
+		put_transaction(cur_trans);
+		mutex_unlock(&root->fs_info->trans_mutex);
+		return 0;
+	}
+	while (trans->transaction->num_writers > 1) {
+		prepare_to_wait(&trans->transaction->writer_wait, &wait,
+				TASK_UNINTERRUPTIBLE);
+		if (trans->transaction->num_writers <= 1)
+			break;
+		mutex_unlock(&root->fs_info->trans_mutex);
+		schedule();
+		mutex_lock(&root->fs_info->trans_mutex);
+	}
+	finish_wait(&trans->transaction->writer_wait, &wait);
+
+	cur_trans = root->fs_info->running_transaction;
+	root->fs_info->running_transaction = NULL;
+	mutex_unlock(&root->fs_info->trans_mutex);
+
+	memcpy(&snap_key, &root->root_key, sizeof(snap_key));
+	root->root_key.offset++;
+
+	if (btrfs_root_blocknr(&root->root_item) != root->node->b_blocknr) {
+		btrfs_set_root_blocknr(&root->root_item, root->node->b_blocknr);
+		ret = btrfs_insert_root(trans, root->fs_info->tree_root,
+					&root->root_key, &root->root_item);
+		BUG_ON(ret);
+	}
+
+	ret = btrfs_commit_tree_roots(trans, root);
+	BUG_ON(ret);
+
+	ret = btrfs_write_and_wait_transaction(trans, root);
+	BUG_ON(ret);
+
+	write_ctree_super(trans, root);
+	btrfs_finish_extent_commit(trans, root->fs_info->extent_root);
+	btrfs_finish_extent_commit(trans, root->fs_info->tree_root);
+	put_transaction(cur_trans);
+	kfree(trans);
+
+	if (root->node != root->commit_root) {
+		trans = btrfs_start_transaction(root, 1);
+		root->commit_root = root->node;
+		get_bh(root->node);
+		ret = btrfs_drop_snapshot(trans, root, snap);
+		BUG_ON(ret);
+
+		ret = btrfs_del_root(trans, root->fs_info->tree_root,
+				     &snap_key);
+		BUG_ON(ret);
+		root->fs_info->generation = root->root_key.offset + 1;
+		ret = btrfs_end_transaction(trans, root);
+		BUG_ON(ret);
+	}
+
+	return ret;
+}
+
diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h
index 9ab27b7491c5..7e6c08a0accd 100644
--- a/fs/btrfs/transaction.h
+++ b/fs/btrfs/transaction.h
@@ -1,27 +1,30 @@
 #ifndef __TRANSACTION__
 #define __TRANSACTION__
 
+struct btrfs_transaction {
+	u64 transid;
+	unsigned long num_writers;
+	int in_commit;
+	int use_count;
+	int commit_done;
+	wait_queue_head_t writer_wait;
+	wait_queue_head_t commit_wait;
+};
+
 struct btrfs_trans_handle {
 	u64 transid;
 	unsigned long blocks_reserved;
 	unsigned long blocks_used;
+	struct btrfs_transaction *transaction;
 };
 
-static inline struct btrfs_trans_handle *
-btrfs_start_transaction(struct btrfs_root *root, int num_blocks)
-{
-	struct btrfs_trans_handle *h = kmalloc(sizeof(*h), GFP_NOFS);
-	h->transid = root->root_key.offset;
-	h->blocks_reserved = num_blocks;
-	h->blocks_used = 0;
-	return h;
-}
-
-static inline void btrfs_free_transaction(struct btrfs_root *root,
-					  struct btrfs_trans_handle *handle)
-{
-	memset(handle, 0, sizeof(*handle));
-	kfree(handle);
-}
 
+int btrfs_end_transaction(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root);
+struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
+						   int num_blocks);
+int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
+				     struct btrfs_root *root);
+int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
+			    struct btrfs_root *root);
 #endif
-- 
cgit v1.2.3


From d98237b3ede7ab98892f7fa62201a13694c526e2 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Wed, 28 Mar 2007 13:57:48 -0400
Subject: Btrfs: use a btree inode instead of sb_getblk

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.h       |   1 +
 fs/btrfs/disk-io.c     | 174 +++++++++++++++++++++++++++++++++++++++++++++----
 fs/btrfs/disk-io.h     |   5 +-
 fs/btrfs/extent-tree.c |   8 +--
 fs/btrfs/super.c       |   8 ++-
 fs/btrfs/transaction.c |   2 +-
 6 files changed, 174 insertions(+), 24 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 1897f3a65b4f..b2ebed756c13 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -233,6 +233,7 @@ struct btrfs_fs_info {
 	struct btrfs_super_block *disk_super;
 	struct buffer_head *sb_buffer;
 	struct super_block *sb;
+	struct inode *btree_inode;
 	struct mutex trans_mutex;
 	struct mutex fs_mutex;
 };
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 758a62aba063..970103f2cacd 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1,14 +1,17 @@
 #include <linux/module.h>
 #include <linux/fs.h>
+#include <linux/blkdev.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
 
+
 static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
 {
 	struct btrfs_node *node = btrfs_buffer_node(buf);
-	if (buf->b_blocknr != btrfs_header_blocknr(&node->header))
+	if (buf->b_blocknr != btrfs_header_blocknr(&node->header)) {
 		BUG();
+	}
 	if (root->node && btrfs_header_parentid(&node->header) !=
 	    btrfs_header_parentid(btrfs_buffer_header(root->node))) {
 		BUG();
@@ -16,25 +19,154 @@ static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
 	return 0;
 }
 
-struct buffer_head *alloc_tree_block(struct btrfs_root *root, u64 blocknr)
+struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr)
+{
+	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
+	int blockbits = root->fs_info->sb->s_blocksize_bits;
+	unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
+	struct page *page;
+	struct buffer_head *bh;
+	struct buffer_head *head;
+	struct buffer_head *ret = NULL;
+
+	page = find_lock_page(mapping, index);
+	if (!page)
+		return NULL;
+
+	if (!page_has_buffers(page))
+		goto out_unlock;
+
+	head = page_buffers(page);
+	bh = head;
+	do {
+		if (buffer_mapped(bh) && bh->b_blocknr == blocknr) {
+			ret = bh;
+			get_bh(bh);
+			goto out_unlock;
+		}
+		bh = bh->b_this_page;
+	} while (bh != head);
+out_unlock:
+	unlock_page(page);
+	page_cache_release(page);
+	return ret;
+}
+
+struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
+						 u64 blocknr)
+{
+	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
+	int blockbits = root->fs_info->sb->s_blocksize_bits;
+	unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
+	struct page *page;
+	struct buffer_head *bh;
+	struct buffer_head *head;
+	struct buffer_head *ret = NULL;
+	u64 first_block = index << (PAGE_CACHE_SHIFT - blockbits);
+	page = grab_cache_page(mapping, index);
+	if (!page)
+		return NULL;
+
+	wait_on_page_writeback(page);
+	if (!page_has_buffers(page))
+		create_empty_buffers(page, root->fs_info->sb->s_blocksize, 0);
+	head = page_buffers(page);
+	bh = head;
+	do {
+		if (!buffer_mapped(bh)) {
+			bh->b_bdev = root->fs_info->sb->s_bdev;
+			bh->b_blocknr = first_block;
+			set_buffer_mapped(bh);
+		}
+		if (bh->b_blocknr == blocknr) {
+			ret = bh;
+			get_bh(bh);
+			goto out_unlock;
+		}
+		bh = bh->b_this_page;
+		first_block++;
+	} while (bh != head);
+out_unlock:
+	unlock_page(page);
+	page_cache_release(page);
+	return ret;
+}
+
+static sector_t max_block(struct block_device *bdev)
+{
+	sector_t retval = ~((sector_t)0);
+	loff_t sz = i_size_read(bdev->bd_inode);
+
+	if (sz) {
+		unsigned int size = block_size(bdev);
+		unsigned int sizebits = blksize_bits(size);
+		retval = (sz >> sizebits);
+	}
+	return retval;
+}
+
+static int btree_get_block(struct inode *inode, sector_t iblock,
+			   struct buffer_head *bh, int create)
+{
+	if (iblock >= max_block(inode->i_sb->s_bdev)) {
+		if (create)
+			return -EIO;
+
+		/*
+		 * for reads, we're just trying to fill a partial page.
+		 * return a hole, they will have to call get_block again
+		 * before they can fill it, and they will get -EIO at that
+		 * time
+		 */
+		return 0;
+	}
+	bh->b_bdev = inode->i_sb->s_bdev;
+	bh->b_blocknr = iblock;
+	set_buffer_mapped(bh);
+	return 0;
+}
+
+static int btree_writepage(struct page *page, struct writeback_control *wbc)
 {
-	return sb_getblk(root->fs_info->sb, blocknr);
+	return block_write_full_page(page, btree_get_block, wbc);
 }
 
-struct buffer_head *find_tree_block(struct btrfs_root *root, u64 blocknr)
+static int btree_readpage(struct file * file, struct page * page)
 {
-	return sb_getblk(root->fs_info->sb, blocknr);
+	return block_read_full_page(page, btree_get_block);
 }
 
+static struct address_space_operations btree_aops = {
+	.readpage	= btree_readpage,
+	.writepage	= btree_writepage,
+	.sync_page	= block_sync_page,
+};
+
 struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr)
 {
-	struct buffer_head *buf = sb_bread(root->fs_info->sb, blocknr);
+	struct buffer_head *bh = NULL;
 
-	if (!buf)
-		return buf;
-	if (check_tree_block(root, buf))
+	bh = btrfs_find_create_tree_block(root, blocknr);
+	if (!bh)
+		return bh;
+	lock_buffer(bh);
+	if (!buffer_uptodate(bh)) {
+		get_bh(bh);
+		bh->b_end_io = end_buffer_read_sync;
+		submit_bh(READ, bh);
+		wait_on_buffer(bh);
+		if (!buffer_uptodate(bh))
+			goto fail;
+	} else {
+		unlock_buffer(bh);
+	}
+	if (check_tree_block(root, bh))
 		BUG();
-	return buf;
+	return bh;
+fail:
+	brelse(bh);
+	return NULL;
+
 }
 
 int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
@@ -101,11 +233,11 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 						GFP_NOFS);
 	int ret;
 
-	/* FIXME: don't be stupid */
 	if (!btrfs_super_root(disk_super))
 		return NULL;
 	init_bit_radix(&fs_info->pinned_radix);
 	init_bit_radix(&fs_info->pending_del_radix);
+	sb_set_blocksize(sb, sb_buffer->b_size);
 	fs_info->running_transaction = NULL;
 	fs_info->fs_root = root;
 	fs_info->tree_root = tree_root;
@@ -114,14 +246,30 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	fs_info->last_inode_alloc = 0;
 	fs_info->last_inode_alloc_dirid = 0;
 	fs_info->disk_super = disk_super;
-	fs_info->sb_buffer = sb_buffer;
 	fs_info->sb = sb;
+	fs_info->btree_inode = new_inode(sb);
+	fs_info->btree_inode->i_ino = 1;
+	fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
+	fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
+	mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
+
 	mutex_init(&fs_info->trans_mutex);
 	mutex_init(&fs_info->fs_mutex);
 	memset(&fs_info->current_insert, 0, sizeof(fs_info->current_insert));
 	memset(&fs_info->last_insert, 0, sizeof(fs_info->last_insert));
 
 	__setup_root(disk_super, tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID);
+
+	fs_info->sb_buffer = read_tree_block(tree_root, sb_buffer->b_blocknr);
+
+	if (!fs_info->sb_buffer)
+		return NULL;
+
+	brelse(sb_buffer);
+	sb_buffer = NULL;
+	disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
+	fs_info->disk_super = disk_super;
+
 	tree_root->node = read_tree_block(tree_root,
 					  btrfs_super_root(disk_super));
 	BUG_ON(!tree_root->node);
@@ -137,7 +285,6 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	ret = find_and_setup_root(disk_super, tree_root, fs_info,
 				  BTRFS_FS_TREE_OBJECTID, root);
 	BUG_ON(ret);
-
 	root->commit_root = root->node;
 	get_bh(root->node);
 	root->ref_cows = 1;
@@ -191,6 +338,7 @@ int close_ctree(struct btrfs_root *root)
 				    root->fs_info->tree_root->node);
 	btrfs_block_release(root, root->commit_root);
 	btrfs_block_release(root, root->fs_info->sb_buffer);
+	iput(root->fs_info->btree_inode);
 	kfree(root->fs_info->extent_root);
 	kfree(root->fs_info->inode_root);
 	kfree(root->fs_info->tree_root);
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 099f7eea0ec7..c2c38bda704d 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -21,7 +21,8 @@ static inline struct btrfs_header *btrfs_buffer_header(struct buffer_head *bh)
 }
 
 struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr);
-struct buffer_head *find_tree_block(struct btrfs_root *root, u64 blocknr);
+struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
+						 u64 blocknr);
 int write_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		     struct buffer_head *buf);
 int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
@@ -37,5 +38,5 @@ int close_ctree(struct btrfs_root *root);
 void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf);
 int write_ctree_super(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root);
-int mkfs(int fd, u64 num_blocks, u32 blocksize);
+struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr);
 #endif
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 176c6dc534ba..b2ae8e768b6c 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -172,8 +172,8 @@ static int pin_down_block(struct btrfs_root *root, u64 blocknr, int pending)
 	struct buffer_head *bh;
 
 	if (!pending) {
-		bh = sb_find_get_block(root->fs_info->sb, blocknr);
-		if (bh) {
+		bh = btrfs_find_tree_block(root, blocknr);
+		if (bh && buffer_uptodate(bh)) {
 			header = btrfs_buffer_header(bh);
 			if (btrfs_header_generation(header) ==
 			    root->fs_info->running_transaction->transid) {
@@ -291,12 +291,10 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, u64 blocknr, u64 num_blocks, int pin)
 {
 	struct btrfs_root *extent_root = root->fs_info->extent_root;
-	struct buffer_head *t;
 	int pending_ret;
 	int ret;
 
 	if (root == extent_root) {
-		t = find_tree_block(root, blocknr);
 		pin_down_block(root, blocknr, 1);
 		return 0;
 	}
@@ -482,7 +480,7 @@ struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 		BUG();
 		return NULL;
 	}
-	buf = find_tree_block(root, ins.objectid);
+	buf = btrfs_find_create_tree_block(root, ins.objectid);
 	set_buffer_uptodate(buf);
 	return buf;
 }
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 67659b6ce962..8f07f462236a 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -474,6 +474,8 @@ static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
 	disk_super = (struct btrfs_super_block *)bh->b_data;
 	root = open_ctree(sb, bh, disk_super);
 	sb->s_fs_info = root;
+	disk_super = root->fs_info->disk_super;
+
 	if (!root) {
 		printk("btrfs: open_ctree failed\n");
 		return -EIO;
@@ -734,15 +736,15 @@ static int btrfs_sync_fs(struct super_block *sb, int wait)
 	struct btrfs_trans_handle *trans;
 	struct btrfs_root *root;
 	int ret;
+	root = btrfs_sb(sb);
 
 	sb->s_dirt = 0;
 	if (!wait) {
-		filemap_flush(sb->s_bdev->bd_inode->i_mapping);
+		filemap_flush(root->fs_info->btree_inode->i_mapping);
 		return 0;
 	}
-	filemap_write_and_wait(sb->s_bdev->bd_inode->i_mapping);
+	filemap_write_and_wait(root->fs_info->btree_inode->i_mapping);
 
-	root = btrfs_sb(sb);
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
 	ret = btrfs_commit_transaction(trans, root);
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 46a596e345f0..b20fb53a0d27 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -74,7 +74,7 @@ int btrfs_end_transaction(struct btrfs_trans_handle *trans,
 int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
 				     struct btrfs_root *root)
 {
-	filemap_write_and_wait(root->fs_info->sb->s_bdev->bd_inode->i_mapping);
+	filemap_write_and_wait(root->fs_info->btree_inode->i_mapping);
 	return 0;
 }
 
-- 
cgit v1.2.3


From f254e52c1ce550fdaa0d31f5e068f0d67c2485d4 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Thu, 29 Mar 2007 15:15:27 -0400
Subject: Btrfs: verify csums on read

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.c     |   9 ++-
 fs/btrfs/ctree.h     |  47 ++++++++-----
 fs/btrfs/disk-io.c   |  37 ++++++++---
 fs/btrfs/disk-io.h   |   2 +
 fs/btrfs/file-item.c |  60 +++++++++++++++++
 fs/btrfs/super.c     | 184 +++++++++++++++++++++++++++++++++++++++++++--------
 6 files changed, 279 insertions(+), 60 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 48c611948d11..1eb333e2e18b 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -109,14 +109,14 @@ static int comp_keys(struct btrfs_disk_key *disk, struct btrfs_key *k2)
 		return 1;
 	if (k1.objectid < k2->objectid)
 		return -1;
-	if (k1.flags > k2->flags)
-		return 1;
-	if (k1.flags < k2->flags)
-		return -1;
 	if (k1.offset > k2->offset)
 		return 1;
 	if (k1.offset < k2->offset)
 		return -1;
+	if (k1.flags > k2->flags)
+		return 1;
+	if (k1.flags < k2->flags)
+		return -1;
 	return 0;
 }
 
@@ -1165,7 +1165,6 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, struct btrfs_root
 		BUG();
 	ret = btrfs_search_slot(trans, root, cpu_key, path, data_size, 1);
 	if (ret == 0) {
-		btrfs_release_path(root, path);
 		return -EEXIST;
 	}
 	if (ret < 0)
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 6ff87f44c5d8..df1a025a771c 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -21,6 +21,9 @@ struct btrfs_transaction;
  */
 #define BTRFS_NAME_LEN 255
 
+/* 32 bytes in various csum fields */
+#define BTRFS_CSUM_SIZE 32
+
 /*
  * the key defines the order in the tree, and so it also defines (optimal)
  * block layout.  objectid corresonds to the inode number.  The flags
@@ -37,21 +40,21 @@ struct btrfs_transaction;
  */
 struct btrfs_disk_key {
 	__le64 objectid;
-	__le32 flags;
 	__le64 offset;
+	__le32 flags;
 } __attribute__ ((__packed__));
 
 struct btrfs_key {
 	u64 objectid;
-	u32 flags;
 	u64 offset;
+	u32 flags;
 } __attribute__ ((__packed__));
 
 /*
  * every tree block (leaf or node) starts with this header.
  */
 struct btrfs_header {
-	__le32 csum[8];
+	u8 csum[BTRFS_CSUM_SIZE];
 	u8 fsid[16]; /* FS specific uuid */
 	__le64 blocknr; /* which block this node is supposed to live in */
 	__le64 generation;
@@ -75,7 +78,7 @@ struct buffer_head;
  * it currently lacks any block count etc etc
  */
 struct btrfs_super_block {
-	__le32 csum[8];
+	u8 csum[BTRFS_CSUM_SIZE];
 	/* the first 3 fields must match struct btrfs_header */
 	u8 fsid[16];    /* FS specific uuid */
 	__le64 blocknr; /* this block number */
@@ -147,7 +150,7 @@ struct btrfs_extent_item {
 } __attribute__ ((__packed__));
 
 struct btrfs_inode_timespec {
-	__le32 sec;
+	__le64 sec;
 	__le32 nsec;
 } __attribute__ ((__packed__));
 
@@ -214,6 +217,10 @@ struct btrfs_file_extent_item {
 	__le64 num_blocks;
 } __attribute__ ((__packed__));
 
+struct btrfs_csum_item {
+	u8 csum[BTRFS_CSUM_SIZE];
+} __attribute__ ((__packed__));
+
 struct btrfs_inode_map_item {
 	struct btrfs_disk_key key;
 } __attribute__ ((__packed__));
@@ -283,27 +290,32 @@ struct btrfs_root {
  * a (hopefully) huge chunk of disk
  */
 #define BTRFS_EXTENT_DATA_KEY	4
+/*
+ * csum items have the checksums for data in the extents
+ */
+#define BTRFS_CSUM_ITEM_KEY	5
+
 /*
  * root items point to tree roots.  There are typically in the root
  * tree used by the super block to find all the other trees
  */
-#define BTRFS_ROOT_ITEM_KEY	5
+#define BTRFS_ROOT_ITEM_KEY	6
 /*
  * extent items are in the extent map tree.  These record which blocks
  * are used, and how many references there are to each block
  */
-#define BTRFS_EXTENT_ITEM_KEY	6
+#define BTRFS_EXTENT_ITEM_KEY	7
 
 /*
  * the inode map records which inode numbers are in use and where
  * they actually live on disk
  */
-#define BTRFS_INODE_MAP_ITEM_KEY 7
+#define BTRFS_INODE_MAP_ITEM_KEY 8
 /*
  * string items are for debugging.  They just store a short string of
  * data in the FS
  */
-#define BTRFS_STRING_ITEM_KEY	8
+#define BTRFS_STRING_ITEM_KEY	9
 
 static inline u64 btrfs_inode_generation(struct btrfs_inode_item *i)
 {
@@ -407,15 +419,15 @@ static inline void btrfs_set_inode_compat_flags(struct btrfs_inode_item *i,
 	i->compat_flags = cpu_to_le16(val);
 }
 
-static inline u32 btrfs_timespec_sec(struct btrfs_inode_timespec *ts)
+static inline u64 btrfs_timespec_sec(struct btrfs_inode_timespec *ts)
 {
-	return le32_to_cpu(ts->sec);
+	return le64_to_cpu(ts->sec);
 }
 
 static inline void btrfs_set_timespec_sec(struct btrfs_inode_timespec *ts,
-					  u32 val)
+					  u64 val)
 {
-	ts->sec = cpu_to_le32(val);
+	ts->sec = cpu_to_le64(val);
 }
 
 static inline u32 btrfs_timespec_nsec(struct btrfs_inode_timespec *ts)
@@ -429,8 +441,6 @@ static inline void btrfs_set_timespec_nsec(struct btrfs_inode_timespec *ts,
 	ts->nsec = cpu_to_le32(val);
 }
 
-
-
 static inline u64 btrfs_extent_owner(struct btrfs_extent_item *ei)
 {
 	return le64_to_cpu(ei->owner);
@@ -905,4 +915,11 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root,
 			     struct btrfs_path *path, u64 objectid,
 			     u64 blocknr, int mod);
+int btrfs_csum_file_block(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root,
+			  u64 objectid, u64 offset,
+			  char *data, size_t len);
+int btrfs_csum_verify_file_block(struct btrfs_root *root,
+				 u64 objectid, u64 offset,
+				 char *data, size_t len);
 #endif
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index f2c1dd40c087..f5db2b936502 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -127,31 +127,46 @@ static int btree_get_block(struct inode *inode, sector_t iblock,
 	return 0;
 }
 
-static int csum_tree_block(struct btrfs_root * root, struct buffer_head *bh,
-			    int verify)
+int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
+		    char *result)
 {
-	struct btrfs_node *node = btrfs_buffer_node(bh);
 	struct scatterlist sg;
 	struct crypto_hash *tfm = root->fs_info->hash_tfm;
 	struct hash_desc desc;
 	int ret;
-	char result[32];
 
 	desc.tfm = tfm;
 	desc.flags = 0;
-	sg_init_one(&sg, bh->b_data + 32, bh->b_size - 32);
+	sg_init_one(&sg, data, len);
 	spin_lock(&root->fs_info->hash_lock);
-	ret = crypto_hash_digest(&desc, &sg, bh->b_size - 32, result);
+	ret = crypto_hash_digest(&desc, &sg, len, result);
 	spin_unlock(&root->fs_info->hash_lock);
 	if (ret) {
 		printk("sha256 digest failed\n");
 	}
+	return ret;
+}
+static int csum_tree_block(struct btrfs_root *root, struct buffer_head *bh,
+			   int verify)
+{
+	char result[BTRFS_CSUM_SIZE];
+	int ret;
+	struct btrfs_node *node;
+
+	ret = btrfs_csum_data(root, bh->b_data + BTRFS_CSUM_SIZE,
+			      bh->b_size - BTRFS_CSUM_SIZE, result);
+	if (ret)
+		return ret;
 	if (verify) {
-		if (memcmp(node->header.csum, result, sizeof(result)))
-			printk("csum verify failed on %Lu\n", bh->b_blocknr);
-		return -EINVAL;
-	} else
-		memcpy(node->header.csum, result, sizeof(node->header.csum));
+		if (memcmp(bh->b_data, result, BTRFS_CSUM_SIZE)) {
+			printk("checksum verify failed on %lu\n",
+			       bh->b_blocknr);
+			return 1;
+		}
+	} else {
+		node = btrfs_buffer_node(bh);
+		memcpy(&node->header.csum, result, BTRFS_CSUM_SIZE);
+	}
 	return 0;
 }
 
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index c2c38bda704d..f6998e2192ce 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -39,4 +39,6 @@ void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf);
 int write_ctree_super(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root);
 struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr);
+int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
+		    char *result);
 #endif
diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c
index 5230a44cb19b..2d2c23ca7cbf 100644
--- a/fs/btrfs/file-item.c
+++ b/fs/btrfs/file-item.c
@@ -57,3 +57,63 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans,
 	ret = btrfs_search_slot(trans, root, &file_key, path, ins_len, cow);
 	return ret;
 }
+
+int btrfs_csum_file_block(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root,
+			  u64 objectid, u64 offset,
+			  char *data, size_t len)
+{
+	int ret;
+	struct btrfs_key file_key;
+	struct btrfs_path path;
+	struct btrfs_csum_item *item;
+
+	btrfs_init_path(&path);
+	file_key.objectid = objectid;
+	file_key.offset = offset;
+	file_key.flags = 0;
+	btrfs_set_key_type(&file_key, BTRFS_CSUM_ITEM_KEY);
+	ret = btrfs_insert_empty_item(trans, root, &path, &file_key,
+				      BTRFS_CSUM_SIZE);
+	if (ret != 0 && ret != -EEXIST)
+		goto fail;
+	item = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]), path.slots[0],
+			      struct btrfs_csum_item);
+	ret = 0;
+	ret = btrfs_csum_data(root, data, len, item->csum);
+	mark_buffer_dirty(path.nodes[0]);
+fail:
+	btrfs_release_path(root, &path);
+	return ret;
+}
+
+int btrfs_csum_verify_file_block(struct btrfs_root *root,
+				 u64 objectid, u64 offset,
+				 char *data, size_t len)
+{
+	int ret;
+	struct btrfs_key file_key;
+	struct btrfs_path path;
+	struct btrfs_csum_item *item;
+	char result[BTRFS_CSUM_SIZE];
+
+	btrfs_init_path(&path);
+	file_key.objectid = objectid;
+	file_key.offset = offset;
+	file_key.flags = 0;
+	btrfs_set_key_type(&file_key, BTRFS_CSUM_ITEM_KEY);
+	ret = btrfs_search_slot(NULL, root, &file_key, &path, 0, 0);
+	if (ret)
+		goto fail;
+	item = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]), path.slots[0],
+			      struct btrfs_csum_item);
+	ret = 0;
+	ret = btrfs_csum_data(root, data, len, result);
+	WARN_ON(ret);
+	if (memcmp(result, item->csum, BTRFS_CSUM_SIZE))
+		ret = 1;
+fail:
+	btrfs_release_path(root, &path);
+	return ret;
+}
+
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 7914b31f5bcd..04428137d75f 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -249,15 +249,16 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 	struct btrfs_key key;
 	struct btrfs_disk_key *found_key;
 	struct btrfs_leaf *leaf;
-	struct btrfs_file_extent_item *fi;
-	u64 extent_start;
-	u64 extent_num_blocks;
+	struct btrfs_file_extent_item *fi = NULL;
+	u64 extent_start = 0;
+	u64 extent_num_blocks = 0;
+	int found_extent;
 
 	/* FIXME, add redo link to tree so we don't leak on crash */
 	key.objectid = inode->i_ino;
 	key.offset = (u64)-1;
 	key.flags = 0;
-	btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
+	btrfs_set_key_type(&key, BTRFS_CSUM_ITEM_KEY);
 	while(1) {
 		btrfs_init_path(&path);
 		ret = btrfs_search_slot(trans, root, &key, &path, -1, 1);
@@ -273,25 +274,32 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 		found_key = &leaf->items[path.slots[0]].key;
 		if (btrfs_disk_key_objectid(found_key) != inode->i_ino)
 			break;
-		if (btrfs_disk_key_type(found_key) != BTRFS_EXTENT_DATA_KEY)
+		if (btrfs_disk_key_type(found_key) != BTRFS_CSUM_ITEM_KEY &&
+		    btrfs_disk_key_type(found_key) != BTRFS_EXTENT_DATA_KEY)
 			break;
 		if (btrfs_disk_key_offset(found_key) < inode->i_size)
 			break;
-		fi = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]),
-				    path.slots[0],
-				    struct btrfs_file_extent_item);
-		extent_start = btrfs_file_extent_disk_blocknr(fi);
-		extent_num_blocks = btrfs_file_extent_disk_num_blocks(fi);
-		key.offset = btrfs_disk_key_offset(found_key) - 1;
+		if (btrfs_disk_key_type(found_key) == BTRFS_EXTENT_DATA_KEY) {
+			fi = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]),
+					    path.slots[0],
+					    struct btrfs_file_extent_item);
+			extent_start = btrfs_file_extent_disk_blocknr(fi);
+			extent_num_blocks =
+				btrfs_file_extent_disk_num_blocks(fi);
+			inode->i_blocks -=
+				btrfs_file_extent_num_blocks(fi) >> 9;
+			found_extent = 1;
+		} else {
+			found_extent = 0;
+		}
 		ret = btrfs_del_item(trans, root, &path);
 		BUG_ON(ret);
-		inode->i_blocks -= btrfs_file_extent_num_blocks(fi) >> 9;
 		btrfs_release_path(root, &path);
-		ret = btrfs_free_extent(trans, root, extent_start,
-					extent_num_blocks, 0);
-		BUG_ON(ret);
-		if (key.offset + 1 == 0)
-			break;
+		if (found_extent) {
+			ret = btrfs_free_extent(trans, root, extent_start,
+						extent_num_blocks, 0);
+			BUG_ON(ret);
+		}
 	}
 	btrfs_release_path(root, &path);
 	ret = 0;
@@ -975,10 +983,24 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 	int err = 0;
 	int ret;
 	int this_write;
+	struct inode *inode = file->f_path.dentry->d_inode;
 
 	for (i = 0; i < num_pages; i++) {
 		offset = pos & (PAGE_CACHE_SIZE -1);
 		this_write = min(PAGE_CACHE_SIZE - offset, write_bytes);
+		/* FIXME, one block at a time */
+
+		mutex_lock(&root->fs_info->fs_mutex);
+		trans = btrfs_start_transaction(root, 1);
+		btrfs_csum_file_block(trans, root, inode->i_ino,
+				      pages[i]->index << PAGE_CACHE_SHIFT,
+				      kmap(pages[i]), PAGE_CACHE_SIZE);
+		kunmap(pages[i]);
+		SetPageChecked(pages[i]);
+		ret = btrfs_end_transaction(trans, root);
+		BUG_ON(ret);
+		mutex_unlock(&root->fs_info->fs_mutex);
+
 		ret = nobh_commit_write(file, pages[i], offset,
 					 offset + this_write);
 		pos += this_write;
@@ -1022,7 +1044,7 @@ static int prepare_pages(struct btrfs_trans_handle *trans,
 		this_write = min(PAGE_CACHE_SIZE - offset, write_bytes);
 		ret = nobh_prepare_write(pages[i], offset,
 					 offset + this_write,
-					 btrfs_get_block_lock);
+					 btrfs_get_block);
 		pos += this_write;
 		if (ret) {
 			err = ret;
@@ -1051,7 +1073,6 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 	size_t num_written = 0;
 	int err = 0;
 	int ret = 0;
-	struct btrfs_trans_handle *trans;
 	struct inode *inode = file->f_path.dentry->d_inode;
 	struct btrfs_root *root = btrfs_sb(inode->i_sb);
 	struct page *pages[1];
@@ -1077,25 +1098,18 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 		size_t write_bytes = min(count, PAGE_CACHE_SIZE - offset);
 		size_t num_pages = (write_bytes + PAGE_CACHE_SIZE - 1) >>
 					PAGE_CACHE_SHIFT;
-		mutex_lock(&root->fs_info->fs_mutex);
-		trans = btrfs_start_transaction(root, 1);
-
-		ret = prepare_pages(trans, root, file, pages, num_pages,
+		ret = prepare_pages(NULL, root, file, pages, num_pages,
 				    pos, write_bytes);
 		BUG_ON(ret);
 		ret = btrfs_copy_from_user(pos, num_pages,
 					   write_bytes, pages, buf);
 		BUG_ON(ret);
 
-		mutex_unlock(&root->fs_info->fs_mutex);
-
-		ret = dirty_and_release_pages(trans, root, file, pages,
+		ret = dirty_and_release_pages(NULL, root, file, pages,
 					      num_pages, pos, write_bytes);
 		BUG_ON(ret);
 		btrfs_drop_pages(pages, num_pages);
 
-		ret = btrfs_end_transaction(trans, root);
-
 		buf += write_bytes;
 		count -= write_bytes;
 		pos += write_bytes;
@@ -1111,6 +1125,118 @@ out:
 	return num_written ? num_written : err;
 }
 
+static int btrfs_read_actor(read_descriptor_t *desc, struct page *page,
+			unsigned long offset, unsigned long size)
+{
+	char *kaddr;
+	unsigned long left, count = desc->count;
+
+	if (size > count)
+		size = count;
+
+	if (!PageChecked(page)) {
+		/* FIXME, do it per block */
+		struct btrfs_root *root = btrfs_sb(page->mapping->host->i_sb);
+		int ret = btrfs_csum_verify_file_block(root,
+					  page->mapping->host->i_ino,
+					  page->index << PAGE_CACHE_SHIFT,
+					  kmap(page), PAGE_CACHE_SIZE);
+		if (ret) {
+			printk("failed to verify ino %lu page %lu\n",
+			       page->mapping->host->i_ino,
+			       page->index);
+			memset(page_address(page), 0, PAGE_CACHE_SIZE);
+		}
+		SetPageChecked(page);
+		kunmap(page);
+	}
+	/*
+	 * Faults on the destination of a read are common, so do it before
+	 * taking the kmap.
+	 */
+	if (!fault_in_pages_writeable(desc->arg.buf, size)) {
+		kaddr = kmap_atomic(page, KM_USER0);
+		left = __copy_to_user_inatomic(desc->arg.buf,
+						kaddr + offset, size);
+		kunmap_atomic(kaddr, KM_USER0);
+		if (left == 0)
+			goto success;
+	}
+
+	/* Do it the slow way */
+	kaddr = kmap(page);
+	left = __copy_to_user(desc->arg.buf, kaddr + offset, size);
+	kunmap(page);
+
+	if (left) {
+		size -= left;
+		desc->error = -EFAULT;
+	}
+success:
+	desc->count = count - size;
+	desc->written += size;
+	desc->arg.buf += size;
+	return size;
+}
+
+/**
+ * btrfs_file_aio_read - filesystem read routine
+ * @iocb:	kernel I/O control block
+ * @iov:	io vector request
+ * @nr_segs:	number of segments in the iovec
+ * @pos:	current file position
+ */
+static ssize_t btrfs_file_aio_read(struct kiocb *iocb, const struct iovec *iov,
+				   unsigned long nr_segs, loff_t pos)
+{
+	struct file *filp = iocb->ki_filp;
+	ssize_t retval;
+	unsigned long seg;
+	size_t count;
+	loff_t *ppos = &iocb->ki_pos;
+
+	count = 0;
+	for (seg = 0; seg < nr_segs; seg++) {
+		const struct iovec *iv = &iov[seg];
+
+		/*
+		 * If any segment has a negative length, or the cumulative
+		 * length ever wraps negative then return -EINVAL.
+		 */
+		count += iv->iov_len;
+		if (unlikely((ssize_t)(count|iv->iov_len) < 0))
+			return -EINVAL;
+		if (access_ok(VERIFY_WRITE, iv->iov_base, iv->iov_len))
+			continue;
+		if (seg == 0)
+			return -EFAULT;
+		nr_segs = seg;
+		count -= iv->iov_len;	/* This segment is no good */
+		break;
+	}
+	retval = 0;
+	if (count) {
+		for (seg = 0; seg < nr_segs; seg++) {
+			read_descriptor_t desc;
+
+			desc.written = 0;
+			desc.arg.buf = iov[seg].iov_base;
+			desc.count = iov[seg].iov_len;
+			if (desc.count == 0)
+				continue;
+			desc.error = 0;
+			do_generic_file_read(filp, ppos, &desc,
+					     btrfs_read_actor);
+			retval += desc.written;
+			if (desc.error) {
+				retval = retval ?: desc.error;
+				break;
+			}
+		}
+	}
+	return retval;
+}
+
 static int btrfs_get_sb(struct file_system_type *fs_type,
 	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
 {
@@ -1166,7 +1292,7 @@ static struct inode_operations btrfs_file_inode_operations = {
 static struct file_operations btrfs_file_operations = {
 	.llseek		= generic_file_llseek,
 	.read		= do_sync_read,
-	.aio_read       = generic_file_aio_read,
+	.aio_read       = btrfs_file_aio_read,
 	.write		= btrfs_file_write,
 	.mmap		= generic_file_mmap,
 	.open		= generic_file_open,
-- 
cgit v1.2.3


From 2c90e5d658424bc71b111eb5a972240d5d06fe86 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Mon, 2 Apr 2007 10:50:19 -0400
Subject: Btrfs: still corruption hunting

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/bit-radix.c   |   7 +--
 fs/btrfs/btrfs_inode.h |  14 +++++
 fs/btrfs/ctree.c       |  42 ++++++++++----
 fs/btrfs/ctree.h       |   3 +
 fs/btrfs/disk-io.c     |  84 ++++++++++++++--------------
 fs/btrfs/disk-io.h     |   4 +-
 fs/btrfs/extent-tree.c |  18 ++++--
 fs/btrfs/super.c       | 146 ++++++++++++++++++++++++++++++++++++++++++++-----
 fs/btrfs/transaction.c |  39 +++++++++----
 fs/btrfs/transaction.h |   3 +
 10 files changed, 265 insertions(+), 95 deletions(-)
 create mode 100644 fs/btrfs/btrfs_inode.h

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/bit-radix.c b/fs/btrfs/bit-radix.c
index 845422368f48..a50a1c44eabd 100644
--- a/fs/btrfs/bit-radix.c
+++ b/fs/btrfs/bit-radix.c
@@ -4,6 +4,7 @@
 #define BIT_ARRAY_BYTES 256
 #define BIT_RADIX_BITS_PER_ARRAY ((BIT_ARRAY_BYTES - sizeof(unsigned long)) * 8)
 
+extern struct kmem_cache *btrfs_bit_radix_cachep;
 int set_radix_bit(struct radix_tree_root *radix, unsigned long bit)
 {
 	unsigned long *bits;
@@ -16,7 +17,7 @@ int set_radix_bit(struct radix_tree_root *radix, unsigned long bit)
 
 	bits = radix_tree_lookup(radix, slot);
 	if (!bits) {
-		bits = kmalloc(BIT_ARRAY_BYTES, GFP_NOFS);
+		bits = kmem_cache_alloc(btrfs_bit_radix_cachep, GFP_NOFS);
 		if (!bits)
 			return -ENOMEM;
 		memset(bits + 1, 0, BIT_ARRAY_BYTES - sizeof(unsigned long));
@@ -68,12 +69,10 @@ int clear_radix_bit(struct radix_tree_root *radix, unsigned long bit)
 			break;
 		}
 	}
-
 	if (empty) {
 		bits = radix_tree_delete(radix, slot);
-		synchronize_rcu();
 		BUG_ON(!bits);
-		kfree(bits);
+		kmem_cache_free(btrfs_bit_radix_cachep, bits);
 	}
 #endif
 	return 0;
diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
new file mode 100644
index 000000000000..e159841650a4
--- /dev/null
+++ b/fs/btrfs/btrfs_inode.h
@@ -0,0 +1,14 @@
+#ifndef __BTRFS_I__
+#define __BTRFS_I__
+
+struct btrfs_inode {
+	u32 magic;
+	struct inode vfs_inode;
+	u32 magic2;
+};
+static inline struct btrfs_inode *BTRFS_I(struct inode *inode)
+{
+	return container_of(inode, struct btrfs_inode, vfs_inode);
+}
+
+#endif
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index a0dfa2d6cb9b..453ce835209a 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -16,6 +16,16 @@ static int balance_node_right(struct btrfs_trans_handle *trans, struct
 static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		   struct btrfs_path *path, int level, int slot);
 
+struct btrfs_path *btrfs_alloc_path(void)
+{
+	return kmem_cache_alloc(btrfs_path_cachep, GFP_NOFS);
+}
+
+void btrfs_free_path(struct btrfs_path *p)
+{
+	kmem_cache_free(btrfs_path_cachep, p);
+}
+
 inline void btrfs_init_path(struct btrfs_path *p)
 {
 	memset(p, 0, sizeof(*p));
@@ -47,17 +57,18 @@ static int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root
 	}
 	cow = btrfs_alloc_free_block(trans, root);
 	cow_node = btrfs_buffer_node(cow);
+	if (buf->b_size != root->blocksize || cow->b_size != root->blocksize)
+		WARN_ON(1);
 	memcpy(cow_node, btrfs_buffer_node(buf), root->blocksize);
 	btrfs_set_header_blocknr(&cow_node->header, cow->b_blocknr);
 	btrfs_set_header_generation(&cow_node->header, trans->transid);
-	*cow_ret = cow;
-	btrfs_mark_buffer_dirty(cow);
 	btrfs_inc_ref(trans, root, buf);
 	if (buf == root->node) {
 		root->node = cow;
 		get_bh(cow);
-		if (buf != root->commit_root)
+		if (buf != root->commit_root) {
 			btrfs_free_extent(trans, root, buf->b_blocknr, 1, 1);
+		}
 		btrfs_block_release(root, buf);
 	} else {
 		btrfs_set_node_blockptr(btrfs_buffer_node(parent), parent_slot,
@@ -66,6 +77,7 @@ static int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root
 		btrfs_free_extent(trans, root, buf->b_blocknr, 1, 1);
 	}
 	btrfs_block_release(root, buf);
+	*cow_ret = cow;
 	return 0;
 }
 
@@ -477,9 +489,12 @@ again:
 					       p->slots[level + 1],
 					       &cow_buf);
 			b = cow_buf;
+			c = btrfs_buffer_node(b);
 		}
 		BUG_ON(!cow && ins_len);
-		c = btrfs_buffer_node(b);
+		if (level != btrfs_header_level(&c->header))
+			WARN_ON(1);
+		level = btrfs_header_level(&c->header);
 		p->nodes[level] = b;
 		ret = check_block(root, p, level);
 		if (ret)
@@ -1257,19 +1272,22 @@ int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root
 		      data_size)
 {
 	int ret = 0;
-	struct btrfs_path path;
+	struct btrfs_path *path;
 	u8 *ptr;
 
-	btrfs_init_path(&path);
-	ret = btrfs_insert_empty_item(trans, root, &path, cpu_key, data_size);
+	path = btrfs_alloc_path();
+	BUG_ON(!path);
+	btrfs_init_path(path);
+	ret = btrfs_insert_empty_item(trans, root, path, cpu_key, data_size);
 	if (!ret) {
-		ptr = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]),
-				     path.slots[0], u8);
-		btrfs_memcpy(root, path.nodes[0]->b_data,
+		ptr = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
+				     path->slots[0], u8);
+		btrfs_memcpy(root, path->nodes[0]->b_data,
 			     ptr, data, data_size);
-		btrfs_mark_buffer_dirty(path.nodes[0]);
+		btrfs_mark_buffer_dirty(path->nodes[0]);
 	}
-	btrfs_release_path(root, &path);
+	btrfs_release_path(root, path);
+	btrfs_free_path(path);
 	return ret;
 }
 
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 9ec0d65ebe9b..d8e03bd797ff 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -7,6 +7,7 @@
 
 struct btrfs_trans_handle;
 struct btrfs_transaction;
+extern struct kmem_cache *btrfs_path_cachep;
 
 #define BTRFS_MAGIC "_BtRfS_M"
 
@@ -888,6 +889,8 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_key *key, struct btrfs_path *p, int
 		      ins_len, int cow);
 void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p);
+struct btrfs_path *btrfs_alloc_path(void);
+void btrfs_free_path(struct btrfs_path *p);
 void btrfs_init_path(struct btrfs_path *p);
 int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		   struct btrfs_path *path);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index bb133104e2e9..2dbd55084a4e 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -8,18 +8,6 @@
 #include "disk-io.h"
 #include "transaction.h"
 
-#define PATTERN 0xDEADBEEFUL
-static inline void check_pattern(struct buffer_head *buf)
-{
-	if (buf->b_private != (void *)PATTERN)
-		WARN_ON(1);
-}
-
-static inline void set_pattern(struct buffer_head *buf)
-{
-	buf->b_private = (void *)PATTERN;
-}
-
 static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
 {
 	struct btrfs_node *node = btrfs_buffer_node(buf);
@@ -35,6 +23,8 @@ static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
 
 struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr)
 {
+	return sb_find_get_block(root->fs_info->sb, blocknr);
+#if 0
 	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
 	int blockbits = root->fs_info->sb->s_blocksize_bits;
 	unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
@@ -43,6 +33,7 @@ struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr)
 	struct buffer_head *head;
 	struct buffer_head *ret = NULL;
 
+
 	page = find_lock_page(mapping, index);
 	if (!page)
 		return NULL;
@@ -64,15 +55,17 @@ out_unlock:
 	unlock_page(page);
 	if (ret) {
 		touch_buffer(ret);
-		check_pattern(ret);
 	}
 	page_cache_release(page);
 	return ret;
+#endif
 }
 
 struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
 						 u64 blocknr)
 {
+	return sb_getblk(root->fs_info->sb, blocknr);
+#if 0
 	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
 	int blockbits = root->fs_info->sb->s_blocksize_bits;
 	unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
@@ -95,7 +88,6 @@ struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
 			bh->b_bdev = root->fs_info->sb->s_bdev;
 			bh->b_blocknr = first_block;
 			set_buffer_mapped(bh);
-			set_pattern(bh);
 		}
 		if (bh->b_blocknr == blocknr) {
 			ret = bh;
@@ -111,6 +103,7 @@ out_unlock:
 		touch_buffer(ret);
 	page_cache_release(page);
 	return ret;
+#endif
 }
 
 static sector_t max_block(struct block_device *bdev)
@@ -225,6 +218,8 @@ static struct address_space_operations btree_aops = {
 
 struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr)
 {
+	return sb_bread(root->fs_info->sb, blocknr);
+#if 0
 	struct buffer_head *bh = NULL;
 
 	bh = btrfs_find_create_tree_block(root, blocknr);
@@ -239,7 +234,6 @@ struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr)
 		if (!buffer_uptodate(bh))
 			goto fail;
 		csum_tree_block(root, bh, 1);
-		set_pattern(bh);
 	} else {
 		unlock_buffer(bh);
 	}
@@ -250,6 +244,7 @@ fail:
 	brelse(bh);
 	return NULL;
 
+#endif
 }
 
 int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
@@ -268,14 +263,14 @@ int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	return 0;
 }
 
-static int __setup_root(struct btrfs_super_block *super,
+static int __setup_root(int blocksize,
 			struct btrfs_root *root,
 			struct btrfs_fs_info *fs_info,
 			u64 objectid)
 {
 	root->node = NULL;
 	root->commit_root = NULL;
-	root->blocksize = btrfs_super_blocksize(super);
+	root->blocksize = blocksize;
 	root->ref_cows = 0;
 	root->fs_info = fs_info;
 	memset(&root->root_key, 0, sizeof(root->root_key));
@@ -283,7 +278,7 @@ static int __setup_root(struct btrfs_super_block *super,
 	return 0;
 }
 
-static int find_and_setup_root(struct btrfs_super_block *super,
+static int find_and_setup_root(int blocksize,
 			       struct btrfs_root *tree_root,
 			       struct btrfs_fs_info *fs_info,
 			       u64 objectid,
@@ -291,7 +286,7 @@ static int find_and_setup_root(struct btrfs_super_block *super,
 {
 	int ret;
 
-	__setup_root(super, root, fs_info, objectid);
+	__setup_root(blocksize, root, fs_info, objectid);
 	ret = btrfs_find_last_root(tree_root, objectid,
 				   &root->root_item, &root->root_key);
 	BUG_ON(ret);
@@ -302,9 +297,7 @@ static int find_and_setup_root(struct btrfs_super_block *super,
 	return 0;
 }
 
-struct btrfs_root *open_ctree(struct super_block *sb,
-			      struct buffer_head *sb_buffer,
-			      struct btrfs_super_block *disk_super)
+struct btrfs_root *open_ctree(struct super_block *sb)
 {
 	struct btrfs_root *root = kmalloc(sizeof(struct btrfs_root),
 					  GFP_NOFS);
@@ -317,13 +310,11 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
 						GFP_NOFS);
 	int ret;
+	struct btrfs_super_block *disk_super;
 
-	if (!btrfs_super_root(disk_super)) {
-		return NULL;
-	}
 	init_bit_radix(&fs_info->pinned_radix);
 	init_bit_radix(&fs_info->pending_del_radix);
-	sb_set_blocksize(sb, sb_buffer->b_size);
+	sb_set_blocksize(sb, 4096);
 	fs_info->running_transaction = NULL;
 	fs_info->fs_root = root;
 	fs_info->tree_root = tree_root;
@@ -331,55 +322,59 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	fs_info->inode_root = inode_root;
 	fs_info->last_inode_alloc = 0;
 	fs_info->last_inode_alloc_dirid = 0;
-	fs_info->disk_super = disk_super;
 	fs_info->sb = sb;
+	fs_info->btree_inode = NULL;
+#if 0
 	fs_info->btree_inode = new_inode(sb);
 	fs_info->btree_inode->i_ino = 1;
+	fs_info->btree_inode->i_nlink = 1;
 	fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
 	fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
 	insert_inode_hash(fs_info->btree_inode);
-
 	mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
+#endif
 	fs_info->hash_tfm = crypto_alloc_hash("sha256", 0, CRYPTO_ALG_ASYNC);
 	spin_lock_init(&fs_info->hash_lock);
-
 	if (!fs_info->hash_tfm || IS_ERR(fs_info->hash_tfm)) {
 		printk("failed to allocate sha256 hash\n");
 		return NULL;
 	}
-
 	mutex_init(&fs_info->trans_mutex);
 	mutex_init(&fs_info->fs_mutex);
 	memset(&fs_info->current_insert, 0, sizeof(fs_info->current_insert));
 	memset(&fs_info->last_insert, 0, sizeof(fs_info->last_insert));
 
-	__setup_root(disk_super, tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID);
-
-	fs_info->sb_buffer = read_tree_block(tree_root, sb_buffer->b_blocknr);
+	__setup_root(sb->s_blocksize, tree_root,
+		     fs_info, BTRFS_ROOT_TREE_OBJECTID);
+	fs_info->sb_buffer = read_tree_block(tree_root,
+					     BTRFS_SUPER_INFO_OFFSET /
+					     sb->s_blocksize);
 
 	if (!fs_info->sb_buffer) {
 printk("failed2\n");
 		return NULL;
 	}
-	brelse(sb_buffer);
-	sb_buffer = NULL;
 	disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
+	if (!btrfs_super_root(disk_super)) {
+		return NULL;
+	}
 	fs_info->disk_super = disk_super;
-
 	tree_root->node = read_tree_block(tree_root,
 					  btrfs_super_root(disk_super));
 	BUG_ON(!tree_root->node);
 
-	ret = find_and_setup_root(disk_super, tree_root, fs_info,
+	mutex_lock(&fs_info->fs_mutex);
+	ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
 				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
 	BUG_ON(ret);
 
-	ret = find_and_setup_root(disk_super, tree_root, fs_info,
+	ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
 				  BTRFS_INODE_MAP_OBJECTID, inode_root);
 	BUG_ON(ret);
 
-	ret = find_and_setup_root(disk_super, tree_root, fs_info,
+	ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
 				  BTRFS_FS_TREE_OBJECTID, root);
+	mutex_unlock(&fs_info->fs_mutex);
 	BUG_ON(ret);
 	root->commit_root = root->node;
 	get_bh(root->node);
@@ -392,9 +387,11 @@ int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root)
 {
 	struct buffer_head *bh = root->fs_info->sb_buffer;
+
 	btrfs_set_super_root(root->fs_info->disk_super,
 			     root->fs_info->tree_root->node->b_blocknr);
 	lock_buffer(bh);
+	WARN_ON(atomic_read(&bh->b_count) < 1);
 	clear_buffer_dirty(bh);
 	csum_tree_block(root, bh, 0);
 	bh->b_end_io = end_buffer_write_sync;
@@ -413,6 +410,7 @@ int close_ctree(struct btrfs_root *root)
 	int ret;
 	struct btrfs_trans_handle *trans;
 
+	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
 	btrfs_commit_transaction(trans, root);
 	/* run commit again to  drop the original snapshot */
@@ -421,6 +419,7 @@ int close_ctree(struct btrfs_root *root)
 	ret = btrfs_write_and_wait_transaction(NULL, root);
 	BUG_ON(ret);
 	write_ctree_super(NULL, root);
+	mutex_unlock(&root->fs_info->fs_mutex);
 
 	if (root->node)
 		btrfs_block_release(root, root->node);
@@ -436,8 +435,8 @@ int close_ctree(struct btrfs_root *root)
 	btrfs_block_release(root, root->commit_root);
 	btrfs_block_release(root, root->fs_info->sb_buffer);
 	crypto_free_hash(root->fs_info->hash_tfm);
-	truncate_inode_pages(root->fs_info->btree_inode->i_mapping, 0);
-	iput(root->fs_info->btree_inode);
+	// truncate_inode_pages(root->fs_info->btree_inode->i_mapping, 0);
+	// iput(root->fs_info->btree_inode);
 	kfree(root->fs_info->extent_root);
 	kfree(root->fs_info->inode_root);
 	kfree(root->fs_info->tree_root);
@@ -448,7 +447,6 @@ int close_ctree(struct btrfs_root *root)
 
 void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf)
 {
-	check_pattern(buf);
-	brelse(buf);
+	// brelse(buf);
 }
 
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index f6998e2192ce..ac6764ba8aa6 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -31,9 +31,7 @@ int clean_tree_block(struct btrfs_trans_handle *trans,
 		     struct btrfs_root *root, struct buffer_head *buf);
 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root);
-struct btrfs_root *open_ctree(struct super_block *sb,
-			      struct buffer_head *sb_buffer,
-			      struct btrfs_super_block *disk_super);
+struct btrfs_root *open_ctree(struct super_block *sb);
 int close_ctree(struct btrfs_root *root);
 void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf);
 int write_ctree_super(struct btrfs_trans_handle *trans,
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index c4194dab7a33..37b87e28a2f3 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -173,12 +173,16 @@ static int pin_down_block(struct btrfs_root *root, u64 blocknr, int pending)
 
 	if (!pending) {
 		bh = btrfs_find_tree_block(root, blocknr);
-		if (bh && buffer_uptodate(bh)) {
-			header = btrfs_buffer_header(bh);
-			if (btrfs_header_generation(header) ==
-			    root->fs_info->running_transaction->transid) {
-				btrfs_block_release(root, bh);
-				return 0;
+		if (bh) {
+			if (buffer_uptodate(bh)) {
+				u64 transid =
+				    root->fs_info->running_transaction->transid;
+				header = btrfs_buffer_header(bh);
+				if (btrfs_header_generation(header) ==
+				    transid) {
+					btrfs_block_release(root, bh);
+					return 0;
+				}
 			}
 			btrfs_block_release(root, bh);
 		}
@@ -539,6 +543,8 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 	 */
 	while(*level >= 0) {
 		cur = path->nodes[*level];
+		if (btrfs_header_level(btrfs_buffer_header(cur)) != *level)
+			WARN_ON(1);
 		if (path->slots[*level] >=
 		    btrfs_header_nritems(btrfs_buffer_header(cur)))
 			break;
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 894a70bf26d2..6969b672b570 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -14,6 +14,7 @@
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
+#include "btrfs_inode.h"
 
 #define BTRFS_SUPER_MAGIC 0x9123682E
 
@@ -24,6 +25,14 @@ static struct inode_operations btrfs_file_inode_operations;
 static struct address_space_operations btrfs_aops;
 static struct file_operations btrfs_file_operations;
 
+static int check_inode(struct inode *inode)
+{
+	struct btrfs_inode *ei = BTRFS_I(inode);
+	WARN_ON(ei->magic != 0xDEADBEEF);
+	WARN_ON(ei->magic2 != 0xDEADBEAF);
+	return 0;
+}
+
 static void btrfs_read_locked_inode(struct inode *inode)
 {
 	struct btrfs_path path;
@@ -34,6 +43,7 @@ static void btrfs_read_locked_inode(struct inode *inode)
 	btrfs_init_path(&path);
 	mutex_lock(&root->fs_info->fs_mutex);
 
+	check_inode(inode);
 	ret = btrfs_lookup_inode(NULL, root, &path, inode->i_ino, 0);
 	if (ret) {
 		btrfs_release_path(root, &path);
@@ -41,6 +51,7 @@ static void btrfs_read_locked_inode(struct inode *inode)
 		make_bad_inode(inode);
 		return;
 	}
+	check_inode(inode);
 	inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]),
 				  path.slots[0],
 				  struct btrfs_inode_item);
@@ -60,6 +71,7 @@ static void btrfs_read_locked_inode(struct inode *inode)
 	inode->i_generation = btrfs_inode_generation(inode_item);
 	btrfs_release_path(root, &path);
 	mutex_unlock(&root->fs_info->fs_mutex);
+	check_inode(inode);
 	switch (inode->i_mode & S_IFMT) {
 #if 0
 	default:
@@ -80,6 +92,7 @@ static void btrfs_read_locked_inode(struct inode *inode)
 		// inode->i_op = &page_symlink_inode_operations;
 		break;
 	}
+	check_inode(inode);
 	return;
 }
 
@@ -347,6 +360,7 @@ static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
 				    namelen, 0);
 	if (ret || !btrfs_match_dir_item_name(root, &path, name, namelen)) {
 		*ino = 0;
+		ret = 0;
 		goto out;
 	}
 	di = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]), path.slots[0],
@@ -354,6 +368,7 @@ static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
 	*ino = btrfs_dir_objectid(di);
 out:
 	btrfs_release_path(root, &path);
+	check_inode(dir);
 	return ret;
 }
 
@@ -367,7 +382,6 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
 
 	if (dentry->d_name.len > BTRFS_NAME_LEN)
 		return ERR_PTR(-ENAMETOOLONG);
-
 	mutex_lock(&root->fs_info->fs_mutex);
 	ret = btrfs_inode_by_name(dir, dentry, &ino);
 	mutex_unlock(&root->fs_info->fs_mutex);
@@ -378,7 +392,9 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
 		inode = iget(dir->i_sb, ino);
 		if (!inode)
 			return ERR_PTR(-EACCES);
+		check_inode(inode);
 	}
+	check_inode(dir);
 	return d_splice_alias(inode, dentry);
 }
 
@@ -471,23 +487,14 @@ static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
 	struct inode * inode;
 	struct dentry * root_dentry;
 	struct btrfs_super_block *disk_super;
-	struct buffer_head *bh;
 	struct btrfs_root *root;
 
 	sb->s_maxbytes = MAX_LFS_FILESIZE;
-	sb->s_blocksize = PAGE_CACHE_SIZE;
-	sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
 	sb->s_magic = BTRFS_SUPER_MAGIC;
 	sb->s_op = &btrfs_super_ops;
 	sb->s_time_gran = 1;
 
-	bh = sb_bread(sb, BTRFS_SUPER_INFO_OFFSET / sb->s_blocksize);
-	if (!bh) {
-		printk("btrfs: unable to read on disk super\n");
-		return -EIO;
-	}
-	disk_super = (struct btrfs_super_block *)bh->b_data;
-	root = open_ctree(sb, bh, disk_super);
+	root = open_ctree(sb);
 
 	if (!root) {
 		printk("btrfs: open_ctree failed\n");
@@ -533,6 +540,7 @@ static void fill_inode_item(struct btrfs_inode_item *item,
 	btrfs_set_timespec_nsec(&item->ctime, inode->i_ctime.tv_nsec);
 	btrfs_set_inode_nblocks(item, inode->i_blocks);
 	btrfs_set_inode_generation(item, inode->i_generation);
+	check_inode(inode);
 }
 
 static int btrfs_update_inode(struct btrfs_trans_handle *trans,
@@ -560,6 +568,7 @@ static int btrfs_update_inode(struct btrfs_trans_handle *trans,
 	btrfs_mark_buffer_dirty(path.nodes[0]);
 failed:
 	btrfs_release_path(root, &path);
+	check_inode(inode);
 	return 0;
 }
 
@@ -577,6 +586,7 @@ static int btrfs_write_inode(struct inode *inode, int wait)
 	else
 		btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
+	check_inode(inode);
 	return ret;
 }
 
@@ -594,6 +604,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 	if (!inode)
 		return ERR_PTR(-ENOMEM);
 
+	check_inode(inode);
 	ret = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
 	BUG_ON(ret);
 
@@ -616,6 +627,8 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 	BUG_ON(ret);
 
 	insert_inode_hash(inode);
+	check_inode(inode);
+	check_inode(dir);
 	return inode;
 }
 
@@ -632,7 +645,8 @@ static int btrfs_add_link(struct btrfs_trans_handle *trans,
 		ret = btrfs_update_inode(trans, btrfs_sb(inode->i_sb),
 					 dentry->d_parent->d_inode);
 	}
-
+	check_inode(inode);
+	check_inode(dentry->d_parent->d_inode);
 	return ret;
 }
 
@@ -644,6 +658,9 @@ static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
 		d_instantiate(dentry, inode);
 		return 0;
 	}
+	if (err > 0)
+		err = -EEXIST;
+	check_inode(inode);
 	return err;
 }
 
@@ -675,6 +692,9 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry,
 out_unlock:
 	btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
+	check_inode(inode);
+	check_inode(dir);
+
 	if (drop_inode) {
 		inode_dec_link_count(inode);
 		iput(inode);
@@ -755,11 +775,11 @@ static int btrfs_sync_fs(struct super_block *sb, int wait)
 
 	sb->s_dirt = 0;
 	if (!wait) {
-		filemap_flush(root->fs_info->btree_inode->i_mapping);
+		// filemap_flush(root->fs_info->btree_inode->i_mapping);
+		filemap_flush(root->fs_info->sb->s_bdev->bd_inode->i_mapping);
 		return 0;
 	}
-	filemap_write_and_wait(root->fs_info->btree_inode->i_mapping);
-
+	filemap_write_and_wait(root->fs_info->sb->s_bdev->bd_inode->i_mapping);
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
 	ret = btrfs_commit_transaction(trans, root);
@@ -1242,6 +1262,95 @@ static ssize_t btrfs_file_aio_read(struct kiocb *iocb, const struct iovec *iov,
 	return retval;
 }
 
+static struct kmem_cache *btrfs_inode_cachep;
+struct kmem_cache *btrfs_trans_handle_cachep;
+struct kmem_cache *btrfs_transaction_cachep;
+struct kmem_cache *btrfs_bit_radix_cachep;
+struct kmem_cache *btrfs_path_cachep;
+
+/*
+ * Called inside transaction, so use GFP_NOFS
+ */
+static struct inode *btrfs_alloc_inode(struct super_block *sb)
+{
+	struct btrfs_inode *ei;
+
+	ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
+	if (!ei)
+		return NULL;
+	ei->magic = 0xDEADBEEF;
+	ei->magic2 = 0xDEADBEAF;
+	return &ei->vfs_inode;
+}
+
+static void btrfs_destroy_inode(struct inode *inode)
+{
+	struct btrfs_inode *ei = BTRFS_I(inode);
+	WARN_ON(ei->magic != 0xDEADBEEF);
+	WARN_ON(ei->magic2 != 0xDEADBEAF);
+	WARN_ON(!list_empty(&inode->i_dentry));
+	WARN_ON(inode->i_ino == 1);
+	WARN_ON(inode->i_data.nrpages);
+
+	ei->magic = 0;
+	ei->magic2 = 0;
+	kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
+}
+
+static void init_once(void * foo, struct kmem_cache * cachep,
+		      unsigned long flags)
+{
+	struct btrfs_inode *ei = (struct btrfs_inode *) foo;
+
+	if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
+	    SLAB_CTOR_CONSTRUCTOR) {
+		inode_init_once(&ei->vfs_inode);
+	}
+}
+
+static int init_inodecache(void)
+{
+	btrfs_inode_cachep = kmem_cache_create("btrfs_inode_cache",
+					     sizeof(struct btrfs_inode),
+					     0, (SLAB_RECLAIM_ACCOUNT|
+						SLAB_MEM_SPREAD),
+					     init_once, NULL);
+	btrfs_trans_handle_cachep = kmem_cache_create("btrfs_trans_handle_cache",
+					     sizeof(struct btrfs_trans_handle),
+					     0, (SLAB_RECLAIM_ACCOUNT|
+						SLAB_MEM_SPREAD),
+					     NULL, NULL);
+	btrfs_transaction_cachep = kmem_cache_create("btrfs_transaction_cache",
+					     sizeof(struct btrfs_transaction),
+					     0, (SLAB_RECLAIM_ACCOUNT|
+						SLAB_MEM_SPREAD),
+					     NULL, NULL);
+	btrfs_path_cachep = kmem_cache_create("btrfs_path_cache",
+					     sizeof(struct btrfs_transaction),
+					     0, (SLAB_RECLAIM_ACCOUNT|
+						SLAB_MEM_SPREAD),
+					     NULL, NULL);
+	btrfs_bit_radix_cachep = kmem_cache_create("btrfs_radix",
+					     256,
+					     0, (SLAB_RECLAIM_ACCOUNT|
+						SLAB_MEM_SPREAD |
+						SLAB_DESTROY_BY_RCU),
+					     NULL, NULL);
+	if (btrfs_inode_cachep == NULL || btrfs_trans_handle_cachep == NULL ||
+	    btrfs_transaction_cachep == NULL || btrfs_bit_radix_cachep == NULL)
+		return -ENOMEM;
+	return 0;
+}
+
+static void destroy_inodecache(void)
+{
+	kmem_cache_destroy(btrfs_inode_cachep);
+	kmem_cache_destroy(btrfs_trans_handle_cachep);
+	kmem_cache_destroy(btrfs_transaction_cachep);
+	kmem_cache_destroy(btrfs_bit_radix_cachep);
+	kmem_cache_destroy(btrfs_path_cachep);
+}
+
 static int btrfs_get_sb(struct file_system_type *fs_type,
 	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
 {
@@ -1265,6 +1374,8 @@ static struct super_operations btrfs_super_ops = {
 	.write_super	= btrfs_write_super,
 	.sync_fs	= btrfs_sync_fs,
 	.write_inode	= btrfs_write_inode,
+	.alloc_inode	= btrfs_alloc_inode,
+	.destroy_inode	= btrfs_destroy_inode,
 };
 
 static struct inode_operations btrfs_dir_inode_operations = {
@@ -1305,12 +1416,17 @@ static struct file_operations btrfs_file_operations = {
 
 static int __init init_btrfs_fs(void)
 {
+	int err;
 	printk("btrfs loaded!\n");
+	err = init_inodecache();
+	if (err)
+		return err;
 	return register_filesystem(&btrfs_fs_type);
 }
 
 static void __exit exit_btrfs_fs(void)
 {
+	destroy_inodecache();
 	unregister_filesystem(&btrfs_fs_type);
 	printk("btrfs unloaded\n");
 }
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 84c4e278ce2e..72b52e1e0b1b 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -5,13 +5,20 @@
 #include "transaction.h"
 
 static int total_trans = 0;
+extern struct kmem_cache *btrfs_trans_handle_cachep;
+extern struct kmem_cache *btrfs_transaction_cachep;
+
+#define TRANS_MAGIC 0xE1E10E
 static void put_transaction(struct btrfs_transaction *transaction)
 {
+	WARN_ON(transaction->use_count == 0);
 	transaction->use_count--;
+	WARN_ON(transaction->magic != TRANS_MAGIC);
 	if (transaction->use_count == 0) {
 		WARN_ON(total_trans == 0);
 		total_trans--;
-		kfree(transaction);
+		memset(transaction, 0, sizeof(*transaction));
+		kmem_cache_free(btrfs_transaction_cachep, transaction);
 	}
 }
 
@@ -20,7 +27,8 @@ static int join_transaction(struct btrfs_root *root)
 	struct btrfs_transaction *cur_trans;
 	cur_trans = root->fs_info->running_transaction;
 	if (!cur_trans) {
-		cur_trans = kmalloc(sizeof(*cur_trans), GFP_NOFS);
+		cur_trans = kmem_cache_alloc(btrfs_transaction_cachep,
+					     GFP_NOFS);
 		total_trans++;
 		BUG_ON(!cur_trans);
 		root->fs_info->running_transaction = cur_trans;
@@ -28,6 +36,7 @@ static int join_transaction(struct btrfs_root *root)
 		cur_trans->transid = root->root_key.offset + 1;
 		init_waitqueue_head(&cur_trans->writer_wait);
 		init_waitqueue_head(&cur_trans->commit_wait);
+		cur_trans->magic = TRANS_MAGIC;
 		cur_trans->in_commit = 0;
 		cur_trans->use_count = 1;
 		cur_trans->commit_done = 0;
@@ -39,7 +48,8 @@ static int join_transaction(struct btrfs_root *root)
 struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
 						   int num_blocks)
 {
-	struct btrfs_trans_handle *h = kmalloc(sizeof(*h), GFP_NOFS);
+	struct btrfs_trans_handle *h =
+		kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS);
 	int ret;
 
 	mutex_lock(&root->fs_info->trans_mutex);
@@ -51,6 +61,7 @@ struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
 	h->blocks_used = 0;
 	root->fs_info->running_transaction->use_count++;
 	mutex_unlock(&root->fs_info->trans_mutex);
+	h->magic = h->magic2 = TRANS_MAGIC;
 	return h;
 }
 
@@ -58,6 +69,8 @@ int btrfs_end_transaction(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *root)
 {
 	struct btrfs_transaction *cur_trans;
+	WARN_ON(trans->magic != TRANS_MAGIC);
+	WARN_ON(trans->magic2 != TRANS_MAGIC);
 	mutex_lock(&root->fs_info->trans_mutex);
 	cur_trans = root->fs_info->running_transaction;
 	WARN_ON(cur_trans->num_writers < 1);
@@ -67,7 +80,7 @@ int btrfs_end_transaction(struct btrfs_trans_handle *trans,
 	put_transaction(cur_trans);
 	mutex_unlock(&root->fs_info->trans_mutex);
 	memset(trans, 0, sizeof(*trans));
-	kfree(trans);
+	kmem_cache_free(btrfs_trans_handle_cachep, trans);
 	return 0;
 }
 
@@ -75,7 +88,7 @@ int btrfs_end_transaction(struct btrfs_trans_handle *trans,
 int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
 				     struct btrfs_root *root)
 {
-	filemap_write_and_wait(root->fs_info->btree_inode->i_mapping);
+	filemap_write_and_wait(root->fs_info->sb->s_bdev->bd_inode->i_mapping);
 	return 0;
 }
 
@@ -137,6 +150,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 
 	mutex_lock(&root->fs_info->trans_mutex);
 	if (trans->transaction->in_commit) {
+printk("already in commit!, waiting\n");
 		cur_trans = trans->transaction;
 		trans->transaction->use_count++;
 		btrfs_end_transaction(trans, root);
@@ -146,7 +160,10 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 		mutex_unlock(&root->fs_info->trans_mutex);
 		return 0;
 	}
+	cur_trans = trans->transaction;
+	trans->transaction->in_commit = 1;
 	while (trans->transaction->num_writers > 1) {
+		WARN_ON(cur_trans != trans->transaction);
 		prepare_to_wait(&trans->transaction->writer_wait, &wait,
 				TASK_UNINTERRUPTIBLE);
 		if (trans->transaction->num_writers <= 1)
@@ -154,15 +171,15 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 		mutex_unlock(&root->fs_info->trans_mutex);
 		schedule();
 		mutex_lock(&root->fs_info->trans_mutex);
+		finish_wait(&trans->transaction->writer_wait, &wait);
 	}
 	finish_wait(&trans->transaction->writer_wait, &wait);
-
+	WARN_ON(cur_trans != trans->transaction);
 	if (root->node != root->commit_root) {
 		memcpy(&snap_key, &root->root_key, sizeof(snap_key));
 		root->root_key.offset++;
 	}
 
-
 	if (btrfs_root_blocknr(&root->root_item) != root->node->b_blocknr) {
 		btrfs_set_root_blocknr(&root->root_item, root->node->b_blocknr);
 		ret = btrfs_insert_root(trans, root->fs_info->tree_root,
@@ -172,22 +189,21 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 
 	ret = btrfs_commit_tree_roots(trans, root);
 	BUG_ON(ret);
-
 	cur_trans = root->fs_info->running_transaction;
 	root->fs_info->running_transaction = NULL;
 	mutex_unlock(&root->fs_info->trans_mutex);
-
 	ret = btrfs_write_and_wait_transaction(trans, root);
 	BUG_ON(ret);
 
 	write_ctree_super(trans, root);
 	btrfs_finish_extent_commit(trans, root);
 	mutex_lock(&root->fs_info->trans_mutex);
+	cur_trans->commit_done = 1;
+	wake_up(&cur_trans->commit_wait);
 	put_transaction(cur_trans);
 	put_transaction(cur_trans);
 	mutex_unlock(&root->fs_info->trans_mutex);
-	kfree(trans);
-
+	kmem_cache_free(btrfs_trans_handle_cachep, trans);
 	if (root->node != root->commit_root) {
 		trans = btrfs_start_transaction(root, 1);
 		snap = root->commit_root;
@@ -203,7 +219,6 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 		ret = btrfs_end_transaction(trans, root);
 		BUG_ON(ret);
 	}
-
 	return ret;
 }
 
diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h
index 7e6c08a0accd..4f1496ae6f24 100644
--- a/fs/btrfs/transaction.h
+++ b/fs/btrfs/transaction.h
@@ -7,15 +7,18 @@ struct btrfs_transaction {
 	int in_commit;
 	int use_count;
 	int commit_done;
+	int magic;
 	wait_queue_head_t writer_wait;
 	wait_queue_head_t commit_wait;
 };
 
 struct btrfs_trans_handle {
+	int magic;
 	u64 transid;
 	unsigned long blocks_reserved;
 	unsigned long blocks_used;
 	struct btrfs_transaction *transaction;
+	int magic2;
 };
 
 
-- 
cgit v1.2.3


From 0f7d52f4431c530b4f39c524448c688bb7754de5 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Mon, 9 Apr 2007 10:42:37 -0400
Subject: Btrfs: groundwork for subvolume and snapshot roots

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/bit-radix.c   |   2 +-
 fs/btrfs/ctree.h       |  20 +++++-
 fs/btrfs/disk-io.c     | 178 +++++++++++++++++++++++++++++++++++++------------
 fs/btrfs/disk-io.h     |   2 +
 fs/btrfs/super.c       |  39 ++++++-----
 fs/btrfs/transaction.c | 140 ++++++++++++++++++++++++++++----------
 6 files changed, 282 insertions(+), 99 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/bit-radix.c b/fs/btrfs/bit-radix.c
index 783f54c2a16c..9fc42e99c7df 100644
--- a/fs/btrfs/bit-radix.c
+++ b/fs/btrfs/bit-radix.c
@@ -86,7 +86,7 @@ int find_first_radix_bit(struct radix_tree_root *radix, unsigned long *retbits,
 	int i;
 	int total_found = 0;
 
-	ret = radix_tree_gang_lookup(radix, (void *)&gang, 0, ARRAY_SIZE(gang));
+	ret = radix_tree_gang_lookup(radix, (void **)gang, 0, ARRAY_SIZE(gang));
 	for (i = 0; i < ret && nr > 0; i++) {
 		found = 0;
 		bits = gang[i];
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 1ff5b99af682..5460030c9e6a 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -232,12 +232,12 @@ struct btrfs_inode_map_item {
 
 struct crypto_hash;
 struct btrfs_fs_info {
-	struct btrfs_root *fs_root;
 	struct btrfs_root *extent_root;
 	struct btrfs_root *tree_root;
 	struct btrfs_root *inode_root;
 	struct btrfs_key current_insert;
 	struct btrfs_key last_insert;
+	struct radix_tree_root fs_roots_radix;
 	struct radix_tree_root pending_del_radix;
 	struct radix_tree_root pinned_radix;
 	u64 last_inode_alloc;
@@ -266,6 +266,9 @@ struct btrfs_root {
 	struct btrfs_root_item root_item;
 	struct btrfs_key root_key;
 	struct btrfs_fs_info *fs_info;
+	struct inode *inode;
+	u64 objectid;
+	u64 last_trans;
 	u32 blocksize;
 	int ref_cows;
 	u32 type;
@@ -595,7 +598,7 @@ static inline u32 btrfs_key_overflow(struct btrfs_key *key)
 
 static inline void btrfs_set_key_overflow(struct btrfs_key *key, u32 over)
 {
-	BUG_ON(over > BTRFS_KEY_OVERFLOW_MAX);
+	BUG_ON(over >= BTRFS_KEY_OVERFLOW_MAX);
 	over = over << BTRFS_KEY_OVERFLOW_SHIFT;
 	key->flags = (key->flags & ~((u64)BTRFS_KEY_OVERFLOW_MASK)) | over;
 }
@@ -634,7 +637,7 @@ static inline void btrfs_set_disK_key_overflow(struct btrfs_disk_key *key,
 					       u32 over)
 {
 	u32 flags = btrfs_disk_key_flags(key);
-	BUG_ON(over > BTRFS_KEY_OVERFLOW_MAX);
+	BUG_ON(over >= BTRFS_KEY_OVERFLOW_MAX);
 	over = over << BTRFS_KEY_OVERFLOW_SHIFT;
 	flags = (flags & ~((u64)BTRFS_KEY_OVERFLOW_MASK)) | over;
 	btrfs_set_disk_key_flags(key, flags);
@@ -748,6 +751,17 @@ static inline void btrfs_set_super_blocknr(struct btrfs_super_block *s, u64 val)
 	s->blocknr = cpu_to_le64(val);
 }
 
+static inline u64 btrfs_super_generation(struct btrfs_super_block *s)
+{
+	return le64_to_cpu(s->generation);
+}
+
+static inline void btrfs_set_super_generation(struct btrfs_super_block *s,
+					      u64 val)
+{
+	s->generation = cpu_to_le64(val);
+}
+
 static inline u64 btrfs_super_root(struct btrfs_super_block *s)
 {
 	return le64_to_cpu(s->root);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index b9301a5e4608..b557bdd1e26a 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -4,9 +4,11 @@
 #include <linux/crypto.h>
 #include <linux/scatterlist.h>
 #include <linux/swap.h>
+#include <linux/radix-tree.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
+#include "btrfs_inode.h"
 
 static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
 {
@@ -180,7 +182,7 @@ static int csum_tree_block(struct btrfs_root *root, struct buffer_head *bh,
 static int btree_writepage(struct page *page, struct writeback_control *wbc)
 {
 	struct buffer_head *bh;
-	struct btrfs_root *root = btrfs_sb(page->mapping->host->i_sb);
+	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
 	struct buffer_head *head;
 	if (!page_has_buffers(page)) {
 		create_empty_buffers(page, root->fs_info->sb->s_blocksize,
@@ -259,10 +261,13 @@ static int __setup_root(int blocksize,
 			u64 objectid)
 {
 	root->node = NULL;
+	root->inode = NULL;
 	root->commit_root = NULL;
 	root->blocksize = blocksize;
 	root->ref_cows = 0;
 	root->fs_info = fs_info;
+	root->objectid = objectid;
+	root->last_trans = 0;
 	memset(&root->root_key, 0, sizeof(root->root_key));
 	memset(&root->root_item, 0, sizeof(root->root_item));
 	return 0;
@@ -287,10 +292,78 @@ static int find_and_setup_root(int blocksize,
 	return 0;
 }
 
+struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
+				      struct btrfs_key *location)
+{
+	struct btrfs_root *root;
+	struct btrfs_root *tree_root = fs_info->tree_root;
+	struct btrfs_path *path;
+	struct btrfs_leaf *l;
+	int ret = 0;
+
+printk("read_fs_root looking for %Lu %Lu %u\n", location->objectid, location->offset, location->flags);
+	root = kmalloc(sizeof(*root), GFP_NOFS);
+	if (!root) {
+printk("failed1\n");
+		return ERR_PTR(-ENOMEM);
+	}
+	if (location->offset == (u64)-1) {
+		ret = find_and_setup_root(fs_info->sb->s_blocksize,
+					  fs_info->tree_root, fs_info,
+					  location->objectid, root);
+		if (ret) {
+printk("failed2\n");
+			kfree(root);
+			return ERR_PTR(ret);
+		}
+		goto insert;
+	}
+
+	__setup_root(fs_info->sb->s_blocksize, root, fs_info,
+		     location->objectid);
+
+	path = btrfs_alloc_path();
+	BUG_ON(!path);
+	ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
+	if (ret != 0) {
+printk("internal search_slot gives us %d\n", ret);
+		if (ret > 0)
+			ret = -ENOENT;
+		goto out;
+	}
+	l = btrfs_buffer_leaf(path->nodes[0]);
+	memcpy(&root->root_item,
+	       btrfs_item_ptr(l, path->slots[0], struct btrfs_root_item),
+	       sizeof(root->root_item));
+	memcpy(&root->root_key, location, sizeof(*location));
+	ret = 0;
+out:
+	btrfs_release_path(root, path);
+	btrfs_free_path(path);
+	if (ret) {
+		kfree(root);
+		return ERR_PTR(ret);
+	}
+	root->node = read_tree_block(root,
+				     btrfs_root_blocknr(&root->root_item));
+	BUG_ON(!root->node);
+insert:
+printk("inserting %p\n", root);
+	root->ref_cows = 1;
+	ret = radix_tree_insert(&fs_info->fs_roots_radix, (unsigned long)root,
+				root);
+	if (ret) {
+printk("radix_tree_insert gives us %d\n", ret);
+		brelse(root->node);
+		kfree(root);
+		return ERR_PTR(ret);
+	}
+printk("all worked\n");
+	return root;
+}
+
 struct btrfs_root *open_ctree(struct super_block *sb)
 {
-	struct btrfs_root *root = kmalloc(sizeof(struct btrfs_root),
-					  GFP_NOFS);
 	struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
 						 GFP_NOFS);
 	struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
@@ -304,9 +377,9 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 
 	init_bit_radix(&fs_info->pinned_radix);
 	init_bit_radix(&fs_info->pending_del_radix);
+	INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
 	sb_set_blocksize(sb, 4096);
 	fs_info->running_transaction = NULL;
-	fs_info->fs_root = root;
 	fs_info->tree_root = tree_root;
 	fs_info->extent_root = extent_root;
 	fs_info->inode_root = inode_root;
@@ -318,6 +391,9 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	fs_info->btree_inode->i_nlink = 1;
 	fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
 	fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
+	BTRFS_I(fs_info->btree_inode)->root = tree_root;
+	memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
+	       sizeof(struct btrfs_key));
 	insert_inode_hash(fs_info->btree_inode);
 	mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
 	fs_info->hash_tfm = crypto_alloc_hash("sha256", 0, CRYPTO_ALG_ASYNC);
@@ -337,13 +413,12 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 					     BTRFS_SUPER_INFO_OFFSET /
 					     sb->s_blocksize);
 
-	if (!fs_info->sb_buffer) {
+	if (!fs_info->sb_buffer)
 		return NULL;
-	}
 	disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
-	if (!btrfs_super_root(disk_super)) {
+	if (!btrfs_super_root(disk_super))
 		return NULL;
-	}
+
 	fs_info->disk_super = disk_super;
 	tree_root->node = read_tree_block(tree_root,
 					  btrfs_super_root(disk_super));
@@ -358,14 +433,8 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 				  BTRFS_INODE_MAP_OBJECTID, inode_root);
 	BUG_ON(ret);
 
-	ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
-				  BTRFS_FS_TREE_OBJECTID, root);
-	BUG_ON(ret);
-	root->commit_root = root->node;
-	get_bh(root->node);
-	root->ref_cows = 1;
-	root->fs_info->generation = root->root_key.offset + 1;
-	ret = btrfs_find_highest_inode(root, &root->fs_info->last_inode_alloc);
+	fs_info->generation = btrfs_super_generation(disk_super) + 1;
+	ret = btrfs_find_highest_inode(tree_root, &fs_info->last_inode_alloc);
 	if (ret == 0)
 		fs_info->highest_inode = fs_info->last_inode_alloc;
 	memset(&fs_info->kobj, 0, sizeof(fs_info->kobj));
@@ -373,7 +442,7 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	kobject_set_name(&fs_info->kobj, "%s", sb->s_id);
 	kobject_register(&fs_info->kobj);
 	mutex_unlock(&fs_info->fs_mutex);
-	return root;
+	return tree_root;
 }
 
 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
@@ -398,12 +467,42 @@ int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
 	return 0;
 }
 
+int del_fs_roots(struct btrfs_fs_info *fs_info)
+{
+	int ret;
+	struct btrfs_root *gang[8];
+	int i;
+
+	while(1) {
+		ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
+					     (void **)gang, 0,
+					     ARRAY_SIZE(gang));
+		if (!ret)
+			break;
+		for (i = 0; i < ret; i++) {
+			radix_tree_delete(&fs_info->fs_roots_radix,
+					  (unsigned long)gang[i]);
+			if (gang[i]->inode)
+				iput(gang[i]->inode);
+			else
+				printk("no inode for root %p\n", gang[i]);
+			if (gang[i]->node)
+				brelse(gang[i]->node);
+			if (gang[i]->commit_root)
+				brelse(gang[i]->commit_root);
+			kfree(gang[i]);
+		}
+	}
+	return 0;
+}
+
 int close_ctree(struct btrfs_root *root)
 {
 	int ret;
 	struct btrfs_trans_handle *trans;
+	struct btrfs_fs_info *fs_info = root->fs_info;
 
-	mutex_lock(&root->fs_info->fs_mutex);
+	mutex_lock(&fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
 	btrfs_commit_transaction(trans, root);
 	/* run commit again to  drop the original snapshot */
@@ -412,29 +511,26 @@ int close_ctree(struct btrfs_root *root)
 	ret = btrfs_write_and_wait_transaction(NULL, root);
 	BUG_ON(ret);
 	write_ctree_super(NULL, root);
-	mutex_unlock(&root->fs_info->fs_mutex);
-
-	if (root->node)
-		btrfs_block_release(root, root->node);
-	if (root->fs_info->extent_root->node)
-		btrfs_block_release(root->fs_info->extent_root,
-				    root->fs_info->extent_root->node);
-	if (root->fs_info->inode_root->node)
-		btrfs_block_release(root->fs_info->inode_root,
-				    root->fs_info->inode_root->node);
-	if (root->fs_info->tree_root->node)
-		btrfs_block_release(root->fs_info->tree_root,
-				    root->fs_info->tree_root->node);
-	btrfs_block_release(root, root->commit_root);
-	btrfs_block_release(root, root->fs_info->sb_buffer);
-	crypto_free_hash(root->fs_info->hash_tfm);
-	truncate_inode_pages(root->fs_info->btree_inode->i_mapping, 0);
-	iput(root->fs_info->btree_inode);
-	kfree(root->fs_info->extent_root);
-	kfree(root->fs_info->inode_root);
-	kfree(root->fs_info->tree_root);
-	kobject_unregister(&root->fs_info->kobj);
-	kfree(root);
+	mutex_unlock(&fs_info->fs_mutex);
+
+	if (fs_info->extent_root->node)
+		btrfs_block_release(fs_info->extent_root,
+				    fs_info->extent_root->node);
+	if (fs_info->inode_root->node)
+		btrfs_block_release(fs_info->inode_root,
+				    fs_info->inode_root->node);
+	if (fs_info->tree_root->node)
+		btrfs_block_release(fs_info->tree_root,
+				    fs_info->tree_root->node);
+	btrfs_block_release(root, fs_info->sb_buffer);
+	crypto_free_hash(fs_info->hash_tfm);
+	truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
+	iput(fs_info->btree_inode);
+	del_fs_roots(fs_info);
+	kfree(fs_info->extent_root);
+	kfree(fs_info->inode_root);
+	kfree(fs_info->tree_root);
+	kobject_unregister(&fs_info->kobj);
 	return 0;
 }
 
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index ac6764ba8aa6..0ef6e6f714af 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -39,4 +39,6 @@ int write_ctree_super(struct btrfs_trans_handle *trans,
 struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr);
 int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
 		    char *result);
+struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
+				      struct btrfs_key *location);
 #endif
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 66d9fb2288c3..3c9236ca889c 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -436,7 +436,6 @@ int fixup_tree_root_location(struct btrfs_root *root,
 {
 	struct btrfs_path *path;
 	struct btrfs_root_item *ri;
-	int ret;
 
 	if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
 		return 0;
@@ -447,22 +446,19 @@ int fixup_tree_root_location(struct btrfs_root *root,
 	BUG_ON(!path);
 	mutex_lock(&root->fs_info->fs_mutex);
 
-	ret = btrfs_lookup_inode(NULL, root, path, location, 0);
-	if (ret)
-		goto out;
-	ri = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
-			  path->slots[0],
-			  struct btrfs_root_item);
+	*sub_root = btrfs_read_fs_root(root->fs_info, location);
+	if (IS_ERR(*sub_root))
+		return PTR_ERR(*sub_root);
+
+	ri = &(*sub_root)->root_item;
 	location->objectid = btrfs_root_dirid(ri);
 	location->flags = 0;
 	btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
 	location->offset = 0;
-	/* FIXME properly select the root */
-	*sub_root = root->fs_info->fs_root;
-out:
+
 	btrfs_free_path(path);
 	mutex_unlock(&root->fs_info->fs_mutex);
-	return ret;
+	return 0;
 }
 
 
@@ -494,6 +490,15 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
 		if (!inode)
 			return ERR_PTR(-EACCES);
 		if (inode->i_state & I_NEW) {
+			if (sub_root != root) {
+				ret = radix_tree_insert(
+						&root->fs_info->fs_roots_radix,
+						(unsigned long)sub_root,
+						sub_root);
+printk("adding new root for inode %lu\n", inode->i_ino);
+				igrab(inode);
+				sub_root->inode = inode;
+			}
 			BTRFS_I(inode)->root = sub_root;
 			memcpy(&BTRFS_I(inode)->location, &location,
 			       sizeof(location));
@@ -605,7 +610,7 @@ static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
 	struct inode * inode;
 	struct dentry * root_dentry;
 	struct btrfs_super_block *disk_super;
-	struct btrfs_root *root;
+	struct btrfs_root *tree_root;
 	struct btrfs_inode *bi;
 
 	sb->s_maxbytes = MAX_LFS_FILESIZE;
@@ -613,14 +618,14 @@ static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
 	sb->s_op = &btrfs_super_ops;
 	sb->s_time_gran = 1;
 
-	root = open_ctree(sb);
+	tree_root = open_ctree(sb);
 
-	if (!root) {
+	if (!tree_root) {
 		printk("btrfs: open_ctree failed\n");
 		return -EIO;
 	}
-	sb->s_fs_info = root;
-	disk_super = root->fs_info->disk_super;
+	sb->s_fs_info = tree_root;
+	disk_super = tree_root->fs_info->disk_super;
 	printk("read in super total blocks %Lu root %Lu\n",
 	       btrfs_super_total_blocks(disk_super),
 	       btrfs_super_root_dir(disk_super));
@@ -630,7 +635,7 @@ static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
 	bi->location.objectid = inode->i_ino;
 	bi->location.offset = 0;
 	bi->location.flags = 0;
-	bi->root = root->fs_info->tree_root;
+	bi->root = tree_root;
 	btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY);
 
 	if (!inode)
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 20d84bd03223..83a0194ab163 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -8,6 +8,8 @@ static int total_trans = 0;
 extern struct kmem_cache *btrfs_trans_handle_cachep;
 extern struct kmem_cache *btrfs_transaction_cachep;
 
+#define BTRFS_ROOT_TRANS_TAG 0
+
 #define TRANS_MAGIC 0xE1E10E
 static void put_transaction(struct btrfs_transaction *transaction)
 {
@@ -31,9 +33,10 @@ static int join_transaction(struct btrfs_root *root)
 					     GFP_NOFS);
 		total_trans++;
 		BUG_ON(!cur_trans);
+		root->fs_info->generation++;
 		root->fs_info->running_transaction = cur_trans;
 		cur_trans->num_writers = 0;
-		cur_trans->transid = root->root_key.offset + 1;
+		cur_trans->transid = root->fs_info->generation;
 		init_waitqueue_head(&cur_trans->writer_wait);
 		init_waitqueue_head(&cur_trans->commit_wait);
 		cur_trans->magic = TRANS_MAGIC;
@@ -51,13 +54,22 @@ struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
 	struct btrfs_trans_handle *h =
 		kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS);
 	int ret;
+	u64 running_trans_id;
 
-	/* FIXME, use the right root */
-	root = root->fs_info->fs_root;
 	mutex_lock(&root->fs_info->trans_mutex);
 	ret = join_transaction(root);
 	BUG_ON(ret);
-	h->transid = root->fs_info->running_transaction->transid;
+	running_trans_id = root->fs_info->running_transaction->transid;
+
+	if (root != root->fs_info->tree_root && root->last_trans <
+	    running_trans_id) {
+		radix_tree_tag_set(&root->fs_info->fs_roots_radix,
+				   (unsigned long)root, BTRFS_ROOT_TRANS_TAG);
+		root->commit_root = root->node;
+		get_bh(root->node);
+	}
+	root->last_trans = running_trans_id;
+	h->transid = running_trans_id;
 	h->transaction = root->fs_info->running_transaction;
 	h->blocks_reserved = num_blocks;
 	h->blocks_used = 0;
@@ -72,9 +84,6 @@ int btrfs_end_transaction(struct btrfs_trans_handle *trans,
 {
 	struct btrfs_transaction *cur_trans;
 
-	/* FIXME, use the right root */
-	root = root->fs_info->fs_root;
-
 	WARN_ON(trans->magic != TRANS_MAGIC);
 	WARN_ON(trans->magic2 != TRANS_MAGIC);
 	mutex_lock(&root->fs_info->trans_mutex);
@@ -145,17 +154,96 @@ static int wait_for_commit(struct btrfs_root *root,
 	return 0;
 }
 
+struct dirty_root {
+	struct list_head list;
+	struct btrfs_key snap_key;
+	struct buffer_head *commit_root;
+	struct btrfs_root *root;
+};
+
+int add_dirty_roots(struct btrfs_trans_handle *trans,
+		    struct radix_tree_root *radix, struct list_head *list)
+{
+	struct dirty_root *dirty;
+	struct btrfs_root *gang[8];
+	struct btrfs_root *root;
+	int i;
+	int ret;
+	int err;
+printk("add dirty\n");
+	while(1) {
+		ret = radix_tree_gang_lookup_tag(radix, (void **)gang, 0,
+						 ARRAY_SIZE(gang),
+						 BTRFS_ROOT_TRANS_TAG);
+		if (ret == 0)
+			break;
+		for (i = 0; i < ret; i++) {
+			root = gang[i];
+			radix_tree_tag_clear(radix, (unsigned long)root,
+					     BTRFS_ROOT_TRANS_TAG);
+			if (root->commit_root == root->node) {
+				WARN_ON(root->node->b_blocknr !=
+					btrfs_root_blocknr(&root->root_item));
+				brelse(root->commit_root);
+				root->commit_root = NULL;
+				continue;
+			}
+			dirty = kmalloc(sizeof(*dirty), GFP_NOFS);
+			BUG_ON(!dirty);
+			memcpy(&dirty->snap_key, &root->root_key,
+			       sizeof(root->root_key));
+			dirty->commit_root = root->commit_root;
+			root->commit_root = NULL;
+			dirty->root = root;
+printk("adding dirty root %Lu gen %Lu blocknr %Lu\n", root->root_key.objectid, root->root_key.offset, dirty->commit_root->b_blocknr);
+			root->root_key.offset = root->fs_info->generation;
+			btrfs_set_root_blocknr(&root->root_item,
+					       root->node->b_blocknr);
+			err = btrfs_insert_root(trans, root->fs_info->tree_root,
+						&root->root_key,
+						&root->root_item);
+			BUG_ON(err);
+			list_add(&dirty->list, list);
+		}
+	}
+printk("add dirty done\n");
+	return 0;
+}
+
+int drop_dirty_roots(struct btrfs_root *tree_root, struct list_head *list)
+{
+	struct dirty_root *dirty;
+	struct btrfs_trans_handle *trans;
+	int ret;
+
+	while(!list_empty(list)) {
+		dirty = list_entry(list->next, struct dirty_root, list);
+		list_del_init(&dirty->list);
+		trans = btrfs_start_transaction(tree_root, 1);
+printk("drop snapshot root %p, commit_root blocknr %Lu generation %Lu\n", dirty->root, dirty->commit_root->b_blocknr, dirty->snap_key.offset);
+		ret = btrfs_drop_snapshot(trans, dirty->root,
+					  dirty->commit_root);
+		BUG_ON(ret);
+
+printk("del root objectid %Lu, offset %Lu\n", dirty->snap_key.objectid, dirty->snap_key.offset);
+		ret = btrfs_del_root(trans, tree_root, &dirty->snap_key);
+		BUG_ON(ret);
+		ret = btrfs_end_transaction(trans, tree_root);
+		BUG_ON(ret);
+		kfree(dirty);
+	}
+	return 0;
+}
+
 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root)
 {
 	int ret = 0;
-	struct buffer_head *snap;
-	struct btrfs_key snap_key;
 	struct btrfs_transaction *cur_trans;
+	struct list_head dirty_fs_roots;
 	DEFINE_WAIT(wait);
 
-	/* FIXME, use the right root */
-	root = root->fs_info->fs_root;
+	INIT_LIST_HEAD(&dirty_fs_roots);
 
 	mutex_lock(&root->fs_info->trans_mutex);
 	if (trans->transaction->in_commit) {
@@ -184,22 +272,13 @@ printk("already in commit!, waiting\n");
 	}
 	finish_wait(&trans->transaction->writer_wait, &wait);
 	WARN_ON(cur_trans != trans->transaction);
-	if (root->node != root->commit_root) {
-		memcpy(&snap_key, &root->root_key, sizeof(snap_key));
-		root->root_key.offset++;
-	}
-
-	if (btrfs_root_blocknr(&root->root_item) != root->node->b_blocknr) {
-		btrfs_set_root_blocknr(&root->root_item, root->node->b_blocknr);
-		ret = btrfs_insert_root(trans, root->fs_info->tree_root,
-					&root->root_key, &root->root_item);
-		BUG_ON(ret);
-	}
-
+	add_dirty_roots(trans, &root->fs_info->fs_roots_radix, &dirty_fs_roots);
 	ret = btrfs_commit_tree_roots(trans, root);
 	BUG_ON(ret);
 	cur_trans = root->fs_info->running_transaction;
 	root->fs_info->running_transaction = NULL;
+	btrfs_set_super_generation(root->fs_info->disk_super,
+				   root->fs_info->generation + 1);
 	mutex_unlock(&root->fs_info->trans_mutex);
 	ret = btrfs_write_and_wait_transaction(trans, root);
 	BUG_ON(ret);
@@ -213,21 +292,8 @@ printk("already in commit!, waiting\n");
 	put_transaction(cur_trans);
 	mutex_unlock(&root->fs_info->trans_mutex);
 	kmem_cache_free(btrfs_trans_handle_cachep, trans);
-	if (root->node != root->commit_root) {
-		trans = btrfs_start_transaction(root, 1);
-		snap = root->commit_root;
-		root->commit_root = root->node;
-		get_bh(root->node);
-		ret = btrfs_drop_snapshot(trans, root, snap);
-		BUG_ON(ret);
 
-		ret = btrfs_del_root(trans, root->fs_info->tree_root,
-				     &snap_key);
-		BUG_ON(ret);
-		root->fs_info->generation = root->root_key.offset + 1;
-		ret = btrfs_end_transaction(trans, root);
-		BUG_ON(ret);
-	}
+	drop_dirty_roots(root->fs_info->tree_root, &dirty_fs_roots);
 	return ret;
 }
 
-- 
cgit v1.2.3


From 7eccb903a817e890c947ba4bc90c6a9af9b4219a Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Wed, 11 Apr 2007 15:53:25 -0400
Subject: Btrfs: create a logical->phsyical block number mapping scheme

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.c       |  38 +++++++-------
 fs/btrfs/ctree.h       |   1 +
 fs/btrfs/disk-io.c     | 140 +++++++++++++++++++++++++++++++++++--------------
 fs/btrfs/disk-io.h     |   1 +
 fs/btrfs/extent-tree.c |   8 +--
 fs/btrfs/super.c       |   6 +--
 fs/btrfs/transaction.c |   8 +--
 7 files changed, 134 insertions(+), 68 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index f31ac843bf2e..e2fc6f0d0243 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -66,21 +66,21 @@ static int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root
 	if (buf->b_size != root->blocksize || cow->b_size != root->blocksize)
 		WARN_ON(1);
 	memcpy(cow_node, btrfs_buffer_node(buf), root->blocksize);
-	btrfs_set_header_blocknr(&cow_node->header, cow->b_blocknr);
+	btrfs_set_header_blocknr(&cow_node->header, bh_blocknr(cow));
 	btrfs_set_header_generation(&cow_node->header, trans->transid);
 	btrfs_inc_ref(trans, root, buf);
 	if (buf == root->node) {
 		root->node = cow;
 		get_bh(cow);
 		if (buf != root->commit_root) {
-			btrfs_free_extent(trans, root, buf->b_blocknr, 1, 1);
+			btrfs_free_extent(trans, root, bh_blocknr(buf), 1, 1);
 		}
 		btrfs_block_release(root, buf);
 	} else {
 		btrfs_set_node_blockptr(btrfs_buffer_node(parent), parent_slot,
-					cow->b_blocknr);
+					bh_blocknr(cow));
 		btrfs_mark_buffer_dirty(parent);
-		btrfs_free_extent(trans, root, buf->b_blocknr, 1, 1);
+		btrfs_free_extent(trans, root, bh_blocknr(buf), 1, 1);
 	}
 	btrfs_block_release(root, buf);
 	mark_buffer_dirty(cow);
@@ -312,7 +312,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 	 */
 	if (!parent_buf) {
 		struct buffer_head *child;
-		u64 blocknr = mid_buf->b_blocknr;
+		u64 blocknr = bh_blocknr(mid_buf);
 
 		if (btrfs_header_nritems(&mid->header) != 1)
 			return 0;
@@ -361,7 +361,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 		if (wret < 0)
 			ret = wret;
 		if (btrfs_header_nritems(&right->header) == 0) {
-			u64 blocknr = right_buf->b_blocknr;
+			u64 blocknr = bh_blocknr(right_buf);
 			clean_tree_block(trans, root, right_buf);
 			wait_on_buffer(right_buf);
 			btrfs_block_release(root, right_buf);
@@ -400,7 +400,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 	}
 	if (btrfs_header_nritems(&mid->header) == 0) {
 		/* we've managed to empty the middle node, drop it */
-		u64 blocknr = mid_buf->b_blocknr;
+		u64 blocknr = bh_blocknr(mid_buf);
 		clean_tree_block(trans, root, mid_buf);
 		wait_on_buffer(mid_buf);
 		btrfs_block_release(root, mid_buf);
@@ -686,7 +686,7 @@ static int insert_new_root(struct btrfs_trans_handle *trans, struct btrfs_root
 	memset(c, 0, root->blocksize);
 	btrfs_set_header_nritems(&c->header, 1);
 	btrfs_set_header_level(&c->header, level);
-	btrfs_set_header_blocknr(&c->header, t->b_blocknr);
+	btrfs_set_header_blocknr(&c->header, bh_blocknr(t));
 	btrfs_set_header_generation(&c->header, trans->transid);
 	lower = btrfs_buffer_node(path->nodes[level-1]);
 	memcpy(c->header.fsid, root->fs_info->disk_super->fsid,
@@ -697,7 +697,7 @@ static int insert_new_root(struct btrfs_trans_handle *trans, struct btrfs_root
 		lower_key = &lower->ptrs[0].key;
 	btrfs_memcpy(root, c, &c->ptrs[0].key, lower_key,
 		     sizeof(struct btrfs_disk_key));
-	btrfs_set_node_blockptr(c, 0, path->nodes[level - 1]->b_blocknr);
+	btrfs_set_node_blockptr(c, 0, bh_blocknr(path->nodes[level - 1]));
 
 	btrfs_mark_buffer_dirty(t);
 
@@ -780,7 +780,7 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 	split = btrfs_buffer_node(split_buffer);
 	btrfs_set_header_flags(&split->header, btrfs_header_flags(&c->header));
 	btrfs_set_header_level(&split->header, btrfs_header_level(&c->header));
-	btrfs_set_header_blocknr(&split->header, split_buffer->b_blocknr);
+	btrfs_set_header_blocknr(&split->header, bh_blocknr(split_buffer));
 	btrfs_set_header_generation(&split->header, trans->transid);
 	memcpy(split->header.fsid, root->fs_info->disk_super->fsid,
 	       sizeof(split->header.fsid));
@@ -794,7 +794,7 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_mark_buffer_dirty(t);
 	btrfs_mark_buffer_dirty(split_buffer);
 	wret = insert_ptr(trans, root, path, &split->ptrs[0].key,
-			  split_buffer->b_blocknr, path->slots[level + 1] + 1,
+			  bh_blocknr(split_buffer), path->slots[level + 1] + 1,
 			  level + 1);
 	if (wret)
 		ret = wret;
@@ -1138,7 +1138,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 	BUG_ON(!right_buffer);
 	right = btrfs_buffer_leaf(right_buffer);
 	memset(&right->header, 0, sizeof(right->header));
-	btrfs_set_header_blocknr(&right->header, right_buffer->b_blocknr);
+	btrfs_set_header_blocknr(&right->header, bh_blocknr(right_buffer));
 	btrfs_set_header_generation(&right->header, trans->transid);
 	btrfs_set_header_level(&right->header, 0);
 	memcpy(right->header.fsid, root->fs_info->disk_super->fsid,
@@ -1152,7 +1152,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 				btrfs_set_header_nritems(&right->header, 0);
 				wret = insert_ptr(trans, root, path,
 						  &disk_key,
-						  right_buffer->b_blocknr,
+						  bh_blocknr(right_buffer),
 						  path->slots[1] + 1, 1);
 				if (wret)
 					ret = wret;
@@ -1173,7 +1173,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 				btrfs_set_header_nritems(&right->header, 0);
 				wret = insert_ptr(trans, root, path,
 						  &disk_key,
-						  right_buffer->b_blocknr,
+						  bh_blocknr(right_buffer),
 						  path->slots[1] - 1, 1);
 				if (wret)
 					ret = wret;
@@ -1207,7 +1207,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_set_header_nritems(&l->header, mid);
 	ret = 0;
 	wret = insert_ptr(trans, root, path, &right->items[0].key,
-			  right_buffer->b_blocknr, path->slots[1] + 1, 1);
+			  bh_blocknr(right_buffer), path->slots[1] + 1, 1);
 	if (wret)
 		ret = wret;
 	btrfs_mark_buffer_dirty(right_buffer);
@@ -1228,7 +1228,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 	BUG_ON(!right_buffer);
 	right = btrfs_buffer_leaf(right_buffer);
 	memset(&right->header, 0, sizeof(right->header));
-	btrfs_set_header_blocknr(&right->header, right_buffer->b_blocknr);
+	btrfs_set_header_blocknr(&right->header, bh_blocknr(right_buffer));
 	btrfs_set_header_generation(&right->header, trans->transid);
 	btrfs_set_header_level(&right->header, 0);
 	memcpy(right->header.fsid, root->fs_info->disk_super->fsid,
@@ -1237,7 +1237,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_set_header_nritems(&right->header, 0);
 	wret = insert_ptr(trans, root, path,
 			  &disk_key,
-			  right_buffer->b_blocknr,
+			  bh_blocknr(right_buffer),
 			  path->slots[1], 1);
 	if (wret)
 		ret = wret;
@@ -1456,7 +1456,7 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 			if (wret)
 				ret = wret;
 			wret = btrfs_free_extent(trans, root,
-						 leaf_buf->b_blocknr, 1, 1);
+						 bh_blocknr(leaf_buf), 1, 1);
 			if (wret)
 				ret = wret;
 		}
@@ -1487,7 +1487,7 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 					ret = wret;
 			}
 			if (btrfs_header_nritems(&leaf->header) == 0) {
-				u64 blocknr = leaf_buf->b_blocknr;
+				u64 blocknr = bh_blocknr(leaf_buf);
 				clean_tree_block(trans, root, leaf_buf);
 				wait_on_buffer(leaf_buf);
 				wret = del_ptr(trans, root, path, 1, slot);
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 01310de2bf4a..454eb88611bb 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -242,6 +242,7 @@ struct btrfs_fs_info {
 	struct radix_tree_root fs_roots_radix;
 	struct radix_tree_root pending_del_radix;
 	struct radix_tree_root pinned_radix;
+	struct radix_tree_root dev_radix;
 	u64 generation;
 	struct btrfs_transaction *running_transaction;
 	struct btrfs_super_block *disk_super;
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index e09233262af6..c872a7e67abd 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -10,10 +10,30 @@
 #include "transaction.h"
 #include "btrfs_inode.h"
 
+struct dev_lookup {
+	u64 block_start;
+	u64 num_blocks;
+	struct block_device *bdev;
+};
+
+u64 bh_blocknr(struct buffer_head *bh)
+{
+	int blkbits = bh->b_page->mapping->host->i_blkbits;
+	u64 blocknr = bh->b_page->index << (PAGE_CACHE_SHIFT - blkbits);
+	unsigned long offset;
+
+	if (PageHighMem(bh->b_page))
+		offset = (unsigned long)bh->b_data;
+	else
+		offset = bh->b_data - (char *)page_address(bh->b_page);
+	blocknr += offset >> (PAGE_CACHE_SHIFT - blkbits);
+	return blocknr;
+}
+
 static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
 {
 	struct btrfs_node *node = btrfs_buffer_node(buf);
-	if (buf->b_blocknr != btrfs_header_blocknr(&node->header)) {
+	if (bh_blocknr(buf) != btrfs_header_blocknr(&node->header)) {
 		BUG();
 	}
 	return 0;
@@ -40,7 +60,7 @@ struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr)
 	head = page_buffers(page);
 	bh = head;
 	do {
-		if (buffer_mapped(bh) && bh->b_blocknr == blocknr) {
+		if (buffer_mapped(bh) && bh_blocknr(bh) == blocknr) {
 			ret = bh;
 			get_bh(bh);
 			goto out_unlock;
@@ -56,6 +76,33 @@ out_unlock:
 	return ret;
 }
 
+static int map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
+			     u64 logical)
+{
+	struct dev_lookup *lookup[2];
+	char b[BDEVNAME_SIZE];
+
+	int ret;
+
+	root = root->fs_info->dev_root;
+	ret = radix_tree_gang_lookup(&root->fs_info->dev_radix,
+				     (void **)lookup,
+				     (unsigned long)logical,
+				     ARRAY_SIZE(lookup));
+	if (ret == 0 || lookup[0]->block_start > logical ||
+	    lookup[0]->block_start + lookup[0]->num_blocks <= logical) {
+		ret = -ENOENT;
+		goto out;
+	}
+	bh->b_bdev = lookup[0]->bdev;
+	bh->b_blocknr = logical - lookup[0]->block_start;
+printk("logical mapping %Lu to %lu bdev  %s\n", logical, bh->b_blocknr, bdevname(bh->b_bdev, b));
+	set_buffer_mapped(bh);
+	ret = 0;
+out:
+	return ret;
+}
+
 struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
 						 u64 blocknr)
 {
@@ -66,6 +113,7 @@ struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
 	struct buffer_head *bh;
 	struct buffer_head *head;
 	struct buffer_head *ret = NULL;
+	int err;
 	u64 first_block = index << (PAGE_CACHE_SHIFT - blockbits);
 
 	page = grab_cache_page(mapping, index);
@@ -78,11 +126,10 @@ struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
 	bh = head;
 	do {
 		if (!buffer_mapped(bh)) {
-			bh->b_bdev = root->fs_info->sb->s_bdev;
-			bh->b_blocknr = first_block;
-			set_buffer_mapped(bh);
+			err = map_bh_to_logical(root, bh, first_block);
+			BUG_ON(err);
 		}
-		if (bh->b_blocknr == blocknr) {
+		if (bh_blocknr(bh) == blocknr) {
 			ret = bh;
 			get_bh(bh);
 			goto out_unlock;
@@ -98,38 +145,13 @@ out_unlock:
 	return ret;
 }
 
-static sector_t max_block(struct block_device *bdev)
-{
-	sector_t retval = ~((sector_t)0);
-	loff_t sz = i_size_read(bdev->bd_inode);
-
-	if (sz) {
-		unsigned int size = block_size(bdev);
-		unsigned int sizebits = blksize_bits(size);
-		retval = (sz >> sizebits);
-	}
-	return retval;
-}
-
 static int btree_get_block(struct inode *inode, sector_t iblock,
 			   struct buffer_head *bh, int create)
 {
-	if (iblock >= max_block(inode->i_sb->s_bdev)) {
-		if (create)
-			return -EIO;
-
-		/*
-		 * for reads, we're just trying to fill a partial page.
-		 * return a hole, they will have to call get_block again
-		 * before they can fill it, and they will get -EIO at that
-		 * time
-		 */
-		return 0;
-	}
-	bh->b_bdev = inode->i_sb->s_bdev;
-	bh->b_blocknr = iblock;
-	set_buffer_mapped(bh);
-	return 0;
+	int err;
+	struct btrfs_root *root = BTRFS_I(bh->b_page->mapping->host)->root;
+	err = map_bh_to_logical(root, bh, iblock);
+	return err;
 }
 
 int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
@@ -164,8 +186,8 @@ static int csum_tree_block(struct btrfs_root *root, struct buffer_head *bh,
 		return ret;
 	if (verify) {
 		if (memcmp(bh->b_data, result, BTRFS_CSUM_SIZE)) {
-			printk("checksum verify failed on %lu\n",
-			       bh->b_blocknr);
+			printk("checksum verify failed on %Lu\n",
+			       bh_blocknr(bh));
 			return 1;
 		}
 	} else {
@@ -386,10 +408,12 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 						GFP_NOFS);
 	int ret;
 	struct btrfs_super_block *disk_super;
+	struct dev_lookup *dev_lookup;
 
 	init_bit_radix(&fs_info->pinned_radix);
 	init_bit_radix(&fs_info->pending_del_radix);
 	INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
+	INIT_RADIX_TREE(&fs_info->dev_radix, GFP_NOFS);
 	sb_set_blocksize(sb, 4096);
 	fs_info->running_transaction = NULL;
 	fs_info->tree_root = tree_root;
@@ -422,6 +446,13 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 
 	__setup_root(sb->s_blocksize, tree_root,
 		     fs_info, BTRFS_ROOT_TREE_OBJECTID);
+
+	dev_lookup = kmalloc(sizeof(*dev_lookup), GFP_NOFS);
+	dev_lookup->block_start = 0;
+	dev_lookup->num_blocks = (u32)-2;
+	dev_lookup->bdev = sb->s_bdev;
+	ret = radix_tree_insert(&fs_info->dev_radix, (u32)-2, dev_lookup);
+	BUG_ON(ret);
 	fs_info->sb_buffer = read_tree_block(tree_root,
 					     BTRFS_SUPER_INFO_OFFSET /
 					     sb->s_blocksize);
@@ -432,6 +463,14 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	if (!btrfs_super_root(disk_super))
 		return NULL;
 
+	radix_tree_delete(&fs_info->dev_radix, (u32)-2);
+	dev_lookup->block_start = btrfs_super_device_block_start(disk_super);
+	dev_lookup->num_blocks = btrfs_super_device_num_blocks(disk_super);
+	ret = radix_tree_insert(&fs_info->dev_radix,
+				dev_lookup->block_start +
+				dev_lookup->num_blocks, dev_lookup);
+	BUG_ON(ret);
+
 	fs_info->disk_super = disk_super;
 	dev_root->node = read_tree_block(tree_root,
 					  btrfs_super_device_root(disk_super));
@@ -459,7 +498,7 @@ int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
 	struct buffer_head *bh = root->fs_info->sb_buffer;
 
 	btrfs_set_super_root(root->fs_info->disk_super,
-			     root->fs_info->tree_root->node->b_blocknr);
+			     bh_blocknr(root->fs_info->tree_root->node));
 	lock_buffer(bh);
 	WARN_ON(atomic_read(&bh->b_count) < 1);
 	clear_buffer_dirty(bh);
@@ -506,6 +545,29 @@ int del_fs_roots(struct btrfs_fs_info *fs_info)
 	}
 	return 0;
 }
+static int free_dev_radix(struct btrfs_fs_info *fs_info)
+{
+	struct dev_lookup *lookup[8];
+	struct block_device *super_bdev = fs_info->sb->s_bdev;
+	int ret;
+	int i;
+	while(1) {
+		ret = radix_tree_gang_lookup(&fs_info->dev_radix,
+					     (void **)lookup, 0,
+					     ARRAY_SIZE(lookup));
+		if (!ret)
+			break;
+		for (i = 0; i < ret; i++) {
+			if (lookup[i]->bdev != super_bdev)
+				close_bdev_excl(lookup[i]->bdev);
+			radix_tree_delete(&fs_info->dev_radix,
+					  lookup[i]->block_start +
+					  lookup[i]->num_blocks);
+			kfree(lookup[i]);
+		}
+	}
+	return 0;
+}
 
 int close_ctree(struct btrfs_root *root)
 {
@@ -537,6 +599,8 @@ int close_ctree(struct btrfs_root *root)
 	crypto_free_hash(fs_info->hash_tfm);
 	truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
 	iput(fs_info->btree_inode);
+
+	free_dev_radix(fs_info);
 	del_fs_roots(fs_info);
 	kfree(fs_info->extent_root);
 	kfree(fs_info->tree_root);
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 0ef6e6f714af..46cc5dc9d572 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -41,4 +41,5 @@ int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
 		    char *result);
 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 				      struct btrfs_key *location);
+u64 bh_blocknr(struct buffer_head *bh);
 #endif
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index be9630df506c..d560831c10a7 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -80,7 +80,7 @@ static int lookup_block_ref(struct btrfs_trans_handle *trans, struct btrfs_root
 int btrfs_inc_root_ref(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root)
 {
-	return inc_block_ref(trans, root, root->node->b_blocknr, 1);
+	return inc_block_ref(trans, root, bh_blocknr(root->node), 1);
 }
 
 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
@@ -563,7 +563,7 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	WARN_ON(*level < 0);
 	WARN_ON(*level >= BTRFS_MAX_LEVEL);
-	ret = lookup_block_ref(trans, root, path->nodes[*level]->b_blocknr,
+	ret = lookup_block_ref(trans, root, bh_blocknr(path->nodes[*level]),
 			       1, &refs);
 	BUG_ON(ret);
 	if (refs > 1)
@@ -607,7 +607,7 @@ out:
 	WARN_ON(*level < 0);
 	WARN_ON(*level >= BTRFS_MAX_LEVEL);
 	ret = btrfs_free_extent(trans, root,
-				path->nodes[*level]->b_blocknr, 1, 1);
+				bh_blocknr(path->nodes[*level]), 1, 1);
 	btrfs_block_release(root, path->nodes[*level]);
 	path->nodes[*level] = NULL;
 	*level += 1;
@@ -635,7 +635,7 @@ static int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 			return 0;
 		} else {
 			ret = btrfs_free_extent(trans, root,
-						path->nodes[*level]->b_blocknr,
+						bh_blocknr(path->nodes[*level]),
 						1, 1);
 			BUG_ON(ret);
 			btrfs_block_release(root, path->nodes[*level]);
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 2972aaa9eb7e..4ff0cc1efb13 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -1700,7 +1700,7 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 	leaf = btrfs_buffer_leaf(subvol);
 	btrfs_set_header_nritems(&leaf->header, 0);
 	btrfs_set_header_level(&leaf->header, 0);
-	btrfs_set_header_blocknr(&leaf->header, subvol->b_blocknr);
+	btrfs_set_header_blocknr(&leaf->header, bh_blocknr(subvol));
 	btrfs_set_header_generation(&leaf->header, trans->transid);
 	memcpy(leaf->header.fsid, root->fs_info->disk_super->fsid,
 	       sizeof(leaf->header.fsid));
@@ -1713,7 +1713,7 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 	btrfs_set_inode_nblocks(inode_item, 1);
 	btrfs_set_inode_mode(inode_item, S_IFDIR | 0755);
 
-	btrfs_set_root_blocknr(&root_item, subvol->b_blocknr);
+	btrfs_set_root_blocknr(&root_item, bh_blocknr(subvol));
 	btrfs_set_root_refs(&root_item, 1);
 
 	mark_buffer_dirty(subvol);
@@ -1803,7 +1803,7 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 	key.offset = 1;
 	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
-	btrfs_set_root_blocknr(&new_root_item, root->node->b_blocknr);
+	btrfs_set_root_blocknr(&new_root_item, bh_blocknr(root->node));
 
 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
 				&new_root_item);
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index e15a072407bf..3a15943ea8ed 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -119,10 +119,10 @@ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
 
 	while(1) {
 		old_extent_block = btrfs_root_blocknr(&extent_root->root_item);
-		if (old_extent_block == extent_root->node->b_blocknr)
+		if (old_extent_block == bh_blocknr(extent_root->node))
 			break;
 		btrfs_set_root_blocknr(&extent_root->root_item,
-				       extent_root->node->b_blocknr);
+				       bh_blocknr(extent_root->node));
 		ret = btrfs_update_root(trans, tree_root,
 					&extent_root->root_key,
 					&extent_root->root_item);
@@ -176,7 +176,7 @@ int add_dirty_roots(struct btrfs_trans_handle *trans,
 				     (unsigned long)root->root_key.objectid,
 				     BTRFS_ROOT_TRANS_TAG);
 			if (root->commit_root == root->node) {
-				WARN_ON(root->node->b_blocknr !=
+				WARN_ON(bh_blocknr(root->node) !=
 					btrfs_root_blocknr(&root->root_item));
 				brelse(root->commit_root);
 				root->commit_root = NULL;
@@ -191,7 +191,7 @@ int add_dirty_roots(struct btrfs_trans_handle *trans,
 			dirty->root = root;
 			root->root_key.offset = root->fs_info->generation;
 			btrfs_set_root_blocknr(&root->root_item,
-					       root->node->b_blocknr);
+					       bh_blocknr(root->node));
 			err = btrfs_insert_root(trans, root->fs_info->tree_root,
 						&root->root_key,
 						&root->root_item);
-- 
cgit v1.2.3


From 8352d8a473ac84bf7a1c69690b626946d744ca58 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Thu, 12 Apr 2007 10:43:05 -0400
Subject: Btrfs: add disk ioctl, mostly working

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/disk-io.c     | 128 ++++++++++++++++++++++++++++++++++++++++++++++---
 fs/btrfs/disk-io.h     |   6 +++
 fs/btrfs/ioctl.h       |   2 +
 fs/btrfs/super.c       |  96 +++++++++++++++++++++++++++++++++----
 fs/btrfs/transaction.c |   6 +++
 5 files changed, 221 insertions(+), 17 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index c872a7e67abd..3ba4df20c913 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -16,6 +16,28 @@ struct dev_lookup {
 	struct block_device *bdev;
 };
 
+int btrfs_insert_dev_radix(struct btrfs_root *root,
+			   struct block_device *bdev,
+			   u64 block_start,
+			   u64 num_blocks)
+{
+	struct dev_lookup *lookup;
+	char b[BDEVNAME_SIZE];
+	int ret;
+
+	lookup = kmalloc(sizeof(*lookup), GFP_NOFS);
+	if (!lookup)
+		return -ENOMEM;
+	lookup->block_start = block_start;
+	lookup->num_blocks = num_blocks;
+	lookup->bdev = bdev;
+printk("inserting %s into dev radix %Lu %Lu\n", bdevname(bdev, b), block_start, num_blocks);
+
+	ret = radix_tree_insert(&root->fs_info->dev_radix, block_start +
+				num_blocks - 1, lookup);
+	return ret;
+}
+
 u64 bh_blocknr(struct buffer_head *bh)
 {
 	int blkbits = bh->b_page->mapping->host->i_blkbits;
@@ -34,6 +56,8 @@ static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
 {
 	struct btrfs_node *node = btrfs_buffer_node(buf);
 	if (bh_blocknr(buf) != btrfs_header_blocknr(&node->header)) {
+		printk(KERN_CRIT "bh_blocknr(buf) is %Lu, header is %Lu\n",
+		       bh_blocknr(buf), btrfs_header_blocknr(&node->header));
 		BUG();
 	}
 	return 0;
@@ -76,11 +100,10 @@ out_unlock:
 	return ret;
 }
 
-static int map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
+int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
 			     u64 logical)
 {
 	struct dev_lookup *lookup[2];
-	char b[BDEVNAME_SIZE];
 
 	int ret;
 
@@ -96,7 +119,6 @@ static int map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
 	}
 	bh->b_bdev = lookup[0]->bdev;
 	bh->b_blocknr = logical - lookup[0]->block_start;
-printk("logical mapping %Lu to %lu bdev  %s\n", logical, bh->b_blocknr, bdevname(bh->b_bdev, b));
 	set_buffer_mapped(bh);
 	ret = 0;
 out:
@@ -126,7 +148,7 @@ struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
 	bh = head;
 	do {
 		if (!buffer_mapped(bh)) {
-			err = map_bh_to_logical(root, bh, first_block);
+			err = btrfs_map_bh_to_logical(root, bh, first_block);
 			BUG_ON(err);
 		}
 		if (bh_blocknr(bh) == blocknr) {
@@ -150,7 +172,7 @@ static int btree_get_block(struct inode *inode, sector_t iblock,
 {
 	int err;
 	struct btrfs_root *root = BTRFS_I(bh->b_page->mapping->host)->root;
-	err = map_bh_to_logical(root, bh, iblock);
+	err = btrfs_map_bh_to_logical(root, bh, iblock);
 	return err;
 }
 
@@ -396,6 +418,89 @@ printk("all worked\n");
 	return root;
 }
 
+int btrfs_open_disk(struct btrfs_root *root, u64 block_start, u64 num_blocks,
+		    char *filename, int name_len)
+{
+	char *null_filename;
+	struct block_device *bdev;
+	int ret;
+
+	if (block_start == 0) {
+printk("skipping disk with block_start == 0\n");
+return 0;
+	}
+	null_filename = kmalloc(name_len + 1, GFP_NOFS);
+	if (!null_filename)
+		return -ENOMEM;
+	memcpy(null_filename, filename, name_len);
+	null_filename[name_len] = '\0';
+
+	bdev = open_bdev_excl(null_filename, O_RDWR, root->fs_info->sb);
+	if (IS_ERR(bdev)) {
+		ret = PTR_ERR(bdev);
+		goto out;
+	}
+	set_blocksize(bdev, root->fs_info->sb->s_blocksize);
+	ret = btrfs_insert_dev_radix(root, bdev, block_start, num_blocks);
+	BUG_ON(ret);
+	ret = 0;
+out:
+	kfree(null_filename);
+	return ret;
+}
+
+static int read_device_info(struct btrfs_root *root)
+{
+	struct btrfs_path *path;
+	int ret;
+	struct btrfs_key key;
+	struct btrfs_leaf *leaf;
+	struct btrfs_device_item *dev_item;
+	int nritems;
+	int slot;
+
+	root = root->fs_info->dev_root;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+	key.objectid = 0;
+	key.offset = 0;
+	key.flags = 0;
+	btrfs_set_key_type(&key, BTRFS_DEV_ITEM_KEY);
+
+	mutex_lock(&root->fs_info->fs_mutex);
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	leaf = btrfs_buffer_leaf(path->nodes[0]);
+	nritems = btrfs_header_nritems(&leaf->header);
+	while(1) {
+		slot = path->slots[0];
+		if (slot >= nritems) {
+			ret = btrfs_next_leaf(root, path);
+			if (ret)
+				break;
+			leaf = btrfs_buffer_leaf(path->nodes[0]);
+			nritems = btrfs_header_nritems(&leaf->header);
+			slot = path->slots[0];
+		}
+		btrfs_disk_key_to_cpu(&key, &leaf->items[slot].key);
+		if (btrfs_key_type(&key) != BTRFS_DEV_ITEM_KEY) {
+			path->slots[0]++;
+			continue;
+		}
+		dev_item = btrfs_item_ptr(leaf, slot, struct btrfs_device_item);
+printk("found key %Lu %Lu\n", key.objectid, key.offset);
+		ret = btrfs_open_disk(root, key.objectid, key.offset,
+				      (char *)(dev_item + 1),
+				      btrfs_device_pathlen(dev_item));
+		BUG_ON(ret);
+		path->slots[0]++;
+	}
+	btrfs_free_path(path);
+	mutex_unlock(&root->fs_info->fs_mutex);
+	return 0;
+}
+
 struct btrfs_root *open_ctree(struct super_block *sb)
 {
 	struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
@@ -463,17 +568,26 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	if (!btrfs_super_root(disk_super))
 		return NULL;
 
+	i_size_write(fs_info->btree_inode,
+		     btrfs_super_total_blocks(disk_super) <<
+		     fs_info->btree_inode->i_blkbits);
+
 	radix_tree_delete(&fs_info->dev_radix, (u32)-2);
 	dev_lookup->block_start = btrfs_super_device_block_start(disk_super);
 	dev_lookup->num_blocks = btrfs_super_device_num_blocks(disk_super);
 	ret = radix_tree_insert(&fs_info->dev_radix,
 				dev_lookup->block_start +
-				dev_lookup->num_blocks, dev_lookup);
+				dev_lookup->num_blocks - 1, dev_lookup);
 	BUG_ON(ret);
 
 	fs_info->disk_super = disk_super;
+
 	dev_root->node = read_tree_block(tree_root,
 					  btrfs_super_device_root(disk_super));
+
+	ret = read_device_info(dev_root);
+	BUG_ON(ret);
+
 	tree_root->node = read_tree_block(tree_root,
 					  btrfs_super_root(disk_super));
 	BUG_ON(!tree_root->node);
@@ -562,7 +676,7 @@ static int free_dev_radix(struct btrfs_fs_info *fs_info)
 				close_bdev_excl(lookup[i]->bdev);
 			radix_tree_delete(&fs_info->dev_radix,
 					  lookup[i]->block_start +
-					  lookup[i]->num_blocks);
+					  lookup[i]->num_blocks - 1);
 			kfree(lookup[i]);
 		}
 	}
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 46cc5dc9d572..aa9f56c38419 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -42,4 +42,10 @@ int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 				      struct btrfs_key *location);
 u64 bh_blocknr(struct buffer_head *bh);
+int btrfs_insert_dev_radix(struct btrfs_root *root,
+			   struct block_device *bdev,
+			   u64 block_start,
+			   u64 num_blocks);
+int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
+			     u64 logical);
 #endif
diff --git a/fs/btrfs/ioctl.h b/fs/btrfs/ioctl.h
index 201fb3270723..497fadd98c3a 100644
--- a/fs/btrfs/ioctl.h
+++ b/fs/btrfs/ioctl.h
@@ -10,4 +10,6 @@ struct btrfs_ioctl_vol_args {
 
 #define BTRFS_IOC_SNAP_CREATE _IOW(BTRFS_IOCTL_MAGIC, 1, \
 				   struct btrfs_ioctl_vol_args)
+#define BTRFS_IOC_ADD_DISK _IOW(BTRFS_IOCTL_MAGIC, 2, \
+				   struct btrfs_ioctl_vol_args)
 #endif
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 4ff0cc1efb13..c46d7eafcf62 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -1073,6 +1073,7 @@ static int btrfs_get_block_lock(struct inode *inode, sector_t iblock,
 
 		}
 		map_bh(result, inode->i_sb, blocknr);
+		btrfs_map_bh_to_logical(root, result, blocknr);
 		goto out;
 	}
 
@@ -1092,7 +1093,8 @@ static int btrfs_get_block_lock(struct inode *inode, sector_t iblock,
 	extent_end = extent_start + btrfs_file_extent_num_blocks(item);
 	if (iblock >= extent_start && iblock < extent_end) {
 		err = 0;
-		map_bh(result, inode->i_sb, blocknr + iblock - extent_start);
+		btrfs_map_bh_to_logical(root, result, blocknr + iblock -
+					extent_start);
 		goto out;
 	}
 allocate:
@@ -1112,6 +1114,7 @@ allocate:
 	set_buffer_new(result);
 	map_bh(result, inode->i_sb, blocknr);
 
+	btrfs_map_bh_to_logical(root, result, blocknr);
 out:
 	btrfs_release_path(root, path);
 	btrfs_free_path(path);
@@ -1153,12 +1156,6 @@ static int btrfs_readpage(struct file *file, struct page *page)
 	return mpage_readpage(page, btrfs_get_block);
 }
 
-static int btrfs_readpages(struct file *file, struct address_space *mapping,
-			   struct list_head *pages, unsigned nr_pages)
-{
-	return mpage_readpages(mapping, pages, nr_pages, btrfs_get_block);
-}
-
 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
 {
 	return nobh_writepage(page, btrfs_get_block, wbc);
@@ -1831,12 +1828,81 @@ printk("adding snapshot name %.*s root %Lu %Lu %u\n", namelen, name, key.objecti
 	return 0;
 }
 
+static int add_disk(struct btrfs_root *root, char *name, int namelen)
+{
+	struct block_device *bdev;
+	struct btrfs_path *path;
+	struct super_block *sb = root->fs_info->sb;
+	struct btrfs_root *dev_root = root->fs_info->dev_root;
+	struct btrfs_trans_handle *trans;
+	struct btrfs_device_item *dev_item;
+	struct btrfs_key key;
+	u16 item_size;
+	u64 num_blocks;
+	u64 new_blocks;
+	int ret;
+printk("adding disk %s\n", name);
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+	num_blocks = btrfs_super_total_blocks(root->fs_info->disk_super);
+	bdev = open_bdev_excl(name, O_RDWR, sb);
+	if (IS_ERR(bdev)) {
+		ret = PTR_ERR(bdev);
+printk("open bdev excl failed ret %d\n", ret);
+		goto out_nolock;
+	}
+	set_blocksize(bdev, sb->s_blocksize);
+	new_blocks = bdev->bd_inode->i_size >> sb->s_blocksize_bits;
+	key.objectid = num_blocks;
+	key.offset = new_blocks;
+	key.flags = 0;
+	btrfs_set_key_type(&key, BTRFS_DEV_ITEM_KEY);
+
+	mutex_lock(&dev_root->fs_info->fs_mutex);
+	trans = btrfs_start_transaction(dev_root, 1);
+	item_size = sizeof(*dev_item) + namelen;
+printk("insert empty on %Lu %Lu %u size %d\n", num_blocks, new_blocks, key.flags, item_size);
+	ret = btrfs_insert_empty_item(trans, dev_root, path, &key, item_size);
+	if (ret) {
+printk("insert failed %d\n", ret);
+		close_bdev_excl(bdev);
+		if (ret > 0)
+			ret = -EEXIST;
+		goto out;
+	}
+	dev_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
+				  path->slots[0], struct btrfs_device_item);
+	btrfs_set_device_pathlen(dev_item, namelen);
+	memcpy(dev_item + 1, name, namelen);
+	mark_buffer_dirty(path->nodes[0]);
+
+	ret = btrfs_insert_dev_radix(root, bdev, num_blocks, new_blocks);
+
+	if (!ret) {
+		btrfs_set_super_total_blocks(root->fs_info->disk_super,
+					     num_blocks + new_blocks);
+		i_size_write(root->fs_info->btree_inode,
+			     (num_blocks + new_blocks) <<
+			     root->fs_info->btree_inode->i_blkbits);
+	}
+
+out:
+	ret = btrfs_commit_transaction(trans, dev_root);
+	BUG_ON(ret);
+	mutex_unlock(&root->fs_info->fs_mutex);
+out_nolock:
+	btrfs_free_path(path);
+
+	return ret;
+}
+
 static int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int
 		       cmd, unsigned long arg)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_ioctl_vol_args vol_args;
-	int ret;
+	int ret = 0;
 	int namelen;
 	struct btrfs_path *path;
 	u64 root_dirid;
@@ -1869,10 +1935,21 @@ static int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int
 			ret = create_snapshot(root, vol_args.name, namelen);
 		WARN_ON(ret);
 		break;
+	case BTRFS_IOC_ADD_DISK:
+		if (copy_from_user(&vol_args,
+				   (struct btrfs_ioctl_vol_args __user *)arg,
+				   sizeof(vol_args)))
+			return -EFAULT;
+		namelen = strlen(vol_args.name);
+		if (namelen > BTRFS_VOL_NAME_MAX)
+			return -EINVAL;
+		vol_args.name[namelen] = '\0';
+		ret = add_disk(root, vol_args.name, namelen);
+		break;
 	default:
 		return -ENOTTY;
 	}
-	return 0;
+	return ret;
 }
 
 static struct kmem_cache *btrfs_inode_cachep;
@@ -2004,7 +2081,6 @@ static struct file_operations btrfs_dir_file_operations = {
 
 static struct address_space_operations btrfs_aops = {
 	.readpage	= btrfs_readpage,
-	.readpages	= btrfs_readpages,
 	.writepage	= btrfs_writepage,
 	.sync_page	= block_sync_page,
 	.prepare_write	= btrfs_prepare_write,
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 3a15943ea8ed..8740752f3845 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -116,7 +116,13 @@ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
 	struct btrfs_fs_info *fs_info = root->fs_info;
 	struct btrfs_root *tree_root = fs_info->tree_root;
 	struct btrfs_root *extent_root = fs_info->extent_root;
+	struct btrfs_root *dev_root = fs_info->dev_root;
 
+	if (btrfs_super_device_root(fs_info->disk_super) !=
+	    bh_blocknr(dev_root->node)) {
+		btrfs_set_super_device_root(fs_info->disk_super,
+					    bh_blocknr(dev_root->node));
+	}
 	while(1) {
 		old_extent_block = btrfs_root_blocknr(&extent_root->root_item);
 		if (old_extent_block == bh_blocknr(extent_root->node))
-- 
cgit v1.2.3


From b4100d64241fed93a3f821ddf59d11ab4443a3ba Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Thu, 12 Apr 2007 12:14:00 -0400
Subject: Btrfs: add a device id to device items

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.h   | 36 ++++++++++++++++++++++++++++++++++++
 fs/btrfs/disk-io.c | 31 ++++++++++++++++++++-----------
 fs/btrfs/disk-io.h |  1 +
 fs/btrfs/super.c   |  9 ++++++++-
 4 files changed, 65 insertions(+), 12 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 454eb88611bb..ef3583cf74d6 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -91,10 +91,12 @@ struct btrfs_super_block {
 	__le64 total_blocks;
 	__le64 blocks_used;
 	__le64 root_dir_objectid;
+	__le64 last_device_id;
 	/* fields below here vary with the underlying disk */
 	__le64 device_block_start;
 	__le64 device_num_blocks;
 	__le64 device_root;
+	__le64 device_id;
 } __attribute__ ((__packed__));
 
 /*
@@ -230,6 +232,7 @@ struct btrfs_csum_item {
 
 struct btrfs_device_item {
 	__le16 pathlen;
+	__le64 device_id;
 } __attribute__ ((__packed__));
 
 struct crypto_hash;
@@ -798,6 +801,28 @@ static inline void btrfs_set_super_root_dir(struct btrfs_super_block *s, u64
 	s->root_dir_objectid = cpu_to_le64(val);
 }
 
+static inline u64 btrfs_super_last_device_id(struct btrfs_super_block *s)
+{
+	return le64_to_cpu(s->last_device_id);
+}
+
+static inline void btrfs_set_super_last_device_id(struct btrfs_super_block *s,
+						  u64 val)
+{
+	s->last_device_id = cpu_to_le64(val);
+}
+
+static inline u64 btrfs_super_device_id(struct btrfs_super_block *s)
+{
+	return le64_to_cpu(s->device_id);
+}
+
+static inline void btrfs_set_super_device_id(struct btrfs_super_block *s,
+						  u64 val)
+{
+	s->device_id = cpu_to_le64(val);
+}
+
 static inline u64 btrfs_super_device_block_start(struct btrfs_super_block *s)
 {
 	return le64_to_cpu(s->device_block_start);
@@ -910,6 +935,17 @@ static inline void btrfs_set_device_pathlen(struct btrfs_device_item *d,
 	d->pathlen = cpu_to_le16(val);
 }
 
+static inline u64 btrfs_device_id(struct btrfs_device_item *d)
+{
+	return le64_to_cpu(d->device_id);
+}
+
+static inline void btrfs_set_device_id(struct btrfs_device_item *d,
+						u64 val)
+{
+	d->device_id = cpu_to_le64(val);
+}
+
 static inline struct btrfs_root *btrfs_sb(struct super_block *sb)
 {
 	return sb->s_fs_info;
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 3ba4df20c913..06b969c14625 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -13,11 +13,13 @@
 struct dev_lookup {
 	u64 block_start;
 	u64 num_blocks;
+	u64 device_id;
 	struct block_device *bdev;
 };
 
 int btrfs_insert_dev_radix(struct btrfs_root *root,
 			   struct block_device *bdev,
+			   u64 device_id,
 			   u64 block_start,
 			   u64 num_blocks)
 {
@@ -31,6 +33,7 @@ int btrfs_insert_dev_radix(struct btrfs_root *root,
 	lookup->block_start = block_start;
 	lookup->num_blocks = num_blocks;
 	lookup->bdev = bdev;
+	lookup->device_id = device_id;
 printk("inserting %s into dev radix %Lu %Lu\n", bdevname(bdev, b), block_start, num_blocks);
 
 	ret = radix_tree_insert(&root->fs_info->dev_radix, block_start +
@@ -418,17 +421,14 @@ printk("all worked\n");
 	return root;
 }
 
-int btrfs_open_disk(struct btrfs_root *root, u64 block_start, u64 num_blocks,
-		    char *filename, int name_len)
+static int btrfs_open_disk(struct btrfs_root *root, u64 device_id,
+			   u64 block_start, u64 num_blocks,
+			   char *filename, int name_len)
 {
 	char *null_filename;
 	struct block_device *bdev;
 	int ret;
 
-	if (block_start == 0) {
-printk("skipping disk with block_start == 0\n");
-return 0;
-	}
 	null_filename = kmalloc(name_len + 1, GFP_NOFS);
 	if (!null_filename)
 		return -ENOMEM;
@@ -441,7 +441,8 @@ return 0;
 		goto out;
 	}
 	set_blocksize(bdev, root->fs_info->sb->s_blocksize);
-	ret = btrfs_insert_dev_radix(root, bdev, block_start, num_blocks);
+	ret = btrfs_insert_dev_radix(root, bdev, device_id,
+				     block_start, num_blocks);
 	BUG_ON(ret);
 	ret = 0;
 out:
@@ -490,10 +491,14 @@ static int read_device_info(struct btrfs_root *root)
 		}
 		dev_item = btrfs_item_ptr(leaf, slot, struct btrfs_device_item);
 printk("found key %Lu %Lu\n", key.objectid, key.offset);
-		ret = btrfs_open_disk(root, key.objectid, key.offset,
-				      (char *)(dev_item + 1),
-				      btrfs_device_pathlen(dev_item));
-		BUG_ON(ret);
+		if (btrfs_device_id(dev_item) !=
+		    btrfs_super_device_id(root->fs_info->disk_super)) {
+			ret = btrfs_open_disk(root, btrfs_device_id(dev_item),
+					      key.objectid, key.offset,
+					      (char *)(dev_item + 1),
+					      btrfs_device_pathlen(dev_item));
+			BUG_ON(ret);
+		}
 		path->slots[0]++;
 	}
 	btrfs_free_path(path);
@@ -556,6 +561,7 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	dev_lookup->block_start = 0;
 	dev_lookup->num_blocks = (u32)-2;
 	dev_lookup->bdev = sb->s_bdev;
+	dev_lookup->device_id = 0;
 	ret = radix_tree_insert(&fs_info->dev_radix, (u32)-2, dev_lookup);
 	BUG_ON(ret);
 	fs_info->sb_buffer = read_tree_block(tree_root,
@@ -575,6 +581,8 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	radix_tree_delete(&fs_info->dev_radix, (u32)-2);
 	dev_lookup->block_start = btrfs_super_device_block_start(disk_super);
 	dev_lookup->num_blocks = btrfs_super_device_num_blocks(disk_super);
+	dev_lookup->device_id = btrfs_super_device_id(disk_super);
+
 	ret = radix_tree_insert(&fs_info->dev_radix,
 				dev_lookup->block_start +
 				dev_lookup->num_blocks - 1, dev_lookup);
@@ -659,6 +667,7 @@ int del_fs_roots(struct btrfs_fs_info *fs_info)
 	}
 	return 0;
 }
+
 static int free_dev_radix(struct btrfs_fs_info *fs_info)
 {
 	struct dev_lookup *lookup[8];
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index aa9f56c38419..444ebb0141ae 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -44,6 +44,7 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 u64 bh_blocknr(struct buffer_head *bh);
 int btrfs_insert_dev_radix(struct btrfs_root *root,
 			   struct block_device *bdev,
+			   u64 device_id,
 			   u64 block_start,
 			   u64 num_blocks);
 int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index c46d7eafcf62..8dcf600e39ef 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -1840,7 +1840,9 @@ static int add_disk(struct btrfs_root *root, char *name, int namelen)
 	u16 item_size;
 	u64 num_blocks;
 	u64 new_blocks;
+	u64 device_id;
 	int ret;
+
 printk("adding disk %s\n", name);
 	path = btrfs_alloc_path();
 	if (!path)
@@ -1875,9 +1877,14 @@ printk("insert failed %d\n", ret);
 				  path->slots[0], struct btrfs_device_item);
 	btrfs_set_device_pathlen(dev_item, namelen);
 	memcpy(dev_item + 1, name, namelen);
+
+	device_id = btrfs_super_last_device_id(root->fs_info->disk_super) + 1;
+	btrfs_set_super_last_device_id(root->fs_info->disk_super, device_id);
+	btrfs_set_device_id(dev_item, device_id);
 	mark_buffer_dirty(path->nodes[0]);
 
-	ret = btrfs_insert_dev_radix(root, bdev, num_blocks, new_blocks);
+	ret = btrfs_insert_dev_radix(root, bdev, device_id, num_blocks,
+				     new_blocks);
 
 	if (!ret) {
 		btrfs_set_super_total_blocks(root->fs_info->disk_super,
-- 
cgit v1.2.3


From 090d18753c7fb73f7d846c3a89a50cd35136d144 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Tue, 1 May 2007 08:53:32 -0400
Subject: Btrfs: directory readahead

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/disk-io.c     | 29 ++++++++++++++++++++++++++++-
 fs/btrfs/disk-io.h     |  6 ++++++
 fs/btrfs/extent-tree.c |  1 +
 fs/btrfs/super.c       | 28 ++++++++++++++++++++++++++++
 4 files changed, 63 insertions(+), 1 deletion(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index e1b6e13a5ae8..354524adf984 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -5,6 +5,7 @@
 #include <linux/scatterlist.h>
 #include <linux/swap.h>
 #include <linux/radix-tree.h>
+#include <linux/file.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
@@ -253,6 +254,29 @@ static struct address_space_operations btree_aops = {
 	.sync_page	= block_sync_page,
 };
 
+int readahead_tree_block(struct btrfs_root *root, u64 blocknr)
+{
+	struct buffer_head *bh = NULL;
+
+	bh = btrfs_find_create_tree_block(root, blocknr);
+	if (!bh)
+		return 0;
+	if (buffer_uptodate(bh))
+		goto done;
+	if (test_set_buffer_locked(bh))
+		goto done;
+	if (!buffer_uptodate(bh)) {
+		get_bh(bh);
+		bh->b_end_io = end_buffer_read_sync;
+		submit_bh(READ, bh);
+	} else {
+		unlock_buffer(bh);
+	}
+done:
+	brelse(bh);
+	return 0;
+}
+
 struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr)
 {
 	struct buffer_head *bh = NULL;
@@ -270,11 +294,14 @@ struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr)
 		wait_on_buffer(bh);
 		if (!buffer_uptodate(bh))
 			goto fail;
-		csum_tree_block(root, bh, 1);
 	} else {
 		unlock_buffer(bh);
 	}
 uptodate:
+	if (!buffer_checked(bh)) {
+		csum_tree_block(root, bh, 1);
+		set_buffer_checked(bh);
+	}
 	if (check_tree_block(root, bh))
 		BUG();
 	return bh;
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 444ebb0141ae..1ee7d2a55b5b 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -5,6 +5,11 @@
 
 #define BTRFS_SUPER_INFO_OFFSET (16 * 1024)
 
+enum btrfs_bh_state_bits {
+	BH_Checked = BH_PrivateStart,
+};
+BUFFER_FNS(Checked, checked);
+
 static inline struct btrfs_node *btrfs_buffer_node(struct buffer_head *bh)
 {
 	return (struct btrfs_node *)bh->b_data;
@@ -21,6 +26,7 @@ static inline struct btrfs_header *btrfs_buffer_header(struct buffer_head *bh)
 }
 
 struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr);
+int readahead_tree_block(struct btrfs_root *root, u64 blocknr);
 struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
 						 u64 blocknr);
 int write_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 8b8cbe25fffb..2837fddf7b1c 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -812,6 +812,7 @@ struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 	BUG_ON(ret);
 	buf = btrfs_find_create_tree_block(root, ins.objectid);
 	set_buffer_uptodate(buf);
+	set_buffer_checked(buf);
 	set_radix_bit(&trans->transaction->dirty_pages, buf->b_page->index);
 	return buf;
 }
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 7ecbe7c86186..a29a781b86c2 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -585,6 +585,31 @@ printk("adding new root for inode %lu root %p (found %p)\n", inode->i_ino, sub_r
 	return d_splice_alias(inode, dentry);
 }
 
+static void reada_leaves(struct btrfs_root *root, struct btrfs_path *path)
+{
+	struct btrfs_node *node;
+	int i;
+	int nritems;
+	u64 objectid;
+	u64 item_objectid;
+	u64 blocknr;
+	int slot;
+
+	if (!path->nodes[1])
+		return;
+	node = btrfs_buffer_node(path->nodes[1]);
+	slot = path->slots[1];
+	objectid = btrfs_disk_key_objectid(&node->ptrs[slot].key);
+	nritems = btrfs_header_nritems(&node->header);
+	for (i = slot; i < nritems; i++) {
+		item_objectid = btrfs_disk_key_objectid(&node->ptrs[i].key);
+		if (item_objectid != objectid)
+			break;
+		blocknr = btrfs_node_blockptr(node, i);
+		readahead_tree_block(root, blocknr);
+	}
+}
+
 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
 {
 	struct inode *inode = filp->f_path.dentry->d_inode;
@@ -619,6 +644,7 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
 	if (ret < 0)
 		goto err;
 	advance = 0;
+	reada_leaves(root, path);
 	while(1) {
 		leaf = btrfs_buffer_leaf(path->nodes[0]);
 		nritems = btrfs_header_nritems(&leaf->header);
@@ -631,6 +657,8 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
 				leaf = btrfs_buffer_leaf(path->nodes[0]);
 				nritems = btrfs_header_nritems(&leaf->header);
 				slot = path->slots[0];
+				if (path->slots[1] == 0)
+					reada_leaves(root, path);
 			} else {
 				slot++;
 				path->slots[0]++;
-- 
cgit v1.2.3


From 35b7e476107e3d54f03384e0f2fa3dfd68933353 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Wed, 2 May 2007 15:53:43 -0400
Subject: Btrfs: fix page cache memory leak

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/bit-radix.c   |  2 --
 fs/btrfs/ctree.h       |  3 +++
 fs/btrfs/dir-item.c    | 11 +++++-----
 fs/btrfs/disk-io.c     |  8 ++++++--
 fs/btrfs/disk-io.h     |  2 ++
 fs/btrfs/extent-tree.c | 12 +----------
 fs/btrfs/hash.c        |  1 +
 fs/btrfs/print-tree.c  |  1 +
 fs/btrfs/super.c       | 56 ++++++++++++++++++++++++++++++--------------------
 fs/btrfs/transaction.c |  8 +++++---
 10 files changed, 59 insertions(+), 45 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/bit-radix.c b/fs/btrfs/bit-radix.c
index 9fc42e99c7df..e9ace32d7bae 100644
--- a/fs/btrfs/bit-radix.c
+++ b/fs/btrfs/bit-radix.c
@@ -22,9 +22,7 @@ int set_radix_bit(struct radix_tree_root *radix, unsigned long bit)
 			return -ENOMEM;
 		memset(bits + 1, 0, BIT_ARRAY_BYTES - sizeof(unsigned long));
 		bits[0] = slot;
-		radix_tree_preload(GFP_NOFS);
 		ret = radix_tree_insert(radix, slot, bits);
-		radix_tree_preload_end();
 		if (ret)
 			return ret;
 	}
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index e6bf9919536a..b5855a5365ef 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -8,6 +8,9 @@
 
 struct btrfs_trans_handle;
 struct btrfs_transaction;
+extern struct kmem_cache *btrfs_trans_handle_cachep;
+extern struct kmem_cache *btrfs_transaction_cachep;
+extern struct kmem_cache *btrfs_bit_radix_cachep;
 extern struct kmem_cache *btrfs_path_cachep;
 
 #define BTRFS_MAGIC "_BtRfS_M"
diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c
index 7a7e9846860a..00a28d90fea6 100644
--- a/fs/btrfs/dir-item.c
+++ b/fs/btrfs/dir-item.c
@@ -4,11 +4,12 @@
 #include "hash.h"
 #include "transaction.h"
 
-struct btrfs_dir_item *insert_with_overflow(struct btrfs_trans_handle *trans,
-					    struct btrfs_root *root,
-					    struct btrfs_path *path,
-					    struct btrfs_key *cpu_key,
-					    u32 data_size)
+static struct btrfs_dir_item *insert_with_overflow(struct btrfs_trans_handle
+						   *trans,
+						   struct btrfs_root *root,
+						   struct btrfs_path *path,
+						   struct btrfs_key *cpu_key,
+						   u32 data_size)
 {
 	int ret;
 	char *ptr;
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 354524adf984..5828a104dfef 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -5,7 +5,7 @@
 #include <linux/scatterlist.h>
 #include <linux/swap.h>
 #include <linux/radix-tree.h>
-#include <linux/file.h>
+#include <linux/writeback.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
@@ -694,7 +694,7 @@ static int free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
 	return 0;
 }
 
-int del_fs_roots(struct btrfs_fs_info *fs_info)
+static int del_fs_roots(struct btrfs_fs_info *fs_info)
 {
 	int ret;
 	struct btrfs_root *gang[8];
@@ -781,3 +781,7 @@ void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf)
 	brelse(buf);
 }
 
+void btrfs_btree_balance_dirty(struct btrfs_root *root)
+{
+	balance_dirty_pages_ratelimited(root->fs_info->btree_inode->i_mapping);
+}
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 1ee7d2a55b5b..822ccb8aa4ae 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -55,4 +55,6 @@ int btrfs_insert_dev_radix(struct btrfs_root *root,
 			   u64 num_blocks);
 int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
 			     u64 logical);
+int btrfs_releasepage(struct page *page, gfp_t flags);
+void btrfs_btree_balance_dirty(struct btrfs_root *root);
 #endif
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 2837fddf7b1c..0e20d1c42fca 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -322,18 +322,10 @@ static int update_block_group(struct btrfs_trans_handle *trans,
 	return 0;
 }
 
-static int try_remove_page(struct address_space *mapping, unsigned long index)
-{
-	int ret;
-	ret = invalidate_mapping_pages(mapping, index, index);
-	return ret;
-}
-
 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct
 			       btrfs_root *root)
 {
 	unsigned long gang[8];
-	struct inode *btree_inode = root->fs_info->btree_inode;
 	u64 first = 0;
 	int ret;
 	int i;
@@ -348,9 +340,6 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct
 			first = gang[0];
 		for (i = 0; i < ret; i++) {
 			clear_radix_bit(pinned_radix, gang[i]);
-			try_remove_page(btree_inode->i_mapping,
-					gang[i] << (PAGE_CACHE_SHIFT -
-						    btree_inode->i_blkbits));
 		}
 	}
 	return 0;
@@ -983,6 +972,7 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
 			break;
 		if (wret < 0)
 			ret = wret;
+		btrfs_btree_balance_dirty(root);
 	}
 	for (i = 0; i <= orig_level; i++) {
 		if (path->nodes[i]) {
diff --git a/fs/btrfs/hash.c b/fs/btrfs/hash.c
index 22519b8e0cf2..32de1ea1b64e 100644
--- a/fs/btrfs/hash.c
+++ b/fs/btrfs/hash.c
@@ -11,6 +11,7 @@
  */
 
 #include <linux/types.h>
+#include "hash.h"
 #define DELTA 0x9E3779B9
 
 static void TEA_transform(__u32 buf[2], __u32 const in[])
diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c
index 2f95fc67a036..28813411de66 100644
--- a/fs/btrfs/print-tree.c
+++ b/fs/btrfs/print-tree.c
@@ -1,6 +1,7 @@
 #include <linux/module.h>
 #include "ctree.h"
 #include "disk-io.h"
+#include "print-tree.h"
 
 void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l)
 {
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index a29a781b86c2..130a1d3d9f5f 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -18,14 +18,14 @@
 #include "btrfs_inode.h"
 #include "ioctl.h"
 
-void btrfs_fsinfo_release(struct kobject *obj)
+static void btrfs_fsinfo_release(struct kobject *obj)
 {
 	struct btrfs_fs_info *fsinfo = container_of(obj,
 					    struct btrfs_fs_info, kobj);
 	kfree(fsinfo);
 }
 
-struct kobj_type btrfs_fsinfo_ktype = {
+static struct kobj_type btrfs_fsinfo_ktype = {
 	.release = btrfs_fsinfo_release,
 };
 
@@ -148,7 +148,6 @@ static void fill_inode_item(struct btrfs_inode_item *item,
 				    BTRFS_I(inode)->block_group->key.objectid);
 }
 
-
 static int btrfs_update_inode(struct btrfs_trans_handle *trans,
 			      struct btrfs_root *root,
 			      struct inode *inode)
@@ -251,6 +250,7 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
 	ret = btrfs_unlink_trans(trans, root, dir, dentry);
 	btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
+	btrfs_btree_balance_dirty(root);
 	return ret;
 }
 
@@ -324,6 +324,7 @@ out:
 	btrfs_free_path(path);
 	mutex_unlock(&root->fs_info->fs_mutex);
 	ret = btrfs_end_transaction(trans, root);
+	btrfs_btree_balance_dirty(root);
 	if (ret && !err)
 		err = ret;
 	return err;
@@ -449,6 +450,7 @@ static void btrfs_delete_inode(struct inode *inode)
 	btrfs_free_inode(trans, root, inode);
 	btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
+	btrfs_btree_balance_dirty(root);
 	return;
 no_delete:
 	clear_inode(inode);
@@ -481,7 +483,7 @@ out:
 	return ret;
 }
 
-int fixup_tree_root_location(struct btrfs_root *root,
+static int fixup_tree_root_location(struct btrfs_root *root,
 			     struct btrfs_key *location,
 			     struct btrfs_root **sub_root)
 {
@@ -512,7 +514,7 @@ int fixup_tree_root_location(struct btrfs_root *root,
 	return 0;
 }
 
-int btrfs_init_locked_inode(struct inode *inode, void *p)
+static int btrfs_init_locked_inode(struct inode *inode, void *p)
 {
 	struct btrfs_iget_args *args = p;
 	inode->i_ino = args->ino;
@@ -520,15 +522,15 @@ int btrfs_init_locked_inode(struct inode *inode, void *p)
 	return 0;
 }
 
-int btrfs_find_actor(struct inode *inode, void *opaque)
+static int btrfs_find_actor(struct inode *inode, void *opaque)
 {
 	struct btrfs_iget_args *args = opaque;
 	return (args->ino == inode->i_ino &&
 		args->root == BTRFS_I(inode)->root);
 }
 
-struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
-				struct btrfs_root *root)
+static struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
+				       struct btrfs_root *root)
 {
 	struct inode *inode;
 	struct btrfs_iget_args args;
@@ -790,6 +792,7 @@ static void btrfs_dirty_inode(struct inode *inode)
 	btrfs_update_inode(trans, root, inode);
 	btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
+	btrfs_btree_balance_dirty(root);
 }
 
 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
@@ -913,6 +916,7 @@ out_unlock:
 		inode_dec_link_count(inode);
 		iput(inode);
 	}
+	btrfs_btree_balance_dirty(root);
 	return err;
 }
 
@@ -1002,6 +1006,7 @@ out_unlock:
 	mutex_unlock(&root->fs_info->fs_mutex);
 	if (drop_on_err)
 		iput(inode);
+	btrfs_btree_balance_dirty(root);
 	return err;
 }
 
@@ -1099,7 +1104,6 @@ static int btrfs_get_block_lock(struct inode *inode, sector_t iblock,
 	    found_type != BTRFS_EXTENT_DATA_KEY) {
 		extent_end = 0;
 		extent_start = 0;
-		btrfs_release_path(root, path);
 		goto out;
 	}
 	found_type = btrfs_file_extent_type(item);
@@ -1135,7 +1139,6 @@ static int btrfs_get_block_lock(struct inode *inode, sector_t iblock,
 		btrfs_map_bh_to_logical(root, result, 0);
 	}
 out:
-	btrfs_release_path(root, path);
 	btrfs_free_path(path);
 	return err;
 }
@@ -1231,13 +1234,13 @@ static int __btrfs_write_full_page(struct inode *inode, struct page *page,
 		} else if (!buffer_mapped(bh) && buffer_dirty(bh)) {
 			WARN_ON(bh->b_size != blocksize);
 			err = btrfs_get_block(inode, block, bh, 0);
-			if (err)
+			if (err) {
+printk("writepage going to recovery err %d\n", err);
 				goto recover;
+			}
 			if (buffer_new(bh)) {
 				/* blockdev mappings never come here */
 				clear_buffer_new(bh);
-				unmap_underlying_metadata(bh->b_bdev,
-							bh->b_blocknr);
 			}
 		}
 		bh = bh->b_this_page;
@@ -1303,11 +1306,6 @@ done:
 		if (uptodate)
 			SetPageUptodate(page);
 		end_page_writeback(page);
-		/*
-		 * The page and buffer_heads can be released at any time from
-		 * here on.
-		 */
-		wbc->pages_skipped++;	/* We didn't write this page */
 	}
 	return err;
 
@@ -1409,10 +1407,11 @@ static void btrfs_truncate(struct inode *inode)
 	btrfs_set_trans_block_group(trans, inode);
 	ret = btrfs_truncate_in_trans(trans, root, inode);
 	BUG_ON(ret);
+	btrfs_update_inode(trans, root, inode);
 	ret = btrfs_end_transaction(trans, root);
 	BUG_ON(ret);
 	mutex_unlock(&root->fs_info->fs_mutex);
-	mark_inode_dirty(inode);
+	btrfs_btree_balance_dirty(root);
 }
 
 /*
@@ -1777,10 +1776,15 @@ static int prepare_pages(struct btrfs_root *root,
 			err = -ENOMEM;
 			goto failed_release;
 		}
+		cancel_dirty_page(pages[i], PAGE_CACHE_SIZE);
+		wait_on_page_writeback(pages[i]);
 		offset = pos & (PAGE_CACHE_SIZE -1);
 		this_write = min(PAGE_CACHE_SIZE - offset, write_bytes);
-		create_empty_buffers(pages[i], root->fs_info->sb->s_blocksize,
-				     (1 << BH_Uptodate));
+		if (!page_has_buffers(pages[i])) {
+			create_empty_buffers(pages[i],
+					     root->fs_info->sb->s_blocksize,
+					     (1 << BH_Uptodate));
+		}
 		head = page_buffers(pages[i]);
 		bh = head;
 		do {
@@ -1820,7 +1824,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 	struct inode *inode = file->f_path.dentry->d_inode;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct page *pages[8];
-	struct page *pinned[2] = { NULL, NULL };
+	struct page *pinned[2];
 	unsigned long first_index;
 	unsigned long last_index;
 	u64 start_pos;
@@ -1829,6 +1833,8 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 	struct btrfs_trans_handle *trans;
 	struct btrfs_key ins;
 
+	pinned[0] = NULL;
+	pinned[1] = NULL;
 	if (file->f_flags & O_DIRECT)
 		return -EINVAL;
 	pos = *ppos;
@@ -1858,6 +1864,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 		if (!PageUptodate(pinned[0])) {
 			ret = mpage_readpage(pinned[0], btrfs_get_block);
 			BUG_ON(ret);
+			wait_on_page_locked(pinned[0]);
 		} else {
 			unlock_page(pinned[0]);
 		}
@@ -1869,6 +1876,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 		if (!PageUptodate(pinned[1])) {
 			ret = mpage_readpage(pinned[1], btrfs_get_block);
 			BUG_ON(ret);
+			wait_on_page_locked(pinned[1]);
 		} else {
 			unlock_page(pinned[1]);
 		}
@@ -1940,6 +1948,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 		num_written += write_bytes;
 
 		balance_dirty_pages_ratelimited(inode->i_mapping);
+		btrfs_btree_balance_dirty(root);
 		cond_resched();
 	}
 out_unlock:
@@ -2165,6 +2174,7 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 	iput(inode);
 
 	mutex_unlock(&root->fs_info->fs_mutex);
+	btrfs_btree_balance_dirty(root);
 	return 0;
 }
 
@@ -2220,6 +2230,7 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 	ret = btrfs_commit_transaction(trans, root);
 	BUG_ON(ret);
 	mutex_unlock(&root->fs_info->fs_mutex);
+	btrfs_btree_balance_dirty(root);
 	return 0;
 }
 
@@ -2295,6 +2306,7 @@ out:
 	mutex_unlock(&root->fs_info->fs_mutex);
 out_nolock:
 	btrfs_free_path(path);
+	btrfs_btree_balance_dirty(root);
 
 	return ret;
 }
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index bdbf514c111f..8bbe9107ff7e 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -199,8 +199,9 @@ struct dirty_root {
 	struct btrfs_root *root;
 };
 
-int add_dirty_roots(struct btrfs_trans_handle *trans,
-		    struct radix_tree_root *radix, struct list_head *list)
+static int add_dirty_roots(struct btrfs_trans_handle *trans,
+			   struct radix_tree_root *radix,
+			   struct list_head *list)
 {
 	struct dirty_root *dirty;
 	struct btrfs_root *gang[8];
@@ -246,7 +247,8 @@ int add_dirty_roots(struct btrfs_trans_handle *trans,
 	return 0;
 }
 
-int drop_dirty_roots(struct btrfs_root *tree_root, struct list_head *list)
+static int drop_dirty_roots(struct btrfs_root *tree_root,
+			    struct list_head *list)
 {
 	struct dirty_root *dirty;
 	struct btrfs_trans_handle *trans;
-- 
cgit v1.2.3


From 6cbd55707802b98843f953d1ae6d8f5bcd9a76c0 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Tue, 12 Jun 2007 09:07:21 -0400
Subject: Btrfs: add GPLv2

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/COPYING       | 356 +++++++++++++++++++++++++++++++++++++++++++++++++
 fs/btrfs/TODO          |   1 -
 fs/btrfs/bit-radix.c   |  18 +++
 fs/btrfs/bit-radix.h   |  18 +++
 fs/btrfs/btrfs_inode.h |  18 +++
 fs/btrfs/ctree.c       |  18 +++
 fs/btrfs/ctree.h       |  18 +++
 fs/btrfs/dir-item.c    |  18 +++
 fs/btrfs/disk-io.c     |  18 +++
 fs/btrfs/disk-io.h     |  18 +++
 fs/btrfs/extent-tree.c |  18 +++
 fs/btrfs/file-item.c   |  18 +++
 fs/btrfs/file.c        |  18 +++
 fs/btrfs/hash.c        |  18 +++
 fs/btrfs/hash.h        |  18 +++
 fs/btrfs/inode-item.c  |  18 +++
 fs/btrfs/inode-map.c   |  18 +++
 fs/btrfs/inode.c       |  18 +++
 fs/btrfs/ioctl.h       |  18 +++
 fs/btrfs/print-tree.c  |  18 +++
 fs/btrfs/print-tree.h  |  18 +++
 fs/btrfs/root-tree.c   |  18 +++
 fs/btrfs/super.c       |  18 +++
 fs/btrfs/sysfs.c       |  18 +++
 fs/btrfs/transaction.c |  18 +++
 fs/btrfs/transaction.h |  18 +++
 26 files changed, 788 insertions(+), 1 deletion(-)
 create mode 100644 fs/btrfs/COPYING

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/COPYING b/fs/btrfs/COPYING
new file mode 100644
index 000000000000..ca442d313d86
--- /dev/null
+++ b/fs/btrfs/COPYING
@@ -0,0 +1,356 @@
+
+   NOTE! This copyright does *not* cover user programs that use kernel
+ services by normal system calls - this is merely considered normal use
+ of the kernel, and does *not* fall under the heading of "derived work".
+ Also note that the GPL below is copyrighted by the Free Software
+ Foundation, but the instance of code that it refers to (the Linux
+ kernel) is copyrighted by me and others who actually wrote it.
+
+ Also note that the only valid version of the GPL as far as the kernel
+ is concerned is _this_ particular version of the license (ie v2, not
+ v2.2 or v3.x or whatever), unless explicitly otherwise stated.
+
+			Linus Torvalds
+
+----------------------------------------
+
+		    GNU GENERAL PUBLIC LICENSE
+		       Version 2, June 1991
+
+ Copyright (C) 1989, 1991 Free Software Foundation, Inc.
+                       51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+ Everyone is permitted to copy and distribute verbatim copies
+ of this license document, but changing it is not allowed.
+
+			    Preamble
+
+  The licenses for most software are designed to take away your
+freedom to share and change it.  By contrast, the GNU General Public
+License is intended to guarantee your freedom to share and change free
+software--to make sure the software is free for all its users.  This
+General Public License applies to most of the Free Software
+Foundation's software and to any other program whose authors commit to
+using it.  (Some other Free Software Foundation software is covered by
+the GNU Library General Public License instead.)  You can apply it to
+your programs, too.
+
+  When we speak of free software, we are referring to freedom, not
+price.  Our General Public Licenses are designed to make sure that you
+have the freedom to distribute copies of free software (and charge for
+this service if you wish), that you receive source code or can get it
+if you want it, that you can change the software or use pieces of it
+in new free programs; and that you know you can do these things.
+
+  To protect your rights, we need to make restrictions that forbid
+anyone to deny you these rights or to ask you to surrender the rights.
+These restrictions translate to certain responsibilities for you if you
+distribute copies of the software, or if you modify it.
+
+  For example, if you distribute copies of such a program, whether
+gratis or for a fee, you must give the recipients all the rights that
+you have.  You must make sure that they, too, receive or can get the
+source code.  And you must show them these terms so they know their
+rights.
+
+  We protect your rights with two steps: (1) copyright the software, and
+(2) offer you this license which gives you legal permission to copy,
+distribute and/or modify the software.
+
+  Also, for each author's protection and ours, we want to make certain
+that everyone understands that there is no warranty for this free
+software.  If the software is modified by someone else and passed on, we
+want its recipients to know that what they have is not the original, so
+that any problems introduced by others will not reflect on the original
+authors' reputations.
+
+  Finally, any free program is threatened constantly by software
+patents.  We wish to avoid the danger that redistributors of a free
+program will individually obtain patent licenses, in effect making the
+program proprietary.  To prevent this, we have made it clear that any
+patent must be licensed for everyone's free use or not licensed at all.
+
+  The precise terms and conditions for copying, distribution and
+modification follow.
+
+		    GNU GENERAL PUBLIC LICENSE
+   TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
+
+  0. This License applies to any program or other work which contains
+a notice placed by the copyright holder saying it may be distributed
+under the terms of this General Public License.  The "Program", below,
+refers to any such program or work, and a "work based on the Program"
+means either the Program or any derivative work under copyright law:
+that is to say, a work containing the Program or a portion of it,
+either verbatim or with modifications and/or translated into another
+language.  (Hereinafter, translation is included without limitation in
+the term "modification".)  Each licensee is addressed as "you".
+
+Activities other than copying, distribution and modification are not
+covered by this License; they are outside its scope.  The act of
+running the Program is not restricted, and the output from the Program
+is covered only if its contents constitute a work based on the
+Program (independent of having been made by running the Program).
+Whether that is true depends on what the Program does.
+
+  1. You may copy and distribute verbatim copies of the Program's
+source code as you receive it, in any medium, provided that you
+conspicuously and appropriately publish on each copy an appropriate
+copyright notice and disclaimer of warranty; keep intact all the
+notices that refer to this License and to the absence of any warranty;
+and give any other recipients of the Program a copy of this License
+along with the Program.
+
+You may charge a fee for the physical act of transferring a copy, and
+you may at your option offer warranty protection in exchange for a fee.
+
+  2. You may modify your copy or copies of the Program or any portion
+of it, thus forming a work based on the Program, and copy and
+distribute such modifications or work under the terms of Section 1
+above, provided that you also meet all of these conditions:
+
+    a) You must cause the modified files to carry prominent notices
+    stating that you changed the files and the date of any change.
+
+    b) You must cause any work that you distribute or publish, that in
+    whole or in part contains or is derived from the Program or any
+    part thereof, to be licensed as a whole at no charge to all third
+    parties under the terms of this License.
+
+    c) If the modified program normally reads commands interactively
+    when run, you must cause it, when started running for such
+    interactive use in the most ordinary way, to print or display an
+    announcement including an appropriate copyright notice and a
+    notice that there is no warranty (or else, saying that you provide
+    a warranty) and that users may redistribute the program under
+    these conditions, and telling the user how to view a copy of this
+    License.  (Exception: if the Program itself is interactive but
+    does not normally print such an announcement, your work based on
+    the Program is not required to print an announcement.)
+
+These requirements apply to the modified work as a whole.  If
+identifiable sections of that work are not derived from the Program,
+and can be reasonably considered independent and separate works in
+themselves, then this License, and its terms, do not apply to those
+sections when you distribute them as separate works.  But when you
+distribute the same sections as part of a whole which is a work based
+on the Program, the distribution of the whole must be on the terms of
+this License, whose permissions for other licensees extend to the
+entire whole, and thus to each and every part regardless of who wrote it.
+
+Thus, it is not the intent of this section to claim rights or contest
+your rights to work written entirely by you; rather, the intent is to
+exercise the right to control the distribution of derivative or
+collective works based on the Program.
+
+In addition, mere aggregation of another work not based on the Program
+with the Program (or with a work based on the Program) on a volume of
+a storage or distribution medium does not bring the other work under
+the scope of this License.
+
+  3. You may copy and distribute the Program (or a work based on it,
+under Section 2) in object code or executable form under the terms of
+Sections 1 and 2 above provided that you also do one of the following:
+
+    a) Accompany it with the complete corresponding machine-readable
+    source code, which must be distributed under the terms of Sections
+    1 and 2 above on a medium customarily used for software interchange; or,
+
+    b) Accompany it with a written offer, valid for at least three
+    years, to give any third party, for a charge no more than your
+    cost of physically performing source distribution, a complete
+    machine-readable copy of the corresponding source code, to be
+    distributed under the terms of Sections 1 and 2 above on a medium
+    customarily used for software interchange; or,
+
+    c) Accompany it with the information you received as to the offer
+    to distribute corresponding source code.  (This alternative is
+    allowed only for noncommercial distribution and only if you
+    received the program in object code or executable form with such
+    an offer, in accord with Subsection b above.)
+
+The source code for a work means the preferred form of the work for
+making modifications to it.  For an executable work, complete source
+code means all the source code for all modules it contains, plus any
+associated interface definition files, plus the scripts used to
+control compilation and installation of the executable.  However, as a
+special exception, the source code distributed need not include
+anything that is normally distributed (in either source or binary
+form) with the major components (compiler, kernel, and so on) of the
+operating system on which the executable runs, unless that component
+itself accompanies the executable.
+
+If distribution of executable or object code is made by offering
+access to copy from a designated place, then offering equivalent
+access to copy the source code from the same place counts as
+distribution of the source code, even though third parties are not
+compelled to copy the source along with the object code.
+
+  4. You may not copy, modify, sublicense, or distribute the Program
+except as expressly provided under this License.  Any attempt
+otherwise to copy, modify, sublicense or distribute the Program is
+void, and will automatically terminate your rights under this License.
+However, parties who have received copies, or rights, from you under
+this License will not have their licenses terminated so long as such
+parties remain in full compliance.
+
+  5. You are not required to accept this License, since you have not
+signed it.  However, nothing else grants you permission to modify or
+distribute the Program or its derivative works.  These actions are
+prohibited by law if you do not accept this License.  Therefore, by
+modifying or distributing the Program (or any work based on the
+Program), you indicate your acceptance of this License to do so, and
+all its terms and conditions for copying, distributing or modifying
+the Program or works based on it.
+
+  6. Each time you redistribute the Program (or any work based on the
+Program), the recipient automatically receives a license from the
+original licensor to copy, distribute or modify the Program subject to
+these terms and conditions.  You may not impose any further
+restrictions on the recipients' exercise of the rights granted herein.
+You are not responsible for enforcing compliance by third parties to
+this License.
+
+  7. If, as a consequence of a court judgment or allegation of patent
+infringement or for any other reason (not limited to patent issues),
+conditions are imposed on you (whether by court order, agreement or
+otherwise) that contradict the conditions of this License, they do not
+excuse you from the conditions of this License.  If you cannot
+distribute so as to satisfy simultaneously your obligations under this
+License and any other pertinent obligations, then as a consequence you
+may not distribute the Program at all.  For example, if a patent
+license would not permit royalty-free redistribution of the Program by
+all those who receive copies directly or indirectly through you, then
+the only way you could satisfy both it and this License would be to
+refrain entirely from distribution of the Program.
+
+If any portion of this section is held invalid or unenforceable under
+any particular circumstance, the balance of the section is intended to
+apply and the section as a whole is intended to apply in other
+circumstances.
+
+It is not the purpose of this section to induce you to infringe any
+patents or other property right claims or to contest validity of any
+such claims; this section has the sole purpose of protecting the
+integrity of the free software distribution system, which is
+implemented by public license practices.  Many people have made
+generous contributions to the wide range of software distributed
+through that system in reliance on consistent application of that
+system; it is up to the author/donor to decide if he or she is willing
+to distribute software through any other system and a licensee cannot
+impose that choice.
+
+This section is intended to make thoroughly clear what is believed to
+be a consequence of the rest of this License.
+
+  8. If the distribution and/or use of the Program is restricted in
+certain countries either by patents or by copyrighted interfaces, the
+original copyright holder who places the Program under this License
+may add an explicit geographical distribution limitation excluding
+those countries, so that distribution is permitted only in or among
+countries not thus excluded.  In such case, this License incorporates
+the limitation as if written in the body of this License.
+
+  9. The Free Software Foundation may publish revised and/or new versions
+of the General Public License from time to time.  Such new versions will
+be similar in spirit to the present version, but may differ in detail to
+address new problems or concerns.
+
+Each version is given a distinguishing version number.  If the Program
+specifies a version number of this License which applies to it and "any
+later version", you have the option of following the terms and conditions
+either of that version or of any later version published by the Free
+Software Foundation.  If the Program does not specify a version number of
+this License, you may choose any version ever published by the Free Software
+Foundation.
+
+  10. If you wish to incorporate parts of the Program into other free
+programs whose distribution conditions are different, write to the author
+to ask for permission.  For software which is copyrighted by the Free
+Software Foundation, write to the Free Software Foundation; we sometimes
+make exceptions for this.  Our decision will be guided by the two goals
+of preserving the free status of all derivatives of our free software and
+of promoting the sharing and reuse of software generally.
+
+			    NO WARRANTY
+
+  11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
+FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW.  EXCEPT WHEN
+OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
+PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
+OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.  THE ENTIRE RISK AS
+TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU.  SHOULD THE
+PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
+REPAIR OR CORRECTION.
+
+  12. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
+WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
+REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
+INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
+OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
+TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
+YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
+PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
+POSSIBILITY OF SUCH DAMAGES.
+
+		     END OF TERMS AND CONDITIONS
+
+	    How to Apply These Terms to Your New Programs
+
+  If you develop a new program, and you want it to be of the greatest
+possible use to the public, the best way to achieve this is to make it
+free software which everyone can redistribute and change under these terms.
+
+  To do so, attach the following notices to the program.  It is safest
+to attach them to the start of each source file to most effectively
+convey the exclusion of warranty; and each file should have at least
+the "copyright" line and a pointer to where the full notice is found.
+
+    <one line to give the program's name and a brief idea of what it does.>
+    Copyright (C) <year>  <name of author>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License
+    along with this program; if not, write to the Free Software
+    Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+
+
+Also add information on how to contact you by electronic and paper mail.
+
+If the program is interactive, make it output a short notice like this
+when it starts in an interactive mode:
+
+    Gnomovision version 69, Copyright (C) year name of author
+    Gnomovision comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
+    This is free software, and you are welcome to redistribute it
+    under certain conditions; type `show c' for details.
+
+The hypothetical commands `show w' and `show c' should show the appropriate
+parts of the General Public License.  Of course, the commands you use may
+be called something other than `show w' and `show c'; they could even be
+mouse-clicks or menu items--whatever suits your program.
+
+You should also get your employer (if you work as a programmer) or your
+school, if any, to sign a "copyright disclaimer" for the program, if
+necessary.  Here is a sample; alter the names:
+
+  Yoyodyne, Inc., hereby disclaims all copyright interest in the program
+  `Gnomovision' (which makes passes at compilers) written by James Hacker.
+
+  <signature of Ty Coon>, 1 April 1989
+  Ty Coon, President of Vice
+
+This General Public License does not permit incorporating your program into
+proprietary programs.  If your program is a subroutine library, you may
+consider it more useful to permit linking proprietary applications with the
+library.  If this is what you want to do, use the GNU Library General
+Public License instead of this License.
diff --git a/fs/btrfs/TODO b/fs/btrfs/TODO
index 4a48a2365d57..d9b6d38c603a 100644
--- a/fs/btrfs/TODO
+++ b/fs/btrfs/TODO
@@ -1,4 +1,3 @@
-* fix printk warnings
 * cleanup, add more error checking, get rid of BUG_ONs
 * Fix ENOSPC handling
 * Make allocator smarter
diff --git a/fs/btrfs/bit-radix.c b/fs/btrfs/bit-radix.c
index 8f9cd4277231..a4ddf4de9797 100644
--- a/fs/btrfs/bit-radix.c
+++ b/fs/btrfs/bit-radix.c
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #include <linux/module.h>
 #include "bit-radix.h"
 
diff --git a/fs/btrfs/bit-radix.h b/fs/btrfs/bit-radix.h
index 4e717e30db4f..c100f54d5c32 100644
--- a/fs/btrfs/bit-radix.h
+++ b/fs/btrfs/bit-radix.h
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #ifndef __BIT_RADIX__
 #define __BIT_RADIX__
 #include <linux/radix-tree.h>
diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
index 6b50076b02d5..783f06b63259 100644
--- a/fs/btrfs/btrfs_inode.h
+++ b/fs/btrfs/btrfs_inode.h
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #ifndef __BTRFS_I__
 #define __BTRFS_I__
 
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 20e01f95ce6e..4736155af419 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #include <linux/module.h>
 #include "ctree.h"
 #include "disk-io.h"
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 24410d42e3ce..4bd648d68e88 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #ifndef __BTRFS__
 #define __BTRFS__
 
diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c
index b408a3d20c7b..0d179a44ebf4 100644
--- a/fs/btrfs/dir-item.c
+++ b/fs/btrfs/dir-item.c
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #include <linux/module.h>
 #include "ctree.h"
 #include "disk-io.h"
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 878101172546..8c68a64c0ceb 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #include <linux/module.h>
 #include <linux/fs.h>
 #include <linux/blkdev.h>
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 822ccb8aa4ae..7b76ccc48754 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #ifndef __DISKIO__
 #define __DISKIO__
 
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index fe02fbfa6d59..975a0eb9241b 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #include <linux/module.h>
 #include "ctree.h"
 #include "disk-io.h"
diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c
index d5a98827e38a..37dd3801a107 100644
--- a/fs/btrfs/file-item.c
+++ b/fs/btrfs/file-item.c
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #include <linux/module.h>
 #include "ctree.h"
 #include "disk-io.h"
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 0325dc038592..566e526b60ff 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #include <linux/module.h>
 #include <linux/buffer_head.h>
 #include <linux/fs.h>
diff --git a/fs/btrfs/hash.c b/fs/btrfs/hash.c
index 32de1ea1b64e..e5c76903d722 100644
--- a/fs/btrfs/hash.c
+++ b/fs/btrfs/hash.c
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 /*
  *  Original copy from:
  *  linux/fs/ext3/hash.c
diff --git a/fs/btrfs/hash.h b/fs/btrfs/hash.h
index e56c4670093d..d3be02670585 100644
--- a/fs/btrfs/hash.h
+++ b/fs/btrfs/hash.h
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #ifndef __HASH__
 #define __HASH__
 int btrfs_name_hash(const char *name, int len, u64 *hash_result);
diff --git a/fs/btrfs/inode-item.c b/fs/btrfs/inode-item.c
index c3b990b661c2..449e9dbdf957 100644
--- a/fs/btrfs/inode-item.c
+++ b/fs/btrfs/inode-item.c
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #include <linux/module.h>
 #include "ctree.h"
 #include "disk-io.h"
diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c
index 318e27a6378b..58930fc48be6 100644
--- a/fs/btrfs/inode-map.c
+++ b/fs/btrfs/inode-map.c
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #include <linux/module.h>
 #include "ctree.h"
 #include "disk-io.h"
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 1cabc1951b0e..46f869d6d0fb 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #include <linux/module.h>
 #include <linux/buffer_head.h>
 #include <linux/fs.h>
diff --git a/fs/btrfs/ioctl.h b/fs/btrfs/ioctl.h
index 497fadd98c3a..23bed48c51a7 100644
--- a/fs/btrfs/ioctl.h
+++ b/fs/btrfs/ioctl.h
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #ifndef __IOCTL_
 #define __IOCTL_
 #include <linux/ioctl.h>
diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c
index 21791f037561..1fa705683008 100644
--- a/fs/btrfs/print-tree.c
+++ b/fs/btrfs/print-tree.c
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #include <linux/module.h>
 #include "ctree.h"
 #include "disk-io.h"
diff --git a/fs/btrfs/print-tree.h b/fs/btrfs/print-tree.h
index 396041a05cfa..205cd03601ab 100644
--- a/fs/btrfs/print-tree.h
+++ b/fs/btrfs/print-tree.h
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #ifndef __PRINT_TREE_
 #define __PRINT_TREE_
 void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l);
diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c
index 13c30284fd84..7a2d63c1ee80 100644
--- a/fs/btrfs/root-tree.c
+++ b/fs/btrfs/root-tree.c
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #include <linux/module.h>
 #include "ctree.h"
 #include "disk-io.h"
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 94b3535a2a11..ef8ad18b275c 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #include <linux/module.h>
 #include <linux/buffer_head.h>
 #include <linux/fs.h>
diff --git a/fs/btrfs/sysfs.c b/fs/btrfs/sysfs.c
index ed8b7e20cfbc..ead9dc8051be 100644
--- a/fs/btrfs/sysfs.c
+++ b/fs/btrfs/sysfs.c
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #include <linux/module.h>
 #include "ctree.h"
 #include "disk-io.h"
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 4f3c8ddcb401..30914521cf01 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #include <linux/module.h>
 #include <linux/fs.h>
 #include "ctree.h"
diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h
index f25b4900db45..5fb1d322aa10 100644
--- a/fs/btrfs/transaction.h
+++ b/fs/btrfs/transaction.h
@@ -1,3 +1,21 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
 #ifndef __TRANSACTION__
 #define __TRANSACTION__
 #include "btrfs_inode.h"
-- 
cgit v1.2.3


From 5eda7b5e9b0bed864dd18284c7df9b3c8207dad7 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Fri, 22 Jun 2007 14:16:25 -0400
Subject: Btrfs: Add the ability to find and remove dead roots after a crash.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.h       | 17 ++++++++++--
 fs/btrfs/disk-io.c     | 44 +++++++++++++++++++----------
 fs/btrfs/disk-io.h     |  3 ++
 fs/btrfs/inode.c       |  2 ++
 fs/btrfs/root-tree.c   | 75 ++++++++++++++++++++++++++++++++++++++++++++------
 fs/btrfs/transaction.c | 54 +++++++++++++++++++++++++++++++-----
 fs/btrfs/transaction.h |  1 +
 7 files changed, 164 insertions(+), 32 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 77071f273977..fb6fffb71dd0 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -122,12 +122,12 @@ struct btrfs_super_block {
 	u8 fsid[16];    /* FS specific uuid */
 	__le64 blocknr; /* this block number */
 	__le64 magic;
-	__le32 blocksize;
 	__le64 generation;
 	__le64 root;
 	__le64 total_blocks;
 	__le64 blocks_used;
 	__le64 root_dir_objectid;
+	__le32 blocksize;
 } __attribute__ ((__packed__));
 
 /*
@@ -226,10 +226,12 @@ struct btrfs_root_item {
 	struct btrfs_inode_item inode;
 	__le64 root_dirid;
 	__le64 blocknr;
-	__le32 flags;
 	__le64 block_limit;
 	__le64 blocks_used;
+	__le32 flags;
 	__le32 refs;
+	struct btrfs_disk_key drop_progress;
+	u8 drop_level;
 } __attribute__ ((__packed__));
 
 #define BTRFS_FILE_EXTENT_REG 0
@@ -800,6 +802,16 @@ static inline void btrfs_set_root_refs(struct btrfs_root_item *item, u32 val)
 	item->refs = cpu_to_le32(val);
 }
 
+static inline u32 btrfs_root_flags(struct btrfs_root_item *item)
+{
+	return le32_to_cpu(item->flags);
+}
+
+static inline void btrfs_set_root_flags(struct btrfs_root_item *item, u32 val)
+{
+	item->flags = cpu_to_le32(val);
+}
+
 static inline u64 btrfs_super_blocknr(struct btrfs_super_block *s)
 {
 	return le64_to_cpu(s->blocknr);
@@ -1076,6 +1088,7 @@ int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *item);
 int btrfs_find_last_root(struct btrfs_root *root, u64 objectid, struct
 			 btrfs_root_item *item, struct btrfs_key *key);
+int btrfs_find_dead_roots(struct btrfs_root *root);
 /* dir-item.c */
 int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
 			  *root, const char *name, int name_len, u64 dir,
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 865a284aa06c..d7615e1578cc 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -326,8 +326,8 @@ static int find_and_setup_root(int blocksize,
 	return 0;
 }
 
-struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
-				      struct btrfs_key *location)
+struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
+					       struct btrfs_key *location)
 {
 	struct btrfs_root *root;
 	struct btrfs_root *tree_root = fs_info->tree_root;
@@ -336,11 +336,7 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 	u64 highest_inode;
 	int ret = 0;
 
-	root = radix_tree_lookup(&fs_info->fs_roots_radix,
-				 (unsigned long)location->objectid);
-	if (root)
-		return root;
-	root = kmalloc(sizeof(*root), GFP_NOFS);
+	root = kzalloc(sizeof(*root), GFP_NOFS);
 	if (!root)
 		return ERR_PTR(-ENOMEM);
 	if (location->offset == (u64)-1) {
@@ -383,6 +379,28 @@ out:
 	BUG_ON(!root->node);
 insert:
 	root->ref_cows = 1;
+	ret = btrfs_find_highest_inode(root, &highest_inode);
+	if (ret == 0) {
+		root->highest_inode = highest_inode;
+		root->last_inode_alloc = highest_inode;
+	}
+	return root;
+}
+
+struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
+				      struct btrfs_key *location)
+{
+	struct btrfs_root *root;
+	int ret;
+
+	root = radix_tree_lookup(&fs_info->fs_roots_radix,
+				 (unsigned long)location->objectid);
+	if (root)
+		return root;
+
+	root = btrfs_read_fs_root_no_radix(fs_info, location);
+	if (IS_ERR(root))
+		return root;
 	ret = radix_tree_insert(&fs_info->fs_roots_radix,
 				(unsigned long)root->root_key.objectid,
 				root);
@@ -391,11 +409,6 @@ insert:
 		kfree(root);
 		return ERR_PTR(ret);
 	}
-	ret = btrfs_find_highest_inode(root, &highest_inode);
-	if (ret == 0) {
-		root->highest_inode = highest_inode;
-		root->last_inode_alloc = highest_inode;
-	}
 	return root;
 }
 
@@ -489,6 +502,9 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	btrfs_read_block_groups(extent_root);
 
 	fs_info->generation = btrfs_super_generation(disk_super) + 1;
+	ret = btrfs_find_dead_roots(tree_root);
+	if (ret)
+		goto fail_tree_root;
 	mutex_unlock(&fs_info->fs_mutex);
 	return tree_root;
 
@@ -538,7 +554,7 @@ int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
 	return 0;
 }
 
-static int free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
+int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
 {
 	radix_tree_delete(&fs_info->fs_roots_radix,
 			  (unsigned long)root->root_key.objectid);
@@ -565,7 +581,7 @@ static int del_fs_roots(struct btrfs_fs_info *fs_info)
 		if (!ret)
 			break;
 		for (i = 0; i < ret; i++)
-			free_fs_root(fs_info, gang[i]);
+			btrfs_free_fs_root(fs_info, gang[i]);
 	}
 	return 0;
 }
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 7b76ccc48754..c4a695ac44f6 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -65,6 +65,8 @@ int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
 		    char *result);
 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 				      struct btrfs_key *location);
+struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
+					       struct btrfs_key *location);
 u64 bh_blocknr(struct buffer_head *bh);
 int btrfs_insert_dev_radix(struct btrfs_root *root,
 			   struct block_device *bdev,
@@ -75,4 +77,5 @@ int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
 			     u64 logical);
 int btrfs_releasepage(struct page *page, gfp_t flags);
 void btrfs_btree_balance_dirty(struct btrfs_root *root);
+int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root);
 #endif
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 6d031daa7779..9d2a0a3f674d 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -2028,6 +2028,8 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 
 	btrfs_set_root_blocknr(&root_item, bh_blocknr(subvol));
 	btrfs_set_root_refs(&root_item, 1);
+	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
+	root_item.drop_level = 0;
 	brelse(subvol);
 	subvol = NULL;
 
diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c
index ac0fae7780ff..737e5a38d17e 100644
--- a/fs/btrfs/root-tree.c
+++ b/fs/btrfs/root-tree.c
@@ -18,6 +18,7 @@
 
 #include <linux/module.h>
 #include "ctree.h"
+#include "transaction.h"
 #include "disk-io.h"
 #include "print-tree.h"
 
@@ -32,7 +33,7 @@ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid,
 
 	search_key.objectid = objectid;
 	search_key.flags = (u32)-1;
-	search_key.offset = (u32)-1;
+	search_key.offset = (u64)-1;
 
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
@@ -50,6 +51,7 @@ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid,
 	memcpy(item, btrfs_item_ptr(l, slot, struct btrfs_root_item),
 		sizeof(*item));
 	btrfs_disk_key_to_cpu(key, &l->items[slot].key);
+printk("find last finds key %Lu %u %Lu slot %d search for obj %Lu\n", key->objectid, key->flags, key->offset, slot, objectid);
 	ret = 0;
 out:
 	btrfs_release_path(root, path);
@@ -93,6 +95,67 @@ int btrfs_insert_root(struct btrfs_trans_handle *trans, struct btrfs_root
 	return ret;
 }
 
+int btrfs_find_dead_roots(struct btrfs_root *root)
+{
+	struct btrfs_root *dead_root;
+	struct btrfs_item *item;
+	struct btrfs_root_item *ri;
+	struct btrfs_key key;
+	struct btrfs_path *path;
+	int ret;
+	u32 nritems;
+	struct btrfs_leaf *leaf;
+	int slot;
+
+	key.objectid = 0;
+	key.flags = 0;
+	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
+	key.offset = 0;
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		goto err;
+	while(1) {
+		leaf = btrfs_buffer_leaf(path->nodes[0]);
+		nritems = btrfs_header_nritems(&leaf->header);
+		slot = path->slots[0];
+		if (slot >= nritems) {
+			ret = btrfs_next_leaf(root, path);
+			if (ret)
+				break;
+			leaf = btrfs_buffer_leaf(path->nodes[0]);
+			nritems = btrfs_header_nritems(&leaf->header);
+			slot = path->slots[0];
+		}
+		item = leaf->items + slot;
+		btrfs_disk_key_to_cpu(&key, &item->key);
+		if (btrfs_key_type(&key) != BTRFS_ROOT_ITEM_KEY)
+			goto next;
+		ri = btrfs_item_ptr(leaf, slot, struct btrfs_root_item);
+		if (btrfs_root_refs(ri) != 0)
+			goto next;
+		dead_root = btrfs_read_fs_root_no_radix(root->fs_info, &key);
+		if (IS_ERR(root)) {
+			ret = PTR_ERR(root);
+			goto err;
+		}
+printk("found dead root %Lu %u %Lu\n", key.objectid, key.flags, key.offset);
+		ret = btrfs_add_dead_root(dead_root,
+					  &root->fs_info->dead_roots);
+		if (ret)
+			goto err;
+next:
+		slot++;
+		path->slots[0]++;
+	}
+	ret = 0;
+err:
+	btrfs_free_path(path);
+	return ret;
+}
+
 int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		   struct btrfs_key *key)
 {
@@ -111,14 +174,8 @@ int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 			    path->slots[0], struct btrfs_root_item);
 
 	refs = btrfs_root_refs(ri);
-	BUG_ON(refs == 0);
-	if (refs == 1) {
-		ret = btrfs_del_item(trans, root, path);
-	} else {
-		btrfs_set_root_refs(ri, refs - 1);
-		WARN_ON(1);
-		mark_buffer_dirty(path->nodes[0]);
-	}
+	BUG_ON(refs != 0);
+	ret = btrfs_del_item(trans, root, path);
 out:
 	btrfs_release_path(root, path);
 	btrfs_free_path(path);
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 321f8852755b..85a2a5e27148 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -85,11 +85,15 @@ struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
 
 	if (root != root->fs_info->tree_root && root->last_trans <
 	    running_trans_id) {
-		radix_tree_tag_set(&root->fs_info->fs_roots_radix,
-				   (unsigned long)root->root_key.objectid,
-				   BTRFS_ROOT_TRANS_TAG);
-		root->commit_root = root->node;
-		get_bh(root->node);
+		if (root->root_item.refs != 0) {
+			radix_tree_tag_set(&root->fs_info->fs_roots_radix,
+					   (unsigned long)root->root_key.objectid,
+					   BTRFS_ROOT_TRANS_TAG);
+			root->commit_root = root->node;
+			get_bh(root->node);
+		} else {
+			WARN_ON(1);
+		}
 	}
 	root->last_trans = running_trans_id;
 	h->transid = running_trans_id;
@@ -208,8 +212,24 @@ struct dirty_root {
 	struct btrfs_key snap_key;
 	struct buffer_head *commit_root;
 	struct btrfs_root *root;
+	int free_on_drop;
 };
 
+int btrfs_add_dead_root(struct btrfs_root *root, struct list_head *dead_list)
+{
+	struct dirty_root *dirty;
+
+	dirty = kmalloc(sizeof(*dirty), GFP_NOFS);
+	if (!dirty)
+		return -ENOMEM;
+	memcpy(&dirty->snap_key, &root->root_key, sizeof(root->root_key));
+	dirty->commit_root = root->node;
+	dirty->root = root;
+	dirty->free_on_drop = 1;
+	list_add(&dirty->list, dead_list);
+	return 0;
+}
+
 static int add_dirty_roots(struct btrfs_trans_handle *trans,
 			   struct radix_tree_root *radix,
 			   struct list_head *list)
@@ -217,9 +237,11 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans,
 	struct dirty_root *dirty;
 	struct btrfs_root *gang[8];
 	struct btrfs_root *root;
+	struct btrfs_root_item tmp_item;
 	int i;
 	int ret;
 	int err = 0;
+	u32 refs;
 
 	while(1) {
 		ret = radix_tree_gang_lookup_tag(radix, (void **)gang, 0,
@@ -246,6 +268,9 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans,
 			dirty->commit_root = root->commit_root;
 			root->commit_root = NULL;
 			dirty->root = root;
+			dirty->free_on_drop = 0;
+			memcpy(&tmp_item, &root->root_item, sizeof(tmp_item));
+
 			root->root_key.offset = root->fs_info->generation;
 			btrfs_set_root_blocknr(&root->root_item,
 					       bh_blocknr(root->node));
@@ -254,7 +279,18 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans,
 						&root->root_item);
 			if (err)
 				break;
-			list_add(&dirty->list, list);
+
+			refs = btrfs_root_refs(&tmp_item);
+			btrfs_set_root_refs(&tmp_item, refs - 1);
+			err = btrfs_update_root(trans, root->fs_info->tree_root,
+						&dirty->snap_key,
+						&tmp_item);
+
+			BUG_ON(err);
+			if (refs == 1)
+				list_add(&dirty->list, list);
+			else
+				kfree(dirty);
 		}
 	}
 	return err;
@@ -270,16 +306,20 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 		mutex_lock(&tree_root->fs_info->fs_mutex);
 		dirty = list_entry(list->next, struct dirty_root, list);
 		list_del_init(&dirty->list);
+
 		trans = btrfs_start_transaction(tree_root, 1);
+printk("deleting root %Lu %u %Lu\n", dirty->snap_key.objectid, dirty->snap_key.flags, dirty->snap_key.offset);
 		ret = btrfs_drop_snapshot(trans, dirty->root,
 					  dirty->commit_root);
 		BUG_ON(ret);
-
 		ret = btrfs_del_root(trans, tree_root, &dirty->snap_key);
 		if (ret)
 			break;
 		ret = btrfs_end_transaction(trans, tree_root);
 		BUG_ON(ret);
+
+		if (dirty->free_on_drop)
+			kfree(dirty->root);
 		kfree(dirty);
 		mutex_unlock(&tree_root->fs_info->fs_mutex);
 		btrfs_btree_balance_dirty(tree_root);
diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h
index 5fb1d322aa10..ebf44f3e1118 100644
--- a/fs/btrfs/transaction.h
+++ b/fs/btrfs/transaction.h
@@ -69,5 +69,6 @@ void btrfs_transaction_flush_work(struct btrfs_root *root);
 void btrfs_transaction_queue_work(struct btrfs_root *root, int delay);
 void btrfs_init_transaction_sys(void);
 void btrfs_exit_transaction_sys(void);
+int btrfs_add_dead_root(struct btrfs_root *root, struct list_head *dead_list);
 
 #endif
-- 
cgit v1.2.3


From ccd467d60e81b48cdbecae93532b66bcdedca91d Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Thu, 28 Jun 2007 15:57:36 -0400
Subject: Btrfs: crash recovery fixes

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.c       | 13 +++++++++++-
 fs/btrfs/ctree.h       | 12 ++++-------
 fs/btrfs/disk-io.c     | 22 ++++++++++++--------
 fs/btrfs/disk-io.h     |  1 +
 fs/btrfs/extent-tree.c | 37 ++++++++++++++++++++++++++++++----
 fs/btrfs/file.c        | 15 ++++++++++----
 fs/btrfs/inode.c       |  6 +++---
 fs/btrfs/transaction.c | 54 +++++++++++++++++++++++++++++++-------------------
 8 files changed, 112 insertions(+), 48 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 606a19b5916d..9eb646529edc 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -75,6 +75,17 @@ static int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root
 	struct btrfs_node *cow_node;
 	int ret;
 
+	WARN_ON(!buffer_uptodate(buf));
+	if (trans->transaction != root->fs_info->running_transaction) {
+		printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid,
+		       root->fs_info->running_transaction->transid);
+		WARN_ON(1);
+	}
+	if (trans->transid != root->fs_info->generation) {
+		printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid,
+		       root->fs_info->generation);
+		WARN_ON(1);
+	}
 	if (btrfs_header_generation(btrfs_buffer_header(buf)) ==
 				    trans->transid) {
 		*cow_ret = buf;
@@ -107,7 +118,7 @@ static int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root
 		btrfs_free_extent(trans, root, bh_blocknr(buf), 1, 1);
 	}
 	btrfs_block_release(root, buf);
-	mark_buffer_dirty(cow);
+	btrfs_mark_buffer_dirty(cow);
 	*cow_ret = cow;
 	return 0;
 }
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 1998f86df08a..0287bd51d87c 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1013,18 +1013,13 @@ static inline void btrfs_memmove(struct btrfs_root *root,
 	memmove(dst, src, nr);
 }
 
-static inline void btrfs_mark_buffer_dirty(struct buffer_head *bh)
-{
-	WARN_ON(!atomic_read(&bh->b_count));
-	mark_buffer_dirty(bh);
-}
-
 /* helper function to cast into the data area of the leaf. */
 #define btrfs_item_ptr(leaf, slot, type) \
 	((type *)(btrfs_leaf_data(leaf) + \
 	btrfs_item_offset((leaf)->items + (slot))))
 
 /* extent-tree.c */
+int btrfs_copy_pinned(struct btrfs_root *root, struct radix_tree_root *copy);
 struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
 							 btrfs_fs_info *info,
 							 u64 blocknr);
@@ -1044,8 +1039,9 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		  struct buffer_head *buf);
 int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, u64 blocknr, u64 num_blocks, int pin);
-int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct
-			       btrfs_root *root);
+int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root,
+			       struct radix_tree_root *unpin_radix);
 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root,
 				u64 blocknr, u64 num_blocks);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 7081729d5b16..d1bf5bc1bc14 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -270,14 +270,6 @@ fail:
 	return NULL;
 }
 
-int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		     struct buffer_head *buf)
-{
-	WARN_ON(atomic_read(&buf->b_count) == 0);
-	mark_buffer_dirty(buf);
-	return 0;
-}
-
 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		     struct buffer_head *buf)
 {
@@ -621,6 +613,20 @@ int close_ctree(struct btrfs_root *root)
 	return 0;
 }
 
+void btrfs_mark_buffer_dirty(struct buffer_head *bh)
+{
+	struct btrfs_root *root = BTRFS_I(bh->b_page->mapping->host)->root;
+	u64 transid = btrfs_header_generation(btrfs_buffer_header(bh));
+	WARN_ON(!atomic_read(&bh->b_count));
+	if (transid != root->fs_info->generation) {
+		printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
+			(unsigned long long)bh->b_blocknr,
+			transid, root->fs_info->generation);
+		WARN_ON(1);
+	}
+	mark_buffer_dirty(bh);
+}
+
 void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf)
 {
 	brelse(buf);
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index c4a695ac44f6..9e2c261b41ae 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -78,4 +78,5 @@ int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
 int btrfs_releasepage(struct page *page, gfp_t flags);
 void btrfs_btree_balance_dirty(struct btrfs_root *root);
 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root);
+void btrfs_mark_buffer_dirty(struct buffer_head *bh);
 #endif
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 01dc30579287..14b93268920e 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -523,6 +523,7 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	}
 	return 0;
 fail:
+	WARN_ON(1);
 	for (i =0; i < faili; i++) {
 		if (leaf) {
 			u64 disk_blocknr;
@@ -572,7 +573,7 @@ static int write_one_cache_group(struct btrfs_trans_handle *trans,
 	bi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
 			    struct btrfs_block_group_item);
 	memcpy(bi, &cache->item, sizeof(*bi));
-	mark_buffer_dirty(path->nodes[0]);
+	btrfs_mark_buffer_dirty(path->nodes[0]);
 	btrfs_release_path(extent_root, path);
 fail:
 	finish_current_insert(trans, extent_root);
@@ -739,8 +740,30 @@ static int try_remove_page(struct address_space *mapping, unsigned long index)
 	return ret;
 }
 
-int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct
-			       btrfs_root *root)
+int btrfs_copy_pinned(struct btrfs_root *root, struct radix_tree_root *copy)
+{
+	unsigned long gang[8];
+	u64 last = 0;
+	struct radix_tree_root *pinned_radix = &root->fs_info->pinned_radix;
+	int ret;
+	int i;
+
+	while(1) {
+		ret = find_first_radix_bit(pinned_radix, gang, last,
+					   ARRAY_SIZE(gang));
+		if (!ret)
+			break;
+		for (i = 0 ; i < ret; i++) {
+			set_radix_bit(copy, gang[i]);
+			last = gang[i] + 1;
+		}
+	}
+	return 0;
+}
+
+int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root,
+			       struct radix_tree_root *unpin_radix)
 {
 	unsigned long gang[8];
 	struct inode *btree_inode = root->fs_info->btree_inode;
@@ -752,7 +775,7 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct
 	struct radix_tree_root *extent_radix = &root->fs_info->extent_map_radix;
 
 	while(1) {
-		ret = find_first_radix_bit(pinned_radix, gang, 0,
+		ret = find_first_radix_bit(unpin_radix, gang, 0,
 					   ARRAY_SIZE(gang));
 		if (!ret)
 			break;
@@ -760,6 +783,7 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct
 			first = gang[0];
 		for (i = 0; i < ret; i++) {
 			clear_radix_bit(pinned_radix, gang[i]);
+			clear_radix_bit(unpin_radix, gang[i]);
 			block_group = btrfs_lookup_block_group(root->fs_info,
 							       gang[i]);
 			if (block_group) {
@@ -1309,6 +1333,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 	if (data) {
 		ret = find_free_extent(trans, root, 0, 0,
 				       search_end, 0, &prealloc_key, 0, 0, 0);
+		BUG_ON(ret);
 		if (ret)
 			return ret;
 		exclude_nr = info->extent_tree_prealloc_nr;
@@ -1319,6 +1344,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 	ret = find_free_extent(trans, root, num_blocks, search_start,
 			       search_end, hint_block, ins,
 			       exclude_start, exclude_nr, data);
+	BUG_ON(ret);
 	if (ret)
 		return ret;
 
@@ -1334,10 +1360,12 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 	if (!data) {
 		exclude_start = ins->objectid;
 		exclude_nr = ins->offset;
+		hint_block = exclude_start + exclude_nr;
 		ret = find_free_extent(trans, root, 0, search_start,
 				       search_end, hint_block,
 				       &prealloc_key, exclude_start,
 				       exclude_nr, 0);
+		BUG_ON(ret);
 		if (ret)
 			return ret;
 	}
@@ -1348,6 +1376,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 	ret = btrfs_insert_item(trans, extent_root, ins, &extent_item,
 				sizeof(extent_item));
 
+	BUG_ON(ret);
 	finish_current_insert(trans, extent_root);
 	pending_ret = del_pending_extents(trans, extent_root);
 	if (ret) {
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index fef7ba1e707f..2456cc3e1cfd 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -127,7 +127,7 @@ static int insert_inline_extent(struct btrfs_root *root, struct inode *inode,
 		     ptr, kaddr + bh_offset(bh),
 		     size);
 	kunmap_atomic(kaddr, KM_USER0);
-	mark_buffer_dirty(path->nodes[0]);
+	btrfs_mark_buffer_dirty(path->nodes[0]);
 fail:
 	btrfs_free_path(path);
 	ret = btrfs_end_transaction(trans, root);
@@ -211,11 +211,13 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans,
 	int found_type;
 	int found_extent;
 	int found_inline;
+	int recow;
 
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
 	while(1) {
+		recow = 0;
 		btrfs_release_path(root, path);
 		ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino,
 					       search_start, -1);
@@ -244,6 +246,10 @@ next_slot:
 		if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY) {
 			goto out;
 		}
+		if (recow) {
+			search_start = key.offset;
+			continue;
+		}
 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
 			extent = btrfs_item_ptr(leaf, slot,
 						struct btrfs_file_extent_item);
@@ -274,6 +280,7 @@ next_slot:
 				nextret = btrfs_next_leaf(root, path);
 				if (nextret)
 					goto out;
+				recow = 1;
 			} else {
 				path->slots[0]++;
 			}
@@ -321,7 +328,7 @@ next_slot:
 				}
 				btrfs_set_file_extent_num_blocks(extent,
 								 new_num);
-				mark_buffer_dirty(path->nodes[0]);
+				btrfs_mark_buffer_dirty(path->nodes[0]);
 			} else {
 				WARN_ON(1);
 			}
@@ -452,6 +459,8 @@ static int prepare_pages(struct btrfs_root *root,
 			err = -ENOMEM;
 			goto failed_release;
 		}
+		cancel_dirty_page(pages[i], PAGE_CACHE_SIZE);
+		wait_on_page_writeback(pages[i]);
 	}
 
 	mutex_lock(&root->fs_info->fs_mutex);
@@ -522,8 +531,6 @@ static int prepare_pages(struct btrfs_root *root,
 	mutex_unlock(&root->fs_info->fs_mutex);
 
 	for (i = 0; i < num_pages; i++) {
-		cancel_dirty_page(pages[i], PAGE_CACHE_SIZE);
-		wait_on_page_writeback(pages[i]);
 		offset = pos & (PAGE_CACHE_SIZE -1);
 		this_write = min((size_t)PAGE_CACHE_SIZE - offset, write_bytes);
 		if (!page_has_buffers(pages[i])) {
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index eba06e7cf414..4fc0367d54f2 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -506,7 +506,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 							 extent_num_blocks);
 				inode->i_blocks -= (orig_num_blocks -
 					extent_num_blocks) << 3;
-				mark_buffer_dirty(path->nodes[0]);
+				btrfs_mark_buffer_dirty(path->nodes[0]);
 			} else {
 				extent_start =
 					btrfs_file_extent_disk_blocknr(fi);
@@ -2020,7 +2020,7 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 	btrfs_set_header_owner(&leaf->header, root->root_key.objectid);
 	memcpy(leaf->header.fsid, root->fs_info->disk_super->fsid,
 	       sizeof(leaf->header.fsid));
-	mark_buffer_dirty(subvol);
+	btrfs_mark_buffer_dirty(subvol);
 
 	inode_item = &root_item.inode;
 	memset(inode_item, 0, sizeof(*inode_item));
@@ -2497,7 +2497,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
 	ptr = btrfs_file_extent_inline_start(ei);
 	btrfs_memcpy(root, path->nodes[0]->b_data,
 		     ptr, symname, name_len);
-	mark_buffer_dirty(path->nodes[0]);
+	btrfs_mark_buffer_dirty(path->nodes[0]);
 	btrfs_free_path(path);
 	inode->i_op = &btrfs_symlink_inode_operations;
 	inode->i_mapping->a_ops = &btrfs_symlink_aops;
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 3b2face593e9..bec38ae8aa11 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -85,6 +85,8 @@ struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
 
 	if (root != root->fs_info->tree_root && root->last_trans <
 	    running_trans_id) {
+		WARN_ON(root == root->fs_info->extent_root);
+		WARN_ON(root->ref_cows != 1);
 		if (root->root_item.refs != 0) {
 			radix_tree_tag_set(&root->fs_info->fs_roots_radix,
 					   (unsigned long)root->root_key.objectid,
@@ -113,10 +115,11 @@ int btrfs_end_transaction(struct btrfs_trans_handle *trans,
 
 	mutex_lock(&root->fs_info->trans_mutex);
 	cur_trans = root->fs_info->running_transaction;
+	WARN_ON(cur_trans != trans->transaction);
 	WARN_ON(cur_trans->num_writers < 1);
+	cur_trans->num_writers--;
 	if (waitqueue_active(&cur_trans->writer_wait))
 		wake_up(&cur_trans->writer_wait);
-	cur_trans->num_writers--;
 	put_transaction(cur_trans);
 	mutex_unlock(&root->fs_info->trans_mutex);
 	memset(trans, 0, sizeof(*trans));
@@ -194,6 +197,7 @@ static int wait_for_commit(struct btrfs_root *root,
 			   struct btrfs_transaction *commit)
 {
 	DEFINE_WAIT(wait);
+	mutex_lock(&root->fs_info->trans_mutex);
 	while(!commit->commit_done) {
 		prepare_to_wait(&commit->commit_wait, &wait,
 				TASK_UNINTERRUPTIBLE);
@@ -203,6 +207,7 @@ static int wait_for_commit(struct btrfs_root *root,
 		schedule();
 		mutex_lock(&root->fs_info->trans_mutex);
 	}
+	mutex_unlock(&root->fs_info->trans_mutex);
 	finish_wait(&commit->commit_wait, &wait);
 	return 0;
 }
@@ -279,7 +284,6 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans,
 						&root->root_item);
 			if (err)
 				break;
-
 			refs = btrfs_root_refs(&tmp_item);
 			btrfs_set_root_refs(&tmp_item, refs - 1);
 			err = btrfs_update_root(trans, root->fs_info->tree_root,
@@ -333,31 +337,53 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	struct btrfs_transaction *cur_trans;
 	struct btrfs_transaction *prev_trans = NULL;
 	struct list_head dirty_fs_roots;
+	struct radix_tree_root pinned_copy;
 	DEFINE_WAIT(wait);
 
+	init_bit_radix(&pinned_copy);
 	INIT_LIST_HEAD(&dirty_fs_roots);
 
 	mutex_lock(&root->fs_info->trans_mutex);
 	if (trans->transaction->in_commit) {
 		cur_trans = trans->transaction;
 		trans->transaction->use_count++;
+		mutex_unlock(&root->fs_info->trans_mutex);
 		btrfs_end_transaction(trans, root);
+
+		mutex_unlock(&root->fs_info->fs_mutex);
 		ret = wait_for_commit(root, cur_trans);
 		BUG_ON(ret);
 		put_transaction(cur_trans);
-		mutex_unlock(&root->fs_info->trans_mutex);
+		mutex_lock(&root->fs_info->fs_mutex);
 		return 0;
 	}
-	cur_trans = trans->transaction;
 	trans->transaction->in_commit = 1;
+	cur_trans = trans->transaction;
+	if (cur_trans->list.prev != &root->fs_info->trans_list) {
+		prev_trans = list_entry(cur_trans->list.prev,
+					struct btrfs_transaction, list);
+		if (!prev_trans->commit_done) {
+			prev_trans->use_count++;
+			mutex_unlock(&root->fs_info->fs_mutex);
+			mutex_unlock(&root->fs_info->trans_mutex);
+
+			wait_for_commit(root, prev_trans);
+			put_transaction(prev_trans);
+
+			mutex_lock(&root->fs_info->fs_mutex);
+			mutex_lock(&root->fs_info->trans_mutex);
+		}
+	}
 	while (trans->transaction->num_writers > 1) {
 		WARN_ON(cur_trans != trans->transaction);
 		prepare_to_wait(&trans->transaction->writer_wait, &wait,
 				TASK_UNINTERRUPTIBLE);
 		if (trans->transaction->num_writers <= 1)
 			break;
+		mutex_unlock(&root->fs_info->fs_mutex);
 		mutex_unlock(&root->fs_info->trans_mutex);
 		schedule();
+		mutex_lock(&root->fs_info->fs_mutex);
 		mutex_lock(&root->fs_info->trans_mutex);
 		finish_wait(&trans->transaction->writer_wait, &wait);
 	}
@@ -372,34 +398,22 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 
 	cur_trans = root->fs_info->running_transaction;
 	root->fs_info->running_transaction = NULL;
-	if (cur_trans->list.prev != &root->fs_info->trans_list) {
-		prev_trans = list_entry(cur_trans->list.prev,
-					struct btrfs_transaction, list);
-		if (prev_trans->commit_done)
-			prev_trans = NULL;
-		else
-			prev_trans->use_count++;
-	}
 	btrfs_set_super_generation(&root->fs_info->super_copy,
 				   cur_trans->transid);
 	btrfs_set_super_root(&root->fs_info->super_copy,
 			     bh_blocknr(root->fs_info->tree_root->node));
 	memcpy(root->fs_info->disk_super, &root->fs_info->super_copy,
 	       sizeof(root->fs_info->super_copy));
+
+	btrfs_copy_pinned(root, &pinned_copy);
+
 	mutex_unlock(&root->fs_info->trans_mutex);
 	mutex_unlock(&root->fs_info->fs_mutex);
 	ret = btrfs_write_and_wait_transaction(trans, root);
-	if (prev_trans) {
-		mutex_lock(&root->fs_info->trans_mutex);
-		wait_for_commit(root, prev_trans);
-		put_transaction(prev_trans);
-		mutex_unlock(&root->fs_info->trans_mutex);
-	}
 	BUG_ON(ret);
 	write_ctree_super(trans, root);
-
 	mutex_lock(&root->fs_info->fs_mutex);
-	btrfs_finish_extent_commit(trans, root);
+	btrfs_finish_extent_commit(trans, root, &pinned_copy);
 	mutex_lock(&root->fs_info->trans_mutex);
 	cur_trans->commit_done = 1;
 	wake_up(&cur_trans->commit_wait);
-- 
cgit v1.2.3


From f2183bde1a918d338337955c8e8ba29bd8f5e7b1 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Fri, 10 Aug 2007 14:42:37 -0400
Subject: Btrfs: Add BH_Defrag to mark buffers that are in need of defragging

This allows the tree walking code to defrag only the newly allocated
buffers, it seems to be a good balance between perfect defragging and the
performance hit of repeatedly reallocating blocks.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.c       | 22 ++++++++++++----------
 fs/btrfs/disk-io.h     |  2 ++
 fs/btrfs/extent-tree.c |  2 ++
 fs/btrfs/tree-defrag.c |  3 ++-
 4 files changed, 18 insertions(+), 11 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index ee1ae00d2827..7cf43da5e78e 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -175,6 +175,7 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 	int end_slot;
 	int i;
 	int err = 0;
+	int parent_level;
 
 	if (trans->transaction != root->fs_info->running_transaction) {
 		printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid,
@@ -188,6 +189,7 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 	}
 	parent_node = btrfs_buffer_node(parent);
 	parent_nritems = btrfs_header_nritems(&parent_node->header);
+	parent_level = btrfs_header_level(&parent_node->header);
 
 	start_slot = 0;
 	end_slot = parent_nritems;
@@ -215,13 +217,16 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 
 		cur_bh = btrfs_find_tree_block(root, blocknr);
 		if (!cur_bh || !buffer_uptodate(cur_bh) ||
-		    buffer_locked(cur_bh)) {
+		    buffer_locked(cur_bh) || !buffer_defrag(cur_bh)) {
 			if (cache_only) {
 				brelse(cur_bh);
 				continue;
 			}
-			brelse(cur_bh);
-			cur_bh = read_tree_block(root, blocknr);
+			if (!cur_bh || !buffer_uptodate(cur_bh) ||
+			    buffer_locked(cur_bh)) {
+				brelse(cur_bh);
+				cur_bh = read_tree_block(root, blocknr);
+			}
 		}
 		if (search_start == 0)
 			search_start = last_block & ~((u64)65535);
@@ -232,6 +237,9 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 		if (err)
 			break;
 		search_start = bh_blocknr(tmp_bh);
+		*last_ret = search_start;
+		if (parent_level == 1)
+			clear_buffer_defrag(tmp_bh);
 		brelse(tmp_bh);
 	}
 	return err;
@@ -811,16 +819,10 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
 			clear_radix_bit(&found, blocknr);
 			if (nread > 32)
 				continue;
-			if (direction > 0 && cluster_start <= blocknr &&
-			    cluster_start + 8 > blocknr) {
-				cluster_start = blocknr;
+			if (close_blocks(cluster_start, blocknr)) {
 				readahead_tree_block(root, blocknr);
 				nread++;
-			} else if (direction < 0 && cluster_start >= blocknr &&
-				   blocknr + 8 > cluster_start) {
 				cluster_start = blocknr;
-				readahead_tree_block(root, blocknr);
-				nread++;
 			}
 		}
 	}
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 9e2c261b41ae..81fd18cbd824 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -25,8 +25,10 @@
 
 enum btrfs_bh_state_bits {
 	BH_Checked = BH_PrivateStart,
+	BH_Defrag,
 };
 BUFFER_FNS(Checked, checked);
+BUFFER_FNS(Defrag, defrag);
 
 static inline struct btrfs_node *btrfs_buffer_node(struct buffer_head *bh)
 {
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 3418bb62b996..b3641234473f 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1015,6 +1015,7 @@ check_failed:
 	ins->objectid = search_start;
 	ins->offset = 0;
 	start_found = 0;
+	path->reada = 1;
 
 	ret = btrfs_search_slot(trans, root, ins, path, 0, 0);
 	if (ret < 0)
@@ -1264,6 +1265,7 @@ struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 	WARN_ON(buffer_dirty(buf));
 	set_buffer_uptodate(buf);
 	set_buffer_checked(buf);
+	set_buffer_defrag(buf);
 	set_radix_bit(&trans->transaction->dirty_pages, buf->b_page->index);
 	return buf;
 }
diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c
index a09064a9a41c..35fd20d24645 100644
--- a/fs/btrfs/tree-defrag.c
+++ b/fs/btrfs/tree-defrag.c
@@ -86,7 +86,7 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 		if (cache_only) {
 			next = btrfs_find_tree_block(root, blocknr);
 			if (!next || !buffer_uptodate(next) ||
-			   buffer_locked(next)) {
+			   buffer_locked(next) || !buffer_defrag(next)) {
 				brelse(next);
 				path->slots[*level]++;
 				continue;
@@ -142,6 +142,7 @@ static int defrag_walk_up(struct btrfs_trans_handle *trans,
 			root->defrag_level = i;
 			return 0;
 		} else {
+			clear_buffer_defrag(path->nodes[*level]);
 			btrfs_block_release(root, path->nodes[*level]);
 			path->nodes[*level] = NULL;
 			*level = i + 1;
-- 
cgit v1.2.3


From a52d9a8033c454cd9b4697cfafb467509fc1693f Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Mon, 27 Aug 2007 16:49:44 -0400
Subject: Btrfs: Extent based page cache code.  This uses an rbtree of extents
 and tests instead of buffer heads.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/Makefile      |    3 +-
 fs/btrfs/btrfs_inode.h |    3 +
 fs/btrfs/ctree.h       |    7 +
 fs/btrfs/disk-io.h     |    1 -
 fs/btrfs/extent_map.c  | 1656 ++++++++++++++++++++++++++++++++++++++++++++++++
 fs/btrfs/extent_map.h  |   89 +++
 fs/btrfs/file.c        |  295 ++++-----
 fs/btrfs/inode.c       |  816 +++++++++---------------
 fs/btrfs/super.c       |    2 +
 9 files changed, 2203 insertions(+), 669 deletions(-)
 create mode 100644 fs/btrfs/extent_map.c
 create mode 100644 fs/btrfs/extent_map.h

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 9321438c5eef..ea7a22870350 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -4,7 +4,8 @@ ifneq ($(KERNELRELEASE),)
 obj-m  := btrfs.o
 btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
 	   hash.o file-item.o inode-item.o inode-map.o disk-io.o \
-	   transaction.o bit-radix.o inode.o file.o tree-defrag.o
+	   transaction.o bit-radix.o inode.o file.o tree-defrag.o \
+	   extent_map.o
 
 #btrfs-y := ctree.o disk-io.o radix-tree.o extent-tree.o print-tree.o \
 #	  root-tree.o dir-item.o hash.o file-item.o inode-item.o \
diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
index f446f161f736..d1d5af471c3c 100644
--- a/fs/btrfs/btrfs_inode.h
+++ b/fs/btrfs/btrfs_inode.h
@@ -19,11 +19,14 @@
 #ifndef __BTRFS_I__
 #define __BTRFS_I__
 
+#include "extent_map.h"
+
 /* in memory btrfs inode */
 struct btrfs_inode {
 	struct btrfs_root *root;
 	struct btrfs_block_group_cache *block_group;
 	struct btrfs_key location;
+	struct extent_map_tree extent_tree;
 	struct inode vfs_inode;
 
 	/*
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index c7f5161271d3..947f061ed118 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1176,7 +1176,14 @@ int btrfs_commit_write(struct file *file, struct page *page,
 		       unsigned from, unsigned to);
 int btrfs_get_block(struct inode *inode, sector_t iblock,
 		    struct buffer_head *result, int create);
+struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
+				    size_t page_offset, u64 start, u64 end,
+				    int create);
+int btrfs_update_inode(struct btrfs_trans_handle *trans,
+			      struct btrfs_root *root,
+			      struct inode *inode);
 /* file.c */
+int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end);
 extern struct file_operations btrfs_file_operations;
 int btrfs_drop_extents(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root, struct inode *inode,
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 81fd18cbd824..5261733b8735 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -77,7 +77,6 @@ int btrfs_insert_dev_radix(struct btrfs_root *root,
 			   u64 num_blocks);
 int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
 			     u64 logical);
-int btrfs_releasepage(struct page *page, gfp_t flags);
 void btrfs_btree_balance_dirty(struct btrfs_root *root);
 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root);
 void btrfs_mark_buffer_dirty(struct buffer_head *bh);
diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c
new file mode 100644
index 000000000000..d378edf0964e
--- /dev/null
+++ b/fs/btrfs/extent_map.c
@@ -0,0 +1,1656 @@
+#include <linux/bitops.h>
+#include <linux/slab.h>
+#include <linux/bio.h>
+#include <linux/mm.h>
+#include <linux/gfp.h>
+#include <linux/pagemap.h>
+#include <linux/page-flags.h>
+#include <linux/module.h>
+#include <linux/spinlock.h>
+#include <linux/blkdev.h>
+#include "extent_map.h"
+
+static struct kmem_cache *extent_map_cache;
+static struct kmem_cache *extent_state_cache;
+
+struct tree_entry {
+	u64 start;
+	u64 end;
+	int in_tree;
+	struct rb_node rb_node;
+};
+
+/* bits for the extent state */
+#define EXTENT_DIRTY 1
+#define EXTENT_WRITEBACK (1 << 1)
+#define EXTENT_UPTODATE (1 << 2)
+#define EXTENT_LOCKED (1 << 3)
+#define EXTENT_NEW (1 << 4)
+#define EXTENT_DELALLOC (1 << 5)
+
+#define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK)
+
+static LIST_HEAD(all_states);
+spinlock_t state_lock = SPIN_LOCK_UNLOCKED;
+
+void __init extent_map_init(void)
+{
+	extent_map_cache = kmem_cache_create("extent_map",
+					    sizeof(struct extent_map), 0,
+					    SLAB_RECLAIM_ACCOUNT |
+					    SLAB_DESTROY_BY_RCU,
+					    NULL);
+	extent_state_cache = kmem_cache_create("extent_state",
+					    sizeof(struct extent_state), 0,
+					    SLAB_RECLAIM_ACCOUNT |
+					    SLAB_DESTROY_BY_RCU,
+					    NULL);
+}
+
+void __exit extent_map_exit(void)
+{
+	while(!list_empty(&all_states)) {
+		struct extent_state *state;
+		struct list_head *cur = all_states.next;
+		state = list_entry(cur, struct extent_state, list);
+		printk("found leaked state %Lu %Lu state %d in_tree %d\n",
+		       state->start, state->end, state->state, state->in_tree);
+		list_del(&state->list);
+		kfree(state);
+	}
+	if (extent_map_cache)
+		kmem_cache_destroy(extent_map_cache);
+	if (extent_state_cache)
+		kmem_cache_destroy(extent_state_cache);
+}
+
+void extent_map_tree_init(struct extent_map_tree *tree,
+			  struct address_space *mapping, gfp_t mask)
+{
+	tree->map.rb_node = NULL;
+	tree->state.rb_node = NULL;
+	rwlock_init(&tree->lock);
+	tree->mapping = mapping;
+}
+EXPORT_SYMBOL(extent_map_tree_init);
+
+struct extent_map *alloc_extent_map(gfp_t mask)
+{
+	struct extent_map *em;
+	em = kmem_cache_alloc(extent_map_cache, mask);
+	if (!em || IS_ERR(em))
+		return em;
+	em->in_tree = 0;
+	atomic_set(&em->refs, 1);
+	return em;
+}
+EXPORT_SYMBOL(alloc_extent_map);
+
+void free_extent_map(struct extent_map *em)
+{
+	if (atomic_dec_and_test(&em->refs)) {
+		WARN_ON(em->in_tree);
+		kmem_cache_free(extent_map_cache, em);
+	}
+}
+EXPORT_SYMBOL(free_extent_map);
+
+
+struct extent_state *alloc_extent_state(gfp_t mask)
+{
+	struct extent_state *state;
+	state = kmem_cache_alloc(extent_state_cache, mask);
+	if (!state || IS_ERR(state))
+		return state;
+	state->state = 0;
+	state->in_tree = 0;
+	atomic_set(&state->refs, 1);
+	init_waitqueue_head(&state->wq);
+	spin_lock_irq(&state_lock);
+	list_add(&state->list, &all_states);
+	spin_unlock_irq(&state_lock);
+	return state;
+}
+EXPORT_SYMBOL(alloc_extent_state);
+
+void free_extent_state(struct extent_state *state)
+{
+	if (atomic_dec_and_test(&state->refs)) {
+		WARN_ON(state->in_tree);
+		spin_lock_irq(&state_lock);
+		list_del_init(&state->list);
+		spin_unlock_irq(&state_lock);
+		kmem_cache_free(extent_state_cache, state);
+	}
+}
+EXPORT_SYMBOL(free_extent_state);
+
+static struct rb_node *tree_insert(struct rb_root *root, u64 offset,
+				   struct rb_node *node)
+{
+	struct rb_node ** p = &root->rb_node;
+	struct rb_node * parent = NULL;
+	struct tree_entry *entry;
+
+	while(*p) {
+		parent = *p;
+		entry = rb_entry(parent, struct tree_entry, rb_node);
+
+		if (offset < entry->start)
+			p = &(*p)->rb_left;
+		else if (offset > entry->end)
+			p = &(*p)->rb_right;
+		else
+			return parent;
+	}
+
+	entry = rb_entry(node, struct tree_entry, rb_node);
+	entry->in_tree = 1;
+	rb_link_node(node, parent, p);
+	rb_insert_color(node, root);
+	return NULL;
+}
+
+static struct rb_node *__tree_search(struct rb_root *root, u64 offset,
+				   struct rb_node **prev_ret)
+{
+	struct rb_node * n = root->rb_node;
+	struct rb_node *prev = NULL;
+	struct tree_entry *entry;
+	struct tree_entry *prev_entry = NULL;
+
+	while(n) {
+		entry = rb_entry(n, struct tree_entry, rb_node);
+		prev = n;
+		prev_entry = entry;
+
+		if (offset < entry->start)
+			n = n->rb_left;
+		else if (offset > entry->end)
+			n = n->rb_right;
+		else
+			return n;
+	}
+	if (!prev_ret)
+		return NULL;
+	while(prev && offset > prev_entry->end) {
+		prev = rb_next(prev);
+		prev_entry = rb_entry(prev, struct tree_entry, rb_node);
+	}
+	*prev_ret = prev;
+	return NULL;
+}
+
+static inline struct rb_node *tree_search(struct rb_root *root, u64 offset)
+{
+	struct rb_node *prev;
+	struct rb_node *ret;
+	ret = __tree_search(root, offset, &prev);
+	if (!ret)
+		return prev;
+	return ret;
+}
+
+static int tree_delete(struct rb_root *root, u64 offset)
+{
+	struct rb_node *node;
+	struct tree_entry *entry;
+
+	node = __tree_search(root, offset, NULL);
+	if (!node)
+		return -ENOENT;
+	entry = rb_entry(node, struct tree_entry, rb_node);
+	entry->in_tree = 0;
+	rb_erase(node, root);
+	return 0;
+}
+
+/*
+ * add_extent_mapping tries a simple backward merge with existing
+ * mappings.  The extent_map struct passed in will be inserted into
+ * the tree directly (no copies made, just a reference taken).
+ */
+int add_extent_mapping(struct extent_map_tree *tree,
+		       struct extent_map *em)
+{
+	int ret = 0;
+	struct extent_map *prev = NULL;
+	struct rb_node *rb;
+
+	write_lock_irq(&tree->lock);
+	rb = tree_insert(&tree->map, em->end, &em->rb_node);
+	if (rb) {
+		prev = rb_entry(rb, struct extent_map, rb_node);
+		printk("found extent map %Lu %Lu on insert of %Lu %Lu\n", prev->start, prev->end, em->start, em->end);
+		ret = -EEXIST;
+		goto out;
+	}
+	atomic_inc(&em->refs);
+	if (em->start != 0) {
+		rb = rb_prev(&em->rb_node);
+		if (rb)
+			prev = rb_entry(rb, struct extent_map, rb_node);
+		if (prev && prev->end + 1 == em->start &&
+		    ((em->block_start == 0 && prev->block_start == 0) ||
+			     (em->block_start == prev->block_end + 1))) {
+			em->start = prev->start;
+			em->block_start = prev->block_start;
+			rb_erase(&prev->rb_node, &tree->map);
+			prev->in_tree = 0;
+			free_extent_map(prev);
+		}
+	 }
+out:
+	write_unlock_irq(&tree->lock);
+	return ret;
+}
+EXPORT_SYMBOL(add_extent_mapping);
+
+/*
+ * lookup_extent_mapping returns the first extent_map struct in the
+ * tree that intersects the [start, end] (inclusive) range.  There may
+ * be additional objects in the tree that intersect, so check the object
+ * returned carefully to make sure you don't need additional lookups.
+ */
+struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree,
+					 u64 start, u64 end)
+{
+	struct extent_map *em;
+	struct rb_node *rb_node;
+
+	read_lock_irq(&tree->lock);
+	rb_node = tree_search(&tree->map, start);
+	if (!rb_node) {
+		em = NULL;
+		goto out;
+	}
+	if (IS_ERR(rb_node)) {
+		em = ERR_PTR(PTR_ERR(rb_node));
+		goto out;
+	}
+	em = rb_entry(rb_node, struct extent_map, rb_node);
+	if (em->end < start || em->start > end) {
+		em = NULL;
+		goto out;
+	}
+	atomic_inc(&em->refs);
+out:
+	read_unlock_irq(&tree->lock);
+	return em;
+}
+EXPORT_SYMBOL(lookup_extent_mapping);
+
+/*
+ * removes an extent_map struct from the tree.  No reference counts are
+ * dropped, and no checks are done to  see if the range is in use
+ */
+int remove_extent_mapping(struct extent_map_tree *tree, struct extent_map *em)
+{
+	int ret;
+
+	write_lock_irq(&tree->lock);
+	ret = tree_delete(&tree->map, em->end);
+	write_unlock_irq(&tree->lock);
+	return ret;
+}
+EXPORT_SYMBOL(remove_extent_mapping);
+
+/*
+ * utility function to look for merge candidates inside a given range.
+ * Any extents with matching state are merged together into a single
+ * extent in the tree.  Extents with EXTENT_IO in their state field
+ * are not merged because the end_io handlers need to be able to do
+ * operations on them without sleeping (or doing allocations/splits).
+ *
+ * This should be called with the tree lock held.
+ */
+static int merge_state(struct extent_map_tree *tree,
+		       struct extent_state *state)
+{
+	struct extent_state *other;
+	struct rb_node *other_node;
+
+	if (state->state & EXTENT_IOBITS)
+		return 0;
+
+	other_node = rb_prev(&state->rb_node);
+	if (other_node) {
+		other = rb_entry(other_node, struct extent_state, rb_node);
+		if (other->end == state->start - 1 &&
+		    other->state == state->state) {
+			state->start = other->start;
+			other->in_tree = 0;
+			rb_erase(&other->rb_node, &tree->state);
+			free_extent_state(other);
+		}
+	}
+	other_node = rb_next(&state->rb_node);
+	if (other_node) {
+		other = rb_entry(other_node, struct extent_state, rb_node);
+		if (other->start == state->end + 1 &&
+		    other->state == state->state) {
+			other->start = state->start;
+			state->in_tree = 0;
+			rb_erase(&state->rb_node, &tree->state);
+			free_extent_state(state);
+		}
+	}
+	return 0;
+}
+
+/*
+ * insert an extent_state struct into the tree.  'bits' are set on the
+ * struct before it is inserted.
+ *
+ * This may return -EEXIST if the extent is already there, in which case the
+ * state struct is freed.
+ *
+ * The tree lock is not taken internally.  This is a utility function and
+ * probably isn't what you want to call (see set/clear_extent_bit).
+ */
+static int insert_state(struct extent_map_tree *tree,
+			struct extent_state *state, u64 start, u64 end,
+			int bits)
+{
+	struct rb_node *node;
+
+	if (end < start) {
+		printk("end < start %Lu %Lu\n", end, start);
+		WARN_ON(1);
+	}
+	state->state |= bits;
+	state->start = start;
+	state->end = end;
+	if ((end & 4095) == 0) {
+		printk("insert state %Lu %Lu strange end\n", start, end);
+		WARN_ON(1);
+	}
+	node = tree_insert(&tree->state, end, &state->rb_node);
+	if (node) {
+		struct extent_state *found;
+		found = rb_entry(node, struct extent_state, rb_node);
+printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, start, end);
+		free_extent_state(state);
+		return -EEXIST;
+	}
+	merge_state(tree, state);
+	return 0;
+}
+
+/*
+ * split a given extent state struct in two, inserting the preallocated
+ * struct 'prealloc' as the newly created second half.  'split' indicates an
+ * offset inside 'orig' where it should be split.
+ *
+ * Before calling,
+ * the tree has 'orig' at [orig->start, orig->end].  After calling, there
+ * are two extent state structs in the tree:
+ * prealloc: [orig->start, split - 1]
+ * orig: [ split, orig->end ]
+ *
+ * The tree locks are not taken by this function. They need to be held
+ * by the caller.
+ */
+static int split_state(struct extent_map_tree *tree, struct extent_state *orig,
+		       struct extent_state *prealloc, u64 split)
+{
+	struct rb_node *node;
+	prealloc->start = orig->start;
+	prealloc->end = split - 1;
+	prealloc->state = orig->state;
+	orig->start = split;
+	if ((prealloc->end & 4095) == 0) {
+		printk("insert state %Lu %Lu strange end\n", prealloc->start,
+		       prealloc->end);
+		WARN_ON(1);
+	}
+	node = tree_insert(&tree->state, prealloc->end, &prealloc->rb_node);
+	if (node) {
+		struct extent_state *found;
+		found = rb_entry(node, struct extent_state, rb_node);
+printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, prealloc->start, prealloc->end);
+		free_extent_state(prealloc);
+		return -EEXIST;
+	}
+	return 0;
+}
+
+/*
+ * utility function to clear some bits in an extent state struct.
+ * it will optionally wake up any one waiting on this state (wake == 1), or
+ * forcibly remove the state from the tree (delete == 1).
+ *
+ * If no bits are set on the state struct after clearing things, the
+ * struct is freed and removed from the tree
+ */
+static int clear_state_bit(struct extent_map_tree *tree,
+			    struct extent_state *state, int bits, int wake,
+			    int delete)
+{
+	int ret = state->state & bits;
+	state->state &= ~bits;
+	if (wake)
+		wake_up(&state->wq);
+	if (delete || state->state == 0) {
+		if (state->in_tree) {
+			rb_erase(&state->rb_node, &tree->state);
+			state->in_tree = 0;
+			free_extent_state(state);
+		} else {
+			WARN_ON(1);
+		}
+	} else {
+		merge_state(tree, state);
+	}
+	return ret;
+}
+
+/*
+ * clear some bits on a range in the tree.  This may require splitting
+ * or inserting elements in the tree, so the gfp mask is used to
+ * indicate which allocations or sleeping are allowed.
+ *
+ * pass 'wake' == 1 to kick any sleepers, and 'delete' == 1 to remove
+ * the given range from the tree regardless of state (ie for truncate).
+ *
+ * the range [start, end] is inclusive.
+ *
+ * This takes the tree lock, and returns < 0 on error, > 0 if any of the
+ * bits were already set, or zero if none of the bits were already set.
+ */
+int clear_extent_bit(struct extent_map_tree *tree, u64 start, u64 end,
+		     int bits, int wake, int delete, gfp_t mask)
+{
+	struct extent_state *state;
+	struct extent_state *prealloc = NULL;
+	struct rb_node *node;
+	int err;
+	int set = 0;
+
+again:
+	if (!prealloc && (mask & __GFP_WAIT)) {
+		prealloc = alloc_extent_state(mask);
+		if (!prealloc)
+			return -ENOMEM;
+	}
+
+	write_lock_irq(&tree->lock);
+	/*
+	 * this search will find the extents that end after
+	 * our range starts
+	 */
+	node = tree_search(&tree->state, start);
+	if (!node)
+		goto out;
+	state = rb_entry(node, struct extent_state, rb_node);
+	if (state->start > end)
+		goto out;
+	WARN_ON(state->end < start);
+
+	/*
+	 *     | ---- desired range ---- |
+	 *  | state | or
+	 *  | ------------- state -------------- |
+	 *
+	 * We need to split the extent we found, and may flip
+	 * bits on second half.
+	 *
+	 * If the extent we found extends past our range, we
+	 * just split and search again.  It'll get split again
+	 * the next time though.
+	 *
+	 * If the extent we found is inside our range, we clear
+	 * the desired bit on it.
+	 */
+
+	if (state->start < start) {
+		err = split_state(tree, state, prealloc, start);
+		BUG_ON(err == -EEXIST);
+		prealloc = NULL;
+		if (err)
+			goto out;
+		if (state->end <= end) {
+			start = state->end + 1;
+			set |= clear_state_bit(tree, state, bits,
+					wake, delete);
+		} else {
+			start = state->start;
+		}
+		goto search_again;
+	}
+	/*
+	 * | ---- desired range ---- |
+	 *                        | state |
+	 * We need to split the extent, and clear the bit
+	 * on the first half
+	 */
+	if (state->start <= end && state->end > end) {
+		err = split_state(tree, state, prealloc, end + 1);
+		BUG_ON(err == -EEXIST);
+
+		if (wake)
+			wake_up(&state->wq);
+		set |= clear_state_bit(tree, prealloc, bits,
+				       wake, delete);
+		prealloc = NULL;
+		goto out;
+	}
+
+	start = state->end + 1;
+	set |= clear_state_bit(tree, state, bits, wake, delete);
+	goto search_again;
+
+out:
+	write_unlock_irq(&tree->lock);
+	if (prealloc)
+		free_extent_state(prealloc);
+
+	return set;
+
+search_again:
+	if (start >= end)
+		goto out;
+	write_unlock_irq(&tree->lock);
+	if (mask & __GFP_WAIT)
+		cond_resched();
+	goto again;
+}
+EXPORT_SYMBOL(clear_extent_bit);
+
+static int wait_on_state(struct extent_map_tree *tree,
+			 struct extent_state *state)
+{
+	DEFINE_WAIT(wait);
+	prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE);
+	read_unlock_irq(&tree->lock);
+	schedule();
+	read_lock_irq(&tree->lock);
+	finish_wait(&state->wq, &wait);
+	return 0;
+}
+
+/*
+ * waits for one or more bits to clear on a range in the state tree.
+ * The range [start, end] is inclusive.
+ * The tree lock is taken by this function
+ */
+int wait_extent_bit(struct extent_map_tree *tree, u64 start, u64 end, int bits)
+{
+	struct extent_state *state;
+	struct rb_node *node;
+
+	read_lock_irq(&tree->lock);
+again:
+	while (1) {
+		/*
+		 * this search will find all the extents that end after
+		 * our range starts
+		 */
+		node = tree_search(&tree->state, start);
+		if (!node)
+			break;
+
+		state = rb_entry(node, struct extent_state, rb_node);
+
+		if (state->start > end)
+			goto out;
+
+		if (state->state & bits) {
+			start = state->start;
+			atomic_inc(&state->refs);
+			wait_on_state(tree, state);
+			free_extent_state(state);
+			goto again;
+		}
+		start = state->end + 1;
+
+		if (start > end)
+			break;
+
+		if (need_resched()) {
+			read_unlock_irq(&tree->lock);
+			cond_resched();
+			read_lock_irq(&tree->lock);
+		}
+	}
+out:
+	read_unlock_irq(&tree->lock);
+	return 0;
+}
+EXPORT_SYMBOL(wait_extent_bit);
+
+/*
+ * set some bits on a range in the tree.  This may require allocations
+ * or sleeping, so the gfp mask is used to indicate what is allowed.
+ *
+ * If 'exclusive' == 1, this will fail with -EEXIST if some part of the
+ * range already has the desired bits set.  The start of the existing
+ * range is returned in failed_start in this case.
+ *
+ * [start, end] is inclusive
+ * This takes the tree lock.
+ */
+int set_extent_bit(struct extent_map_tree *tree, u64 start, u64 end, int bits,
+		   int exclusive, u64 *failed_start, gfp_t mask)
+{
+	struct extent_state *state;
+	struct extent_state *prealloc = NULL;
+	struct rb_node *node;
+	int err = 0;
+	int set;
+	u64 last_start;
+	u64 last_end;
+again:
+	if (!prealloc && (mask & __GFP_WAIT)) {
+		prealloc = alloc_extent_state(mask);
+		if (!prealloc)
+			return -ENOMEM;
+	}
+
+	write_lock_irq(&tree->lock);
+	/*
+	 * this search will find all the extents that end after
+	 * our range starts.
+	 */
+	node = tree_search(&tree->state, start);
+	if (!node) {
+		err = insert_state(tree, prealloc, start, end, bits);
+		prealloc = NULL;
+		BUG_ON(err == -EEXIST);
+		goto out;
+	}
+
+	state = rb_entry(node, struct extent_state, rb_node);
+	last_start = state->start;
+	last_end = state->end;
+
+	/*
+	 * | ---- desired range ---- |
+	 * | state |
+	 *
+	 * Just lock what we found and keep going
+	 */
+	if (state->start == start && state->end <= end) {
+		set = state->state & bits;
+		if (set && exclusive) {
+			*failed_start = state->start;
+			err = -EEXIST;
+			goto out;
+		}
+		state->state |= bits;
+		start = state->end + 1;
+		merge_state(tree, state);
+		goto search_again;
+	}
+
+	/*
+	 *     | ---- desired range ---- |
+	 * | state |
+	 *   or
+	 * | ------------- state -------------- |
+	 *
+	 * We need to split the extent we found, and may flip bits on
+	 * second half.
+	 *
+	 * If the extent we found extends past our
+	 * range, we just split and search again.  It'll get split
+	 * again the next time though.
+	 *
+	 * If the extent we found is inside our range, we set the
+	 * desired bit on it.
+	 */
+	if (state->start < start) {
+		set = state->state & bits;
+		if (exclusive && set) {
+			*failed_start = start;
+			err = -EEXIST;
+			goto out;
+		}
+		err = split_state(tree, state, prealloc, start);
+		BUG_ON(err == -EEXIST);
+		prealloc = NULL;
+		if (err)
+			goto out;
+		if (state->end <= end) {
+			state->state |= bits;
+			start = state->end + 1;
+			merge_state(tree, state);
+		} else {
+			start = state->start;
+		}
+		goto search_again;
+	}
+	/*
+	 * | ---- desired range ---- |
+	 *                        | state |
+	 * We need to split the extent, and set the bit
+	 * on the first half
+	 */
+	if (state->start <= end && state->end > end) {
+		set = state->state & bits;
+		if (exclusive && set) {
+			*failed_start = start;
+			err = -EEXIST;
+			goto out;
+		}
+		err = split_state(tree, state, prealloc, end + 1);
+		BUG_ON(err == -EEXIST);
+
+		prealloc->state |= bits;
+		merge_state(tree, prealloc);
+		prealloc = NULL;
+		goto out;
+	}
+
+	/*
+	 * | ---- desired range ---- |
+	 *     | state | or               | state |
+	 *
+	 * There's a hole, we need to insert something in it and
+	 * ignore the extent we found.
+	 */
+	if (state->start > start) {
+		u64 this_end;
+		if (end < last_start)
+			this_end = end;
+		else
+			this_end = last_start -1;
+		err = insert_state(tree, prealloc, start, this_end,
+				   bits);
+		prealloc = NULL;
+		BUG_ON(err == -EEXIST);
+		if (err)
+			goto out;
+		start = this_end + 1;
+		goto search_again;
+	}
+	goto search_again;
+
+out:
+	write_unlock_irq(&tree->lock);
+	if (prealloc)
+		free_extent_state(prealloc);
+
+	return err;
+
+search_again:
+	if (start > end)
+		goto out;
+	write_unlock_irq(&tree->lock);
+	if (mask & __GFP_WAIT)
+		cond_resched();
+	goto again;
+}
+EXPORT_SYMBOL(set_extent_bit);
+
+/* wrappers around set/clear extent bit */
+int set_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end,
+		     gfp_t mask)
+{
+	return set_extent_bit(tree, start, end, EXTENT_DIRTY, 0, NULL,
+			      mask);
+}
+EXPORT_SYMBOL(set_extent_dirty);
+
+int clear_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end,
+		       gfp_t mask)
+{
+	return clear_extent_bit(tree, start, end, EXTENT_DIRTY, 0, 0, mask);
+}
+EXPORT_SYMBOL(clear_extent_dirty);
+
+int set_extent_new(struct extent_map_tree *tree, u64 start, u64 end,
+		     gfp_t mask)
+{
+	return set_extent_bit(tree, start, end, EXTENT_NEW, 0, NULL,
+			      mask);
+}
+EXPORT_SYMBOL(set_extent_new);
+
+int clear_extent_new(struct extent_map_tree *tree, u64 start, u64 end,
+		       gfp_t mask)
+{
+	return clear_extent_bit(tree, start, end, EXTENT_NEW, 0, 0, mask);
+}
+EXPORT_SYMBOL(clear_extent_new);
+
+int set_extent_uptodate(struct extent_map_tree *tree, u64 start, u64 end,
+			gfp_t mask)
+{
+	return set_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, NULL,
+			      mask);
+}
+EXPORT_SYMBOL(set_extent_uptodate);
+
+int clear_extent_uptodate(struct extent_map_tree *tree, u64 start, u64 end,
+			  gfp_t mask)
+{
+	return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0, mask);
+}
+EXPORT_SYMBOL(clear_extent_uptodate);
+
+int set_extent_writeback(struct extent_map_tree *tree, u64 start, u64 end,
+			 gfp_t mask)
+{
+	return set_extent_bit(tree, start, end, EXTENT_WRITEBACK,
+			      0, NULL, mask);
+}
+EXPORT_SYMBOL(set_extent_writeback);
+
+int clear_extent_writeback(struct extent_map_tree *tree, u64 start, u64 end,
+			   gfp_t mask)
+{
+	return clear_extent_bit(tree, start, end, EXTENT_WRITEBACK, 1, 0, mask);
+}
+EXPORT_SYMBOL(clear_extent_writeback);
+
+int wait_on_extent_writeback(struct extent_map_tree *tree, u64 start, u64 end)
+{
+	return wait_extent_bit(tree, start, end, EXTENT_WRITEBACK);
+}
+EXPORT_SYMBOL(wait_on_extent_writeback);
+
+/*
+ * locks a range in ascending order, waiting for any locked regions
+ * it hits on the way.  [start,end] are inclusive, and this will sleep.
+ */
+int lock_extent(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask)
+{
+	int err;
+	u64 failed_start;
+	while (1) {
+		err = set_extent_bit(tree, start, end, EXTENT_LOCKED, 1,
+				     &failed_start, mask);
+		if (err == -EEXIST && (mask & __GFP_WAIT)) {
+			wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED);
+			start = failed_start;
+		} else {
+			break;
+		}
+		WARN_ON(start > end);
+	}
+	return err;
+}
+EXPORT_SYMBOL(lock_extent);
+
+int unlock_extent(struct extent_map_tree *tree, u64 start, u64 end,
+		  gfp_t mask)
+{
+	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, mask);
+}
+EXPORT_SYMBOL(unlock_extent);
+
+/*
+ * helper function to set pages and extents in the tree dirty
+ */
+int set_range_dirty(struct extent_map_tree *tree, u64 start, u64 end)
+{
+	unsigned long index = start >> PAGE_CACHE_SHIFT;
+	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
+	struct page *page;
+
+	while (index <= end_index) {
+		page = find_get_page(tree->mapping, index);
+		BUG_ON(!page);
+		__set_page_dirty_nobuffers(page);
+		page_cache_release(page);
+		index++;
+	}
+	set_extent_dirty(tree, start, end, GFP_NOFS);
+	return 0;
+}
+EXPORT_SYMBOL(set_range_dirty);
+
+/*
+ * helper function to set both pages and extents in the tree writeback
+ */
+int set_range_writeback(struct extent_map_tree *tree, u64 start, u64 end)
+{
+	unsigned long index = start >> PAGE_CACHE_SHIFT;
+	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
+	struct page *page;
+
+	while (index <= end_index) {
+		page = find_get_page(tree->mapping, index);
+		BUG_ON(!page);
+		set_page_writeback(page);
+		page_cache_release(page);
+		index++;
+	}
+	set_extent_writeback(tree, start, end, GFP_NOFS);
+	return 0;
+}
+EXPORT_SYMBOL(set_range_writeback);
+
+/*
+ * helper function to lock both pages and extents in the tree.
+ * pages must be locked first.
+ */
+int lock_range(struct extent_map_tree *tree, u64 start, u64 end)
+{
+	unsigned long index = start >> PAGE_CACHE_SHIFT;
+	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
+	struct page *page;
+	int err;
+
+	while (index <= end_index) {
+		page = grab_cache_page(tree->mapping, index);
+		if (!page) {
+			err = -ENOMEM;
+			goto failed;
+		}
+		if (IS_ERR(page)) {
+			err = PTR_ERR(page);
+			goto failed;
+		}
+		index++;
+	}
+	lock_extent(tree, start, end, GFP_NOFS);
+	return 0;
+
+failed:
+	/*
+	 * we failed above in getting the page at 'index', so we undo here
+	 * up to but not including the page at 'index'
+	 */
+	end_index = index;
+	index = start >> PAGE_CACHE_SHIFT;
+	while (index < end_index) {
+		page = find_get_page(tree->mapping, index);
+		unlock_page(page);
+		page_cache_release(page);
+		index++;
+	}
+	return err;
+}
+EXPORT_SYMBOL(lock_range);
+
+/*
+ * helper function to unlock both pages and extents in the tree.
+ */
+int unlock_range(struct extent_map_tree *tree, u64 start, u64 end)
+{
+	unsigned long index = start >> PAGE_CACHE_SHIFT;
+	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
+	struct page *page;
+
+	while (index <= end_index) {
+		page = find_get_page(tree->mapping, index);
+		unlock_page(page);
+		page_cache_release(page);
+		index++;
+	}
+	unlock_extent(tree, start, end, GFP_NOFS);
+	return 0;
+}
+EXPORT_SYMBOL(unlock_range);
+
+/*
+ * searches a range in the state tree for a given mask.
+ * If 'filled' == 1, this returns 1 only if ever extent in the tree
+ * has the bits set.  Otherwise, 1 is returned if any bit in the
+ * range is found set.
+ */
+static int test_range_bit(struct extent_map_tree *tree, u64 start, u64 end,
+			  int bits, int filled)
+{
+	struct extent_state *state = NULL;
+	struct rb_node *node;
+	int bitset = 0;
+
+	read_lock_irq(&tree->lock);
+	node = tree_search(&tree->state, start);
+	while (node && start <= end) {
+		state = rb_entry(node, struct extent_state, rb_node);
+		if (state->start > end)
+			break;
+
+		if (filled && state->start > start) {
+			bitset = 0;
+			break;
+		}
+		if (state->state & bits) {
+			bitset = 1;
+			if (!filled)
+				break;
+		} else if (filled) {
+			bitset = 0;
+			break;
+		}
+		start = state->end + 1;
+		if (start > end)
+			break;
+		node = rb_next(node);
+	}
+	read_unlock_irq(&tree->lock);
+	return bitset;
+}
+
+/*
+ * helper function to set a given page up to date if all the
+ * extents in the tree for that page are up to date
+ */
+static int check_page_uptodate(struct extent_map_tree *tree,
+			       struct page *page)
+{
+	u64 start = page->index << PAGE_CACHE_SHIFT;
+	u64 end = start + PAGE_CACHE_SIZE - 1;
+	if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1))
+		SetPageUptodate(page);
+	return 0;
+}
+
+/*
+ * helper function to unlock a page if all the extents in the tree
+ * for that page are unlocked
+ */
+static int check_page_locked(struct extent_map_tree *tree,
+			     struct page *page)
+{
+	u64 start = page->index << PAGE_CACHE_SHIFT;
+	u64 end = start + PAGE_CACHE_SIZE - 1;
+	if (!test_range_bit(tree, start, end, EXTENT_LOCKED, 0))
+		unlock_page(page);
+	return 0;
+}
+
+/*
+ * helper function to end page writeback if all the extents
+ * in the tree for that page are done with writeback
+ */
+static int check_page_writeback(struct extent_map_tree *tree,
+			     struct page *page)
+{
+	u64 start = page->index << PAGE_CACHE_SHIFT;
+	u64 end = start + PAGE_CACHE_SIZE - 1;
+	if (!test_range_bit(tree, start, end, EXTENT_WRITEBACK, 0))
+		end_page_writeback(page);
+	return 0;
+}
+
+/* lots and lots of room for performance fixes in the end_bio funcs */
+
+/*
+ * after a writepage IO is done, we need to:
+ * clear the uptodate bits on error
+ * clear the writeback bits in the extent tree for this IO
+ * end_page_writeback if the page has no more pending IO
+ *
+ * Scheduling is not allowed, so the extent state tree is expected
+ * to have one and only one object corresponding to this IO.
+ */
+static int end_bio_extent_writepage(struct bio *bio,
+				   unsigned int bytes_done, int err)
+{
+	const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
+	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
+	struct extent_map_tree *tree = bio->bi_private;
+	u64 start;
+	u64 end;
+	int whole_page;
+
+	if (bio->bi_size)
+		return 1;
+
+	do {
+		struct page *page = bvec->bv_page;
+		start = (page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset;
+		end = start + bvec->bv_len - 1;
+
+		if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE)
+			whole_page = 1;
+		else
+			whole_page = 0;
+
+		if (--bvec >= bio->bi_io_vec)
+			prefetchw(&bvec->bv_page->flags);
+
+		if (!uptodate) {
+			clear_extent_uptodate(tree, start, end, GFP_ATOMIC);
+			ClearPageUptodate(page);
+			SetPageError(page);
+		}
+		clear_extent_writeback(tree, start, end, GFP_ATOMIC);
+
+		if (whole_page)
+			end_page_writeback(page);
+		else
+			check_page_writeback(tree, page);
+	} while (bvec >= bio->bi_io_vec);
+
+	bio_put(bio);
+	return 0;
+}
+
+/*
+ * after a readpage IO is done, we need to:
+ * clear the uptodate bits on error
+ * set the uptodate bits if things worked
+ * set the page up to date if all extents in the tree are uptodate
+ * clear the lock bit in the extent tree
+ * unlock the page if there are no other extents locked for it
+ *
+ * Scheduling is not allowed, so the extent state tree is expected
+ * to have one and only one object corresponding to this IO.
+ */
+static int end_bio_extent_readpage(struct bio *bio,
+				   unsigned int bytes_done, int err)
+{
+	const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
+	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
+	struct extent_map_tree *tree = bio->bi_private;
+	u64 start;
+	u64 end;
+	int whole_page;
+
+	if (bio->bi_size)
+		return 1;
+
+	do {
+		struct page *page = bvec->bv_page;
+		start = (page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset;
+		end = start + bvec->bv_len - 1;
+
+		if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE)
+			whole_page = 1;
+		else
+			whole_page = 0;
+
+		if (--bvec >= bio->bi_io_vec)
+			prefetchw(&bvec->bv_page->flags);
+
+		if (uptodate) {
+			set_extent_uptodate(tree, start, end, GFP_ATOMIC);
+			if (whole_page)
+				SetPageUptodate(page);
+			else
+				check_page_uptodate(tree, page);
+		} else {
+			ClearPageUptodate(page);
+			SetPageError(page);
+		}
+
+		unlock_extent(tree, start, end, GFP_ATOMIC);
+
+		if (whole_page)
+			unlock_page(page);
+		else
+			check_page_locked(tree, page);
+	} while (bvec >= bio->bi_io_vec);
+
+	bio_put(bio);
+	return 0;
+}
+
+/*
+ * IO done from prepare_write is pretty simple, we just unlock
+ * the structs in the extent tree when done, and set the uptodate bits
+ * as appropriate.
+ */
+static int end_bio_extent_preparewrite(struct bio *bio,
+				       unsigned int bytes_done, int err)
+{
+	const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
+	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
+	struct extent_map_tree *tree = bio->bi_private;
+	u64 start;
+	u64 end;
+
+	if (bio->bi_size)
+		return 1;
+
+	do {
+		struct page *page = bvec->bv_page;
+		start = (page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset;
+		end = start + bvec->bv_len - 1;
+
+		if (--bvec >= bio->bi_io_vec)
+			prefetchw(&bvec->bv_page->flags);
+
+		if (uptodate) {
+			set_extent_uptodate(tree, start, end, GFP_ATOMIC);
+		} else {
+			ClearPageUptodate(page);
+			SetPageError(page);
+		}
+
+		unlock_extent(tree, start, end, GFP_ATOMIC);
+
+	} while (bvec >= bio->bi_io_vec);
+
+	bio_put(bio);
+	return 0;
+}
+
+static int submit_extent_page(int rw, struct extent_map_tree *tree,
+			      struct page *page, sector_t sector,
+			      size_t size, unsigned long offset,
+			      struct block_device *bdev,
+			      bio_end_io_t end_io_func)
+{
+	struct bio *bio;
+	int ret = 0;
+
+	bio = bio_alloc(GFP_NOIO, 1);
+
+	bio->bi_sector = sector;
+	bio->bi_bdev = bdev;
+	bio->bi_io_vec[0].bv_page = page;
+	bio->bi_io_vec[0].bv_len = size;
+	bio->bi_io_vec[0].bv_offset = offset;
+
+	bio->bi_vcnt = 1;
+	bio->bi_idx = 0;
+	bio->bi_size = size;
+
+	bio->bi_end_io = end_io_func;
+	bio->bi_private = tree;
+
+	bio_get(bio);
+	submit_bio(rw, bio);
+
+	if (bio_flagged(bio, BIO_EOPNOTSUPP))
+		ret = -EOPNOTSUPP;
+
+	bio_put(bio);
+	return ret;
+}
+
+/*
+ * basic readpage implementation.  Locked extent state structs are inserted
+ * into the tree that are removed when the IO is done (by the end_io
+ * handlers)
+ */
+int extent_read_full_page(struct extent_map_tree *tree, struct page *page,
+			  get_extent_t *get_extent)
+{
+	struct inode *inode = page->mapping->host;
+	u64 start = page->index << PAGE_CACHE_SHIFT;
+	u64 page_end = start + PAGE_CACHE_SIZE - 1;
+	u64 end;
+	u64 cur = start;
+	u64 extent_offset;
+	u64 last_byte = i_size_read(inode);
+	u64 block_start;
+	u64 cur_end;
+	sector_t sector;
+	struct extent_map *em;
+	struct block_device *bdev;
+	int ret;
+	int nr = 0;
+	size_t page_offset = 0;
+	size_t iosize;
+	size_t blocksize = inode->i_sb->s_blocksize;
+
+	if (!PagePrivate(page)) {
+		SetPagePrivate(page);
+		set_page_private(page, 1);
+		page_cache_get(page);
+	}
+
+	end = page_end;
+	lock_extent(tree, start, end, GFP_NOFS);
+
+	while (cur <= end) {
+		if (cur >= last_byte) {
+			iosize = PAGE_CACHE_SIZE - page_offset;
+			zero_user_page(page, page_offset, iosize, KM_USER0);
+			set_extent_uptodate(tree, cur, cur + iosize - 1,
+					    GFP_NOFS);
+			unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
+			break;
+		}
+		em = get_extent(inode, page, page_offset, cur, end, 0);
+		if (IS_ERR(em) || !em) {
+			SetPageError(page);
+			unlock_extent(tree, cur, end, GFP_NOFS);
+			break;
+		}
+
+		extent_offset = cur - em->start;
+		BUG_ON(em->end < cur);
+		BUG_ON(end < cur);
+
+		iosize = min(em->end - cur, end - cur) + 1;
+		cur_end = min(em->end, end);
+		iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1);
+		sector = (em->block_start + extent_offset) >> 9;
+		bdev = em->bdev;
+		block_start = em->block_start;
+		free_extent_map(em);
+		em = NULL;
+
+		/* we've found a hole, just zero and go on */
+		if (block_start == 0) {
+			zero_user_page(page, page_offset, iosize, KM_USER0);
+			set_extent_uptodate(tree, cur, cur + iosize - 1,
+					    GFP_NOFS);
+			unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
+			cur = cur + iosize;
+			page_offset += iosize;
+			continue;
+		}
+		/* the get_extent function already copied into the page */
+		if (test_range_bit(tree, cur, cur_end, EXTENT_UPTODATE, 1)) {
+			unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
+			cur = cur + iosize;
+			page_offset += iosize;
+			continue;
+		}
+
+		ret = submit_extent_page(READ, tree, page,
+					 sector, iosize, page_offset, bdev,
+					 end_bio_extent_readpage);
+		if (ret)
+			SetPageError(page);
+		cur = cur + iosize;
+		page_offset += iosize;
+		nr++;
+	}
+	if (!nr) {
+		if (!PageError(page))
+			SetPageUptodate(page);
+		unlock_page(page);
+	}
+	return 0;
+}
+EXPORT_SYMBOL(extent_read_full_page);
+
+/*
+ * the writepage semantics are similar to regular writepage.  extent
+ * records are inserted to lock ranges in the tree, and as dirty areas
+ * are found, they are marked writeback.  Then the lock bits are removed
+ * and the end_io handler clears the writeback ranges
+ */
+int extent_write_full_page(struct extent_map_tree *tree, struct page *page,
+			  get_extent_t *get_extent,
+			  struct writeback_control *wbc)
+{
+	struct inode *inode = page->mapping->host;
+	u64 start = page->index << PAGE_CACHE_SHIFT;
+	u64 page_end = start + PAGE_CACHE_SIZE - 1;
+	u64 end;
+	u64 cur = start;
+	u64 extent_offset;
+	u64 last_byte = i_size_read(inode);
+	u64 block_start;
+	sector_t sector;
+	struct extent_map *em;
+	struct block_device *bdev;
+	int ret;
+	int nr = 0;
+	size_t page_offset = 0;
+	size_t iosize;
+	size_t blocksize;
+	loff_t i_size = i_size_read(inode);
+	unsigned long end_index = i_size >> PAGE_CACHE_SHIFT;
+
+	if (page->index > end_index) {
+		clear_extent_dirty(tree, start, page_end, GFP_NOFS);
+		unlock_page(page);
+		return 0;
+	}
+
+	if (page->index == end_index) {
+		size_t offset = i_size & (PAGE_CACHE_SIZE - 1);
+		zero_user_page(page, offset,
+			       PAGE_CACHE_SIZE - offset, KM_USER0);
+	}
+
+	if (!PagePrivate(page)) {
+		SetPagePrivate(page);
+		set_page_private(page, 1);
+		page_cache_get(page);
+	}
+
+	end = page_end;
+	lock_extent(tree, start, page_end, GFP_NOFS);
+
+	if (last_byte <= start) {
+		clear_extent_dirty(tree, start, page_end, GFP_NOFS);
+		goto done;
+	}
+
+	set_extent_uptodate(tree, start, page_end, GFP_NOFS);
+	blocksize = inode->i_sb->s_blocksize;
+
+	while (cur <= end) {
+		if (cur >= last_byte) {
+			clear_extent_dirty(tree, cur, page_end, GFP_NOFS);
+			break;
+		}
+		em = get_extent(inode, page, page_offset, cur, end, 1);
+		if (IS_ERR(em) || !em) {
+			SetPageError(page);
+			break;
+		}
+
+		extent_offset = cur - em->start;
+		BUG_ON(em->end < cur);
+		BUG_ON(end < cur);
+		iosize = min(em->end - cur, end - cur) + 1;
+		iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1);
+		sector = (em->block_start + extent_offset) >> 9;
+		bdev = em->bdev;
+		block_start = em->block_start;
+		free_extent_map(em);
+		em = NULL;
+
+		if (block_start == 0 || block_start == EXTENT_MAP_INLINE) {
+			clear_extent_dirty(tree, cur,
+					   cur + iosize - 1, GFP_NOFS);
+			cur = cur + iosize;
+			page_offset += iosize;
+			continue;
+		}
+
+		/* leave this out until we have a page_mkwrite call */
+		if (0 && !test_range_bit(tree, cur, cur + iosize - 1,
+				   EXTENT_DIRTY, 0)) {
+			cur = cur + iosize;
+			page_offset += iosize;
+			continue;
+		}
+		clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS);
+		set_range_writeback(tree, cur, cur + iosize - 1);
+		ret = submit_extent_page(WRITE, tree, page,
+					 sector, iosize, page_offset, bdev,
+					 end_bio_extent_writepage);
+		if (ret)
+			SetPageError(page);
+		cur = cur + iosize;
+		page_offset += iosize;
+		nr++;
+	}
+done:
+	WARN_ON(test_range_bit(tree, start, page_end, EXTENT_DIRTY, 0));
+	unlock_extent(tree, start, page_end, GFP_NOFS);
+	unlock_page(page);
+	return 0;
+}
+EXPORT_SYMBOL(extent_write_full_page);
+
+/*
+ * basic invalidatepage code, this waits on any locked or writeback
+ * ranges corresponding to the page, and then deletes any extent state
+ * records from the tree
+ */
+int extent_invalidatepage(struct extent_map_tree *tree,
+			  struct page *page, unsigned long offset)
+{
+	u64 start = (page->index << PAGE_CACHE_SHIFT);
+	u64 end = start + PAGE_CACHE_SIZE - 1;
+	size_t blocksize = page->mapping->host->i_sb->s_blocksize;
+
+	start += (offset + blocksize -1) & ~(blocksize - 1);
+	if (start > end)
+		return 0;
+
+	lock_extent(tree, start, end, GFP_NOFS);
+	wait_on_extent_writeback(tree, start, end);
+	clear_extent_bit(tree, start, end, EXTENT_LOCKED | EXTENT_DIRTY,
+			 1, 1, GFP_NOFS);
+	return 0;
+}
+EXPORT_SYMBOL(extent_invalidatepage);
+
+/*
+ * simple commit_write call, set_range_dirty is used to mark both
+ * the pages and the extent records as dirty
+ */
+int extent_commit_write(struct extent_map_tree *tree,
+			struct inode *inode, struct page *page,
+			unsigned from, unsigned to)
+{
+	loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;
+
+	if (!PagePrivate(page)) {
+		SetPagePrivate(page);
+		set_page_private(page, 1);
+		page_cache_get(page);
+	}
+
+	set_page_dirty(page);
+
+	if (pos > inode->i_size) {
+		i_size_write(inode, pos);
+		mark_inode_dirty(inode);
+	}
+	return 0;
+}
+EXPORT_SYMBOL(extent_commit_write);
+
+int extent_prepare_write(struct extent_map_tree *tree,
+			 struct inode *inode, struct page *page,
+			 unsigned from, unsigned to, get_extent_t *get_extent)
+{
+	u64 page_start = page->index << PAGE_CACHE_SHIFT;
+	u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
+	u64 block_start;
+	u64 orig_block_start;
+	u64 block_end;
+	u64 cur_end;
+	struct extent_map *em;
+	unsigned blocksize = 1 << inode->i_blkbits;
+	size_t page_offset = 0;
+	size_t block_off_start;
+	size_t block_off_end;
+	int err = 0;
+	int iocount = 0;
+	int ret = 0;
+	int isnew;
+
+	if (!PagePrivate(page)) {
+		SetPagePrivate(page);
+		set_page_private(page, 1);
+		page_cache_get(page);
+	}
+	block_start = (page_start + from) & ~((u64)blocksize - 1);
+	block_end = (page_start + to - 1) | (blocksize - 1);
+	orig_block_start = block_start;
+
+	lock_extent(tree, page_start, page_end, GFP_NOFS);
+	while(block_start <= block_end) {
+		em = get_extent(inode, page, page_offset, block_start,
+				block_end, 1);
+		if (IS_ERR(em) || !em) {
+			goto err;
+		}
+		cur_end = min(block_end, em->end);
+		block_off_start = block_start & (PAGE_CACHE_SIZE - 1);
+		block_off_end = block_off_start + blocksize;
+		isnew = clear_extent_new(tree, block_start, cur_end, GFP_NOFS);
+
+		if (!PageUptodate(page) && isnew &&
+		    (block_off_end > to || block_off_start < from)) {
+			void *kaddr;
+
+			kaddr = kmap_atomic(page, KM_USER0);
+			if (block_off_end > to)
+				memset(kaddr + to, 0, block_off_end - to);
+			if (block_off_start < from)
+				memset(kaddr + block_off_start, 0,
+				       from - block_off_start);
+			flush_dcache_page(page);
+			kunmap_atomic(kaddr, KM_USER0);
+		}
+		if (!isnew && !PageUptodate(page) &&
+		    (block_off_end > to || block_off_start < from) &&
+		    !test_range_bit(tree, block_start, cur_end,
+				    EXTENT_UPTODATE, 1)) {
+			u64 sector;
+			u64 extent_offset = block_start - em->start;
+			size_t iosize;
+			sector = (em->block_start + extent_offset) >> 9;
+			iosize = (cur_end - block_start + blocksize - 1) &
+				~((u64)blocksize - 1);
+			/*
+			 * we've already got the extent locked, but we
+			 * need to split the state such that our end_bio
+			 * handler can clear the lock.
+			 */
+			set_extent_bit(tree, block_start,
+				       block_start + iosize - 1,
+				       EXTENT_LOCKED, 0, NULL, GFP_NOFS);
+			ret = submit_extent_page(READ, tree, page,
+					 sector, iosize, page_offset, em->bdev,
+					 end_bio_extent_preparewrite);
+			iocount++;
+			block_start = block_start + iosize;
+		} else {
+			set_extent_uptodate(tree, block_start, cur_end,
+					    GFP_NOFS);
+			unlock_extent(tree, block_start, cur_end, GFP_NOFS);
+			block_start = cur_end + 1;
+		}
+		page_offset = block_start & (PAGE_CACHE_SIZE - 1);
+		free_extent_map(em);
+	}
+	if (iocount) {
+		wait_extent_bit(tree, orig_block_start,
+				block_end, EXTENT_LOCKED);
+	}
+	check_page_uptodate(tree, page);
+err:
+	/* FIXME, zero out newly allocated blocks on error */
+	return err;
+}
+EXPORT_SYMBOL(extent_prepare_write);
+
+/*
+ * a helper for releasepage.  As long as there are no locked extents
+ * in the range corresponding to the page, both state records and extent
+ * map records are removed
+ */
+int try_release_extent_mapping(struct extent_map_tree *tree, struct page *page)
+{
+	struct extent_map *em;
+	u64 start = page->index << PAGE_CACHE_SHIFT;
+	u64 end = start + PAGE_CACHE_SIZE - 1;
+	u64 orig_start = start;
+
+	while (start <= end) {
+		em = lookup_extent_mapping(tree, start, end);
+		if (!em || IS_ERR(em))
+			break;
+		if (test_range_bit(tree, em->start, em->end,
+				   EXTENT_LOCKED, 0)) {
+			free_extent_map(em);
+			start = em->end + 1;
+printk("range still locked %Lu %Lu\n", em->start, em->end);
+			break;
+		}
+		remove_extent_mapping(tree, em);
+		start = em->end + 1;
+		/* once for the rb tree */
+		free_extent_map(em);
+		/* once for us */
+		free_extent_map(em);
+	}
+	WARN_ON(test_range_bit(tree, orig_start, end, EXTENT_WRITEBACK, 0));
+	clear_extent_bit(tree, orig_start, end, EXTENT_UPTODATE,
+			 1, 1, GFP_NOFS);
+	return 1;
+}
+EXPORT_SYMBOL(try_release_extent_mapping);
+
diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h
new file mode 100644
index 000000000000..108944aab4bd
--- /dev/null
+++ b/fs/btrfs/extent_map.h
@@ -0,0 +1,89 @@
+#ifndef __EXTENTMAP__
+#define __EXTENTMAP__
+
+#include <linux/rbtree.h>
+
+#define EXTENT_MAP_INLINE (u64)-2
+#define EXTENT_MAP_DELALLOC (u64)-1
+
+struct extent_map_tree {
+	struct rb_root map;
+	struct rb_root state;
+	struct address_space *mapping;
+	rwlock_t lock;
+};
+
+/* note, this must start with the same fields as fs/extent_map.c:tree_entry */
+struct extent_map {
+	u64 start;
+	u64 end; /* inclusive */
+	int in_tree;
+	struct rb_node rb_node;
+	/* block_start and block_end are in bytes */
+	u64 block_start;
+	u64 block_end; /* inclusive */
+	struct block_device *bdev;
+	atomic_t refs;
+};
+
+/* note, this must start with the same fields as fs/extent_map.c:tree_entry */
+struct extent_state {
+	u64 start;
+	u64 end; /* inclusive */
+	int in_tree;
+	struct rb_node rb_node;
+	wait_queue_head_t wq;
+	atomic_t refs;
+	unsigned long state;
+	struct list_head list;
+};
+
+struct extent_buffer {
+	u64 start;
+	u64 end; /* inclusive */
+	char *addr;
+	struct page *pages[];
+};
+
+typedef struct extent_map *(get_extent_t)(struct inode *inode,
+					  struct page *page,
+					  size_t page_offset,
+					  u64 start, u64 end,
+					  int create);
+
+void extent_map_tree_init(struct extent_map_tree *tree,
+			  struct address_space *mapping, gfp_t mask);
+struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree,
+					 u64 start, u64 end);
+int add_extent_mapping(struct extent_map_tree *tree,
+		       struct extent_map *em);
+int remove_extent_mapping(struct extent_map_tree *tree, struct extent_map *em);
+int try_release_extent_mapping(struct extent_map_tree *tree, struct page *page);
+int lock_extent(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask);
+int unlock_extent(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask);
+struct extent_map *alloc_extent_map(gfp_t mask);
+void free_extent_map(struct extent_map *em);
+int extent_read_full_page(struct extent_map_tree *tree, struct page *page,
+			  get_extent_t *get_extent);
+void __init extent_map_init(void);
+void __exit extent_map_exit(void);
+int extent_clean_all_trees(struct extent_map_tree *tree);
+int set_extent_uptodate(struct extent_map_tree *tree, u64 start, u64 end,
+			gfp_t mask);
+int set_extent_new(struct extent_map_tree *tree, u64 start, u64 end,
+		   gfp_t mask);
+int set_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end,
+		     gfp_t mask);
+int extent_invalidatepage(struct extent_map_tree *tree,
+			  struct page *page, unsigned long offset);
+int extent_write_full_page(struct extent_map_tree *tree, struct page *page,
+			  get_extent_t *get_extent,
+			  struct writeback_control *wbc);
+int extent_prepare_write(struct extent_map_tree *tree,
+			 struct inode *inode, struct page *page,
+			 unsigned from, unsigned to, get_extent_t *get_extent);
+int extent_commit_write(struct extent_map_tree *tree,
+			struct inode *inode, struct page *page,
+			unsigned from, unsigned to);
+int set_range_dirty(struct extent_map_tree *tree, u64 start, u64 end);
+#endif
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 6933ab11a5cd..71a481894ab6 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -81,14 +81,14 @@ static void btrfs_drop_pages(struct page **pages, size_t num_pages)
 	}
 }
 
-static int insert_inline_extent(struct btrfs_root *root, struct inode *inode,
+static int insert_inline_extent(struct btrfs_trans_handle *trans,
+				struct btrfs_root *root, struct inode *inode,
 				u64 offset, ssize_t size,
-				struct buffer_head *bh)
+				struct page *page, size_t page_offset)
 {
 	struct btrfs_key key;
 	struct btrfs_path *path;
 	char *ptr, *kaddr;
-	struct btrfs_trans_handle *trans;
 	struct btrfs_file_extent_item *ei;
 	u32 datasize;
 	int err = 0;
@@ -98,8 +98,6 @@ static int insert_inline_extent(struct btrfs_root *root, struct inode *inode,
 	if (!path)
 		return -ENOMEM;
 
-	mutex_lock(&root->fs_info->fs_mutex);
-	trans = btrfs_start_transaction(root, 1);
 	btrfs_set_trans_block_group(trans, inode);
 
 	key.objectid = inode->i_ino;
@@ -122,18 +120,13 @@ static int insert_inline_extent(struct btrfs_root *root, struct inode *inode,
 				   BTRFS_FILE_EXTENT_INLINE);
 	ptr = btrfs_file_extent_inline_start(ei);
 
-	kaddr = kmap_atomic(bh->b_page, KM_USER0);
+	kaddr = kmap_atomic(page, KM_USER0);
 	btrfs_memcpy(root, path->nodes[0]->b_data,
-		     ptr, kaddr + bh_offset(bh),
-		     size);
+		     ptr, kaddr + page_offset, size);
 	kunmap_atomic(kaddr, KM_USER0);
 	btrfs_mark_buffer_dirty(path->nodes[0]);
 fail:
 	btrfs_free_path(path);
-	ret = btrfs_end_transaction(trans, root);
-	if (ret && !err)
-		err = ret;
-	mutex_unlock(&root->fs_info->fs_mutex);
 	return err;
 }
 
@@ -145,45 +138,143 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 				   loff_t pos,
 				   size_t write_bytes)
 {
-	int i;
-	int offset;
 	int err = 0;
-	int ret;
-	int this_write;
+	int i;
 	struct inode *inode = file->f_path.dentry->d_inode;
-	struct buffer_head *bh;
+	struct extent_map *em;
+	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+	struct btrfs_key ins;
+	u64 hint_block;
+	u64 num_blocks;
+	u64 start_pos;
+	u64 end_of_last_block;
+	u64 end_pos = pos + write_bytes;
+	loff_t isize = i_size_read(inode);
 
-	for (i = 0; i < num_pages; i++) {
-		offset = pos & (PAGE_CACHE_SIZE -1);
-		this_write = min((size_t)PAGE_CACHE_SIZE - offset, write_bytes);
+	em = alloc_extent_map(GFP_NOFS);
+	if (!em)
+		return -ENOMEM;
 
-		/* FIXME, one block at a time */
-		bh = page_buffers(pages[i]);
+	em->bdev = inode->i_sb->s_bdev;
 
-		if (buffer_mapped(bh) && bh->b_blocknr == 0) {
-			ret = insert_inline_extent(root, inode,
-					pages[i]->index << PAGE_CACHE_SHIFT,
-					offset + this_write, bh);
-			if (ret) {
-				err = ret;
-				goto failed;
-			}
-		}
+	start_pos = pos & ~((u64)root->blocksize - 1);
+	num_blocks = (write_bytes + pos - start_pos + root->blocksize - 1) >>
+			inode->i_blkbits;
 
-		ret = btrfs_commit_write(file, pages[i], offset,
-					 offset + this_write);
-		pos += this_write;
-		if (ret) {
-			err = ret;
+	end_of_last_block = start_pos + (num_blocks << inode->i_blkbits) - 1;
+	mutex_lock(&root->fs_info->fs_mutex);
+	trans = btrfs_start_transaction(root, 1);
+	if (!trans) {
+		err = -ENOMEM;
+		goto out_unlock;
+	}
+	btrfs_set_trans_block_group(trans, inode);
+	inode->i_blocks += num_blocks << 3;
+	hint_block = 0;
+
+	if ((end_of_last_block & 4095) == 0) {
+		printk("strange end of last %Lu %lu %Lu\n", start_pos, write_bytes, end_of_last_block);
+	}
+	set_extent_uptodate(em_tree, start_pos, end_of_last_block, GFP_NOFS);
+
+	/* FIXME...EIEIO, ENOSPC and more */
+
+	/* step one, delete the existing extents in this range */
+	/* FIXME blocksize != pagesize */
+	if (start_pos < inode->i_size) {
+		err = btrfs_drop_extents(trans, root, inode,
+			 start_pos, (pos + write_bytes + root->blocksize -1) &
+			 ~((u64)root->blocksize - 1), &hint_block);
+		if (err)
+			goto failed;
+	}
+
+	/* insert any holes we need to create */
+	if (inode->i_size < start_pos) {
+		u64 last_pos_in_file;
+		u64 hole_size;
+		u64 mask = root->blocksize - 1;
+		last_pos_in_file = (isize + mask) & ~mask;
+		hole_size = (start_pos - last_pos_in_file + mask) & ~mask;
+		hole_size >>= inode->i_blkbits;
+		if (last_pos_in_file < start_pos) {
+			err = btrfs_insert_file_extent(trans, root,
+						       inode->i_ino,
+						       last_pos_in_file,
+						       0, 0, hole_size);
+		}
+		if (err)
 			goto failed;
+	}
+
+	/*
+	 * either allocate an extent for the new bytes or setup the key
+	 * to show we are doing inline data in the extent
+	 */
+	if (isize >= PAGE_CACHE_SIZE || pos + write_bytes < inode->i_size ||
+	    pos + write_bytes - start_pos > BTRFS_MAX_INLINE_DATA_SIZE(root)) {
+		err = btrfs_alloc_extent(trans, root, inode->i_ino,
+					 num_blocks, 0, hint_block, (u64)-1,
+					 &ins, 1);
+		BUG_ON(err);
+		err = btrfs_insert_file_extent(trans, root, inode->i_ino,
+				       start_pos, ins.objectid, ins.offset,
+				       ins.offset);
+		BUG_ON(err);
+		em->start = start_pos;
+		em->end = end_of_last_block;
+		em->block_start = ins.objectid << inode->i_blkbits;
+		em->block_end = em->block_start +
+			(ins.offset << inode->i_blkbits) - 1;
+		set_extent_dirty(em_tree, start_pos, end_of_last_block,
+				 GFP_NOFS);
+		err = add_extent_mapping(em_tree, em);
+		for (i = 0; i < num_pages; i++) {
+			struct page *p = pages[i];
+			SetPageUptodate(p);
+			__set_page_dirty_nobuffers(p);
 		}
-		WARN_ON(this_write > write_bytes);
-		write_bytes -= this_write;
+	} else {
+		struct page *p = pages[0];
+		err = insert_inline_extent(trans, root, inode, start_pos,
+					   end_pos - start_pos, p, 0);
+		BUG_ON(err);
+		em->start = start_pos;
+		em->end = end_pos;
+		em->block_start = EXTENT_MAP_INLINE;
+		em->block_end = EXTENT_MAP_INLINE;
+		add_extent_mapping(em_tree, em);
+	}
+	if (end_pos > isize) {
+		i_size_write(inode, end_pos);
+		btrfs_update_inode(trans, root, inode);
 	}
 failed:
+	err = btrfs_end_transaction(trans, root);
+out_unlock:
+	mutex_unlock(&root->fs_info->fs_mutex);
+	free_extent_map(em);
 	return err;
 }
 
+int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end)
+{
+	struct extent_map *em;
+	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+
+	while(1) {
+		em = lookup_extent_mapping(em_tree, start, end);
+		if (!em)
+			break;
+		remove_extent_mapping(em_tree, em);
+		/* once for us */
+		free_extent_map(em);
+		/* once for the tree*/
+		free_extent_map(em);
+	}
+	return 0;
+}
+
 /*
  * this is very complex, but the basic idea is to drop all extents
  * in the range start - end.  hint_block is filled in with a block number
@@ -213,6 +304,8 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans,
 	int found_inline;
 	int recow;
 
+	btrfs_drop_extent_cache(inode, start, end - 1);
+
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
@@ -434,18 +527,9 @@ static int prepare_pages(struct btrfs_root *root,
 	int i;
 	unsigned long index = pos >> PAGE_CACHE_SHIFT;
 	struct inode *inode = file->f_path.dentry->d_inode;
-	int offset;
 	int err = 0;
-	int this_write;
-	struct buffer_head *bh;
-	struct buffer_head *head;
-	loff_t isize = i_size_read(inode);
-	struct btrfs_trans_handle *trans;
-	u64 hint_block;
 	u64 num_blocks;
-	u64 alloc_extent_start;
 	u64 start_pos;
-	struct btrfs_key ins;
 
 	start_pos = pos & ~((u64)PAGE_CACHE_SIZE - 1);
 	num_blocks = (write_bytes + pos - start_pos + root->blocksize - 1) >>
@@ -457,119 +541,17 @@ static int prepare_pages(struct btrfs_root *root,
 		pages[i] = grab_cache_page(inode->i_mapping, index + i);
 		if (!pages[i]) {
 			err = -ENOMEM;
-			goto failed_release;
+			BUG_ON(1);
 		}
 		cancel_dirty_page(pages[i], PAGE_CACHE_SIZE);
 		wait_on_page_writeback(pages[i]);
-	}
-
-	mutex_lock(&root->fs_info->fs_mutex);
-	trans = btrfs_start_transaction(root, 1);
-	if (!trans) {
-		err = -ENOMEM;
-		mutex_unlock(&root->fs_info->fs_mutex);
-		goto out_unlock;
-	}
-	btrfs_set_trans_block_group(trans, inode);
-	/* FIXME blocksize != 4096 */
-	inode->i_blocks += num_blocks << 3;
-	hint_block = 0;
-
-	/* FIXME...EIEIO, ENOSPC and more */
-
-	/* step one, delete the existing extents in this range */
-	/* FIXME blocksize != pagesize */
-	if (start_pos < inode->i_size) {
-		err = btrfs_drop_extents(trans, root, inode,
-			 start_pos, (pos + write_bytes + root->blocksize -1) &
-			 ~((u64)root->blocksize - 1), &hint_block);
-		if (err)
-			goto failed_release;
-	}
-
-	/* insert any holes we need to create */
-	if (inode->i_size < start_pos) {
-		u64 last_pos_in_file;
-		u64 hole_size;
-		u64 mask = root->blocksize - 1;
-		last_pos_in_file = (isize + mask) & ~mask;
-		hole_size = (start_pos - last_pos_in_file + mask) & ~mask;
-		hole_size >>= inode->i_blkbits;
-		if (last_pos_in_file < start_pos) {
-			err = btrfs_insert_file_extent(trans, root,
-						       inode->i_ino,
-						       last_pos_in_file,
-						       0, 0, hole_size);
-		}
-		if (err)
-			goto failed_release;
-	}
-
-	/*
-	 * either allocate an extent for the new bytes or setup the key
-	 * to show we are doing inline data in the extent
-	 */
-	if (isize >= PAGE_CACHE_SIZE || pos + write_bytes < inode->i_size ||
-	    pos + write_bytes - start_pos > BTRFS_MAX_INLINE_DATA_SIZE(root)) {
-		err = btrfs_alloc_extent(trans, root, inode->i_ino,
-					 num_blocks, 0, hint_block, (u64)-1,
-					 &ins, 1);
-		if (err)
-			goto failed_truncate;
-		err = btrfs_insert_file_extent(trans, root, inode->i_ino,
-				       start_pos, ins.objectid, ins.offset,
-				       ins.offset);
-		if (err)
-			goto failed_truncate;
-	} else {
-		ins.offset = 0;
-		ins.objectid = 0;
-	}
-	BUG_ON(err);
-	alloc_extent_start = ins.objectid;
-	err = btrfs_end_transaction(trans, root);
-	mutex_unlock(&root->fs_info->fs_mutex);
-
-	for (i = 0; i < num_pages; i++) {
-		offset = pos & (PAGE_CACHE_SIZE -1);
-		this_write = min((size_t)PAGE_CACHE_SIZE - offset, write_bytes);
-		if (!page_has_buffers(pages[i])) {
-			create_empty_buffers(pages[i],
-					     root->fs_info->sb->s_blocksize,
-					     (1 << BH_Uptodate));
+		if (!PagePrivate(pages[i])) {
+			SetPagePrivate(pages[i]);
+			set_page_private(pages[i], 1);
+			page_cache_get(pages[i]);
 		}
-		head = page_buffers(pages[i]);
-		bh = head;
-		do {
-			err = btrfs_map_bh_to_logical(root, bh,
-						      alloc_extent_start);
-			BUG_ON(err);
-			if (err)
-				goto failed_truncate;
-			bh = bh->b_this_page;
-			if (alloc_extent_start)
-				alloc_extent_start++;
-		} while (bh != head);
-		pos += this_write;
-		WARN_ON(this_write > write_bytes);
-		write_bytes -= this_write;
 	}
 	return 0;
-
-failed_release:
-	btrfs_drop_pages(pages, num_pages);
-	return err;
-
-failed_truncate:
-	btrfs_drop_pages(pages, num_pages);
-	if (pos > isize)
-		vmtruncate(inode, isize);
-	return err;
-
-out_unlock:
-	mutex_unlock(&root->fs_info->fs_mutex);
-	goto failed_release;
-
 }
 
 static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
@@ -685,7 +667,6 @@ out:
 		page_cache_release(pinned[1]);
 	*ppos = pos;
 	current->backing_dev_info = NULL;
-	mark_inode_dirty(inode);
 	return num_written ? num_written : err;
 }
 
@@ -714,8 +695,8 @@ static int btrfs_sync_file(struct file *file,
 	mutex_unlock(&root->fs_info->trans_mutex);
 
 	/*
- 	 * ok we haven't committed the transaction yet, lets do a commit
- 	 */
+	 * ok we haven't committed the transaction yet, lets do a commit
+	 */
 	trans = btrfs_start_transaction(root, 1);
 	if (!trans) {
 		ret = -ENOMEM;
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 504d37dfa6c9..fd4cb65fcf1f 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -121,6 +121,8 @@ void btrfs_read_locked_inode(struct inode *inode)
 	switch (inode->i_mode & S_IFMT) {
 	case S_IFREG:
 		inode->i_mapping->a_ops = &btrfs_aops;
+		extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
+				     inode->i_mapping, GFP_NOFS);
 		inode->i_fop = &btrfs_file_operations;
 		inode->i_op = &btrfs_file_inode_operations;
 		break;
@@ -169,7 +171,7 @@ static void fill_inode_item(struct btrfs_inode_item *item,
 				    BTRFS_I(inode)->block_group->key.objectid);
 }
 
-static int btrfs_update_inode(struct btrfs_trans_handle *trans,
+int btrfs_update_inode(struct btrfs_trans_handle *trans,
 			      struct btrfs_root *root,
 			      struct inode *inode)
 {
@@ -400,6 +402,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 	int found_extent;
 	int del_item;
 
+	btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
 	path = btrfs_alloc_path();
 	path->reada = -1;
 	BUG_ON(!path);
@@ -511,6 +514,50 @@ error:
 	return ret;
 }
 
+static int btrfs_cow_one_page(struct btrfs_trans_handle *trans,
+			      struct inode *inode, struct page *page,
+			      size_t zero_start)
+{
+	char *kaddr;
+	int ret = 0;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	u64 alloc_hint = 0;
+	u64 page_start = page->index << PAGE_CACHE_SHIFT;
+	struct btrfs_key ins;
+
+	if (!PagePrivate(page)) {
+		SetPagePrivate(page);
+		set_page_private(page, 1);
+		page_cache_get(page);
+	}
+
+	btrfs_set_trans_block_group(trans, inode);
+
+	ret = btrfs_drop_extents(trans, root, inode,
+				 page_start, page_start + PAGE_CACHE_SIZE,
+				 &alloc_hint);
+	if (ret)
+		goto out;
+	ret = btrfs_alloc_extent(trans, root, inode->i_ino, 1, 0,
+				 alloc_hint, (u64)-1, &ins, 1);
+	if (ret)
+		goto out;
+	ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
+				       page_start, ins.objectid, 1, 1);
+	if (ret)
+		goto out;
+	SetPageChecked(page);
+	kaddr = kmap(page);
+	if (zero_start != PAGE_CACHE_SIZE) {
+		memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
+		flush_dcache_page(page);
+	}
+	kunmap(page);
+
+out:
+	return ret;
+}
+
 /*
  * taken from block_truncate_page, but does cow as it zeros out
  * any bytes left in the last page in the file.
@@ -518,16 +565,14 @@ error:
 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
 {
 	struct inode *inode = mapping->host;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_trans_handle *trans;
 	unsigned blocksize = 1 << inode->i_blkbits;
 	pgoff_t index = from >> PAGE_CACHE_SHIFT;
 	unsigned offset = from & (PAGE_CACHE_SIZE-1);
 	struct page *page;
-	char *kaddr;
 	int ret = 0;
-	struct btrfs_root *root = BTRFS_I(inode)->root;
-	u64 alloc_hint = 0;
-	struct btrfs_key ins;
-	struct btrfs_trans_handle *trans;
+	u64 page_start;
 
 	if ((offset & (blocksize - 1)) == 0)
 		goto out;
@@ -536,7 +581,6 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
 	page = grab_cache_page(mapping, index);
 	if (!page)
 		goto out;
-
 	if (!PageUptodate(page)) {
 		ret = btrfs_readpage(NULL, page);
 		lock_page(page);
@@ -545,37 +589,24 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
 			goto out;
 		}
 	}
+	page_start = page->index << PAGE_CACHE_SHIFT;
+
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
-	btrfs_set_trans_block_group(trans, inode);
-
-	ret = btrfs_drop_extents(trans, root, inode,
-				 page->index << PAGE_CACHE_SHIFT,
-				 (page->index + 1) << PAGE_CACHE_SHIFT,
-				 &alloc_hint);
-	if (ret)
-		goto out;
-	ret = btrfs_alloc_extent(trans, root, inode->i_ino, 1, 0,
-				 alloc_hint, (u64)-1, &ins, 1);
-	if (ret)
-		goto out;
-	ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
-				       page->index << PAGE_CACHE_SHIFT,
-				       ins.objectid, 1, 1);
-	if (ret)
-		goto out;
-	SetPageChecked(page);
-	kaddr = kmap(page);
-	memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
-	flush_dcache_page(page);
-	ret = btrfs_csum_file_block(trans, root, inode->i_ino,
-			      page->index << PAGE_CACHE_SHIFT,
-			      kaddr, PAGE_CACHE_SIZE);
-	kunmap(page);
+	ret = btrfs_cow_one_page(trans, inode, page, offset);
+	if (!ret) {
+		char *kaddr = kmap(page);
+		ret = btrfs_csum_file_block(trans, root, inode->i_ino,
+					    page_start, kaddr, PAGE_CACHE_SIZE);
+		kunmap(page);
+	}
+	set_extent_dirty(&BTRFS_I(inode)->extent_tree,
+			 page_start, page_start + PAGE_CACHE_SIZE - 1,
+			 GFP_NOFS);
+	set_page_dirty(page);
 	btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
 
-	set_page_dirty(page);
 	unlock_page(page);
 	page_cache_release(page);
 out:
@@ -1095,6 +1126,8 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry,
 		inode->i_mapping->a_ops = &btrfs_aops;
 		inode->i_fop = &btrfs_file_operations;
 		inode->i_op = &btrfs_file_inode_operations;
+		extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
+				     inode->i_mapping, GFP_NOFS);
 	}
 	dir->i_sb->s_dirt = 1;
 	btrfs_update_inode_block_group(trans, inode);
@@ -1238,6 +1271,182 @@ out_unlock:
 	return err;
 }
 
+struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
+				    size_t page_offset, u64 start, u64 end,
+				    int create)
+{
+	int ret;
+	int err = 0;
+	u64 blocknr;
+	u64 extent_start = 0;
+	u64 extent_end = 0;
+	u64 objectid = inode->i_ino;
+	u32 found_type;
+	int failed_insert = 0;
+	struct btrfs_path *path;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_file_extent_item *item;
+	struct btrfs_leaf *leaf;
+	struct btrfs_disk_key *found_key;
+	struct extent_map *em = NULL;
+	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+	struct btrfs_trans_handle *trans = NULL;
+
+	path = btrfs_alloc_path();
+	BUG_ON(!path);
+	mutex_lock(&root->fs_info->fs_mutex);
+
+again:
+	em = lookup_extent_mapping(em_tree, start, end);
+	if (em) {
+		goto out;
+	}
+	if (!em) {
+		em = alloc_extent_map(GFP_NOFS);
+		if (!em) {
+			err = -ENOMEM;
+			goto out;
+		}
+		em->start = 0;
+		em->end = 0;
+	}
+	em->bdev = inode->i_sb->s_bdev;
+	ret = btrfs_lookup_file_extent(NULL, root, path,
+				       objectid, start, 0);
+	if (ret < 0) {
+		err = ret;
+		goto out;
+	}
+
+	if (ret != 0) {
+		if (path->slots[0] == 0)
+			goto not_found;
+		path->slots[0]--;
+	}
+
+	item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
+			      struct btrfs_file_extent_item);
+	leaf = btrfs_buffer_leaf(path->nodes[0]);
+	blocknr = btrfs_file_extent_disk_blocknr(item);
+	blocknr += btrfs_file_extent_offset(item);
+
+	/* are we inside the extent that was found? */
+	found_key = &leaf->items[path->slots[0]].key;
+	found_type = btrfs_disk_key_type(found_key);
+	if (btrfs_disk_key_objectid(found_key) != objectid ||
+	    found_type != BTRFS_EXTENT_DATA_KEY) {
+		goto not_found;
+	}
+
+	found_type = btrfs_file_extent_type(item);
+	extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key);
+	if (found_type == BTRFS_FILE_EXTENT_REG) {
+		extent_end = extent_start +
+		       (btrfs_file_extent_num_blocks(item) << inode->i_blkbits);
+		err = 0;
+		if (start < extent_start || start > extent_end) {
+			em->start = start;
+			if (start < extent_start) {
+				em->end = extent_end - 1;
+			} else {
+				em->end = end;
+			}
+			goto not_found_em;
+		}
+		if (btrfs_file_extent_disk_blocknr(item) == 0) {
+			em->start = extent_start;
+			em->end = extent_end - 1;
+			em->block_start = 0;
+			em->block_end = 0;
+			goto insert;
+		}
+		em->block_start = blocknr << inode->i_blkbits;
+		em->block_end = em->block_start +
+			(btrfs_file_extent_num_blocks(item) <<
+			 inode->i_blkbits) - 1;
+		em->start = extent_start;
+		em->end = extent_end - 1;
+		goto insert;
+	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
+		char *ptr;
+		char *map;
+		u32 size;
+
+		size = btrfs_file_extent_inline_len(leaf->items +
+						    path->slots[0]);
+		extent_end = extent_start + size;
+		if (start < extent_start || start > extent_end) {
+			em->start = start;
+			if (start < extent_start) {
+				em->end = extent_end - 1;
+			} else {
+				em->end = end;
+			}
+			goto not_found_em;
+		}
+		em->block_start = EXTENT_MAP_INLINE;
+		em->block_end = EXTENT_MAP_INLINE;
+		em->start = extent_start;
+		em->end = extent_end - 1;
+		if (!page) {
+			goto insert;
+		}
+		ptr = btrfs_file_extent_inline_start(item);
+		map = kmap(page);
+		memcpy(map + page_offset, ptr, size);
+		flush_dcache_page(result->b_page);
+		kunmap(page);
+		set_extent_uptodate(em_tree, extent_start,
+				    extent_end, GFP_NOFS);
+		goto insert;
+	} else {
+		printk("unkknown found_type %d\n", found_type);
+		WARN_ON(1);
+	}
+not_found:
+	em->start = start;
+	em->end = end;
+not_found_em:
+	em->block_start = 0;
+	em->block_end = 0;
+insert:
+	btrfs_release_path(root, path);
+	if (em->start > start || em->end < start) {
+		printk("bad extent! %Lu %Lu start %Lu end %Lu\n", em->start, em->end, start, end);
+		WARN_ON(1);
+		err = -EIO;
+		goto out;
+	}
+	ret = add_extent_mapping(em_tree, em);
+	if (ret == -EEXIST) {
+		free_extent_map(em);
+		failed_insert++;
+		if (failed_insert > 5) {
+			printk("failing to insert %Lu %Lu\n", start, end);
+			err = -EIO;
+			goto out;
+		}
+		em = NULL;
+		goto again;
+	}
+	err = 0;
+out:
+	btrfs_free_path(path);
+	if (trans) {
+		ret = btrfs_end_transaction(trans, root);
+		if (!err)
+			err = ret;
+	}
+	mutex_unlock(&root->fs_info->fs_mutex);
+	if (err) {
+		free_extent_map(em);
+		WARN_ON(1);
+		return ERR_PTR(err);
+	}
+	return em;
+}
+
+
 /*
  * FIBMAP and others want to pass in a fake buffer head.  They need to
  * use BTRFS_GET_BLOCK_NO_DIRECT to make sure we don't try to memcpy
@@ -1398,46 +1607,22 @@ int btrfs_get_block(struct inode *inode, sector_t iblock,
 	return err;
 }
 
-static int btrfs_get_block_csum(struct inode *inode, sector_t iblock,
-				struct buffer_head *result, int create)
-{
-	int ret;
-	struct btrfs_root *root = BTRFS_I(inode)->root;
-	struct page *page = result->b_page;
-	u64 offset = (page->index << PAGE_CACHE_SHIFT) + bh_offset(result);
-	struct btrfs_csum_item *item;
-	struct btrfs_path *path = NULL;
-
-	mutex_lock(&root->fs_info->fs_mutex);
-	ret = btrfs_get_block_lock(inode, iblock, result, create);
-	if (ret)
-		goto out;
-
-	path = btrfs_alloc_path();
-	item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, offset, 0);
-	if (IS_ERR(item)) {
-		ret = PTR_ERR(item);
-		/* a csum that isn't present is a preallocated region. */
-		if (ret == -ENOENT || ret == -EFBIG)
-			ret = 0;
-		result->b_private = NULL;
-		goto out;
-	}
-	memcpy((char *)&result->b_private, &item->csum, BTRFS_CRC32_SIZE);
-out:
-	if (path)
-		btrfs_free_path(path);
-	mutex_unlock(&root->fs_info->fs_mutex);
-	return ret;
-}
-
 static int btrfs_get_block_bmap(struct inode *inode, sector_t iblock,
 			   struct buffer_head *result, int create)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
-	mutex_lock(&root->fs_info->fs_mutex);
-	btrfs_get_block_lock(inode, iblock, result, BTRFS_GET_BLOCK_NO_DIRECT);
-	mutex_unlock(&root->fs_info->fs_mutex);
+	u64 start = iblock << inode->i_blkbits;
+	u64 end = start + root->blocksize -1;
+	struct extent_map *em;
+
+	em = btrfs_get_extent(inode, NULL, 0, start, end, 0);
+	if (em && !IS_ERR(em) && em->block_start != EXTENT_MAP_INLINE &&
+	    em->block_start != 0) {
+		u64 offset;
+		offset = start - em->start;
+		start = (em->block_start + offset) >> inode->i_blkbits;
+		btrfs_map_bh_to_logical(root, result, start);
+	}
 	return 0;
 }
 
@@ -1449,442 +1634,50 @@ static sector_t btrfs_bmap(struct address_space *as, sector_t block)
 static int btrfs_prepare_write(struct file *file, struct page *page,
 			       unsigned from, unsigned to)
 {
-	return block_prepare_write(page, from, to, btrfs_get_block);
+	return extent_prepare_write(&BTRFS_I(page->mapping->host)->extent_tree,
+				    page->mapping->host, page, from, to,
+				    btrfs_get_extent);
 }
 
-static void buffer_io_error(struct buffer_head *bh)
+int btrfs_readpage(struct file *file, struct page *page)
 {
-	char b[BDEVNAME_SIZE];
-
-	printk(KERN_ERR "Buffer I/O error on device %s, logical block %Lu\n",
-			bdevname(bh->b_bdev, b),
-			(unsigned long long)bh->b_blocknr);
+	struct extent_map_tree *tree;
+	tree = &BTRFS_I(page->mapping->host)->extent_tree;
+	return extent_read_full_page(tree, page, btrfs_get_extent);
 }
-
-/*
- * I/O completion handler for block_read_full_page() - pages
- * which come unlocked at the end of I/O.
- */
-static void btrfs_end_buffer_async_read(struct buffer_head *bh, int uptodate)
+static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
 {
-	unsigned long flags;
-	struct buffer_head *first;
-	struct buffer_head *tmp;
-	struct page *page;
-	int page_uptodate = 1;
-	struct inode *inode;
-	int ret;
-
-	BUG_ON(!buffer_async_read(bh));
-
-	page = bh->b_page;
-	inode = page->mapping->host;
-	if (uptodate) {
-		void *kaddr;
-		struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
-		if (bh->b_private) {
-			char csum[BTRFS_CRC32_SIZE];
-			kaddr = kmap_atomic(page, KM_IRQ0);
-			ret = btrfs_csum_data(root, kaddr + bh_offset(bh),
-					      bh->b_size, csum);
-			BUG_ON(ret);
-			if (memcmp(csum, &bh->b_private, BTRFS_CRC32_SIZE)) {
-				u64 offset;
-				offset = (page->index << PAGE_CACHE_SHIFT) +
-					bh_offset(bh);
-				printk("btrfs csum failed ino %lu off %llu\n",
-				       page->mapping->host->i_ino,
-				       (unsigned long long)offset);
-				memset(kaddr + bh_offset(bh), 1, bh->b_size);
-				flush_dcache_page(page);
-			}
-			kunmap_atomic(kaddr, KM_IRQ0);
-		}
-		set_buffer_uptodate(bh);
-	} else {
-		clear_buffer_uptodate(bh);
-		if (printk_ratelimit())
-			buffer_io_error(bh);
-		SetPageError(page);
-	}
-
-	/*
-	 * Be _very_ careful from here on. Bad things can happen if
-	 * two buffer heads end IO at almost the same time and both
-	 * decide that the page is now completely done.
-	 */
-	first = page_buffers(page);
-	local_irq_save(flags);
-	bit_spin_lock(BH_Uptodate_Lock, &first->b_state);
-	clear_buffer_async_read(bh);
-	unlock_buffer(bh);
-	tmp = bh;
-	do {
-		if (!buffer_uptodate(tmp))
-			page_uptodate = 0;
-		if (buffer_async_read(tmp)) {
-			BUG_ON(!buffer_locked(tmp));
-			goto still_busy;
-		}
-		tmp = tmp->b_this_page;
-	} while (tmp != bh);
-	bit_spin_unlock(BH_Uptodate_Lock, &first->b_state);
-	local_irq_restore(flags);
-
-	/*
-	 * If none of the buffers had errors and they are all
-	 * uptodate then we can set the page uptodate.
-	 */
-	if (page_uptodate && !PageError(page))
-		SetPageUptodate(page);
-	unlock_page(page);
-	return;
-
-still_busy:
-	bit_spin_unlock(BH_Uptodate_Lock, &first->b_state);
-	local_irq_restore(flags);
-	return;
+	struct extent_map_tree *tree;
+	tree = &BTRFS_I(page->mapping->host)->extent_tree;
+	return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
 }
 
-/*
- * Generic "read page" function for block devices that have the normal
- * get_block functionality. This is most of the block device filesystems.
- * Reads the page asynchronously --- the unlock_buffer() and
- * set/clear_buffer_uptodate() functions propagate buffer state into the
- * page struct once IO has completed.
- */
-int btrfs_readpage(struct file *file, struct page *page)
+static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
 {
-	struct inode *inode = page->mapping->host;
-	sector_t iblock, lblock;
-	struct buffer_head *bh, *head, *arr[MAX_BUF_PER_PAGE];
-	unsigned int blocksize;
-	int nr, i;
-	int fully_mapped = 1;
-
-	BUG_ON(!PageLocked(page));
-	blocksize = 1 << inode->i_blkbits;
-	if (!page_has_buffers(page))
-		create_empty_buffers(page, blocksize, 0);
-	head = page_buffers(page);
-
-	iblock = (sector_t)page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
-	lblock = (i_size_read(inode)+blocksize-1) >> inode->i_blkbits;
-	bh = head;
-	nr = 0;
-	i = 0;
-
-	do {
-		if (buffer_uptodate(bh))
-			continue;
-
-		if (!buffer_mapped(bh)) {
-			int err = 0;
-
-			fully_mapped = 0;
-			if (iblock < lblock) {
-				WARN_ON(bh->b_size != blocksize);
-				err = btrfs_get_block_csum(inode, iblock,
-							   bh, 0);
-				if (err)
-					SetPageError(page);
-			}
-			if (!buffer_mapped(bh)) {
-				void *kaddr = kmap_atomic(page, KM_USER0);
-				memset(kaddr + i * blocksize, 0, blocksize);
-				flush_dcache_page(page);
-				kunmap_atomic(kaddr, KM_USER0);
-				if (!err)
-					set_buffer_uptodate(bh);
-				continue;
-			}
-			/*
-			 * get_block() might have updated the buffer
-			 * synchronously
-			 */
-			if (buffer_uptodate(bh))
-				continue;
-		}
-		arr[nr++] = bh;
-	} while (i++, iblock++, (bh = bh->b_this_page) != head);
-
-	if (fully_mapped)
-		SetPageMappedToDisk(page);
-
-	if (!nr) {
-		/*
-		 * All buffers are uptodate - we can set the page uptodate
-		 * as well. But not if get_block() returned an error.
-		 */
-		if (!PageError(page))
-			SetPageUptodate(page);
-		unlock_page(page);
-		return 0;
-	}
-
-	/* Stage two: lock the buffers */
-	for (i = 0; i < nr; i++) {
-		bh = arr[i];
-		lock_buffer(bh);
-		bh->b_end_io = btrfs_end_buffer_async_read;
-		set_buffer_async_read(bh);
-	}
-
-	/*
-	 * Stage 3: start the IO.  Check for uptodateness
-	 * inside the buffer lock in case another process reading
-	 * the underlying blockdev brought it uptodate (the sct fix).
-	 */
-	for (i = 0; i < nr; i++) {
-		bh = arr[i];
-		if (buffer_uptodate(bh))
-			btrfs_end_buffer_async_read(bh, 1);
-		else
-			submit_bh(READ, bh);
-	}
-	return 0;
-}
-
-/*
- * Aside from a tiny bit of packed file data handling, this is the
- * same as the generic code.
- *
- * While block_write_full_page is writing back the dirty buffers under
- * the page lock, whoever dirtied the buffers may decide to clean them
- * again at any time.  We handle that by only looking at the buffer
- * state inside lock_buffer().
- *
- * If block_write_full_page() is called for regular writeback
- * (wbc->sync_mode == WB_SYNC_NONE) then it will redirty a page which has a
- * locked buffer.   This only can happen if someone has written the buffer
- * directly, with submit_bh().  At the address_space level PageWriteback
- * prevents this contention from occurring.
- */
-static int __btrfs_write_full_page(struct inode *inode, struct page *page,
-				   struct writeback_control *wbc)
-{
-	int err;
-	sector_t block;
-	sector_t last_block;
-	struct buffer_head *bh, *head;
-	const unsigned blocksize = 1 << inode->i_blkbits;
-	int nr_underway = 0;
-	struct btrfs_root *root = BTRFS_I(inode)->root;
-
-	BUG_ON(!PageLocked(page));
-
-	last_block = (i_size_read(inode) - 1) >> inode->i_blkbits;
-
-	/* no csumming allowed when from PF_MEMALLOC */
-	if (current->flags & PF_MEMALLOC) {
-		redirty_page_for_writepage(wbc, page);
-		unlock_page(page);
-		return 0;
-	}
+	struct extent_map_tree *tree;
+	int ret;
 
-	if (!page_has_buffers(page)) {
-		create_empty_buffers(page, blocksize,
-					(1 << BH_Dirty)|(1 << BH_Uptodate));
+	if (page->private != 1) {
+		WARN_ON(1);
+		return try_to_free_buffers(page);
 	}
-
-	/*
-	 * Be very careful.  We have no exclusion from __set_page_dirty_buffers
-	 * here, and the (potentially unmapped) buffers may become dirty at
-	 * any time.  If a buffer becomes dirty here after we've inspected it
-	 * then we just miss that fact, and the page stays dirty.
-	 *
-	 * Buffers outside i_size may be dirtied by __set_page_dirty_buffers;
-	 * handle that here by just cleaning them.
-	 */
-
-	block = (sector_t)page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
-	head = page_buffers(page);
-	bh = head;
-
-	/*
-	 * Get all the dirty buffers mapped to disk addresses and
-	 * handle any aliases from the underlying blockdev's mapping.
-	 */
-	do {
-		if (block > last_block) {
-			/*
-			 * mapped buffers outside i_size will occur, because
-			 * this page can be outside i_size when there is a
-			 * truncate in progress.
-			 */
-			/*
-			 * The buffer was zeroed by block_write_full_page()
-			 */
-			clear_buffer_dirty(bh);
-			set_buffer_uptodate(bh);
-		} else if (!buffer_mapped(bh) && buffer_dirty(bh)) {
-			WARN_ON(bh->b_size != blocksize);
-			err = btrfs_get_block(inode, block, bh, 0);
-			if (err) {
-				goto recover;
-			}
-			if (buffer_new(bh)) {
-				/* blockdev mappings never come here */
-				clear_buffer_new(bh);
-			}
-		}
-		bh = bh->b_this_page;
-		block++;
-	} while (bh != head);
-
-	do {
-		if (!buffer_mapped(bh))
-			continue;
-		/*
-		 * If it's a fully non-blocking write attempt and we cannot
-		 * lock the buffer then redirty the page.  Note that this can
-		 * potentially cause a busy-wait loop from pdflush and kswapd
-		 * activity, but those code paths have their own higher-level
-		 * throttling.
-		 */
-		if (wbc->sync_mode != WB_SYNC_NONE || !wbc->nonblocking) {
-			lock_buffer(bh);
-		} else if (test_set_buffer_locked(bh)) {
-			redirty_page_for_writepage(wbc, page);
-			continue;
-		}
-		if (test_clear_buffer_dirty(bh) && bh->b_blocknr != 0) {
-			struct btrfs_trans_handle *trans;
-			int ret;
-			u64 off = page->index << PAGE_CACHE_SHIFT;
-			char *kaddr;
-
-			off += bh_offset(bh);
-			mutex_lock(&root->fs_info->fs_mutex);
-			trans = btrfs_start_transaction(root, 1);
-			btrfs_set_trans_block_group(trans, inode);
-			kaddr = kmap(page);
-			btrfs_csum_file_block(trans, root, inode->i_ino,
-						    off, kaddr + bh_offset(bh),
-						    bh->b_size);
-			kunmap(page);
-			ret = btrfs_end_transaction(trans, root);
-			BUG_ON(ret);
-			mutex_unlock(&root->fs_info->fs_mutex);
-			mark_buffer_async_write(bh);
-		} else {
-			unlock_buffer(bh);
-		}
-	} while ((bh = bh->b_this_page) != head);
-
-	/*
-	 * The page and its buffers are protected by PageWriteback(), so we can
-	 * drop the bh refcounts early.
-	 */
-	BUG_ON(PageWriteback(page));
-	set_page_writeback(page);
-
-	do {
-		struct buffer_head *next = bh->b_this_page;
-		if (buffer_async_write(bh)) {
-			submit_bh(WRITE, bh);
-			nr_underway++;
-		}
-		bh = next;
-	} while (bh != head);
-	unlock_page(page);
-
-	err = 0;
-done:
-	if (nr_underway == 0) {
-		/*
-		 * The page was marked dirty, but the buffers were
-		 * clean.  Someone wrote them back by hand with
-		 * ll_rw_block/submit_bh.  A rare case.
-		 */
-		int uptodate = 1;
-		do {
-			if (!buffer_uptodate(bh)) {
-				uptodate = 0;
-				break;
-			}
-			bh = bh->b_this_page;
-		} while (bh != head);
-		if (uptodate)
-			SetPageUptodate(page);
-		end_page_writeback(page);
+	tree = &BTRFS_I(page->mapping->host)->extent_tree;
+	ret = try_release_extent_mapping(tree, page);
+	if (ret == 1) {
+		ClearPagePrivate(page);
+		set_page_private(page, 0);
+		page_cache_release(page);
 	}
-	return err;
-
-recover:
-	/*
-	 * ENOSPC, or some other error.  We may already have added some
-	 * blocks to the file, so we need to write these out to avoid
-	 * exposing stale data.
-	 * The page is currently locked and not marked for writeback
-	 */
-	bh = head;
-	/* Recovery: lock and submit the mapped buffers */
-	do {
-		if (buffer_mapped(bh) && buffer_dirty(bh)) {
-			lock_buffer(bh);
-			mark_buffer_async_write(bh);
-		} else {
-			/*
-			 * The buffer may have been set dirty during
-			 * attachment to a dirty page.
-			 */
-			clear_buffer_dirty(bh);
-		}
-	} while ((bh = bh->b_this_page) != head);
-	SetPageError(page);
-	BUG_ON(PageWriteback(page));
-	set_page_writeback(page);
-	do {
-		struct buffer_head *next = bh->b_this_page;
-		if (buffer_async_write(bh)) {
-			clear_buffer_dirty(bh);
-			submit_bh(WRITE, bh);
-			nr_underway++;
-		}
-		bh = next;
-	} while (bh != head);
-	unlock_page(page);
-	goto done;
+	return ret;
 }
 
-static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
+static void btrfs_invalidatepage(struct page *page, unsigned long offset)
 {
-	struct inode * const inode = page->mapping->host;
-	loff_t i_size = i_size_read(inode);
-	const pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
-	unsigned offset;
-	void *kaddr;
-
-	/* Is the page fully inside i_size? */
-	if (page->index < end_index)
-		return __btrfs_write_full_page(inode, page, wbc);
-
-	/* Is the page fully outside i_size? (truncate in progress) */
-	offset = i_size & (PAGE_CACHE_SIZE-1);
-	if (page->index >= end_index+1 || !offset) {
-		/*
-		 * The page may have dirty, unmapped buffers.  For example,
-		 * they may have been added in ext3_writepage().  Make them
-		 * freeable here, so the page does not leak.
-		 */
-		block_invalidatepage(page, 0);
-		unlock_page(page);
-		return 0; /* don't care */
-	}
+	struct extent_map_tree *tree;
 
-	/*
-	 * The page straddles i_size.  It must be zeroed out on each and every
-	 * writepage invokation because it may be mmapped.  "A file is mapped
-	 * in multiples of the page size.  For a file that is not a multiple of
-	 * the  page size, the remaining memory is zeroed when mapped, and
-	 * writes to that region are not written out to the file."
-	 */
-	kaddr = kmap_atomic(page, KM_USER0);
-	memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
-	flush_dcache_page(page);
-	kunmap_atomic(kaddr, KM_USER0);
-	return __btrfs_write_full_page(inode, page, wbc);
+	tree = &BTRFS_I(page->mapping->host)->extent_tree;
+	extent_invalidatepage(tree, page, offset);
+	btrfs_releasepage(page, GFP_NOFS);
 }
 
 /*
@@ -1905,28 +1698,39 @@ static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
 {
 	struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_trans_handle *trans;
 	unsigned long end;
 	loff_t size;
 	int ret = -EINVAL;
+	u64 page_start;
 
 	lock_page(page);
 	wait_on_page_writeback(page);
 	size = i_size_read(inode);
+	page_start = page->index << PAGE_CACHE_SHIFT;
+
 	if ((page->mapping != inode->i_mapping) ||
-	    ((page->index << PAGE_CACHE_SHIFT) > size)) {
+	    (page_start > size)) {
 		/* page got truncated out from underneath us */
 		goto out_unlock;
 	}
 
 	/* page is wholly or partially inside EOF */
-	if (((page->index + 1) << PAGE_CACHE_SHIFT) > size)
+	if (page_start + PAGE_CACHE_SIZE > size)
 		end = size & ~PAGE_CACHE_MASK;
 	else
 		end = PAGE_CACHE_SIZE;
 
-	ret = btrfs_prepare_write(NULL, page, 0, end);
-	if (!ret)
-		ret = btrfs_commit_write(NULL, page, 0, end);
+	mutex_lock(&root->fs_info->fs_mutex);
+	trans = btrfs_start_transaction(root, 1);
+	ret = btrfs_cow_one_page(trans, inode, page, end);
+	btrfs_end_transaction(trans, root);
+	mutex_unlock(&root->fs_info->fs_mutex);
+	set_extent_dirty(&BTRFS_I(inode)->extent_tree,
+			 page_start, page_start + PAGE_CACHE_SIZE - 1,
+			 GFP_NOFS);
+	set_page_dirty(page);
 
 out_unlock:
 	unlock_page(page);
@@ -1962,21 +1766,8 @@ static void btrfs_truncate(struct inode *inode)
 int btrfs_commit_write(struct file *file, struct page *page,
 		       unsigned from, unsigned to)
 {
-	struct inode *inode = page->mapping->host;
-	struct buffer_head *bh;
-	loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;
-
-	SetPageUptodate(page);
-	bh = page_buffers(page);
-	set_buffer_uptodate(bh);
-	if (buffer_mapped(bh) && bh->b_blocknr != 0) {
-		set_page_dirty(page);
-	}
-	if (pos > inode->i_size) {
-		i_size_write(inode, pos);
-		mark_inode_dirty(inode);
-	}
-	return 0;
+	return extent_commit_write(&BTRFS_I(page->mapping->host)->extent_tree,
+				   page->mapping->host, page, from, to);
 }
 
 static int create_subvol(struct btrfs_root *root, char *name, int namelen)
@@ -2471,6 +2262,8 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
 		inode->i_mapping->a_ops = &btrfs_aops;
 		inode->i_fop = &btrfs_file_operations;
 		inode->i_op = &btrfs_file_inode_operations;
+		extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
+				     inode->i_mapping, GFP_NOFS);
 	}
 	dir->i_sb->s_dirt = 1;
 	btrfs_update_inode_block_group(trans, inode);
@@ -2553,6 +2346,9 @@ static struct address_space_operations btrfs_aops = {
 	.prepare_write	= btrfs_prepare_write,
 	.commit_write	= btrfs_commit_write,
 	.bmap		= btrfs_bmap,
+	.invalidatepage = btrfs_invalidatepage,
+	.releasepage	= btrfs_releasepage,
+	.set_page_dirty	= __set_page_dirty_nobuffers,
 };
 
 static struct address_space_operations btrfs_symlink_aops = {
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 74f3de47423c..7b7c32c73a2c 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -185,6 +185,7 @@ static int __init init_btrfs_fs(void)
 	err = btrfs_init_cachep();
 	if (err)
 		return err;
+	extent_map_init();
 	return register_filesystem(&btrfs_fs_type);
 }
 
@@ -192,6 +193,7 @@ static void __exit exit_btrfs_fs(void)
 {
 	btrfs_exit_transaction_sys();
 	btrfs_destroy_cachep();
+	extent_map_exit();
 	unregister_filesystem(&btrfs_fs_type);
 }
 
-- 
cgit v1.2.3


From 58176a9604c5db1784d2c979aea472b3be40b6f0 Mon Sep 17 00:00:00 2001
From: Josef Bacik <jbacik@redhat.com>
Date: Wed, 29 Aug 2007 15:47:34 -0400
Subject: Btrfs: Add per-root block accounting and sysfs entries

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/Makefile      |   2 +-
 fs/btrfs/ctree.h       |  54 +++++++++++
 fs/btrfs/disk-io.c     |  31 ++++++-
 fs/btrfs/disk-io.h     |   3 +-
 fs/btrfs/extent-tree.c |  17 +++-
 fs/btrfs/inode.c       |  13 ++-
 fs/btrfs/super.c       |  14 +++
 fs/btrfs/sysfs.c       | 236 +++++++++++++++++++++++++++++++++++++++++++++++++
 fs/btrfs/transaction.c |  34 ++++++-
 9 files changed, 393 insertions(+), 11 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index ea7a22870350..ee505c87ecf3 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -5,7 +5,7 @@ obj-m  := btrfs.o
 btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
 	   hash.o file-item.o inode-item.o inode-map.o disk-io.o \
 	   transaction.o bit-radix.o inode.o file.o tree-defrag.o \
-	   extent_map.o
+	   extent_map.o sysfs.o
 
 #btrfs-y := ctree.o disk-io.o radix-tree.o extent-tree.o print-tree.o \
 #	  root-tree.o dir-item.o hash.o file-item.o inode-item.o \
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 947f061ed118..ffeee546a0d7 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -22,6 +22,7 @@
 #include <linux/fs.h>
 #include <linux/buffer_head.h>
 #include <linux/workqueue.h>
+#include <linux/completion.h>
 #include "bit-radix.h"
 
 struct btrfs_trans_handle;
@@ -313,6 +314,8 @@ struct btrfs_fs_info {
 	struct list_head trans_list;
 	struct list_head dead_roots;
 	struct delayed_work trans_work;
+	struct kobject super_kobj;
+	struct completion kobj_unregister;
 	int do_barriers;
 	int closing;
 };
@@ -328,6 +331,8 @@ struct btrfs_root {
 	struct btrfs_key root_key;
 	struct btrfs_fs_info *fs_info;
 	struct inode *inode;
+	struct kobject root_kobj;
+	struct completion kobj_unregister;
 	u64 objectid;
 	u64 last_trans;
 	u32 blocksize;
@@ -338,6 +343,7 @@ struct btrfs_root {
 	struct btrfs_key defrag_progress;
 	int defrag_running;
 	int defrag_level;
+	char *name;
 };
 
 /* the lower bits in the key flags defines the item type */
@@ -814,6 +820,28 @@ static inline void btrfs_set_root_flags(struct btrfs_root_item *item, u32 val)
 	item->flags = cpu_to_le32(val);
 }
 
+static inline void btrfs_set_root_blocks_used(struct btrfs_root_item *item,
+						   u64 val)
+{
+	item->blocks_used = cpu_to_le64(val);
+}
+
+static inline u64 btrfs_root_blocks_used(struct btrfs_root_item *item)
+{
+	return le64_to_cpu(item->blocks_used);
+}
+
+static inline void btrfs_set_root_block_limit(struct btrfs_root_item *item,
+						u64 val)
+{
+	item->block_limit = cpu_to_le64(val);
+}
+
+static inline u64 btrfs_root_block_limit(struct btrfs_root_item *item)
+{
+	return le64_to_cpu(item->block_limit);
+}
+
 static inline u64 btrfs_super_blocknr(struct btrfs_super_block *s)
 {
 	return le64_to_cpu(s->blocknr);
@@ -1014,6 +1042,23 @@ static inline void btrfs_memmove(struct btrfs_root *root,
 	memmove(dst, src, nr);
 }
 
+static inline int btrfs_set_root_name(struct btrfs_root *root,
+				      const char *name, int len)
+{
+	/* if we already have a name just free it */
+	if (root->name)
+		kfree(root->name);
+
+	root->name = kmalloc(len+1, GFP_KERNEL);
+	if (!root->name)
+		return -ENOMEM;
+
+	memcpy(root->name, name, len);
+	root->name[len] ='\0';
+
+	return 0;
+}
+
 /* helper function to cast into the data area of the leaf. */
 #define btrfs_item_ptr(leaf, slot, type) \
 	((type *)(btrfs_leaf_data(leaf) + \
@@ -1191,4 +1236,13 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans,
 /* tree-defrag.c */
 int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root, int cache_only);
+
+/* sysfs.c */
+int btrfs_init_sysfs(void);
+void btrfs_exit_sysfs(void);
+int btrfs_sysfs_add_super(struct btrfs_fs_info *fs);
+int btrfs_sysfs_add_root(struct btrfs_root *root);
+void btrfs_sysfs_del_root(struct btrfs_root *root);
+void btrfs_sysfs_del_super(struct btrfs_fs_info *root);
+
 #endif
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index b2f79878d51a..c25ef0a68f18 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -294,9 +294,12 @@ static int __setup_root(int blocksize,
 	root->last_trans = 0;
 	root->highest_inode = 0;
 	root->last_inode_alloc = 0;
+	root->name = NULL;
 	memset(&root->root_key, 0, sizeof(root->root_key));
 	memset(&root->root_item, 0, sizeof(root->root_item));
 	memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
+	memset(&root->root_kobj, 0, sizeof(root->root_kobj));
+	init_completion(&root->kobj_unregister);
 	root->defrag_running = 0;
 	root->defrag_level = 0;
 	root->root_key.objectid = objectid;
@@ -384,7 +387,8 @@ insert:
 }
 
 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
-				      struct btrfs_key *location)
+				      struct btrfs_key *location,
+				      const char *name, int namelen)
 {
 	struct btrfs_root *root;
 	int ret;
@@ -405,6 +409,22 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 		kfree(root);
 		return ERR_PTR(ret);
 	}
+
+	ret = btrfs_set_root_name(root, name, namelen);
+	if (ret) {
+		brelse(root->node);
+		kfree(root);
+		return ERR_PTR(ret);
+	}
+
+	ret = btrfs_sysfs_add_root(root);
+	if (ret) {
+		brelse(root->node);
+		kfree(root->name);
+		kfree(root);
+		return ERR_PTR(ret);
+	}
+
 	return root;
 }
 
@@ -433,6 +453,8 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	INIT_RADIX_TREE(&fs_info->block_group_data_radix, GFP_KERNEL);
 	INIT_LIST_HEAD(&fs_info->trans_list);
 	INIT_LIST_HEAD(&fs_info->dead_roots);
+	memset(&fs_info->super_kobj, 0, sizeof(fs_info->super_kobj));
+	init_completion(&fs_info->kobj_unregister);
 	sb_set_blocksize(sb, 4096);
 	fs_info->running_transaction = NULL;
 	fs_info->last_trans_committed = 0;
@@ -500,8 +522,10 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 
 	fs_info->generation = btrfs_super_generation(disk_super) + 1;
 	ret = btrfs_find_dead_roots(tree_root);
-	if (ret)
+	if (ret) {
+		mutex_unlock(&fs_info->fs_mutex);
 		goto fail_tree_root;
+	}
 	mutex_unlock(&fs_info->fs_mutex);
 	return tree_root;
 
@@ -553,12 +577,15 @@ int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
 {
 	radix_tree_delete(&fs_info->fs_roots_radix,
 			  (unsigned long)root->root_key.objectid);
+	btrfs_sysfs_del_root(root);
 	if (root->inode)
 		iput(root->inode);
 	if (root->node)
 		brelse(root->node);
 	if (root->commit_root)
 		brelse(root->commit_root);
+	if (root->name)
+		kfree(root->name);
 	kfree(root);
 	return 0;
 }
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 5261733b8735..da6bb72750f2 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -66,7 +66,8 @@ struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr);
 int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
 		    char *result);
 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
-				      struct btrfs_key *location);
+				      struct btrfs_key *location,
+				      const char *name, int namelen);
 struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
 					       struct btrfs_key *location);
 u64 bh_blocknr(struct buffer_head *bh);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index c31e84d42653..ff3f7c2be605 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -858,16 +858,23 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_set_extent_refs(ei, refs);
 	btrfs_mark_buffer_dirty(path->nodes[0]);
 	if (refs == 0) {
-		u64 super_blocks_used;
+		u64 super_blocks_used, root_blocks_used;
 
 		if (pin) {
 			ret = pin_down_block(root, blocknr, 0);
 			BUG_ON(ret);
 		}
 
+		/* block accounting for super block */
 		super_blocks_used = btrfs_super_blocks_used(&info->super_copy);
 		btrfs_set_super_blocks_used(&info->super_copy,
 					    super_blocks_used - num_blocks);
+
+		/* block accounting for root item */
+		root_blocks_used = btrfs_root_blocks_used(&root->root_item);
+		btrfs_set_root_blocks_used(&root->root_item,
+					   root_blocks_used - num_blocks);
+
 		ret = btrfs_del_item(trans, extent_root, path);
 		if (ret) {
 			return ret;
@@ -1175,7 +1182,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 {
 	int ret;
 	int pending_ret;
-	u64 super_blocks_used;
+	u64 super_blocks_used, root_blocks_used;
 	u64 search_start = 0;
 	struct btrfs_fs_info *info = root->fs_info;
 	struct btrfs_root *extent_root = info->extent_root;
@@ -1193,10 +1200,16 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 	if (ret)
 		return ret;
 
+	/* block accounting for super block */
 	super_blocks_used = btrfs_super_blocks_used(&info->super_copy);
 	btrfs_set_super_blocks_used(&info->super_copy, super_blocks_used +
 				    num_blocks);
 
+	/* block accounting for root item */
+	root_blocks_used = btrfs_root_blocks_used(&root->root_item);
+	btrfs_set_root_blocks_used(&root->root_item, root_blocks_used +
+				   num_blocks);
+
 	if (root == extent_root) {
 		BUG_ON(num_blocks != 1);
 		set_radix_bit(&root->fs_info->extent_ins_radix, ins->objectid);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index dc181089aa74..2e3918e6049e 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -733,7 +733,8 @@ out:
  */
 static int fixup_tree_root_location(struct btrfs_root *root,
 			     struct btrfs_key *location,
-			     struct btrfs_root **sub_root)
+			     struct btrfs_root **sub_root,
+			     struct dentry *dentry)
 {
 	struct btrfs_path *path;
 	struct btrfs_root_item *ri;
@@ -747,7 +748,9 @@ static int fixup_tree_root_location(struct btrfs_root *root,
 	BUG_ON(!path);
 	mutex_lock(&root->fs_info->fs_mutex);
 
-	*sub_root = btrfs_read_fs_root(root->fs_info, location);
+	*sub_root = btrfs_read_fs_root(root->fs_info, location,
+					dentry->d_name.name,
+					dentry->d_name.len);
 	if (IS_ERR(*sub_root))
 		return PTR_ERR(*sub_root);
 
@@ -812,7 +815,8 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
 		return ERR_PTR(ret);
 	inode = NULL;
 	if (location.objectid) {
-		ret = fixup_tree_root_location(root, &location, &sub_root);
+		ret = fixup_tree_root_location(root, &location, &sub_root,
+						dentry);
 		if (ret < 0)
 			return ERR_PTR(ret);
 		if (ret > 0)
@@ -1829,6 +1833,7 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 
 	btrfs_set_root_blocknr(&root_item, bh_blocknr(subvol));
 	btrfs_set_root_refs(&root_item, 1);
+	btrfs_set_root_blocks_used(&root_item, 0);
 	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
 	root_item.drop_level = 0;
 	brelse(subvol);
@@ -1865,7 +1870,7 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 	if (ret)
 		goto fail_commit;
 
-	new_root = btrfs_read_fs_root(root->fs_info, &key);
+	new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
 	BUG_ON(!new_root);
 
 	trans = btrfs_start_transaction(new_root, 1);
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 7da95538745f..a68101ae1a3b 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -45,12 +45,14 @@ static struct super_operations btrfs_super_ops;
 static void btrfs_put_super (struct super_block * sb)
 {
 	struct btrfs_root *root = btrfs_sb(sb);
+	struct btrfs_fs_info *fs = root->fs_info;
 	int ret;
 
 	ret = close_ctree(root);
 	if (ret) {
 		printk("close ctree returns %d\n", ret);
 	}
+	btrfs_sysfs_del_super(fs);
 	sb->s_fs_info = NULL;
 }
 
@@ -101,6 +103,12 @@ static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
 		err = -ENOMEM;
 		goto fail_close;
 	}
+
+	/* this does the super kobj at the same time */
+	err = btrfs_sysfs_add_super(tree_root->fs_info);
+	if (err)
+		goto fail_close;
+
 	sb->s_root = root_dentry;
 	btrfs_transaction_queue_work(tree_root, HZ * 30);
 	return 0;
@@ -182,6 +190,11 @@ static struct super_operations btrfs_super_ops = {
 static int __init init_btrfs_fs(void)
 {
 	int err;
+
+	err = btrfs_init_sysfs();
+	if (err)
+		return err;
+
 	btrfs_init_transaction_sys();
 	err = btrfs_init_cachep();
 	if (err)
@@ -196,6 +209,7 @@ static void __exit exit_btrfs_fs(void)
 	btrfs_destroy_cachep();
 	extent_map_exit();
 	unregister_filesystem(&btrfs_fs_type);
+	btrfs_exit_sysfs();
 }
 
 module_init(init_btrfs_fs)
diff --git a/fs/btrfs/sysfs.c b/fs/btrfs/sysfs.c
index db8917e5b256..2058783373eb 100644
--- a/fs/btrfs/sysfs.c
+++ b/fs/btrfs/sysfs.c
@@ -16,6 +16,242 @@
  * Boston, MA 021110-1307, USA.
  */
 
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/completion.h>
+#include <linux/buffer_head.h>
+#include <linux/module.h>
+#include <linux/kobject.h>
+
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
+
+static ssize_t root_blocks_used_show(struct btrfs_root *root, char *buf)
+{
+	return snprintf(buf, PAGE_SIZE, "%llu\n",
+		(unsigned long long)btrfs_root_blocks_used(&root->root_item));
+}
+
+static ssize_t root_block_limit_show(struct btrfs_root *root, char *buf)
+{
+	return snprintf(buf, PAGE_SIZE, "%llu\n",
+		(unsigned long long)btrfs_root_block_limit(&root->root_item));
+}
+
+static ssize_t super_blocks_used_show(struct btrfs_fs_info *fs, char *buf)
+{
+	return snprintf(buf, PAGE_SIZE, "%llu\n",
+		(unsigned long long)btrfs_super_blocks_used(fs->disk_super));
+}
+
+static ssize_t super_total_blocks_show(struct btrfs_fs_info *fs, char *buf)
+{
+	return snprintf(buf, PAGE_SIZE, "%llu\n",
+		(unsigned long long)btrfs_super_total_blocks(fs->disk_super));
+}
+
+static ssize_t super_blocksize_show(struct btrfs_fs_info *fs, char *buf)
+{
+	return snprintf(buf, PAGE_SIZE, "%llu\n",
+		(unsigned long long)btrfs_super_blocksize(fs->disk_super));
+}
+
+/* this is for root attrs (subvols/snapshots) */
+struct btrfs_root_attr {
+	struct attribute attr;
+	ssize_t (*show)(struct btrfs_root *, char *);
+	ssize_t (*store)(struct btrfs_root *, const char *, size_t);
+};
+
+#define ROOT_ATTR(name, mode, show, store) \
+static struct btrfs_root_attr btrfs_root_attr_##name = __ATTR(name, mode, show, store)
+
+ROOT_ATTR(blocks_used,	0444,	root_blocks_used_show,	NULL);
+ROOT_ATTR(block_limit,	0644,	root_block_limit_show,	NULL);
+
+static struct attribute *btrfs_root_attrs[] = {
+	&btrfs_root_attr_blocks_used.attr,
+	&btrfs_root_attr_block_limit.attr,
+	NULL,
+};
+
+/* this is for super attrs (actual full fs) */
+struct btrfs_super_attr {
+	struct attribute attr;
+	ssize_t (*show)(struct btrfs_fs_info *, char *);
+	ssize_t (*store)(struct btrfs_fs_info *, const char *, size_t);
+};
+
+#define SUPER_ATTR(name, mode, show, store) \
+static struct btrfs_super_attr btrfs_super_attr_##name = __ATTR(name, mode, show, store)
+
+SUPER_ATTR(blocks_used,		0444,	super_blocks_used_show,		NULL);
+SUPER_ATTR(total_blocks,	0444,	super_total_blocks_show,	NULL);
+SUPER_ATTR(blocksize,		0444,	super_blocksize_show,		NULL);
+
+static struct attribute *btrfs_super_attrs[] = {
+	&btrfs_super_attr_blocks_used.attr,
+	&btrfs_super_attr_total_blocks.attr,
+	&btrfs_super_attr_blocksize.attr,
+	NULL,
+};
+
+static ssize_t btrfs_super_attr_show(struct kobject *kobj,
+				    struct attribute *attr, char *buf)
+{
+	struct btrfs_fs_info *fs = container_of(kobj, struct btrfs_fs_info,
+						super_kobj);
+	struct btrfs_super_attr *a = container_of(attr,
+						  struct btrfs_super_attr,
+						  attr);
+
+	return a->show ? a->show(fs, buf) : 0;
+}
+
+static ssize_t btrfs_super_attr_store(struct kobject *kobj,
+				     struct attribute *attr,
+				     const char *buf, size_t len)
+{
+	struct btrfs_fs_info *fs = container_of(kobj, struct btrfs_fs_info,
+						super_kobj);
+	struct btrfs_super_attr *a = container_of(attr,
+						  struct btrfs_super_attr,
+						  attr);
+
+	return a->store ? a->store(fs, buf, len) : 0;
+}
+
+static ssize_t btrfs_root_attr_show(struct kobject *kobj,
+				    struct attribute *attr, char *buf)
+{
+	struct btrfs_root *root = container_of(kobj, struct btrfs_root,
+						root_kobj);
+	struct btrfs_root_attr *a = container_of(attr,
+						 struct btrfs_root_attr,
+						 attr);
+
+	return a->show ? a->show(root, buf) : 0;
+}
+
+static ssize_t btrfs_root_attr_store(struct kobject *kobj,
+				     struct attribute *attr,
+				     const char *buf, size_t len)
+{
+	struct btrfs_root *root = container_of(kobj, struct btrfs_root,
+						root_kobj);
+	struct btrfs_root_attr *a = container_of(attr,
+						 struct btrfs_root_attr,
+						 attr);
+	return a->store ? a->store(root, buf, len) : 0;
+}
+
+static void btrfs_super_release(struct kobject *kobj)
+{
+	struct btrfs_fs_info *fs = container_of(kobj, struct btrfs_fs_info,
+						super_kobj);
+	complete(&fs->kobj_unregister);
+}
+
+static void btrfs_root_release(struct kobject *kobj)
+{
+	struct btrfs_root *root = container_of(kobj, struct btrfs_root,
+						root_kobj);
+	complete(&root->kobj_unregister);
+}
+
+static struct sysfs_ops btrfs_super_attr_ops = {
+	.show	= btrfs_super_attr_show,
+	.store	= btrfs_super_attr_store,
+};
+
+static struct sysfs_ops btrfs_root_attr_ops = {
+	.show	= btrfs_root_attr_show,
+	.store	= btrfs_root_attr_store,
+};
+
+static struct kobj_type btrfs_root_ktype = {
+	.default_attrs	= btrfs_root_attrs,
+	.sysfs_ops	= &btrfs_root_attr_ops,
+	.release	= btrfs_root_release,
+};
+
+static struct kobj_type btrfs_super_ktype = {
+	.default_attrs	= btrfs_super_attrs,
+	.sysfs_ops	= &btrfs_super_attr_ops,
+	.release	= btrfs_super_release,
+};
+
+static struct kset btrfs_kset = {
+	.kobj	= {.name = "btrfs"},
+};
+
+int btrfs_sysfs_add_super(struct btrfs_fs_info *fs)
+{
+	int error;
+
+	fs->super_kobj.kset = &btrfs_kset;
+	fs->super_kobj.ktype = &btrfs_super_ktype;
+
+	error = kobject_set_name(&fs->super_kobj, "%s",
+				 fs->sb->s_id);
+	if (error)
+		goto fail;
+
+	error = kobject_register(&fs->super_kobj);
+	if (error)
+		goto fail;
+
+	return 0;
+
+fail:
+	printk(KERN_ERR "btrfs: sysfs creation for super failed\n");
+	return error;
+}
+
+int btrfs_sysfs_add_root(struct btrfs_root *root)
+{
+	int error;
+
+	root->root_kobj.ktype = &btrfs_root_ktype;
+	root->root_kobj.parent = &root->fs_info->super_kobj;
+
+	error = kobject_set_name(&root->root_kobj, "%s", root->name);
+	if (error) {
+		goto fail;
+	}
+
+	error = kobject_register(&root->root_kobj);
+	if (error)
+		goto fail;
+
+	return 0;
+
+fail:
+	printk(KERN_ERR "btrfs: sysfs creation for root failed\n");
+	return error;
+}
+
+void btrfs_sysfs_del_root(struct btrfs_root *root)
+{
+	kobject_unregister(&root->root_kobj);
+	wait_for_completion(&root->kobj_unregister);
+}
+
+void btrfs_sysfs_del_super(struct btrfs_fs_info *fs)
+{
+	kobject_unregister(&fs->super_kobj);
+	wait_for_completion(&fs->kobj_unregister);
+}
+
+int btrfs_init_sysfs()
+{
+	kobj_set_kset_s(&btrfs_kset, fs_subsys);
+	return kset_register(&btrfs_kset);
+}
+
+void btrfs_exit_sysfs()
+{
+	kset_unregister(&btrfs_kset);
+}
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 18abea802794..2b15daa3a9f2 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -236,6 +236,7 @@ static int wait_for_commit(struct btrfs_root *root,
 struct dirty_root {
 	struct list_head list;
 	struct btrfs_root *root;
+	struct btrfs_root *latest_root;
 };
 
 int btrfs_add_dead_root(struct btrfs_root *root, struct list_head *dead_list)
@@ -278,6 +279,15 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans,
 					btrfs_root_blocknr(&root->root_item));
 				brelse(root->commit_root);
 				root->commit_root = NULL;
+
+				/* make sure to update the root on disk
+				 * so we get any updates to the block used
+				 * counts
+				 */
+				err = btrfs_update_root(trans,
+						root->fs_info->tree_root,
+						&root->root_key,
+						&root->root_item);
 				continue;
 			}
 			dirty = kmalloc(sizeof(*dirty), GFP_NOFS);
@@ -291,6 +301,7 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans,
 
 			memcpy(dirty->root, root, sizeof(*root));
 			dirty->root->node = root->commit_root;
+			dirty->latest_root = root;
 			root->commit_root = NULL;
 
 			root->root_key.offset = root->fs_info->generation;
@@ -384,20 +395,29 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 {
 	struct dirty_root *dirty;
 	struct btrfs_trans_handle *trans;
+	u64 num_blocks;
+	u64 blocks_used;
 	int ret = 0;
 	int err;
 
 	while(!list_empty(list)) {
+		struct btrfs_root *root;
+
 		mutex_lock(&tree_root->fs_info->fs_mutex);
 		dirty = list_entry(list->next, struct dirty_root, list);
 		list_del_init(&dirty->list);
 
+		num_blocks = btrfs_root_blocks_used(&dirty->root->root_item);
+		root = dirty->latest_root;
+
 		while(1) {
 			trans = btrfs_start_transaction(tree_root, 1);
+
 			ret = btrfs_drop_snapshot(trans, dirty->root);
 			if (ret != -EAGAIN) {
 				break;
 			}
+
 			err = btrfs_update_root(trans,
 					tree_root,
 					&dirty->root->root_key,
@@ -414,9 +434,19 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 			mutex_lock(&tree_root->fs_info->fs_mutex);
 		}
 		BUG_ON(ret);
+
+		num_blocks -= btrfs_root_blocks_used(&dirty->root->root_item);
+		blocks_used = btrfs_root_blocks_used(&root->root_item);
+		if (num_blocks) {
+			record_root_in_trans(root);
+			btrfs_set_root_blocks_used(&root->root_item,
+						   blocks_used - num_blocks);
+		}
 		ret = btrfs_del_root(trans, tree_root, &dirty->root->root_key);
-		if (ret)
+		if (ret) {
+			BUG();
 			break;
+		}
 		ret = btrfs_end_transaction(trans, tree_root);
 		BUG_ON(ret);
 
@@ -534,10 +564,12 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	wake_up(&cur_trans->commit_wait);
 	put_transaction(cur_trans);
 	put_transaction(cur_trans);
+
 	if (root->fs_info->closing)
 		list_splice_init(&root->fs_info->dead_roots, &dirty_fs_roots);
 	else
 		list_splice_init(&dirty_fs_roots, &root->fs_info->dead_roots);
+
 	mutex_unlock(&root->fs_info->trans_mutex);
 	kmem_cache_free(btrfs_trans_handle_cachep, trans);
 
-- 
cgit v1.2.3


From 86479a04eef8f304a13aeb8b64bcc8e506a68268 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Mon, 10 Sep 2007 19:58:16 -0400
Subject: Add support for defragging files via btrfsctl -d.  Avoid OOM on
 extent tree defrag.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.c       |  4 +++
 fs/btrfs/disk-io.h     |  2 ++
 fs/btrfs/extent_map.c  | 16 +++++----
 fs/btrfs/inode.c       | 93 +++++++++++++++++++++++++++++++++++++++++++-------
 fs/btrfs/tree-defrag.c |  3 ++
 5 files changed, 100 insertions(+), 18 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 89e282c35ce2..b41f48ade419 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -217,6 +217,9 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 		       root->fs_info->generation);
 		WARN_ON(1);
 	}
+	if (buffer_defrag_done(parent))
+		return 0;
+
 	parent_node = btrfs_buffer_node(parent);
 	parent_nritems = btrfs_header_nritems(&parent_node->header);
 	parent_level = btrfs_header_level(&parent_node->header);
@@ -274,6 +277,7 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 		*last_ret = search_start;
 		if (parent_level == 1)
 			clear_buffer_defrag(tmp_bh);
+		set_buffer_defrag_done(tmp_bh);
 		brelse(tmp_bh);
 	}
 	return err;
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index da6bb72750f2..714fa877438b 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -26,9 +26,11 @@
 enum btrfs_bh_state_bits {
 	BH_Checked = BH_PrivateStart,
 	BH_Defrag,
+	BH_DefragDone,
 };
 BUFFER_FNS(Checked, checked);
 BUFFER_FNS(Defrag, defrag);
+BUFFER_FNS(DefragDone, defrag_done);
 
 static inline struct btrfs_node *btrfs_buffer_node(struct buffer_head *bh)
 {
diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c
index 8ad6f8efc5a0..33f7a18dddf7 100644
--- a/fs/btrfs/extent_map.c
+++ b/fs/btrfs/extent_map.c
@@ -10,6 +10,12 @@
 #include <linux/blkdev.h>
 #include "extent_map.h"
 
+/* temporary define until extent_map moves out of btrfs */
+struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
+				       unsigned long extra_flags,
+				       void (*ctor)(void *, struct kmem_cache *,
+						    unsigned long));
+
 static struct kmem_cache *extent_map_cache;
 static struct kmem_cache *extent_state_cache;
 
@@ -32,14 +38,12 @@ struct tree_entry {
 
 void __init extent_map_init(void)
 {
-	extent_map_cache = kmem_cache_create("extent_map",
-					    sizeof(struct extent_map), 0,
-					    SLAB_RECLAIM_ACCOUNT |
+	extent_map_cache = btrfs_cache_create("extent_map",
+					    sizeof(struct extent_map),
 					    SLAB_DESTROY_BY_RCU,
 					    NULL);
-	extent_state_cache = kmem_cache_create("extent_state",
-					    sizeof(struct extent_state), 0,
-					    SLAB_RECLAIM_ACCOUNT |
+	extent_state_cache = btrfs_cache_create("extent_state",
+					    sizeof(struct extent_state),
 					    SLAB_DESTROY_BY_RCU,
 					    NULL);
 }
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 64710fa77d01..6b3e4404dc6a 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -1904,6 +1904,70 @@ fail:
 	return ret;
 }
 
+static unsigned long force_ra(struct address_space *mapping,
+			      struct file_ra_state *ra, struct file *file,
+			      pgoff_t offset, pgoff_t last_index)
+{
+	pgoff_t req_size;
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
+	req_size = last_index - offset + 1;
+	offset = page_cache_readahead(mapping, ra, file, offset, req_size);
+	return offset;
+#else
+	req_size = min(last_index - offset + 1, (pgoff_t)128);
+	page_cache_sync_readahead(mapping, ra, file, offset, req_size);
+	return offset + req_size;
+#endif
+}
+
+int btrfs_defrag_file(struct file *file) {
+	struct inode *inode = file->f_path.dentry->d_inode;
+	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+	struct page *page;
+	unsigned long last_index;
+	unsigned long ra_index = 0;
+	u64 page_start;
+	u64 page_end;
+	unsigned long i;
+
+	mutex_lock(&inode->i_mutex);
+	last_index = inode->i_size >> PAGE_CACHE_SHIFT;
+	for (i = 0; i <= last_index; i++) {
+		if (i == ra_index) {
+			ra_index = force_ra(inode->i_mapping, &file->f_ra,
+					    file, ra_index, last_index);
+		}
+		page = grab_cache_page(inode->i_mapping, i);
+		if (!page)
+			goto out_unlock;
+		if (!PageUptodate(page)) {
+			btrfs_readpage(NULL, page);
+			lock_page(page);
+			if (!PageUptodate(page)) {
+				unlock_page(page);
+				page_cache_release(page);
+				goto out_unlock;
+			}
+		}
+		page_start = page->index << PAGE_CACHE_SHIFT;
+		page_end = page_start + PAGE_CACHE_SIZE - 1;
+
+		lock_extent(em_tree, page_start, page_end, GFP_NOFS);
+		set_extent_delalloc(em_tree, page_start,
+				    page_end, GFP_NOFS);
+		unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
+		set_page_dirty(page);
+		unlock_page(page);
+		page_cache_release(page);
+		balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
+	}
+
+out_unlock:
+	mutex_unlock(&inode->i_mutex);
+	return 0;
+}
+
 int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int
 		cmd, unsigned long arg)
 {
@@ -1948,10 +2012,14 @@ int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int
 		break;
 
 	case BTRFS_IOC_DEFRAG:
-		mutex_lock(&root->fs_info->fs_mutex);
-		btrfs_defrag_root(root, 0);
-		btrfs_defrag_root(root->fs_info->extent_root, 0);
-		mutex_unlock(&root->fs_info->fs_mutex);
+		if (S_ISDIR(inode->i_mode)) {
+			mutex_lock(&root->fs_info->fs_mutex);
+			btrfs_defrag_root(root, 0);
+			btrfs_defrag_root(root->fs_info->extent_root, 0);
+			mutex_unlock(&root->fs_info->fs_mutex);
+		} else if (S_ISREG(inode->i_mode)) {
+			btrfs_defrag_file(filp);
+		}
 		ret = 0;
 		break;
 	default:
@@ -2018,7 +2086,7 @@ void btrfs_destroy_cachep(void)
 		kmem_cache_destroy(btrfs_path_cachep);
 }
 
-static struct kmem_cache *cache_create(const char *name, size_t size,
+struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
 				       unsigned long extra_flags,
 				       void (*ctor)(void *, struct kmem_cache *,
 						    unsigned long))
@@ -2033,27 +2101,28 @@ static struct kmem_cache *cache_create(const char *name, size_t size,
 
 int btrfs_init_cachep(void)
 {
-	btrfs_inode_cachep = cache_create("btrfs_inode_cache",
+	btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
 					  sizeof(struct btrfs_inode),
 					  0, init_once);
 	if (!btrfs_inode_cachep)
 		goto fail;
-	btrfs_trans_handle_cachep = cache_create("btrfs_trans_handle_cache",
-					     sizeof(struct btrfs_trans_handle),
-					     0, NULL);
+	btrfs_trans_handle_cachep =
+			btrfs_cache_create("btrfs_trans_handle_cache",
+					   sizeof(struct btrfs_trans_handle),
+					   0, NULL);
 	if (!btrfs_trans_handle_cachep)
 		goto fail;
-	btrfs_transaction_cachep = cache_create("btrfs_transaction_cache",
+	btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
 					     sizeof(struct btrfs_transaction),
 					     0, NULL);
 	if (!btrfs_transaction_cachep)
 		goto fail;
-	btrfs_path_cachep = cache_create("btrfs_path_cache",
+	btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
 					 sizeof(struct btrfs_transaction),
 					 0, NULL);
 	if (!btrfs_path_cachep)
 		goto fail;
-	btrfs_bit_radix_cachep = cache_create("btrfs_radix", 256,
+	btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
 					      SLAB_DESTROY_BY_RCU, NULL);
 	if (!btrfs_bit_radix_cachep)
 		goto fail;
diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c
index 35fd20d24645..420597127ed1 100644
--- a/fs/btrfs/tree-defrag.c
+++ b/fs/btrfs/tree-defrag.c
@@ -113,6 +113,8 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 	}
 	WARN_ON(*level < 0);
 	WARN_ON(*level >= BTRFS_MAX_LEVEL);
+	clear_buffer_defrag(path->nodes[*level]);
+	clear_buffer_defrag_done(path->nodes[*level]);
 	btrfs_block_release(root, path->nodes[*level]);
 	path->nodes[*level] = NULL;
 	*level += 1;
@@ -143,6 +145,7 @@ static int defrag_walk_up(struct btrfs_trans_handle *trans,
 			return 0;
 		} else {
 			clear_buffer_defrag(path->nodes[*level]);
+			clear_buffer_defrag_done(path->nodes[*level]);
 			btrfs_block_release(root, path->nodes[*level]);
 			path->nodes[*level] = NULL;
 			*level = i + 1;
-- 
cgit v1.2.3


From d3c2fdcf7b79079f60ac64e61d886964d4647910 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Mon, 17 Sep 2007 10:58:06 -0400
Subject: Btrfs: Use balance_dirty_pages_nr on btree blocks

btrfs_btree_balance_dirty is changed to pass the number of pages dirtied
for more accurate dirty throttling.  This lets the VM make better decisions
about when to force some writeback.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/disk-io.c     |  5 +++--
 fs/btrfs/disk-io.h     |  2 +-
 fs/btrfs/extent-tree.c |  1 +
 fs/btrfs/file.c        |  2 +-
 fs/btrfs/inode.c       | 47 +++++++++++++++++++++++++++++++++++------------
 fs/btrfs/transaction.c | 14 ++++++++++----
 6 files changed, 51 insertions(+), 20 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 4296839eea1a..ef6d76f2ec20 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -666,7 +666,8 @@ void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf)
 	brelse(buf);
 }
 
-void btrfs_btree_balance_dirty(struct btrfs_root *root)
+void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
 {
-	balance_dirty_pages_ratelimited(root->fs_info->btree_inode->i_mapping);
+	balance_dirty_pages_ratelimited_nr(
+			root->fs_info->btree_inode->i_mapping, nr);
 }
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 714fa877438b..fc6b520c6e0b 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -80,7 +80,7 @@ int btrfs_insert_dev_radix(struct btrfs_root *root,
 			   u64 num_blocks);
 int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
 			     u64 logical);
-void btrfs_btree_balance_dirty(struct btrfs_root *root);
+void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr);
 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root);
 void btrfs_mark_buffer_dirty(struct buffer_head *bh);
 #endif
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 9151850266b0..fe772f9b06cf 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1275,6 +1275,7 @@ struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 	set_buffer_checked(buf);
 	set_buffer_defrag(buf);
 	set_radix_bit(&trans->transaction->dirty_pages, buf->b_page->index);
+	trans->blocks_used++;
 	return buf;
 }
 
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 26826a543355..698eaea612f1 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -655,7 +655,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 		num_written += write_bytes;
 
 		balance_dirty_pages_ratelimited_nr(inode->i_mapping, num_pages);
-		btrfs_btree_balance_dirty(root);
+		btrfs_btree_balance_dirty(root, 1);
 		cond_resched();
 	}
 	mutex_unlock(&inode->i_mutex);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index e24b875327da..bd00df33fb3f 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -387,15 +387,17 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
 	struct btrfs_root *root;
 	struct btrfs_trans_handle *trans;
 	int ret;
+	unsigned long nr;
 
 	root = BTRFS_I(dir)->root;
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
 	btrfs_set_trans_block_group(trans, dir);
 	ret = btrfs_unlink_trans(trans, root, dir, dentry);
+	nr = trans->blocks_used;
 	btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
-	btrfs_btree_balance_dirty(root);
+	btrfs_btree_balance_dirty(root, nr);
 	return ret;
 }
 
@@ -412,6 +414,7 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
 	int found_type;
 	struct btrfs_leaf *leaf;
 	char *goodnames = "..";
+	unsigned long nr;
 
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
@@ -467,8 +470,9 @@ out:
 	btrfs_release_path(root, path);
 	btrfs_free_path(path);
 	mutex_unlock(&root->fs_info->fs_mutex);
+	nr = trans->blocks_used;
 	ret = btrfs_end_transaction(trans, root);
-	btrfs_btree_balance_dirty(root);
+	btrfs_btree_balance_dirty(root, nr);
 	if (ret && !err)
 		err = ret;
 	return err;
@@ -764,6 +768,7 @@ void btrfs_delete_inode(struct inode *inode)
 {
 	struct btrfs_trans_handle *trans;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
+	unsigned long nr;
 	int ret;
 
 	truncate_inode_pages(&inode->i_data, 0);
@@ -780,15 +785,17 @@ void btrfs_delete_inode(struct inode *inode)
 	ret = btrfs_free_inode(trans, root, inode);
 	if (ret)
 		goto no_delete_lock;
+	nr = trans->blocks_used;
 	btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
-	btrfs_btree_balance_dirty(root);
+	btrfs_btree_balance_dirty(root, nr);
 	return;
 
 no_delete_lock:
+	nr = trans->blocks_used;
 	btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
-	btrfs_btree_balance_dirty(root);
+	btrfs_btree_balance_dirty(root, nr);
 no_delete:
 	clear_inode(inode);
 }
@@ -1165,6 +1172,7 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
 	int err;
 	int drop_inode = 0;
 	u64 objectid;
+	unsigned long nr;
 
 	if (!new_valid_dev(rdev))
 		return -EINVAL;
@@ -1198,6 +1206,7 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
 	btrfs_update_inode_block_group(trans, inode);
 	btrfs_update_inode_block_group(trans, dir);
 out_unlock:
+	nr = trans->blocks_used;
 	btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
 
@@ -1205,7 +1214,7 @@ out_unlock:
 		inode_dec_link_count(inode);
 		iput(inode);
 	}
-	btrfs_btree_balance_dirty(root);
+	btrfs_btree_balance_dirty(root, nr);
 	return err;
 }
 
@@ -1217,6 +1226,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry,
 	struct inode *inode;
 	int err;
 	int drop_inode = 0;
+	unsigned long nr;
 	u64 objectid;
 
 	mutex_lock(&root->fs_info->fs_mutex);
@@ -1251,6 +1261,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry,
 	btrfs_update_inode_block_group(trans, inode);
 	btrfs_update_inode_block_group(trans, dir);
 out_unlock:
+	nr = trans->blocks_used;
 	btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
 
@@ -1258,7 +1269,7 @@ out_unlock:
 		inode_dec_link_count(inode);
 		iput(inode);
 	}
-	btrfs_btree_balance_dirty(root);
+	btrfs_btree_balance_dirty(root, nr);
 	return err;
 }
 
@@ -1268,6 +1279,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
 	struct btrfs_trans_handle *trans;
 	struct btrfs_root *root = BTRFS_I(dir)->root;
 	struct inode *inode = old_dentry->d_inode;
+	unsigned long nr;
 	int err;
 	int drop_inode = 0;
 
@@ -1288,6 +1300,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
 	if (err)
 		drop_inode = 1;
 
+	nr = trans->blocks_used;
 	btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
 
@@ -1295,7 +1308,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
 		inode_dec_link_count(inode);
 		iput(inode);
 	}
-	btrfs_btree_balance_dirty(root);
+	btrfs_btree_balance_dirty(root, nr);
 	return err;
 }
 
@@ -1336,6 +1349,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 	int err = 0;
 	int drop_on_err = 0;
 	u64 objectid;
+	unsigned long nr = 1;
 
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
@@ -1380,12 +1394,13 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 	btrfs_update_inode_block_group(trans, dir);
 
 out_fail:
+	nr = trans->blocks_used;
 	btrfs_end_transaction(trans, root);
 out_unlock:
 	mutex_unlock(&root->fs_info->fs_mutex);
 	if (drop_on_err)
 		iput(inode);
-	btrfs_btree_balance_dirty(root);
+	btrfs_btree_balance_dirty(root, nr);
 	return err;
 }
 
@@ -1682,6 +1697,7 @@ static void btrfs_truncate(struct inode *inode)
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	int ret;
 	struct btrfs_trans_handle *trans;
+	unsigned long nr;
 
 	if (!S_ISREG(inode->i_mode))
 		return;
@@ -1697,10 +1713,11 @@ static void btrfs_truncate(struct inode *inode)
 	/* FIXME, add redo link to tree so we don't leak on crash */
 	ret = btrfs_truncate_in_trans(trans, root, inode);
 	btrfs_update_inode(trans, root, inode);
+	nr = trans->blocks_used;
 	ret = btrfs_end_transaction(trans, root);
 	BUG_ON(ret);
 	mutex_unlock(&root->fs_info->fs_mutex);
-	btrfs_btree_balance_dirty(root);
+	btrfs_btree_balance_dirty(root, nr);
 }
 
 int btrfs_commit_write(struct file *file, struct page *page,
@@ -1725,6 +1742,7 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 	int err;
 	u64 objectid;
 	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
+	unsigned long nr = 1;
 
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
@@ -1814,12 +1832,13 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 	if (ret)
 		goto fail;
 fail:
+	nr = trans->blocks_used;
 	err = btrfs_commit_transaction(trans, root);
 	if (err && !ret)
 		ret = err;
 fail_commit:
 	mutex_unlock(&root->fs_info->fs_mutex);
-	btrfs_btree_balance_dirty(root);
+	btrfs_btree_balance_dirty(root, nr);
 	return ret;
 }
 
@@ -1832,6 +1851,7 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 	int ret;
 	int err;
 	u64 objectid;
+	unsigned long nr;
 
 	if (!root->ref_cows)
 		return -EINVAL;
@@ -1884,12 +1904,13 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 	if (ret)
 		goto fail;
 fail:
+	nr = trans->blocks_used;
 	err = btrfs_commit_transaction(trans, root);
 	if (err && !ret)
 		ret = err;
 	mutex_unlock(&root->fs_info->fs_mutex);
 	up_write(&root->snap_sem);
-	btrfs_btree_balance_dirty(root);
+	btrfs_btree_balance_dirty(root, nr);
 	return ret;
 }
 
@@ -2240,6 +2261,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
 	int datasize;
 	char *ptr;
 	struct btrfs_file_extent_item *ei;
+	unsigned long nr;
 
 	name_len = strlen(symname) + 1;
 	if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
@@ -2309,13 +2331,14 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
 		drop_inode = 1;
 
 out_unlock:
+	nr = trans->blocks_used;
 	btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
 	if (drop_inode) {
 		inode_dec_link_count(inode);
 		iput(inode);
 	}
-	btrfs_btree_balance_dirty(root);
+	btrfs_btree_balance_dirty(root, nr);
 	return err;
 }
 
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 29755593de65..087074db0bd5 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -18,6 +18,7 @@
 
 #include <linux/fs.h>
 #include <linux/sched.h>
+#include <linux/writeback.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
@@ -340,6 +341,7 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly)
 	struct btrfs_fs_info *info = root->fs_info;
 	int ret;
 	struct btrfs_trans_handle *trans;
+	unsigned long nr;
 
 	if (root->defrag_running)
 		return 0;
@@ -348,10 +350,11 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly)
 	while (1) {
 		root->defrag_running = 1;
 		ret = btrfs_defrag_leaves(trans, root, cacheonly);
+		nr = trans->blocks_used;
 		btrfs_end_transaction(trans, root);
 		mutex_unlock(&info->fs_mutex);
 
-		btrfs_btree_balance_dirty(root);
+		btrfs_btree_balance_dirty(info->tree_root, nr);
 		cond_resched();
 
 		mutex_lock(&info->fs_mutex);
@@ -398,6 +401,7 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 {
 	struct dirty_root *dirty;
 	struct btrfs_trans_handle *trans;
+	unsigned long nr;
 	u64 num_blocks;
 	u64 blocks_used;
 	int ret = 0;
@@ -426,11 +430,11 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 					&dirty->root->root_item);
 			if (err)
 				ret = err;
+			nr = trans->blocks_used;
 			ret = btrfs_end_transaction(trans, tree_root);
 			BUG_ON(ret);
 			mutex_unlock(&tree_root->fs_info->fs_mutex);
-
-			btrfs_btree_balance_dirty(tree_root);
+			btrfs_btree_balance_dirty(tree_root, nr);
 			schedule();
 
 			mutex_lock(&tree_root->fs_info->fs_mutex);
@@ -449,13 +453,15 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 			BUG();
 			break;
 		}
+		nr = trans->blocks_used;
 		ret = btrfs_end_transaction(trans, tree_root);
 		BUG_ON(ret);
 
 		kfree(dirty->root);
 		kfree(dirty);
 		mutex_unlock(&tree_root->fs_info->fs_mutex);
-		btrfs_btree_balance_dirty(tree_root);
+
+		btrfs_btree_balance_dirty(tree_root, nr);
 		schedule();
 	}
 	return ret;
-- 
cgit v1.2.3


From 5f39d397dfbe140a14edecd4e73c34ce23c4f9ee Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Mon, 15 Oct 2007 16:14:19 -0400
Subject: Btrfs: Create extent_buffer interface for large blocksizes

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/Makefile      |    2 +
 fs/btrfs/ctree.c       | 1484 ++++++++++++++++++++++++++----------------------
 fs/btrfs/ctree.h       |  849 ++++++++++-----------------
 fs/btrfs/dir-item.c    |  102 ++--
 fs/btrfs/disk-io.c     |  391 ++++++-------
 fs/btrfs/disk-io.h     |   55 +-
 fs/btrfs/extent-tree.c |  294 +++++-----
 fs/btrfs/extent_map.c  |  672 +++++++++++++++++++++-
 fs/btrfs/extent_map.h  |   75 ++-
 fs/btrfs/file-item.c   |   63 +-
 fs/btrfs/file.c        |  129 +++--
 fs/btrfs/inode-item.c  |   22 +-
 fs/btrfs/inode-map.c   |   17 +-
 fs/btrfs/inode.c       |  435 ++++++++------
 fs/btrfs/print-tree.c  |   96 ++--
 fs/btrfs/print-tree.h  |    4 +-
 fs/btrfs/root-tree.c   |   51 +-
 fs/btrfs/super.c       |    7 +-
 fs/btrfs/sysfs.c       |   10 +-
 fs/btrfs/transaction.c |   61 +-
 fs/btrfs/transaction.h |    5 +-
 fs/btrfs/tree-defrag.c |   58 +-
 22 files changed, 2794 insertions(+), 2088 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 6341cf7a9f91..a3b51085d7f0 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -23,4 +23,6 @@ modules_install:
 clean:
 	$(MAKE) -C $(KERNELDIR) M=`pwd` clean
 
+tester:
+	$(MAKE) -C $(KERNELDIR) M=`pwd` tree-defrag.o transaction.o sysfs.o super.o root-tree.o inode-map.o inode-item.o inode.o file-item.o file.o extent_map.o disk-io.o ctree.o dir-item.o extent-tree.o
 endif
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index b41f48ade419..f60920e8a0e0 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -16,21 +16,24 @@
  * Boston, MA 021110-1307, USA.
  */
 
+#include <linux/highmem.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
+#include "print-tree.h"
 
 static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_path *path, int level);
 static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_key *ins_key,
 		      struct btrfs_path *path, int data_size);
-static int push_node_left(struct btrfs_trans_handle *trans, struct btrfs_root
-			  *root, struct buffer_head *dst, struct buffer_head
-			  *src);
-static int balance_node_right(struct btrfs_trans_handle *trans, struct
-			      btrfs_root *root, struct buffer_head *dst_buf,
-			      struct buffer_head *src_buf);
+static int push_node_left(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root, struct extent_buffer *dst,
+			  struct extent_buffer *src);
+static int balance_node_right(struct btrfs_trans_handle *trans,
+			      struct btrfs_root *root,
+			      struct extent_buffer *dst_buf,
+			      struct extent_buffer *src_buf);
 static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		   struct btrfs_path *path, int level, int slot);
 
@@ -62,40 +65,38 @@ void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p)
 	for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
 		if (!p->nodes[i])
 			break;
-		btrfs_block_release(root, p->nodes[i]);
+		free_extent_buffer(p->nodes[i]);
 	}
 	memset(p, 0, sizeof(*p));
 }
 
-static int __btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root
-			   *root, struct buffer_head *buf, struct buffer_head
-			   *parent, int parent_slot, struct buffer_head
-			   **cow_ret, u64 search_start, u64 empty_size)
+static int __btrfs_cow_block(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root,
+			     struct extent_buffer *buf,
+			     struct extent_buffer *parent, int parent_slot,
+			     struct extent_buffer **cow_ret,
+			     u64 search_start, u64 empty_size)
 {
-	struct buffer_head *cow;
-	struct btrfs_node *cow_node;
+	struct extent_buffer *cow;
 	int ret = 0;
 	int different_trans = 0;
 
 	WARN_ON(root->ref_cows && trans->transid != root->last_trans);
-	WARN_ON(!buffer_uptodate(buf));
+
 	cow = btrfs_alloc_free_block(trans, root, search_start, empty_size);
 	if (IS_ERR(cow))
 		return PTR_ERR(cow);
 
-	cow_node = btrfs_buffer_node(cow);
-	if (buf->b_size != root->blocksize || cow->b_size != root->blocksize)
+	if (buf->len != root->sectorsize || cow->len != root->sectorsize)
 		WARN_ON(1);
 
-	memcpy(cow_node, btrfs_buffer_node(buf), root->blocksize);
-	btrfs_set_header_blocknr(&cow_node->header, bh_blocknr(cow));
-	btrfs_set_header_generation(&cow_node->header, trans->transid);
-	btrfs_set_header_owner(&cow_node->header, root->root_key.objectid);
+	copy_extent_buffer(cow, buf, 0, 0, cow->len);
+	btrfs_set_header_blocknr(cow, extent_buffer_blocknr(cow));
+	btrfs_set_header_generation(cow, trans->transid);
+	btrfs_set_header_owner(cow, root->root_key.objectid);
 
-	WARN_ON(btrfs_header_generation(btrfs_buffer_header(buf)) >
-		trans->transid);
-	if (btrfs_header_generation(btrfs_buffer_header(buf)) !=
-				    trans->transid) {
+	WARN_ON(btrfs_header_generation(buf) > trans->transid);
+	if (btrfs_header_generation(buf) != trans->transid) {
 		different_trans = 1;
 		ret = btrfs_inc_ref(trans, root, buf);
 		if (ret)
@@ -106,29 +107,29 @@ static int __btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	if (buf == root->node) {
 		root->node = cow;
-		get_bh(cow);
+		extent_buffer_get(cow);
 		if (buf != root->commit_root) {
-			btrfs_free_extent(trans, root, bh_blocknr(buf), 1, 1);
+			btrfs_free_extent(trans, root,
+					  extent_buffer_blocknr(buf), 1, 1);
 		}
-		btrfs_block_release(root, buf);
+		free_extent_buffer(buf);
 	} else {
-		btrfs_set_node_blockptr(btrfs_buffer_node(parent), parent_slot,
-					bh_blocknr(cow));
+		btrfs_set_node_blockptr(parent, parent_slot,
+					extent_buffer_blocknr(cow));
 		btrfs_mark_buffer_dirty(parent);
-		WARN_ON(btrfs_header_generation(btrfs_buffer_header(parent)) !=
-				    trans->transid);
-		btrfs_free_extent(trans, root, bh_blocknr(buf), 1, 1);
+		WARN_ON(btrfs_header_generation(parent) != trans->transid);
+		btrfs_free_extent(trans, root, extent_buffer_blocknr(buf),1,1);
 	}
-	btrfs_block_release(root, buf);
+	free_extent_buffer(buf);
 	btrfs_mark_buffer_dirty(cow);
 	*cow_ret = cow;
 	return 0;
 }
 
-int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root
-			   *root, struct buffer_head *buf, struct buffer_head
-			   *parent, int parent_slot, struct buffer_head
-			   **cow_ret)
+int btrfs_cow_block(struct btrfs_trans_handle *trans,
+		    struct btrfs_root *root, struct extent_buffer *buf,
+		    struct extent_buffer *parent, int parent_slot,
+		    struct extent_buffer **cow_ret)
 {
 	u64 search_start;
 	if (trans->transaction != root->fs_info->running_transaction) {
@@ -141,13 +142,12 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root
 		       root->fs_info->generation);
 		WARN_ON(1);
 	}
-	if (btrfs_header_generation(btrfs_buffer_header(buf)) ==
-				    trans->transid) {
+	if (btrfs_header_generation(buf) == trans->transid) {
 		*cow_ret = buf;
 		return 0;
 	}
 
-	search_start = bh_blocknr(buf) & ~((u64)65535);
+	search_start = extent_buffer_blocknr(buf) & ~((u64)65535);
 	return __btrfs_cow_block(trans, root, buf, parent,
 				 parent_slot, cow_ret, search_start, 0);
 }
@@ -161,9 +161,11 @@ static int close_blocks(u64 blocknr, u64 other)
 	return 0;
 }
 
-static int should_defrag_leaf(struct buffer_head *bh)
+#if 0
+static int should_defrag_leaf(struct extent_buffer *eb)
 {
-	struct btrfs_leaf *leaf = btrfs_buffer_leaf(bh);
+	return 0;
+	struct btrfs_leaf *leaf = btrfs_buffer_leaf(eb);
 	struct btrfs_disk_key *key;
 	u32 nritems;
 
@@ -188,14 +190,17 @@ static int should_defrag_leaf(struct buffer_head *bh)
 	}
 	return 0;
 }
+#endif
 
 int btrfs_realloc_node(struct btrfs_trans_handle *trans,
-		       struct btrfs_root *root, struct buffer_head *parent,
+		       struct btrfs_root *root, struct extent_buffer *parent,
 		       int cache_only, u64 *last_ret)
 {
+	return 0;
+#if 0
 	struct btrfs_node *parent_node;
-	struct buffer_head *cur_bh;
-	struct buffer_head *tmp_bh;
+	struct extent_buffer *cur_eb;
+	struct extent_buffer *tmp_eb;
 	u64 blocknr;
 	u64 search_start = *last_ret;
 	u64 last_block = 0;
@@ -281,6 +286,7 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 		brelse(tmp_bh);
 	}
 	return err;
+#endif
 }
 
 /*
@@ -289,12 +295,12 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
  * which is the stop of the leaf data stack
  */
 static inline unsigned int leaf_data_end(struct btrfs_root *root,
-					 struct btrfs_leaf *leaf)
+					 struct extent_buffer *leaf)
 {
-	u32 nr = btrfs_header_nritems(&leaf->header);
+	u32 nr = btrfs_header_nritems(leaf);
 	if (nr == 0)
 		return BTRFS_LEAF_DATA_SIZE(root);
-	return btrfs_item_offset(leaf->items + nr - 1);
+	return btrfs_item_offset_nr(leaf, nr - 1);
 }
 
 /*
@@ -310,9 +316,9 @@ static int comp_keys(struct btrfs_disk_key *disk, struct btrfs_key *k2)
 		return 1;
 	if (k1.objectid < k2->objectid)
 		return -1;
-	if (k1.flags > k2->flags)
+	if (k1.type > k2->type)
 		return 1;
-	if (k1.flags < k2->flags)
+	if (k1.type < k2->type)
 		return -1;
 	if (k1.offset > k2->offset)
 		return 1;
@@ -324,37 +330,39 @@ static int comp_keys(struct btrfs_disk_key *disk, struct btrfs_key *k2)
 static int check_node(struct btrfs_root *root, struct btrfs_path *path,
 		      int level)
 {
-	struct btrfs_node *parent = NULL;
-	struct btrfs_node *node = btrfs_buffer_node(path->nodes[level]);
+	struct extent_buffer *parent = NULL;
+	struct extent_buffer *node = path->nodes[level];
+	struct btrfs_disk_key parent_key;
+	struct btrfs_disk_key node_key;
 	int parent_slot;
 	int slot;
 	struct btrfs_key cpukey;
-	u32 nritems = btrfs_header_nritems(&node->header);
+	u32 nritems = btrfs_header_nritems(node);
 
 	if (path->nodes[level + 1])
-		parent = btrfs_buffer_node(path->nodes[level + 1]);
+		parent = path->nodes[level + 1];
 
 	slot = path->slots[level];
-	BUG_ON(!buffer_uptodate(path->nodes[level]));
 	BUG_ON(nritems == 0);
 	if (parent) {
-		struct btrfs_disk_key *parent_key;
-
 		parent_slot = path->slots[level + 1];
-		parent_key = &parent->ptrs[parent_slot].key;
-		BUG_ON(memcmp(parent_key, &node->ptrs[0].key,
+		btrfs_node_key(parent, &parent_key, parent_slot);
+		btrfs_node_key(node, &node_key, 0);
+		BUG_ON(memcmp(&parent_key, &node_key,
 			      sizeof(struct btrfs_disk_key)));
 		BUG_ON(btrfs_node_blockptr(parent, parent_slot) !=
-		       btrfs_header_blocknr(&node->header));
+		       btrfs_header_blocknr(node));
 	}
 	BUG_ON(nritems > BTRFS_NODEPTRS_PER_BLOCK(root));
 	if (slot != 0) {
-		btrfs_disk_key_to_cpu(&cpukey, &node->ptrs[slot - 1].key);
-		BUG_ON(comp_keys(&node->ptrs[slot].key, &cpukey) <= 0);
+		btrfs_node_key_to_cpu(node, &cpukey, slot - 1);
+		btrfs_node_key(node, &node_key, slot);
+		BUG_ON(comp_keys(&node_key, &cpukey) <= 0);
 	}
 	if (slot < nritems - 1) {
-		btrfs_disk_key_to_cpu(&cpukey, &node->ptrs[slot + 1].key);
-		BUG_ON(comp_keys(&node->ptrs[slot].key, &cpukey) >= 0);
+		btrfs_node_key_to_cpu(node, &cpukey, slot + 1);
+		btrfs_node_key(node, &node_key, slot);
+		BUG_ON(comp_keys(&node_key, &cpukey) >= 0);
 	}
 	return 0;
 }
@@ -362,83 +370,172 @@ static int check_node(struct btrfs_root *root, struct btrfs_path *path,
 static int check_leaf(struct btrfs_root *root, struct btrfs_path *path,
 		      int level)
 {
-	struct btrfs_leaf *leaf = btrfs_buffer_leaf(path->nodes[level]);
-	struct btrfs_node *parent = NULL;
+	struct extent_buffer *leaf = path->nodes[level];
+	struct extent_buffer *parent = NULL;
 	int parent_slot;
-	int slot = path->slots[0];
 	struct btrfs_key cpukey;
+	struct btrfs_disk_key parent_key;
+	struct btrfs_disk_key leaf_key;
+	int slot = path->slots[0];
 
-	u32 nritems = btrfs_header_nritems(&leaf->header);
+	u32 nritems = btrfs_header_nritems(leaf);
 
 	if (path->nodes[level + 1])
-		parent = btrfs_buffer_node(path->nodes[level + 1]);
-
-	BUG_ON(btrfs_leaf_free_space(root, leaf) < 0);
+		parent = path->nodes[level + 1];
 
 	if (nritems == 0)
 		return 0;
 
 	if (parent) {
-		struct btrfs_disk_key *parent_key;
-
 		parent_slot = path->slots[level + 1];
-		parent_key = &parent->ptrs[parent_slot].key;
+		btrfs_node_key(parent, &parent_key, parent_slot);
+		btrfs_item_key(leaf, &leaf_key, 0);
 
-		BUG_ON(memcmp(parent_key, &leaf->items[0].key,
+		BUG_ON(memcmp(&parent_key, &leaf_key,
 		       sizeof(struct btrfs_disk_key)));
 		BUG_ON(btrfs_node_blockptr(parent, parent_slot) !=
-		       btrfs_header_blocknr(&leaf->header));
+		       btrfs_header_blocknr(leaf));
+	}
+#if 0
+	for (i = 0; nritems > 1 && i < nritems - 2; i++) {
+		btrfs_item_key_to_cpu(leaf, &cpukey, i + 1);
+		btrfs_item_key(leaf, &leaf_key, i);
+		if (comp_keys(&leaf_key, &cpukey) >= 0) {
+			btrfs_print_leaf(root, leaf);
+			printk("slot %d offset bad key\n", i);
+			BUG_ON(1);
+		}
+		if (btrfs_item_offset_nr(leaf, i) !=
+			btrfs_item_end_nr(leaf, i + 1)) {
+			btrfs_print_leaf(root, leaf);
+			printk("slot %d offset bad\n", i);
+			BUG_ON(1);
+		}
+		if (i == 0) {
+			if (btrfs_item_offset_nr(leaf, i) +
+			       btrfs_item_size_nr(leaf, i) !=
+			       BTRFS_LEAF_DATA_SIZE(root)) {
+				btrfs_print_leaf(root, leaf);
+				printk("slot %d first offset bad\n", i);
+				BUG_ON(1);
+			}
+		}
 	}
-	if (slot != 0) {
-		btrfs_disk_key_to_cpu(&cpukey, &leaf->items[slot - 1].key);
-		BUG_ON(comp_keys(&leaf->items[slot].key, &cpukey) <= 0);
-		BUG_ON(btrfs_item_offset(leaf->items + slot - 1) !=
-			btrfs_item_end(leaf->items + slot));
+	if (nritems > 0) {
+		if (btrfs_item_size_nr(leaf, nritems - 1) > 4096) {
+				btrfs_print_leaf(root, leaf);
+				printk("slot %d bad size \n", nritems - 1);
+				BUG_ON(1);
+		}
+	}
+#endif
+	if (slot != 0 && slot < nritems - 1) {
+		btrfs_item_key(leaf, &leaf_key, slot);
+		btrfs_item_key_to_cpu(leaf, &cpukey, slot - 1);
+		if (comp_keys(&leaf_key, &cpukey) <= 0) {
+			btrfs_print_leaf(root, leaf);
+			printk("slot %d offset bad key\n", slot);
+			BUG_ON(1);
+		}
+		if (btrfs_item_offset_nr(leaf, slot - 1) !=
+		       btrfs_item_end_nr(leaf, slot)) {
+			btrfs_print_leaf(root, leaf);
+			printk("slot %d offset bad\n", slot);
+			BUG_ON(1);
+		}
 	}
 	if (slot < nritems - 1) {
-		btrfs_disk_key_to_cpu(&cpukey, &leaf->items[slot + 1].key);
-		BUG_ON(comp_keys(&leaf->items[slot].key, &cpukey) >= 0);
-		BUG_ON(btrfs_item_offset(leaf->items + slot) !=
-			btrfs_item_end(leaf->items + slot + 1));
+		btrfs_item_key(leaf, &leaf_key, slot);
+		btrfs_item_key_to_cpu(leaf, &cpukey, slot + 1);
+		BUG_ON(comp_keys(&leaf_key, &cpukey) >= 0);
+		if (btrfs_item_offset_nr(leaf, slot) !=
+			btrfs_item_end_nr(leaf, slot + 1)) {
+			btrfs_print_leaf(root, leaf);
+			printk("slot %d offset bad\n", slot);
+			BUG_ON(1);
+		}
 	}
-	BUG_ON(btrfs_item_offset(leaf->items) +
-	       btrfs_item_size(leaf->items) != BTRFS_LEAF_DATA_SIZE(root));
+	BUG_ON(btrfs_item_offset_nr(leaf, 0) +
+	       btrfs_item_size_nr(leaf, 0) != BTRFS_LEAF_DATA_SIZE(root));
 	return 0;
 }
 
 static int check_block(struct btrfs_root *root, struct btrfs_path *path,
 			int level)
 {
-	struct btrfs_node *node = btrfs_buffer_node(path->nodes[level]);
-	if (memcmp(node->header.fsid, root->fs_info->disk_super->fsid,
-		   sizeof(node->header.fsid)))
-		BUG();
+	struct extent_buffer *buf = path->nodes[level];
+	char fsid[BTRFS_FSID_SIZE];
+
+	read_extent_buffer(buf, fsid, (unsigned long)btrfs_header_fsid(buf),
+			   BTRFS_FSID_SIZE);
+
+	if (memcmp(fsid, root->fs_info->fsid, BTRFS_FSID_SIZE)) {
+		int i = 0;
+		printk("warning bad block %Lu\n", buf->start);
+		if (!btrfs_buffer_uptodate(buf)) {
+			WARN_ON(1);
+		}
+		for (i = 0; i < BTRFS_FSID_SIZE; i++) {
+			printk("%x:%x ", root->fs_info->fsid[i], fsid[i]);
+		}
+		printk("\n");
+		// BUG();
+	}
 	if (level == 0)
 		return check_leaf(root, path, level);
 	return check_node(root, path, level);
 }
 
 /*
- * search for key in the array p.  items p are item_size apart
- * and there are 'max' items in p
+ * search for key in the extent_buffer.  The items start at offset p,
+ * and they are item_size apart.  There are 'max' items in p.
+ *
  * the slot in the array is returned via slot, and it points to
  * the place where you would insert key if it is not found in
  * the array.
  *
  * slot may point to max if the key is bigger than all of the keys
  */
-static int generic_bin_search(char *p, int item_size, struct btrfs_key *key,
-		       int max, int *slot)
+static int generic_bin_search(struct extent_buffer *eb, unsigned long p,
+			      int item_size, struct btrfs_key *key,
+			      int max, int *slot)
 {
 	int low = 0;
 	int high = max;
 	int mid;
 	int ret;
 	struct btrfs_disk_key *tmp;
+	struct btrfs_disk_key unaligned;
+	unsigned long offset;
+	char *map_token = NULL;
+	char *kaddr = NULL;
+	unsigned long map_start = 0;
+	unsigned long map_len = 0;
 
 	while(low < high) {
 		mid = (low + high) / 2;
-		tmp = (struct btrfs_disk_key *)(p + mid * item_size);
+		offset = p + mid * item_size;
+
+		if (!map_token || offset < map_start ||
+		    (offset + sizeof(struct btrfs_disk_key)) >
+		    map_start + map_len) {
+			if (map_token)
+				unmap_extent_buffer(eb, map_token, KM_USER0);
+			map_extent_buffer(eb, offset, &map_token, &kaddr,
+					  &map_start, &map_len, KM_USER0);
+
+		}
+		if (offset + sizeof(struct btrfs_disk_key) >
+		    map_start + map_len) {
+			unmap_extent_buffer(eb, map_token, KM_USER0);
+			read_extent_buffer(eb, &unaligned,
+					   offset, sizeof(unaligned));
+			map_token = NULL;
+			tmp = &unaligned;
+		} else {
+			tmp = (struct btrfs_disk_key *)(kaddr + offset -
+							map_start);
+		}
 		ret = comp_keys(tmp, key);
 
 		if (ret < 0)
@@ -447,10 +544,13 @@ static int generic_bin_search(char *p, int item_size, struct btrfs_key *key,
 			high = mid;
 		else {
 			*slot = mid;
+			unmap_extent_buffer(eb, map_token, KM_USER0);
 			return 0;
 		}
 	}
 	*slot = low;
+	if (map_token)
+		unmap_extent_buffer(eb, map_token, KM_USER0);
 	return 1;
 }
 
@@ -458,46 +558,42 @@ static int generic_bin_search(char *p, int item_size, struct btrfs_key *key,
  * simple bin_search frontend that does the right thing for
  * leaves vs nodes
  */
-static int bin_search(struct btrfs_node *c, struct btrfs_key *key, int *slot)
+static int bin_search(struct extent_buffer *eb, struct btrfs_key *key,
+		      int level, int *slot)
 {
-	if (btrfs_is_leaf(c)) {
-		struct btrfs_leaf *l = (struct btrfs_leaf *)c;
-		return generic_bin_search((void *)l->items,
+	if (level == 0) {
+		return generic_bin_search(eb,
+					  offsetof(struct btrfs_leaf, items),
 					  sizeof(struct btrfs_item),
-					  key, btrfs_header_nritems(&c->header),
+					  key, btrfs_header_nritems(eb),
 					  slot);
 	} else {
-		return generic_bin_search((void *)c->ptrs,
+		return generic_bin_search(eb,
+					  offsetof(struct btrfs_node, ptrs),
 					  sizeof(struct btrfs_key_ptr),
-					  key, btrfs_header_nritems(&c->header),
+					  key, btrfs_header_nritems(eb),
 					  slot);
 	}
 	return -1;
 }
 
-static struct buffer_head *read_node_slot(struct btrfs_root *root,
-				   struct buffer_head *parent_buf,
-				   int slot)
+static struct extent_buffer *read_node_slot(struct btrfs_root *root,
+				   struct extent_buffer *parent, int slot)
 {
-	struct btrfs_node *node = btrfs_buffer_node(parent_buf);
 	if (slot < 0)
 		return NULL;
-	if (slot >= btrfs_header_nritems(&node->header))
+	if (slot >= btrfs_header_nritems(parent))
 		return NULL;
-	return read_tree_block(root, btrfs_node_blockptr(node, slot));
+	return read_tree_block(root, btrfs_node_blockptr(parent, slot));
 }
 
 static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 			 *root, struct btrfs_path *path, int level)
 {
-	struct buffer_head *right_buf;
-	struct buffer_head *mid_buf;
-	struct buffer_head *left_buf;
-	struct buffer_head *parent_buf = NULL;
-	struct btrfs_node *right = NULL;
-	struct btrfs_node *mid;
-	struct btrfs_node *left = NULL;
-	struct btrfs_node *parent = NULL;
+	struct extent_buffer *right = NULL;
+	struct extent_buffer *mid;
+	struct extent_buffer *left = NULL;
+	struct extent_buffer *parent = NULL;
 	int ret = 0;
 	int wret;
 	int pslot;
@@ -508,60 +604,57 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 	if (level == 0)
 		return 0;
 
-	mid_buf = path->nodes[level];
-	mid = btrfs_buffer_node(mid_buf);
+	mid = path->nodes[level];
 	orig_ptr = btrfs_node_blockptr(mid, orig_slot);
 
 	if (level < BTRFS_MAX_LEVEL - 1)
-		parent_buf = path->nodes[level + 1];
+		parent = path->nodes[level + 1];
 	pslot = path->slots[level + 1];
 
 	/*
 	 * deal with the case where there is only one pointer in the root
 	 * by promoting the node below to a root
 	 */
-	if (!parent_buf) {
-		struct buffer_head *child;
-		u64 blocknr = bh_blocknr(mid_buf);
+	if (!parent) {
+		struct extent_buffer *child;
+		u64 blocknr = extent_buffer_blocknr(mid);
 
-		if (btrfs_header_nritems(&mid->header) != 1)
+		if (btrfs_header_nritems(mid) != 1)
 			return 0;
 
 		/* promote the child to a root */
-		child = read_node_slot(root, mid_buf, 0);
+		child = read_node_slot(root, mid, 0);
 		BUG_ON(!child);
 		root->node = child;
 		path->nodes[level] = NULL;
-		clean_tree_block(trans, root, mid_buf);
-		wait_on_buffer(mid_buf);
+		clean_tree_block(trans, root, mid);
+		wait_on_tree_block_writeback(root, mid);
 		/* once for the path */
-		btrfs_block_release(root, mid_buf);
+		free_extent_buffer(mid);
 		/* once for the root ptr */
-		btrfs_block_release(root, mid_buf);
+		free_extent_buffer(mid);
 		return btrfs_free_extent(trans, root, blocknr, 1, 1);
 	}
-	parent = btrfs_buffer_node(parent_buf);
-
-	if (btrfs_header_nritems(&mid->header) >
+	if (btrfs_header_nritems(mid) >
 	    BTRFS_NODEPTRS_PER_BLOCK(root) / 4)
 		return 0;
 
-	if (btrfs_header_nritems(&mid->header) < 2)
+	if (btrfs_header_nritems(mid) < 2)
 		err_on_enospc = 1;
 
-	left_buf = read_node_slot(root, parent_buf, pslot - 1);
-	if (left_buf) {
-		wret = btrfs_cow_block(trans, root, left_buf,
-				       parent_buf, pslot - 1, &left_buf);
+	left = read_node_slot(root, parent, pslot - 1);
+	if (left) {
+		wret = btrfs_cow_block(trans, root, left,
+				       parent, pslot - 1, &left);
 		if (wret) {
 			ret = wret;
 			goto enospc;
 		}
 	}
-	right_buf = read_node_slot(root, parent_buf, pslot + 1);
-	if (right_buf) {
-		wret = btrfs_cow_block(trans, root, right_buf,
-				       parent_buf, pslot + 1, &right_buf);
+	right = read_node_slot(root, parent, pslot + 1);
+	if (right) {
+		wret = btrfs_cow_block(trans, root, right,
+				       parent, pslot + 1, &right);
 		if (wret) {
 			ret = wret;
 			goto enospc;
@@ -569,30 +662,27 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 	}
 
 	/* first, try to make some room in the middle buffer */
-	if (left_buf) {
-		left = btrfs_buffer_node(left_buf);
-		orig_slot += btrfs_header_nritems(&left->header);
-		wret = push_node_left(trans, root, left_buf, mid_buf);
+	if (left) {
+		orig_slot += btrfs_header_nritems(left);
+		wret = push_node_left(trans, root, left, mid);
 		if (wret < 0)
 			ret = wret;
-		if (btrfs_header_nritems(&mid->header) < 2)
+		if (btrfs_header_nritems(mid) < 2)
 			err_on_enospc = 1;
 	}
 
 	/*
 	 * then try to empty the right most buffer into the middle
 	 */
-	if (right_buf) {
-		right = btrfs_buffer_node(right_buf);
-		wret = push_node_left(trans, root, mid_buf, right_buf);
+	if (right) {
+		wret = push_node_left(trans, root, mid, right);
 		if (wret < 0 && wret != -ENOSPC)
 			ret = wret;
-		if (btrfs_header_nritems(&right->header) == 0) {
-			u64 blocknr = bh_blocknr(right_buf);
-			clean_tree_block(trans, root, right_buf);
-			wait_on_buffer(right_buf);
-			btrfs_block_release(root, right_buf);
-			right_buf = NULL;
+		if (btrfs_header_nritems(right) == 0) {
+			u64 blocknr = extent_buffer_blocknr(right);
+			clean_tree_block(trans, root, right);
+			wait_on_tree_block_writeback(root, right);
+			free_extent_buffer(right);
 			right = NULL;
 			wret = del_ptr(trans, root, path, level + 1, pslot +
 				       1);
@@ -602,14 +692,13 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 			if (wret)
 				ret = wret;
 		} else {
-			btrfs_memcpy(root, parent,
-				     &parent->ptrs[pslot + 1].key,
-				     &right->ptrs[0].key,
-				     sizeof(struct btrfs_disk_key));
-			btrfs_mark_buffer_dirty(parent_buf);
+			struct btrfs_disk_key right_key;
+			btrfs_node_key(right, &right_key, 0);
+			btrfs_set_node_key(parent, &right_key, pslot + 1);
+			btrfs_mark_buffer_dirty(parent);
 		}
 	}
-	if (btrfs_header_nritems(&mid->header) == 1) {
+	if (btrfs_header_nritems(mid) == 1) {
 		/*
 		 * we're not allowed to leave a node with one item in the
 		 * tree during a delete.  A deletion from lower in the tree
@@ -619,21 +708,20 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 		 * otherwise we would have pulled some pointers from the
 		 * right
 		 */
-		BUG_ON(!left_buf);
-		wret = balance_node_right(trans, root, mid_buf, left_buf);
+		BUG_ON(!left);
+		wret = balance_node_right(trans, root, mid, left);
 		if (wret < 0) {
 			ret = wret;
 			goto enospc;
 		}
 		BUG_ON(wret == 1);
 	}
-	if (btrfs_header_nritems(&mid->header) == 0) {
+	if (btrfs_header_nritems(mid) == 0) {
 		/* we've managed to empty the middle node, drop it */
-		u64 blocknr = bh_blocknr(mid_buf);
-		clean_tree_block(trans, root, mid_buf);
-		wait_on_buffer(mid_buf);
-		btrfs_block_release(root, mid_buf);
-		mid_buf = NULL;
+		u64 blocknr = extent_buffer_blocknr(mid);
+		clean_tree_block(trans, root, mid);
+		wait_on_tree_block_writeback(root, mid);
+		free_extent_buffer(mid);
 		mid = NULL;
 		wret = del_ptr(trans, root, path, level + 1, pslot);
 		if (wret)
@@ -643,37 +731,36 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 			ret = wret;
 	} else {
 		/* update the parent key to reflect our changes */
-		btrfs_memcpy(root, parent,
-			     &parent->ptrs[pslot].key, &mid->ptrs[0].key,
-			     sizeof(struct btrfs_disk_key));
-		btrfs_mark_buffer_dirty(parent_buf);
+		struct btrfs_disk_key mid_key;
+		btrfs_node_key(mid, &mid_key, 0);
+		btrfs_set_node_key(parent, &mid_key, pslot);
+		btrfs_mark_buffer_dirty(parent);
 	}
 
 	/* update the path */
-	if (left_buf) {
-		if (btrfs_header_nritems(&left->header) > orig_slot) {
-			get_bh(left_buf);
-			path->nodes[level] = left_buf;
+	if (left) {
+		if (btrfs_header_nritems(left) > orig_slot) {
+			extent_buffer_get(left);
+			path->nodes[level] = left;
 			path->slots[level + 1] -= 1;
 			path->slots[level] = orig_slot;
-			if (mid_buf)
-				btrfs_block_release(root, mid_buf);
+			if (mid)
+				free_extent_buffer(mid);
 		} else {
-			orig_slot -= btrfs_header_nritems(&left->header);
+			orig_slot -= btrfs_header_nritems(left);
 			path->slots[level] = orig_slot;
 		}
 	}
 	/* double check we haven't messed things up */
 	check_block(root, path, level);
 	if (orig_ptr !=
-	    btrfs_node_blockptr(btrfs_buffer_node(path->nodes[level]),
-				path->slots[level]))
+	    btrfs_node_blockptr(path->nodes[level], path->slots[level]))
 		BUG();
 enospc:
-	if (right_buf)
-		btrfs_block_release(root, right_buf);
-	if (left_buf)
-		btrfs_block_release(root, left_buf);
+	if (right)
+		free_extent_buffer(right);
+	if (left)
+		free_extent_buffer(left);
 	return ret;
 }
 
@@ -682,14 +769,10 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root,
 				struct btrfs_path *path, int level)
 {
-	struct buffer_head *right_buf;
-	struct buffer_head *mid_buf;
-	struct buffer_head *left_buf;
-	struct buffer_head *parent_buf = NULL;
-	struct btrfs_node *right = NULL;
-	struct btrfs_node *mid;
-	struct btrfs_node *left = NULL;
-	struct btrfs_node *parent = NULL;
+	struct extent_buffer *right = NULL;
+	struct extent_buffer *mid;
+	struct extent_buffer *left = NULL;
+	struct extent_buffer *parent = NULL;
 	int ret = 0;
 	int wret;
 	int pslot;
@@ -699,107 +782,101 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 	if (level == 0)
 		return 1;
 
-	mid_buf = path->nodes[level];
-	mid = btrfs_buffer_node(mid_buf);
+	mid = path->nodes[level];
 	orig_ptr = btrfs_node_blockptr(mid, orig_slot);
 
 	if (level < BTRFS_MAX_LEVEL - 1)
-		parent_buf = path->nodes[level + 1];
+		parent = path->nodes[level + 1];
 	pslot = path->slots[level + 1];
 
-	if (!parent_buf)
+	if (!parent)
 		return 1;
-	parent = btrfs_buffer_node(parent_buf);
 
-	left_buf = read_node_slot(root, parent_buf, pslot - 1);
+	left = read_node_slot(root, parent, pslot - 1);
 
 	/* first, try to make some room in the middle buffer */
-	if (left_buf) {
+	if (left) {
 		u32 left_nr;
-		left = btrfs_buffer_node(left_buf);
-		left_nr = btrfs_header_nritems(&left->header);
+		left_nr = btrfs_header_nritems(left);
 		if (left_nr >= BTRFS_NODEPTRS_PER_BLOCK(root) - 1) {
 			wret = 1;
 		} else {
-			ret = btrfs_cow_block(trans, root, left_buf, parent_buf,
-					      pslot - 1, &left_buf);
+			ret = btrfs_cow_block(trans, root, left, parent,
+					      pslot - 1, &left);
 			if (ret)
 				wret = 1;
 			else {
-				left = btrfs_buffer_node(left_buf);
 				wret = push_node_left(trans, root,
-						      left_buf, mid_buf);
+						      left, mid);
 			}
 		}
 		if (wret < 0)
 			ret = wret;
 		if (wret == 0) {
+			struct btrfs_disk_key disk_key;
 			orig_slot += left_nr;
-			btrfs_memcpy(root, parent,
-				     &parent->ptrs[pslot].key,
-				     &mid->ptrs[0].key,
-				     sizeof(struct btrfs_disk_key));
-			btrfs_mark_buffer_dirty(parent_buf);
-			if (btrfs_header_nritems(&left->header) > orig_slot) {
-				path->nodes[level] = left_buf;
+			btrfs_node_key(mid, &disk_key, 0);
+			btrfs_set_node_key(parent, &disk_key, pslot);
+			btrfs_mark_buffer_dirty(parent);
+			if (btrfs_header_nritems(left) > orig_slot) {
+				path->nodes[level] = left;
 				path->slots[level + 1] -= 1;
 				path->slots[level] = orig_slot;
-				btrfs_block_release(root, mid_buf);
+				free_extent_buffer(mid);
 			} else {
 				orig_slot -=
-					btrfs_header_nritems(&left->header);
+					btrfs_header_nritems(left);
 				path->slots[level] = orig_slot;
-				btrfs_block_release(root, left_buf);
+				free_extent_buffer(left);
 			}
 			check_node(root, path, level);
 			return 0;
 		}
-		btrfs_block_release(root, left_buf);
+		free_extent_buffer(left);
 	}
-	right_buf = read_node_slot(root, parent_buf, pslot + 1);
+	right= read_node_slot(root, parent, pslot + 1);
 
 	/*
 	 * then try to empty the right most buffer into the middle
 	 */
-	if (right_buf) {
+	if (right) {
 		u32 right_nr;
-		right = btrfs_buffer_node(right_buf);
-		right_nr = btrfs_header_nritems(&right->header);
+		right_nr = btrfs_header_nritems(right);
 		if (right_nr >= BTRFS_NODEPTRS_PER_BLOCK(root) - 1) {
 			wret = 1;
 		} else {
-			ret = btrfs_cow_block(trans, root, right_buf,
-					      parent_buf, pslot + 1,
-					      &right_buf);
+			ret = btrfs_cow_block(trans, root, right,
+					      parent, pslot + 1,
+					      &right);
 			if (ret)
 				wret = 1;
 			else {
-				right = btrfs_buffer_node(right_buf);
 				wret = balance_node_right(trans, root,
-							  right_buf, mid_buf);
+							  right, mid);
 			}
 		}
 		if (wret < 0)
 			ret = wret;
 		if (wret == 0) {
-			btrfs_memcpy(root, parent,
-				     &parent->ptrs[pslot + 1].key,
-				     &right->ptrs[0].key,
-				     sizeof(struct btrfs_disk_key));
-			btrfs_mark_buffer_dirty(parent_buf);
-			if (btrfs_header_nritems(&mid->header) <= orig_slot) {
-				path->nodes[level] = right_buf;
+			struct btrfs_disk_key disk_key;
+
+			btrfs_node_key(right, &disk_key, 0);
+			btrfs_set_node_key(parent, &disk_key, pslot + 1);
+			btrfs_mark_buffer_dirty(parent);
+
+			if (btrfs_header_nritems(mid) <= orig_slot) {
+				path->nodes[level] = right;
 				path->slots[level + 1] += 1;
 				path->slots[level] = orig_slot -
-					btrfs_header_nritems(&mid->header);
-				btrfs_block_release(root, mid_buf);
+					btrfs_header_nritems(mid);
+				free_extent_buffer(mid);
 			} else {
-				btrfs_block_release(root, right_buf);
+				free_extent_buffer(right);
 			}
 			check_node(root, path, level);
 			return 0;
 		}
-		btrfs_block_release(root, right_buf);
+		free_extent_buffer(right);
 	}
 	check_node(root, path, level);
 	return 1;
@@ -811,10 +888,9 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
 			     int level, int slot)
 {
-	struct btrfs_node *node;
+	struct extent_buffer *node;
 	int i;
 	u32 nritems;
-	u64 item_objectid;
 	u64 blocknr;
 	u64 search;
 	u64 cluster_start;
@@ -823,7 +899,7 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
 	int direction = path->reada;
 	struct radix_tree_root found;
 	unsigned long gang[8];
-	struct buffer_head *bh;
+	struct extent_buffer *eb;
 
 	if (level == 0)
 		return;
@@ -831,18 +907,17 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
 	if (!path->nodes[level])
 		return;
 
-	node = btrfs_buffer_node(path->nodes[level]);
+	node = path->nodes[level];
 	search = btrfs_node_blockptr(node, slot);
-	bh = btrfs_find_tree_block(root, search);
-	if (bh) {
-		brelse(bh);
+	eb = btrfs_find_tree_block(root, search);
+	if (eb) {
+		free_extent_buffer(eb);
 		return;
 	}
 
 	init_bit_radix(&found);
-	nritems = btrfs_header_nritems(&node->header);
+	nritems = btrfs_header_nritems(node);
 	for (i = slot; i < nritems; i++) {
-		item_objectid = btrfs_disk_key_objectid(&node->ptrs[i].key);
 		blocknr = btrfs_node_blockptr(node, i);
 		set_radix_bit(&found, blocknr);
 	}
@@ -886,8 +961,7 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_key *key, struct btrfs_path *p, int
 		      ins_len, int cow)
 {
-	struct buffer_head *b;
-	struct btrfs_node *c;
+	struct extent_buffer *b;
 	u64 blocknr;
 	int slot;
 	int ret;
@@ -901,10 +975,9 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
 	WARN_ON(!mutex_is_locked(&root->fs_info->fs_mutex));
 again:
 	b = root->node;
-	get_bh(b);
+	extent_buffer_get(b);
 	while (b) {
-		c = btrfs_buffer_node(b);
-		level = btrfs_header_level(&c->header);
+		level = btrfs_header_level(b);
 		if (cow) {
 			int wret;
 			wret = btrfs_cow_block(trans, root, b,
@@ -912,32 +985,30 @@ again:
 					       p->slots[level + 1],
 					       &b);
 			if (wret) {
-				btrfs_block_release(root, b);
+				free_extent_buffer(b);
 				return wret;
 			}
-			c = btrfs_buffer_node(b);
 		}
 		BUG_ON(!cow && ins_len);
-		if (level != btrfs_header_level(&c->header))
+		if (level != btrfs_header_level(b))
 			WARN_ON(1);
-		level = btrfs_header_level(&c->header);
+		level = btrfs_header_level(b);
 		p->nodes[level] = b;
 		ret = check_block(root, p, level);
 		if (ret)
 			return -1;
-		ret = bin_search(c, key, &slot);
-		if (!btrfs_is_leaf(c)) {
+		ret = bin_search(b, key, level, &slot);
+		if (level != 0) {
 			if (ret && slot > 0)
 				slot -= 1;
 			p->slots[level] = slot;
-			if (ins_len > 0 && btrfs_header_nritems(&c->header) >=
+			if (ins_len > 0 && btrfs_header_nritems(b) >=
 			    BTRFS_NODEPTRS_PER_BLOCK(root) - 1) {
 				int sret = split_node(trans, root, p, level);
 				BUG_ON(sret > 0);
 				if (sret)
 					return sret;
 				b = p->nodes[level];
-				c = btrfs_buffer_node(b);
 				slot = p->slots[level];
 			} else if (ins_len < 0) {
 				int sret = balance_level(trans, root, p,
@@ -947,22 +1018,19 @@ again:
 				b = p->nodes[level];
 				if (!b)
 					goto again;
-				c = btrfs_buffer_node(b);
 				slot = p->slots[level];
-				BUG_ON(btrfs_header_nritems(&c->header) == 1);
+				BUG_ON(btrfs_header_nritems(b) == 1);
 			}
 			/* this is only true while dropping a snapshot */
 			if (level == lowest_level)
 				break;
-			blocknr = btrfs_node_blockptr(c, slot);
+			blocknr = btrfs_node_blockptr(b, slot);
 			if (should_reada)
 				reada_for_search(root, p, level, slot);
-			b = read_tree_block(root, btrfs_node_blockptr(c, slot));
-
+			b = read_tree_block(root, btrfs_node_blockptr(b, slot));
 		} else {
-			struct btrfs_leaf *l = (struct btrfs_leaf *)c;
 			p->slots[level] = slot;
-			if (ins_len > 0 && btrfs_leaf_free_space(root, l) <
+			if (ins_len > 0 && btrfs_leaf_free_space(root, b) <
 			    sizeof(struct btrfs_item) + ins_len) {
 				int sret = split_leaf(trans, root, key,
 						      p, ins_len);
@@ -986,19 +1054,20 @@ again:
  * If this fails to write a tree block, it returns -1, but continues
  * fixing up the blocks in ram so the tree is consistent.
  */
-static int fixup_low_keys(struct btrfs_trans_handle *trans, struct btrfs_root
-			  *root, struct btrfs_path *path, struct btrfs_disk_key
-			  *key, int level)
+static int fixup_low_keys(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root, struct btrfs_path *path,
+			  struct btrfs_disk_key *key, int level)
 {
 	int i;
 	int ret = 0;
+	struct extent_buffer *t;
+
 	for (i = level; i < BTRFS_MAX_LEVEL; i++) {
-		struct btrfs_node *t;
 		int tslot = path->slots[i];
 		if (!path->nodes[i])
 			break;
-		t = btrfs_buffer_node(path->nodes[i]);
-		btrfs_memcpy(root, t, &t->ptrs[tslot].key, key, sizeof(*key));
+		t = path->nodes[i];
+		btrfs_set_node_key(t, key, tslot);
 		btrfs_mark_buffer_dirty(path->nodes[i]);
 		if (tslot != 0)
 			break;
@@ -1014,18 +1083,16 @@ static int fixup_low_keys(struct btrfs_trans_handle *trans, struct btrfs_root
  * error, and > 0 if there was no room in the left hand block.
  */
 static int push_node_left(struct btrfs_trans_handle *trans, struct btrfs_root
-			  *root, struct buffer_head *dst_buf, struct
-			  buffer_head *src_buf)
+			  *root, struct extent_buffer *dst,
+			  struct extent_buffer *src)
 {
-	struct btrfs_node *src = btrfs_buffer_node(src_buf);
-	struct btrfs_node *dst = btrfs_buffer_node(dst_buf);
 	int push_items = 0;
 	int src_nritems;
 	int dst_nritems;
 	int ret = 0;
 
-	src_nritems = btrfs_header_nritems(&src->header);
-	dst_nritems = btrfs_header_nritems(&dst->header);
+	src_nritems = btrfs_header_nritems(src);
+	dst_nritems = btrfs_header_nritems(dst);
 	push_items = BTRFS_NODEPTRS_PER_BLOCK(root) - dst_nritems;
 
 	if (push_items <= 0) {
@@ -1035,17 +1102,21 @@ static int push_node_left(struct btrfs_trans_handle *trans, struct btrfs_root
 	if (src_nritems < push_items)
 		push_items = src_nritems;
 
-	btrfs_memcpy(root, dst, dst->ptrs + dst_nritems, src->ptrs,
-		     push_items * sizeof(struct btrfs_key_ptr));
+	copy_extent_buffer(dst, src,
+			   btrfs_node_key_ptr_offset(dst_nritems),
+			   btrfs_node_key_ptr_offset(0),
+		           push_items * sizeof(struct btrfs_key_ptr));
+
 	if (push_items < src_nritems) {
-		btrfs_memmove(root, src, src->ptrs, src->ptrs + push_items,
-			(src_nritems - push_items) *
-			sizeof(struct btrfs_key_ptr));
-	}
-	btrfs_set_header_nritems(&src->header, src_nritems - push_items);
-	btrfs_set_header_nritems(&dst->header, dst_nritems + push_items);
-	btrfs_mark_buffer_dirty(src_buf);
-	btrfs_mark_buffer_dirty(dst_buf);
+		memmove_extent_buffer(src, btrfs_node_key_ptr_offset(0),
+				      btrfs_node_key_ptr_offset(push_items),
+				      (src_nritems - push_items) *
+				      sizeof(struct btrfs_key_ptr));
+	}
+	btrfs_set_header_nritems(src, src_nritems - push_items);
+	btrfs_set_header_nritems(dst, dst_nritems + push_items);
+	btrfs_mark_buffer_dirty(src);
+	btrfs_mark_buffer_dirty(dst);
 	return ret;
 }
 
@@ -1058,24 +1129,22 @@ static int push_node_left(struct btrfs_trans_handle *trans, struct btrfs_root
  *
  * this will  only push up to 1/2 the contents of the left node over
  */
-static int balance_node_right(struct btrfs_trans_handle *trans, struct
-			      btrfs_root *root, struct buffer_head *dst_buf,
-			      struct buffer_head *src_buf)
+static int balance_node_right(struct btrfs_trans_handle *trans,
+			      struct btrfs_root *root,
+			      struct extent_buffer *dst,
+			      struct extent_buffer *src)
 {
-	struct btrfs_node *src = btrfs_buffer_node(src_buf);
-	struct btrfs_node *dst = btrfs_buffer_node(dst_buf);
 	int push_items = 0;
 	int max_push;
 	int src_nritems;
 	int dst_nritems;
 	int ret = 0;
 
-	src_nritems = btrfs_header_nritems(&src->header);
-	dst_nritems = btrfs_header_nritems(&dst->header);
+	src_nritems = btrfs_header_nritems(src);
+	dst_nritems = btrfs_header_nritems(dst);
 	push_items = BTRFS_NODEPTRS_PER_BLOCK(root) - dst_nritems;
-	if (push_items <= 0) {
+	if (push_items <= 0)
 		return 1;
-	}
 
 	max_push = src_nritems / 2 + 1;
 	/* don't try to empty the node */
@@ -1085,18 +1154,21 @@ static int balance_node_right(struct btrfs_trans_handle *trans, struct
 	if (max_push < push_items)
 		push_items = max_push;
 
-	btrfs_memmove(root, dst, dst->ptrs + push_items, dst->ptrs,
-		      dst_nritems * sizeof(struct btrfs_key_ptr));
+	memmove_extent_buffer(dst, btrfs_node_key_ptr_offset(push_items),
+				      btrfs_node_key_ptr_offset(0),
+				      (dst_nritems) *
+				      sizeof(struct btrfs_key_ptr));
 
-	btrfs_memcpy(root, dst, dst->ptrs,
-		     src->ptrs + src_nritems - push_items,
-		     push_items * sizeof(struct btrfs_key_ptr));
+	copy_extent_buffer(dst, src,
+			   btrfs_node_key_ptr_offset(0),
+			   btrfs_node_key_ptr_offset(src_nritems - push_items),
+		           push_items * sizeof(struct btrfs_key_ptr));
 
-	btrfs_set_header_nritems(&src->header, src_nritems - push_items);
-	btrfs_set_header_nritems(&dst->header, dst_nritems + push_items);
+	btrfs_set_header_nritems(src, src_nritems - push_items);
+	btrfs_set_header_nritems(dst, dst_nritems + push_items);
 
-	btrfs_mark_buffer_dirty(src_buf);
-	btrfs_mark_buffer_dirty(dst_buf);
+	btrfs_mark_buffer_dirty(src);
+	btrfs_mark_buffer_dirty(dst);
 	return ret;
 }
 
@@ -1107,45 +1179,46 @@ static int balance_node_right(struct btrfs_trans_handle *trans, struct
  *
  * returns zero on success or < 0 on failure.
  */
-static int insert_new_root(struct btrfs_trans_handle *trans, struct btrfs_root
-			   *root, struct btrfs_path *path, int level)
+static int insert_new_root(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *root,
+			   struct btrfs_path *path, int level)
 {
-	struct buffer_head *t;
-	struct btrfs_node *lower;
-	struct btrfs_node *c;
-	struct btrfs_disk_key *lower_key;
+	struct extent_buffer *lower;
+	struct extent_buffer *c;
+	struct btrfs_disk_key lower_key;
 
 	BUG_ON(path->nodes[level]);
 	BUG_ON(path->nodes[level-1] != root->node);
 
-	t = btrfs_alloc_free_block(trans, root, root->node->b_blocknr, 0);
-	if (IS_ERR(t))
-		return PTR_ERR(t);
-	c = btrfs_buffer_node(t);
-	memset(c, 0, root->blocksize);
-	btrfs_set_header_nritems(&c->header, 1);
-	btrfs_set_header_level(&c->header, level);
-	btrfs_set_header_blocknr(&c->header, bh_blocknr(t));
-	btrfs_set_header_generation(&c->header, trans->transid);
-	btrfs_set_header_owner(&c->header, root->root_key.objectid);
-	lower = btrfs_buffer_node(path->nodes[level-1]);
-	memcpy(c->header.fsid, root->fs_info->disk_super->fsid,
-	       sizeof(c->header.fsid));
-	if (btrfs_is_leaf(lower))
-		lower_key = &((struct btrfs_leaf *)lower)->items[0].key;
+	c = btrfs_alloc_free_block(trans, root,
+				   extent_buffer_blocknr(root->node), 0);
+	if (IS_ERR(c))
+		return PTR_ERR(c);
+	memset_extent_buffer(c, 0, 0, root->nodesize);
+	btrfs_set_header_nritems(c, 1);
+	btrfs_set_header_level(c, level);
+	btrfs_set_header_blocknr(c, extent_buffer_blocknr(c));
+	btrfs_set_header_generation(c, trans->transid);
+	btrfs_set_header_owner(c, root->root_key.objectid);
+	lower = path->nodes[level-1];
+
+	write_extent_buffer(c, root->fs_info->fsid,
+			    (unsigned long)btrfs_header_fsid(c),
+			    BTRFS_FSID_SIZE);
+	if (level == 1)
+		btrfs_item_key(lower, &lower_key, 0);
 	else
-		lower_key = &lower->ptrs[0].key;
-	btrfs_memcpy(root, c, &c->ptrs[0].key, lower_key,
-		     sizeof(struct btrfs_disk_key));
-	btrfs_set_node_blockptr(c, 0, bh_blocknr(path->nodes[level - 1]));
+		btrfs_node_key(lower, &lower_key, 0);
+	btrfs_set_node_key(c, &lower_key, 0);
+	btrfs_set_node_blockptr(c, 0, extent_buffer_blocknr(lower));
 
-	btrfs_mark_buffer_dirty(t);
+	btrfs_mark_buffer_dirty(c);
 
 	/* the super has an extra ref to root->node */
-	btrfs_block_release(root, root->node);
-	root->node = t;
-	get_bh(t);
-	path->nodes[level] = t;
+	free_extent_buffer(root->node);
+	root->node = c;
+	extent_buffer_get(c);
+	path->nodes[level] = c;
 	path->slots[level] = 0;
 	return 0;
 }
@@ -1163,26 +1236,26 @@ static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_path *path, struct btrfs_disk_key
 		      *key, u64 blocknr, int slot, int level)
 {
-	struct btrfs_node *lower;
+	struct extent_buffer *lower;
 	int nritems;
 
 	BUG_ON(!path->nodes[level]);
-	lower = btrfs_buffer_node(path->nodes[level]);
-	nritems = btrfs_header_nritems(&lower->header);
+	lower = path->nodes[level];
+	nritems = btrfs_header_nritems(lower);
 	if (slot > nritems)
 		BUG();
 	if (nritems == BTRFS_NODEPTRS_PER_BLOCK(root))
 		BUG();
 	if (slot != nritems) {
-		btrfs_memmove(root, lower, lower->ptrs + slot + 1,
-			      lower->ptrs + slot,
+		memmove_extent_buffer(lower,
+			      btrfs_node_key_ptr_offset(slot + 1),
+			      btrfs_node_key_ptr_offset(slot),
 			      (nritems - slot) * sizeof(struct btrfs_key_ptr));
 	}
-	btrfs_memcpy(root, lower, &lower->ptrs[slot].key,
-		     key, sizeof(struct btrfs_disk_key));
+	btrfs_set_node_key(lower, key, slot);
 	btrfs_set_node_blockptr(lower, slot, blocknr);
-	btrfs_set_header_nritems(&lower->header, nritems + 1);
-	btrfs_mark_buffer_dirty(path->nodes[level]);
+	btrfs_set_header_nritems(lower, nritems + 1);
+	btrfs_mark_buffer_dirty(lower);
 	check_node(root, path, level);
 	return 0;
 }
@@ -1199,69 +1272,73 @@ static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root
 static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_path *path, int level)
 {
-	struct buffer_head *t;
-	struct btrfs_node *c;
-	struct buffer_head *split_buffer;
-	struct btrfs_node *split;
+	struct extent_buffer *c;
+	struct extent_buffer *split;
+	struct btrfs_disk_key disk_key;
 	int mid;
 	int ret;
 	int wret;
 	u32 c_nritems;
 
-	t = path->nodes[level];
-	c = btrfs_buffer_node(t);
-	if (t == root->node) {
+	c = path->nodes[level];
+	if (c == root->node) {
 		/* trying to split the root, lets make a new one */
 		ret = insert_new_root(trans, root, path, level + 1);
 		if (ret)
 			return ret;
 	} else {
 		ret = push_nodes_for_insert(trans, root, path, level);
-		t = path->nodes[level];
-		c = btrfs_buffer_node(t);
-		if (!ret &&
-		    btrfs_header_nritems(&c->header) <
+		c = path->nodes[level];
+		if (!ret && btrfs_header_nritems(c) <
 		    BTRFS_NODEPTRS_PER_BLOCK(root) - 1)
 			return 0;
 		if (ret < 0)
 			return ret;
 	}
 
-	c_nritems = btrfs_header_nritems(&c->header);
-	split_buffer = btrfs_alloc_free_block(trans, root, t->b_blocknr, 0);
-	if (IS_ERR(split_buffer))
-		return PTR_ERR(split_buffer);
+	c_nritems = btrfs_header_nritems(c);
+	split = btrfs_alloc_free_block(trans, root,
+				       extent_buffer_blocknr(c), 0);
+	if (IS_ERR(split))
+		return PTR_ERR(split);
+
+	btrfs_set_header_flags(split, btrfs_header_flags(c));
+	btrfs_set_header_level(split, btrfs_header_level(c));
+	btrfs_set_header_blocknr(split, extent_buffer_blocknr(split));
+	btrfs_set_header_generation(split, trans->transid);
+	btrfs_set_header_owner(split, root->root_key.objectid);
+	write_extent_buffer(split, root->fs_info->fsid,
+			    (unsigned long)btrfs_header_fsid(split),
+			    BTRFS_FSID_SIZE);
 
-	split = btrfs_buffer_node(split_buffer);
-	btrfs_set_header_flags(&split->header, btrfs_header_flags(&c->header));
-	btrfs_set_header_level(&split->header, btrfs_header_level(&c->header));
-	btrfs_set_header_blocknr(&split->header, bh_blocknr(split_buffer));
-	btrfs_set_header_generation(&split->header, trans->transid);
-	btrfs_set_header_owner(&split->header, root->root_key.objectid);
-	memcpy(split->header.fsid, root->fs_info->disk_super->fsid,
-	       sizeof(split->header.fsid));
 	mid = (c_nritems + 1) / 2;
-	btrfs_memcpy(root, split, split->ptrs, c->ptrs + mid,
-		     (c_nritems - mid) * sizeof(struct btrfs_key_ptr));
-	btrfs_set_header_nritems(&split->header, c_nritems - mid);
-	btrfs_set_header_nritems(&c->header, mid);
+
+	copy_extent_buffer(split, c,
+			   btrfs_node_key_ptr_offset(0),
+			   btrfs_node_key_ptr_offset(mid),
+			   (c_nritems - mid) * sizeof(struct btrfs_key_ptr));
+	btrfs_set_header_nritems(split, c_nritems - mid);
+	btrfs_set_header_nritems(c, mid);
 	ret = 0;
 
-	btrfs_mark_buffer_dirty(t);
-	btrfs_mark_buffer_dirty(split_buffer);
-	wret = insert_ptr(trans, root, path, &split->ptrs[0].key,
-			  bh_blocknr(split_buffer), path->slots[level + 1] + 1,
+	btrfs_mark_buffer_dirty(c);
+	btrfs_mark_buffer_dirty(split);
+
+	btrfs_node_key(split, &disk_key, 0);
+	wret = insert_ptr(trans, root, path, &disk_key,
+			  extent_buffer_blocknr(split),
+			  path->slots[level + 1] + 1,
 			  level + 1);
 	if (wret)
 		ret = wret;
 
 	if (path->slots[level] >= mid) {
 		path->slots[level] -= mid;
-		btrfs_block_release(root, t);
-		path->nodes[level] = split_buffer;
+		free_extent_buffer(c);
+		path->nodes[level] = split;
 		path->slots[level + 1] += 1;
 	} else {
-		btrfs_block_release(root, split_buffer);
+		free_extent_buffer(split);
 	}
 	return ret;
 }
@@ -1271,16 +1348,16 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
  * and nr indicate which items in the leaf to check.  This totals up the
  * space used both by the item structs and the item data
  */
-static int leaf_space_used(struct btrfs_leaf *l, int start, int nr)
+static int leaf_space_used(struct extent_buffer *l, int start, int nr)
 {
 	int data_len;
-	int nritems = btrfs_header_nritems(&l->header);
+	int nritems = btrfs_header_nritems(l);
 	int end = min(nritems, start + nr) - 1;
 
 	if (!nr)
 		return 0;
-	data_len = btrfs_item_end(l->items + start);
-	data_len = data_len - btrfs_item_offset(l->items + end);
+	data_len = btrfs_item_end_nr(l, start);
+	data_len = data_len - btrfs_item_offset_nr(l, end);
 	data_len += sizeof(struct btrfs_item) * nr;
 	WARN_ON(data_len < 0);
 	return data_len;
@@ -1291,10 +1368,17 @@ static int leaf_space_used(struct btrfs_leaf *l, int start, int nr)
  * the start of the leaf data.  IOW, how much room
  * the leaf has left for both items and data
  */
-int btrfs_leaf_free_space(struct btrfs_root *root, struct btrfs_leaf *leaf)
+int btrfs_leaf_free_space(struct btrfs_root *root, struct extent_buffer *leaf)
 {
-	int nritems = btrfs_header_nritems(&leaf->header);
-	return BTRFS_LEAF_DATA_SIZE(root) - leaf_space_used(leaf, 0, nritems);
+	int nritems = btrfs_header_nritems(leaf);
+	int ret;
+	ret = BTRFS_LEAF_DATA_SIZE(root) - leaf_space_used(leaf, 0, nritems);
+	if (ret < 0) {
+		printk("leaf free space ret %d, leaf data size %lu, used %d nritems %d\n",
+		       ret, BTRFS_LEAF_DATA_SIZE(root),
+		       leaf_space_used(leaf, 0, nritems), nritems);
+	}
+	return ret;
 }
 
 /*
@@ -1307,12 +1391,10 @@ int btrfs_leaf_free_space(struct btrfs_root *root, struct btrfs_leaf *leaf)
 static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 			   *root, struct btrfs_path *path, int data_size)
 {
-	struct buffer_head *left_buf = path->nodes[0];
-	struct btrfs_leaf *left = btrfs_buffer_leaf(left_buf);
-	struct btrfs_leaf *right;
-	struct buffer_head *right_buf;
-	struct buffer_head *upper;
-	struct btrfs_node *upper_node;
+	struct extent_buffer *left = path->nodes[0];
+	struct extent_buffer *right;
+	struct extent_buffer *upper;
+	struct btrfs_disk_key disk_key;
 	int slot;
 	int i;
 	int free_space;
@@ -1321,6 +1403,7 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 	struct btrfs_item *item;
 	u32 left_nritems;
 	u32 right_nritems;
+	u32 data_end;
 	int ret;
 
 	slot = path->slots[1];
@@ -1328,102 +1411,109 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 		return 1;
 	}
 	upper = path->nodes[1];
-	upper_node = btrfs_buffer_node(upper);
-	if (slot >= btrfs_header_nritems(&upper_node->header) - 1) {
+	if (slot >= btrfs_header_nritems(upper) - 1)
 		return 1;
-	}
-	right_buf = read_tree_block(root,
-		    btrfs_node_blockptr(btrfs_buffer_node(upper), slot + 1));
-	right = btrfs_buffer_leaf(right_buf);
+
+	right = read_tree_block(root, btrfs_node_blockptr(upper, slot + 1));
 	free_space = btrfs_leaf_free_space(root, right);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
-		btrfs_block_release(root, right_buf);
+		free_extent_buffer(right);
 		return 1;
 	}
+
 	/* cow and double check */
-	ret = btrfs_cow_block(trans, root, right_buf, upper,
-			      slot + 1, &right_buf);
+	ret = btrfs_cow_block(trans, root, right, upper,
+			      slot + 1, &right);
 	if (ret) {
-		btrfs_block_release(root, right_buf);
+		free_extent_buffer(right);
 		return 1;
 	}
-	right = btrfs_buffer_leaf(right_buf);
 	free_space = btrfs_leaf_free_space(root, right);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
-		btrfs_block_release(root, right_buf);
+		free_extent_buffer(right);
 		return 1;
 	}
 
-	left_nritems = btrfs_header_nritems(&left->header);
+	left_nritems = btrfs_header_nritems(left);
 	if (left_nritems == 0) {
-		btrfs_block_release(root, right_buf);
+		free_extent_buffer(right);
 		return 1;
 	}
+
 	for (i = left_nritems - 1; i >= 1; i--) {
-		item = left->items + i;
+		item = btrfs_item_nr(left, i);
 		if (path->slots[0] == i)
 			push_space += data_size + sizeof(*item);
-		if (btrfs_item_size(item) + sizeof(*item) + push_space >
+		if (btrfs_item_size(left, item) + sizeof(*item) + push_space >
 		    free_space)
 			break;
 		push_items++;
-		push_space += btrfs_item_size(item) + sizeof(*item);
+		push_space += btrfs_item_size(left, item) + sizeof(*item);
 	}
+
 	if (push_items == 0) {
-		btrfs_block_release(root, right_buf);
+		free_extent_buffer(right);
 		return 1;
 	}
+
 	if (push_items == left_nritems)
 		WARN_ON(1);
-	right_nritems = btrfs_header_nritems(&right->header);
+
 	/* push left to right */
-	push_space = btrfs_item_end(left->items + left_nritems - push_items);
+	right_nritems = btrfs_header_nritems(right);
+	push_space = btrfs_item_end_nr(left, left_nritems - push_items);
 	push_space -= leaf_data_end(root, left);
+
 	/* make room in the right data area */
-	btrfs_memmove(root, right, btrfs_leaf_data(right) +
-		      leaf_data_end(root, right) - push_space,
-		      btrfs_leaf_data(right) +
-		      leaf_data_end(root, right), BTRFS_LEAF_DATA_SIZE(root) -
-		      leaf_data_end(root, right));
+	data_end = leaf_data_end(root, right);
+	memmove_extent_buffer(right,
+			      btrfs_leaf_data(right) + data_end - push_space,
+			      btrfs_leaf_data(right) + data_end,
+			      BTRFS_LEAF_DATA_SIZE(root) - data_end);
+
 	/* copy from the left data area */
-	btrfs_memcpy(root, right, btrfs_leaf_data(right) +
+	copy_extent_buffer(right, left, btrfs_leaf_data(right) +
 		     BTRFS_LEAF_DATA_SIZE(root) - push_space,
 		     btrfs_leaf_data(left) + leaf_data_end(root, left),
 		     push_space);
-	btrfs_memmove(root, right, right->items + push_items, right->items,
-		right_nritems * sizeof(struct btrfs_item));
+
+	memmove_extent_buffer(right, btrfs_item_nr_offset(push_items),
+			      btrfs_item_nr_offset(0),
+			      right_nritems * sizeof(struct btrfs_item));
+
 	/* copy the items from left to right */
-	btrfs_memcpy(root, right, right->items, left->items +
-		     left_nritems - push_items,
-		     push_items * sizeof(struct btrfs_item));
+	copy_extent_buffer(right, left, btrfs_item_nr_offset(0),
+		   btrfs_item_nr_offset(left_nritems - push_items),
+		   push_items * sizeof(struct btrfs_item));
 
 	/* update the item pointers */
 	right_nritems += push_items;
-	btrfs_set_header_nritems(&right->header, right_nritems);
+	btrfs_set_header_nritems(right, right_nritems);
 	push_space = BTRFS_LEAF_DATA_SIZE(root);
 	for (i = 0; i < right_nritems; i++) {
-		btrfs_set_item_offset(right->items + i, push_space -
-				      btrfs_item_size(right->items + i));
-		push_space = btrfs_item_offset(right->items + i);
+		item = btrfs_item_nr(right, i);
+		btrfs_set_item_offset(right, item, push_space -
+				      btrfs_item_size(right, item));
+		push_space = btrfs_item_offset(right, item);
 	}
 	left_nritems -= push_items;
-	btrfs_set_header_nritems(&left->header, left_nritems);
+	btrfs_set_header_nritems(left, left_nritems);
 
-	btrfs_mark_buffer_dirty(left_buf);
-	btrfs_mark_buffer_dirty(right_buf);
+	btrfs_mark_buffer_dirty(left);
+	btrfs_mark_buffer_dirty(right);
 
-	btrfs_memcpy(root, upper_node, &upper_node->ptrs[slot + 1].key,
-		&right->items[0].key, sizeof(struct btrfs_disk_key));
+	btrfs_item_key(right, &disk_key, 0);
+	btrfs_set_node_key(upper, &disk_key, slot + 1);
 	btrfs_mark_buffer_dirty(upper);
 
 	/* then fixup the leaf pointer in the path */
 	if (path->slots[0] >= left_nritems) {
 		path->slots[0] -= left_nritems;
-		btrfs_block_release(root, path->nodes[0]);
-		path->nodes[0] = right_buf;
+		free_extent_buffer(path->nodes[0]);
+		path->nodes[0] = right;
 		path->slots[1] += 1;
 	} else {
-		btrfs_block_release(root, right_buf);
+		free_extent_buffer(right);
 	}
 	if (path->nodes[1])
 		check_node(root, path, 1);
@@ -1436,10 +1526,9 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 			  *root, struct btrfs_path *path, int data_size)
 {
-	struct buffer_head *right_buf = path->nodes[0];
-	struct btrfs_leaf *right = btrfs_buffer_leaf(right_buf);
-	struct buffer_head *t;
-	struct btrfs_leaf *left;
+	struct btrfs_disk_key disk_key;
+	struct extent_buffer *right = path->nodes[0];
+	struct extent_buffer *left;
 	int slot;
 	int i;
 	int free_space;
@@ -1447,119 +1536,128 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 	int push_items = 0;
 	struct btrfs_item *item;
 	u32 old_left_nritems;
+	u32 right_nritems;
 	int ret = 0;
 	int wret;
 
 	slot = path->slots[1];
-	if (slot == 0) {
+	if (slot == 0)
 		return 1;
-	}
-	if (!path->nodes[1]) {
+	if (!path->nodes[1])
 		return 1;
-	}
-	t = read_tree_block(root,
-	    btrfs_node_blockptr(btrfs_buffer_node(path->nodes[1]), slot - 1));
-	left = btrfs_buffer_leaf(t);
+
+	left = read_tree_block(root, btrfs_node_blockptr(path->nodes[1],
+							 slot - 1));
 	free_space = btrfs_leaf_free_space(root, left);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
-		btrfs_block_release(root, t);
+		free_extent_buffer(left);
 		return 1;
 	}
 
 	/* cow and double check */
-	ret = btrfs_cow_block(trans, root, t, path->nodes[1], slot - 1, &t);
+	ret = btrfs_cow_block(trans, root, left,
+			      path->nodes[1], slot - 1, &left);
 	if (ret) {
 		/* we hit -ENOSPC, but it isn't fatal here */
-		btrfs_block_release(root, t);
+		free_extent_buffer(left);
 		return 1;
 	}
-	left = btrfs_buffer_leaf(t);
 	free_space = btrfs_leaf_free_space(root, left);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
-		btrfs_block_release(root, t);
+		free_extent_buffer(left);
 		return 1;
 	}
 
-	if (btrfs_header_nritems(&right->header) == 0) {
-		btrfs_block_release(root, t);
+	right_nritems = btrfs_header_nritems(right);
+	if (right_nritems == 0) {
+		free_extent_buffer(left);
 		return 1;
 	}
 
-	for (i = 0; i < btrfs_header_nritems(&right->header) - 1; i++) {
-		item = right->items + i;
+	for (i = 0; i < right_nritems - 1; i++) {
+		item = btrfs_item_nr(right, i);
 		if (path->slots[0] == i)
 			push_space += data_size + sizeof(*item);
-		if (btrfs_item_size(item) + sizeof(*item) + push_space >
+		if (btrfs_item_size(right, item) + sizeof(*item) + push_space >
 		    free_space)
 			break;
 		push_items++;
-		push_space += btrfs_item_size(item) + sizeof(*item);
+		push_space += btrfs_item_size(right, item) + sizeof(*item);
 	}
 	if (push_items == 0) {
-		btrfs_block_release(root, t);
+		free_extent_buffer(left);
 		return 1;
 	}
-	if (push_items == btrfs_header_nritems(&right->header))
+	if (push_items == btrfs_header_nritems(right))
 		WARN_ON(1);
+
 	/* push data from right to left */
-	btrfs_memcpy(root, left, left->items +
-		     btrfs_header_nritems(&left->header),
-		     right->items, push_items * sizeof(struct btrfs_item));
+	copy_extent_buffer(left, right,
+			   btrfs_item_nr_offset(btrfs_header_nritems(left)),
+			   btrfs_item_nr_offset(0),
+			   push_items * sizeof(struct btrfs_item));
+
 	push_space = BTRFS_LEAF_DATA_SIZE(root) -
-		     btrfs_item_offset(right->items + push_items -1);
-	btrfs_memcpy(root, left, btrfs_leaf_data(left) +
+		     btrfs_item_offset_nr(right, push_items -1);
+
+	copy_extent_buffer(left, right, btrfs_leaf_data(left) +
 		     leaf_data_end(root, left) - push_space,
 		     btrfs_leaf_data(right) +
-		     btrfs_item_offset(right->items + push_items - 1),
+		     btrfs_item_offset_nr(right, push_items - 1),
 		     push_space);
-	old_left_nritems = btrfs_header_nritems(&left->header);
+	old_left_nritems = btrfs_header_nritems(left);
 	BUG_ON(old_left_nritems < 0);
 
 	for (i = old_left_nritems; i < old_left_nritems + push_items; i++) {
-		u32 ioff = btrfs_item_offset(left->items + i);
-		btrfs_set_item_offset(left->items + i, ioff -
-				     (BTRFS_LEAF_DATA_SIZE(root) -
-				      btrfs_item_offset(left->items +
-						        old_left_nritems - 1)));
+		u32 ioff;
+		item = btrfs_item_nr(left, i);
+		ioff = btrfs_item_offset(left, item);
+		btrfs_set_item_offset(left, item,
+		      ioff - (BTRFS_LEAF_DATA_SIZE(root) -
+		      btrfs_item_offset_nr(left, old_left_nritems - 1)));
 	}
-	btrfs_set_header_nritems(&left->header, old_left_nritems + push_items);
+	btrfs_set_header_nritems(left, old_left_nritems + push_items);
 
 	/* fixup right node */
-	push_space = btrfs_item_offset(right->items + push_items - 1) -
-		     leaf_data_end(root, right);
-	btrfs_memmove(root, right, btrfs_leaf_data(right) +
-		      BTRFS_LEAF_DATA_SIZE(root) - push_space,
-		      btrfs_leaf_data(right) +
-		      leaf_data_end(root, right), push_space);
-	btrfs_memmove(root, right, right->items, right->items + push_items,
-		(btrfs_header_nritems(&right->header) - push_items) *
-		sizeof(struct btrfs_item));
-	btrfs_set_header_nritems(&right->header,
-				 btrfs_header_nritems(&right->header) -
-				 push_items);
+	push_space = btrfs_item_offset_nr(right, push_items - 1) -
+					  leaf_data_end(root, right);
+	memmove_extent_buffer(right, btrfs_leaf_data(right) +
+			      BTRFS_LEAF_DATA_SIZE(root) - push_space,
+			      btrfs_leaf_data(right) +
+			      leaf_data_end(root, right), push_space);
+
+	memmove_extent_buffer(right, btrfs_item_nr_offset(0),
+			      btrfs_item_nr_offset(push_items),
+			     (btrfs_header_nritems(right) - push_items) *
+			     sizeof(struct btrfs_item));
+
+	right_nritems = btrfs_header_nritems(right) - push_items;
+	btrfs_set_header_nritems(right, right_nritems);
 	push_space = BTRFS_LEAF_DATA_SIZE(root);
 
-	for (i = 0; i < btrfs_header_nritems(&right->header); i++) {
-		btrfs_set_item_offset(right->items + i, push_space -
-				      btrfs_item_size(right->items + i));
-		push_space = btrfs_item_offset(right->items + i);
+	for (i = 0; i < right_nritems; i++) {
+		item = btrfs_item_nr(right, i);
+		btrfs_set_item_offset(right, item, push_space -
+				      btrfs_item_size(right, item));
+		push_space = btrfs_item_offset(right, item);
 	}
 
-	btrfs_mark_buffer_dirty(t);
-	btrfs_mark_buffer_dirty(right_buf);
+	btrfs_mark_buffer_dirty(left);
+	btrfs_mark_buffer_dirty(right);
 
-	wret = fixup_low_keys(trans, root, path, &right->items[0].key, 1);
+	btrfs_item_key(right, &disk_key, 0);
+	wret = fixup_low_keys(trans, root, path, &disk_key, 1);
 	if (wret)
 		ret = wret;
 
 	/* then fixup the leaf pointer in the path */
 	if (path->slots[0] < push_items) {
 		path->slots[0] += old_left_nritems;
-		btrfs_block_release(root, path->nodes[0]);
-		path->nodes[0] = t;
+		free_extent_buffer(path->nodes[0]);
+		path->nodes[0] = left;
 		path->slots[1] -= 1;
 	} else {
-		btrfs_block_release(root, t);
+		free_extent_buffer(left);
 		path->slots[0] -= push_items;
 	}
 	BUG_ON(path->slots[0] < 0);
@@ -1578,13 +1676,11 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_key *ins_key,
 		      struct btrfs_path *path, int data_size)
 {
-	struct buffer_head *l_buf;
-	struct btrfs_leaf *l;
+	struct extent_buffer *l;
 	u32 nritems;
 	int mid;
 	int slot;
-	struct btrfs_leaf *right;
-	struct buffer_head *right_buffer;
+	struct extent_buffer *right;
 	int space_needed = data_size + sizeof(struct btrfs_item);
 	int data_copy_size;
 	int rt_data_off;
@@ -1603,8 +1699,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 		if (wret < 0)
 			return wret;
 	}
-	l_buf = path->nodes[0];
-	l = btrfs_buffer_leaf(l_buf);
+	l = path->nodes[0];
 
 	/* did the pushes work? */
 	if (btrfs_leaf_free_space(root, l) >=
@@ -1617,36 +1712,38 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 			return ret;
 	}
 	slot = path->slots[0];
-	nritems = btrfs_header_nritems(&l->header);
+	nritems = btrfs_header_nritems(l);
 	mid = (nritems + 1)/ 2;
 
-	right_buffer = btrfs_alloc_free_block(trans, root, l_buf->b_blocknr, 0);
-	if (IS_ERR(right_buffer))
-		return PTR_ERR(right_buffer);
-
-	right = btrfs_buffer_leaf(right_buffer);
-	memset(&right->header, 0, sizeof(right->header));
-	btrfs_set_header_blocknr(&right->header, bh_blocknr(right_buffer));
-	btrfs_set_header_generation(&right->header, trans->transid);
-	btrfs_set_header_owner(&right->header, root->root_key.objectid);
-	btrfs_set_header_level(&right->header, 0);
-	memcpy(right->header.fsid, root->fs_info->disk_super->fsid,
-	       sizeof(right->header.fsid));
+	right = btrfs_alloc_free_block(trans, root,
+					      extent_buffer_blocknr(l), 0);
+	if (IS_ERR(right))
+		return PTR_ERR(right);
+
+	memset_extent_buffer(right, 0, 0, sizeof(struct btrfs_header));
+	btrfs_set_header_blocknr(right, extent_buffer_blocknr(right));
+	btrfs_set_header_generation(right, trans->transid);
+	btrfs_set_header_owner(right, root->root_key.objectid);
+	btrfs_set_header_level(right, 0);
+	write_extent_buffer(right, root->fs_info->fsid,
+			    (unsigned long)btrfs_header_fsid(right),
+			    BTRFS_FSID_SIZE);
+
 	if (mid <= slot) {
 		if (nritems == 1 ||
 		    leaf_space_used(l, mid, nritems - mid) + space_needed >
 			BTRFS_LEAF_DATA_SIZE(root)) {
 			if (slot >= nritems) {
 				btrfs_cpu_key_to_disk(&disk_key, ins_key);
-				btrfs_set_header_nritems(&right->header, 0);
+				btrfs_set_header_nritems(right, 0);
 				wret = insert_ptr(trans, root, path,
 						  &disk_key,
-						  bh_blocknr(right_buffer),
+						  extent_buffer_blocknr(right),
 						  path->slots[1] + 1, 1);
 				if (wret)
 					ret = wret;
-				btrfs_block_release(root, path->nodes[0]);
-				path->nodes[0] = right_buffer;
+				free_extent_buffer(path->nodes[0]);
+				path->nodes[0] = right;
 				path->slots[0] = 0;
 				path->slots[1] += 1;
 				return ret;
@@ -1659,15 +1756,15 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 			BTRFS_LEAF_DATA_SIZE(root)) {
 			if (slot == 0) {
 				btrfs_cpu_key_to_disk(&disk_key, ins_key);
-				btrfs_set_header_nritems(&right->header, 0);
+				btrfs_set_header_nritems(right, 0);
 				wret = insert_ptr(trans, root, path,
 						  &disk_key,
-						  bh_blocknr(right_buffer),
+						  extent_buffer_blocknr(right),
 						  path->slots[1], 1);
 				if (wret)
 					ret = wret;
-				btrfs_block_release(root, path->nodes[0]);
-				path->nodes[0] = right_buffer;
+				free_extent_buffer(path->nodes[0]);
+				path->nodes[0] = right;
 				path->slots[0] = 0;
 				if (path->slots[1] == 0) {
 					wret = fixup_low_keys(trans, root,
@@ -1681,61 +1778,74 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 			double_split = 1;
 		}
 	}
-	btrfs_set_header_nritems(&right->header, nritems - mid);
-	data_copy_size = btrfs_item_end(l->items + mid) -
-			 leaf_data_end(root, l);
-	btrfs_memcpy(root, right, right->items, l->items + mid,
-		     (nritems - mid) * sizeof(struct btrfs_item));
-	btrfs_memcpy(root, right,
+	nritems = nritems - mid;
+	btrfs_set_header_nritems(right, nritems);
+	data_copy_size = btrfs_item_end_nr(l, mid) - leaf_data_end(root, l);
+
+	copy_extent_buffer(right, l, btrfs_item_nr_offset(0),
+			   btrfs_item_nr_offset(mid),
+			   nritems * sizeof(struct btrfs_item));
+
+	copy_extent_buffer(right, l,
 		     btrfs_leaf_data(right) + BTRFS_LEAF_DATA_SIZE(root) -
 		     data_copy_size, btrfs_leaf_data(l) +
 		     leaf_data_end(root, l), data_copy_size);
+
 	rt_data_off = BTRFS_LEAF_DATA_SIZE(root) -
-		      btrfs_item_end(l->items + mid);
+		      btrfs_item_end_nr(l, mid);
 
-	for (i = 0; i < btrfs_header_nritems(&right->header); i++) {
-		u32 ioff = btrfs_item_offset(right->items + i);
-		btrfs_set_item_offset(right->items + i, ioff + rt_data_off);
+	for (i = 0; i < nritems; i++) {
+		struct btrfs_item *item = btrfs_item_nr(right, i);
+		u32 ioff = btrfs_item_offset(right, item);
+		btrfs_set_item_offset(right, item, ioff + rt_data_off);
 	}
 
-	btrfs_set_header_nritems(&l->header, mid);
+	btrfs_set_header_nritems(l, mid);
 	ret = 0;
-	wret = insert_ptr(trans, root, path, &right->items[0].key,
-			  bh_blocknr(right_buffer), path->slots[1] + 1, 1);
+	btrfs_item_key(right, &disk_key, 0);
+	wret = insert_ptr(trans, root, path, &disk_key,
+			  extent_buffer_blocknr(right), path->slots[1] + 1, 1);
 	if (wret)
 		ret = wret;
-	btrfs_mark_buffer_dirty(right_buffer);
-	btrfs_mark_buffer_dirty(l_buf);
+
+	btrfs_mark_buffer_dirty(right);
+	btrfs_mark_buffer_dirty(l);
 	BUG_ON(path->slots[0] != slot);
+
 	if (mid <= slot) {
-		btrfs_block_release(root, path->nodes[0]);
-		path->nodes[0] = right_buffer;
+		free_extent_buffer(path->nodes[0]);
+		path->nodes[0] = right;
 		path->slots[0] -= mid;
 		path->slots[1] += 1;
 	} else
-		btrfs_block_release(root, right_buffer);
+		free_extent_buffer(right);
+
 	BUG_ON(path->slots[0] < 0);
 	check_node(root, path, 1);
+	check_leaf(root, path, 0);
 
 	if (!double_split)
 		return ret;
-	right_buffer = btrfs_alloc_free_block(trans, root, l_buf->b_blocknr, 0);
-	if (IS_ERR(right_buffer))
-		return PTR_ERR(right_buffer);
-
-	right = btrfs_buffer_leaf(right_buffer);
-	memset(&right->header, 0, sizeof(right->header));
-	btrfs_set_header_blocknr(&right->header, bh_blocknr(right_buffer));
-	btrfs_set_header_generation(&right->header, trans->transid);
-	btrfs_set_header_owner(&right->header, root->root_key.objectid);
-	btrfs_set_header_level(&right->header, 0);
-	memcpy(right->header.fsid, root->fs_info->disk_super->fsid,
-	       sizeof(right->header.fsid));
+
+	right = btrfs_alloc_free_block(trans, root,
+				       extent_buffer_blocknr(l), 0);
+	if (IS_ERR(right))
+		return PTR_ERR(right);
+
+	memset_extent_buffer(right, 0, 0, sizeof(struct btrfs_header));
+	btrfs_set_header_blocknr(right, extent_buffer_blocknr(right));
+	btrfs_set_header_generation(right, trans->transid);
+	btrfs_set_header_owner(right, root->root_key.objectid);
+	btrfs_set_header_level(right, 0);
+	write_extent_buffer(right, root->fs_info->fsid,
+			    (unsigned long)btrfs_header_fsid(right),
+			    BTRFS_FSID_SIZE);
+
 	btrfs_cpu_key_to_disk(&disk_key, ins_key);
-	btrfs_set_header_nritems(&right->header, 0);
+	btrfs_set_header_nritems(right, 0);
 	wret = insert_ptr(trans, root, path,
 			  &disk_key,
-			  bh_blocknr(right_buffer),
+			  extent_buffer_blocknr(right),
 			  path->slots[1], 1);
 	if (wret)
 		ret = wret;
@@ -1744,8 +1854,8 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 		if (wret)
 			ret = wret;
 	}
-	btrfs_block_release(root, path->nodes[0]);
-	path->nodes[0] = right_buffer;
+	free_extent_buffer(path->nodes[0]);
+	path->nodes[0] = right;
 	path->slots[0] = 0;
 	check_node(root, path, 1);
 	check_leaf(root, path, 0);
@@ -1760,8 +1870,8 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans,
 	int ret = 0;
 	int slot;
 	int slot_orig;
-	struct btrfs_leaf *leaf;
-	struct buffer_head *leaf_buf;
+	struct extent_buffer *leaf;
+	struct btrfs_item *item;
 	u32 nritems;
 	unsigned int data_end;
 	unsigned int old_data_start;
@@ -1770,15 +1880,14 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans,
 	int i;
 
 	slot_orig = path->slots[0];
-	leaf_buf = path->nodes[0];
-	leaf = btrfs_buffer_leaf(leaf_buf);
+	leaf = path->nodes[0];
 
-	nritems = btrfs_header_nritems(&leaf->header);
+	nritems = btrfs_header_nritems(leaf);
 	data_end = leaf_data_end(root, leaf);
 
 	slot = path->slots[0];
-	old_data_start = btrfs_item_offset(leaf->items + slot);
-	old_size = btrfs_item_size(leaf->items + slot);
+	old_data_start = btrfs_item_offset_nr(leaf, slot);
+	old_size = btrfs_item_size_nr(leaf, slot);
 	BUG_ON(old_size <= new_size);
 	size_diff = old_size - new_size;
 
@@ -1790,32 +1899,38 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans,
 	 */
 	/* first correct the data pointers */
 	for (i = slot; i < nritems; i++) {
-		u32 ioff = btrfs_item_offset(leaf->items + i);
-		btrfs_set_item_offset(leaf->items + i,
-				      ioff + size_diff);
+		u32 ioff;
+		item = btrfs_item_nr(leaf, i);
+		ioff = btrfs_item_offset(leaf, item);
+		btrfs_set_item_offset(leaf, item, ioff + size_diff);
 	}
 	/* shift the data */
-	btrfs_memmove(root, leaf, btrfs_leaf_data(leaf) +
+	memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) +
 		      data_end + size_diff, btrfs_leaf_data(leaf) +
 		      data_end, old_data_start + new_size - data_end);
-	btrfs_set_item_size(leaf->items + slot, new_size);
-	btrfs_mark_buffer_dirty(leaf_buf);
+
+	item = btrfs_item_nr(leaf, slot);
+	btrfs_set_item_size(leaf, item, new_size);
+	btrfs_mark_buffer_dirty(leaf);
 
 	ret = 0;
-	if (btrfs_leaf_free_space(root, leaf) < 0)
+	if (btrfs_leaf_free_space(root, leaf) < 0) {
+		btrfs_print_leaf(root, leaf);
 		BUG();
+	}
 	check_leaf(root, path, 0);
 	return ret;
 }
 
-int btrfs_extend_item(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, struct btrfs_path *path, u32 data_size)
+int btrfs_extend_item(struct btrfs_trans_handle *trans,
+		      struct btrfs_root *root, struct btrfs_path *path,
+		      u32 data_size)
 {
 	int ret = 0;
 	int slot;
 	int slot_orig;
-	struct btrfs_leaf *leaf;
-	struct buffer_head *leaf_buf;
+	struct extent_buffer *leaf;
+	struct btrfs_item *item;
 	u32 nritems;
 	unsigned int data_end;
 	unsigned int old_data;
@@ -1823,16 +1938,17 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans, struct btrfs_root
 	int i;
 
 	slot_orig = path->slots[0];
-	leaf_buf = path->nodes[0];
-	leaf = btrfs_buffer_leaf(leaf_buf);
+	leaf = path->nodes[0];
 
-	nritems = btrfs_header_nritems(&leaf->header);
+	nritems = btrfs_header_nritems(leaf);
 	data_end = leaf_data_end(root, leaf);
 
-	if (btrfs_leaf_free_space(root, leaf) < data_size)
+	if (btrfs_leaf_free_space(root, leaf) < data_size) {
+		btrfs_print_leaf(root, leaf);
 		BUG();
+	}
 	slot = path->slots[0];
-	old_data = btrfs_item_end(leaf->items + slot);
+	old_data = btrfs_item_end_nr(leaf, slot);
 
 	BUG_ON(slot < 0);
 	BUG_ON(slot >= nritems);
@@ -1842,22 +1958,28 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans, struct btrfs_root
 	 */
 	/* first correct the data pointers */
 	for (i = slot; i < nritems; i++) {
-		u32 ioff = btrfs_item_offset(leaf->items + i);
-		btrfs_set_item_offset(leaf->items + i,
-				      ioff - data_size);
+		u32 ioff;
+		item = btrfs_item_nr(leaf, i);
+		ioff = btrfs_item_offset(leaf, item);
+		btrfs_set_item_offset(leaf, item, ioff - data_size);
 	}
+
 	/* shift the data */
-	btrfs_memmove(root, leaf, btrfs_leaf_data(leaf) +
+	memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) +
 		      data_end - data_size, btrfs_leaf_data(leaf) +
 		      data_end, old_data - data_end);
+
 	data_end = old_data;
-	old_size = btrfs_item_size(leaf->items + slot);
-	btrfs_set_item_size(leaf->items + slot, old_size + data_size);
-	btrfs_mark_buffer_dirty(leaf_buf);
+	old_size = btrfs_item_size_nr(leaf, slot);
+	item = btrfs_item_nr(leaf, slot);
+	btrfs_set_item_size(leaf, item, old_size + data_size);
+	btrfs_mark_buffer_dirty(leaf);
 
 	ret = 0;
-	if (btrfs_leaf_free_space(root, leaf) < 0)
+	if (btrfs_leaf_free_space(root, leaf) < 0) {
+		btrfs_print_leaf(root, leaf);
 		BUG();
+	}
 	check_leaf(root, path, 0);
 	return ret;
 }
@@ -1866,15 +1988,16 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans, struct btrfs_root
  * Given a key and some data, insert an item into the tree.
  * This does all the path init required, making room in the tree if needed.
  */
-int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, struct btrfs_root
-			    *root, struct btrfs_path *path, struct btrfs_key
-			    *cpu_key, u32 data_size)
+int btrfs_insert_empty_item(struct btrfs_trans_handle *trans,
+			    struct btrfs_root *root,
+			    struct btrfs_path *path,
+			    struct btrfs_key *cpu_key, u32 data_size)
 {
+	struct extent_buffer *leaf;
+	struct btrfs_item *item;
 	int ret = 0;
 	int slot;
 	int slot_orig;
-	struct btrfs_leaf *leaf;
-	struct buffer_head *leaf_buf;
 	u32 nritems;
 	unsigned int data_end;
 	struct btrfs_disk_key disk_key;
@@ -1884,6 +2007,7 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, struct btrfs_root
 	/* create a root if there isn't one */
 	if (!root->node)
 		BUG();
+
 	ret = btrfs_search_slot(trans, root, cpu_key, path, data_size, 1);
 	if (ret == 0) {
 		return -EEXIST;
@@ -1892,57 +2016,68 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, struct btrfs_root
 		goto out;
 
 	slot_orig = path->slots[0];
-	leaf_buf = path->nodes[0];
-	leaf = btrfs_buffer_leaf(leaf_buf);
+	leaf = path->nodes[0];
 
-	nritems = btrfs_header_nritems(&leaf->header);
+	nritems = btrfs_header_nritems(leaf);
 	data_end = leaf_data_end(root, leaf);
 
 	if (btrfs_leaf_free_space(root, leaf) <
 	    sizeof(struct btrfs_item) + data_size) {
 		BUG();
 	}
+
 	slot = path->slots[0];
 	BUG_ON(slot < 0);
+
 	if (slot != nritems) {
 		int i;
-		unsigned int old_data = btrfs_item_end(leaf->items + slot);
+		unsigned int old_data = btrfs_item_end_nr(leaf, slot);
 
+		if (old_data < data_end) {
+			btrfs_print_leaf(root, leaf);
+			printk("slot %d old_data %d data_end %d\n",
+			       slot, old_data, data_end);
+			BUG_ON(1);
+		}
 		/*
 		 * item0..itemN ... dataN.offset..dataN.size .. data0.size
 		 */
 		/* first correct the data pointers */
 		for (i = slot; i < nritems; i++) {
-			u32 ioff = btrfs_item_offset(leaf->items + i);
-			btrfs_set_item_offset(leaf->items + i,
-					      ioff - data_size);
+			u32 ioff;
+			item = btrfs_item_nr(leaf, i);
+			ioff = btrfs_item_offset(leaf, item);
+			btrfs_set_item_offset(leaf, item, ioff - data_size);
 		}
 
 		/* shift the items */
-		btrfs_memmove(root, leaf, leaf->items + slot + 1,
-			      leaf->items + slot,
+		memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + 1),
+			      btrfs_item_nr_offset(slot),
 			      (nritems - slot) * sizeof(struct btrfs_item));
 
 		/* shift the data */
-		btrfs_memmove(root, leaf, btrfs_leaf_data(leaf) +
+		memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) +
 			      data_end - data_size, btrfs_leaf_data(leaf) +
 			      data_end, old_data - data_end);
 		data_end = old_data;
 	}
+
 	/* setup the item for the new data */
-	btrfs_memcpy(root, leaf, &leaf->items[slot].key, &disk_key,
-		     sizeof(struct btrfs_disk_key));
-	btrfs_set_item_offset(leaf->items + slot, data_end - data_size);
-	btrfs_set_item_size(leaf->items + slot, data_size);
-	btrfs_set_header_nritems(&leaf->header, nritems + 1);
-	btrfs_mark_buffer_dirty(leaf_buf);
+	btrfs_set_item_key(leaf, &disk_key, slot);
+	item = btrfs_item_nr(leaf, slot);
+	btrfs_set_item_offset(leaf, item, data_end - data_size);
+	btrfs_set_item_size(leaf, item, data_size);
+	btrfs_set_header_nritems(leaf, nritems + 1);
+	btrfs_mark_buffer_dirty(leaf);
 
 	ret = 0;
 	if (slot == 0)
 		ret = fixup_low_keys(trans, root, path, &disk_key, 1);
 
-	if (btrfs_leaf_free_space(root, leaf) < 0)
+	if (btrfs_leaf_free_space(root, leaf) < 0) {
+		btrfs_print_leaf(root, leaf);
 		BUG();
+	}
 	check_leaf(root, path, 0);
 out:
 	return ret;
@@ -1958,17 +2093,17 @@ int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root
 {
 	int ret = 0;
 	struct btrfs_path *path;
-	u8 *ptr;
+	struct extent_buffer *leaf;
+	unsigned long ptr;
 
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
 	ret = btrfs_insert_empty_item(trans, root, path, cpu_key, data_size);
 	if (!ret) {
-		ptr = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
-				     path->slots[0], u8);
-		btrfs_memcpy(root, path->nodes[0]->b_data,
-			     ptr, data, data_size);
-		btrfs_mark_buffer_dirty(path->nodes[0]);
+		leaf = path->nodes[0];
+		ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
+		write_extent_buffer(leaf, data, ptr, data_size);
+		btrfs_mark_buffer_dirty(leaf);
 	}
 	btrfs_free_path(path);
 	return ret;
@@ -1984,30 +2119,30 @@ int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root
 static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		   struct btrfs_path *path, int level, int slot)
 {
-	struct btrfs_node *node;
-	struct buffer_head *parent = path->nodes[level];
+	struct extent_buffer *parent = path->nodes[level];
 	u32 nritems;
 	int ret = 0;
 	int wret;
 
-	node = btrfs_buffer_node(parent);
-	nritems = btrfs_header_nritems(&node->header);
+	nritems = btrfs_header_nritems(parent);
 	if (slot != nritems -1) {
-		btrfs_memmove(root, node, node->ptrs + slot,
-			      node->ptrs + slot + 1,
+		memmove_extent_buffer(parent,
+			      btrfs_node_key_ptr_offset(slot),
+			      btrfs_node_key_ptr_offset(slot + 1),
 			      sizeof(struct btrfs_key_ptr) *
 			      (nritems - slot - 1));
 	}
 	nritems--;
-	btrfs_set_header_nritems(&node->header, nritems);
+	btrfs_set_header_nritems(parent, nritems);
 	if (nritems == 0 && parent == root->node) {
-		struct btrfs_header *header = btrfs_buffer_header(root->node);
-		BUG_ON(btrfs_header_level(header) != 1);
+		BUG_ON(btrfs_header_level(root->node) != 1);
 		/* just turn the root into a leaf and break */
-		btrfs_set_header_level(header, 0);
+		btrfs_set_header_level(root->node, 0);
 	} else if (slot == 0) {
-		wret = fixup_low_keys(trans, root, path, &node->ptrs[0].key,
-				      level + 1);
+		struct btrfs_disk_key disk_key;
+
+		btrfs_node_key(parent, &disk_key, 0);
+		wret = fixup_low_keys(trans, root, path, &disk_key, level + 1);
 		if (wret)
 			ret = wret;
 	}
@@ -2023,59 +2158,67 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		   struct btrfs_path *path)
 {
 	int slot;
-	struct btrfs_leaf *leaf;
-	struct buffer_head *leaf_buf;
+	struct extent_buffer *leaf;
+	struct btrfs_item *item;
 	int doff;
 	int dsize;
 	int ret = 0;
 	int wret;
 	u32 nritems;
 
-	leaf_buf = path->nodes[0];
-	leaf = btrfs_buffer_leaf(leaf_buf);
+	leaf = path->nodes[0];
 	slot = path->slots[0];
-	doff = btrfs_item_offset(leaf->items + slot);
-	dsize = btrfs_item_size(leaf->items + slot);
-	nritems = btrfs_header_nritems(&leaf->header);
+	doff = btrfs_item_offset_nr(leaf, slot);
+	dsize = btrfs_item_size_nr(leaf, slot);
+	nritems = btrfs_header_nritems(leaf);
 
 	if (slot != nritems - 1) {
 		int i;
 		int data_end = leaf_data_end(root, leaf);
-		btrfs_memmove(root, leaf, btrfs_leaf_data(leaf) +
+
+		memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) +
 			      data_end + dsize,
 			      btrfs_leaf_data(leaf) + data_end,
 			      doff - data_end);
+
 		for (i = slot + 1; i < nritems; i++) {
-			u32 ioff = btrfs_item_offset(leaf->items + i);
-			btrfs_set_item_offset(leaf->items + i, ioff + dsize);
+			u32 ioff;
+			item = btrfs_item_nr(leaf, i);
+			ioff = btrfs_item_offset(leaf, item);
+			btrfs_set_item_offset(leaf, item, ioff + dsize);
 		}
-		btrfs_memmove(root, leaf, leaf->items + slot,
-			      leaf->items + slot + 1,
+		memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot),
+			      btrfs_item_nr_offset(slot + 1),
 			      sizeof(struct btrfs_item) *
 			      (nritems - slot - 1));
 	}
-	btrfs_set_header_nritems(&leaf->header, nritems - 1);
+	btrfs_set_header_nritems(leaf, nritems - 1);
 	nritems--;
+
 	/* delete the leaf if we've emptied it */
 	if (nritems == 0) {
-		if (leaf_buf == root->node) {
-			btrfs_set_header_level(&leaf->header, 0);
+		if (leaf == root->node) {
+			btrfs_set_header_level(leaf, 0);
 		} else {
-			clean_tree_block(trans, root, leaf_buf);
-			wait_on_buffer(leaf_buf);
+			clean_tree_block(trans, root, leaf);
+			wait_on_tree_block_writeback(root, leaf);
 			wret = del_ptr(trans, root, path, 1, path->slots[1]);
 			if (wret)
 				ret = wret;
 			wret = btrfs_free_extent(trans, root,
-						 bh_blocknr(leaf_buf), 1, 1);
+						 extent_buffer_blocknr(leaf),
+						 1, 1);
 			if (wret)
 				ret = wret;
 		}
 	} else {
 		int used = leaf_space_used(leaf, 0, nritems);
 		if (slot == 0) {
+			struct btrfs_disk_key disk_key;
+
+			btrfs_item_key(leaf, &disk_key, 0);
 			wret = fixup_low_keys(trans, root, path,
-					      &leaf->items[0].key, 1);
+					      &disk_key, 1);
 			if (wret)
 				ret = wret;
 		}
@@ -2087,34 +2230,40 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 			 * for possible call to del_ptr below
 			 */
 			slot = path->slots[1];
-			get_bh(leaf_buf);
+			extent_buffer_get(leaf);
+
 			wret = push_leaf_left(trans, root, path, 1);
 			if (wret < 0 && wret != -ENOSPC)
 				ret = wret;
-			if (path->nodes[0] == leaf_buf &&
-			    btrfs_header_nritems(&leaf->header)) {
+
+			if (path->nodes[0] == leaf &&
+			    btrfs_header_nritems(leaf)) {
 				wret = push_leaf_right(trans, root, path, 1);
 				if (wret < 0 && wret != -ENOSPC)
 					ret = wret;
 			}
-			if (btrfs_header_nritems(&leaf->header) == 0) {
-				u64 blocknr = bh_blocknr(leaf_buf);
-				clean_tree_block(trans, root, leaf_buf);
-				wait_on_buffer(leaf_buf);
+
+			if (btrfs_header_nritems(leaf) == 0) {
+				u64 blocknr = extent_buffer_blocknr(leaf);
+
+				clean_tree_block(trans, root, leaf);
+				wait_on_tree_block_writeback(root, leaf);
+
 				wret = del_ptr(trans, root, path, 1, slot);
 				if (wret)
 					ret = wret;
-				btrfs_block_release(root, leaf_buf);
+
+				free_extent_buffer(leaf);
 				wret = btrfs_free_extent(trans, root, blocknr,
 							 1, 1);
 				if (wret)
 					ret = wret;
 			} else {
-				btrfs_mark_buffer_dirty(leaf_buf);
-				btrfs_block_release(root, leaf_buf);
+				btrfs_mark_buffer_dirty(leaf);
+				free_extent_buffer(leaf);
 			}
 		} else {
-			btrfs_mark_buffer_dirty(leaf_buf);
+			btrfs_mark_buffer_dirty(leaf);
 		}
 	}
 	return ret;
@@ -2130,25 +2279,27 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 	int slot;
 	int level = 1;
 	u64 blocknr;
-	struct buffer_head *c;
-	struct btrfs_node *c_node;
-	struct buffer_head *next = NULL;
+	struct extent_buffer *c;
+	struct extent_buffer *next = NULL;
 
 	while(level < BTRFS_MAX_LEVEL) {
 		if (!path->nodes[level])
 			return 1;
+
 		slot = path->slots[level] + 1;
 		c = path->nodes[level];
-		c_node = btrfs_buffer_node(c);
-		if (slot >= btrfs_header_nritems(&c_node->header)) {
+		if (slot >= btrfs_header_nritems(c)) {
 			level++;
 			continue;
 		}
-		blocknr = btrfs_node_blockptr(c_node, slot);
+
+		blocknr = btrfs_node_blockptr(c, slot);
 		if (next)
-			btrfs_block_release(root, next);
+			free_extent_buffer(next);
+
 		if (path->reada)
 			reada_for_search(root, path, level, slot);
+
 		next = read_tree_block(root, blocknr);
 		break;
 	}
@@ -2156,15 +2307,14 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 	while(1) {
 		level--;
 		c = path->nodes[level];
-		btrfs_block_release(root, c);
+		free_extent_buffer(c);
 		path->nodes[level] = next;
 		path->slots[level] = 0;
 		if (!level)
 			break;
 		if (path->reada)
 			reada_for_search(root, path, level, 0);
-		next = read_tree_block(root,
-		       btrfs_node_blockptr(btrfs_buffer_node(next), 0));
+		next = read_tree_block(root, btrfs_node_blockptr(next, 0));
 	}
 	return 0;
 }
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 800a3499cc3f..c4b829806855 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -20,10 +20,10 @@
 #define __BTRFS__
 
 #include <linux/fs.h>
-#include <linux/buffer_head.h>
 #include <linux/workqueue.h>
 #include <linux/completion.h>
 #include "bit-radix.h"
+#include "extent_map.h"
 
 struct btrfs_trans_handle;
 struct btrfs_transaction;
@@ -32,7 +32,7 @@ extern struct kmem_cache *btrfs_transaction_cachep;
 extern struct kmem_cache *btrfs_bit_radix_cachep;
 extern struct kmem_cache *btrfs_path_cachep;
 
-#define BTRFS_MAGIC "_BtRfS_M"
+#define BTRFS_MAGIC "_B2RfS_M"
 
 #define BTRFS_ROOT_TREE_OBJECTID 1ULL
 #define BTRFS_EXTENT_TREE_OBJECTID 2ULL
@@ -78,41 +78,41 @@ extern struct kmem_cache *btrfs_path_cachep;
  */
 struct btrfs_disk_key {
 	__le64 objectid;
-	__le32 flags;
+	u8 type;
 	__le64 offset;
 } __attribute__ ((__packed__));
 
 struct btrfs_key {
 	u64 objectid;
-	u32 flags;
+	u8 type;
 	u64 offset;
 } __attribute__ ((__packed__));
 
+#define BTRFS_FSID_SIZE 16
 /*
  * every tree block (leaf or node) starts with this header.
  */
 struct btrfs_header {
 	u8 csum[BTRFS_CSUM_SIZE];
-	u8 fsid[16]; /* FS specific uuid */
+	u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
 	__le64 blocknr; /* which block this node is supposed to live in */
 	__le64 generation;
 	__le64 owner;
-	__le16 nritems;
+	__le32 nritems;
 	__le16 flags;
 	u8 level;
 } __attribute__ ((__packed__));
 
 #define BTRFS_MAX_LEVEL 8
-#define BTRFS_NODEPTRS_PER_BLOCK(r) (((r)->blocksize - \
+#define BTRFS_NODEPTRS_PER_BLOCK(r) (((r)->nodesize - \
 			        sizeof(struct btrfs_header)) / \
 			       (sizeof(struct btrfs_disk_key) + sizeof(u64)))
 #define __BTRFS_LEAF_DATA_SIZE(bs) ((bs) - sizeof(struct btrfs_header))
-#define BTRFS_LEAF_DATA_SIZE(r) (__BTRFS_LEAF_DATA_SIZE(r->blocksize))
+#define BTRFS_LEAF_DATA_SIZE(r) (__BTRFS_LEAF_DATA_SIZE(r->leafsize))
 #define BTRFS_MAX_INLINE_DATA_SIZE(r) (BTRFS_LEAF_DATA_SIZE(r) - \
 					sizeof(struct btrfs_item) - \
 					sizeof(struct btrfs_file_extent_item))
 
-struct buffer_head;
 /*
  * the super block basically lists the main trees of the FS
  * it currently lacks any block count etc etc
@@ -128,7 +128,9 @@ struct btrfs_super_block {
 	__le64 total_blocks;
 	__le64 blocks_used;
 	__le64 root_dir_objectid;
-	__le32 blocksize;
+	__le32 sectorsize;
+	__le32 nodesize;
+	__le32 leafsize;
 } __attribute__ ((__packed__));
 
 /*
@@ -138,7 +140,7 @@ struct btrfs_super_block {
 struct btrfs_item {
 	struct btrfs_disk_key key;
 	__le32 offset;
-	__le16 size;
+	__le32 size;
 } __attribute__ ((__packed__));
 
 /*
@@ -176,7 +178,7 @@ struct btrfs_node {
  * used while walking the tree.
  */
 struct btrfs_path {
-	struct buffer_head *nodes[BTRFS_MAX_LEVEL];
+	struct extent_buffer *nodes[BTRFS_MAX_LEVEL];
 	int slots[BTRFS_MAX_LEVEL];
 	int reada;
 	int lowest_level;
@@ -292,6 +294,7 @@ struct btrfs_block_group_cache {
 };
 
 struct btrfs_fs_info {
+	u8 fsid[BTRFS_FSID_SIZE];
 	struct btrfs_root *extent_root;
 	struct btrfs_root *tree_root;
 	struct radix_tree_root fs_roots_radix;
@@ -304,9 +307,8 @@ struct btrfs_fs_info {
 	u64 generation;
 	u64 last_trans_committed;
 	struct btrfs_transaction *running_transaction;
-	struct btrfs_super_block *disk_super;
 	struct btrfs_super_block super_copy;
-	struct buffer_head *sb_buffer;
+	struct extent_buffer *sb_buffer;
 	struct super_block *sb;
 	struct inode *btree_inode;
 	struct mutex trans_mutex;
@@ -325,8 +327,8 @@ struct btrfs_fs_info {
  * and for the extent tree extent_root root.
  */
 struct btrfs_root {
-	struct buffer_head *node;
-	struct buffer_head *commit_root;
+	struct extent_buffer *node;
+	struct extent_buffer *commit_root;
 	struct btrfs_root_item root_item;
 	struct btrfs_key root_key;
 	struct btrfs_fs_info *fs_info;
@@ -336,7 +338,16 @@ struct btrfs_root {
 	struct rw_semaphore snap_sem;
 	u64 objectid;
 	u64 last_trans;
-	u32 blocksize;
+
+	/* data allocations are done in sectorsize units */
+	u32 sectorsize;
+
+	/* node allocations are done in nodesize units */
+	u32 nodesize;
+
+	/* leaf allocations are done in leafsize units */
+	u32 leafsize;
+
 	u32 type;
 	u64 highest_inode;
 	u64 last_inode_alloc;
@@ -347,12 +358,6 @@ struct btrfs_root {
 	char *name;
 };
 
-/* the lower bits in the key flags defines the item type */
-#define BTRFS_KEY_TYPE_MAX	256
-#define BTRFS_KEY_TYPE_SHIFT	24
-#define BTRFS_KEY_TYPE_MASK	(((u32)BTRFS_KEY_TYPE_MAX - 1) << \
-				  BTRFS_KEY_TYPE_SHIFT)
-
 /*
  * inode items have the data typically returned from stat and store other
  * info about object characteristics.  There is one for every file and dir in
@@ -402,246 +407,253 @@ struct btrfs_root {
  */
 #define BTRFS_STRING_ITEM_KEY	253
 
+/* some macros to generate set/get funcs for the struct fields.  This
+ * assumes there is a lefoo_to_cpu for every type, so lets make a simple
+ * one for u8:
+ */
+#define le8_to_cpu(v) (v)
+#define cpu_to_le8(v) (v)
+#define __le8 u8
+
+#define read_eb_member(eb, ptr, type, member, result) (			\
+	read_extent_buffer(eb, (char *)(result),			\
+			   ((unsigned long)(ptr)) +			\
+			    offsetof(type, member),			\
+			   sizeof(((type *)0)->member)))
+
+#define write_eb_member(eb, ptr, type, member, result) (		\
+	write_extent_buffer(eb, (char *)(result),			\
+			   ((unsigned long)(ptr)) +			\
+			    offsetof(type, member),			\
+			   sizeof(((type *)0)->member)))
+
+#define BTRFS_SETGET_FUNCS(name, type, member, bits)			\
+static inline u##bits btrfs_##name(struct extent_buffer *eb,		\
+				   type *s)				\
+{									\
+	__le##bits res;							\
+	read_eb_member(eb, s, type, member, &res);			\
+	return le##bits##_to_cpu(res);					\
+}									\
+static inline void btrfs_set_##name(struct extent_buffer *eb,		\
+				    type *s, u##bits val)		\
+{									\
+	val = cpu_to_le##bits(val);					\
+	write_eb_member(eb, s, type, member, &val);			\
+}
+
+#define BTRFS_SETGET_HEADER_FUNCS(name, type, member, bits)		\
+static inline u##bits btrfs_##name(struct extent_buffer *eb)		\
+{									\
+	__le##bits res;							\
+	read_eb_member(eb, NULL, type, member, &res);			\
+	return le##bits##_to_cpu(res);					\
+}									\
+static inline void btrfs_set_##name(struct extent_buffer *eb,		\
+				    u##bits val)			\
+{									\
+	val = cpu_to_le##bits(val);					\
+	write_eb_member(eb, NULL, type, member, &val);			\
+}
 
-static inline u64 btrfs_block_group_used(struct btrfs_block_group_item *bi)
-{
-	return le64_to_cpu(bi->used);
-}
-
-static inline void btrfs_set_block_group_used(struct
-						   btrfs_block_group_item *bi,
-						   u64 val)
-{
-	bi->used = cpu_to_le64(val);
-}
-
-static inline u64 btrfs_inode_generation(struct btrfs_inode_item *i)
-{
-	return le64_to_cpu(i->generation);
-}
-
-static inline void btrfs_set_inode_generation(struct btrfs_inode_item *i,
-					      u64 val)
-{
-	i->generation = cpu_to_le64(val);
-}
-
-static inline u64 btrfs_inode_size(struct btrfs_inode_item *i)
-{
-	return le64_to_cpu(i->size);
-}
-
-static inline void btrfs_set_inode_size(struct btrfs_inode_item *i, u64 val)
-{
-	i->size = cpu_to_le64(val);
-}
-
-static inline u64 btrfs_inode_nblocks(struct btrfs_inode_item *i)
-{
-	return le64_to_cpu(i->nblocks);
-}
-
-static inline void btrfs_set_inode_nblocks(struct btrfs_inode_item *i, u64 val)
-{
-	i->nblocks = cpu_to_le64(val);
-}
-
-static inline u64 btrfs_inode_block_group(struct btrfs_inode_item *i)
-{
-	return le64_to_cpu(i->block_group);
-}
-
-static inline void btrfs_set_inode_block_group(struct btrfs_inode_item *i,
-						u64 val)
-{
-	i->block_group = cpu_to_le64(val);
-}
-
-static inline u32 btrfs_inode_nlink(struct btrfs_inode_item *i)
-{
-	return le32_to_cpu(i->nlink);
-}
-
-static inline void btrfs_set_inode_nlink(struct btrfs_inode_item *i, u32 val)
-{
-	i->nlink = cpu_to_le32(val);
-}
-
-static inline u32 btrfs_inode_uid(struct btrfs_inode_item *i)
-{
-	return le32_to_cpu(i->uid);
-}
-
-static inline void btrfs_set_inode_uid(struct btrfs_inode_item *i, u32 val)
-{
-	i->uid = cpu_to_le32(val);
-}
-
-static inline u32 btrfs_inode_gid(struct btrfs_inode_item *i)
-{
-	return le32_to_cpu(i->gid);
-}
-
-static inline void btrfs_set_inode_gid(struct btrfs_inode_item *i, u32 val)
-{
-	i->gid = cpu_to_le32(val);
-}
-
-static inline u32 btrfs_inode_mode(struct btrfs_inode_item *i)
-{
-	return le32_to_cpu(i->mode);
-}
-
-static inline void btrfs_set_inode_mode(struct btrfs_inode_item *i, u32 val)
-{
-	i->mode = cpu_to_le32(val);
+#define BTRFS_SETGET_STACK_FUNCS(name, type, member, bits)		\
+static inline u##bits btrfs_##name(type *s)				\
+{									\
+	return le##bits##_to_cpu(s->member);				\
+}									\
+static inline void btrfs_set_##name(type *s, u##bits val)		\
+{									\
+	s->member = cpu_to_le##bits(val);				\
 }
 
-static inline u32 btrfs_inode_rdev(struct btrfs_inode_item *i)
-{
-	return le32_to_cpu(i->rdev);
-}
+/* struct btrfs_block_group_item */
+BTRFS_SETGET_STACK_FUNCS(block_group_used, struct btrfs_block_group_item,
+			 used, 64);
+BTRFS_SETGET_FUNCS(disk_block_group_used, struct btrfs_block_group_item,
+			 used, 64);
 
-static inline void btrfs_set_inode_rdev(struct btrfs_inode_item *i, u32 val)
-{
-	i->rdev = cpu_to_le32(val);
-}
+/* struct btrfs_inode_item */
+BTRFS_SETGET_FUNCS(inode_generation, struct btrfs_inode_item, generation, 64);
+BTRFS_SETGET_FUNCS(inode_size, struct btrfs_inode_item, size, 64);
+BTRFS_SETGET_FUNCS(inode_nblocks, struct btrfs_inode_item, nblocks, 64);
+BTRFS_SETGET_FUNCS(inode_block_group, struct btrfs_inode_item, block_group, 64);
+BTRFS_SETGET_FUNCS(inode_nlink, struct btrfs_inode_item, nlink, 32);
+BTRFS_SETGET_FUNCS(inode_uid, struct btrfs_inode_item, uid, 32);
+BTRFS_SETGET_FUNCS(inode_gid, struct btrfs_inode_item, gid, 32);
+BTRFS_SETGET_FUNCS(inode_mode, struct btrfs_inode_item, mode, 32);
+BTRFS_SETGET_FUNCS(inode_rdev, struct btrfs_inode_item, rdev, 32);
+BTRFS_SETGET_FUNCS(inode_flags, struct btrfs_inode_item, flags, 16);
+BTRFS_SETGET_FUNCS(inode_compat_flags, struct btrfs_inode_item,
+		   compat_flags, 16);
 
-static inline u16 btrfs_inode_flags(struct btrfs_inode_item *i)
+static inline struct btrfs_inode_timespec *
+btrfs_inode_atime(struct btrfs_inode_item *inode_item)
 {
-	return le16_to_cpu(i->flags);
+	unsigned long ptr = (unsigned long)inode_item;
+	ptr += offsetof(struct btrfs_inode_item, atime);
+	return (struct btrfs_inode_timespec *)ptr;
 }
 
-static inline void btrfs_set_inode_flags(struct btrfs_inode_item *i, u16 val)
+static inline struct btrfs_inode_timespec *
+btrfs_inode_mtime(struct btrfs_inode_item *inode_item)
 {
-	i->flags = cpu_to_le16(val);
+	unsigned long ptr = (unsigned long)inode_item;
+	ptr += offsetof(struct btrfs_inode_item, mtime);
+	return (struct btrfs_inode_timespec *)ptr;
 }
 
-static inline u16 btrfs_inode_compat_flags(struct btrfs_inode_item *i)
+static inline struct btrfs_inode_timespec *
+btrfs_inode_ctime(struct btrfs_inode_item *inode_item)
 {
-	return le16_to_cpu(i->compat_flags);
+	unsigned long ptr = (unsigned long)inode_item;
+	ptr += offsetof(struct btrfs_inode_item, ctime);
+	return (struct btrfs_inode_timespec *)ptr;
 }
 
-static inline void btrfs_set_inode_compat_flags(struct btrfs_inode_item *i,
-						u16 val)
+static inline struct btrfs_inode_timespec *
+btrfs_inode_otime(struct btrfs_inode_item *inode_item)
 {
-	i->compat_flags = cpu_to_le16(val);
+	unsigned long ptr = (unsigned long)inode_item;
+	ptr += offsetof(struct btrfs_inode_item, otime);
+	return (struct btrfs_inode_timespec *)ptr;
 }
 
-static inline u64 btrfs_timespec_sec(struct btrfs_inode_timespec *ts)
-{
-	return le64_to_cpu(ts->sec);
-}
+BTRFS_SETGET_FUNCS(timespec_sec, struct btrfs_inode_timespec, sec, 64);
+BTRFS_SETGET_FUNCS(timespec_nsec, struct btrfs_inode_timespec, nsec, 32);
 
-static inline void btrfs_set_timespec_sec(struct btrfs_inode_timespec *ts,
-					  u64 val)
-{
-	ts->sec = cpu_to_le64(val);
-}
+/* struct btrfs_extent_item */
+BTRFS_SETGET_FUNCS(extent_refs, struct btrfs_extent_item, refs, 32);
+BTRFS_SETGET_FUNCS(extent_owner, struct btrfs_extent_item, owner, 32);
 
-static inline u32 btrfs_timespec_nsec(struct btrfs_inode_timespec *ts)
-{
-	return le32_to_cpu(ts->nsec);
-}
+BTRFS_SETGET_STACK_FUNCS(stack_extent_refs, struct btrfs_extent_item,
+			 refs, 32);
+BTRFS_SETGET_STACK_FUNCS(stack_extent_owner, struct btrfs_extent_item,
+			 owner, 32);
 
-static inline void btrfs_set_timespec_nsec(struct btrfs_inode_timespec *ts,
-					  u32 val)
-{
-	ts->nsec = cpu_to_le32(val);
-}
+/* struct btrfs_node */
+BTRFS_SETGET_FUNCS(key_blockptr, struct btrfs_key_ptr, blockptr, 64);
 
-static inline u32 btrfs_extent_refs(struct btrfs_extent_item *ei)
+static inline u64 btrfs_node_blockptr(struct extent_buffer *eb, int nr)
 {
-	return le32_to_cpu(ei->refs);
+	unsigned long ptr;
+	ptr = offsetof(struct btrfs_node, ptrs) +
+		sizeof(struct btrfs_key_ptr) * nr;
+	return btrfs_key_blockptr(eb, (struct btrfs_key_ptr *)ptr);
 }
 
-static inline void btrfs_set_extent_refs(struct btrfs_extent_item *ei, u32 val)
+static inline void btrfs_set_node_blockptr(struct extent_buffer *eb,
+					   int nr, u64 val)
 {
-	ei->refs = cpu_to_le32(val);
+	unsigned long ptr;
+	ptr = offsetof(struct btrfs_node, ptrs) +
+		sizeof(struct btrfs_key_ptr) * nr;
+	btrfs_set_key_blockptr(eb, (struct btrfs_key_ptr *)ptr, val);
 }
 
-static inline u64 btrfs_extent_owner(struct btrfs_extent_item *ei)
+static unsigned long btrfs_node_key_ptr_offset(int nr)
 {
-	return le64_to_cpu(ei->owner);
+	return offsetof(struct btrfs_node, ptrs) +
+		sizeof(struct btrfs_key_ptr) * nr;
 }
 
-static inline void btrfs_set_extent_owner(struct btrfs_extent_item *ei, u64 val)
+static void btrfs_node_key(struct extent_buffer *eb,
+			   struct btrfs_disk_key *disk_key, int nr)
 {
-	ei->owner = cpu_to_le64(val);
+	unsigned long ptr;
+	ptr = btrfs_node_key_ptr_offset(nr);
+	read_eb_member(eb, (struct btrfs_key_ptr *)ptr,
+		       struct btrfs_key_ptr, key, disk_key);
 }
-
-static inline u64 btrfs_node_blockptr(struct btrfs_node *n, int nr)
+static inline void btrfs_set_node_key(struct extent_buffer *eb,
+				      struct btrfs_disk_key *disk_key, int nr)
 {
-	return le64_to_cpu(n->ptrs[nr].blockptr);
+	unsigned long ptr;
+	ptr = btrfs_node_key_ptr_offset(nr);
+	write_eb_member(eb, (struct btrfs_key_ptr *)ptr,
+		       struct btrfs_key_ptr, key, disk_key);
 }
 
+/* struct btrfs_item */
+BTRFS_SETGET_FUNCS(item_offset, struct btrfs_item, offset, 32);
+BTRFS_SETGET_FUNCS(item_size, struct btrfs_item, size, 32);
 
-static inline void btrfs_set_node_blockptr(struct btrfs_node *n, int nr,
-					   u64 val)
+static inline unsigned long btrfs_item_nr_offset(int nr)
 {
-	n->ptrs[nr].blockptr = cpu_to_le64(val);
+	return offsetof(struct btrfs_leaf, items) +
+		sizeof(struct btrfs_item) * nr;
 }
 
-static inline u32 btrfs_item_offset(struct btrfs_item *item)
+static inline struct btrfs_item *btrfs_item_nr(struct extent_buffer *eb,
+					       int nr)
 {
-	return le32_to_cpu(item->offset);
+	return (struct btrfs_item *)btrfs_item_nr_offset(nr);
 }
 
-static inline void btrfs_set_item_offset(struct btrfs_item *item, u32 val)
+static inline u32 btrfs_item_end(struct extent_buffer *eb,
+				 struct btrfs_item *item)
 {
-	item->offset = cpu_to_le32(val);
+	return btrfs_item_offset(eb, item) + btrfs_item_size(eb, item);
 }
 
-static inline u32 btrfs_item_end(struct btrfs_item *item)
+static inline u32 btrfs_item_end_nr(struct extent_buffer *eb, int nr)
 {
-	return le32_to_cpu(item->offset) + le16_to_cpu(item->size);
+	return btrfs_item_end(eb, btrfs_item_nr(eb, nr));
 }
 
-static inline u16 btrfs_item_size(struct btrfs_item *item)
+static inline u32 btrfs_item_offset_nr(struct extent_buffer *eb, int nr)
 {
-	return le16_to_cpu(item->size);
+	return btrfs_item_offset(eb, btrfs_item_nr(eb, nr));
 }
 
-static inline void btrfs_set_item_size(struct btrfs_item *item, u16 val)
+static inline u32 btrfs_item_size_nr(struct extent_buffer *eb, int nr)
 {
-	item->size = cpu_to_le16(val);
+	return btrfs_item_size(eb, btrfs_item_nr(eb, nr));
 }
 
-static inline u16 btrfs_dir_flags(struct btrfs_dir_item *d)
+static inline void btrfs_item_key(struct extent_buffer *eb,
+			   struct btrfs_disk_key *disk_key, int nr)
 {
-	return le16_to_cpu(d->flags);
+	struct btrfs_item *item = btrfs_item_nr(eb, nr);
+	read_eb_member(eb, item, struct btrfs_item, key, disk_key);
 }
 
-static inline void btrfs_set_dir_flags(struct btrfs_dir_item *d, u16 val)
+static inline void btrfs_set_item_key(struct extent_buffer *eb,
+			       struct btrfs_disk_key *disk_key, int nr)
 {
-	d->flags = cpu_to_le16(val);
+	struct btrfs_item *item = btrfs_item_nr(eb, nr);
+	write_eb_member(eb, item, struct btrfs_item, key, disk_key);
 }
 
-static inline u8 btrfs_dir_type(struct btrfs_dir_item *d)
-{
-	return d->type;
-}
+/* struct btrfs_dir_item */
+BTRFS_SETGET_FUNCS(dir_flags, struct btrfs_dir_item, flags, 16);
+BTRFS_SETGET_FUNCS(dir_type, struct btrfs_dir_item, type, 8);
+BTRFS_SETGET_FUNCS(dir_name_len, struct btrfs_dir_item, name_len, 16);
 
-static inline void btrfs_set_dir_type(struct btrfs_dir_item *d, u8 val)
+static inline void btrfs_dir_item_key(struct extent_buffer *eb,
+				      struct btrfs_dir_item *item,
+				      struct btrfs_disk_key *key)
 {
-	d->type = val;
+	read_eb_member(eb, item, struct btrfs_dir_item, location, key);
 }
 
-static inline u16 btrfs_dir_name_len(struct btrfs_dir_item *d)
+static inline void btrfs_set_dir_item_key(struct extent_buffer *eb,
+					  struct btrfs_dir_item *item,
+					  struct btrfs_disk_key *key)
 {
-	return le16_to_cpu(d->name_len);
+	write_eb_member(eb, item, struct btrfs_dir_item, location, key);
 }
 
-static inline void btrfs_set_dir_name_len(struct btrfs_dir_item *d, u16 val)
-{
-	d->name_len = cpu_to_le16(val);
-}
+/* struct btrfs_disk_key */
+BTRFS_SETGET_STACK_FUNCS(disk_key_objectid, struct btrfs_disk_key,
+			 objectid, 64);
+BTRFS_SETGET_STACK_FUNCS(disk_key_offset, struct btrfs_disk_key, offset, 64);
+BTRFS_SETGET_STACK_FUNCS(disk_key_type, struct btrfs_disk_key, type, 8);
 
 static inline void btrfs_disk_key_to_cpu(struct btrfs_key *cpu,
 					 struct btrfs_disk_key *disk)
 {
 	cpu->offset = le64_to_cpu(disk->offset);
-	cpu->flags = le32_to_cpu(disk->flags);
+	cpu->type = disk->type;
 	cpu->objectid = le64_to_cpu(disk->objectid);
 }
 
@@ -649,400 +661,167 @@ static inline void btrfs_cpu_key_to_disk(struct btrfs_disk_key *disk,
 					 struct btrfs_key *cpu)
 {
 	disk->offset = cpu_to_le64(cpu->offset);
-	disk->flags = cpu_to_le32(cpu->flags);
+	disk->type = cpu->type;
 	disk->objectid = cpu_to_le64(cpu->objectid);
 }
 
-static inline u64 btrfs_disk_key_objectid(struct btrfs_disk_key *disk)
-{
-	return le64_to_cpu(disk->objectid);
-}
-
-static inline void btrfs_set_disk_key_objectid(struct btrfs_disk_key *disk,
-					       u64 val)
-{
-	disk->objectid = cpu_to_le64(val);
-}
-
-static inline u64 btrfs_disk_key_offset(struct btrfs_disk_key *disk)
-{
-	return le64_to_cpu(disk->offset);
-}
-
-static inline void btrfs_set_disk_key_offset(struct btrfs_disk_key *disk,
-					     u64 val)
-{
-	disk->offset = cpu_to_le64(val);
-}
-
-static inline u32 btrfs_disk_key_flags(struct btrfs_disk_key *disk)
-{
-	return le32_to_cpu(disk->flags);
-}
-
-static inline void btrfs_set_disk_key_flags(struct btrfs_disk_key *disk,
-					    u32 val)
-{
-	disk->flags = cpu_to_le32(val);
-}
-
-static inline u32 btrfs_disk_key_type(struct btrfs_disk_key *key)
-{
-	return le32_to_cpu(key->flags) >> BTRFS_KEY_TYPE_SHIFT;
-}
-
-static inline void btrfs_set_disk_key_type(struct btrfs_disk_key *key,
-					       u32 val)
-{
-	u32 flags = btrfs_disk_key_flags(key);
-	BUG_ON(val >= BTRFS_KEY_TYPE_MAX);
-	val = val << BTRFS_KEY_TYPE_SHIFT;
-	flags = (flags & ~BTRFS_KEY_TYPE_MASK) | val;
-	btrfs_set_disk_key_flags(key, flags);
-}
-
-static inline u32 btrfs_key_type(struct btrfs_key *key)
-{
-	return key->flags >> BTRFS_KEY_TYPE_SHIFT;
-}
-
-static inline void btrfs_set_key_type(struct btrfs_key *key, u32 val)
-{
-	BUG_ON(val >= BTRFS_KEY_TYPE_MAX);
-	val = val << BTRFS_KEY_TYPE_SHIFT;
-	key->flags = (key->flags & ~(BTRFS_KEY_TYPE_MASK)) | val;
-}
-
-static inline u64 btrfs_header_blocknr(struct btrfs_header *h)
-{
-	return le64_to_cpu(h->blocknr);
-}
-
-static inline void btrfs_set_header_blocknr(struct btrfs_header *h, u64 blocknr)
-{
-	h->blocknr = cpu_to_le64(blocknr);
-}
-
-static inline u64 btrfs_header_generation(struct btrfs_header *h)
+static inline void btrfs_node_key_to_cpu(struct extent_buffer *eb,
+				  struct btrfs_key *key, int nr)
 {
-	return le64_to_cpu(h->generation);
+	struct btrfs_disk_key disk_key;
+	btrfs_node_key(eb, &disk_key, nr);
+	btrfs_disk_key_to_cpu(key, &disk_key);
 }
 
-static inline void btrfs_set_header_generation(struct btrfs_header *h,
-					       u64 val)
+static inline void btrfs_item_key_to_cpu(struct extent_buffer *eb,
+				  struct btrfs_key *key, int nr)
 {
-	h->generation = cpu_to_le64(val);
+	struct btrfs_disk_key disk_key;
+	btrfs_item_key(eb, &disk_key, nr);
+	btrfs_disk_key_to_cpu(key, &disk_key);
 }
 
-static inline u64 btrfs_header_owner(struct btrfs_header *h)
+static inline void btrfs_dir_item_key_to_cpu(struct extent_buffer *eb,
+				      struct btrfs_dir_item *item,
+				      struct btrfs_key *key)
 {
-	return le64_to_cpu(h->owner);
+	struct btrfs_disk_key disk_key;
+	btrfs_dir_item_key(eb, item, &disk_key);
+	btrfs_disk_key_to_cpu(key, &disk_key);
 }
 
-static inline void btrfs_set_header_owner(struct btrfs_header *h,
-					       u64 val)
-{
-	h->owner = cpu_to_le64(val);
-}
-
-static inline u16 btrfs_header_nritems(struct btrfs_header *h)
-{
-	return le16_to_cpu(h->nritems);
-}
-
-static inline void btrfs_set_header_nritems(struct btrfs_header *h, u16 val)
-{
-	h->nritems = cpu_to_le16(val);
-}
-
-static inline u16 btrfs_header_flags(struct btrfs_header *h)
-{
-	return le16_to_cpu(h->flags);
-}
-
-static inline void btrfs_set_header_flags(struct btrfs_header *h, u16 val)
-{
-	h->flags = cpu_to_le16(val);
-}
-
-static inline int btrfs_header_level(struct btrfs_header *h)
-{
-	return h->level;
-}
-
-static inline void btrfs_set_header_level(struct btrfs_header *h, int level)
-{
-	BUG_ON(level > BTRFS_MAX_LEVEL);
-	h->level = level;
-}
-
-static inline int btrfs_is_leaf(struct btrfs_node *n)
-{
-	return (btrfs_header_level(&n->header) == 0);
-}
-
-static inline u64 btrfs_root_blocknr(struct btrfs_root_item *item)
-{
-	return le64_to_cpu(item->blocknr);
-}
-
-static inline void btrfs_set_root_blocknr(struct btrfs_root_item *item, u64 val)
-{
-	item->blocknr = cpu_to_le64(val);
-}
-
-static inline u64 btrfs_root_dirid(struct btrfs_root_item *item)
-{
-	return le64_to_cpu(item->root_dirid);
-}
-
-static inline void btrfs_set_root_dirid(struct btrfs_root_item *item, u64 val)
-{
-	item->root_dirid = cpu_to_le64(val);
-}
-
-static inline u32 btrfs_root_refs(struct btrfs_root_item *item)
-{
-	return le32_to_cpu(item->refs);
-}
-
-static inline void btrfs_set_root_refs(struct btrfs_root_item *item, u32 val)
-{
-	item->refs = cpu_to_le32(val);
-}
-
-static inline u32 btrfs_root_flags(struct btrfs_root_item *item)
-{
-	return le32_to_cpu(item->flags);
-}
-
-static inline void btrfs_set_root_flags(struct btrfs_root_item *item, u32 val)
-{
-	item->flags = cpu_to_le32(val);
-}
-
-static inline void btrfs_set_root_blocks_used(struct btrfs_root_item *item,
-						   u64 val)
-{
-	item->blocks_used = cpu_to_le64(val);
-}
-
-static inline u64 btrfs_root_blocks_used(struct btrfs_root_item *item)
-{
-	return le64_to_cpu(item->blocks_used);
-}
-
-static inline void btrfs_set_root_block_limit(struct btrfs_root_item *item,
-						u64 val)
-{
-	item->block_limit = cpu_to_le64(val);
-}
-
-static inline u64 btrfs_root_block_limit(struct btrfs_root_item *item)
-{
-	return le64_to_cpu(item->block_limit);
-}
 
-static inline u64 btrfs_super_blocknr(struct btrfs_super_block *s)
+static inline u8 btrfs_key_type(struct btrfs_key *key)
 {
-	return le64_to_cpu(s->blocknr);
+	return key->type;
 }
 
-static inline void btrfs_set_super_blocknr(struct btrfs_super_block *s, u64 val)
+static inline void btrfs_set_key_type(struct btrfs_key *key, u8 val)
 {
-	s->blocknr = cpu_to_le64(val);
+	key->type = val;
 }
 
-static inline u64 btrfs_super_generation(struct btrfs_super_block *s)
-{
-	return le64_to_cpu(s->generation);
-}
+/* struct btrfs_header */
+BTRFS_SETGET_HEADER_FUNCS(header_blocknr, struct btrfs_header, blocknr, 64);
+BTRFS_SETGET_HEADER_FUNCS(header_generation, struct btrfs_header,
+			  generation, 64);
+BTRFS_SETGET_HEADER_FUNCS(header_owner, struct btrfs_header, owner, 64);
+BTRFS_SETGET_HEADER_FUNCS(header_nritems, struct btrfs_header, nritems, 32);
+BTRFS_SETGET_HEADER_FUNCS(header_flags, struct btrfs_header, flags, 16);
+BTRFS_SETGET_HEADER_FUNCS(header_level, struct btrfs_header, level, 8);
 
-static inline void btrfs_set_super_generation(struct btrfs_super_block *s,
-					      u64 val)
+static inline u8 *btrfs_header_fsid(struct extent_buffer *eb)
 {
-	s->generation = cpu_to_le64(val);
+	unsigned long ptr = offsetof(struct btrfs_header, fsid);
+	return (u8 *)ptr;
 }
 
-static inline u64 btrfs_super_root(struct btrfs_super_block *s)
+static inline u8 *btrfs_super_fsid(struct extent_buffer *eb)
 {
-	return le64_to_cpu(s->root);
+	unsigned long ptr = offsetof(struct btrfs_super_block, fsid);
+	return (u8 *)ptr;
 }
 
-static inline void btrfs_set_super_root(struct btrfs_super_block *s, u64 val)
+static inline u8 *btrfs_header_csum(struct extent_buffer *eb)
 {
-	s->root = cpu_to_le64(val);
+	unsigned long ptr = offsetof(struct btrfs_header, csum);
+	return (u8 *)ptr;
 }
 
-static inline u64 btrfs_super_total_blocks(struct btrfs_super_block *s)
+static inline struct btrfs_node *btrfs_buffer_node(struct extent_buffer *eb)
 {
-	return le64_to_cpu(s->total_blocks);
+	return NULL;
 }
 
-static inline void btrfs_set_super_total_blocks(struct btrfs_super_block *s,
-						u64 val)
+static inline struct btrfs_leaf *btrfs_buffer_leaf(struct extent_buffer *eb)
 {
-	s->total_blocks = cpu_to_le64(val);
+	return NULL;
 }
 
-static inline u64 btrfs_super_blocks_used(struct btrfs_super_block *s)
+static inline struct btrfs_header *btrfs_buffer_header(struct extent_buffer *eb)
 {
-	return le64_to_cpu(s->blocks_used);
+	return NULL;
 }
 
-static inline void btrfs_set_super_blocks_used(struct btrfs_super_block *s,
-						u64 val)
+static inline int btrfs_is_leaf(struct extent_buffer *eb)
 {
-	s->blocks_used = cpu_to_le64(val);
+	return (btrfs_header_level(eb) == 0);
 }
 
-static inline u32 btrfs_super_blocksize(struct btrfs_super_block *s)
-{
-	return le32_to_cpu(s->blocksize);
-}
+/* struct btrfs_root_item */
+BTRFS_SETGET_FUNCS(disk_root_refs, struct btrfs_root_item, refs, 32);
+BTRFS_SETGET_FUNCS(disk_root_blocknr, struct btrfs_root_item, blocknr, 64);
 
-static inline void btrfs_set_super_blocksize(struct btrfs_super_block *s,
-						u32 val)
-{
-	s->blocksize = cpu_to_le32(val);
-}
+BTRFS_SETGET_STACK_FUNCS(root_blocknr, struct btrfs_root_item, blocknr, 64);
+BTRFS_SETGET_STACK_FUNCS(root_dirid, struct btrfs_root_item, root_dirid, 64);
+BTRFS_SETGET_STACK_FUNCS(root_refs, struct btrfs_root_item, refs, 32);
+BTRFS_SETGET_STACK_FUNCS(root_flags, struct btrfs_root_item, flags, 32);
+BTRFS_SETGET_STACK_FUNCS(root_used, struct btrfs_root_item, blocks_used, 64);
+BTRFS_SETGET_STACK_FUNCS(root_limit, struct btrfs_root_item, block_limit, 64);
 
-static inline u64 btrfs_super_root_dir(struct btrfs_super_block *s)
-{
-	return le64_to_cpu(s->root_dir_objectid);
-}
+/* struct btrfs_super_block */
+BTRFS_SETGET_STACK_FUNCS(super_blocknr, struct btrfs_super_block, blocknr, 64);
+BTRFS_SETGET_STACK_FUNCS(super_generation, struct btrfs_super_block,
+			 generation, 64);
+BTRFS_SETGET_STACK_FUNCS(super_root, struct btrfs_super_block, root, 64);
+BTRFS_SETGET_STACK_FUNCS(super_total_blocks, struct btrfs_super_block,
+		   total_blocks, 64);
+BTRFS_SETGET_STACK_FUNCS(super_blocks_used, struct btrfs_super_block,
+		   blocks_used, 64);
+BTRFS_SETGET_STACK_FUNCS(super_sectorsize, struct btrfs_super_block,
+			 sectorsize, 32);
+BTRFS_SETGET_STACK_FUNCS(super_nodesize, struct btrfs_super_block,
+			 nodesize, 32);
+BTRFS_SETGET_STACK_FUNCS(super_leafsize, struct btrfs_super_block,
+			 leafsize, 32);
+BTRFS_SETGET_STACK_FUNCS(super_root_dir, struct btrfs_super_block,
+			 root_dir_objectid, 64);
 
-static inline void btrfs_set_super_root_dir(struct btrfs_super_block *s, u64
-					    val)
+static inline unsigned long btrfs_leaf_data(struct extent_buffer *l)
 {
-	s->root_dir_objectid = cpu_to_le64(val);
+	return offsetof(struct btrfs_leaf, items);
 }
 
-static inline u8 *btrfs_leaf_data(struct btrfs_leaf *l)
-{
-	return (u8 *)l->items;
-}
+/* struct btrfs_file_extent_item */
+BTRFS_SETGET_FUNCS(file_extent_type, struct btrfs_file_extent_item, type, 8);
 
-static inline int btrfs_file_extent_type(struct btrfs_file_extent_item *e)
-{
-	return e->type;
-}
-static inline void btrfs_set_file_extent_type(struct btrfs_file_extent_item *e,
-					      u8 val)
-{
-	e->type = val;
-}
-
-static inline char *btrfs_file_extent_inline_start(struct
+static inline unsigned long btrfs_file_extent_inline_start(struct
 						   btrfs_file_extent_item *e)
 {
-	return (char *)(&e->disk_blocknr);
+	unsigned long offset = (unsigned long)e;
+	offset += offsetof(struct btrfs_file_extent_item, disk_blocknr);
+	return offset;
 }
 
 static inline u32 btrfs_file_extent_calc_inline_size(u32 datasize)
 {
-	return (unsigned long)(&((struct
-		  btrfs_file_extent_item *)NULL)->disk_blocknr) + datasize;
-}
-
-static inline u32 btrfs_file_extent_inline_len(struct btrfs_item *e)
-{
-	struct btrfs_file_extent_item *fe = NULL;
-	return btrfs_item_size(e) - (unsigned long)(&fe->disk_blocknr);
-}
-
-static inline u64 btrfs_file_extent_disk_blocknr(struct btrfs_file_extent_item
-						 *e)
-{
-	return le64_to_cpu(e->disk_blocknr);
+	return offsetof(struct btrfs_file_extent_item, disk_blocknr) + datasize;
 }
 
-static inline void btrfs_set_file_extent_disk_blocknr(struct
-						      btrfs_file_extent_item
-						      *e, u64 val)
+static inline u32 btrfs_file_extent_inline_len(struct extent_buffer *eb,
+					       struct btrfs_item *e)
 {
-	e->disk_blocknr = cpu_to_le64(val);
+	unsigned long offset;
+	offset = offsetof(struct btrfs_file_extent_item, disk_blocknr);
+	return btrfs_item_size(eb, e) - offset;
 }
 
-static inline u64 btrfs_file_extent_generation(struct btrfs_file_extent_item *e)
-{
-	return le64_to_cpu(e->generation);
-}
-
-static inline void btrfs_set_file_extent_generation(struct
-						    btrfs_file_extent_item *e,
-						    u64 val)
-{
-	e->generation = cpu_to_le64(val);
-}
-
-static inline u64 btrfs_file_extent_disk_num_blocks(struct
-						    btrfs_file_extent_item *e)
-{
-	return le64_to_cpu(e->disk_num_blocks);
-}
-
-static inline void btrfs_set_file_extent_disk_num_blocks(struct
-							 btrfs_file_extent_item
-							 *e, u64 val)
-{
-	e->disk_num_blocks = cpu_to_le64(val);
-}
-
-static inline u64 btrfs_file_extent_offset(struct btrfs_file_extent_item *e)
-{
-	return le64_to_cpu(e->offset);
-}
-
-static inline void btrfs_set_file_extent_offset(struct btrfs_file_extent_item
-						*e, u64 val)
-{
-	e->offset = cpu_to_le64(val);
-}
-
-static inline u64 btrfs_file_extent_num_blocks(struct btrfs_file_extent_item
-					       *e)
-{
-	return le64_to_cpu(e->num_blocks);
-}
-
-static inline void btrfs_set_file_extent_num_blocks(struct
-						    btrfs_file_extent_item *e,
-						    u64 val)
-{
-	e->num_blocks = cpu_to_le64(val);
-}
+BTRFS_SETGET_FUNCS(file_extent_disk_blocknr, struct btrfs_file_extent_item,
+		   disk_blocknr, 64);
+BTRFS_SETGET_FUNCS(file_extent_generation, struct btrfs_file_extent_item,
+		   generation, 64);
+BTRFS_SETGET_FUNCS(file_extent_disk_num_blocks, struct btrfs_file_extent_item,
+		   disk_num_blocks, 64);
+BTRFS_SETGET_FUNCS(file_extent_offset, struct btrfs_file_extent_item,
+		  offset, 64);
+BTRFS_SETGET_FUNCS(file_extent_num_blocks, struct btrfs_file_extent_item,
+		   num_blocks, 64);
 
 static inline struct btrfs_root *btrfs_sb(struct super_block *sb)
 {
 	return sb->s_fs_info;
 }
 
-static inline void btrfs_check_bounds(void *vptr, size_t len,
-				     void *vcontainer, size_t container_len)
-{
-	char *ptr = vptr;
-	char *container = vcontainer;
-	WARN_ON(ptr < container);
-	WARN_ON(ptr + len > container + container_len);
-}
-
-static inline void btrfs_memcpy(struct btrfs_root *root,
-				void *dst_block,
-				void *dst, const void *src, size_t nr)
-{
-	btrfs_check_bounds(dst, nr, dst_block, root->fs_info->sb->s_blocksize);
-	memcpy(dst, src, nr);
-}
-
-static inline void btrfs_memmove(struct btrfs_root *root,
-				void *dst_block,
-				void *dst, void *src, size_t nr)
-{
-	btrfs_check_bounds(dst, nr, dst_block, root->fs_info->sb->s_blocksize);
-	memmove(dst, src, nr);
-}
-
 static inline int btrfs_set_root_name(struct btrfs_root *root,
 				      const char *name, int len)
 {
@@ -1063,7 +842,11 @@ static inline int btrfs_set_root_name(struct btrfs_root *root,
 /* helper function to cast into the data area of the leaf. */
 #define btrfs_item_ptr(leaf, slot, type) \
 	((type *)(btrfs_leaf_data(leaf) + \
-	btrfs_item_offset((leaf)->items + (slot))))
+	btrfs_item_offset_nr(leaf, slot)))
+
+#define btrfs_item_ptr_offset(leaf, slot) \
+	((unsigned long)(btrfs_leaf_data(leaf) + \
+	btrfs_item_offset_nr(leaf, slot)))
 
 /* mount option defines and helpers */
 #define BTRFS_MOUNT_SUBVOL		0x000001
@@ -1084,7 +867,7 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
 						 int data, int owner);
 int btrfs_inc_root_ref(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root);
-struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
+struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 					    struct btrfs_root *root, u64 hint,
 					    u64 empty_size);
 int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
@@ -1092,7 +875,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 		       u64 num_blocks, u64 empty_size, u64 search_start,
 		       u64 search_end, struct btrfs_key *ins, int data);
 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		  struct buffer_head *buf);
+		  struct extent_buffer *buf);
 int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, u64 blocknr, u64 num_blocks, int pin);
 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
@@ -1106,10 +889,10 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
 int btrfs_free_block_groups(struct btrfs_fs_info *info);
 int btrfs_read_block_groups(struct btrfs_root *root);
 /* ctree.c */
-int btrfs_cow_block(struct btrfs_trans_handle *trans, struct btrfs_root
-			   *root, struct buffer_head *buf, struct buffer_head
-			   *parent, int parent_slot, struct buffer_head
-			   **cow_ret);
+int btrfs_cow_block(struct btrfs_trans_handle *trans,
+		    struct btrfs_root *root, struct extent_buffer *buf,
+		    struct extent_buffer *parent, int parent_slot,
+		    struct extent_buffer **cow_ret);
 int btrfs_extend_item(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_path *path, u32 data_size);
 int btrfs_truncate_item(struct btrfs_trans_handle *trans,
@@ -1120,7 +903,7 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_key *key, struct btrfs_path *p, int
 		      ins_len, int cow);
 int btrfs_realloc_node(struct btrfs_trans_handle *trans,
-		       struct btrfs_root *root, struct buffer_head *parent,
+		       struct btrfs_root *root, struct extent_buffer *parent,
 		       int cache_only, u64 *last_ret);
 void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p);
 struct btrfs_path *btrfs_alloc_path(void);
@@ -1134,7 +917,7 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans, struct btrfs_root
 			    *root, struct btrfs_path *path, struct btrfs_key
 			    *cpu_key, u32 data_size);
 int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path);
-int btrfs_leaf_free_space(struct btrfs_root *root, struct btrfs_leaf *leaf);
+int btrfs_leaf_free_space(struct btrfs_root *root, struct extent_buffer *leaf);
 int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
 			*root);
 /* root-item.c */
@@ -1179,9 +962,9 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans,
 int btrfs_find_highest_inode(struct btrfs_root *fs_root, u64 *objectid);
 
 /* inode-item.c */
-int btrfs_insert_inode(struct btrfs_trans_handle *trans, struct btrfs_root
-		       *root, u64 objectid, struct btrfs_inode_item
-		       *inode_item);
+int btrfs_insert_empty_inode(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root,
+			     struct btrfs_path *path, u64 objectid);
 int btrfs_lookup_inode(struct btrfs_trans_handle *trans, struct btrfs_root
 		       *root, struct btrfs_path *path,
 		       struct btrfs_key *location, int mod);
@@ -1224,8 +1007,6 @@ struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
 				struct btrfs_root *root);
 int btrfs_commit_write(struct file *file, struct page *page,
 		       unsigned from, unsigned to);
-int btrfs_get_block(struct inode *inode, sector_t iblock,
-		    struct buffer_head *result, int create);
 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
 				    size_t page_offset, u64 start, u64 end,
 				    int create);
diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c
index 49db5fa7ced3..6f19de41b878 100644
--- a/fs/btrfs/dir-item.c
+++ b/fs/btrfs/dir-item.c
@@ -33,7 +33,7 @@ static struct btrfs_dir_item *insert_with_overflow(struct btrfs_trans_handle
 	int ret;
 	char *ptr;
 	struct btrfs_item *item;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 
 	ret = btrfs_insert_empty_item(trans, root, path, cpu_key, data_size);
 	if (ret == -EEXIST) {
@@ -49,11 +49,11 @@ static struct btrfs_dir_item *insert_with_overflow(struct btrfs_trans_handle
 	if (ret < 0)
 		return ERR_PTR(ret);
 	WARN_ON(ret > 0);
-	leaf = btrfs_buffer_leaf(path->nodes[0]);
-	item = leaf->items + path->slots[0];
+	leaf = path->nodes[0];
+	item = btrfs_item_nr(leaf, path->slots[0]);
 	ptr = btrfs_item_ptr(leaf, path->slots[0], char);
-	BUG_ON(data_size > btrfs_item_size(item));
-	ptr += btrfs_item_size(item) - data_size;
+	BUG_ON(data_size > btrfs_item_size(leaf, item));
+	ptr += btrfs_item_size(leaf, item) - data_size;
 	return (struct btrfs_dir_item *)ptr;
 }
 
@@ -65,12 +65,13 @@ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
 	int ret2 = 0;
 	struct btrfs_path *path;
 	struct btrfs_dir_item *dir_item;
-	char *name_ptr;
+	struct extent_buffer *leaf;
+	unsigned long name_ptr;
 	struct btrfs_key key;
+	struct btrfs_disk_key disk_key;
 	u32 data_size;
 
 	key.objectid = dir;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY);
 	ret = btrfs_name_hash(name, name_len, &key.offset);
 	BUG_ON(ret);
@@ -85,14 +86,16 @@ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
 		goto out;
 	}
 
-	btrfs_cpu_key_to_disk(&dir_item->location, location);
-	btrfs_set_dir_type(dir_item, type);
-	btrfs_set_dir_flags(dir_item, 0);
-	btrfs_set_dir_name_len(dir_item, name_len);
-	name_ptr = (char *)(dir_item + 1);
+	leaf = path->nodes[0];
+	btrfs_cpu_key_to_disk(&disk_key, location);
+	btrfs_set_dir_item_key(leaf, dir_item, &disk_key);
+	btrfs_set_dir_type(leaf, dir_item, type);
+	btrfs_set_dir_flags(leaf, dir_item, 0);
+	btrfs_set_dir_name_len(leaf, dir_item, name_len);
+	name_ptr = (unsigned long)(dir_item + 1);
 
-	btrfs_memcpy(root, path->nodes[0]->b_data, name_ptr, name, name_len);
-	btrfs_mark_buffer_dirty(path->nodes[0]);
+	write_extent_buffer(leaf, name, name_ptr, name_len);
+	btrfs_mark_buffer_dirty(leaf);
 
 second_insert:
 	/* FIXME, use some real flag for selecting the extra index */
@@ -110,13 +113,15 @@ second_insert:
 		ret2 = PTR_ERR(dir_item);
 		goto out;
 	}
-	btrfs_cpu_key_to_disk(&dir_item->location, location);
-	btrfs_set_dir_type(dir_item, type);
-	btrfs_set_dir_flags(dir_item, 0);
-	btrfs_set_dir_name_len(dir_item, name_len);
-	name_ptr = (char *)(dir_item + 1);
-	btrfs_memcpy(root, path->nodes[0]->b_data, name_ptr, name, name_len);
-	btrfs_mark_buffer_dirty(path->nodes[0]);
+	leaf = path->nodes[0];
+	btrfs_cpu_key_to_disk(&disk_key, location);
+	btrfs_set_dir_item_key(leaf, dir_item, &disk_key);
+	btrfs_set_dir_type(leaf, dir_item, type);
+	btrfs_set_dir_flags(leaf, dir_item, 0);
+	btrfs_set_dir_name_len(leaf, dir_item, name_len);
+	name_ptr = (unsigned long)(dir_item + 1);
+	write_extent_buffer(leaf, name, name_ptr, name_len);
+	btrfs_mark_buffer_dirty(leaf);
 out:
 	btrfs_free_path(path);
 	if (ret)
@@ -136,14 +141,15 @@ struct btrfs_dir_item *btrfs_lookup_dir_item(struct btrfs_trans_handle *trans,
 	struct btrfs_key key;
 	int ins_len = mod < 0 ? -1 : 0;
 	int cow = mod != 0;
-	struct btrfs_disk_key *found_key;
-	struct btrfs_leaf *leaf;
+	struct btrfs_key found_key;
+	struct extent_buffer *leaf;
 
 	key.objectid = dir;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY);
+
 	ret = btrfs_name_hash(name, name_len, &key.offset);
 	BUG_ON(ret);
+
 	ret = btrfs_search_slot(trans, root, &key, path, ins_len, cow);
 	if (ret < 0)
 		return ERR_PTR(ret);
@@ -152,12 +158,13 @@ struct btrfs_dir_item *btrfs_lookup_dir_item(struct btrfs_trans_handle *trans,
 			return NULL;
 		path->slots[0]--;
 	}
-	leaf = btrfs_buffer_leaf(path->nodes[0]);
-	found_key = &leaf->items[path->slots[0]].key;
 
-	if (btrfs_disk_key_objectid(found_key) != dir ||
-	    btrfs_disk_key_type(found_key) != BTRFS_DIR_ITEM_KEY ||
-	    btrfs_disk_key_offset(found_key) != key.offset)
+	leaf = path->nodes[0];
+	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+
+	if (found_key.objectid != dir ||
+	    btrfs_key_type(&found_key) != BTRFS_DIR_ITEM_KEY ||
+	    found_key.offset != key.offset)
 		return NULL;
 
 	return btrfs_match_dir_item_name(root, path, name, name_len);
@@ -176,7 +183,6 @@ btrfs_lookup_dir_index_item(struct btrfs_trans_handle *trans,
 	int cow = mod != 0;
 
 	key.objectid = dir;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_DIR_INDEX_KEY);
 	key.offset = objectid;
 
@@ -193,21 +199,22 @@ struct btrfs_dir_item *btrfs_match_dir_item_name(struct btrfs_root *root,
 			      const char *name, int name_len)
 {
 	struct btrfs_dir_item *dir_item;
-	char *name_ptr;
+	unsigned long name_ptr;
 	u32 total_len;
 	u32 cur = 0;
 	u32 this_len;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 
-	leaf = btrfs_buffer_leaf(path->nodes[0]);
+	leaf = path->nodes[0];
 	dir_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dir_item);
-	total_len = btrfs_item_size(leaf->items + path->slots[0]);
+	total_len = btrfs_item_size_nr(leaf, path->slots[0]);
 	while(cur < total_len) {
-		this_len = sizeof(*dir_item) + btrfs_dir_name_len(dir_item);
-		name_ptr = (char *)(dir_item + 1);
+		this_len = sizeof(*dir_item) +
+			btrfs_dir_name_len(leaf, dir_item);
+		name_ptr = (unsigned long)(dir_item + 1);
 
-		if (btrfs_dir_name_len(dir_item) == name_len &&
-		    memcmp(name_ptr, name, name_len) == 0)
+		if (btrfs_dir_name_len(leaf, dir_item) == name_len &&
+		    memcmp_extent_buffer(leaf, name, name_ptr, name_len) == 0)
 			return dir_item;
 
 		cur += this_len;
@@ -223,20 +230,23 @@ int btrfs_delete_one_dir_name(struct btrfs_trans_handle *trans,
 			      struct btrfs_dir_item *di)
 {
 
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	u32 sub_item_len;
 	u32 item_len;
 	int ret = 0;
 
-	leaf = btrfs_buffer_leaf(path->nodes[0]);
-	sub_item_len = sizeof(*di) + btrfs_dir_name_len(di);
-	item_len = btrfs_item_size(leaf->items + path->slots[0]);
-	if (sub_item_len == btrfs_item_size(leaf->items + path->slots[0])) {
+	leaf = path->nodes[0];
+	sub_item_len = sizeof(*di) + btrfs_dir_name_len(leaf, di);
+	item_len = btrfs_item_size_nr(leaf, path->slots[0]);
+	if (sub_item_len == item_len) {
 		ret = btrfs_del_item(trans, root, path);
 	} else {
-		char *ptr = (char *)di;
-		char *start = btrfs_item_ptr(leaf, path->slots[0], char);
-		btrfs_memmove(root, leaf, ptr, ptr + sub_item_len,
+		/* MARKER */
+		unsigned long ptr = (unsigned long)di;
+		unsigned long start;
+
+		start = btrfs_item_ptr_offset(leaf, path->slots[0]);
+		memmove_extent_buffer(leaf, ptr, ptr + sub_item_len,
 			item_len - (ptr + sub_item_len - start));
 		ret = btrfs_truncate_item(trans, root, path,
 					  item_len - sub_item_len);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index ef6d76f2ec20..0c1f90cbedb0 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -23,138 +23,132 @@
 #include <linux/swap.h>
 #include <linux/radix-tree.h>
 #include <linux/writeback.h>
+#include <linux/buffer_head.h> // for block_sync_page
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
 #include "btrfs_inode.h"
 
-u64 bh_blocknr(struct buffer_head *bh)
+#if 0
+static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
 {
-	return bh->b_blocknr;
-}
-
-static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
-{
-	struct btrfs_node *node = btrfs_buffer_node(buf);
-	if (bh_blocknr(buf) != btrfs_header_blocknr(&node->header)) {
-		printk(KERN_CRIT "bh_blocknr(buf) is %llu, header is %llu\n",
-		       (unsigned long long)bh_blocknr(buf),
-		       (unsigned long long)btrfs_header_blocknr(&node->header));
+	if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) {
+		printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n",
+		       (unsigned long long)extent_buffer_blocknr(buf),
+		       (unsigned long long)btrfs_header_blocknr(buf));
 		return 1;
 	}
 	return 0;
 }
+#endif
 
-struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr)
+struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
+					    u64 blocknr)
 {
-	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
-	int blockbits = root->fs_info->sb->s_blocksize_bits;
-	unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
-	struct page *page;
-	struct buffer_head *bh;
-	struct buffer_head *head;
-	struct buffer_head *ret = NULL;
-
-
-	page = find_lock_page(mapping, index);
-	if (!page)
-		return NULL;
-
-	if (!page_has_buffers(page))
-		goto out_unlock;
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	return find_extent_buffer(&BTRFS_I(btree_inode)->extent_tree,
+				   blocknr * root->sectorsize,
+				   root->sectorsize, GFP_NOFS);
+}
 
-	head = page_buffers(page);
-	bh = head;
-	do {
-		if (buffer_mapped(bh) && bh_blocknr(bh) == blocknr) {
-			ret = bh;
-			get_bh(bh);
-			goto out_unlock;
-		}
-		bh = bh->b_this_page;
-	} while (bh != head);
-out_unlock:
-	unlock_page(page);
-	page_cache_release(page);
-	return ret;
+struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
+						 u64 blocknr)
+{
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	return alloc_extent_buffer(&BTRFS_I(btree_inode)->extent_tree,
+				   blocknr * root->sectorsize,
+				   root->sectorsize, GFP_NOFS);
 }
 
-int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
-			     u64 logical)
+struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
+				    size_t page_offset, u64 start, u64 end,
+				    int create)
 {
-	if (logical == 0) {
-		bh->b_bdev = NULL;
-		bh->b_blocknr = 0;
-		set_buffer_mapped(bh);
-	} else {
-		map_bh(bh, root->fs_info->sb, logical);
+	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+	struct extent_map *em;
+	int ret;
+
+again:
+	em = lookup_extent_mapping(em_tree, start, end);
+	if (em) {
+		goto out;
 	}
-	return 0;
+	em = alloc_extent_map(GFP_NOFS);
+	if (!em) {
+		em = ERR_PTR(-ENOMEM);
+		goto out;
+	}
+	em->start = 0;
+	em->end = (i_size_read(inode) & ~((u64)PAGE_CACHE_SIZE -1)) - 1;
+	em->block_start = 0;
+	em->block_end = em->end;
+	em->bdev = inode->i_sb->s_bdev;
+	ret = add_extent_mapping(em_tree, em);
+	if (ret == -EEXIST) {
+		free_extent_map(em);
+		em = NULL;
+		goto again;
+	} else if (ret) {
+		em = ERR_PTR(ret);
+	}
+out:
+	return em;
 }
 
-struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
-						 u64 blocknr)
+static int btree_writepage(struct page *page, struct writeback_control *wbc)
 {
-	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
-	int blockbits = root->fs_info->sb->s_blocksize_bits;
-	unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
-	struct page *page;
-	struct buffer_head *bh;
-	struct buffer_head *head;
-	struct buffer_head *ret = NULL;
-	int err;
-	u64 first_block = index << (PAGE_CACHE_SHIFT - blockbits);
+	struct extent_map_tree *tree;
+	tree = &BTRFS_I(page->mapping->host)->extent_tree;
+	return extent_write_full_page(tree, page, btree_get_extent, wbc);
+}
+int btree_readpage(struct file *file, struct page *page)
+{
+	struct extent_map_tree *tree;
+	tree = &BTRFS_I(page->mapping->host)->extent_tree;
+	return extent_read_full_page(tree, page, btree_get_extent);
+}
 
-	page = find_or_create_page(mapping, index, GFP_NOFS);
-	if (!page)
-		return NULL;
+static int btree_releasepage(struct page *page, gfp_t unused_gfp_flags)
+{
+	struct extent_map_tree *tree;
+	int ret;
 
-	if (!page_has_buffers(page))
-		create_empty_buffers(page, root->fs_info->sb->s_blocksize, 0);
-	head = page_buffers(page);
-	bh = head;
-	do {
-		if (!buffer_mapped(bh)) {
-			err = btrfs_map_bh_to_logical(root, bh, first_block);
-			BUG_ON(err);
-		}
-		if (bh_blocknr(bh) == blocknr) {
-			ret = bh;
-			get_bh(bh);
-			goto out_unlock;
-		}
-		bh = bh->b_this_page;
-		first_block++;
-	} while (bh != head);
-out_unlock:
-	unlock_page(page);
-	if (ret)
-		touch_buffer(ret);
-	page_cache_release(page);
+	BUG_ON(page->private != 1);
+	tree = &BTRFS_I(page->mapping->host)->extent_tree;
+	ret = try_release_extent_mapping(tree, page);
+	if (ret == 1) {
+		ClearPagePrivate(page);
+		set_page_private(page, 0);
+		page_cache_release(page);
+	}
 	return ret;
 }
 
-static int btree_get_block(struct inode *inode, sector_t iblock,
-			   struct buffer_head *bh, int create)
+static void btree_invalidatepage(struct page *page, unsigned long offset)
 {
-	int err;
-	struct btrfs_root *root = BTRFS_I(bh->b_page->mapping->host)->root;
-	err = btrfs_map_bh_to_logical(root, bh, iblock);
-	return err;
+	struct extent_map_tree *tree;
+	tree = &BTRFS_I(page->mapping->host)->extent_tree;
+	extent_invalidatepage(tree, page, offset);
+	btree_releasepage(page, GFP_NOFS);
 }
 
 int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
 		    char *result)
 {
+	return 0;
+#if 0
 	u32 crc;
 	crc = crc32c(0, data, len);
 	memcpy(result, &crc, BTRFS_CRC32_SIZE);
 	return 0;
+#endif
 }
 
-static int csum_tree_block(struct btrfs_root *root, struct buffer_head *bh,
+#if 0
+static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
 			   int verify)
 {
+	return 0;
 	char result[BTRFS_CRC32_SIZE];
 	int ret;
 	struct btrfs_node *node;
@@ -176,7 +170,9 @@ static int csum_tree_block(struct btrfs_root *root, struct buffer_head *bh,
 	}
 	return 0;
 }
+#endif
 
+#if 0
 static int btree_writepage(struct page *page, struct writeback_control *wbc)
 {
 	struct buffer_head *bh;
@@ -195,87 +191,65 @@ static int btree_writepage(struct page *page, struct writeback_control *wbc)
 	} while (bh != head);
 	return block_write_full_page(page, btree_get_block, wbc);
 }
-
-static int btree_readpage(struct file * file, struct page * page)
-{
-	return block_read_full_page(page, btree_get_block);
-}
+#endif
 
 static struct address_space_operations btree_aops = {
 	.readpage	= btree_readpage,
 	.writepage	= btree_writepage,
+	.releasepage	= btree_releasepage,
+	.invalidatepage = btree_invalidatepage,
 	.sync_page	= block_sync_page,
 };
 
 int readahead_tree_block(struct btrfs_root *root, u64 blocknr)
 {
-	struct buffer_head *bh = NULL;
+	struct extent_buffer *buf = NULL;
+	struct inode *btree_inode = root->fs_info->btree_inode;
 	int ret = 0;
 
-	bh = btrfs_find_create_tree_block(root, blocknr);
-	if (!bh)
+	buf = btrfs_find_create_tree_block(root, blocknr);
+	if (!buf)
 		return 0;
-	if (buffer_uptodate(bh)) {
-		ret = 1;
-		goto done;
-	}
-	if (test_set_buffer_locked(bh)) {
-		ret = 1;
-		goto done;
-	}
-	if (!buffer_uptodate(bh)) {
-		get_bh(bh);
-		bh->b_end_io = end_buffer_read_sync;
-		submit_bh(READ, bh);
-	} else {
-		unlock_buffer(bh);
-		ret = 1;
-	}
-done:
-	brelse(bh);
+	read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
+				 buf, 0);
+	free_extent_buffer(buf);
 	return ret;
 }
 
-struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr)
+struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr)
 {
-	struct buffer_head *bh = NULL;
-
-	bh = btrfs_find_create_tree_block(root, blocknr);
-	if (!bh)
-		return bh;
-	if (buffer_uptodate(bh))
-		goto uptodate;
-	lock_buffer(bh);
-	if (!buffer_uptodate(bh)) {
-		get_bh(bh);
-		bh->b_end_io = end_buffer_read_sync;
-		submit_bh(READ, bh);
-		wait_on_buffer(bh);
-		if (!buffer_uptodate(bh))
-			goto fail;
-	} else {
-		unlock_buffer(bh);
-	}
-uptodate:
-	if (!buffer_checked(bh)) {
-		csum_tree_block(root, bh, 1);
-		set_buffer_checked(bh);
-	}
-	if (check_tree_block(root, bh))
-		goto fail;
-	return bh;
-fail:
-	brelse(bh);
-	return NULL;
+	struct extent_buffer *buf = NULL;
+	struct inode *btree_inode = root->fs_info->btree_inode;
+
+	buf = btrfs_find_create_tree_block(root, blocknr);
+	if (!buf)
+		return NULL;
+	read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
+				 buf, 1);
+	return buf;
 }
 
 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		     struct buffer_head *buf)
+		     struct extent_buffer *buf)
 {
-	WARN_ON(atomic_read(&buf->b_count) == 0);
-	lock_buffer(buf);
-	clear_buffer_dirty(buf);
-	unlock_buffer(buf);
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf);
+	return 0;
+}
+
+int wait_on_tree_block_writeback(struct btrfs_root *root,
+				 struct extent_buffer *buf)
+{
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->extent_tree,
+					buf);
+	return 0;
+}
+
+int set_tree_block_dirty(struct btrfs_root *root, struct extent_buffer *buf)
+{
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf);
 	return 0;
 }
 
@@ -287,7 +261,9 @@ static int __setup_root(int blocksize,
 	root->node = NULL;
 	root->inode = NULL;
 	root->commit_root = NULL;
-	root->blocksize = blocksize;
+	root->sectorsize = blocksize;
+	root->nodesize = blocksize;
+	root->leafsize = blocksize;
 	root->ref_cows = 0;
 	root->fs_info = fs_info;
 	root->objectid = objectid;
@@ -332,7 +308,7 @@ struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
 	struct btrfs_root *root;
 	struct btrfs_root *tree_root = fs_info->tree_root;
 	struct btrfs_path *path;
-	struct btrfs_leaf *l;
+	struct extent_buffer *l;
 	u64 highest_inode;
 	int ret = 0;
 
@@ -361,11 +337,10 @@ struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
 			ret = -ENOENT;
 		goto out;
 	}
-	l = btrfs_buffer_leaf(path->nodes[0]);
-	memcpy(&root->root_item,
-	       btrfs_item_ptr(l, path->slots[0], struct btrfs_root_item),
+	l = path->nodes[0];
+	read_extent_buffer(l, &root->root_item,
+	       btrfs_item_ptr_offset(l, path->slots[0]),
 	       sizeof(root->root_item));
-	memcpy(&root->root_key, location, sizeof(*location));
 	ret = 0;
 out:
 	btrfs_release_path(root, path);
@@ -406,21 +381,21 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 				(unsigned long)root->root_key.objectid,
 				root);
 	if (ret) {
-		brelse(root->node);
+		free_extent_buffer(root->node);
 		kfree(root);
 		return ERR_PTR(ret);
 	}
 
 	ret = btrfs_set_root_name(root, name, namelen);
 	if (ret) {
-		brelse(root->node);
+		free_extent_buffer(root->node);
 		kfree(root);
 		return ERR_PTR(ret);
 	}
 
 	ret = btrfs_sysfs_add_root(root);
 	if (ret) {
-		brelse(root->node);
+		free_extent_buffer(root->node);
 		kfree(root->name);
 		kfree(root);
 		return ERR_PTR(ret);
@@ -471,6 +446,9 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	fs_info->btree_inode->i_nlink = 1;
 	fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
 	fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
+	extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
+			     fs_info->btree_inode->i_mapping,
+			     GFP_NOFS);
 	fs_info->do_barriers = 1;
 	fs_info->closing = 0;
 
@@ -493,10 +471,14 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 
 	if (!fs_info->sb_buffer)
 		goto fail_iput;
-	disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
-	fs_info->disk_super = disk_super;
-	memcpy(&fs_info->super_copy, disk_super, sizeof(fs_info->super_copy));
 
+	read_extent_buffer(fs_info->sb_buffer, &fs_info->super_copy, 0,
+			   sizeof(fs_info->super_copy));
+
+	read_extent_buffer(fs_info->sb_buffer, fs_info->fsid,
+			   (unsigned long)btrfs_super_fsid(fs_info->sb_buffer),
+			   BTRFS_FSID_SIZE);
+	disk_super = &fs_info->super_copy;
 	if (!btrfs_super_root(disk_super))
 		goto fail_sb_buffer;
 
@@ -530,9 +512,9 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	return tree_root;
 
 fail_tree_root:
-	btrfs_block_release(tree_root, tree_root->node);
+	free_extent_buffer(tree_root->node);
 fail_sb_buffer:
-	btrfs_block_release(tree_root, fs_info->sb_buffer);
+	free_extent_buffer(fs_info->sb_buffer);
 fail_iput:
 	iput(fs_info->btree_inode);
 fail:
@@ -546,31 +528,13 @@ int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root)
 {
 	int ret;
-	struct buffer_head *bh = root->fs_info->sb_buffer;
-
-	lock_buffer(bh);
-	WARN_ON(atomic_read(&bh->b_count) < 1);
-	clear_buffer_dirty(bh);
-	csum_tree_block(root, bh, 0);
-	bh->b_end_io = end_buffer_write_sync;
-	get_bh(bh);
-	if (root->fs_info->do_barriers)
-		ret = submit_bh(WRITE_BARRIER, bh);
-	else
-		ret = submit_bh(WRITE, bh);
-	if (ret == -EOPNOTSUPP) {
-		get_bh(bh);
-		lock_buffer(bh);
-		set_buffer_uptodate(bh);
-		root->fs_info->do_barriers = 0;
-		ret = submit_bh(WRITE, bh);
-	}
-	wait_on_buffer(bh);
-	if (!buffer_uptodate(bh)) {
-		WARN_ON(1);
-		return -EIO;
-	}
-	return 0;
+	struct extent_buffer *super = root->fs_info->sb_buffer;
+	struct inode *btree_inode = root->fs_info->btree_inode;
+
+	set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, super);
+	ret = sync_page_range_nolock(btree_inode, btree_inode->i_mapping,
+				     super->start, super->len);
+	return ret;
 }
 
 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
@@ -581,9 +545,9 @@ int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
 	if (root->inode)
 		iput(root->inode);
 	if (root->node)
-		brelse(root->node);
+		free_extent_buffer(root->node);
 	if (root->commit_root)
-		brelse(root->commit_root);
+		free_extent_buffer(root->commit_root);
 	if (root->name)
 		kfree(root->name);
 	kfree(root);
@@ -629,12 +593,10 @@ int close_ctree(struct btrfs_root *root)
 	mutex_unlock(&fs_info->fs_mutex);
 
 	if (fs_info->extent_root->node)
-		btrfs_block_release(fs_info->extent_root,
-				    fs_info->extent_root->node);
+		free_extent_buffer(fs_info->extent_root->node);
 	if (fs_info->tree_root->node)
-		btrfs_block_release(fs_info->tree_root,
-				    fs_info->tree_root->node);
-	btrfs_block_release(root, fs_info->sb_buffer);
+		free_extent_buffer(fs_info->tree_root->node);
+	free_extent_buffer(fs_info->sb_buffer);
 	truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
 	iput(fs_info->btree_inode);
 
@@ -645,25 +607,32 @@ int close_ctree(struct btrfs_root *root)
 	return 0;
 }
 
-void btrfs_mark_buffer_dirty(struct buffer_head *bh)
+int btrfs_buffer_uptodate(struct extent_buffer *buf)
+{
+	struct inode *btree_inode = buf->pages[0]->mapping->host;
+	return extent_buffer_uptodate(&BTRFS_I(btree_inode)->extent_tree, buf);
+}
+
+int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
 {
-	struct btrfs_root *root = BTRFS_I(bh->b_page->mapping->host)->root;
-	u64 transid = btrfs_header_generation(btrfs_buffer_header(bh));
+	struct inode *btree_inode = buf->pages[0]->mapping->host;
+	return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->extent_tree,
+					  buf);
+}
 
-	WARN_ON(!atomic_read(&bh->b_count));
+void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
+{
+	struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
+	u64 transid = btrfs_header_generation(buf);
+	struct inode *btree_inode = root->fs_info->btree_inode;
 
 	if (transid != root->fs_info->generation) {
 		printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
-			(unsigned long long)bh->b_blocknr,
+			(unsigned long long)extent_buffer_blocknr(buf),
 			transid, root->fs_info->generation);
 		WARN_ON(1);
 	}
-	mark_buffer_dirty(bh);
-}
-
-void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf)
-{
-	brelse(buf);
+	set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf);
 }
 
 void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index fc6b520c6e0b..70d9413c599f 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -19,68 +19,35 @@
 #ifndef __DISKIO__
 #define __DISKIO__
 
-#include <linux/buffer_head.h>
-
 #define BTRFS_SUPER_INFO_OFFSET (16 * 1024)
 
-enum btrfs_bh_state_bits {
-	BH_Checked = BH_PrivateStart,
-	BH_Defrag,
-	BH_DefragDone,
-};
-BUFFER_FNS(Checked, checked);
-BUFFER_FNS(Defrag, defrag);
-BUFFER_FNS(DefragDone, defrag_done);
-
-static inline struct btrfs_node *btrfs_buffer_node(struct buffer_head *bh)
-{
-	return (struct btrfs_node *)bh->b_data;
-}
-
-static inline struct btrfs_leaf *btrfs_buffer_leaf(struct buffer_head *bh)
-{
-	return (struct btrfs_leaf *)bh->b_data;
-}
-
-static inline struct btrfs_header *btrfs_buffer_header(struct buffer_head *bh)
-{
-	return &((struct btrfs_node *)bh->b_data)->header;
-}
-
-struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr);
+struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr);
 int readahead_tree_block(struct btrfs_root *root, u64 blocknr);
-struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
-						 u64 blocknr);
-int write_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		     struct buffer_head *buf);
-int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		     struct buffer_head *buf);
+struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
+						   u64 blocknr);
 int clean_tree_block(struct btrfs_trans_handle *trans,
-		     struct btrfs_root *root, struct buffer_head *buf);
-int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
-			     struct btrfs_root *root);
+		     struct btrfs_root *root, struct extent_buffer *buf);
 struct btrfs_root *open_ctree(struct super_block *sb);
 int close_ctree(struct btrfs_root *root);
-void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf);
 int write_ctree_super(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root);
-struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr);
-int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
-		    char *result);
+struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
+					    u64 blocknr);
 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 				      struct btrfs_key *location,
 				      const char *name, int namelen);
 struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
 					       struct btrfs_key *location);
-u64 bh_blocknr(struct buffer_head *bh);
 int btrfs_insert_dev_radix(struct btrfs_root *root,
 			   struct block_device *bdev,
 			   u64 device_id,
 			   u64 block_start,
 			   u64 num_blocks);
-int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
-			     u64 logical);
 void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr);
 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root);
-void btrfs_mark_buffer_dirty(struct buffer_head *bh);
+void btrfs_mark_buffer_dirty(struct extent_buffer *buf);
+int btrfs_buffer_uptodate(struct extent_buffer *buf);
+int btrfs_set_buffer_uptodate(struct extent_buffer *buf);
+int wait_on_tree_block_writeback(struct btrfs_root *root,
+				 struct extent_buffer *buf);
 #endif
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index f261a8326cdf..089c41cbca74 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -33,7 +33,7 @@ static int cache_block_group(struct btrfs_root *root,
 	struct btrfs_path *path;
 	int ret;
 	struct btrfs_key key;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	struct radix_tree_root *extent_radix;
 	int slot;
 	u64 i;
@@ -56,7 +56,6 @@ static int cache_block_group(struct btrfs_root *root,
 	path->reada = 2;
 	first_free = block_group->key.objectid;
 	key.objectid = block_group->key.objectid;
-	key.flags = 0;
 	key.offset = 0;
 
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
@@ -69,9 +68,9 @@ static int cache_block_group(struct btrfs_root *root,
 		path->slots[0]--;
 
 	while(1) {
-		leaf = btrfs_buffer_leaf(path->nodes[0]);
+		leaf = path->nodes[0];
 		slot = path->slots[0];
-		if (slot >= btrfs_header_nritems(&leaf->header)) {
+		if (slot >= btrfs_header_nritems(leaf)) {
 			ret = btrfs_next_leaf(root, path);
 			if (ret < 0)
 				goto err;
@@ -82,7 +81,7 @@ static int cache_block_group(struct btrfs_root *root,
 			}
 		}
 
-		btrfs_disk_key_to_cpu(&key, &leaf->items[slot].key);
+		btrfs_item_key_to_cpu(leaf, &key, slot);
 		if (key.objectid < block_group->key.objectid) {
 			if (key.objectid + key.offset > first_free)
 				first_free = key.objectid + key.offset;
@@ -116,8 +115,7 @@ next:
 		hole_size = block_group->key.objectid +
 			block_group->key.offset - last;
 		for (i = 0; i < hole_size; i++) {
-			set_radix_bit(extent_radix,
-					last + i);
+			set_radix_bit(extent_radix, last + i);
 		}
 	}
 	block_group->cached = 1;
@@ -366,7 +364,7 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 	struct btrfs_path *path;
 	int ret;
 	struct btrfs_key key;
-	struct btrfs_leaf *l;
+	struct extent_buffer *l;
 	struct btrfs_extent_item *item;
 	u32 refs;
 
@@ -375,7 +373,6 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 		return -ENOMEM;
 
 	key.objectid = blocknr;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 	key.offset = num_blocks;
 	ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
@@ -386,10 +383,10 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 		BUG();
 	}
 	BUG_ON(ret != 0);
-	l = btrfs_buffer_leaf(path->nodes[0]);
+	l = path->nodes[0];
 	item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
-	refs = btrfs_extent_refs(item);
-	btrfs_set_extent_refs(item, refs + 1);
+	refs = btrfs_extent_refs(l, item);
+	btrfs_set_extent_refs(l, item, refs + 1);
 	btrfs_mark_buffer_dirty(path->nodes[0]);
 
 	btrfs_release_path(root->fs_info->extent_root, path);
@@ -414,23 +411,25 @@ static int lookup_extent_ref(struct btrfs_trans_handle *trans,
 	struct btrfs_path *path;
 	int ret;
 	struct btrfs_key key;
-	struct btrfs_leaf *l;
+	struct extent_buffer *l;
 	struct btrfs_extent_item *item;
 
 	path = btrfs_alloc_path();
 	key.objectid = blocknr;
 	key.offset = num_blocks;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 	ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
 				0, 0);
 	if (ret < 0)
 		goto out;
-	if (ret != 0)
+	if (ret != 0) {
+		btrfs_print_leaf(root, path->nodes[0]);
+		printk("failed to find block number %Lu\n", blocknr);
 		BUG();
-	l = btrfs_buffer_leaf(path->nodes[0]);
+	}
+	l = path->nodes[0];
 	item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
-	*refs = btrfs_extent_refs(item);
+	*refs = btrfs_extent_refs(l, item);
 out:
 	btrfs_free_path(path);
 	return 0;
@@ -439,16 +438,16 @@ out:
 int btrfs_inc_root_ref(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root)
 {
-	return btrfs_inc_extent_ref(trans, root, bh_blocknr(root->node), 1);
+	return btrfs_inc_extent_ref(trans, root,
+				    extent_buffer_blocknr(root->node), 1);
 }
 
 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		  struct buffer_head *buf)
+		  struct extent_buffer *buf)
 {
 	u64 blocknr;
-	struct btrfs_node *buf_node;
-	struct btrfs_leaf *buf_leaf;
-	struct btrfs_disk_key *key;
+	u32 nritems;
+	struct btrfs_key key;
 	struct btrfs_file_extent_item *fi;
 	int i;
 	int leaf;
@@ -458,31 +457,31 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 
 	if (!root->ref_cows)
 		return 0;
-	buf_node = btrfs_buffer_node(buf);
-	leaf = btrfs_is_leaf(buf_node);
-	buf_leaf = btrfs_buffer_leaf(buf);
-	for (i = 0; i < btrfs_header_nritems(&buf_node->header); i++) {
+
+	leaf = btrfs_is_leaf(buf);
+	nritems = btrfs_header_nritems(buf);
+	for (i = 0; i < nritems; i++) {
 		if (leaf) {
 			u64 disk_blocknr;
-			key = &buf_leaf->items[i].key;
-			if (btrfs_disk_key_type(key) != BTRFS_EXTENT_DATA_KEY)
+			btrfs_item_key_to_cpu(buf, &key, i);
+			if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
 				continue;
-			fi = btrfs_item_ptr(buf_leaf, i,
+			fi = btrfs_item_ptr(buf, i,
 					    struct btrfs_file_extent_item);
-			if (btrfs_file_extent_type(fi) ==
+			if (btrfs_file_extent_type(buf, fi) ==
 			    BTRFS_FILE_EXTENT_INLINE)
 				continue;
-			disk_blocknr = btrfs_file_extent_disk_blocknr(fi);
+			disk_blocknr = btrfs_file_extent_disk_blocknr(buf, fi);
 			if (disk_blocknr == 0)
 				continue;
 			ret = btrfs_inc_extent_ref(trans, root, disk_blocknr,
-				    btrfs_file_extent_disk_num_blocks(fi));
+				    btrfs_file_extent_disk_num_blocks(buf, fi));
 			if (ret) {
 				faili = i;
 				goto fail;
 			}
 		} else {
-			blocknr = btrfs_node_blockptr(buf_node, i);
+			blocknr = btrfs_node_blockptr(buf, i);
 			ret = btrfs_inc_extent_ref(trans, root, blocknr, 1);
 			if (ret) {
 				faili = i;
@@ -496,22 +495,23 @@ fail:
 	for (i =0; i < faili; i++) {
 		if (leaf) {
 			u64 disk_blocknr;
-			key = &buf_leaf->items[i].key;
-			if (btrfs_disk_key_type(key) != BTRFS_EXTENT_DATA_KEY)
+			btrfs_item_key_to_cpu(buf, &key, i);
+			if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
 				continue;
-			fi = btrfs_item_ptr(buf_leaf, i,
+			fi = btrfs_item_ptr(buf, i,
 					    struct btrfs_file_extent_item);
-			if (btrfs_file_extent_type(fi) ==
+			if (btrfs_file_extent_type(buf, fi) ==
 			    BTRFS_FILE_EXTENT_INLINE)
 				continue;
-			disk_blocknr = btrfs_file_extent_disk_blocknr(fi);
+			disk_blocknr = btrfs_file_extent_disk_blocknr(buf, fi);
 			if (disk_blocknr == 0)
 				continue;
 			err = btrfs_free_extent(trans, root, disk_blocknr,
-				    btrfs_file_extent_disk_num_blocks(fi), 0);
+				    btrfs_file_extent_disk_num_blocks(buf,
+								      fi), 0);
 			BUG_ON(err);
 		} else {
-			blocknr = btrfs_node_blockptr(buf_node, i);
+			blocknr = btrfs_node_blockptr(buf, i);
 			err = btrfs_free_extent(trans, root, blocknr, 1, 0);
 			BUG_ON(err);
 		}
@@ -527,16 +527,18 @@ static int write_one_cache_group(struct btrfs_trans_handle *trans,
 	int ret;
 	int pending_ret;
 	struct btrfs_root *extent_root = root->fs_info->extent_root;
-	struct btrfs_block_group_item *bi;
+	unsigned long bi;
+	struct extent_buffer *leaf;
 
 	ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
 	if (ret < 0)
 		goto fail;
 	BUG_ON(ret);
-	bi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
-			    struct btrfs_block_group_item);
-	memcpy(bi, &cache->item, sizeof(*bi));
-	btrfs_mark_buffer_dirty(path->nodes[0]);
+
+	leaf = path->nodes[0];
+	bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
+	write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
+	btrfs_mark_buffer_dirty(leaf);
 	btrfs_release_path(extent_root, path);
 fail:
 	finish_current_insert(trans, extent_root);
@@ -768,11 +770,11 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct
 	unsigned long gang[8];
 	struct btrfs_fs_info *info = extent_root->fs_info;
 
-	btrfs_set_extent_refs(&extent_item, 1);
+	btrfs_set_stack_extent_refs(&extent_item, 1);
 	ins.offset = 1;
-	ins.flags = 0;
 	btrfs_set_key_type(&ins, BTRFS_EXTENT_ITEM_KEY);
-	btrfs_set_extent_owner(&extent_item, extent_root->root_key.objectid);
+	btrfs_set_stack_extent_owner(&extent_item,
+				     extent_root->root_key.objectid);
 
 	while(1) {
 		ret = find_first_radix_bit(&info->extent_ins_radix, gang, 0,
@@ -795,23 +797,20 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct
 static int pin_down_block(struct btrfs_root *root, u64 blocknr, int pending)
 {
 	int err;
-	struct btrfs_header *header;
-	struct buffer_head *bh;
+	struct extent_buffer *buf;
 
 	if (!pending) {
-		bh = btrfs_find_tree_block(root, blocknr);
-		if (bh) {
-			if (buffer_uptodate(bh)) {
+		buf = btrfs_find_tree_block(root, blocknr);
+		if (buf) {
+			if (btrfs_buffer_uptodate(buf)) {
 				u64 transid =
 				    root->fs_info->running_transaction->transid;
-				header = btrfs_buffer_header(bh);
-				if (btrfs_header_generation(header) ==
-				    transid) {
-					btrfs_block_release(root, bh);
+				if (btrfs_header_generation(buf) == transid) {
+					free_extent_buffer(buf);
 					return 0;
 				}
 			}
-			btrfs_block_release(root, bh);
+			free_extent_buffer(buf);
 		}
 		err = set_radix_bit(&root->fs_info->pinned_radix, blocknr);
 		if (!err) {
@@ -839,12 +838,12 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	struct btrfs_key key;
 	struct btrfs_fs_info *info = root->fs_info;
 	struct btrfs_root *extent_root = info->extent_root;
+	struct extent_buffer *leaf;
 	int ret;
 	struct btrfs_extent_item *ei;
 	u32 refs;
 
 	key.objectid = blocknr;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 	key.offset = num_blocks;
 
@@ -856,12 +855,16 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	if (ret < 0)
 		return ret;
 	BUG_ON(ret);
-	ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
+
+	leaf = path->nodes[0];
+	ei = btrfs_item_ptr(leaf, path->slots[0],
 			    struct btrfs_extent_item);
-	BUG_ON(ei->refs == 0);
-	refs = btrfs_extent_refs(ei) - 1;
-	btrfs_set_extent_refs(ei, refs);
-	btrfs_mark_buffer_dirty(path->nodes[0]);
+	refs = btrfs_extent_refs(leaf, ei);
+	BUG_ON(refs == 0);
+	refs -= 1;
+	btrfs_set_extent_refs(leaf, ei, refs);
+	btrfs_mark_buffer_dirty(leaf);
+
 	if (refs == 0) {
 		u64 super_blocks_used, root_blocks_used;
 
@@ -876,8 +879,8 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 					    super_blocks_used - num_blocks);
 
 		/* block accounting for root item */
-		root_blocks_used = btrfs_root_blocks_used(&root->root_item);
-		btrfs_set_root_blocks_used(&root->root_item,
+		root_blocks_used = btrfs_root_used(&root->root_item);
+		btrfs_set_root_used(&root->root_item,
 					   root_blocks_used - num_blocks);
 
 		ret = btrfs_del_item(trans, extent_root, path);
@@ -984,7 +987,7 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	u64 test_block;
 	u64 orig_search_start = search_start;
 	int start_found;
-	struct btrfs_leaf *l;
+	struct extent_buffer *l;
 	struct btrfs_root * root = orig_root->fs_info->extent_root;
 	struct btrfs_fs_info *info = root->fs_info;
 	int total_needed = num_blocks;
@@ -994,10 +997,10 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	int wrapped = 0;
 
 	WARN_ON(num_blocks < 1);
-	ins->flags = 0;
 	btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
 
-	level = btrfs_header_level(btrfs_buffer_header(root->node));
+	level = btrfs_header_level(root->node);
+
 	if (search_end == (u64)-1)
 		search_end = btrfs_super_total_blocks(&info->super_copy);
 	if (hint_block) {
@@ -1034,8 +1037,9 @@ check_failed:
 		path->slots[0]--;
 	}
 
-	l = btrfs_buffer_leaf(path->nodes[0]);
-	btrfs_disk_key_to_cpu(&key, &l->items[path->slots[0]].key);
+	l = path->nodes[0];
+	btrfs_item_key_to_cpu(l, &key, path->slots[0]);
+
 	/*
 	 * a rare case, go back one key if we hit a block group item
 	 * instead of an extent item
@@ -1055,9 +1059,9 @@ check_failed:
 	}
 
 	while (1) {
-		l = btrfs_buffer_leaf(path->nodes[0]);
+		l = path->nodes[0];
 		slot = path->slots[0];
-		if (slot >= btrfs_header_nritems(&l->header)) {
+		if (slot >= btrfs_header_nritems(l)) {
 			ret = btrfs_next_leaf(root, path);
 			if (ret == 0)
 				continue;
@@ -1075,7 +1079,7 @@ check_failed:
 			goto check_pending;
 		}
 
-		btrfs_disk_key_to_cpu(&key, &l->items[slot].key);
+		btrfs_item_key_to_cpu(l, &key, slot);
 		if (key.objectid >= search_start && key.objectid > last_block &&
 		    start_found) {
 			if (last_block < search_start)
@@ -1183,8 +1187,8 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 	struct btrfs_root *extent_root = info->extent_root;
 	struct btrfs_extent_item extent_item;
 
-	btrfs_set_extent_refs(&extent_item, 1);
-	btrfs_set_extent_owner(&extent_item, owner);
+	btrfs_set_stack_extent_refs(&extent_item, 1);
+	btrfs_set_stack_extent_owner(&extent_item, owner);
 
 	WARN_ON(num_blocks < 1);
 	ret = find_free_extent(trans, root, num_blocks, empty_size,
@@ -1201,8 +1205,8 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 				    num_blocks);
 
 	/* block accounting for root item */
-	root_blocks_used = btrfs_root_blocks_used(&root->root_item);
-	btrfs_set_root_blocks_used(&root->root_item, root_blocks_used +
+	root_blocks_used = btrfs_root_used(&root->root_item);
+	btrfs_set_root_used(&root->root_item, root_blocks_used +
 				   num_blocks);
 
 	if (root == extent_root) {
@@ -1241,13 +1245,13 @@ update_block:
  * helper function to allocate a block for a given tree
  * returns the tree buffer or NULL.
  */
-struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
-					   struct btrfs_root *root, u64 hint,
-					   u64 empty_size)
+struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
+					     struct btrfs_root *root, u64 hint,
+					     u64 empty_size)
 {
 	struct btrfs_key ins;
 	int ret;
-	struct buffer_head *buf;
+	struct extent_buffer *buf;
 
 	ret = btrfs_alloc_extent(trans, root, root->root_key.objectid,
 				 1, empty_size, hint, (u64)-1, &ins, 0);
@@ -1260,53 +1264,57 @@ struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 		btrfs_free_extent(trans, root, ins.objectid, 1, 0);
 		return ERR_PTR(-ENOMEM);
 	}
-	WARN_ON(buffer_dirty(buf));
-	set_buffer_uptodate(buf);
+	btrfs_set_buffer_uptodate(buf);
+	set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
+			 buf->start + buf->len - 1, GFP_NOFS);
+	/*
 	set_buffer_checked(buf);
 	set_buffer_defrag(buf);
-	set_radix_bit(&trans->transaction->dirty_pages, buf->b_page->index);
+	*/
+	/* FIXME!!!!!!!!!!!!!!!!
+	set_radix_bit(&trans->transaction->dirty_pages, buf->pages[0]->index);
+	*/
 	trans->blocks_used++;
 	return buf;
 }
 
 static int drop_leaf_ref(struct btrfs_trans_handle *trans,
-			 struct btrfs_root *root, struct buffer_head *cur)
+			 struct btrfs_root *root, struct extent_buffer *leaf)
 {
-	struct btrfs_disk_key *key;
-	struct btrfs_leaf *leaf;
+	struct btrfs_key key;
 	struct btrfs_file_extent_item *fi;
 	int i;
 	int nritems;
 	int ret;
 
-	BUG_ON(!btrfs_is_leaf(btrfs_buffer_node(cur)));
-	leaf = btrfs_buffer_leaf(cur);
-	nritems = btrfs_header_nritems(&leaf->header);
+	BUG_ON(!btrfs_is_leaf(leaf));
+	nritems = btrfs_header_nritems(leaf);
 	for (i = 0; i < nritems; i++) {
 		u64 disk_blocknr;
-		key = &leaf->items[i].key;
-		if (btrfs_disk_key_type(key) != BTRFS_EXTENT_DATA_KEY)
+
+		btrfs_item_key_to_cpu(leaf, &key, i);
+		if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
 			continue;
 		fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
-		if (btrfs_file_extent_type(fi) == BTRFS_FILE_EXTENT_INLINE)
+		if (btrfs_file_extent_type(leaf, fi) ==
+		    BTRFS_FILE_EXTENT_INLINE)
 			continue;
 		/*
 		 * FIXME make sure to insert a trans record that
 		 * repeats the snapshot del on crash
 		 */
-		disk_blocknr = btrfs_file_extent_disk_blocknr(fi);
+		disk_blocknr = btrfs_file_extent_disk_blocknr(leaf, fi);
 		if (disk_blocknr == 0)
 			continue;
 		ret = btrfs_free_extent(trans, root, disk_blocknr,
-					btrfs_file_extent_disk_num_blocks(fi),
-					0);
+				btrfs_file_extent_disk_num_blocks(leaf, fi), 0);
 		BUG_ON(ret);
 	}
 	return 0;
 }
 
 static void reada_walk_down(struct btrfs_root *root,
-			    struct btrfs_node *node)
+			    struct extent_buffer *node)
 {
 	int i;
 	u32 nritems;
@@ -1314,7 +1322,7 @@ static void reada_walk_down(struct btrfs_root *root,
 	int ret;
 	u32 refs;
 
-	nritems = btrfs_header_nritems(&node->header);
+	nritems = btrfs_header_nritems(node);
 	for (i = 0; i < nritems; i++) {
 		blocknr = btrfs_node_blockptr(node, i);
 		ret = lookup_extent_ref(NULL, root, blocknr, 1, &refs);
@@ -1337,16 +1345,17 @@ static void reada_walk_down(struct btrfs_root *root,
 static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 			  *root, struct btrfs_path *path, int *level)
 {
-	struct buffer_head *next;
-	struct buffer_head *cur;
+	struct extent_buffer *next;
+	struct extent_buffer *cur;
 	u64 blocknr;
 	int ret;
 	u32 refs;
 
 	WARN_ON(*level < 0);
 	WARN_ON(*level >= BTRFS_MAX_LEVEL);
-	ret = lookup_extent_ref(trans, root, bh_blocknr(path->nodes[*level]),
-			       1, &refs);
+	ret = lookup_extent_ref(trans, root,
+				extent_buffer_blocknr(path->nodes[*level]),
+				1, &refs);
 	BUG_ON(ret);
 	if (refs > 1)
 		goto out;
@@ -1360,21 +1369,20 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 		cur = path->nodes[*level];
 
 		if (*level > 0 && path->slots[*level] == 0)
-			reada_walk_down(root, btrfs_buffer_node(cur));
+			reada_walk_down(root, cur);
 
-		if (btrfs_header_level(btrfs_buffer_header(cur)) != *level)
+		if (btrfs_header_level(cur) != *level)
 			WARN_ON(1);
 
 		if (path->slots[*level] >=
-		    btrfs_header_nritems(btrfs_buffer_header(cur)))
+		    btrfs_header_nritems(cur))
 			break;
 		if (*level == 0) {
 			ret = drop_leaf_ref(trans, root, cur);
 			BUG_ON(ret);
 			break;
 		}
-		blocknr = btrfs_node_blockptr(btrfs_buffer_node(cur),
-					      path->slots[*level]);
+		blocknr = btrfs_node_blockptr(cur, path->slots[*level]);
 		ret = lookup_extent_ref(trans, root, blocknr, 1, &refs);
 		BUG_ON(ret);
 		if (refs != 1) {
@@ -1384,8 +1392,8 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 			continue;
 		}
 		next = btrfs_find_tree_block(root, blocknr);
-		if (!next || !buffer_uptodate(next)) {
-			brelse(next);
+		if (!next || !btrfs_buffer_uptodate(next)) {
+			free_extent_buffer(next);
 			mutex_unlock(&root->fs_info->fs_mutex);
 			next = read_tree_block(root, blocknr);
 			mutex_lock(&root->fs_info->fs_mutex);
@@ -1395,7 +1403,7 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 			BUG_ON(ret);
 			if (refs != 1) {
 				path->slots[*level]++;
-				brelse(next);
+				free_extent_buffer(next);
 				ret = btrfs_free_extent(trans, root,
 							blocknr, 1, 1);
 				BUG_ON(ret);
@@ -1404,17 +1412,17 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 		}
 		WARN_ON(*level <= 0);
 		if (path->nodes[*level-1])
-			btrfs_block_release(root, path->nodes[*level-1]);
+			free_extent_buffer(path->nodes[*level-1]);
 		path->nodes[*level-1] = next;
-		*level = btrfs_header_level(btrfs_buffer_header(next));
+		*level = btrfs_header_level(next);
 		path->slots[*level] = 0;
 	}
 out:
 	WARN_ON(*level < 0);
 	WARN_ON(*level >= BTRFS_MAX_LEVEL);
 	ret = btrfs_free_extent(trans, root,
-				bh_blocknr(path->nodes[*level]), 1, 1);
-	btrfs_block_release(root, path->nodes[*level]);
+			extent_buffer_blocknr(path->nodes[*level]), 1, 1);
+	free_extent_buffer(path->nodes[*level]);
 	path->nodes[*level] = NULL;
 	*level += 1;
 	BUG_ON(ret);
@@ -1436,24 +1444,24 @@ static int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
 		slot = path->slots[i];
-		if (slot < btrfs_header_nritems(
-		    btrfs_buffer_header(path->nodes[i])) - 1) {
-			struct btrfs_node *node;
-			node = btrfs_buffer_node(path->nodes[i]);
+		if (slot < btrfs_header_nritems(path->nodes[i]) - 1) {
+			struct extent_buffer *node;
+			struct btrfs_disk_key disk_key;
+			node = path->nodes[i];
 			path->slots[i]++;
 			*level = i;
 			WARN_ON(*level == 0);
+			btrfs_node_key(node, &disk_key, path->slots[i]);
 			memcpy(&root_item->drop_progress,
-			       &node->ptrs[path->slots[i]].key,
-			       sizeof(root_item->drop_progress));
+			       &disk_key, sizeof(disk_key));
 			root_item->drop_level = i;
 			return 0;
 		} else {
 			ret = btrfs_free_extent(trans, root,
-						bh_blocknr(path->nodes[*level]),
-						1, 1);
+				    extent_buffer_blocknr(path->nodes[*level]),
+				    1, 1);
 			BUG_ON(ret);
-			btrfs_block_release(root, path->nodes[*level]);
+			free_extent_buffer(path->nodes[*level]);
 			path->nodes[*level] = NULL;
 			*level = i + 1;
 		}
@@ -1480,15 +1488,15 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
 
-	level = btrfs_header_level(btrfs_buffer_header(root->node));
+	level = btrfs_header_level(root->node);
 	orig_level = level;
 	if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
 		path->nodes[level] = root->node;
 		path->slots[level] = 0;
 	} else {
 		struct btrfs_key key;
-		struct btrfs_disk_key *found_key;
-		struct btrfs_node *node;
+		struct btrfs_disk_key found_key;
+		struct extent_buffer *node;
 
 		btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
 		level = root_item->drop_level;
@@ -1498,10 +1506,10 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
 			ret = wret;
 			goto out;
 		}
-		node = btrfs_buffer_node(path->nodes[level]);
-		found_key = &node->ptrs[path->slots[level]].key;
-		WARN_ON(memcmp(found_key, &root_item->drop_progress,
-			       sizeof(*found_key)));
+		node = path->nodes[level];
+		btrfs_node_key(node, &found_key, path->slots[level]);
+		WARN_ON(memcmp(&found_key, &root_item->drop_progress,
+			       sizeof(found_key)));
 	}
 	while(1) {
 		wret = walk_down_tree(trans, root, path, &level);
@@ -1516,12 +1524,12 @@ int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
 		if (wret < 0)
 			ret = wret;
 		ret = -EAGAIN;
-		get_bh(root->node);
+		extent_buffer_get(root->node);
 		break;
 	}
 	for (i = 0; i <= orig_level; i++) {
 		if (path->nodes[i]) {
-			btrfs_block_release(root, path->nodes[i]);
+			free_extent_buffer(path->nodes[i]);
 			path->nodes[i] = 0;
 		}
 	}
@@ -1581,13 +1589,12 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 	struct btrfs_path *path;
 	int ret;
 	int err = 0;
-	struct btrfs_block_group_item *bi;
 	struct btrfs_block_group_cache *cache;
 	struct btrfs_fs_info *info = root->fs_info;
 	struct radix_tree_root *radix;
 	struct btrfs_key key;
 	struct btrfs_key found_key;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	u64 group_size_blocks;
 	u64 used;
 
@@ -1596,7 +1603,6 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 	root = info->extent_root;
 	key.objectid = 0;
 	key.offset = group_size_blocks;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
 
 	path = btrfs_alloc_path();
@@ -1610,18 +1616,18 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 			err = ret;
 			break;
 		}
-		leaf = btrfs_buffer_leaf(path->nodes[0]);
-		btrfs_disk_key_to_cpu(&found_key,
-				      &leaf->items[path->slots[0]].key);
+		leaf = path->nodes[0];
+		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
 		cache = kmalloc(sizeof(*cache), GFP_NOFS);
 		if (!cache) {
 			err = -1;
 			break;
 		}
 
-		bi = btrfs_item_ptr(leaf, path->slots[0],
-				    struct btrfs_block_group_item);
-		if (bi->flags & BTRFS_BLOCK_GROUP_DATA) {
+		read_extent_buffer(leaf, &cache->item,
+				   btrfs_item_ptr_offset(leaf, path->slots[0]),
+				   sizeof(cache->item));
+		if (cache->item.flags & BTRFS_BLOCK_GROUP_DATA) {
 			radix = &info->block_group_data_radix;
 			cache->data = 1;
 		} else {
@@ -1629,7 +1635,6 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 			cache->data = 0;
 		}
 
-		memcpy(&cache->item, bi, sizeof(*bi));
 		memcpy(&cache->key, &found_key, sizeof(found_key));
 		cache->last_alloc = cache->key.objectid;
 		cache->first_free = cache->key.objectid;
@@ -1640,11 +1645,12 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 
 		key.objectid = found_key.objectid + found_key.offset;
 		btrfs_release_path(root, path);
+
 		ret = radix_tree_insert(radix, found_key.objectid +
 					found_key.offset - 1,
 					(void *)cache);
 		BUG_ON(ret);
-		used = btrfs_block_group_used(bi);
+		used = btrfs_block_group_used(&cache->item);
 		if (used < div_factor(key.offset, 8)) {
 			radix_tree_tag_set(radix, found_key.objectid +
 					   found_key.offset - 1,
diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c
index 0ab368e091f9..55fefdfab84c 100644
--- a/fs/btrfs/extent_map.c
+++ b/fs/btrfs/extent_map.c
@@ -26,16 +26,6 @@ struct tree_entry {
 	struct rb_node rb_node;
 };
 
-/* bits for the extent state */
-#define EXTENT_DIRTY 1
-#define EXTENT_WRITEBACK (1 << 1)
-#define EXTENT_UPTODATE (1 << 2)
-#define EXTENT_LOCKED (1 << 3)
-#define EXTENT_NEW (1 << 4)
-#define EXTENT_DELALLOC (1 << 5)
-
-#define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK)
-
 void __init extent_map_init(void)
 {
 	extent_map_cache = btrfs_cache_create("extent_map",
@@ -223,7 +213,8 @@ int add_extent_mapping(struct extent_map_tree *tree,
 		if (rb)
 			prev = rb_entry(rb, struct extent_map, rb_node);
 		if (prev && prev->end + 1 == em->start &&
-		    ((em->block_start == 0 && prev->block_start == 0) ||
+		    ((em->block_start == EXTENT_MAP_HOLE &&
+		      prev->block_start == EXTENT_MAP_HOLE) ||
 			     (em->block_start == prev->block_end + 1))) {
 			em->start = prev->start;
 			em->block_start = prev->block_start;
@@ -926,6 +917,40 @@ int set_range_writeback(struct extent_map_tree *tree, u64 start, u64 end)
 }
 EXPORT_SYMBOL(set_range_writeback);
 
+int find_first_extent_bit(struct extent_map_tree *tree, u64 start,
+			  u64 *start_ret, u64 *end_ret, int bits)
+{
+	struct rb_node *node;
+	struct extent_state *state;
+	int ret = 1;
+
+	write_lock_irq(&tree->lock);
+	/*
+	 * this search will find all the extents that end after
+	 * our range starts.
+	 */
+	node = tree_search(&tree->state, start);
+	if (!node || IS_ERR(node)) {
+		goto out;
+	}
+
+	while(1) {
+		state = rb_entry(node, struct extent_state, rb_node);
+		if (state->state & bits) {
+			*start_ret = state->start;
+			*end_ret = state->end;
+			ret = 0;
+		}
+		node = rb_next(node);
+		if (!node)
+			break;
+	}
+out:
+	write_unlock_irq(&tree->lock);
+	return ret;
+}
+EXPORT_SYMBOL(find_first_extent_bit);
+
 u64 find_lock_delalloc_range(struct extent_map_tree *tree,
 			     u64 start, u64 lock_start, u64 *end, u64 max_bytes)
 {
@@ -1450,7 +1475,7 @@ int extent_read_full_page(struct extent_map_tree *tree, struct page *page,
 		em = NULL;
 
 		/* we've found a hole, just zero and go on */
-		if (block_start == 0) {
+		if (block_start == EXTENT_MAP_HOLE) {
 			zero_user_page(page, page_offset, iosize, KM_USER0);
 			set_extent_uptodate(tree, cur, cur + iosize - 1,
 					    GFP_NOFS);
@@ -1593,7 +1618,8 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page,
 		free_extent_map(em);
 		em = NULL;
 
-		if (block_start == 0 || block_start == EXTENT_MAP_INLINE) {
+		if (block_start == EXTENT_MAP_HOLE ||
+		    block_start == EXTENT_MAP_INLINE) {
 			clear_extent_dirty(tree, cur,
 					   cur + iosize - 1, GFP_NOFS);
 			cur = cur + iosize;
@@ -1630,7 +1656,6 @@ int extent_write_full_page(struct extent_map_tree *tree, struct page *page,
 		nr++;
 	}
 done:
-	WARN_ON(test_range_bit(tree, start, page_end, EXTENT_DIRTY, 0));
 	unlock_extent(tree, start, page_end, GFP_NOFS);
 	unlock_page(page);
 	return 0;
@@ -1827,8 +1852,623 @@ sector_t extent_bmap(struct address_space *mapping, sector_t iblock,
 
 	// XXX(hch): block 0 is valid in some cases, e.g. XFS RT device
 	if (em->block_start == EXTENT_MAP_INLINE ||
-	    em->block_start == 0)
-	    	return 0;
+	    em->block_start == EXTENT_MAP_HOLE)
+		return 0;
 
 	return (em->block_start + start - em->start) >> inode->i_blkbits;
 }
+
+struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree,
+					  u64 start, unsigned long len,
+					  gfp_t mask)
+{
+	unsigned long num_pages = ((start + len - 1) >> PAGE_CACHE_SHIFT) -
+				  (start >> PAGE_CACHE_SHIFT) + 1;
+	unsigned long i;
+	unsigned long index = start >> PAGE_CACHE_SHIFT;
+	struct extent_buffer *eb;
+	struct page *p;
+	struct address_space *mapping = tree->mapping;
+	int uptodate = 0;
+
+	eb = kzalloc(EXTENT_BUFFER_SIZE(num_pages), mask);
+	if (!eb || IS_ERR(eb))
+		return NULL;
+
+	eb->start = start;
+	eb->len = len;
+	atomic_set(&eb->refs, 1);
+
+	for (i = 0; i < num_pages; i++, index++) {
+		p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM);
+		if (!p)
+			goto fail;
+		eb->pages[i] = p;
+		if (!PageUptodate(p))
+			uptodate = 0;
+		unlock_page(p);
+	}
+	if (uptodate)
+		eb->flags |= EXTENT_UPTODATE;
+	return eb;
+fail:
+	free_extent_buffer(eb);
+	return NULL;
+}
+EXPORT_SYMBOL(alloc_extent_buffer);
+
+struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree,
+					 u64 start, unsigned long len,
+					  gfp_t mask)
+{
+	unsigned long num_pages = ((start + len - 1) >> PAGE_CACHE_SHIFT) -
+				  (start >> PAGE_CACHE_SHIFT) + 1;
+	unsigned long i;
+	unsigned long index = start >> PAGE_CACHE_SHIFT;
+	struct extent_buffer *eb;
+	struct page *p;
+	struct address_space *mapping = tree->mapping;
+
+	eb = kzalloc(EXTENT_BUFFER_SIZE(num_pages), mask);
+	if (!eb || IS_ERR(eb))
+		return NULL;
+
+	eb->start = start;
+	eb->len = len;
+	atomic_set(&eb->refs, 1);
+
+	for (i = 0; i < num_pages; i++, index++) {
+		p = find_get_page(mapping, index);
+		if (!p)
+			goto fail;
+		eb->pages[i] = p;
+	}
+	return eb;
+fail:
+	free_extent_buffer(eb);
+	return NULL;
+}
+EXPORT_SYMBOL(find_extent_buffer);
+
+void free_extent_buffer(struct extent_buffer *eb)
+{
+	unsigned long i;
+	unsigned long num_pages;
+
+	if (!eb)
+		return;
+
+	if (!atomic_dec_and_test(&eb->refs))
+		return;
+
+	num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
+		(eb->start >> PAGE_CACHE_SHIFT) + 1;
+
+	for (i = 0; i < num_pages; i++) {
+		if (eb->pages[i])
+			page_cache_release(eb->pages[i]);
+	}
+	kfree(eb);
+}
+EXPORT_SYMBOL(free_extent_buffer);
+
+int clear_extent_buffer_dirty(struct extent_map_tree *tree,
+			      struct extent_buffer *eb)
+{
+	int set;
+	unsigned long i;
+	unsigned long num_pages;
+	struct page *page;
+
+	u64 start = eb->start;
+	u64 end = start + eb->len - 1;
+
+	set = clear_extent_dirty(tree, start, end, GFP_NOFS);
+	num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
+		(eb->start >> PAGE_CACHE_SHIFT) + 1;
+
+	for (i = 0; i < num_pages; i++) {
+		page = eb->pages[i];
+		lock_page(page);
+		/*
+		 * if we're on the last page or the first page and the
+		 * block isn't aligned on a page boundary, do extra checks
+		 * to make sure we don't clean page that is partially dirty
+		 */
+		if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) ||
+		    ((i == num_pages - 1) &&
+		     ((eb->start + eb->len - 1) & (PAGE_CACHE_SIZE - 1)))) {
+			start = page->index << PAGE_CACHE_SHIFT;
+			end  = start + PAGE_CACHE_SIZE - 1;
+			if (test_range_bit(tree, start, end,
+					   EXTENT_DIRTY, 0)) {
+				unlock_page(page);
+				continue;
+			}
+		}
+		clear_page_dirty_for_io(page);
+		unlock_page(page);
+	}
+	return 0;
+}
+EXPORT_SYMBOL(clear_extent_buffer_dirty);
+
+int wait_on_extent_buffer_writeback(struct extent_map_tree *tree,
+				    struct extent_buffer *eb)
+{
+	return wait_on_extent_writeback(tree, eb->start,
+					eb->start + eb->len - 1);
+}
+EXPORT_SYMBOL(wait_on_extent_buffer_writeback);
+
+int set_extent_buffer_dirty(struct extent_map_tree *tree,
+			     struct extent_buffer *eb)
+{
+	return set_range_dirty(tree, eb->start, eb->start + eb->len - 1);
+}
+EXPORT_SYMBOL(set_extent_buffer_dirty);
+
+int set_extent_buffer_uptodate(struct extent_map_tree *tree,
+				struct extent_buffer *eb)
+{
+	unsigned long i;
+	struct page *page;
+	unsigned long num_pages;
+
+	num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
+		(eb->start >> PAGE_CACHE_SHIFT) + 1;
+
+	set_extent_uptodate(tree, eb->start, eb->start + eb->len - 1,
+			    GFP_NOFS);
+	for (i = 0; i < num_pages; i++) {
+		page = eb->pages[i];
+		if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) ||
+		    ((i == num_pages - 1) &&
+		     ((eb->start + eb->len - 1) & (PAGE_CACHE_SIZE - 1)))) {
+			check_page_uptodate(tree, page);
+			continue;
+		}
+		SetPageUptodate(page);
+	}
+	return 0;
+}
+EXPORT_SYMBOL(set_extent_buffer_uptodate);
+
+int extent_buffer_uptodate(struct extent_map_tree *tree,
+			     struct extent_buffer *eb)
+{
+	if (eb->flags & EXTENT_UPTODATE)
+		return 1;
+	return test_range_bit(tree, eb->start, eb->start + eb->len - 1,
+			   EXTENT_UPTODATE, 1);
+}
+EXPORT_SYMBOL(extent_buffer_uptodate);
+
+int read_extent_buffer_pages(struct extent_map_tree *tree,
+			     struct extent_buffer *eb, int wait)
+{
+	unsigned long i;
+	struct page *page;
+	int err;
+	int ret = 0;
+	unsigned long num_pages;
+
+	if (eb->flags & EXTENT_UPTODATE)
+		return 0;
+
+	if (test_range_bit(tree, eb->start, eb->start + eb->len - 1,
+			   EXTENT_UPTODATE, 1)) {
+		return 0;
+	}
+
+	num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
+		(eb->start >> PAGE_CACHE_SHIFT) + 1;
+	for (i = 0; i < num_pages; i++) {
+		page = eb->pages[i];
+		if (PageUptodate(page)) {
+			continue;
+		}
+		if (!wait) {
+			if (TestSetPageLocked(page)) {
+				continue;
+			}
+		} else {
+			lock_page(page);
+		}
+		if (!PageUptodate(page)) {
+			err = page->mapping->a_ops->readpage(NULL, page);
+			if (err) {
+				ret = err;
+			}
+		} else {
+			unlock_page(page);
+		}
+	}
+
+	if (ret || !wait) {
+		return ret;
+	}
+
+	for (i = 0; i < num_pages; i++) {
+		page = eb->pages[i];
+		wait_on_page_locked(page);
+		if (!PageUptodate(page)) {
+			ret = -EIO;
+		}
+	}
+	eb->flags |= EXTENT_UPTODATE;
+	return ret;
+}
+EXPORT_SYMBOL(read_extent_buffer_pages);
+
+void read_extent_buffer(struct extent_buffer *eb, void *dstv,
+			unsigned long start,
+			unsigned long len)
+{
+	size_t cur;
+	size_t offset;
+	struct page *page;
+	char *kaddr;
+	char *dst = (char *)dstv;
+	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
+	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
+
+	WARN_ON(start > eb->len);
+	WARN_ON(start + len > eb->start + eb->len);
+
+	page = eb->pages[i];
+	offset = start & ((unsigned long)PAGE_CACHE_SIZE - 1);
+	if (i == 0)
+		offset += start_offset;
+
+	while(len > 0) {
+		WARN_ON(!PageUptodate(page));
+
+		cur = min(len, (PAGE_CACHE_SIZE - offset));
+		// kaddr = kmap_atomic(page, KM_USER0);
+		kaddr = page_address(page);
+		memcpy(dst, kaddr + offset, cur);
+		// kunmap_atomic(kaddr, KM_USER0);
+
+		dst += cur;
+		len -= cur;
+		offset = 0;
+		i++;
+		page = eb->pages[i];
+	}
+}
+EXPORT_SYMBOL(read_extent_buffer);
+
+int map_extent_buffer(struct extent_buffer *eb, unsigned long start,
+		      char **token, char **map,
+		      unsigned long *map_start,
+		      unsigned long *map_len, int km)
+{
+	size_t offset;
+	char *kaddr;
+	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
+	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
+
+	WARN_ON(start > eb->len);
+
+	if (i == 0) {
+		offset = start_offset;
+		*map_start = 0;
+	} else {
+		offset = 0;
+		*map_start = (i << PAGE_CACHE_SHIFT) - offset;
+	}
+
+	// kaddr = kmap_atomic(eb->pages[i], km);
+	kaddr = page_address(eb->pages[i]);
+	*token = kaddr;
+	*map = kaddr + offset;
+	*map_len = PAGE_CACHE_SIZE - offset;
+	return 0;
+}
+EXPORT_SYMBOL(map_extent_buffer);
+
+void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km)
+{
+	// kunmap_atomic(token, km);
+}
+EXPORT_SYMBOL(unmap_extent_buffer);
+
+int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv,
+			  unsigned long start,
+			  unsigned long len)
+{
+	size_t cur;
+	size_t offset;
+	struct page *page;
+	char *kaddr;
+	char *ptr = (char *)ptrv;
+	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
+	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
+	int ret = 0;
+
+	WARN_ON(start > eb->len);
+	WARN_ON(start + len > eb->start + eb->len);
+
+	page = eb->pages[i];
+	offset = start & ((unsigned long)PAGE_CACHE_SIZE - 1);
+	if (i == 0)
+		offset += start_offset;
+
+	while(len > 0) {
+		WARN_ON(!PageUptodate(page));
+
+		cur = min(len, (PAGE_CACHE_SIZE - offset));
+
+		// kaddr = kmap_atomic(page, KM_USER0);
+		kaddr = page_address(page);
+		ret = memcmp(ptr, kaddr + offset, cur);
+		// kunmap_atomic(kaddr, KM_USER0);
+		if (ret)
+			break;
+
+		ptr += cur;
+		len -= cur;
+		offset = 0;
+		i++;
+		page = eb->pages[i];
+	}
+	return ret;
+}
+EXPORT_SYMBOL(memcmp_extent_buffer);
+
+void write_extent_buffer(struct extent_buffer *eb, const void *srcv,
+			 unsigned long start, unsigned long len)
+{
+	size_t cur;
+	size_t offset;
+	struct page *page;
+	char *kaddr;
+	char *src = (char *)srcv;
+	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
+	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
+
+	WARN_ON(start > eb->len);
+	WARN_ON(start + len > eb->start + eb->len);
+
+	page = eb->pages[i];
+	offset = start & ((unsigned long)PAGE_CACHE_SIZE - 1);
+	if (i == 0)
+		offset += start_offset;
+
+	while(len > 0) {
+		WARN_ON(!PageUptodate(page));
+
+		cur = min(len, PAGE_CACHE_SIZE - offset);
+		// kaddr = kmap_atomic(page, KM_USER0);
+		kaddr = page_address(page);
+		memcpy(kaddr + offset, src, cur);
+		// kunmap_atomic(kaddr, KM_USER0);
+
+		src += cur;
+		len -= cur;
+		offset = 0;
+		i++;
+		page = eb->pages[i];
+	}
+}
+EXPORT_SYMBOL(write_extent_buffer);
+
+void memset_extent_buffer(struct extent_buffer *eb, char c,
+			  unsigned long start, unsigned long len)
+{
+	size_t cur;
+	size_t offset;
+	struct page *page;
+	char *kaddr;
+	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
+	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
+
+	WARN_ON(start > eb->len);
+	WARN_ON(start + len > eb->start + eb->len);
+
+	page = eb->pages[i];
+	offset = start & ((unsigned long)PAGE_CACHE_SIZE - 1);
+	if (i == 0)
+		offset += start_offset;
+
+	while(len > 0) {
+		WARN_ON(!PageUptodate(page));
+
+		cur = min(len, PAGE_CACHE_SIZE - offset);
+		// kaddr = kmap_atomic(page, KM_USER0);
+		kaddr = page_address(page);
+		memset(kaddr + offset, c, cur);
+		// kunmap_atomic(kaddr, KM_USER0);
+
+		len -= cur;
+		offset = 0;
+		i++;
+		page = eb->pages[i];
+	}
+}
+EXPORT_SYMBOL(memset_extent_buffer);
+
+void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
+			unsigned long dst_offset, unsigned long src_offset,
+			unsigned long len)
+{
+	u64 dst_len = dst->len;
+	size_t cur;
+	size_t offset;
+	struct page *page;
+	char *kaddr;
+	size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
+	unsigned long i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT;
+
+	WARN_ON(src->len != dst_len);
+
+	offset = dst_offset & ((unsigned long)PAGE_CACHE_SIZE - 1);
+	if (i == 0)
+		offset += start_offset;
+
+	while(len > 0) {
+		page = dst->pages[i];
+		WARN_ON(!PageUptodate(page));
+
+		cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - offset));
+
+		// kaddr = kmap_atomic(page, KM_USER1);
+		kaddr = page_address(page);
+		read_extent_buffer(src, kaddr + offset, src_offset, cur);
+		// kunmap_atomic(kaddr, KM_USER1);
+
+		src_offset += cur;
+		len -= cur;
+		offset = 0;
+		i++;
+	}
+}
+EXPORT_SYMBOL(copy_extent_buffer);
+
+static void move_pages(struct page *dst_page, struct page *src_page,
+		       unsigned long dst_off, unsigned long src_off,
+		       unsigned long len)
+{
+	// char *dst_kaddr = kmap_atomic(dst_page, KM_USER0);
+	char *dst_kaddr = page_address(dst_page);
+	if (dst_page == src_page) {
+		memmove(dst_kaddr + dst_off, dst_kaddr + src_off, len);
+	} else {
+		// char *src_kaddr = kmap_atomic(src_page, KM_USER1);
+		char *src_kaddr = page_address(src_page);
+		char *p = dst_kaddr + dst_off + len;
+		char *s = src_kaddr + src_off + len;
+
+		while (len--)
+			*--p = *--s;
+
+		// kunmap_atomic(src_kaddr, KM_USER1);
+	}
+	// kunmap_atomic(dst_kaddr, KM_USER0);
+}
+
+static void copy_pages(struct page *dst_page, struct page *src_page,
+		       unsigned long dst_off, unsigned long src_off,
+		       unsigned long len)
+{
+	//kmap_atomic(dst_page, KM_USER0);
+	char *dst_kaddr = page_address(dst_page);
+	char *src_kaddr;
+
+	if (dst_page != src_page)
+		src_kaddr = page_address(src_page); // kmap_atomic(src_page, KM_USER1);
+	else
+		src_kaddr = dst_kaddr;
+
+	memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len);
+	/*
+	kunmap_atomic(dst_kaddr, KM_USER0);
+	if (dst_page != src_page)
+		kunmap_atomic(src_kaddr, KM_USER1);
+	*/
+}
+
+void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
+			   unsigned long src_offset, unsigned long len)
+{
+	size_t cur;
+	size_t dst_off_in_page;
+	size_t src_off_in_page;
+	size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
+	unsigned long dst_i;
+	unsigned long src_i;
+
+	if (src_offset + len > dst->len) {
+		printk("memmove bogus src_offset %lu move len %lu len %lu\n",
+		       src_offset, len, dst->len);
+		BUG_ON(1);
+	}
+	if (dst_offset + len > dst->len) {
+		printk("memmove bogus dst_offset %lu move len %lu len %lu\n",
+		       dst_offset, len, dst->len);
+		BUG_ON(1);
+	}
+
+	while(len > 0) {
+		dst_off_in_page = dst_offset &
+			((unsigned long)PAGE_CACHE_SIZE - 1);
+		src_off_in_page = src_offset &
+			((unsigned long)PAGE_CACHE_SIZE - 1);
+
+		dst_i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT;
+		src_i = (start_offset + src_offset) >> PAGE_CACHE_SHIFT;
+
+		if (src_i == 0)
+			src_off_in_page += start_offset;
+		if (dst_i == 0)
+			dst_off_in_page += start_offset;
+
+		cur = min(len, (unsigned long)(PAGE_CACHE_SIZE -
+					       src_off_in_page));
+		cur = min(cur, (unsigned long)(PAGE_CACHE_SIZE -
+					       dst_off_in_page));
+
+		copy_pages(dst->pages[dst_i], dst->pages[src_i],
+			   dst_off_in_page, src_off_in_page, cur);
+
+		src_offset += cur;
+		dst_offset += cur;
+		len -= cur;
+	}
+}
+EXPORT_SYMBOL(memcpy_extent_buffer);
+
+void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
+			   unsigned long src_offset, unsigned long len)
+{
+	size_t cur;
+	size_t dst_off_in_page;
+	size_t src_off_in_page;
+	unsigned long dst_end = dst_offset + len - 1;
+	unsigned long src_end = src_offset + len - 1;
+	size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
+	unsigned long dst_i;
+	unsigned long src_i;
+
+	if (src_offset + len > dst->len) {
+		printk("memmove bogus src_offset %lu move len %lu len %lu\n",
+		       src_offset, len, dst->len);
+		BUG_ON(1);
+	}
+	if (dst_offset + len > dst->len) {
+		printk("memmove bogus dst_offset %lu move len %lu len %lu\n",
+		       dst_offset, len, dst->len);
+		BUG_ON(1);
+	}
+	if (dst_offset < src_offset) {
+		memcpy_extent_buffer(dst, dst_offset, src_offset, len);
+		return;
+	}
+	while(len > 0) {
+		dst_i = (start_offset + dst_end) >> PAGE_CACHE_SHIFT;
+		src_i = (start_offset + src_end) >> PAGE_CACHE_SHIFT;
+
+		dst_off_in_page = dst_end &
+			((unsigned long)PAGE_CACHE_SIZE - 1);
+		src_off_in_page = src_end &
+			((unsigned long)PAGE_CACHE_SIZE - 1);
+
+		if (src_i == 0)
+			src_off_in_page += start_offset;
+		if (dst_i == 0)
+			dst_off_in_page += start_offset;
+
+		cur = min(len, src_off_in_page + 1);
+		cur = min(cur, dst_off_in_page + 1);
+// printk("move pages orig dst %lu src %lu len %lu, this %lu %lu %lu\n", dst_offset, src_offset, len, dst_off_in_page - cur + 1, src_off_in_page - cur + 1, cur);
+		move_pages(dst->pages[dst_i], dst->pages[src_i],
+			   dst_off_in_page - cur + 1,
+			   src_off_in_page - cur + 1, cur);
+
+		dst_end -= cur - 1;
+		src_end -= cur - 1;
+		len -= cur;
+	}
+}
+EXPORT_SYMBOL(memmove_extent_buffer);
diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h
index d913ce01248d..430b997a70f6 100644
--- a/fs/btrfs/extent_map.h
+++ b/fs/btrfs/extent_map.h
@@ -3,9 +3,20 @@
 
 #include <linux/rbtree.h>
 
+#define EXTENT_MAP_HOLE (u64)-3
 #define EXTENT_MAP_INLINE (u64)-2
 #define EXTENT_MAP_DELALLOC (u64)-1
 
+/* bits for the extent state */
+#define EXTENT_DIRTY 1
+#define EXTENT_WRITEBACK (1 << 1)
+#define EXTENT_UPTODATE (1 << 2)
+#define EXTENT_LOCKED (1 << 3)
+#define EXTENT_NEW (1 << 4)
+#define EXTENT_DELALLOC (1 << 5)
+#define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK)
+
+
 struct extent_map_ops {
 	int (*fill_delalloc)(struct inode *inode, u64 start, u64 end);
 	int (*writepage_io_hook)(struct page *page, u64 start, u64 end);
@@ -51,10 +62,13 @@ struct extent_state {
 	struct list_head list;
 };
 
+#define EXTENT_BUFFER_SIZE(nr) (sizeof(struct extent_buffer) + \
+			       (nr - 1) * sizeof(struct page *))
 struct extent_buffer {
 	u64 start;
-	u64 end; /* inclusive */
-	char *addr;
+	unsigned long len;
+	atomic_t refs;
+	int flags;
 	struct page *pages[];
 };
 
@@ -87,8 +101,12 @@ int set_extent_new(struct extent_map_tree *tree, u64 start, u64 end,
 		   gfp_t mask);
 int set_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end,
 		     gfp_t mask);
+int clear_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end,
+		       gfp_t mask);
 int set_extent_delalloc(struct extent_map_tree *tree, u64 start, u64 end,
 		     gfp_t mask);
+int find_first_extent_bit(struct extent_map_tree *tree, u64 start,
+			  u64 *start_ret, u64 *end_ret, int bits);
 int extent_invalidatepage(struct extent_map_tree *tree,
 			  struct page *page, unsigned long offset);
 int extent_write_full_page(struct extent_map_tree *tree, struct page *page,
@@ -106,4 +124,57 @@ int set_range_dirty(struct extent_map_tree *tree, u64 start, u64 end);
 int set_state_private(struct extent_map_tree *tree, u64 start, u64 private);
 int get_state_private(struct extent_map_tree *tree, u64 start, u64 *private);
 void set_page_extent_mapped(struct page *page);
+
+struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree,
+					  u64 start, unsigned long len,
+					  gfp_t mask);
+struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree,
+					 u64 start, unsigned long len,
+					  gfp_t mask);
+void free_extent_buffer(struct extent_buffer *eb);
+int read_extent_buffer_pages(struct extent_map_tree *tree,
+			     struct extent_buffer *eb, int wait);
+
+static inline void extent_buffer_get(struct extent_buffer *eb)
+{
+	atomic_inc(&eb->refs);
+}
+
+static inline u64 extent_buffer_blocknr(struct extent_buffer *eb)
+{
+	return eb->start / 4096;
+}
+
+int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv,
+			  unsigned long start,
+			  unsigned long len);
+void read_extent_buffer(struct extent_buffer *eb, void *dst,
+			unsigned long start,
+			unsigned long len);
+void write_extent_buffer(struct extent_buffer *eb, const void *src,
+			 unsigned long start, unsigned long len);
+void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
+			unsigned long dst_offset, unsigned long src_offset,
+			unsigned long len);
+void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
+			   unsigned long src_offset, unsigned long len);
+void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
+			   unsigned long src_offset, unsigned long len);
+void memset_extent_buffer(struct extent_buffer *eb, char c,
+			  unsigned long start, unsigned long len);
+int wait_on_extent_buffer_writeback(struct extent_map_tree *tree,
+				    struct extent_buffer *eb);
+int clear_extent_buffer_dirty(struct extent_map_tree *tree,
+			      struct extent_buffer *eb);
+int set_extent_buffer_dirty(struct extent_map_tree *tree,
+			     struct extent_buffer *eb);
+int set_extent_buffer_uptodate(struct extent_map_tree *tree,
+			       struct extent_buffer *eb);
+int extent_buffer_uptodate(struct extent_map_tree *tree,
+			   struct extent_buffer *eb);
+int map_extent_buffer(struct extent_buffer *eb, unsigned long offset,
+		      char **token, char **map,
+		      unsigned long *map_start,
+		      unsigned long *map_len, int km);
+void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km);
 #endif
diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c
index ea4dd4c5fce3..226f6d028c3f 100644
--- a/fs/btrfs/file-item.c
+++ b/fs/btrfs/file-item.c
@@ -34,12 +34,12 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
 	struct btrfs_file_extent_item *item;
 	struct btrfs_key file_key;
 	struct btrfs_path *path;
+	struct extent_buffer *leaf;
 
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
 	file_key.objectid = objectid;
 	file_key.offset = pos;
-	file_key.flags = 0;
 	btrfs_set_key_type(&file_key, BTRFS_EXTENT_DATA_KEY);
 
 	ret = btrfs_insert_empty_item(trans, root, path, &file_key,
@@ -47,15 +47,16 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
 	if (ret < 0)
 		goto out;
 	BUG_ON(ret);
-	item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
+	leaf = path->nodes[0];
+	item = btrfs_item_ptr(leaf, path->slots[0],
 			      struct btrfs_file_extent_item);
-	btrfs_set_file_extent_disk_blocknr(item, offset);
-	btrfs_set_file_extent_disk_num_blocks(item, disk_num_blocks);
-	btrfs_set_file_extent_offset(item, 0);
-	btrfs_set_file_extent_num_blocks(item, num_blocks);
-	btrfs_set_file_extent_generation(item, trans->transid);
-	btrfs_set_file_extent_type(item, BTRFS_FILE_EXTENT_REG);
-	btrfs_mark_buffer_dirty(path->nodes[0]);
+	btrfs_set_file_extent_disk_blocknr(leaf, item, offset);
+	btrfs_set_file_extent_disk_num_blocks(leaf, item, disk_num_blocks);
+	btrfs_set_file_extent_offset(leaf, item, 0);
+	btrfs_set_file_extent_num_blocks(leaf, item, num_blocks);
+	btrfs_set_file_extent_generation(leaf, item, trans->transid);
+	btrfs_set_file_extent_type(leaf, item, BTRFS_FILE_EXTENT_REG);
+	btrfs_mark_buffer_dirty(leaf);
 out:
 	btrfs_free_path(path);
 	return ret;
@@ -71,32 +72,30 @@ struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans,
 	struct btrfs_key file_key;
 	struct btrfs_key found_key;
 	struct btrfs_csum_item *item;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	u64 csum_offset = 0;
 	int csums_in_item;
 
 	file_key.objectid = objectid;
 	file_key.offset = offset;
-	file_key.flags = 0;
 	btrfs_set_key_type(&file_key, BTRFS_CSUM_ITEM_KEY);
 	ret = btrfs_search_slot(trans, root, &file_key, path, 0, cow);
 	if (ret < 0)
 		goto fail;
-	leaf = btrfs_buffer_leaf(path->nodes[0]);
+	leaf = path->nodes[0];
 	if (ret > 0) {
 		ret = 1;
 		if (path->slots[0] == 0)
 			goto fail;
 		path->slots[0]--;
-		btrfs_disk_key_to_cpu(&found_key,
-				      &leaf->items[path->slots[0]].key);
+		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
 		if (btrfs_key_type(&found_key) != BTRFS_CSUM_ITEM_KEY ||
 		    found_key.objectid != objectid) {
 			goto fail;
 		}
 		csum_offset = (offset - found_key.offset) >>
 				root->fs_info->sb->s_blocksize_bits;
-		csums_in_item = btrfs_item_size(leaf->items + path->slots[0]);
+		csums_in_item = btrfs_item_size_nr(leaf, path->slots[0]);
 		csums_in_item /= BTRFS_CRC32_SIZE;
 
 		if (csum_offset >= csums_in_item) {
@@ -127,7 +126,6 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans,
 
 	file_key.objectid = objectid;
 	file_key.offset = offset;
-	file_key.flags = 0;
 	btrfs_set_key_type(&file_key, BTRFS_EXTENT_DATA_KEY);
 	ret = btrfs_search_slot(trans, root, &file_key, path, ins_len, cow);
 	return ret;
@@ -138,12 +136,14 @@ int btrfs_csum_file_block(struct btrfs_trans_handle *trans,
 			  u64 objectid, u64 offset,
 			  char *data, size_t len)
 {
+	return 0;
+#if 0
 	int ret;
 	struct btrfs_key file_key;
 	struct btrfs_key found_key;
 	struct btrfs_path *path;
 	struct btrfs_csum_item *item;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	u64 csum_offset;
 
 	path = btrfs_alloc_path();
@@ -161,8 +161,8 @@ int btrfs_csum_file_block(struct btrfs_trans_handle *trans,
 	if (ret == -EFBIG) {
 		u32 item_size;
 		/* we found one, but it isn't big enough yet */
-		leaf = btrfs_buffer_leaf(path->nodes[0]);
-		item_size = btrfs_item_size(leaf->items + path->slots[0]);
+		leaf = path->nodes[0];
+		item_size = btrfs_item_size_nr(leaf, path->slots[0]);
 		if ((item_size / BTRFS_CRC32_SIZE) >= MAX_CSUM_ITEMS(root)) {
 			/* already at max size, make a new one */
 			goto insert;
@@ -188,8 +188,8 @@ int btrfs_csum_file_block(struct btrfs_trans_handle *trans,
 		goto insert;
 	}
 	path->slots[0]--;
-	leaf = btrfs_buffer_leaf(path->nodes[0]);
-	btrfs_disk_key_to_cpu(&found_key, &leaf->items[path->slots[0]].key);
+	leaf = path->nodes[0];
+	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
 	csum_offset = (offset - found_key.offset) >>
 			root->fs_info->sb->s_blocksize_bits;
 	if (btrfs_key_type(&found_key) != BTRFS_CSUM_ITEM_KEY ||
@@ -197,10 +197,10 @@ int btrfs_csum_file_block(struct btrfs_trans_handle *trans,
 	    csum_offset >= MAX_CSUM_ITEMS(root)) {
 		goto insert;
 	}
-	if (csum_offset >= btrfs_item_size(leaf->items + path->slots[0]) /
+	if (csum_offset >= btrfs_item_size_nr(leaf, path->slots[0]) /
 	    BTRFS_CRC32_SIZE) {
 		u32 diff = (csum_offset + 1) * BTRFS_CRC32_SIZE;
-		diff = diff - btrfs_item_size(leaf->items + path->slots[0]);
+		diff = diff - btrfs_item_size_nr(leaf, path->slots[0]);
 		if (diff != BTRFS_CRC32_SIZE)
 			goto insert;
 		ret = btrfs_extend_item(trans, root, path, diff);
@@ -220,21 +220,20 @@ insert:
 		goto fail;
 	}
 csum:
-	item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
-			      struct btrfs_csum_item);
+	leaf = path->nodes[0];
+	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item);
 	ret = 0;
 	item = (struct btrfs_csum_item *)((unsigned char *)item +
 					  csum_offset * BTRFS_CRC32_SIZE);
 found:
-	btrfs_check_bounds(&item->csum, BTRFS_CRC32_SIZE,
-			   path->nodes[0]->b_data,
-			   root->fs_info->sb->s_blocksize);
+	/* FIXME!!!!!!!!!!!! */
 	ret = btrfs_csum_data(root, data, len, &item->csum);
 	btrfs_mark_buffer_dirty(path->nodes[0]);
 fail:
 	btrfs_release_path(root, path);
 	btrfs_free_path(path);
 	return ret;
+#endif
 }
 
 int btrfs_csum_truncate(struct btrfs_trans_handle *trans,
@@ -242,21 +241,21 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans,
 			u64 isize)
 {
 	struct btrfs_key key;
-	struct btrfs_leaf *leaf = btrfs_buffer_leaf(path->nodes[0]);
+	struct extent_buffer *leaf = path->nodes[0];
 	int slot = path->slots[0];
 	int ret;
 	u32 new_item_size;
 	u64 new_item_span;
 	u64 blocks;
 
-	btrfs_disk_key_to_cpu(&key, &leaf->items[slot].key);
+	btrfs_item_key_to_cpu(leaf, &key, slot);
 	if (isize <= key.offset)
 		return 0;
 	new_item_span = isize - key.offset;
-	blocks = (new_item_span + root->blocksize - 1) >>
+	blocks = (new_item_span + root->sectorsize - 1) >>
 		root->fs_info->sb->s_blocksize_bits;
 	new_item_size = blocks * BTRFS_CRC32_SIZE;
-	if (new_item_size >= btrfs_item_size(leaf->items + slot))
+	if (new_item_size >= btrfs_item_size_nr(leaf, slot))
 		return 0;
 	ret = btrfs_truncate_item(trans, root, path, new_item_size);
 	BUG_ON(ret);
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 4cc459c943ec..1734ca695555 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -16,7 +16,6 @@
  * Boston, MA 021110-1307, USA.
  */
 
-#include <linux/buffer_head.h>
 #include <linux/fs.h>
 #include <linux/pagemap.h>
 #include <linux/highmem.h>
@@ -88,7 +87,9 @@ static int insert_inline_extent(struct btrfs_trans_handle *trans,
 {
 	struct btrfs_key key;
 	struct btrfs_path *path;
-	char *ptr, *kaddr;
+	struct extent_buffer *leaf;
+	char *kaddr;
+	unsigned long ptr;
 	struct btrfs_file_extent_item *ei;
 	u32 datasize;
 	int err = 0;
@@ -102,7 +103,6 @@ static int insert_inline_extent(struct btrfs_trans_handle *trans,
 
 	key.objectid = inode->i_ino;
 	key.offset = offset;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
 	BUG_ON(size >= PAGE_CACHE_SIZE);
 	datasize = btrfs_file_extent_calc_inline_size(size);
@@ -113,18 +113,17 @@ static int insert_inline_extent(struct btrfs_trans_handle *trans,
 		err = ret;
 		goto fail;
 	}
-	ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
-	       path->slots[0], struct btrfs_file_extent_item);
-	btrfs_set_file_extent_generation(ei, trans->transid);
-	btrfs_set_file_extent_type(ei,
-				   BTRFS_FILE_EXTENT_INLINE);
+	leaf = path->nodes[0];
+	ei = btrfs_item_ptr(leaf, path->slots[0],
+			    struct btrfs_file_extent_item);
+	btrfs_set_file_extent_generation(leaf, ei, trans->transid);
+	btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE);
 	ptr = btrfs_file_extent_inline_start(ei);
 
 	kaddr = kmap_atomic(page, KM_USER0);
-	btrfs_memcpy(root, path->nodes[0]->b_data,
-		     ptr, kaddr + page_offset, size);
+	write_extent_buffer(leaf, kaddr + page_offset, ptr, size);
 	kunmap_atomic(kaddr, KM_USER0);
-	btrfs_mark_buffer_dirty(path->nodes[0]);
+	btrfs_mark_buffer_dirty(leaf);
 fail:
 	btrfs_free_path(path);
 	return err;
@@ -156,8 +155,8 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 
 	em->bdev = inode->i_sb->s_bdev;
 
-	start_pos = pos & ~((u64)root->blocksize - 1);
-	num_blocks = (write_bytes + pos - start_pos + root->blocksize - 1) >>
+	start_pos = pos & ~((u64)root->sectorsize - 1);
+	num_blocks = (write_bytes + pos - start_pos + root->sectorsize - 1) >>
 			inode->i_blkbits;
 
 	down_read(&BTRFS_I(inode)->root->snap_sem);
@@ -184,7 +183,7 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 	if (inode->i_size < start_pos) {
 		u64 last_pos_in_file;
 		u64 hole_size;
-		u64 mask = root->blocksize - 1;
+		u64 mask = root->sectorsize - 1;
 		last_pos_in_file = (isize + mask) & ~mask;
 		hole_size = (start_pos - last_pos_in_file + mask) & ~mask;
 
@@ -227,8 +226,8 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 		/* step one, delete the existing extents in this range */
 		/* FIXME blocksize != pagesize */
 		err = btrfs_drop_extents(trans, root, inode, start_pos,
-			 (pos + write_bytes + root->blocksize -1) &
-			 ~((u64)root->blocksize - 1), &hint_block);
+			 (pos + write_bytes + root->sectorsize -1) &
+			 ~((u64)root->sectorsize - 1), &hint_block);
 		if (err)
 			goto failed;
 
@@ -288,7 +287,7 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans,
 {
 	int ret;
 	struct btrfs_key key;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	int slot;
 	struct btrfs_file_extent_item *extent;
 	u64 extent_end = 0;
@@ -327,10 +326,10 @@ next_slot:
 		found_extent = 0;
 		found_inline = 0;
 		extent = NULL;
-		leaf = btrfs_buffer_leaf(path->nodes[0]);
+		leaf = path->nodes[0];
 		slot = path->slots[0];
 		ret = 0;
-		btrfs_disk_key_to_cpu(&key, &leaf->items[slot].key);
+		btrfs_item_key_to_cpu(leaf, &key, slot);
 		if (key.offset >= end || key.objectid != inode->i_ino) {
 			goto out;
 		}
@@ -344,17 +343,18 @@ next_slot:
 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
 			extent = btrfs_item_ptr(leaf, slot,
 						struct btrfs_file_extent_item);
-			found_type = btrfs_file_extent_type(extent);
+			found_type = btrfs_file_extent_type(leaf, extent);
 			if (found_type == BTRFS_FILE_EXTENT_REG) {
 				extent_end = key.offset +
-					(btrfs_file_extent_num_blocks(extent) <<
+				 (btrfs_file_extent_num_blocks(leaf, extent) <<
 					 inode->i_blkbits);
 				found_extent = 1;
 			} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
+				struct btrfs_item *item;
+				item = btrfs_item_nr(leaf, slot);
 				found_inline = 1;
 				extent_end = key.offset +
-				     btrfs_file_extent_inline_len(leaf->items +
-								  slot);
+				     btrfs_file_extent_inline_len(leaf, item);
 			}
 		} else {
 			extent_end = search_start;
@@ -365,8 +365,7 @@ next_slot:
 		    search_start >= extent_end) {
 			int nextret;
 			u32 nritems;
-			nritems = btrfs_header_nritems(
-					btrfs_buffer_header(path->nodes[0]));
+			nritems = btrfs_header_nritems(leaf);
 			if (slot >= nritems - 1) {
 				nextret = btrfs_next_leaf(root, path);
 				if (nextret)
@@ -380,7 +379,7 @@ next_slot:
 
 		/* FIXME, there's only one inline extent allowed right now */
 		if (found_inline) {
-			u64 mask = root->blocksize - 1;
+			u64 mask = root->sectorsize - 1;
 			search_start = (extent_end + mask) & ~mask;
 		} else
 			search_start = extent_end;
@@ -388,10 +387,13 @@ next_slot:
 		if (end < extent_end && end >= key.offset) {
 			if (found_extent) {
 				u64 disk_blocknr =
-					btrfs_file_extent_disk_blocknr(extent);
+				    btrfs_file_extent_disk_blocknr(leaf,extent);
 				u64 disk_num_blocks =
-				      btrfs_file_extent_disk_num_blocks(extent);
-				memcpy(&old, extent, sizeof(old));
+				    btrfs_file_extent_disk_num_blocks(leaf,
+								      extent);
+				read_extent_buffer(leaf, &old,
+						   (unsigned long)extent,
+						   sizeof(old));
 				if (disk_blocknr != 0) {
 					ret = btrfs_inc_extent_ref(trans, root,
 					         disk_blocknr, disk_num_blocks);
@@ -406,20 +408,24 @@ next_slot:
 			u64 new_num;
 			u64 old_num;
 			keep = 1;
-			WARN_ON(start & (root->blocksize - 1));
+			WARN_ON(start & (root->sectorsize - 1));
 			if (found_extent) {
 				new_num = (start - key.offset) >>
 					inode->i_blkbits;
-				old_num = btrfs_file_extent_num_blocks(extent);
+				old_num = btrfs_file_extent_num_blocks(leaf,
+								       extent);
 				*hint_block =
-					btrfs_file_extent_disk_blocknr(extent);
-				if (btrfs_file_extent_disk_blocknr(extent)) {
+					btrfs_file_extent_disk_blocknr(leaf,
+								       extent);
+				if (btrfs_file_extent_disk_blocknr(leaf,
+								   extent)) {
 					inode->i_blocks -=
 						(old_num - new_num) << 3;
 				}
-				btrfs_set_file_extent_num_blocks(extent,
+				btrfs_set_file_extent_num_blocks(leaf,
+								 extent,
 								 new_num);
-				btrfs_mark_buffer_dirty(path->nodes[0]);
+				btrfs_mark_buffer_dirty(leaf);
 			} else {
 				WARN_ON(1);
 			}
@@ -431,13 +437,17 @@ next_slot:
 			u64 extent_num_blocks = 0;
 			if (found_extent) {
 				disk_blocknr =
-				      btrfs_file_extent_disk_blocknr(extent);
+				      btrfs_file_extent_disk_blocknr(leaf,
+								     extent);
 				disk_num_blocks =
-				      btrfs_file_extent_disk_num_blocks(extent);
+				      btrfs_file_extent_disk_num_blocks(leaf,
+									extent);
 				extent_num_blocks =
-				      btrfs_file_extent_num_blocks(extent);
+				      btrfs_file_extent_num_blocks(leaf,
+								   extent);
 				*hint_block =
-					btrfs_file_extent_disk_blocknr(extent);
+					btrfs_file_extent_disk_blocknr(leaf,
+								       extent);
 			}
 			ret = btrfs_del_item(trans, root, path);
 			/* TODO update progress marker and return */
@@ -464,42 +474,37 @@ next_slot:
 			struct btrfs_key ins;
 			ins.objectid = inode->i_ino;
 			ins.offset = end;
-			ins.flags = 0;
 			btrfs_set_key_type(&ins, BTRFS_EXTENT_DATA_KEY);
 			btrfs_release_path(root, path);
 			ret = btrfs_insert_empty_item(trans, root, path, &ins,
 						      sizeof(*extent));
 
+			leaf = path->nodes[0];
 			if (ret) {
-				btrfs_print_leaf(root, btrfs_buffer_leaf(path->nodes[0]));
-				printk("got %d on inserting %Lu %u %Lu start %Lu end %Lu found %Lu %Lu keep was %d\n", ret , ins.objectid, ins.flags, ins.offset, start, end, key.offset, extent_end, keep);
+				btrfs_print_leaf(root, leaf);
+				printk("got %d on inserting %Lu %u %Lu start %Lu end %Lu found %Lu %Lu keep was %d\n", ret , ins.objectid, ins.type, ins.offset, start, end, key.offset, extent_end, keep);
 			}
 			BUG_ON(ret);
-			extent = btrfs_item_ptr(
-				    btrfs_buffer_leaf(path->nodes[0]),
-				    path->slots[0],
-				    struct btrfs_file_extent_item);
-			btrfs_set_file_extent_disk_blocknr(extent,
-				    btrfs_file_extent_disk_blocknr(&old));
-			btrfs_set_file_extent_disk_num_blocks(extent,
-				    btrfs_file_extent_disk_num_blocks(&old));
-
-			btrfs_set_file_extent_offset(extent,
-				    btrfs_file_extent_offset(&old) +
+			extent = btrfs_item_ptr(leaf, path->slots[0],
+						struct btrfs_file_extent_item);
+			write_extent_buffer(leaf, &old,
+					    (unsigned long)extent, sizeof(old));
+
+			btrfs_set_file_extent_offset(leaf, extent,
+				    le64_to_cpu(old.offset) +
 				    ((end - key.offset) >> inode->i_blkbits));
-			WARN_ON(btrfs_file_extent_num_blocks(&old) <
+			WARN_ON(le64_to_cpu(old.num_blocks) <
 				(extent_end - end) >> inode->i_blkbits);
-			btrfs_set_file_extent_num_blocks(extent,
+			btrfs_set_file_extent_num_blocks(leaf, extent,
 				    (extent_end - end) >> inode->i_blkbits);
 
-			btrfs_set_file_extent_type(extent,
+			btrfs_set_file_extent_type(leaf, extent,
 						   BTRFS_FILE_EXTENT_REG);
-			btrfs_set_file_extent_generation(extent,
-				    btrfs_file_extent_generation(&old));
 			btrfs_mark_buffer_dirty(path->nodes[0]);
-			if (btrfs_file_extent_disk_blocknr(&old) != 0) {
+			if (le64_to_cpu(old.disk_blocknr) != 0) {
 				inode->i_blocks +=
-				      btrfs_file_extent_num_blocks(extent) << 3;
+				      btrfs_file_extent_num_blocks(leaf,
+								   extent) << 3;
 			}
 			ret = 0;
 			goto out;
@@ -529,8 +534,8 @@ static int prepare_pages(struct btrfs_root *root,
 	u64 num_blocks;
 	u64 start_pos;
 
-	start_pos = pos & ~((u64)root->blocksize - 1);
-	num_blocks = (write_bytes + pos - start_pos + root->blocksize - 1) >>
+	start_pos = pos & ~((u64)root->sectorsize - 1);
+	num_blocks = (write_bytes + pos - start_pos + root->sectorsize - 1) >>
 			inode->i_blkbits;
 
 	memset(pages, 0, num_pages * sizeof(struct page *));
diff --git a/fs/btrfs/inode-item.c b/fs/btrfs/inode-item.c
index b78346177539..35d2608f8918 100644
--- a/fs/btrfs/inode-item.c
+++ b/fs/btrfs/inode-item.c
@@ -20,24 +20,18 @@
 #include "disk-io.h"
 #include "transaction.h"
 
-int btrfs_insert_inode(struct btrfs_trans_handle *trans, struct btrfs_root
-		       *root, u64 objectid, struct btrfs_inode_item
-		       *inode_item)
+int btrfs_insert_empty_inode(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root,
+			     struct btrfs_path *path, u64 objectid)
 {
-	struct btrfs_path *path;
 	struct btrfs_key key;
 	int ret;
 	key.objectid = objectid;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
 	key.offset = 0;
 
-	path = btrfs_alloc_path();
-	BUG_ON(!path);
-	ret = btrfs_insert_item(trans, root, &key, inode_item,
-				sizeof(*inode_item));
-	btrfs_release_path(root, path);
-	btrfs_free_path(path);
+	ret = btrfs_insert_empty_item(trans, root, path, &key,
+				      sizeof(struct btrfs_inode_item));
 	if (ret == 0 && objectid > root->highest_inode)
 		root->highest_inode = objectid;
 	return ret;
@@ -51,15 +45,15 @@ int btrfs_lookup_inode(struct btrfs_trans_handle *trans, struct btrfs_root
 	int cow = mod != 0;
 	int ret;
 	int slot;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	struct btrfs_key found_key;
 
 	ret = btrfs_search_slot(trans, root, location, path, ins_len, cow);
 	if (ret > 0 && btrfs_key_type(location) == BTRFS_ROOT_ITEM_KEY &&
 	    location->offset == (u64)-1 && path->slots[0] != 0) {
 		slot = path->slots[0] - 1;
-		leaf = btrfs_buffer_leaf(path->nodes[0]);
-		btrfs_disk_key_to_cpu(&found_key, &leaf->items[slot].key);
+		leaf = path->nodes[0];
+		btrfs_item_key_to_cpu(leaf, &found_key, slot);
 		if (found_key.objectid == location->objectid &&
 		    btrfs_key_type(&found_key) == btrfs_key_type(location)) {
 			path->slots[0]--;
diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c
index 405470866254..ab74977adf5c 100644
--- a/fs/btrfs/inode-map.c
+++ b/fs/btrfs/inode-map.c
@@ -24,8 +24,9 @@ int btrfs_find_highest_inode(struct btrfs_root *root, u64 *objectid)
 {
 	struct btrfs_path *path;
 	int ret;
-	struct btrfs_leaf *l;
+	struct extent_buffer *l;
 	struct btrfs_key search_key;
+	struct btrfs_key found_key;
 	int slot;
 
 	path = btrfs_alloc_path();
@@ -39,8 +40,9 @@ int btrfs_find_highest_inode(struct btrfs_root *root, u64 *objectid)
 	BUG_ON(ret == 0);
 	if (path->slots[0] > 0) {
 		slot = path->slots[0] - 1;
-		l = btrfs_buffer_leaf(path->nodes[0]);
-		*objectid = btrfs_disk_key_objectid(&l->items[slot].key);
+		l = path->nodes[0];
+		btrfs_item_key_to_cpu(l, &found_key, slot);
+		*objectid = found_key.objectid;
 	} else {
 		*objectid = BTRFS_FIRST_FREE_OBJECTID;
 	}
@@ -64,13 +66,12 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans,
 	int slot = 0;
 	u64 last_ino = 0;
 	int start_found;
-	struct btrfs_leaf *l;
+	struct extent_buffer *l;
 	struct btrfs_key search_key;
 	u64 search_start = dirid;
 
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
-	search_key.flags = 0;
 	search_start = root->last_inode_alloc;
 	search_start = max(search_start, BTRFS_FIRST_FREE_OBJECTID);
 	search_key.objectid = search_start;
@@ -86,9 +87,9 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans,
 		path->slots[0]--;
 
 	while (1) {
-		l = btrfs_buffer_leaf(path->nodes[0]);
+		l = path->nodes[0];
 		slot = path->slots[0];
-		if (slot >= btrfs_header_nritems(&l->header)) {
+		if (slot >= btrfs_header_nritems(l)) {
 			ret = btrfs_next_leaf(root, path);
 			if (ret == 0)
 				continue;
@@ -103,7 +104,7 @@ int btrfs_find_free_objectid(struct btrfs_trans_handle *trans,
 				last_ino : search_start;
 			goto found;
 		}
-		btrfs_disk_key_to_cpu(&key, &l->items[slot].key);
+		btrfs_item_key_to_cpu(l, &key, slot);
 		if (key.objectid >= search_start) {
 			if (start_found) {
 				if (last_ino < search_start)
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index b03d40a907ca..fbe2836364e0 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -159,10 +159,8 @@ out:
 
 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end)
 {
-	char csum[BTRFS_CRC32_SIZE];
 	size_t offset = start - (page->index << PAGE_CACHE_SHIFT);
 	struct inode *inode = page->mapping->host;
-	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 	char *kaddr;
 	u64 private;
@@ -173,11 +171,15 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end)
 	if (ret) {
 		goto zeroit;
 	}
+	/*
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	char csum[BTRFS_CRC32_SIZE];
 	ret = btrfs_csum_data(root, kaddr + offset, end - start + 1, csum);
 	BUG_ON(ret);
 	if (memcmp(csum, &private, BTRFS_CRC32_SIZE)) {
 		goto zeroit;
 	}
+	*/
 	kunmap_atomic(kaddr, KM_IRQ0);
 	return 0;
 
@@ -192,7 +194,9 @@ zeroit:
 void btrfs_read_locked_inode(struct inode *inode)
 {
 	struct btrfs_path *path;
+	struct extent_buffer *leaf;
 	struct btrfs_inode_item *inode_item;
+	struct btrfs_inode_timespec *tspec;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_key location;
 	u64 alloc_group_block;
@@ -205,29 +209,37 @@ void btrfs_read_locked_inode(struct inode *inode)
 
 	memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
 	ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
-	if (ret) {
+	if (ret)
 		goto make_bad;
-	}
-	inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
-				  path->slots[0],
-				  struct btrfs_inode_item);
 
-	inode->i_mode = btrfs_inode_mode(inode_item);
-	inode->i_nlink = btrfs_inode_nlink(inode_item);
-	inode->i_uid = btrfs_inode_uid(inode_item);
-	inode->i_gid = btrfs_inode_gid(inode_item);
-	inode->i_size = btrfs_inode_size(inode_item);
-	inode->i_atime.tv_sec = btrfs_timespec_sec(&inode_item->atime);
-	inode->i_atime.tv_nsec = btrfs_timespec_nsec(&inode_item->atime);
-	inode->i_mtime.tv_sec = btrfs_timespec_sec(&inode_item->mtime);
-	inode->i_mtime.tv_nsec = btrfs_timespec_nsec(&inode_item->mtime);
-	inode->i_ctime.tv_sec = btrfs_timespec_sec(&inode_item->ctime);
-	inode->i_ctime.tv_nsec = btrfs_timespec_nsec(&inode_item->ctime);
-	inode->i_blocks = btrfs_inode_nblocks(inode_item);
-	inode->i_generation = btrfs_inode_generation(inode_item);
+	leaf = path->nodes[0];
+	inode_item = btrfs_item_ptr(leaf, path->slots[0],
+				    struct btrfs_inode_item);
+
+	inode->i_mode = btrfs_inode_mode(leaf, inode_item);
+	inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
+	inode->i_uid = btrfs_inode_uid(leaf, inode_item);
+	inode->i_gid = btrfs_inode_gid(leaf, inode_item);
+	inode->i_size = btrfs_inode_size(leaf, inode_item);
+
+	tspec = btrfs_inode_atime(inode_item);
+	inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
+	inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
+
+	tspec = btrfs_inode_mtime(inode_item);
+	inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
+	inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
+
+	tspec = btrfs_inode_ctime(inode_item);
+	inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
+	inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
+
+	inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
+	inode->i_generation = btrfs_inode_generation(leaf, inode_item);
 	inode->i_rdev = 0;
-	rdev = btrfs_inode_rdev(inode_item);
-	alloc_group_block = btrfs_inode_block_group(inode_item);
+	rdev = btrfs_inode_rdev(leaf, inode_item);
+
+	alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
 	BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
 						       alloc_group_block);
 
@@ -267,24 +279,35 @@ make_bad:
 	make_bad_inode(inode);
 }
 
-static void fill_inode_item(struct btrfs_inode_item *item,
+static void fill_inode_item(struct extent_buffer *leaf,
+			    struct btrfs_inode_item *item,
 			    struct inode *inode)
 {
-	btrfs_set_inode_uid(item, inode->i_uid);
-	btrfs_set_inode_gid(item, inode->i_gid);
-	btrfs_set_inode_size(item, inode->i_size);
-	btrfs_set_inode_mode(item, inode->i_mode);
-	btrfs_set_inode_nlink(item, inode->i_nlink);
-	btrfs_set_timespec_sec(&item->atime, inode->i_atime.tv_sec);
-	btrfs_set_timespec_nsec(&item->atime, inode->i_atime.tv_nsec);
-	btrfs_set_timespec_sec(&item->mtime, inode->i_mtime.tv_sec);
-	btrfs_set_timespec_nsec(&item->mtime, inode->i_mtime.tv_nsec);
-	btrfs_set_timespec_sec(&item->ctime, inode->i_ctime.tv_sec);
-	btrfs_set_timespec_nsec(&item->ctime, inode->i_ctime.tv_nsec);
-	btrfs_set_inode_nblocks(item, inode->i_blocks);
-	btrfs_set_inode_generation(item, inode->i_generation);
-	btrfs_set_inode_rdev(item, inode->i_rdev);
-	btrfs_set_inode_block_group(item,
+	btrfs_set_inode_uid(leaf, item, inode->i_uid);
+	btrfs_set_inode_gid(leaf, item, inode->i_gid);
+	btrfs_set_inode_size(leaf, item, inode->i_size);
+	btrfs_set_inode_mode(leaf, item, inode->i_mode);
+	btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
+
+	btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
+			       inode->i_atime.tv_sec);
+	btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
+				inode->i_atime.tv_nsec);
+
+	btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
+			       inode->i_mtime.tv_sec);
+	btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
+				inode->i_mtime.tv_nsec);
+
+	btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
+			       inode->i_ctime.tv_sec);
+	btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
+				inode->i_ctime.tv_nsec);
+
+	btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
+	btrfs_set_inode_generation(leaf, item, inode->i_generation);
+	btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
+	btrfs_set_inode_block_group(leaf, item,
 				    BTRFS_I(inode)->block_group->key.objectid);
 }
 
@@ -294,6 +317,7 @@ int btrfs_update_inode(struct btrfs_trans_handle *trans,
 {
 	struct btrfs_inode_item *inode_item;
 	struct btrfs_path *path;
+	struct extent_buffer *leaf;
 	int ret;
 
 	path = btrfs_alloc_path();
@@ -306,12 +330,12 @@ int btrfs_update_inode(struct btrfs_trans_handle *trans,
 		goto failed;
 	}
 
-	inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
-				  path->slots[0],
+	leaf = path->nodes[0];
+	inode_item = btrfs_item_ptr(leaf, path->slots[0],
 				  struct btrfs_inode_item);
 
-	fill_inode_item(inode_item, inode);
-	btrfs_mark_buffer_dirty(path->nodes[0]);
+	fill_inode_item(leaf, inode_item, inode);
+	btrfs_mark_buffer_dirty(leaf);
 	btrfs_set_inode_last_trans(trans, inode);
 	ret = 0;
 failed:
@@ -330,8 +354,9 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
 	const char *name = dentry->d_name.name;
 	int name_len = dentry->d_name.len;
 	int ret = 0;
-	u64 objectid;
+	struct extent_buffer *leaf;
 	struct btrfs_dir_item *di;
+	struct btrfs_key key;
 
 	path = btrfs_alloc_path();
 	if (!path) {
@@ -349,14 +374,15 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
 		ret = -ENOENT;
 		goto err;
 	}
-	objectid = btrfs_disk_key_objectid(&di->location);
+	leaf = path->nodes[0];
+	btrfs_dir_item_key_to_cpu(leaf, di, &key);
 	ret = btrfs_delete_one_dir_name(trans, root, path, di);
 	if (ret)
 		goto err;
 	btrfs_release_path(root, path);
 
 	di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
-					 objectid, name, name_len, -1);
+					 key.objectid, name, name_len, -1);
 	if (IS_ERR(di)) {
 		ret = PTR_ERR(di);
 		goto err;
@@ -391,12 +417,15 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
 	root = BTRFS_I(dir)->root;
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
+
 	btrfs_set_trans_block_group(trans, dir);
 	ret = btrfs_unlink_trans(trans, root, dir, dentry);
 	nr = trans->blocks_used;
+
 	btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
 	btrfs_btree_balance_dirty(root, nr);
+
 	return ret;
 }
 
@@ -411,7 +440,7 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
 	struct btrfs_trans_handle *trans;
 	struct btrfs_key found_key;
 	int found_type;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	char *goodnames = "..";
 	unsigned long nr;
 
@@ -419,10 +448,11 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
 	BUG_ON(!path);
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
+
 	btrfs_set_trans_block_group(trans, dir);
 	key.objectid = inode->i_ino;
 	key.offset = (u64)-1;
-	key.flags = (u32)-1;
+	key.type = (u8)-1;
 	while(1) {
 		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 		if (ret < 0) {
@@ -435,9 +465,8 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
 			goto out;
 		}
 		path->slots[0]--;
-		leaf = btrfs_buffer_leaf(path->nodes[0]);
-		btrfs_disk_key_to_cpu(&found_key,
-				      &leaf->items[path->slots[0]].key);
+		leaf = path->nodes[0];
+		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
 		found_type = btrfs_key_type(&found_key);
 		if (found_key.objectid != inode->i_ino) {
 			err = -ENOENT;
@@ -513,9 +542,9 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 	int ret;
 	struct btrfs_path *path;
 	struct btrfs_key key;
-	struct btrfs_disk_key *found_key;
+	struct btrfs_key found_key;
 	u32 found_type;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	struct btrfs_file_extent_item *fi;
 	u64 extent_start = 0;
 	u64 extent_num_blocks = 0;
@@ -527,10 +556,12 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 	path = btrfs_alloc_path();
 	path->reada = -1;
 	BUG_ON(!path);
+
 	/* FIXME, add redo link to tree so we don't leak on crash */
 	key.objectid = inode->i_ino;
 	key.offset = (u64)-1;
-	key.flags = (u32)-1;
+	key.type = (u8)-1;
+
 	while(1) {
 		btrfs_init_path(path);
 		fi = NULL;
@@ -542,27 +573,28 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 			BUG_ON(path->slots[0] == 0);
 			path->slots[0]--;
 		}
-		leaf = btrfs_buffer_leaf(path->nodes[0]);
-		found_key = &leaf->items[path->slots[0]].key;
-		found_type = btrfs_disk_key_type(found_key);
+		leaf = path->nodes[0];
+		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+		found_type = btrfs_key_type(&found_key);
 
-		if (btrfs_disk_key_objectid(found_key) != inode->i_ino)
+		if (found_key.objectid != inode->i_ino)
 			break;
+
 		if (found_type != BTRFS_CSUM_ITEM_KEY &&
 		    found_type != BTRFS_DIR_ITEM_KEY &&
 		    found_type != BTRFS_DIR_INDEX_KEY &&
 		    found_type != BTRFS_EXTENT_DATA_KEY)
 			break;
 
-		item_end = btrfs_disk_key_offset(found_key);
+		item_end = found_key.offset;
 		if (found_type == BTRFS_EXTENT_DATA_KEY) {
-			fi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
-					    path->slots[0],
+			fi = btrfs_item_ptr(leaf, path->slots[0],
 					    struct btrfs_file_extent_item);
-			if (btrfs_file_extent_type(fi) !=
+			if (btrfs_file_extent_type(leaf, fi) !=
 			    BTRFS_FILE_EXTENT_INLINE) {
-				item_end += btrfs_file_extent_num_blocks(fi) <<
-						inode->i_blkbits;
+				item_end +=
+				    btrfs_file_extent_num_blocks(leaf, fi) <<
+				    inode->i_blkbits;
 			}
 		}
 		if (found_type == BTRFS_CSUM_ITEM_KEY) {
@@ -583,7 +615,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 			btrfs_set_key_type(&key, found_type);
 			continue;
 		}
-		if (btrfs_disk_key_offset(found_key) >= inode->i_size)
+		if (found_key.offset >= inode->i_size)
 			del_item = 1;
 		else
 			del_item = 0;
@@ -591,30 +623,31 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 
 		/* FIXME, shrink the extent if the ref count is only 1 */
 		if (found_type == BTRFS_EXTENT_DATA_KEY &&
-			   btrfs_file_extent_type(fi) !=
+			   btrfs_file_extent_type(leaf, fi) !=
 			   BTRFS_FILE_EXTENT_INLINE) {
 			u64 num_dec;
-			extent_start = btrfs_file_extent_disk_blocknr(fi);
+			extent_start = btrfs_file_extent_disk_blocknr(leaf, fi);
 			if (!del_item) {
 				u64 orig_num_blocks =
-					btrfs_file_extent_num_blocks(fi);
+					btrfs_file_extent_num_blocks(leaf, fi);
 				extent_num_blocks = inode->i_size -
-					btrfs_disk_key_offset(found_key) +
-					root->blocksize - 1;
+					found_key.offset + root->sectorsize - 1;
 				extent_num_blocks >>= inode->i_blkbits;
-				btrfs_set_file_extent_num_blocks(fi,
+				btrfs_set_file_extent_num_blocks(leaf, fi,
 							 extent_num_blocks);
 				num_dec = (orig_num_blocks -
 					   extent_num_blocks) << 3;
 				if (extent_start != 0) {
 					inode->i_blocks -= num_dec;
 				}
-				btrfs_mark_buffer_dirty(path->nodes[0]);
+				btrfs_mark_buffer_dirty(leaf);
 			} else {
 				extent_num_blocks =
-					btrfs_file_extent_disk_num_blocks(fi);
+					btrfs_file_extent_disk_num_blocks(leaf,
+									  fi);
 				/* FIXME blocksize != 4096 */
-				num_dec = btrfs_file_extent_num_blocks(fi) << 3;
+				num_dec = btrfs_file_extent_num_blocks(leaf,
+								       fi) << 3;
 				if (extent_start != 0) {
 					found_extent = 1;
 					inode->i_blocks -= num_dec;
@@ -725,7 +758,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
 		struct btrfs_root *root = BTRFS_I(inode)->root;
 		struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 
-		u64 mask = root->blocksize - 1;
+		u64 mask = root->sectorsize - 1;
 		u64 pos = (inode->i_size + mask) & ~mask;
 		u64 block_end = attr->ia_size | mask;
 		u64 hole_size;
@@ -771,9 +804,11 @@ void btrfs_delete_inode(struct inode *inode)
 	if (is_bad_inode(inode)) {
 		goto no_delete;
 	}
+
 	inode->i_size = 0;
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
+
 	btrfs_set_trans_block_group(trans, inode);
 	ret = btrfs_truncate_in_trans(trans, root, inode);
 	if (ret)
@@ -782,6 +817,7 @@ void btrfs_delete_inode(struct inode *inode)
 	if (ret)
 		goto no_delete_lock;
 	nr = trans->blocks_used;
+
 	btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
 	btrfs_btree_balance_dirty(root, nr);
@@ -819,7 +855,7 @@ static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
 		ret = 0;
 		goto out;
 	}
-	btrfs_disk_key_to_cpu(location, &di->location);
+	btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
 out:
 	btrfs_release_path(root, path);
 	btrfs_free_path(path);
@@ -856,7 +892,6 @@ static int fixup_tree_root_location(struct btrfs_root *root,
 
 	ri = &(*sub_root)->root_item;
 	location->objectid = btrfs_root_dirid(ri);
-	location->flags = 0;
 	btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
 	location->offset = 0;
 
@@ -908,11 +943,14 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
 
 	if (dentry->d_name.len > BTRFS_NAME_LEN)
 		return ERR_PTR(-ENAMETOOLONG);
+
 	mutex_lock(&root->fs_info->fs_mutex);
 	ret = btrfs_inode_by_name(dir, dentry, &location);
 	mutex_unlock(&root->fs_info->fs_mutex);
+
 	if (ret < 0)
 		return ERR_PTR(ret);
+
 	inode = NULL;
 	if (location.objectid) {
 		ret = fixup_tree_root_location(root, &location, &sub_root,
@@ -952,10 +990,11 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
 	struct btrfs_item *item;
 	struct btrfs_dir_item *di;
 	struct btrfs_key key;
+	struct btrfs_key found_key;
 	struct btrfs_path *path;
 	int ret;
 	u32 nritems;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	int slot;
 	int advance;
 	unsigned char d_type;
@@ -964,15 +1003,19 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
 	u32 di_total;
 	u32 di_len;
 	int key_type = BTRFS_DIR_INDEX_KEY;
+	char tmp_name[32];
+	char *name_ptr;
+	int name_len;
 
 	/* FIXME, use a real flag for deciding about the key type */
 	if (root->fs_info->tree_root == root)
 		key_type = BTRFS_DIR_ITEM_KEY;
+
 	mutex_lock(&root->fs_info->fs_mutex);
 	key.objectid = inode->i_ino;
-	key.flags = 0;
 	btrfs_set_key_type(&key, key_type);
 	key.offset = filp->f_pos;
+
 	path = btrfs_alloc_path();
 	path->reada = 2;
 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
@@ -980,16 +1023,16 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
 		goto err;
 	advance = 0;
 	while(1) {
-		leaf = btrfs_buffer_leaf(path->nodes[0]);
-		nritems = btrfs_header_nritems(&leaf->header);
+		leaf = path->nodes[0];
+		nritems = btrfs_header_nritems(leaf);
 		slot = path->slots[0];
 		if (advance || slot >= nritems) {
 			if (slot >= nritems -1) {
 				ret = btrfs_next_leaf(root, path);
 				if (ret)
 					break;
-				leaf = btrfs_buffer_leaf(path->nodes[0]);
-				nritems = btrfs_header_nritems(&leaf->header);
+				leaf = path->nodes[0];
+				nritems = btrfs_header_nritems(leaf);
 				slot = path->slots[0];
 			} else {
 				slot++;
@@ -997,28 +1040,48 @@ static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
 			}
 		}
 		advance = 1;
-		item = leaf->items + slot;
-		if (btrfs_disk_key_objectid(&item->key) != key.objectid)
+		item = btrfs_item_nr(leaf, slot);
+		btrfs_item_key_to_cpu(leaf, &found_key, slot);
+
+		if (found_key.objectid != key.objectid)
 			break;
-		if (btrfs_disk_key_type(&item->key) != key_type)
+		if (btrfs_key_type(&found_key) != key_type)
 			break;
-		if (btrfs_disk_key_offset(&item->key) < filp->f_pos)
+		if (found_key.offset < filp->f_pos)
 			continue;
-		filp->f_pos = btrfs_disk_key_offset(&item->key);
+
+		filp->f_pos = found_key.offset;
 		advance = 1;
 		di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
 		di_cur = 0;
-		di_total = btrfs_item_size(leaf->items + slot);
+		di_total = btrfs_item_size(leaf, item);
 		while(di_cur < di_total) {
-			d_type = btrfs_filetype_table[btrfs_dir_type(di)];
-			over = filldir(dirent, (const char *)(di + 1),
-				       btrfs_dir_name_len(di),
-				       btrfs_disk_key_offset(&item->key),
-				       btrfs_disk_key_objectid(&di->location),
+			struct btrfs_key location;
+
+			name_len = btrfs_dir_name_len(leaf, di);
+			if (name_len < 32) {
+				name_ptr = tmp_name;
+			} else {
+				name_ptr = kmalloc(name_len, GFP_NOFS);
+				BUG_ON(!name_ptr);
+			}
+			read_extent_buffer(leaf, name_ptr,
+					   (unsigned long)(di + 1), name_len);
+
+			d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
+			btrfs_dir_item_key_to_cpu(leaf, di, &location);
+
+			over = filldir(dirent, name_ptr, name_len,
+				       found_key.offset,
+				       location.objectid,
 				       d_type);
+
+			if (name_ptr != tmp_name)
+				kfree(name_ptr);
+
 			if (over)
 				goto nopos;
-			di_len = btrfs_dir_name_len(di) + sizeof(*di);
+			di_len = btrfs_dir_name_len(leaf, di) + sizeof(*di);
 			di_cur += di_len;
 			di = (struct btrfs_dir_item *)((char *)di + di_len);
 		}
@@ -1075,11 +1138,15 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 				     int mode)
 {
 	struct inode *inode;
-	struct btrfs_inode_item inode_item;
+	struct btrfs_inode_item *inode_item;
 	struct btrfs_key *location;
+	struct btrfs_path *path;
 	int ret;
 	int owner;
 
+	path = btrfs_alloc_path();
+	BUG_ON(!path);
+
 	inode = new_inode(root->fs_info->sb);
 	if (!inode)
 		return ERR_PTR(-ENOMEM);
@@ -1095,24 +1162,32 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 	group = btrfs_find_block_group(root, group, 0, 0, owner);
 	BTRFS_I(inode)->block_group = group;
 
+	ret = btrfs_insert_empty_inode(trans, root, path, objectid);
+	if (ret)
+		goto fail;
+
 	inode->i_uid = current->fsuid;
 	inode->i_gid = current->fsgid;
 	inode->i_mode = mode;
 	inode->i_ino = objectid;
 	inode->i_blocks = 0;
 	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
-	fill_inode_item(&inode_item, inode);
+	inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+				  struct btrfs_inode_item);
+	fill_inode_item(path->nodes[0], inode_item, inode);
+	btrfs_mark_buffer_dirty(path->nodes[0]);
+	btrfs_free_path(path);
+
 	location = &BTRFS_I(inode)->location;
 	location->objectid = objectid;
-	location->flags = 0;
 	location->offset = 0;
 	btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
 
-	ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
-	if (ret)
-		return ERR_PTR(ret);
 	insert_inode_hash(inode);
 	return inode;
+fail:
+	btrfs_free_path(path);
+	return ERR_PTR(ret);
 }
 
 static inline u8 btrfs_inode_type(struct inode *inode)
@@ -1127,8 +1202,8 @@ static int btrfs_add_link(struct btrfs_trans_handle *trans,
 	struct btrfs_key key;
 	struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
 	struct inode *parent_inode;
+
 	key.objectid = inode->i_ino;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
 	key.offset = 0;
 
@@ -1285,14 +1360,18 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
 	inc_nlink(inode);
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
+
 	btrfs_set_trans_block_group(trans, dir);
 	atomic_inc(&inode->i_count);
 	err = btrfs_add_nondir(trans, dentry, inode);
+
 	if (err)
 		drop_inode = 1;
+
 	dir->i_sb->s_dirt = 1;
 	btrfs_update_inode_block_group(trans, dir);
 	err = btrfs_update_inode(trans, root, inode);
+
 	if (err)
 		drop_inode = 1;
 
@@ -1321,13 +1400,13 @@ static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans,
 
 	key.objectid = objectid;
 	key.offset = 0;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
 
 	ret = btrfs_insert_dir_item(trans, root, buf, 1, objectid,
 				    &key, BTRFS_FT_DIR);
 	if (ret)
 		goto error;
+
 	key.objectid = dirid;
 	ret = btrfs_insert_dir_item(trans, root, buf, 2, objectid,
 				    &key, BTRFS_FT_DIR);
@@ -1350,6 +1429,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
 	btrfs_set_trans_block_group(trans, dir);
+
 	if (IS_ERR(trans)) {
 		err = PTR_ERR(trans);
 		goto out_unlock;
@@ -1367,6 +1447,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 		err = PTR_ERR(inode);
 		goto out_fail;
 	}
+
 	drop_on_err = 1;
 	inode->i_op = &btrfs_dir_inode_operations;
 	inode->i_fop = &btrfs_dir_file_operations;
@@ -1380,9 +1461,11 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 	err = btrfs_update_inode(trans, root, inode);
 	if (err)
 		goto out_fail;
+
 	err = btrfs_add_link(trans, dentry, inode);
 	if (err)
 		goto out_fail;
+
 	d_instantiate(dentry, inode);
 	drop_on_err = 0;
 	dir->i_sb->s_dirt = 1;
@@ -1392,6 +1475,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 out_fail:
 	nr = trans->blocks_used;
 	btrfs_end_transaction(trans, root);
+
 out_unlock:
 	mutex_unlock(&root->fs_info->fs_mutex);
 	if (drop_on_err)
@@ -1415,8 +1499,8 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
 	struct btrfs_path *path;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_file_extent_item *item;
-	struct btrfs_leaf *leaf;
-	struct btrfs_disk_key *found_key;
+	struct extent_buffer *leaf;
+	struct btrfs_key found_key;
 	struct extent_map *em = NULL;
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 	struct btrfs_trans_handle *trans = NULL;
@@ -1436,8 +1520,8 @@ again:
 			err = -ENOMEM;
 			goto out;
 		}
-		em->start = 0;
-		em->end = 0;
+		em->start = EXTENT_MAP_HOLE;
+		em->end = EXTENT_MAP_HOLE;
 	}
 	em->bdev = inode->i_sb->s_bdev;
 	ret = btrfs_lookup_file_extent(NULL, root, path,
@@ -1453,25 +1537,27 @@ again:
 		path->slots[0]--;
 	}
 
-	item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
+	leaf = path->nodes[0];
+	item = btrfs_item_ptr(leaf, path->slots[0],
 			      struct btrfs_file_extent_item);
-	leaf = btrfs_buffer_leaf(path->nodes[0]);
-	blocknr = btrfs_file_extent_disk_blocknr(item);
-	blocknr += btrfs_file_extent_offset(item);
+
+	blocknr = btrfs_file_extent_disk_blocknr(leaf, item);
+	blocknr += btrfs_file_extent_offset(leaf, item);
 
 	/* are we inside the extent that was found? */
-	found_key = &leaf->items[path->slots[0]].key;
-	found_type = btrfs_disk_key_type(found_key);
-	if (btrfs_disk_key_objectid(found_key) != objectid ||
+	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+	found_type = btrfs_key_type(&found_key);
+	if (found_key.objectid != objectid ||
 	    found_type != BTRFS_EXTENT_DATA_KEY) {
 		goto not_found;
 	}
 
-	found_type = btrfs_file_extent_type(item);
-	extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key);
+	found_type = btrfs_file_extent_type(leaf, item);
+	extent_start = found_key.offset;
 	if (found_type == BTRFS_FILE_EXTENT_REG) {
 		extent_end = extent_start +
-		       (btrfs_file_extent_num_blocks(item) << inode->i_blkbits);
+		       (btrfs_file_extent_num_blocks(leaf, item) <<
+			inode->i_blkbits);
 		err = 0;
 		if (start < extent_start || start >= extent_end) {
 			em->start = start;
@@ -1484,28 +1570,29 @@ again:
 			}
 			goto not_found_em;
 		}
-		if (btrfs_file_extent_disk_blocknr(item) == 0) {
+		if (btrfs_file_extent_disk_blocknr(leaf, item) == 0) {
 			em->start = extent_start;
 			em->end = extent_end - 1;
-			em->block_start = 0;
-			em->block_end = 0;
+			em->block_start = EXTENT_MAP_HOLE;
+			em->block_end = EXTENT_MAP_HOLE;
 			goto insert;
 		}
 		em->block_start = blocknr << inode->i_blkbits;
 		em->block_end = em->block_start +
-			(btrfs_file_extent_num_blocks(item) <<
+			(btrfs_file_extent_num_blocks(leaf, item) <<
 			 inode->i_blkbits) - 1;
 		em->start = extent_start;
 		em->end = extent_end - 1;
 		goto insert;
 	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
-		char *ptr;
+		unsigned long ptr;
 		char *map;
 		u32 size;
 
-		size = btrfs_file_extent_inline_len(leaf->items +
-						    path->slots[0]);
-		extent_end = extent_start | ((u64)root->blocksize - 1);
+		size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
+						    path->slots[0]));
+
+		extent_end = extent_start | ((u64)root->sectorsize - 1);
 		if (start < extent_start || start >= extent_end) {
 			em->start = start;
 			if (start < extent_start) {
@@ -1517,18 +1604,21 @@ again:
 			}
 			goto not_found_em;
 		}
+
 		em->block_start = EXTENT_MAP_INLINE;
 		em->block_end = EXTENT_MAP_INLINE;
 		em->start = extent_start;
 		em->end = extent_end;
+
 		if (!page) {
 			goto insert;
 		}
+
 		ptr = btrfs_file_extent_inline_start(item);
 		map = kmap(page);
-		memcpy(map + page_offset, ptr, size);
+		read_extent_buffer(leaf, map + page_offset, ptr, size);
 		memset(map + page_offset + size, 0,
-		       root->blocksize - (page_offset + size));
+		       root->sectorsize - (page_offset + size));
 		flush_dcache_page(page);
 		kunmap(page);
 		set_extent_uptodate(em_tree, extent_start,
@@ -1542,8 +1632,8 @@ not_found:
 	em->start = start;
 	em->end = end;
 not_found_em:
-	em->block_start = 0;
-	em->block_end = 0;
+	em->block_start = EXTENT_MAP_HOLE;
+	em->block_end = EXTENT_MAP_HOLE;
 insert:
 	btrfs_release_path(root, path);
 	if (em->start > start || em->end < start) {
@@ -1712,6 +1802,7 @@ static void btrfs_truncate(struct inode *inode)
 	ret = btrfs_truncate_in_trans(trans, root, inode);
 	btrfs_update_inode(trans, root, inode);
 	nr = trans->blocks_used;
+
 	ret = btrfs_end_transaction(trans, root);
 	BUG_ON(ret);
 	mutex_unlock(&root->fs_info->fs_mutex);
@@ -1731,8 +1822,7 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 	struct btrfs_key key;
 	struct btrfs_root_item root_item;
 	struct btrfs_inode_item *inode_item;
-	struct buffer_head *subvol;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	struct btrfs_root *new_root;
 	struct inode *inode;
 	struct inode *dir;
@@ -1746,34 +1836,37 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 	trans = btrfs_start_transaction(root, 1);
 	BUG_ON(!trans);
 
-	subvol = btrfs_alloc_free_block(trans, root, 0, 0);
-	if (IS_ERR(subvol))
-		return PTR_ERR(subvol);
-	leaf = btrfs_buffer_leaf(subvol);
-	btrfs_set_header_nritems(&leaf->header, 0);
-	btrfs_set_header_level(&leaf->header, 0);
-	btrfs_set_header_blocknr(&leaf->header, bh_blocknr(subvol));
-	btrfs_set_header_generation(&leaf->header, trans->transid);
-	btrfs_set_header_owner(&leaf->header, root->root_key.objectid);
-	memcpy(leaf->header.fsid, root->fs_info->disk_super->fsid,
-	       sizeof(leaf->header.fsid));
-	btrfs_mark_buffer_dirty(subvol);
+	leaf = btrfs_alloc_free_block(trans, root, 0, 0);
+	if (IS_ERR(leaf))
+		return PTR_ERR(leaf);
+
+	btrfs_set_header_nritems(leaf, 0);
+	btrfs_set_header_level(leaf, 0);
+	btrfs_set_header_blocknr(leaf, extent_buffer_blocknr(leaf));
+	btrfs_set_header_generation(leaf, trans->transid);
+	btrfs_set_header_owner(leaf, root->root_key.objectid);
+	write_extent_buffer(leaf, root->fs_info->fsid,
+			    (unsigned long)btrfs_header_fsid(leaf),
+			    BTRFS_FSID_SIZE);
+	btrfs_mark_buffer_dirty(leaf);
 
 	inode_item = &root_item.inode;
 	memset(inode_item, 0, sizeof(*inode_item));
-	btrfs_set_inode_generation(inode_item, 1);
-	btrfs_set_inode_size(inode_item, 3);
-	btrfs_set_inode_nlink(inode_item, 1);
-	btrfs_set_inode_nblocks(inode_item, 1);
-	btrfs_set_inode_mode(inode_item, S_IFDIR | 0755);
+	inode_item->generation = cpu_to_le64(1);
+	inode_item->size = cpu_to_le64(3);
+	inode_item->nlink = cpu_to_le32(1);
+	inode_item->nblocks = cpu_to_le64(1);
+	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
 
-	btrfs_set_root_blocknr(&root_item, bh_blocknr(subvol));
+	btrfs_set_root_blocknr(&root_item, extent_buffer_blocknr(leaf));
 	btrfs_set_root_refs(&root_item, 1);
-	btrfs_set_root_blocks_used(&root_item, 0);
+	btrfs_set_root_used(&root_item, 0);
+
 	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
 	root_item.drop_level = 0;
-	brelse(subvol);
-	subvol = NULL;
+
+	free_extent_buffer(leaf);
+	leaf = NULL;
 
 	ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
 				       0, &objectid);
@@ -1784,7 +1877,6 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 
 	key.objectid = objectid;
 	key.offset = 1;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
 				&root_item);
@@ -1845,7 +1937,7 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 	struct btrfs_trans_handle *trans;
 	struct btrfs_key key;
 	struct btrfs_root_item new_root_item;
-	struct buffer_head *tmp;
+	struct extent_buffer *tmp;
 	int ret;
 	int err;
 	u64 objectid;
@@ -1876,10 +1968,11 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 
 	key.objectid = objectid;
 	key.offset = 1;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
+
 	btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
-	btrfs_set_root_blocknr(&new_root_item, bh_blocknr(root->node));
+	btrfs_set_root_blocknr(&new_root_item,
+			       extent_buffer_blocknr(root->node));
 
 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
 				&new_root_item);
@@ -1904,8 +1997,10 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 fail:
 	nr = trans->blocks_used;
 	err = btrfs_commit_transaction(trans, root);
+
 	if (err && !ret)
 		ret = err;
+
 	mutex_unlock(&root->fs_info->fs_mutex);
 	up_write(&root->snap_sem);
 	btrfs_btree_balance_dirty(root, nr);
@@ -1986,7 +2081,7 @@ static int btrfs_ioctl_snap_create(struct btrfs_root *root, void __user *arg)
 
 	if (copy_from_user(&vol_args, arg, sizeof(vol_args)))
 		return -EFAULT;
-	
+
 	namelen = strlen(vol_args.name);
 	if (namelen > BTRFS_VOL_NAME_MAX)
 		return -EINVAL;
@@ -2164,8 +2259,10 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
 	    new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
 		return -ENOTEMPTY;
 	}
+
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
+
 	btrfs_set_trans_block_group(trans, new_dir);
 	path = btrfs_alloc_path();
 	if (!path) {
@@ -2177,9 +2274,10 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
 	old_dir->i_ctime = old_dir->i_mtime = ctime;
 	new_dir->i_ctime = new_dir->i_mtime = ctime;
 	old_inode->i_ctime = ctime;
+
 	if (S_ISDIR(old_inode->i_mode) && old_dir != new_dir) {
 		struct btrfs_key *location = &BTRFS_I(new_dir)->location;
-		u64 old_parent_oid;
+		struct btrfs_key old_parent_key;
 		di = btrfs_lookup_dir_item(trans, root, path, old_inode->i_ino,
 					   "..", 2, -1);
 		if (IS_ERR(di)) {
@@ -2190,7 +2288,7 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
 			ret = -ENOENT;
 			goto out_fail;
 		}
-		old_parent_oid = btrfs_disk_key_objectid(&di->location);
+		btrfs_dir_item_key_to_cpu(path->nodes[0], di, &old_parent_key);
 		ret = btrfs_del_item(trans, root, path);
 		if (ret) {
 			goto out_fail;
@@ -2199,7 +2297,7 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
 
 		di = btrfs_lookup_dir_index_item(trans, root, path,
 						 old_inode->i_ino,
-						 old_parent_oid,
+						 old_parent_key.objectid,
 						 "..", 2, -1);
 		if (IS_ERR(di)) {
 			ret = PTR_ERR(di);
@@ -2257,8 +2355,9 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
 	u64 objectid;
 	int name_len;
 	int datasize;
-	char *ptr;
+	unsigned long ptr;
 	struct btrfs_file_extent_item *ei;
+	struct extent_buffer *leaf;
 	unsigned long nr;
 
 	name_len = strlen(symname) + 1;
@@ -2302,7 +2401,6 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
 	BUG_ON(!path);
 	key.objectid = inode->i_ino;
 	key.offset = 0;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
 	datasize = btrfs_file_extent_calc_inline_size(name_len);
 	err = btrfs_insert_empty_item(trans, root, path, &key,
@@ -2311,16 +2409,17 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
 		drop_inode = 1;
 		goto out_unlock;
 	}
-	ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
-	       path->slots[0], struct btrfs_file_extent_item);
-	btrfs_set_file_extent_generation(ei, trans->transid);
-	btrfs_set_file_extent_type(ei,
+	leaf = path->nodes[0];
+	ei = btrfs_item_ptr(leaf, path->slots[0],
+			    struct btrfs_file_extent_item);
+	btrfs_set_file_extent_generation(leaf, ei, trans->transid);
+	btrfs_set_file_extent_type(leaf, ei,
 				   BTRFS_FILE_EXTENT_INLINE);
 	ptr = btrfs_file_extent_inline_start(ei);
-	btrfs_memcpy(root, path->nodes[0]->b_data,
-		     ptr, symname, name_len);
-	btrfs_mark_buffer_dirty(path->nodes[0]);
+	write_extent_buffer(leaf, symname, ptr, name_len);
+	btrfs_mark_buffer_dirty(leaf);
 	btrfs_free_path(path);
+
 	inode->i_op = &btrfs_symlink_inode_operations;
 	inode->i_mapping->a_ops = &btrfs_symlink_aops;
 	inode->i_size = name_len - 1;
diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c
index bafa78603bd2..a825ce078a54 100644
--- a/fs/btrfs/print-tree.c
+++ b/fs/btrfs/print-tree.c
@@ -20,10 +20,10 @@
 #include "disk-io.h"
 #include "print-tree.h"
 
-void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l)
+void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 {
 	int i;
-	u32 nr = btrfs_header_nritems(&l->header);
+	u32 nr = btrfs_header_nritems(l);
 	struct btrfs_item *item;
 	struct btrfs_extent_item *ei;
 	struct btrfs_root_item *ri;
@@ -31,119 +31,113 @@ void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l)
 	struct btrfs_inode_item *ii;
 	struct btrfs_block_group_item *bi;
 	struct btrfs_file_extent_item *fi;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
 	u32 type;
 
 	printk("leaf %llu total ptrs %d free space %d\n",
-		(unsigned long long)btrfs_header_blocknr(&l->header), nr,
+		(unsigned long long)btrfs_header_blocknr(l), nr,
 		btrfs_leaf_free_space(root, l));
 	for (i = 0 ; i < nr ; i++) {
-		item = l->items + i;
-		type = btrfs_disk_key_type(&item->key);
+		item = btrfs_item_nr(l, i);
+		btrfs_item_key_to_cpu(l, &key, i);
+		type = btrfs_key_type(&key);
 		printk("\titem %d key (%llu %x %llu) itemoff %d itemsize %d\n",
 			i,
-			(unsigned long long)btrfs_disk_key_objectid(&item->key),
-			btrfs_disk_key_flags(&item->key),
-			(unsigned long long)btrfs_disk_key_offset(&item->key),
-			btrfs_item_offset(item),
-			btrfs_item_size(item));
+			(unsigned long long)key.objectid, type,
+			(unsigned long long)key.offset,
+			btrfs_item_offset(l, item), btrfs_item_size(l, item));
 		switch (type) {
 		case BTRFS_INODE_ITEM_KEY:
 			ii = btrfs_item_ptr(l, i, struct btrfs_inode_item);
 			printk("\t\tinode generation %llu size %llu mode %o\n",
-			       (unsigned long long)btrfs_inode_generation(ii),
-			       (unsigned long long)btrfs_inode_size(ii),
-			       btrfs_inode_mode(ii));
+		              (unsigned long long)btrfs_inode_generation(l, ii),
+			      (unsigned long long)btrfs_inode_size(l, ii),
+			       btrfs_inode_mode(l, ii));
 			break;
 		case BTRFS_DIR_ITEM_KEY:
 			di = btrfs_item_ptr(l, i, struct btrfs_dir_item);
+			btrfs_dir_item_key_to_cpu(l, di, &found_key);
 			printk("\t\tdir oid %llu flags %u type %u\n",
-				(unsigned long long)btrfs_disk_key_objectid(
-							    &di->location),
-				btrfs_dir_flags(di),
-				btrfs_dir_type(di));
-			printk("\t\tname %.*s\n",
-			       btrfs_dir_name_len(di),(char *)(di + 1));
+				(unsigned long long)found_key.objectid,
+				btrfs_dir_flags(l, di),
+				btrfs_dir_type(l, di));
 			break;
 		case BTRFS_ROOT_ITEM_KEY:
 			ri = btrfs_item_ptr(l, i, struct btrfs_root_item);
 			printk("\t\troot data blocknr %llu refs %u\n",
-				(unsigned long long)btrfs_root_blocknr(ri),
-				btrfs_root_refs(ri));
+				(unsigned long long)btrfs_disk_root_blocknr(l, ri),
+				btrfs_disk_root_refs(l, ri));
 			break;
 		case BTRFS_EXTENT_ITEM_KEY:
 			ei = btrfs_item_ptr(l, i, struct btrfs_extent_item);
 			printk("\t\textent data refs %u\n",
-				btrfs_extent_refs(ei));
+				btrfs_extent_refs(l, ei));
 			break;
 
 		case BTRFS_EXTENT_DATA_KEY:
 			fi = btrfs_item_ptr(l, i,
 					    struct btrfs_file_extent_item);
-			if (btrfs_file_extent_type(fi) ==
+			if (btrfs_file_extent_type(l, fi) ==
 			    BTRFS_FILE_EXTENT_INLINE) {
 				printk("\t\tinline extent data size %u\n",
-			           btrfs_file_extent_inline_len(l->items + i));
+			           btrfs_file_extent_inline_len(l, item));
 				break;
 			}
 			printk("\t\textent data disk block %llu nr %llu\n",
-			       (unsigned long long)btrfs_file_extent_disk_blocknr(fi),
-			       (unsigned long long)btrfs_file_extent_disk_num_blocks(fi));
+			       (unsigned long long)btrfs_file_extent_disk_blocknr(l, fi),
+			       (unsigned long long)btrfs_file_extent_disk_num_blocks(l, fi));
 			printk("\t\textent data offset %llu nr %llu\n",
-			  (unsigned long long)btrfs_file_extent_offset(fi),
-			  (unsigned long long)btrfs_file_extent_num_blocks(fi));
+			  (unsigned long long)btrfs_file_extent_offset(l, fi),
+			  (unsigned long long)btrfs_file_extent_num_blocks(l, fi));
 			break;
 		case BTRFS_BLOCK_GROUP_ITEM_KEY:
 			bi = btrfs_item_ptr(l, i,
 					    struct btrfs_block_group_item);
 			printk("\t\tblock group used %llu\n",
-			       (unsigned long long)btrfs_block_group_used(bi));
-			break;
-		case BTRFS_STRING_ITEM_KEY:
-			printk("\t\titem data %.*s\n", btrfs_item_size(item),
-				btrfs_leaf_data(l) + btrfs_item_offset(item));
+			       (unsigned long long)btrfs_disk_block_group_used(l, bi));
 			break;
 		};
 	}
 }
 
-void btrfs_print_tree(struct btrfs_root *root, struct buffer_head *t)
+void btrfs_print_tree(struct btrfs_root *root, struct extent_buffer *c)
 {
 	int i;
 	u32 nr;
-	struct btrfs_node *c;
+	struct btrfs_key key;
 
-	if (!t)
+	if (!c)
 		return;
-	c = btrfs_buffer_node(t);
-	nr = btrfs_header_nritems(&c->header);
+	nr = btrfs_header_nritems(c);
 	if (btrfs_is_leaf(c)) {
-		btrfs_print_leaf(root, (struct btrfs_leaf *)c);
+		btrfs_print_leaf(root, c);
 		return;
 	}
 	printk("node %llu level %d total ptrs %d free spc %u\n",
-	       (unsigned long long)btrfs_header_blocknr(&c->header),
-	       btrfs_header_level(&c->header), nr,
+	       (unsigned long long)btrfs_header_blocknr(c),
+	       btrfs_header_level(c), nr,
 	       (u32)BTRFS_NODEPTRS_PER_BLOCK(root) - nr);
 	for (i = 0; i < nr; i++) {
+		btrfs_node_key_to_cpu(c, &key, i);
 		printk("\tkey %d (%llu %u %llu) block %llu\n",
 		       i,
-		       (unsigned long long)c->ptrs[i].key.objectid,
-		       c->ptrs[i].key.flags,
-		       (unsigned long long)c->ptrs[i].key.offset,
+		       (unsigned long long)key.objectid,
+		       key.type,
+		       (unsigned long long)key.offset,
 		       (unsigned long long)btrfs_node_blockptr(c, i));
 	}
 	for (i = 0; i < nr; i++) {
-		struct buffer_head *next_buf = read_tree_block(root,
+		struct extent_buffer *next = read_tree_block(root,
 						btrfs_node_blockptr(c, i));
-		struct btrfs_node *next = btrfs_buffer_node(next_buf);
 		if (btrfs_is_leaf(next) &&
-		    btrfs_header_level(&c->header) != 1)
+		    btrfs_header_level(c) != 1)
 			BUG();
-		if (btrfs_header_level(&next->header) !=
-			btrfs_header_level(&c->header) - 1)
+		if (btrfs_header_level(next) !=
+			btrfs_header_level(c) - 1)
 			BUG();
-		btrfs_print_tree(root, next_buf);
-		btrfs_block_release(root, next_buf);
+		btrfs_print_tree(root, next);
+		free_extent_buffer(next);
 	}
 }
 
diff --git a/fs/btrfs/print-tree.h b/fs/btrfs/print-tree.h
index 205cd03601ab..da75efe534d5 100644
--- a/fs/btrfs/print-tree.h
+++ b/fs/btrfs/print-tree.h
@@ -18,6 +18,6 @@
 
 #ifndef __PRINT_TREE_
 #define __PRINT_TREE_
-void btrfs_print_leaf(struct btrfs_root *root, struct btrfs_leaf *l);
-void btrfs_print_tree(struct btrfs_root *root, struct buffer_head *t);
+void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l);
+void btrfs_print_tree(struct btrfs_root *root, struct extent_buffer *t);
 #endif
diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c
index 3b5926dfbeba..88bcdd33f56e 100644
--- a/fs/btrfs/root-tree.c
+++ b/fs/btrfs/root-tree.c
@@ -26,12 +26,13 @@ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid,
 {
 	struct btrfs_path *path;
 	struct btrfs_key search_key;
-	struct btrfs_leaf *l;
+	struct btrfs_key found_key;
+	struct extent_buffer *l;
 	int ret;
 	int slot;
 
 	search_key.objectid = objectid;
-	search_key.flags = (u32)-1;
+	search_key.type = (u8)-1;
 	search_key.offset = (u64)-1;
 
 	path = btrfs_alloc_path();
@@ -39,17 +40,19 @@ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid,
 	ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
 	if (ret < 0)
 		goto out;
+
 	BUG_ON(ret == 0);
-	l = btrfs_buffer_leaf(path->nodes[0]);
+	l = path->nodes[0];
 	BUG_ON(path->slots[0] == 0);
 	slot = path->slots[0] - 1;
-	if (btrfs_disk_key_objectid(&l->items[slot].key) != objectid) {
+	btrfs_item_key_to_cpu(l, &found_key, slot);
+	if (found_key.objectid != objectid) {
 		ret = 1;
 		goto out;
 	}
-	memcpy(item, btrfs_item_ptr(l, slot, struct btrfs_root_item),
-		sizeof(*item));
-	btrfs_disk_key_to_cpu(key, &l->items[slot].key);
+	read_extent_buffer(l, item, btrfs_item_ptr_offset(l, slot),
+			   sizeof(*item));
+	memcpy(key, &found_key, sizeof(found_key));
 	ret = 0;
 out:
 	btrfs_release_path(root, path);
@@ -62,10 +65,10 @@ int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *item)
 {
 	struct btrfs_path *path;
-	struct btrfs_leaf *l;
+	struct extent_buffer *l;
 	int ret;
 	int slot;
-	struct btrfs_root_item *update_item;
+	unsigned long ptr;
 
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
@@ -73,10 +76,10 @@ int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root
 	if (ret < 0)
 		goto out;
 	BUG_ON(ret != 0);
-	l = btrfs_buffer_leaf(path->nodes[0]);
+	l = path->nodes[0];
 	slot = path->slots[0];
-	update_item = btrfs_item_ptr(l, slot, struct btrfs_root_item);
-	btrfs_memcpy(root, l, update_item, item, sizeof(*item));
+	ptr = btrfs_item_ptr_offset(l, slot);
+	write_extent_buffer(l, item, ptr, sizeof(*item));
 	btrfs_mark_buffer_dirty(path->nodes[0]);
 out:
 	btrfs_release_path(root, path);
@@ -103,11 +106,10 @@ int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid,
 	struct btrfs_path *path;
 	int ret;
 	u32 nritems;
-	struct btrfs_leaf *leaf;
+	struct extent_buffer *leaf;
 	int slot;
 
 	key.objectid = objectid;
-	key.flags = 0;
 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
 	key.offset = 0;
 	path = btrfs_alloc_path();
@@ -117,19 +119,19 @@ int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid,
 	if (ret < 0)
 		goto err;
 	while(1) {
-		leaf = btrfs_buffer_leaf(path->nodes[0]);
-		nritems = btrfs_header_nritems(&leaf->header);
+		leaf = path->nodes[0];
+		nritems = btrfs_header_nritems(leaf);
 		slot = path->slots[0];
 		if (slot >= nritems) {
 			ret = btrfs_next_leaf(root, path);
 			if (ret)
 				break;
-			leaf = btrfs_buffer_leaf(path->nodes[0]);
-			nritems = btrfs_header_nritems(&leaf->header);
+			leaf = path->nodes[0];
+			nritems = btrfs_header_nritems(leaf);
 			slot = path->slots[0];
 		}
-		item = leaf->items + slot;
-		btrfs_disk_key_to_cpu(&key, &item->key);
+		item = btrfs_item_nr(leaf, slot);
+		btrfs_item_key_to_cpu(leaf, &key, slot);
 		if (btrfs_key_type(&key) != BTRFS_ROOT_ITEM_KEY)
 			goto next;
 
@@ -140,7 +142,7 @@ int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid,
 			break;
 
 		ri = btrfs_item_ptr(leaf, slot, struct btrfs_root_item);
-		if (btrfs_root_refs(ri) != 0)
+		if (btrfs_disk_root_refs(leaf, ri) != 0)
 			goto next;
 
 		dead_root = btrfs_read_fs_root_no_radix(root->fs_info, &key);
@@ -170,6 +172,7 @@ int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	int ret;
 	u32 refs;
 	struct btrfs_root_item *ri;
+	struct extent_buffer *leaf;
 
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
@@ -177,10 +180,10 @@ int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	if (ret < 0)
 		goto out;
 	BUG_ON(ret != 0);
-	ri = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
-			    path->slots[0], struct btrfs_root_item);
+	leaf = path->nodes[0];
+	ri = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_item);
 
-	refs = btrfs_root_refs(ri);
+	refs = btrfs_disk_root_refs(leaf, ri);
 	BUG_ON(refs != 0);
 	ret = btrfs_del_item(trans, root, path);
 out:
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 66a01cbbbea1..39a1435c68f1 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -41,7 +41,7 @@
 #include "ioctl.h"
 #include "print-tree.h"
 
-#define BTRFS_SUPER_MAGIC 0x9123682E
+#define BTRFS_SUPER_MAGIC 0x9123683E
 
 static struct super_operations btrfs_super_ops;
 
@@ -115,13 +115,12 @@ static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
 		return -EIO;
 	}
 	sb->s_fs_info = tree_root;
-	disk_super = tree_root->fs_info->disk_super;
+	disk_super = &tree_root->fs_info->super_copy;
 	inode = btrfs_iget_locked(sb, btrfs_super_root_dir(disk_super),
 				  tree_root);
 	bi = BTRFS_I(inode);
 	bi->location.objectid = inode->i_ino;
 	bi->location.offset = 0;
-	bi->location.flags = 0;
 	bi->root = tree_root;
 
 	btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY);
@@ -281,6 +280,7 @@ error_s:
 error_bdev:
 	close_bdev_excl(bdev);
 error:
+printk("get_sb failed\n");
 	return error;
 }
 /* end copy & paste */
@@ -295,6 +295,7 @@ static int btrfs_get_sb(struct file_system_type *fs_type,
 	ret = btrfs_get_sb_bdev(fs_type, flags, dev_name, data,
 			btrfs_fill_super, mnt,
 			subvol_name ? subvol_name : "default");
+printk("btrfs_get_sb returns %d\n", ret);
 	return ret;
 }
 
diff --git a/fs/btrfs/sysfs.c b/fs/btrfs/sysfs.c
index 2058783373eb..9654e90eec89 100644
--- a/fs/btrfs/sysfs.c
+++ b/fs/btrfs/sysfs.c
@@ -31,31 +31,31 @@
 static ssize_t root_blocks_used_show(struct btrfs_root *root, char *buf)
 {
 	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_root_blocks_used(&root->root_item));
+		(unsigned long long)btrfs_root_used(&root->root_item));
 }
 
 static ssize_t root_block_limit_show(struct btrfs_root *root, char *buf)
 {
 	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_root_block_limit(&root->root_item));
+		(unsigned long long)btrfs_root_limit(&root->root_item));
 }
 
 static ssize_t super_blocks_used_show(struct btrfs_fs_info *fs, char *buf)
 {
 	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_super_blocks_used(fs->disk_super));
+		(unsigned long long)btrfs_super_blocks_used(&fs->super_copy));
 }
 
 static ssize_t super_total_blocks_show(struct btrfs_fs_info *fs, char *buf)
 {
 	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_super_total_blocks(fs->disk_super));
+		(unsigned long long)btrfs_super_total_blocks(&fs->super_copy));
 }
 
 static ssize_t super_blocksize_show(struct btrfs_fs_info *fs, char *buf)
 {
 	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_super_blocksize(fs->disk_super));
+		(unsigned long long)btrfs_super_sectorsize(&fs->super_copy));
 }
 
 /* this is for root attrs (subvols/snapshots) */
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 087074db0bd5..750f35a37aae 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -19,6 +19,7 @@
 #include <linux/fs.h>
 #include <linux/sched.h>
 #include <linux/writeback.h>
+#include <linux/pagemap.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
@@ -66,7 +67,9 @@ static int join_transaction(struct btrfs_root *root)
 		cur_trans->commit_done = 0;
 		cur_trans->start_time = get_seconds();
 		list_add_tail(&cur_trans->list, &root->fs_info->trans_list);
-		init_bit_radix(&cur_trans->dirty_pages);
+		extent_map_tree_init(&cur_trans->dirty_pages,
+				     root->fs_info->btree_inode->i_mapping,
+				     GFP_NOFS);
 	} else {
 		cur_trans->num_writers++;
 		cur_trans->num_joined++;
@@ -88,7 +91,7 @@ static int record_root_in_trans(struct btrfs_root *root)
 				   (unsigned long)root->root_key.objectid,
 				   BTRFS_ROOT_DEFRAG_TAG);
 			root->commit_root = root->node;
-			get_bh(root->node);
+			extent_buffer_get(root->node);
 		} else {
 			WARN_ON(1);
 		}
@@ -144,29 +147,30 @@ int btrfs_end_transaction(struct btrfs_trans_handle *trans,
 int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
 				     struct btrfs_root *root)
 {
-	unsigned long gang[16];
 	int ret;
-	int i;
 	int err;
 	int werr = 0;
+	struct extent_map_tree *dirty_pages;
 	struct page *page;
-	struct radix_tree_root *dirty_pages;
 	struct inode *btree_inode = root->fs_info->btree_inode;
+	u64 start;
+	u64 end;
+	unsigned long index;
 
 	if (!trans || !trans->transaction) {
 		return filemap_write_and_wait(btree_inode->i_mapping);
 	}
 	dirty_pages = &trans->transaction->dirty_pages;
 	while(1) {
-		ret = find_first_radix_bit(dirty_pages, gang,
-					   0, ARRAY_SIZE(gang));
-		if (!ret)
+		ret = find_first_extent_bit(dirty_pages, 0, &start, &end,
+					    EXTENT_DIRTY);
+		if (ret)
 			break;
-		for (i = 0; i < ret; i++) {
-			/* FIXME EIO */
-			clear_radix_bit(dirty_pages, gang[i]);
-			page = find_lock_page(btree_inode->i_mapping,
-					      gang[i]);
+		clear_extent_dirty(dirty_pages, start, end, GFP_NOFS);
+		while(start <= end) {
+			index = start >> PAGE_CACHE_SHIFT;
+			start = (index + 1) << PAGE_CACHE_SHIFT;
+			page = find_lock_page(btree_inode->i_mapping, index);
 			if (!page)
 				continue;
 			if (PageWriteback(page)) {
@@ -202,10 +206,11 @@ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
 	btrfs_write_dirty_block_groups(trans, extent_root);
 	while(1) {
 		old_extent_block = btrfs_root_blocknr(&extent_root->root_item);
-		if (old_extent_block == bh_blocknr(extent_root->node))
+		if (old_extent_block ==
+		    extent_buffer_blocknr(extent_root->node))
 			break;
 		btrfs_set_root_blocknr(&extent_root->root_item,
-				       bh_blocknr(extent_root->node));
+			       extent_buffer_blocknr(extent_root->node));
 		ret = btrfs_update_root(trans, tree_root,
 					&extent_root->root_key,
 					&extent_root->root_item);
@@ -279,9 +284,9 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans,
 				     (unsigned long)root->root_key.objectid,
 				     BTRFS_ROOT_TRANS_TAG);
 			if (root->commit_root == root->node) {
-				WARN_ON(bh_blocknr(root->node) !=
+				WARN_ON(extent_buffer_blocknr(root->node) !=
 					btrfs_root_blocknr(&root->root_item));
-				brelse(root->commit_root);
+				free_extent_buffer(root->commit_root);
 				root->commit_root = NULL;
 
 				/* make sure to update the root on disk
@@ -310,7 +315,7 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans,
 
 			root->root_key.offset = root->fs_info->generation;
 			btrfs_set_root_blocknr(&root->root_item,
-					       bh_blocknr(root->node));
+				       extent_buffer_blocknr(root->node));
 			err = btrfs_insert_root(trans, root->fs_info->tree_root,
 						&root->root_key,
 						&root->root_item);
@@ -389,10 +394,10 @@ int btrfs_defrag_dirty_roots(struct btrfs_fs_info *info)
 		for (i = 0; i < ret; i++) {
 			root = gang[i];
 			last = root->root_key.objectid + 1;
-			btrfs_defrag_root(root, 1);
+			// btrfs_defrag_root(root, 1);
 		}
 	}
-	btrfs_defrag_root(info->extent_root, 1);
+	// btrfs_defrag_root(info->extent_root, 1);
 	return err;
 }
 
@@ -414,7 +419,7 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 		dirty = list_entry(list->next, struct dirty_root, list);
 		list_del_init(&dirty->list);
 
-		num_blocks = btrfs_root_blocks_used(&dirty->root->root_item);
+		num_blocks = btrfs_root_used(&dirty->root->root_item);
 		root = dirty->latest_root;
 
 		while(1) {
@@ -441,11 +446,11 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 		}
 		BUG_ON(ret);
 
-		num_blocks -= btrfs_root_blocks_used(&dirty->root->root_item);
-		blocks_used = btrfs_root_blocks_used(&root->root_item);
+		num_blocks -= btrfs_root_used(&dirty->root->root_item);
+		blocks_used = btrfs_root_used(&root->root_item);
 		if (num_blocks) {
 			record_root_in_trans(root);
-			btrfs_set_root_blocks_used(&root->root_item,
+			btrfs_set_root_used(&root->root_item,
 						   blocks_used - num_blocks);
 		}
 		ret = btrfs_del_root(trans, tree_root, &dirty->root->root_key);
@@ -553,9 +558,11 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	btrfs_set_super_generation(&root->fs_info->super_copy,
 				   cur_trans->transid);
 	btrfs_set_super_root(&root->fs_info->super_copy,
-			     bh_blocknr(root->fs_info->tree_root->node));
-	memcpy(root->fs_info->disk_super, &root->fs_info->super_copy,
-	       sizeof(root->fs_info->super_copy));
+		     extent_buffer_blocknr(root->fs_info->tree_root->node));
+
+	write_extent_buffer(root->fs_info->sb_buffer,
+			    &root->fs_info->super_copy, 0,
+			    sizeof(root->fs_info->super_copy));
 
 	btrfs_copy_pinned(root, &pinned_copy);
 
diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h
index 4bc328cbb24c..ae39fcfc169a 100644
--- a/fs/btrfs/transaction.h
+++ b/fs/btrfs/transaction.h
@@ -28,7 +28,7 @@ struct btrfs_transaction {
 	int use_count;
 	int commit_done;
 	struct list_head list;
-	struct radix_tree_root dirty_pages;
+	struct extent_map_tree dirty_pages;
 	unsigned long start_time;
 	wait_queue_head_t writer_wait;
 	wait_queue_head_t commit_wait;
@@ -83,5 +83,6 @@ int btrfs_add_dead_root(struct btrfs_root *root, struct btrfs_root *latest,
 int btrfs_defrag_dirty_roots(struct btrfs_fs_info *info);
 int btrfs_defrag_root(struct btrfs_root *root, int cacheonly);
 int btrfs_clean_old_snapshots(struct btrfs_root *root);
-
+int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root);
 #endif
diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c
index 420597127ed1..daf019afa0a1 100644
--- a/fs/btrfs/tree-defrag.c
+++ b/fs/btrfs/tree-defrag.c
@@ -23,14 +23,14 @@
 #include "transaction.h"
 
 static void reada_defrag(struct btrfs_root *root,
-			 struct btrfs_node *node)
+			 struct extent_buffer *node)
 {
 	int i;
 	u32 nritems;
 	u64 blocknr;
 	int ret;
 
-	nritems = btrfs_header_nritems(&node->header);
+	nritems = btrfs_header_nritems(node);
 	for (i = 0; i < nritems; i++) {
 		blocknr = btrfs_node_blockptr(node, i);
 		ret = readahead_tree_block(root, blocknr);
@@ -44,8 +44,8 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 			    struct btrfs_path *path, int *level,
 			    int cache_only, u64 *last_ret)
 {
-	struct buffer_head *next;
-	struct buffer_head *cur;
+	struct extent_buffer *next;
+	struct extent_buffer *cur;
 	u64 blocknr;
 	int ret = 0;
 	int is_extent = 0;
@@ -62,13 +62,13 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 		cur = path->nodes[*level];
 
 		if (!cache_only && *level > 1 && path->slots[*level] == 0)
-			reada_defrag(root, btrfs_buffer_node(cur));
+			reada_defrag(root, cur);
 
-		if (btrfs_header_level(btrfs_buffer_header(cur)) != *level)
+		if (btrfs_header_level(cur) != *level)
 			WARN_ON(1);
 
 		if (path->slots[*level] >=
-		    btrfs_header_nritems(btrfs_buffer_header(cur)))
+		    btrfs_header_nritems(cur))
 			break;
 
 		if (*level == 1) {
@@ -80,14 +80,13 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 
 			break;
 		}
-		blocknr = btrfs_node_blockptr(btrfs_buffer_node(cur),
-					      path->slots[*level]);
+		blocknr = btrfs_node_blockptr(cur, path->slots[*level]);
 
 		if (cache_only) {
 			next = btrfs_find_tree_block(root, blocknr);
-			if (!next || !buffer_uptodate(next) ||
-			   buffer_locked(next) || !buffer_defrag(next)) {
-				brelse(next);
+			/* FIXME, test for defrag */
+			if (!next || !btrfs_buffer_uptodate(next)) {
+				free_extent_buffer(next);
 				path->slots[*level]++;
 				continue;
 			}
@@ -106,16 +105,18 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 
 		WARN_ON(*level <= 0);
 		if (path->nodes[*level-1])
-			btrfs_block_release(root, path->nodes[*level-1]);
+			free_extent_buffer(path->nodes[*level-1]);
 		path->nodes[*level-1] = next;
-		*level = btrfs_header_level(btrfs_buffer_header(next));
+		*level = btrfs_header_level(next);
 		path->slots[*level] = 0;
 	}
 	WARN_ON(*level < 0);
 	WARN_ON(*level >= BTRFS_MAX_LEVEL);
+#if 0
 	clear_buffer_defrag(path->nodes[*level]);
 	clear_buffer_defrag_done(path->nodes[*level]);
-	btrfs_block_release(root, path->nodes[*level]);
+#endif
+	free_extent_buffer(path->nodes[*level]);
 	path->nodes[*level] = NULL;
 	*level += 1;
 	WARN_ON(ret);
@@ -129,24 +130,25 @@ static int defrag_walk_up(struct btrfs_trans_handle *trans,
 {
 	int i;
 	int slot;
-	struct btrfs_node *node;
+	struct extent_buffer *node;
 
 	for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
 		slot = path->slots[i];
-		if (slot < btrfs_header_nritems(
-		    btrfs_buffer_header(path->nodes[i])) - 1) {
+		if (slot < btrfs_header_nritems(path->nodes[i]) - 1) {
 			path->slots[i]++;
 			*level = i;
-			node = btrfs_buffer_node(path->nodes[i]);
+			node = path->nodes[i];
 			WARN_ON(i == 0);
-			btrfs_disk_key_to_cpu(&root->defrag_progress,
-					      &node->ptrs[path->slots[i]].key);
+			btrfs_node_key_to_cpu(node, &root->defrag_progress,
+					      path->slots[i]);
 			root->defrag_level = i;
 			return 0;
 		} else {
+			/*
 			clear_buffer_defrag(path->nodes[*level]);
 			clear_buffer_defrag_done(path->nodes[*level]);
-			btrfs_block_release(root, path->nodes[*level]);
+			*/
+			free_extent_buffer(path->nodes[*level]);
 			path->nodes[*level] = NULL;
 			*level = i + 1;
 		}
@@ -158,7 +160,7 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root, int cache_only)
 {
 	struct btrfs_path *path = NULL;
-	struct buffer_head *tmp;
+	struct extent_buffer *tmp;
 	int ret = 0;
 	int wret;
 	int level;
@@ -172,17 +174,18 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 
 	if (root->ref_cows == 0 && !is_extent)
 		goto out;
+
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
 
-	level = btrfs_header_level(btrfs_buffer_header(root->node));
+	level = btrfs_header_level(root->node);
 	orig_level = level;
 	if (level == 0) {
 		goto out;
 	}
 	if (root->defrag_progress.objectid == 0) {
-		get_bh(root->node);
+		extent_buffer_get(root->node);
 		ret = btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
 		BUG_ON(ret);
 		ret = btrfs_realloc_node(trans, root, root->node, cache_only,
@@ -200,12 +203,15 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 
 		if (is_extent)
 			btrfs_extent_post_op(trans, root);
+
 		if (wret < 0) {
 			ret = wret;
 			goto out;
 		}
+
 		while(level > 0 && !path->nodes[level])
 			level--;
+
 		if (!path->nodes[level]) {
 			ret = 0;
 			goto out;
@@ -230,7 +236,7 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 	}
 	for (i = 0; i <= orig_level; i++) {
 		if (path->nodes[i]) {
-			btrfs_block_release(root, path->nodes[i]);
+			free_extent_buffer(path->nodes[i]);
 			path->nodes[i] = 0;
 		}
 	}
-- 
cgit v1.2.3


From db94535db75e67fab12ccbb7f5ee548e33fed891 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Mon, 15 Oct 2007 16:15:53 -0400
Subject: Btrfs: Allow tree blocks larger than the page size

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.c       | 319 +++++++++++++++++++++++++++++++++++--------------
 fs/btrfs/ctree.h       | 112 ++++++++++-------
 fs/btrfs/disk-io.c     |  92 ++++++++------
 fs/btrfs/disk-io.h     |   9 +-
 fs/btrfs/extent-tree.c | 285 ++++++++++++++++++++++---------------------
 fs/btrfs/extent_map.c  |  91 ++++++++++----
 fs/btrfs/extent_map.h  |  10 +-
 fs/btrfs/file-item.c   |  10 +-
 fs/btrfs/file.c        | 126 ++++++++++---------
 fs/btrfs/inode.c       |  90 +++++++-------
 fs/btrfs/print-tree.c  |  23 ++--
 fs/btrfs/super.c       |   6 +-
 fs/btrfs/sysfs.c       |   5 +-
 fs/btrfs/transaction.c |  39 +++---
 fs/btrfs/tree-defrag.c |  18 +--
 15 files changed, 738 insertions(+), 497 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 50ef351ef6b3..34de83630ae9 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -83,16 +83,15 @@ static int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 
 	WARN_ON(root->ref_cows && trans->transid != root->last_trans);
 
-	cow = btrfs_alloc_free_block(trans, root, search_start, empty_size);
+	cow = btrfs_alloc_free_block(trans, root, buf->len,
+				     search_start, empty_size);
 	if (IS_ERR(cow))
 		return PTR_ERR(cow);
 
 	cow->alloc_addr = (unsigned long)__builtin_return_address(0);
-	if (buf->len != root->sectorsize || cow->len != root->sectorsize)
-		WARN_ON(1);
 
 	copy_extent_buffer(cow, buf, 0, 0, cow->len);
-	btrfs_set_header_blocknr(cow, extent_buffer_blocknr(cow));
+	btrfs_set_header_bytenr(cow, cow->start);
 	btrfs_set_header_generation(cow, trans->transid);
 	btrfs_set_header_owner(cow, root->root_key.objectid);
 
@@ -110,16 +109,16 @@ static int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 		root->node = cow;
 		extent_buffer_get(cow);
 		if (buf != root->commit_root) {
-			btrfs_free_extent(trans, root,
-					  extent_buffer_blocknr(buf), 1, 1);
+			btrfs_free_extent(trans, root, buf->start,
+					  buf->len, 1);
 		}
 		free_extent_buffer(buf);
 	} else {
 		btrfs_set_node_blockptr(parent, parent_slot,
-					extent_buffer_blocknr(cow));
+					cow->start);
 		btrfs_mark_buffer_dirty(parent);
 		WARN_ON(btrfs_header_generation(parent) != trans->transid);
-		btrfs_free_extent(trans, root, extent_buffer_blocknr(buf),1,1);
+		btrfs_free_extent(trans, root, buf->start, buf->len, 1);
 	}
 	free_extent_buffer(buf);
 	btrfs_mark_buffer_dirty(cow);
@@ -149,13 +148,14 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		return 0;
 	}
 
-	search_start = extent_buffer_blocknr(buf) & ~((u64)65535);
+	search_start = buf->start & ~((u64)BTRFS_BLOCK_GROUP_SIZE - 1);
 	ret = __btrfs_cow_block(trans, root, buf, parent,
 				 parent_slot, cow_ret, search_start, 0);
 	(*cow_ret)->alloc_addr = (unsigned long)__builtin_return_address(0);
 	return ret;
 }
 
+#if 0
 static int close_blocks(u64 blocknr, u64 other)
 {
 	if (blocknr < other && other - blocknr < 8)
@@ -165,7 +165,6 @@ static int close_blocks(u64 blocknr, u64 other)
 	return 0;
 }
 
-#if 0
 static int should_defrag_leaf(struct extent_buffer *eb)
 {
 	return 0;
@@ -355,7 +354,7 @@ static int check_node(struct btrfs_root *root, struct btrfs_path *path,
 		BUG_ON(memcmp(&parent_key, &node_key,
 			      sizeof(struct btrfs_disk_key)));
 		BUG_ON(btrfs_node_blockptr(parent, parent_slot) !=
-		       btrfs_header_blocknr(node));
+		       btrfs_header_bytenr(node));
 	}
 	BUG_ON(nritems > BTRFS_NODEPTRS_PER_BLOCK(root));
 	if (slot != 0) {
@@ -398,7 +397,7 @@ static int check_leaf(struct btrfs_root *root, struct btrfs_path *path,
 		BUG_ON(memcmp(&parent_key, &leaf_key,
 		       sizeof(struct btrfs_disk_key)));
 		BUG_ON(btrfs_node_blockptr(parent, parent_slot) !=
-		       btrfs_header_blocknr(leaf));
+		       btrfs_header_bytenr(leaf));
 	}
 #if 0
 	for (i = 0; nritems > 1 && i < nritems - 2; i++) {
@@ -467,14 +466,16 @@ static int check_leaf(struct btrfs_root *root, struct btrfs_path *path,
 static int check_block(struct btrfs_root *root, struct btrfs_path *path,
 			int level)
 {
+#if 0
 	struct extent_buffer *buf = path->nodes[level];
 
 	if (memcmp_extent_buffer(buf, root->fs_info->fsid,
 				 (unsigned long)btrfs_header_fsid(buf),
 				 BTRFS_FSID_SIZE)) {
 		printk("warning bad block %Lu\n", buf->start);
-		BUG();
+		return 1;
 	}
+#endif
 	if (level == 0)
 		return check_leaf(root, path, level);
 	return check_node(root, path, level);
@@ -585,7 +586,8 @@ static struct extent_buffer *read_node_slot(struct btrfs_root *root,
 		return NULL;
 	if (slot >= btrfs_header_nritems(parent))
 		return NULL;
-	return read_tree_block(root, btrfs_node_blockptr(parent, slot));
+	return read_tree_block(root, btrfs_node_blockptr(parent, slot),
+		       btrfs_level_size(root, btrfs_header_level(parent) - 1));
 }
 
 static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
@@ -618,7 +620,6 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 	 */
 	if (!parent) {
 		struct extent_buffer *child;
-		u64 blocknr = extent_buffer_blocknr(mid);
 
 		if (btrfs_header_nritems(mid) != 1)
 			return 0;
@@ -632,9 +633,10 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 		wait_on_tree_block_writeback(root, mid);
 		/* once for the path */
 		free_extent_buffer(mid);
+		ret = btrfs_free_extent(trans, root, mid->start, mid->len, 1);
 		/* once for the root ptr */
 		free_extent_buffer(mid);
-		return btrfs_free_extent(trans, root, blocknr, 1, 1);
+		return ret;
 	}
 	if (btrfs_header_nritems(mid) >
 	    BTRFS_NODEPTRS_PER_BLOCK(root) / 4)
@@ -680,7 +682,9 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 		if (wret < 0 && wret != -ENOSPC)
 			ret = wret;
 		if (btrfs_header_nritems(right) == 0) {
-			u64 blocknr = extent_buffer_blocknr(right);
+			u64 bytenr = right->start;
+			u32 blocksize = right->len;
+
 			clean_tree_block(trans, root, right);
 			wait_on_tree_block_writeback(root, right);
 			free_extent_buffer(right);
@@ -689,7 +693,8 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 				       1);
 			if (wret)
 				ret = wret;
-			wret = btrfs_free_extent(trans, root, blocknr, 1, 1);
+			wret = btrfs_free_extent(trans, root, bytenr,
+						 blocksize, 1);
 			if (wret)
 				ret = wret;
 		} else {
@@ -719,7 +724,8 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 	}
 	if (btrfs_header_nritems(mid) == 0) {
 		/* we've managed to empty the middle node, drop it */
-		u64 blocknr = extent_buffer_blocknr(mid);
+		u64 bytenr = mid->start;
+		u32 blocksize = mid->len;
 		clean_tree_block(trans, root, mid);
 		wait_on_tree_block_writeback(root, mid);
 		free_extent_buffer(mid);
@@ -727,7 +733,7 @@ static int balance_level(struct btrfs_trans_handle *trans, struct btrfs_root
 		wret = del_ptr(trans, root, path, level + 1, pslot);
 		if (wret)
 			ret = wret;
-		wret = btrfs_free_extent(trans, root, blocknr, 1, 1);
+		wret = btrfs_free_extent(trans, root, bytenr, blocksize, 1);
 		if (wret)
 			ret = wret;
 	} else {
@@ -830,7 +836,6 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 				path->slots[level] = orig_slot;
 				free_extent_buffer(left);
 			}
-			check_node(root, path, level);
 			return 0;
 		}
 		free_extent_buffer(left);
@@ -874,12 +879,10 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 			} else {
 				free_extent_buffer(right);
 			}
-			check_node(root, path, level);
 			return 0;
 		}
 		free_extent_buffer(right);
 	}
-	check_node(root, path, level);
 	return 1;
 }
 
@@ -889,19 +892,23 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
 			     int level, int slot)
 {
+	return;
+#if 0
 	struct extent_buffer *node;
 	int i;
 	u32 nritems;
-	u64 blocknr;
+	u64 bytenr;
 	u64 search;
 	u64 cluster_start;
 	int ret;
 	int nread = 0;
 	int direction = path->reada;
+	int level;
 	struct radix_tree_root found;
 	unsigned long gang[8];
 	struct extent_buffer *eb;
 
+
 	if (level == 0)
 		return;
 
@@ -918,8 +925,9 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
 
 	init_bit_radix(&found);
 	nritems = btrfs_header_nritems(node);
+	level = btrfs_header_level(node) - 1;
 	for (i = slot; i < nritems; i++) {
-		blocknr = btrfs_node_blockptr(node, i);
+		bytenr = btrfs_node_blockptr(node, i);
 		set_radix_bit(&found, blocknr);
 	}
 	if (direction > 0) {
@@ -944,6 +952,7 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
 			}
 		}
 	}
+#endif
 }
 /*
  * look for key in the tree.  path is filled in with nodes along the way
@@ -963,7 +972,7 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
 		      ins_len, int cow)
 {
 	struct extent_buffer *b;
-	u64 blocknr;
+	u64 bytenr;
 	int slot;
 	int ret;
 	int level;
@@ -1027,10 +1036,11 @@ again:
 			/* this is only true while dropping a snapshot */
 			if (level == lowest_level)
 				break;
-			blocknr = btrfs_node_blockptr(b, slot);
+			bytenr = btrfs_node_blockptr(b, slot);
 			if (should_reada)
 				reada_for_search(root, p, level, slot);
-			b = read_tree_block(root, btrfs_node_blockptr(b, slot));
+			b = read_tree_block(root, bytenr,
+					    btrfs_level_size(root, level - 1));
 		} else {
 			p->slots[level] = slot;
 			if (ins_len > 0 && btrfs_leaf_free_space(root, b) <
@@ -1193,14 +1203,14 @@ static int insert_new_root(struct btrfs_trans_handle *trans,
 	BUG_ON(path->nodes[level]);
 	BUG_ON(path->nodes[level-1] != root->node);
 
-	c = btrfs_alloc_free_block(trans, root,
-				   extent_buffer_blocknr(root->node), 0);
+	c = btrfs_alloc_free_block(trans, root, root->nodesize,
+				   root->node->start, 0);
 	if (IS_ERR(c))
 		return PTR_ERR(c);
 	memset_extent_buffer(c, 0, 0, root->nodesize);
 	btrfs_set_header_nritems(c, 1);
 	btrfs_set_header_level(c, level);
-	btrfs_set_header_blocknr(c, extent_buffer_blocknr(c));
+	btrfs_set_header_bytenr(c, c->start);
 	btrfs_set_header_generation(c, trans->transid);
 	btrfs_set_header_owner(c, root->root_key.objectid);
 	lower = path->nodes[level-1];
@@ -1213,7 +1223,7 @@ static int insert_new_root(struct btrfs_trans_handle *trans,
 	else
 		btrfs_node_key(lower, &lower_key, 0);
 	btrfs_set_node_key(c, &lower_key, 0);
-	btrfs_set_node_blockptr(c, 0, extent_buffer_blocknr(lower));
+	btrfs_set_node_blockptr(c, 0, lower->start);
 
 	btrfs_mark_buffer_dirty(c);
 
@@ -1237,7 +1247,7 @@ static int insert_new_root(struct btrfs_trans_handle *trans,
  */
 static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_path *path, struct btrfs_disk_key
-		      *key, u64 blocknr, int slot, int level)
+		      *key, u64 bytenr, int slot, int level)
 {
 	struct extent_buffer *lower;
 	int nritems;
@@ -1256,10 +1266,9 @@ static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root
 			      (nritems - slot) * sizeof(struct btrfs_key_ptr));
 	}
 	btrfs_set_node_key(lower, key, slot);
-	btrfs_set_node_blockptr(lower, slot, blocknr);
+	btrfs_set_node_blockptr(lower, slot, bytenr);
 	btrfs_set_header_nritems(lower, nritems + 1);
 	btrfs_mark_buffer_dirty(lower);
-	check_node(root, path, level);
 	return 0;
 }
 
@@ -1300,14 +1309,14 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 	}
 
 	c_nritems = btrfs_header_nritems(c);
-	split = btrfs_alloc_free_block(trans, root,
-				       extent_buffer_blocknr(c), 0);
+	split = btrfs_alloc_free_block(trans, root, root->nodesize,
+				       c->start, 0);
 	if (IS_ERR(split))
 		return PTR_ERR(split);
 
 	btrfs_set_header_flags(split, btrfs_header_flags(c));
 	btrfs_set_header_level(split, btrfs_header_level(c));
-	btrfs_set_header_blocknr(split, extent_buffer_blocknr(split));
+	btrfs_set_header_bytenr(split, split->start);
 	btrfs_set_header_generation(split, trans->transid);
 	btrfs_set_header_owner(split, root->root_key.objectid);
 	write_extent_buffer(split, root->fs_info->fsid,
@@ -1328,8 +1337,7 @@ static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_mark_buffer_dirty(split);
 
 	btrfs_node_key(split, &disk_key, 0);
-	wret = insert_ptr(trans, root, path, &disk_key,
-			  extent_buffer_blocknr(split),
+	wret = insert_ptr(trans, root, path, &disk_key, split->start,
 			  path->slots[level + 1] + 1,
 			  level + 1);
 	if (wret)
@@ -1407,6 +1415,7 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 	u32 left_nritems;
 	u32 right_nritems;
 	u32 data_end;
+	u32 this_item_size;
 	int ret;
 
 	slot = path->slots[1];
@@ -1417,7 +1426,8 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 	if (slot >= btrfs_header_nritems(upper) - 1)
 		return 1;
 
-	right = read_tree_block(root, btrfs_node_blockptr(upper, slot + 1));
+	right = read_tree_block(root, btrfs_node_blockptr(upper, slot + 1),
+				root->leafsize);
 	free_space = btrfs_leaf_free_space(root, right);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
 		free_extent_buffer(right);
@@ -1445,13 +1455,27 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	for (i = left_nritems - 1; i >= 1; i--) {
 		item = btrfs_item_nr(left, i);
+
 		if (path->slots[0] == i)
 			push_space += data_size + sizeof(*item);
-		if (btrfs_item_size(left, item) + sizeof(*item) + push_space >
-		    free_space)
+
+		if (!left->map_token) {
+			map_extent_buffer(left, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&left->map_token, &left->kaddr,
+					&left->map_start, &left->map_len,
+					KM_USER1);
+		}
+
+		this_item_size = btrfs_item_size(left, item);
+		if (this_item_size + sizeof(*item) + push_space > free_space)
 			break;
 		push_items++;
-		push_space += btrfs_item_size(left, item) + sizeof(*item);
+		push_space += this_item_size + sizeof(*item);
+	}
+	if (left->map_token) {
+		unmap_extent_buffer(left, left->map_token, KM_USER1);
+		left->map_token = NULL;
 	}
 
 	if (push_items == 0) {
@@ -1493,11 +1517,23 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 	right_nritems += push_items;
 	btrfs_set_header_nritems(right, right_nritems);
 	push_space = BTRFS_LEAF_DATA_SIZE(root);
+
 	for (i = 0; i < right_nritems; i++) {
 		item = btrfs_item_nr(right, i);
-		btrfs_set_item_offset(right, item, push_space -
-				      btrfs_item_size(right, item));
-		push_space = btrfs_item_offset(right, item);
+		if (!right->map_token) {
+			map_extent_buffer(right, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&right->map_token, &right->kaddr,
+					&right->map_start, &right->map_len,
+					KM_USER1);
+		}
+		push_space -= btrfs_item_size(right, item);
+		btrfs_set_item_offset(right, item, push_space);
+	}
+
+	if (right->map_token) {
+		unmap_extent_buffer(right, right->map_token, KM_USER1);
+		right->map_token = NULL;
 	}
 	left_nritems -= push_items;
 	btrfs_set_header_nritems(left, left_nritems);
@@ -1518,8 +1554,6 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 	} else {
 		free_extent_buffer(right);
 	}
-	if (path->nodes[1])
-		check_node(root, path, 1);
 	return 0;
 }
 /*
@@ -1542,6 +1576,8 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 	u32 right_nritems;
 	int ret = 0;
 	int wret;
+	u32 this_item_size;
+	u32 old_left_item_size;
 
 	slot = path->slots[1];
 	if (slot == 0)
@@ -1550,7 +1586,7 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 		return 1;
 
 	left = read_tree_block(root, btrfs_node_blockptr(path->nodes[1],
-							 slot - 1));
+			       slot - 1), root->leafsize);
 	free_space = btrfs_leaf_free_space(root, left);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
 		free_extent_buffer(left);
@@ -1579,14 +1615,30 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	for (i = 0; i < right_nritems - 1; i++) {
 		item = btrfs_item_nr(right, i);
+		if (!right->map_token) {
+			map_extent_buffer(right, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&right->map_token, &right->kaddr,
+					&right->map_start, &right->map_len,
+					KM_USER1);
+		}
+
 		if (path->slots[0] == i)
 			push_space += data_size + sizeof(*item);
-		if (btrfs_item_size(right, item) + sizeof(*item) + push_space >
-		    free_space)
+
+		this_item_size = btrfs_item_size(right, item);
+		if (this_item_size + sizeof(*item) + push_space > free_space)
 			break;
+
 		push_items++;
-		push_space += btrfs_item_size(right, item) + sizeof(*item);
+		push_space += this_item_size + sizeof(*item);
+	}
+
+	if (right->map_token) {
+		unmap_extent_buffer(right, right->map_token, KM_USER1);
+		right->map_token = NULL;
 	}
+
 	if (push_items == 0) {
 		free_extent_buffer(left);
 		return 1;
@@ -1611,15 +1663,28 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 	old_left_nritems = btrfs_header_nritems(left);
 	BUG_ON(old_left_nritems < 0);
 
+	old_left_item_size = btrfs_item_offset_nr(left, old_left_nritems - 1);
 	for (i = old_left_nritems; i < old_left_nritems + push_items; i++) {
 		u32 ioff;
+
 		item = btrfs_item_nr(left, i);
+		if (!left->map_token) {
+			map_extent_buffer(left, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&left->map_token, &left->kaddr,
+					&left->map_start, &left->map_len,
+					KM_USER1);
+		}
+
 		ioff = btrfs_item_offset(left, item);
 		btrfs_set_item_offset(left, item,
-		      ioff - (BTRFS_LEAF_DATA_SIZE(root) -
-		      btrfs_item_offset_nr(left, old_left_nritems - 1)));
+		      ioff - (BTRFS_LEAF_DATA_SIZE(root) - old_left_item_size));
 	}
 	btrfs_set_header_nritems(left, old_left_nritems + push_items);
+	if (left->map_token) {
+		unmap_extent_buffer(left, left->map_token, KM_USER1);
+		left->map_token = NULL;
+	}
 
 	/* fixup right node */
 	push_space = btrfs_item_offset_nr(right, push_items - 1) -
@@ -1640,9 +1705,21 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	for (i = 0; i < right_nritems; i++) {
 		item = btrfs_item_nr(right, i);
-		btrfs_set_item_offset(right, item, push_space -
-				      btrfs_item_size(right, item));
-		push_space = btrfs_item_offset(right, item);
+
+		if (!right->map_token) {
+			map_extent_buffer(right, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&right->map_token, &right->kaddr,
+					&right->map_start, &right->map_len,
+					KM_USER1);
+		}
+
+		push_space = push_space - btrfs_item_size(right, item);
+		btrfs_set_item_offset(right, item, push_space);
+	}
+	if (right->map_token) {
+		unmap_extent_buffer(right, right->map_token, KM_USER1);
+		right->map_token = NULL;
 	}
 
 	btrfs_mark_buffer_dirty(left);
@@ -1664,8 +1741,6 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 		path->slots[0] -= push_items;
 	}
 	BUG_ON(path->slots[0] < 0);
-	if (path->nodes[1])
-		check_node(root, path, 1);
 	return ret;
 }
 
@@ -1718,13 +1793,13 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 	nritems = btrfs_header_nritems(l);
 	mid = (nritems + 1)/ 2;
 
-	right = btrfs_alloc_free_block(trans, root,
-					      extent_buffer_blocknr(l), 0);
+	right = btrfs_alloc_free_block(trans, root, root->leafsize,
+				       l->start, 0);
 	if (IS_ERR(right))
 		return PTR_ERR(right);
 
 	memset_extent_buffer(right, 0, 0, sizeof(struct btrfs_header));
-	btrfs_set_header_blocknr(right, extent_buffer_blocknr(right));
+	btrfs_set_header_bytenr(right, right->start);
 	btrfs_set_header_generation(right, trans->transid);
 	btrfs_set_header_owner(right, root->root_key.objectid);
 	btrfs_set_header_level(right, 0);
@@ -1740,8 +1815,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 				btrfs_cpu_key_to_disk(&disk_key, ins_key);
 				btrfs_set_header_nritems(right, 0);
 				wret = insert_ptr(trans, root, path,
-						  &disk_key,
-						  extent_buffer_blocknr(right),
+						  &disk_key, right->start,
 						  path->slots[1] + 1, 1);
 				if (wret)
 					ret = wret;
@@ -1762,7 +1836,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 				btrfs_set_header_nritems(right, 0);
 				wret = insert_ptr(trans, root, path,
 						  &disk_key,
-						  extent_buffer_blocknr(right),
+						  right->start,
 						  path->slots[1], 1);
 				if (wret)
 					ret = wret;
@@ -1799,15 +1873,30 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	for (i = 0; i < nritems; i++) {
 		struct btrfs_item *item = btrfs_item_nr(right, i);
-		u32 ioff = btrfs_item_offset(right, item);
+		u32 ioff;
+
+		if (!right->map_token) {
+			map_extent_buffer(right, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&right->map_token, &right->kaddr,
+					&right->map_start, &right->map_len,
+					KM_USER1);
+		}
+
+		ioff = btrfs_item_offset(right, item);
 		btrfs_set_item_offset(right, item, ioff + rt_data_off);
 	}
 
+	if (right->map_token) {
+		unmap_extent_buffer(right, right->map_token, KM_USER1);
+		right->map_token = NULL;
+	}
+
 	btrfs_set_header_nritems(l, mid);
 	ret = 0;
 	btrfs_item_key(right, &disk_key, 0);
-	wret = insert_ptr(trans, root, path, &disk_key,
-			  extent_buffer_blocknr(right), path->slots[1] + 1, 1);
+	wret = insert_ptr(trans, root, path, &disk_key, right->start,
+			  path->slots[1] + 1, 1);
 	if (wret)
 		ret = wret;
 
@@ -1824,19 +1913,17 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 		free_extent_buffer(right);
 
 	BUG_ON(path->slots[0] < 0);
-	check_node(root, path, 1);
-	check_leaf(root, path, 0);
 
 	if (!double_split)
 		return ret;
 
-	right = btrfs_alloc_free_block(trans, root,
-				       extent_buffer_blocknr(l), 0);
+	right = btrfs_alloc_free_block(trans, root, root->leafsize,
+				       l->start, 0);
 	if (IS_ERR(right))
 		return PTR_ERR(right);
 
 	memset_extent_buffer(right, 0, 0, sizeof(struct btrfs_header));
-	btrfs_set_header_blocknr(right, extent_buffer_blocknr(right));
+	btrfs_set_header_bytenr(right, right->start);
 	btrfs_set_header_generation(right, trans->transid);
 	btrfs_set_header_owner(right, root->root_key.objectid);
 	btrfs_set_header_level(right, 0);
@@ -1847,8 +1934,7 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_cpu_key_to_disk(&disk_key, ins_key);
 	btrfs_set_header_nritems(right, 0);
 	wret = insert_ptr(trans, root, path,
-			  &disk_key,
-			  extent_buffer_blocknr(right),
+			  &disk_key, right->start,
 			  path->slots[1], 1);
 	if (wret)
 		ret = wret;
@@ -1860,8 +1946,6 @@ static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
 	free_extent_buffer(path->nodes[0]);
 	path->nodes[0] = right;
 	path->slots[0] = 0;
-	check_node(root, path, 1);
-	check_leaf(root, path, 0);
 	return ret;
 }
 
@@ -1904,9 +1988,24 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans,
 	for (i = slot; i < nritems; i++) {
 		u32 ioff;
 		item = btrfs_item_nr(leaf, i);
+
+		if (!leaf->map_token) {
+			map_extent_buffer(leaf, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&leaf->map_token, &leaf->kaddr,
+					&leaf->map_start, &leaf->map_len,
+					KM_USER1);
+		}
+
 		ioff = btrfs_item_offset(leaf, item);
 		btrfs_set_item_offset(leaf, item, ioff + size_diff);
 	}
+
+	if (leaf->map_token) {
+		unmap_extent_buffer(leaf, leaf->map_token, KM_USER1);
+		leaf->map_token = NULL;
+	}
+
 	/* shift the data */
 	memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) +
 		      data_end + size_diff, btrfs_leaf_data(leaf) +
@@ -1921,7 +2020,6 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans,
 		btrfs_print_leaf(root, leaf);
 		BUG();
 	}
-	check_leaf(root, path, 0);
 	return ret;
 }
 
@@ -1963,10 +2061,23 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans,
 	for (i = slot; i < nritems; i++) {
 		u32 ioff;
 		item = btrfs_item_nr(leaf, i);
+
+		if (!leaf->map_token) {
+			map_extent_buffer(leaf, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&leaf->map_token, &leaf->kaddr,
+					&leaf->map_start, &leaf->map_len,
+					KM_USER1);
+		}
 		ioff = btrfs_item_offset(leaf, item);
 		btrfs_set_item_offset(leaf, item, ioff - data_size);
 	}
 
+	if (leaf->map_token) {
+		unmap_extent_buffer(leaf, leaf->map_token, KM_USER1);
+		leaf->map_token = NULL;
+	}
+
 	/* shift the data */
 	memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) +
 		      data_end - data_size, btrfs_leaf_data(leaf) +
@@ -1983,7 +2094,6 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans,
 		btrfs_print_leaf(root, leaf);
 		BUG();
 	}
-	check_leaf(root, path, 0);
 	return ret;
 }
 
@@ -2046,12 +2156,26 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans,
 		 * item0..itemN ... dataN.offset..dataN.size .. data0.size
 		 */
 		/* first correct the data pointers */
+		WARN_ON(leaf->map_token);
 		for (i = slot; i < nritems; i++) {
 			u32 ioff;
+
 			item = btrfs_item_nr(leaf, i);
+			if (!leaf->map_token) {
+				map_extent_buffer(leaf, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&leaf->map_token, &leaf->kaddr,
+					&leaf->map_start, &leaf->map_len,
+					KM_USER1);
+			}
+
 			ioff = btrfs_item_offset(leaf, item);
 			btrfs_set_item_offset(leaf, item, ioff - data_size);
 		}
+		if (leaf->map_token) {
+			unmap_extent_buffer(leaf, leaf->map_token, KM_USER1);
+			leaf->map_token = NULL;
+		}
 
 		/* shift the items */
 		memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + 1),
@@ -2081,7 +2205,6 @@ int btrfs_insert_empty_item(struct btrfs_trans_handle *trans,
 		btrfs_print_leaf(root, leaf);
 		BUG();
 	}
-	check_leaf(root, path, 0);
 out:
 	return ret;
 }
@@ -2186,10 +2309,24 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 
 		for (i = slot + 1; i < nritems; i++) {
 			u32 ioff;
+
 			item = btrfs_item_nr(leaf, i);
+			if (!leaf->map_token) {
+				map_extent_buffer(leaf, (unsigned long)item,
+					sizeof(struct btrfs_item),
+					&leaf->map_token, &leaf->kaddr,
+					&leaf->map_start, &leaf->map_len,
+					KM_USER1);
+			}
 			ioff = btrfs_item_offset(leaf, item);
 			btrfs_set_item_offset(leaf, item, ioff + dsize);
 		}
+
+		if (leaf->map_token) {
+			unmap_extent_buffer(leaf, leaf->map_token, KM_USER1);
+			leaf->map_token = NULL;
+		}
+
 		memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot),
 			      btrfs_item_nr_offset(slot + 1),
 			      sizeof(struct btrfs_item) *
@@ -2209,8 +2346,7 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 			if (wret)
 				ret = wret;
 			wret = btrfs_free_extent(trans, root,
-						 extent_buffer_blocknr(leaf),
-						 1, 1);
+						 leaf->start, leaf->len, 1);
 			if (wret)
 				ret = wret;
 		}
@@ -2247,7 +2383,8 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 			}
 
 			if (btrfs_header_nritems(leaf) == 0) {
-				u64 blocknr = extent_buffer_blocknr(leaf);
+				u64 bytenr = leaf->start;
+				u32 blocksize = leaf->len;
 
 				clean_tree_block(trans, root, leaf);
 				wait_on_tree_block_writeback(root, leaf);
@@ -2257,8 +2394,8 @@ int btrfs_del_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 					ret = wret;
 
 				free_extent_buffer(leaf);
-				wret = btrfs_free_extent(trans, root, blocknr,
-							 1, 1);
+				wret = btrfs_free_extent(trans, root, bytenr,
+							 blocksize, 1);
 				if (wret)
 					ret = wret;
 			} else {
@@ -2281,7 +2418,7 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 {
 	int slot;
 	int level = 1;
-	u64 blocknr;
+	u64 bytenr;
 	struct extent_buffer *c;
 	struct extent_buffer *next = NULL;
 
@@ -2296,14 +2433,15 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 			continue;
 		}
 
-		blocknr = btrfs_node_blockptr(c, slot);
+		bytenr = btrfs_node_blockptr(c, slot);
 		if (next)
 			free_extent_buffer(next);
 
 		if (path->reada)
 			reada_for_search(root, path, level, slot);
 
-		next = read_tree_block(root, blocknr);
+		next = read_tree_block(root, bytenr,
+				       btrfs_level_size(root, level -1));
 		break;
 	}
 	path->slots[level] = slot;
@@ -2317,7 +2455,8 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 			break;
 		if (path->reada)
 			reada_for_search(root, path, level, 0);
-		next = read_tree_block(root, btrfs_node_blockptr(next, 0));
+		next = read_tree_block(root, btrfs_node_blockptr(next, 0),
+				       btrfs_level_size(root, level - 1));
 	}
 	return 0;
 }
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 256689551eb0..4d05456ec32f 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -96,7 +96,7 @@ struct btrfs_key {
 struct btrfs_header {
 	u8 csum[BTRFS_CSUM_SIZE];
 	u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
-	__le64 blocknr; /* which block this node is supposed to live in */
+	__le64 bytenr; /* which block this node is supposed to live in */
 	__le64 generation;
 	__le64 owner;
 	__le32 nritems;
@@ -122,16 +122,17 @@ struct btrfs_super_block {
 	u8 csum[BTRFS_CSUM_SIZE];
 	/* the first 3 fields must match struct btrfs_header */
 	u8 fsid[16];    /* FS specific uuid */
-	__le64 blocknr; /* this block number */
+	__le64 bytenr; /* this block number */
 	__le64 magic;
 	__le64 generation;
 	__le64 root;
-	__le64 total_blocks;
-	__le64 blocks_used;
+	__le64 total_bytes;
+	__le64 bytes_used;
 	__le64 root_dir_objectid;
 	__le32 sectorsize;
 	__le32 nodesize;
 	__le32 leafsize;
+	u8 root_level;
 } __attribute__ ((__packed__));
 
 /*
@@ -231,13 +232,14 @@ struct btrfs_dir_item {
 struct btrfs_root_item {
 	struct btrfs_inode_item inode;
 	__le64 root_dirid;
-	__le64 blocknr;
-	__le64 block_limit;
-	__le64 blocks_used;
+	__le64 bytenr;
+	__le64 byte_limit;
+	__le64 bytes_used;
 	__le32 flags;
 	__le32 refs;
 	struct btrfs_disk_key drop_progress;
 	u8 drop_level;
+	u8 level;
 } __attribute__ ((__packed__));
 
 #define BTRFS_FILE_EXTENT_REG 0
@@ -250,8 +252,8 @@ struct btrfs_file_extent_item {
 	 * disk space consumed by the extent, checksum blocks are included
 	 * in these numbers
 	 */
-	__le64 disk_blocknr;
-	__le64 disk_num_blocks;
+	__le64 disk_bytenr;
+	__le64 disk_num_bytes;
 	/*
 	 * the logical offset in file blocks (no csums)
 	 * this extent record is for.  This allows a file extent to point
@@ -263,7 +265,7 @@ struct btrfs_file_extent_item {
 	/*
 	 * the logical number of file blocks (no csums included)
 	 */
-	__le64 num_blocks;
+	__le64 num_bytes;
 } __attribute__ ((__packed__));
 
 struct btrfs_csum_item {
@@ -429,6 +431,7 @@ static inline u##bits btrfs_##name(struct extent_buffer *eb,		\
 	int err;							\
 	char *map_token;						\
 	char *kaddr;							\
+	int unmap_on_exit = (eb->map_token == NULL);			\
 	unsigned long map_start;					\
 	unsigned long map_len;						\
 	unsigned long offset = (unsigned long)s +			\
@@ -436,12 +439,13 @@ static inline u##bits btrfs_##name(struct extent_buffer *eb,		\
 	err = map_extent_buffer(eb, offset,				\
 			        sizeof(((type *)0)->member),		\
 				&map_token, &kaddr,			\
-				&map_start, &map_len, KM_USER0);	\
+				&map_start, &map_len, KM_USER1);	\
 	if (!err) {							\
 		__le##bits *tmp = (__le##bits *)(kaddr + offset -	\
 					       map_start);		\
 		u##bits res = le##bits##_to_cpu(*tmp);			\
-		unmap_extent_buffer(eb, map_token, KM_USER0);		\
+		if (unmap_on_exit)					\
+			unmap_extent_buffer(eb, map_token, KM_USER1);	\
 		return res;						\
 	} else {							\
 		__le##bits res;						\
@@ -457,17 +461,19 @@ static inline void btrfs_set_##name(struct extent_buffer *eb,		\
 	char *kaddr;							\
 	unsigned long map_start;					\
 	unsigned long map_len;						\
+	int unmap_on_exit = (eb->map_token == NULL);			\
 	unsigned long offset = (unsigned long)s +			\
 				offsetof(type, member);			\
 	err = map_extent_buffer(eb, offset,				\
 			        sizeof(((type *)0)->member),		\
 				&map_token, &kaddr,			\
-				&map_start, &map_len, KM_USER0);	\
+				&map_start, &map_len, KM_USER1);	\
 	if (!err) {							\
 		__le##bits *tmp = (__le##bits *)(kaddr + offset -	\
 					       map_start);		\
 		*tmp = cpu_to_le##bits(val);				\
-		unmap_extent_buffer(eb, map_token, KM_USER0);		\
+		if (unmap_on_exit)					\
+			unmap_extent_buffer(eb, map_token, KM_USER1);	\
 	} else {							\
 		val = cpu_to_le##bits(val);				\
 		write_eb_member(eb, s, type, member, &val);		\
@@ -483,15 +489,17 @@ static inline u##bits btrfs_##name(struct extent_buffer *eb)		\
 	unsigned long map_start;					\
 	unsigned long map_len;						\
 	unsigned long offset = offsetof(type, member);			\
+	int unmap_on_exit = (eb->map_token == NULL);			\
 	err = map_extent_buffer(eb, offset,				\
 			        sizeof(((type *)0)->member),		\
 				&map_token, &kaddr,			\
-				&map_start, &map_len, KM_USER0);	\
+				&map_start, &map_len, KM_USER1);	\
 	if (!err) {							\
 		__le##bits *tmp = (__le##bits *)(kaddr + offset -	\
 					       map_start);		\
 		u##bits res = le##bits##_to_cpu(*tmp);			\
-		unmap_extent_buffer(eb, map_token, KM_USER0);		\
+		if (unmap_on_exit)					\
+			unmap_extent_buffer(eb, map_token, KM_USER1);	\
 		return res;						\
 	} else {							\
 		__le##bits res;						\
@@ -508,15 +516,17 @@ static inline void btrfs_set_##name(struct extent_buffer *eb,		\
 	unsigned long map_start;					\
 	unsigned long map_len;						\
 	unsigned long offset = offsetof(type, member);			\
+	int unmap_on_exit = (eb->map_token == NULL);			\
 	err = map_extent_buffer(eb, offset,				\
 			        sizeof(((type *)0)->member),		\
 				&map_token, &kaddr,			\
-				&map_start, &map_len, KM_USER0);	\
+				&map_start, &map_len, KM_USER1);	\
 	if (!err) {							\
 		__le##bits *tmp = (__le##bits *)(kaddr + offset -	\
 					       map_start);		\
 		*tmp = cpu_to_le##bits(val);				\
-		unmap_extent_buffer(eb, map_token, KM_USER0);		\
+		if (unmap_on_exit)					\
+			unmap_extent_buffer(eb, map_token, KM_USER1);	\
 	} else {							\
 		val = cpu_to_le##bits(val);				\
 		write_eb_member(eb, NULL, type, member, &val);		\
@@ -769,7 +779,7 @@ static inline void btrfs_set_key_type(struct btrfs_key *key, u8 val)
 }
 
 /* struct btrfs_header */
-BTRFS_SETGET_HEADER_FUNCS(header_blocknr, struct btrfs_header, blocknr, 64);
+BTRFS_SETGET_HEADER_FUNCS(header_bytenr, struct btrfs_header, bytenr, 64);
 BTRFS_SETGET_HEADER_FUNCS(header_generation, struct btrfs_header,
 			  generation, 64);
 BTRFS_SETGET_HEADER_FUNCS(header_owner, struct btrfs_header, owner, 64);
@@ -817,24 +827,28 @@ static inline int btrfs_is_leaf(struct extent_buffer *eb)
 
 /* struct btrfs_root_item */
 BTRFS_SETGET_FUNCS(disk_root_refs, struct btrfs_root_item, refs, 32);
-BTRFS_SETGET_FUNCS(disk_root_blocknr, struct btrfs_root_item, blocknr, 64);
+BTRFS_SETGET_FUNCS(disk_root_bytenr, struct btrfs_root_item, bytenr, 64);
+BTRFS_SETGET_FUNCS(disk_root_level, struct btrfs_root_item, level, 8);
 
-BTRFS_SETGET_STACK_FUNCS(root_blocknr, struct btrfs_root_item, blocknr, 64);
+BTRFS_SETGET_STACK_FUNCS(root_bytenr, struct btrfs_root_item, bytenr, 64);
+BTRFS_SETGET_STACK_FUNCS(root_level, struct btrfs_root_item, level, 8);
 BTRFS_SETGET_STACK_FUNCS(root_dirid, struct btrfs_root_item, root_dirid, 64);
 BTRFS_SETGET_STACK_FUNCS(root_refs, struct btrfs_root_item, refs, 32);
 BTRFS_SETGET_STACK_FUNCS(root_flags, struct btrfs_root_item, flags, 32);
-BTRFS_SETGET_STACK_FUNCS(root_used, struct btrfs_root_item, blocks_used, 64);
-BTRFS_SETGET_STACK_FUNCS(root_limit, struct btrfs_root_item, block_limit, 64);
+BTRFS_SETGET_STACK_FUNCS(root_used, struct btrfs_root_item, bytes_used, 64);
+BTRFS_SETGET_STACK_FUNCS(root_limit, struct btrfs_root_item, byte_limit, 64);
 
 /* struct btrfs_super_block */
-BTRFS_SETGET_STACK_FUNCS(super_blocknr, struct btrfs_super_block, blocknr, 64);
+BTRFS_SETGET_STACK_FUNCS(super_bytenr, struct btrfs_super_block, bytenr, 64);
 BTRFS_SETGET_STACK_FUNCS(super_generation, struct btrfs_super_block,
 			 generation, 64);
 BTRFS_SETGET_STACK_FUNCS(super_root, struct btrfs_super_block, root, 64);
-BTRFS_SETGET_STACK_FUNCS(super_total_blocks, struct btrfs_super_block,
-		   total_blocks, 64);
-BTRFS_SETGET_STACK_FUNCS(super_blocks_used, struct btrfs_super_block,
-		   blocks_used, 64);
+BTRFS_SETGET_STACK_FUNCS(super_root_level, struct btrfs_super_block,
+			 root_level, 8);
+BTRFS_SETGET_STACK_FUNCS(super_total_bytes, struct btrfs_super_block,
+			 total_bytes, 64);
+BTRFS_SETGET_STACK_FUNCS(super_bytes_used, struct btrfs_super_block,
+			 bytes_used, 64);
 BTRFS_SETGET_STACK_FUNCS(super_sectorsize, struct btrfs_super_block,
 			 sectorsize, 32);
 BTRFS_SETGET_STACK_FUNCS(super_nodesize, struct btrfs_super_block,
@@ -856,33 +870,33 @@ static inline unsigned long btrfs_file_extent_inline_start(struct
 						   btrfs_file_extent_item *e)
 {
 	unsigned long offset = (unsigned long)e;
-	offset += offsetof(struct btrfs_file_extent_item, disk_blocknr);
+	offset += offsetof(struct btrfs_file_extent_item, disk_bytenr);
 	return offset;
 }
 
 static inline u32 btrfs_file_extent_calc_inline_size(u32 datasize)
 {
-	return offsetof(struct btrfs_file_extent_item, disk_blocknr) + datasize;
+	return offsetof(struct btrfs_file_extent_item, disk_bytenr) + datasize;
 }
 
 static inline u32 btrfs_file_extent_inline_len(struct extent_buffer *eb,
 					       struct btrfs_item *e)
 {
 	unsigned long offset;
-	offset = offsetof(struct btrfs_file_extent_item, disk_blocknr);
+	offset = offsetof(struct btrfs_file_extent_item, disk_bytenr);
 	return btrfs_item_size(eb, e) - offset;
 }
 
-BTRFS_SETGET_FUNCS(file_extent_disk_blocknr, struct btrfs_file_extent_item,
-		   disk_blocknr, 64);
+BTRFS_SETGET_FUNCS(file_extent_disk_bytenr, struct btrfs_file_extent_item,
+		   disk_bytenr, 64);
 BTRFS_SETGET_FUNCS(file_extent_generation, struct btrfs_file_extent_item,
 		   generation, 64);
-BTRFS_SETGET_FUNCS(file_extent_disk_num_blocks, struct btrfs_file_extent_item,
-		   disk_num_blocks, 64);
+BTRFS_SETGET_FUNCS(file_extent_disk_num_bytes, struct btrfs_file_extent_item,
+		   disk_num_bytes, 64);
 BTRFS_SETGET_FUNCS(file_extent_offset, struct btrfs_file_extent_item,
 		  offset, 64);
-BTRFS_SETGET_FUNCS(file_extent_num_blocks, struct btrfs_file_extent_item,
-		   num_blocks, 64);
+BTRFS_SETGET_FUNCS(file_extent_num_bytes, struct btrfs_file_extent_item,
+		   num_bytes, 64);
 
 static inline struct btrfs_root *btrfs_sb(struct super_block *sb)
 {
@@ -906,6 +920,12 @@ static inline int btrfs_set_root_name(struct btrfs_root *root,
 	return 0;
 }
 
+static inline u32 btrfs_level_size(struct btrfs_root *root, int level) {
+	if (level == 0)
+		return root->leafsize;
+	return root->nodesize;
+}
+
 /* helper function to cast into the data area of the leaf. */
 #define btrfs_item_ptr(leaf, slot, type) \
 	((type *)(btrfs_leaf_data(leaf) + \
@@ -927,7 +947,7 @@ int btrfs_extent_post_op(struct btrfs_trans_handle *trans,
 int btrfs_copy_pinned(struct btrfs_root *root, struct extent_map_tree *copy);
 struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
 							 btrfs_fs_info *info,
-							 u64 blocknr);
+							 u64 bytenr);
 struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
 						 struct btrfs_block_group_cache
 						 *hint, u64 search_start,
@@ -935,22 +955,22 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
 int btrfs_inc_root_ref(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root);
 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
-					    struct btrfs_root *root, u64 hint,
-					    u64 empty_size);
+					    struct btrfs_root *root, u32 size,
+					    u64 hint, u64 empty_size);
 int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root, u64 owner,
-		       u64 num_blocks, u64 empty_size, u64 search_start,
+		       u64 num_bytes, u64 empty_size, u64 search_start,
 		       u64 search_end, struct btrfs_key *ins, int data);
 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		  struct extent_buffer *buf);
 int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, u64 blocknr, u64 num_blocks, int pin);
+		      *root, u64 bytenr, u64 num_bytes, int pin);
 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
 			       struct btrfs_root *root,
 			       struct extent_map_tree *unpin);
 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root,
-				u64 blocknr, u64 num_blocks);
+				u64 bytenr, u64 num_bytes);
 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
 				    struct btrfs_root *root);
 int btrfs_free_block_groups(struct btrfs_fs_info *info);
@@ -1040,12 +1060,12 @@ int btrfs_lookup_inode(struct btrfs_trans_handle *trans, struct btrfs_root
 int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
 			       struct btrfs_root *root,
 			       u64 objectid, u64 pos, u64 offset,
-			       u64 disk_num_blocks,
-			       u64 num_blocks);
+			       u64 disk_num_bytes,
+			       u64 num_bytes);
 int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root,
 			     struct btrfs_path *path, u64 objectid,
-			     u64 blocknr, int mod);
+			     u64 bytenr, int mod);
 int btrfs_csum_file_block(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *root,
 			  u64 objectid, u64 offset,
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 2b86a1d779b7..fad9298c6962 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -28,6 +28,7 @@
 #include "disk-io.h"
 #include "transaction.h"
 #include "btrfs_inode.h"
+#include "print-tree.h"
 
 #if 0
 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
@@ -43,26 +44,25 @@ static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
 #endif
 
 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
-					    u64 blocknr)
+					    u64 bytenr, u32 blocksize)
 {
 	struct inode *btree_inode = root->fs_info->btree_inode;
 	struct extent_buffer *eb;
 	eb = find_extent_buffer(&BTRFS_I(btree_inode)->extent_tree,
-				   blocknr * root->sectorsize,
-				   root->sectorsize, GFP_NOFS);
+				bytenr, blocksize, GFP_NOFS);
 	if (eb)
 		eb->alloc_addr = (unsigned long)__builtin_return_address(0);
 	return eb;
 }
 
 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
-						 u64 blocknr)
+						 u64 bytenr, u32 blocksize)
 {
 	struct inode *btree_inode = root->fs_info->btree_inode;
 	struct extent_buffer *eb;
+
 	eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->extent_tree,
-				   blocknr * root->sectorsize,
-				   root->sectorsize, GFP_NOFS);
+				 bytenr, blocksize, GFP_NOFS);
 	eb->alloc_addr = (unsigned long)__builtin_return_address(0);
 	return eb;
 }
@@ -208,13 +208,13 @@ static struct address_space_operations btree_aops = {
 	.sync_page	= block_sync_page,
 };
 
-int readahead_tree_block(struct btrfs_root *root, u64 blocknr)
+int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize)
 {
 	struct extent_buffer *buf = NULL;
 	struct inode *btree_inode = root->fs_info->btree_inode;
 	int ret = 0;
 
-	buf = btrfs_find_create_tree_block(root, blocknr);
+	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
 	if (!buf)
 		return 0;
 	read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
@@ -223,12 +223,13 @@ int readahead_tree_block(struct btrfs_root *root, u64 blocknr)
 	return ret;
 }
 
-struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr)
+struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
+				      u32 blocksize)
 {
 	struct extent_buffer *buf = NULL;
 	struct inode *btree_inode = root->fs_info->btree_inode;
 
-	buf = btrfs_find_create_tree_block(root, blocknr);
+	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
 	if (!buf)
 		return NULL;
 	read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
@@ -261,7 +262,7 @@ int set_tree_block_dirty(struct btrfs_root *root, struct extent_buffer *buf)
 	return 0;
 }
 
-static int __setup_root(int blocksize,
+static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 			struct btrfs_root *root,
 			struct btrfs_fs_info *fs_info,
 			u64 objectid)
@@ -269,9 +270,9 @@ static int __setup_root(int blocksize,
 	root->node = NULL;
 	root->inode = NULL;
 	root->commit_root = NULL;
-	root->sectorsize = blocksize;
-	root->nodesize = blocksize;
-	root->leafsize = blocksize;
+	root->sectorsize = sectorsize;
+	root->nodesize = nodesize;
+	root->leafsize = leafsize;
 	root->ref_cows = 0;
 	root->fs_info = fs_info;
 	root->objectid = objectid;
@@ -291,21 +292,23 @@ static int __setup_root(int blocksize,
 	return 0;
 }
 
-static int find_and_setup_root(int blocksize,
-			       struct btrfs_root *tree_root,
+static int find_and_setup_root(struct btrfs_root *tree_root,
 			       struct btrfs_fs_info *fs_info,
 			       u64 objectid,
 			       struct btrfs_root *root)
 {
 	int ret;
+	u32 blocksize;
 
-	__setup_root(blocksize, root, fs_info, objectid);
+	__setup_root(tree_root->nodesize, tree_root->leafsize,
+		     tree_root->sectorsize, root, fs_info, objectid);
 	ret = btrfs_find_last_root(tree_root, objectid,
 				   &root->root_item, &root->root_key);
 	BUG_ON(ret);
 
-	root->node = read_tree_block(root,
-				     btrfs_root_blocknr(&root->root_item));
+	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
+	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
+				     blocksize);
 	BUG_ON(!root->node);
 	return 0;
 }
@@ -318,14 +321,14 @@ struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
 	struct btrfs_path *path;
 	struct extent_buffer *l;
 	u64 highest_inode;
+	u32 blocksize;
 	int ret = 0;
 
 	root = kzalloc(sizeof(*root), GFP_NOFS);
 	if (!root)
 		return ERR_PTR(-ENOMEM);
 	if (location->offset == (u64)-1) {
-		ret = find_and_setup_root(fs_info->sb->s_blocksize,
-					  fs_info->tree_root, fs_info,
+		ret = find_and_setup_root(tree_root, fs_info,
 					  location->objectid, root);
 		if (ret) {
 			kfree(root);
@@ -334,7 +337,8 @@ struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
 		goto insert;
 	}
 
-	__setup_root(fs_info->sb->s_blocksize, root, fs_info,
+	__setup_root(tree_root->nodesize, tree_root->leafsize,
+		     tree_root->sectorsize, root, fs_info,
 		     location->objectid);
 
 	path = btrfs_alloc_path();
@@ -357,8 +361,9 @@ out:
 		kfree(root);
 		return ERR_PTR(ret);
 	}
-	root->node = read_tree_block(root,
-				     btrfs_root_blocknr(&root->root_item));
+	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
+	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
+				     blocksize);
 	BUG_ON(!root->node);
 insert:
 	root->ref_cows = 1;
@@ -418,6 +423,10 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 
 struct btrfs_root *open_ctree(struct super_block *sb)
 {
+	u32 sectorsize;
+	u32 nodesize;
+	u32 leafsize;
+	u32 blocksize;
 	struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
 						 GFP_NOFS);
 	struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
@@ -474,12 +483,12 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	mutex_init(&fs_info->trans_mutex);
 	mutex_init(&fs_info->fs_mutex);
 
-	__setup_root(sb->s_blocksize, tree_root,
+	__setup_root(512, 512, 512, tree_root,
 		     fs_info, BTRFS_ROOT_TREE_OBJECTID);
 
 	fs_info->sb_buffer = read_tree_block(tree_root,
-					     BTRFS_SUPER_INFO_OFFSET /
-					     sb->s_blocksize);
+					     BTRFS_SUPER_INFO_OFFSET,
+					     512);
 
 	if (!fs_info->sb_buffer)
 		goto fail_iput;
@@ -494,9 +503,15 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	if (!btrfs_super_root(disk_super))
 		goto fail_sb_buffer;
 
+	nodesize = btrfs_super_nodesize(disk_super);
+	leafsize = btrfs_super_leafsize(disk_super);
+	sectorsize = btrfs_super_sectorsize(disk_super);
+	tree_root->nodesize = nodesize;
+	tree_root->leafsize = leafsize;
+	tree_root->sectorsize = sectorsize;
+
 	i_size_write(fs_info->btree_inode,
-		     btrfs_super_total_blocks(disk_super) <<
-		     fs_info->btree_inode->i_blkbits);
+		     btrfs_super_total_bytes(disk_super));
 
 
 	if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
@@ -504,13 +519,22 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 		printk("btrfs: valid FS not found on %s\n", sb->s_id);
 		goto fail_sb_buffer;
 	}
+	blocksize = btrfs_level_size(tree_root,
+				     btrfs_super_root_level(disk_super));
 	tree_root->node = read_tree_block(tree_root,
-					  btrfs_super_root(disk_super));
+					  btrfs_super_root(disk_super),
+					  blocksize);
 	if (!tree_root->node)
 		goto fail_sb_buffer;
 
+#if 0
+	btrfs_print_leaf(tree_root, tree_root->node);
+	err = -EIO;
+	goto fail_tree_root;
+#endif
 	mutex_lock(&fs_info->fs_mutex);
-	ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
+
+	ret = find_and_setup_root(tree_root, fs_info,
 				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
 	if (ret) {
 		mutex_unlock(&fs_info->fs_mutex);
@@ -611,11 +635,11 @@ int close_ctree(struct btrfs_root *root)
 		free_extent_buffer(fs_info->tree_root->node);
 
 	free_extent_buffer(fs_info->sb_buffer);
-	truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
-	iput(fs_info->btree_inode);
 
 	btrfs_free_block_groups(root->fs_info);
 	del_fs_roots(fs_info);
+	truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
+	iput(fs_info->btree_inode);
 	kfree(fs_info->extent_root);
 	kfree(fs_info->tree_root);
 	return 0;
@@ -642,7 +666,7 @@ void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
 
 	if (transid != root->fs_info->generation) {
 		printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
-			(unsigned long long)extent_buffer_blocknr(buf),
+			(unsigned long long)buf->start,
 			transid, root->fs_info->generation);
 		WARN_ON(1);
 	}
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 70d9413c599f..fd4db5f810cf 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -21,10 +21,11 @@
 
 #define BTRFS_SUPER_INFO_OFFSET (16 * 1024)
 
-struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr);
-int readahead_tree_block(struct btrfs_root *root, u64 blocknr);
+struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
+				      u32 blocksize);
+int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize);
 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
-						   u64 blocknr);
+						   u64 bytenr, u32 blocksize);
 int clean_tree_block(struct btrfs_trans_handle *trans,
 		     struct btrfs_root *root, struct extent_buffer *buf);
 struct btrfs_root *open_ctree(struct super_block *sb);
@@ -32,7 +33,7 @@ int close_ctree(struct btrfs_root *root);
 int write_ctree_super(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root);
 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
-					    u64 blocknr);
+					    u64 bytenr, u32 blocksize);
 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 				      struct btrfs_key *location,
 				      const char *name, int namelen);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 477466d167a4..1be8f9f04a13 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -129,7 +129,7 @@ err:
 
 struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
 							 btrfs_fs_info *info,
-							 u64 blocknr)
+							 u64 bytenr)
 {
 	struct extent_map_tree *block_group_cache;
 	struct btrfs_block_group_cache *block_group = NULL;
@@ -140,7 +140,7 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
 
 	block_group_cache = &info->block_group_cache;
 	ret = find_first_extent_bit(block_group_cache,
-				    blocknr, &start, &end,
+				    bytenr, &start, &end,
 				    BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA);
 	if (ret) {
 		return NULL;
@@ -152,7 +152,7 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
 	block_group = (struct btrfs_block_group_cache *)ptr;
 
 
-	if (block_group->key.objectid <= blocknr && blocknr <=
+	if (block_group->key.objectid <= bytenr && bytenr <=
 	    block_group->key.objectid + block_group->key.offset)
 		return block_group;
 
@@ -315,7 +315,7 @@ found:
 
 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root,
-				u64 blocknr, u64 num_blocks)
+				u64 bytenr, u64 num_bytes)
 {
 	struct btrfs_path *path;
 	int ret;
@@ -324,13 +324,14 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 	struct btrfs_extent_item *item;
 	u32 refs;
 
+	WARN_ON(num_bytes < root->sectorsize);
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
 
-	key.objectid = blocknr;
+	key.objectid = bytenr;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
-	key.offset = num_blocks;
+	key.offset = num_bytes;
 	ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
 				0, 1);
 	if (ret < 0)
@@ -361,8 +362,8 @@ int btrfs_extent_post_op(struct btrfs_trans_handle *trans,
 }
 
 static int lookup_extent_ref(struct btrfs_trans_handle *trans,
-			     struct btrfs_root *root, u64 blocknr,
-			     u64 num_blocks, u32 *refs)
+			     struct btrfs_root *root, u64 bytenr,
+			     u64 num_bytes, u32 *refs)
 {
 	struct btrfs_path *path;
 	int ret;
@@ -370,9 +371,10 @@ static int lookup_extent_ref(struct btrfs_trans_handle *trans,
 	struct extent_buffer *l;
 	struct btrfs_extent_item *item;
 
+	WARN_ON(num_bytes < root->sectorsize);
 	path = btrfs_alloc_path();
-	key.objectid = blocknr;
-	key.offset = num_blocks;
+	key.objectid = bytenr;
+	key.offset = num_bytes;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 	ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
 				0, 0);
@@ -380,7 +382,7 @@ static int lookup_extent_ref(struct btrfs_trans_handle *trans,
 		goto out;
 	if (ret != 0) {
 		btrfs_print_leaf(root, path->nodes[0]);
-		printk("failed to find block number %Lu\n", blocknr);
+		printk("failed to find block number %Lu\n", bytenr);
 		BUG();
 	}
 	l = path->nodes[0];
@@ -394,19 +396,19 @@ out:
 int btrfs_inc_root_ref(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root)
 {
-	return btrfs_inc_extent_ref(trans, root,
-				    extent_buffer_blocknr(root->node), 1);
+	return btrfs_inc_extent_ref(trans, root, root->node->start,
+				    root->node->len);
 }
 
 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		  struct extent_buffer *buf)
 {
-	u64 blocknr;
+	u64 bytenr;
 	u32 nritems;
 	struct btrfs_key key;
 	struct btrfs_file_extent_item *fi;
 	int i;
-	int leaf;
+	int level;
 	int ret;
 	int faili;
 	int err;
@@ -414,11 +416,11 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	if (!root->ref_cows)
 		return 0;
 
-	leaf = btrfs_is_leaf(buf);
+	level = btrfs_header_level(buf);
 	nritems = btrfs_header_nritems(buf);
 	for (i = 0; i < nritems; i++) {
-		if (leaf) {
-			u64 disk_blocknr;
+		if (level == 0) {
+			u64 disk_bytenr;
 			btrfs_item_key_to_cpu(buf, &key, i);
 			if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
 				continue;
@@ -427,18 +429,19 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 			if (btrfs_file_extent_type(buf, fi) ==
 			    BTRFS_FILE_EXTENT_INLINE)
 				continue;
-			disk_blocknr = btrfs_file_extent_disk_blocknr(buf, fi);
-			if (disk_blocknr == 0)
+			disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
+			if (disk_bytenr == 0)
 				continue;
-			ret = btrfs_inc_extent_ref(trans, root, disk_blocknr,
-				    btrfs_file_extent_disk_num_blocks(buf, fi));
+			ret = btrfs_inc_extent_ref(trans, root, disk_bytenr,
+				    btrfs_file_extent_disk_num_bytes(buf, fi));
 			if (ret) {
 				faili = i;
 				goto fail;
 			}
 		} else {
-			blocknr = btrfs_node_blockptr(buf, i);
-			ret = btrfs_inc_extent_ref(trans, root, blocknr, 1);
+			bytenr = btrfs_node_blockptr(buf, i);
+			ret = btrfs_inc_extent_ref(trans, root, bytenr,
+					   btrfs_level_size(root, level - 1));
 			if (ret) {
 				faili = i;
 				goto fail;
@@ -449,8 +452,8 @@ int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 fail:
 	WARN_ON(1);
 	for (i =0; i < faili; i++) {
-		if (leaf) {
-			u64 disk_blocknr;
+		if (level == 0) {
+			u64 disk_bytenr;
 			btrfs_item_key_to_cpu(buf, &key, i);
 			if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
 				continue;
@@ -459,16 +462,17 @@ fail:
 			if (btrfs_file_extent_type(buf, fi) ==
 			    BTRFS_FILE_EXTENT_INLINE)
 				continue;
-			disk_blocknr = btrfs_file_extent_disk_blocknr(buf, fi);
-			if (disk_blocknr == 0)
+			disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
+			if (disk_bytenr == 0)
 				continue;
-			err = btrfs_free_extent(trans, root, disk_blocknr,
-				    btrfs_file_extent_disk_num_blocks(buf,
+			err = btrfs_free_extent(trans, root, disk_bytenr,
+				    btrfs_file_extent_disk_num_bytes(buf,
 								      fi), 0);
 			BUG_ON(err);
 		} else {
-			blocknr = btrfs_node_blockptr(buf, i);
-			err = btrfs_free_extent(trans, root, blocknr, 1, 0);
+			bytenr = btrfs_node_blockptr(buf, i);
+			err = btrfs_free_extent(trans, root, bytenr,
+					btrfs_level_size(root, level - 1), 0);
 			BUG_ON(err);
 		}
 	}
@@ -558,31 +562,31 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
 
 static int update_block_group(struct btrfs_trans_handle *trans,
 			      struct btrfs_root *root,
-			      u64 blocknr, u64 num, int alloc, int mark_free,
-			      int data)
+			      u64 bytenr, u64 num_bytes, int alloc,
+			      int mark_free, int data)
 {
 	struct btrfs_block_group_cache *cache;
 	struct btrfs_fs_info *info = root->fs_info;
-	u64 total = num;
+	u64 total = num_bytes;
 	u64 old_val;
-	u64 block_in_group;
+	u64 byte_in_group;
 	u64 start;
 	u64 end;
 
 	while(total) {
-		cache = btrfs_lookup_block_group(info, blocknr);
+		cache = btrfs_lookup_block_group(info, bytenr);
 		if (!cache) {
 			return -1;
 		}
-		block_in_group = blocknr - cache->key.objectid;
-		WARN_ON(block_in_group > cache->key.offset);
+		byte_in_group = bytenr - cache->key.objectid;
+		WARN_ON(byte_in_group > cache->key.offset);
 		start = cache->key.objectid;
 		end = start + cache->key.offset - 1;
 		set_extent_bits(&info->block_group_cache, start, end,
 				BLOCK_GROUP_DIRTY, GFP_NOFS);
 
 		old_val = btrfs_block_group_used(&cache->item);
-		num = min(total, cache->key.offset - block_in_group);
+		num_bytes = min(total, cache->key.offset - byte_in_group);
 		if (alloc) {
 			if (cache->data != data &&
 			    old_val < (cache->key.offset >> 1)) {
@@ -608,18 +612,18 @@ static int update_block_group(struct btrfs_trans_handle *trans,
 						start, end, bit_to_set,
 						GFP_NOFS);
 			}
-			old_val += num;
+			old_val += num_bytes;
 		} else {
-			old_val -= num;
+			old_val -= num_bytes;
 			if (mark_free) {
 				set_extent_dirty(&info->free_space_cache,
-						 blocknr, blocknr + num - 1,
+						 bytenr, bytenr + num_bytes - 1,
 						 GFP_NOFS);
 			}
 		}
 		btrfs_set_block_group_used(&cache->item, old_val);
-		total -= num;
-		blocknr += num;
+		total -= num_bytes;
+		bytenr += num_bytes;
 	}
 	return 0;
 }
@@ -701,13 +705,14 @@ static int finish_current_insert(struct btrfs_trans_handle *trans, struct
 	return 0;
 }
 
-static int pin_down_block(struct btrfs_root *root, u64 blocknr, int pending)
+static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes,
+			  int pending)
 {
 	int err = 0;
 	struct extent_buffer *buf;
 
 	if (!pending) {
-		buf = btrfs_find_tree_block(root, blocknr);
+		buf = btrfs_find_tree_block(root, bytenr, num_bytes);
 		if (buf) {
 			if (btrfs_buffer_uptodate(buf)) {
 				u64 transid =
@@ -720,10 +725,11 @@ static int pin_down_block(struct btrfs_root *root, u64 blocknr, int pending)
 			free_extent_buffer(buf);
 		}
 		set_extent_dirty(&root->fs_info->pinned_extents,
-				 blocknr, blocknr, GFP_NOFS);
+				 bytenr, bytenr + num_bytes - 1, GFP_NOFS);
 	} else {
 		set_extent_bits(&root->fs_info->pending_del,
-				  blocknr, blocknr, EXTENT_LOCKED, GFP_NOFS);
+				bytenr, bytenr + num_bytes - 1,
+				EXTENT_LOCKED, GFP_NOFS);
 	}
 	BUG_ON(err < 0);
 	return 0;
@@ -733,7 +739,7 @@ static int pin_down_block(struct btrfs_root *root, u64 blocknr, int pending)
  * remove an extent from the root, returns 0 on success
  */
 static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
-			 *root, u64 blocknr, u64 num_blocks, int pin,
+			 *root, u64 bytenr, u64 num_bytes, int pin,
 			 int mark_free)
 {
 	struct btrfs_path *path;
@@ -745,9 +751,9 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	struct btrfs_extent_item *ei;
 	u32 refs;
 
-	key.objectid = blocknr;
+	key.objectid = bytenr;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
-	key.offset = num_blocks;
+	key.offset = num_bytes;
 
 	path = btrfs_alloc_path();
 	if (!path)
@@ -768,28 +774,29 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_mark_buffer_dirty(leaf);
 
 	if (refs == 0) {
-		u64 super_blocks_used, root_blocks_used;
+		u64 super_used;
+		u64 root_used;
 
 		if (pin) {
-			ret = pin_down_block(root, blocknr, 0);
+			ret = pin_down_bytes(root, bytenr, num_bytes, 0);
 			BUG_ON(ret);
 		}
 
 		/* block accounting for super block */
-		super_blocks_used = btrfs_super_blocks_used(&info->super_copy);
-		btrfs_set_super_blocks_used(&info->super_copy,
-					    super_blocks_used - num_blocks);
+		super_used = btrfs_super_bytes_used(&info->super_copy);
+		btrfs_set_super_bytes_used(&info->super_copy,
+					   super_used - num_bytes);
 
 		/* block accounting for root item */
-		root_blocks_used = btrfs_root_used(&root->root_item);
+		root_used = btrfs_root_used(&root->root_item);
 		btrfs_set_root_used(&root->root_item,
-					   root_blocks_used - num_blocks);
+					   root_used - num_bytes);
 
 		ret = btrfs_del_item(trans, extent_root, path);
 		if (ret) {
 			return ret;
 		}
-		ret = update_block_group(trans, root, blocknr, num_blocks, 0,
+		ret = update_block_group(trans, root, bytenr, num_bytes, 0,
 					 mark_free, 0);
 		BUG_ON(ret);
 	}
@@ -836,17 +843,18 @@ static int del_pending_extents(struct btrfs_trans_handle *trans, struct
  * remove an extent from the root, returns 0 on success
  */
 int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, u64 blocknr, u64 num_blocks, int pin)
+		      *root, u64 bytenr, u64 num_bytes, int pin)
 {
 	struct btrfs_root *extent_root = root->fs_info->extent_root;
 	int pending_ret;
 	int ret;
 
+	WARN_ON(num_bytes < root->sectorsize);
 	if (root == extent_root) {
-		pin_down_block(root, blocknr, 1);
+		pin_down_bytes(root, bytenr, num_bytes, 1);
 		return 0;
 	}
-	ret = __free_extent(trans, root, blocknr, num_blocks, pin, pin == 0);
+	ret = __free_extent(trans, root, bytenr, num_bytes, pin, pin == 0);
 	pending_ret = del_pending_extents(trans, root->fs_info->extent_root);
 	return ret ? ret : pending_ret;
 }
@@ -860,8 +868,8 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
  * Any available blocks before search_start are skipped.
  */
 static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
-			    *orig_root, u64 num_blocks, u64 empty_size,
-			    u64 search_start, u64 search_end, u64 hint_block,
+			    *orig_root, u64 num_bytes, u64 empty_size,
+			    u64 search_start, u64 search_end, u64 hint_byte,
 			    struct btrfs_key *ins, u64 exclude_start,
 			    u64 exclude_nr, int data)
 {
@@ -870,30 +878,29 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 	int ret;
 	u64 hole_size = 0;
 	int slot = 0;
-	u64 last_block = 0;
+	u64 last_byte = 0;
 	u64 orig_search_start = search_start;
 	int start_found;
 	struct extent_buffer *l;
 	struct btrfs_root * root = orig_root->fs_info->extent_root;
 	struct btrfs_fs_info *info = root->fs_info;
-	int total_needed = num_blocks;
+	u64 total_needed = num_bytes;
 	int level;
 	struct btrfs_block_group_cache *block_group;
 	int full_scan = 0;
 	int wrapped = 0;
-	u64 cached_search_start = 0;
 
-	WARN_ON(num_blocks < 1);
+	WARN_ON(num_bytes < root->sectorsize);
 	btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
 
 	level = btrfs_header_level(root->node);
 
 	if (search_end == (u64)-1)
-		search_end = btrfs_super_total_blocks(&info->super_copy);
-	if (hint_block) {
-		block_group = btrfs_lookup_block_group(info, hint_block);
+		search_end = btrfs_super_total_bytes(&info->super_copy);
+	if (hint_byte) {
+		block_group = btrfs_lookup_block_group(info, hint_byte);
 		block_group = btrfs_find_block_group(root, block_group,
-						     hint_block, data, 1);
+						     hint_byte, data, 1);
 	} else {
 		block_group = btrfs_find_block_group(root,
 						     trans->block_group, 0,
@@ -906,7 +913,6 @@ static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 check_failed:
 	search_start = find_search_start(root, &block_group,
 					 search_start, total_needed, data);
-	cached_search_start = search_start;
 
 	btrfs_init_path(path);
 	ins->objectid = search_start;
@@ -958,27 +964,27 @@ check_failed:
 				start_found = 1;
 				goto check_pending;
 			}
-			ins->objectid = last_block > search_start ?
-					last_block : search_start;
+			ins->objectid = last_byte > search_start ?
+					last_byte : search_start;
 			ins->offset = search_end - ins->objectid;
 			goto check_pending;
 		}
 		btrfs_item_key_to_cpu(l, &key, slot);
 
-		if (key.objectid >= search_start && key.objectid > last_block &&
+		if (key.objectid >= search_start && key.objectid > last_byte &&
 		    start_found) {
-			if (last_block < search_start)
-				last_block = search_start;
-			hole_size = key.objectid - last_block;
-			if (hole_size >= num_blocks) {
-				ins->objectid = last_block;
+			if (last_byte < search_start)
+				last_byte = search_start;
+			hole_size = key.objectid - last_byte;
+			if (hole_size >= num_bytes) {
+				ins->objectid = last_byte;
 				ins->offset = hole_size;
 				goto check_pending;
 			}
 		}
 		if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY) {
 			if (!start_found) {
-				last_block = key.objectid;
+				last_byte = key.objectid;
 				start_found = 1;
 			}
 			goto next;
@@ -986,9 +992,9 @@ check_failed:
 
 
 		start_found = 1;
-		last_block = key.objectid + key.offset;
+		last_byte = key.objectid + key.offset;
 
-		if (!full_scan && last_block >= block_group->key.objectid +
+		if (!full_scan && last_byte >= block_group->key.objectid +
 		    block_group->key.offset) {
 			btrfs_release_path(root, path);
 			search_start = block_group->key.objectid +
@@ -1006,20 +1012,20 @@ check_pending:
 	btrfs_release_path(root, path);
 	BUG_ON(ins->objectid < search_start);
 
-	if (ins->objectid + num_blocks >= search_end)
+	if (ins->objectid + num_bytes >= search_end)
 		goto enospc;
 
 	if (test_range_bit(&info->extent_ins, ins->objectid,
-			   ins->objectid + num_blocks -1, EXTENT_LOCKED, 0)) {
-		search_start = ins->objectid + num_blocks;
+			   ins->objectid + num_bytes -1, EXTENT_LOCKED, 0)) {
+		search_start = ins->objectid + num_bytes;
 		goto new_group;
 	}
 	if (test_range_bit(&info->pinned_extents, ins->objectid,
-			   ins->objectid + num_blocks -1, EXTENT_DIRTY, 0)) {
-		search_start = ins->objectid + num_blocks;
+			   ins->objectid + num_bytes -1, EXTENT_DIRTY, 0)) {
+		search_start = ins->objectid + num_bytes;
 		goto new_group;
 	}
-	if (exclude_nr > 0 && (ins->objectid + num_blocks > exclude_start &&
+	if (exclude_nr > 0 && (ins->objectid + num_bytes > exclude_start &&
 	    ins->objectid < exclude_start + exclude_nr)) {
 		search_start = exclude_start + exclude_nr;
 		goto new_group;
@@ -1029,12 +1035,12 @@ check_pending:
 		if (block_group)
 			trans->block_group = block_group;
 	}
-	ins->offset = num_blocks;
+	ins->offset = num_bytes;
 	btrfs_free_path(path);
 	return 0;
 
 new_group:
-	if (search_start + num_blocks >= search_end) {
+	if (search_start + num_bytes >= search_end) {
 enospc:
 		search_start = orig_search_start;
 		if (full_scan) {
@@ -1069,12 +1075,12 @@ error:
  */
 int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root, u64 owner,
-		       u64 num_blocks, u64 empty_size, u64 hint_block,
+		       u64 num_bytes, u64 empty_size, u64 hint_byte,
 		       u64 search_end, struct btrfs_key *ins, int data)
 {
 	int ret;
 	int pending_ret;
-	u64 super_blocks_used, root_blocks_used;
+	u64 super_used, root_used;
 	u64 search_start = 0;
 	struct btrfs_fs_info *info = root->fs_info;
 	struct btrfs_root *extent_root = info->extent_root;
@@ -1083,9 +1089,9 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 	btrfs_set_stack_extent_refs(&extent_item, 1);
 	btrfs_set_stack_extent_owner(&extent_item, owner);
 
-	WARN_ON(num_blocks < 1);
-	ret = find_free_extent(trans, root, num_blocks, empty_size,
-			       search_start, search_end, hint_block, ins,
+	WARN_ON(num_bytes < root->sectorsize);
+	ret = find_free_extent(trans, root, num_bytes, empty_size,
+			       search_start, search_end, hint_byte, ins,
 			       trans->alloc_exclude_start,
 			       trans->alloc_exclude_nr, data);
 	BUG_ON(ret);
@@ -1093,21 +1099,18 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 		return ret;
 
 	/* block accounting for super block */
-	super_blocks_used = btrfs_super_blocks_used(&info->super_copy);
-	btrfs_set_super_blocks_used(&info->super_copy, super_blocks_used +
-				    num_blocks);
+	super_used = btrfs_super_bytes_used(&info->super_copy);
+	btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes);
 
 	/* block accounting for root item */
-	root_blocks_used = btrfs_root_used(&root->root_item);
-	btrfs_set_root_used(&root->root_item, root_blocks_used +
-				   num_blocks);
+	root_used = btrfs_root_used(&root->root_item);
+	btrfs_set_root_used(&root->root_item, root_used + num_bytes);
 
 	clear_extent_dirty(&root->fs_info->free_space_cache,
 			   ins->objectid, ins->objectid + ins->offset - 1,
 			   GFP_NOFS);
 
 	if (root == extent_root) {
-		BUG_ON(num_blocks != 1);
 		set_extent_bits(&root->fs_info->extent_ins, ins->objectid,
 				ins->objectid + ins->offset - 1,
 				EXTENT_LOCKED, GFP_NOFS);
@@ -1146,7 +1149,8 @@ update_block:
  * returns the tree buffer or NULL.
  */
 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
-					     struct btrfs_root *root, u64 hint,
+					     struct btrfs_root *root,
+					     u32 blocksize, u64 hint,
 					     u64 empty_size)
 {
 	struct btrfs_key ins;
@@ -1154,14 +1158,15 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 	struct extent_buffer *buf;
 
 	ret = btrfs_alloc_extent(trans, root, root->root_key.objectid,
-				 1, empty_size, hint, (u64)-1, &ins, 0);
+				 blocksize, empty_size, hint,
+				 (u64)-1, &ins, 0);
 	if (ret) {
 		BUG_ON(ret > 0);
 		return ERR_PTR(ret);
 	}
-	buf = btrfs_find_create_tree_block(root, ins.objectid);
+	buf = btrfs_find_create_tree_block(root, ins.objectid, blocksize);
 	if (!buf) {
-		btrfs_free_extent(trans, root, ins.objectid, 1, 0);
+		btrfs_free_extent(trans, root, ins.objectid, blocksize, 0);
 		return ERR_PTR(-ENOMEM);
 	}
 	btrfs_set_buffer_uptodate(buf);
@@ -1191,7 +1196,7 @@ static int drop_leaf_ref(struct btrfs_trans_handle *trans,
 	BUG_ON(!btrfs_is_leaf(leaf));
 	nritems = btrfs_header_nritems(leaf);
 	for (i = 0; i < nritems; i++) {
-		u64 disk_blocknr;
+		u64 disk_bytenr;
 
 		btrfs_item_key_to_cpu(leaf, &key, i);
 		if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
@@ -1204,11 +1209,11 @@ static int drop_leaf_ref(struct btrfs_trans_handle *trans,
 		 * FIXME make sure to insert a trans record that
 		 * repeats the snapshot del on crash
 		 */
-		disk_blocknr = btrfs_file_extent_disk_blocknr(leaf, fi);
-		if (disk_blocknr == 0)
+		disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
+		if (disk_bytenr == 0)
 			continue;
-		ret = btrfs_free_extent(trans, root, disk_blocknr,
-				btrfs_file_extent_disk_num_blocks(leaf, fi), 0);
+		ret = btrfs_free_extent(trans, root, disk_bytenr,
+				btrfs_file_extent_disk_num_bytes(leaf, fi), 0);
 		BUG_ON(ret);
 	}
 	return 0;
@@ -1219,19 +1224,23 @@ static void reada_walk_down(struct btrfs_root *root,
 {
 	int i;
 	u32 nritems;
-	u64 blocknr;
+	u64 bytenr;
 	int ret;
 	u32 refs;
+	int level;
+	u32 blocksize;
 
 	nritems = btrfs_header_nritems(node);
+	level = btrfs_header_level(node);
 	for (i = 0; i < nritems; i++) {
-		blocknr = btrfs_node_blockptr(node, i);
-		ret = lookup_extent_ref(NULL, root, blocknr, 1, &refs);
+		bytenr = btrfs_node_blockptr(node, i);
+		blocksize = btrfs_level_size(root, level - 1);
+		ret = lookup_extent_ref(NULL, root, bytenr, blocksize, &refs);
 		BUG_ON(ret);
 		if (refs != 1)
 			continue;
 		mutex_unlock(&root->fs_info->fs_mutex);
-		ret = readahead_tree_block(root, blocknr);
+		ret = readahead_tree_block(root, bytenr, blocksize);
 		cond_resched();
 		mutex_lock(&root->fs_info->fs_mutex);
 		if (ret)
@@ -1248,15 +1257,16 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 {
 	struct extent_buffer *next;
 	struct extent_buffer *cur;
-	u64 blocknr;
+	u64 bytenr;
+	u32 blocksize;
 	int ret;
 	u32 refs;
 
 	WARN_ON(*level < 0);
 	WARN_ON(*level >= BTRFS_MAX_LEVEL);
 	ret = lookup_extent_ref(trans, root,
-				extent_buffer_blocknr(path->nodes[*level]),
-				1, &refs);
+				path->nodes[*level]->start,
+				path->nodes[*level]->len, &refs);
 	BUG_ON(ret);
 	if (refs > 1)
 		goto out;
@@ -1283,30 +1293,33 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 			BUG_ON(ret);
 			break;
 		}
-		blocknr = btrfs_node_blockptr(cur, path->slots[*level]);
-		ret = lookup_extent_ref(trans, root, blocknr, 1, &refs);
+		bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
+		blocksize = btrfs_level_size(root, *level - 1);
+		ret = lookup_extent_ref(trans, root, bytenr, blocksize, &refs);
 		BUG_ON(ret);
 		if (refs != 1) {
 			path->slots[*level]++;
-			ret = btrfs_free_extent(trans, root, blocknr, 1, 1);
+			ret = btrfs_free_extent(trans, root, bytenr,
+						blocksize, 1);
 			BUG_ON(ret);
 			continue;
 		}
-		next = btrfs_find_tree_block(root, blocknr);
+		next = btrfs_find_tree_block(root, bytenr, blocksize);
 		if (!next || !btrfs_buffer_uptodate(next)) {
 			free_extent_buffer(next);
 			mutex_unlock(&root->fs_info->fs_mutex);
-			next = read_tree_block(root, blocknr);
+			next = read_tree_block(root, bytenr, blocksize);
 			mutex_lock(&root->fs_info->fs_mutex);
 
 			/* we dropped the lock, check one more time */
-			ret = lookup_extent_ref(trans, root, blocknr, 1, &refs);
+			ret = lookup_extent_ref(trans, root, bytenr,
+						blocksize, &refs);
 			BUG_ON(ret);
 			if (refs != 1) {
 				path->slots[*level]++;
 				free_extent_buffer(next);
 				ret = btrfs_free_extent(trans, root,
-							blocknr, 1, 1);
+							bytenr, blocksize, 1);
 				BUG_ON(ret);
 				continue;
 			}
@@ -1321,8 +1334,8 @@ static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 out:
 	WARN_ON(*level < 0);
 	WARN_ON(*level >= BTRFS_MAX_LEVEL);
-	ret = btrfs_free_extent(trans, root,
-			extent_buffer_blocknr(path->nodes[*level]), 1, 1);
+	ret = btrfs_free_extent(trans, root, path->nodes[*level]->start,
+				path->nodes[*level]->len, 1);
 	free_extent_buffer(path->nodes[*level]);
 	path->nodes[*level] = NULL;
 	*level += 1;
@@ -1359,8 +1372,8 @@ static int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 			return 0;
 		} else {
 			ret = btrfs_free_extent(trans, root,
-				    extent_buffer_blocknr(path->nodes[*level]),
-				    1, 1);
+						path->nodes[*level]->start,
+						path->nodes[*level]->len, 1);
 			BUG_ON(ret);
 			free_extent_buffer(path->nodes[*level]);
 			path->nodes[*level] = NULL;
@@ -1476,16 +1489,12 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 	struct btrfs_key key;
 	struct btrfs_key found_key;
 	struct extent_buffer *leaf;
-	u64 group_size_blocks;
 
 	block_group_cache = &info->block_group_cache;
 
-	group_size_blocks = BTRFS_BLOCK_GROUP_SIZE >>
-		info->sb->s_blocksize_bits;
-
 	root = info->extent_root;
 	key.objectid = 0;
-	key.offset = group_size_blocks;
+	key.offset = BTRFS_BLOCK_GROUP_SIZE;
 	btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
 
 	path = btrfs_alloc_path();
@@ -1532,7 +1541,7 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 				  (u64)cache);
 
 		if (key.objectid >=
-		    btrfs_super_total_blocks(&info->super_copy))
+		    btrfs_super_total_bytes(&info->super_copy))
 			break;
 	}
 
diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c
index e081558d52ff..f658703c42e6 100644
--- a/fs/btrfs/extent_map.c
+++ b/fs/btrfs/extent_map.c
@@ -1963,18 +1963,27 @@ static inline struct page *extent_buffer_page(struct extent_buffer *eb, int i)
 	struct page *p;
 	if (i == 0)
 		return eb->first_page;
+
 	i += eb->start >> PAGE_CACHE_SHIFT;
+	if (eb->last_page && eb->last_page->index == i)
+		return eb->last_page;
+
 	p = find_get_page(eb->first_page->mapping, i);
 	page_cache_release(p);
+	eb->last_page = p;
 	return p;
 }
 
+static inline unsigned long num_extent_pages(u64 start, u64 len)
+{
+	return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) -
+		(start >> PAGE_CACHE_SHIFT);
+}
 struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree,
 					  u64 start, unsigned long len,
 					  gfp_t mask)
 {
-	unsigned long num_pages = ((start + len - 1) >> PAGE_CACHE_SHIFT) -
-				  (start >> PAGE_CACHE_SHIFT) + 1;
+	unsigned long num_pages = num_extent_pages(start, len);
 	unsigned long i;
 	unsigned long index = start >> PAGE_CACHE_SHIFT;
 	struct extent_buffer *eb;
@@ -1986,7 +1995,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree,
 	if (!eb || IS_ERR(eb))
 		return NULL;
 
-	eb->alloc_addr = __builtin_return_address(0);
+	eb->alloc_addr = (unsigned long)__builtin_return_address(0);
 	eb->start = start;
 	eb->len = len;
 	atomic_set(&eb->refs, 1);
@@ -1994,6 +2003,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree,
 	for (i = 0; i < num_pages; i++, index++) {
 		p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM);
 		if (!p) {
+			WARN_ON(1);
 			/* make sure the free only frees the pages we've
 			 * grabbed a reference on
 			 */
@@ -2021,8 +2031,7 @@ struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree,
 					 u64 start, unsigned long len,
 					  gfp_t mask)
 {
-	unsigned long num_pages = ((start + len - 1) >> PAGE_CACHE_SHIFT) -
-				  (start >> PAGE_CACHE_SHIFT) + 1;
+	unsigned long num_pages = num_extent_pages(start, len);
 	unsigned long i;
 	unsigned long index = start >> PAGE_CACHE_SHIFT;
 	struct extent_buffer *eb;
@@ -2033,7 +2042,7 @@ struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree,
 	if (!eb || IS_ERR(eb))
 		return NULL;
 
-	eb->alloc_addr = __builtin_return_address(0);
+	eb->alloc_addr = (unsigned long)__builtin_return_address(0);
 	eb->start = start;
 	eb->len = len;
 	atomic_set(&eb->refs, 1);
@@ -2070,8 +2079,7 @@ void free_extent_buffer(struct extent_buffer *eb)
 	if (!atomic_dec_and_test(&eb->refs))
 		return;
 
-	num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
-		(eb->start >> PAGE_CACHE_SHIFT) + 1;
+	num_pages = num_extent_pages(eb->start, eb->len);
 
 	if (eb->first_page)
 		page_cache_release(eb->first_page);
@@ -2094,8 +2102,7 @@ int clear_extent_buffer_dirty(struct extent_map_tree *tree,
 	u64 end = start + eb->len - 1;
 
 	set = clear_extent_dirty(tree, start, end, GFP_NOFS);
-	num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
-		(eb->start >> PAGE_CACHE_SHIFT) + 1;
+	num_pages = num_extent_pages(eb->start, eb->len);
 
 	for (i = 0; i < num_pages; i++) {
 		page = extent_buffer_page(eb, i);
@@ -2145,8 +2152,7 @@ int set_extent_buffer_uptodate(struct extent_map_tree *tree,
 	struct page *page;
 	unsigned long num_pages;
 
-	num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
-		(eb->start >> PAGE_CACHE_SHIFT) + 1;
+	num_pages = num_extent_pages(eb->start, eb->len);
 
 	set_extent_uptodate(tree, eb->start, eb->start + eb->len - 1,
 			    GFP_NOFS);
@@ -2191,8 +2197,7 @@ int read_extent_buffer_pages(struct extent_map_tree *tree,
 		return 0;
 	}
 
-	num_pages = ((eb->start + eb->len - 1) >> PAGE_CACHE_SHIFT) -
-		(eb->start >> PAGE_CACHE_SHIFT) + 1;
+	num_pages = num_extent_pages(eb->start, eb->len);
 	for (i = 0; i < num_pages; i++) {
 		page = extent_buffer_page(eb, i);
 		if (PageUptodate(page)) {
@@ -2267,14 +2272,14 @@ void read_extent_buffer(struct extent_buffer *eb, void *dstv,
 }
 EXPORT_SYMBOL(read_extent_buffer);
 
-int map_extent_buffer(struct extent_buffer *eb, unsigned long start,
-		      unsigned long min_len,
-		      char **token, char **map,
-		      unsigned long *map_start,
-		      unsigned long *map_len, int km)
+static int __map_extent_buffer(struct extent_buffer *eb, unsigned long start,
+			       unsigned long min_len, char **token, char **map,
+			       unsigned long *map_start,
+			       unsigned long *map_len, int km)
 {
 	size_t offset = start & (PAGE_CACHE_SIZE - 1);
 	char *kaddr;
+	struct page *p;
 	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
 	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
 	unsigned long end_i = (start_offset + start + min_len) >>
@@ -2283,21 +2288,59 @@ int map_extent_buffer(struct extent_buffer *eb, unsigned long start,
 	if (i != end_i)
 		return -EINVAL;
 
-	WARN_ON(start > eb->len);
+	if (start >= eb->len) {
+		printk("bad start in map eb start %Lu len %lu caller start %lu min %lu\n", eb->start, eb->len, start, min_len);
+		WARN_ON(1);
+	}
 
 	if (i == 0) {
 		offset = start_offset;
 		*map_start = 0;
 	} else {
+		offset = 0;
 		*map_start = (i << PAGE_CACHE_SHIFT) - start_offset;
 	}
 
-	kaddr = kmap_atomic(extent_buffer_page(eb, i), km);
+	p = extent_buffer_page(eb, i);
+	WARN_ON(!PageUptodate(p));
+	kaddr = kmap_atomic(p, km);
 	*token = kaddr;
 	*map = kaddr + offset;
 	*map_len = PAGE_CACHE_SIZE - offset;
 	return 0;
 }
+
+int map_extent_buffer(struct extent_buffer *eb, unsigned long start,
+		      unsigned long min_len,
+		      char **token, char **map,
+		      unsigned long *map_start,
+		      unsigned long *map_len, int km)
+{
+	int err;
+	int save = 0;
+	if (eb->map_token) {
+		if (start >= eb->map_start &&
+		    start + min_len <= eb->map_start + eb->map_len) {
+			*token = eb->map_token;
+			*map = eb->kaddr;
+			*map_start = eb->map_start;
+			*map_len = eb->map_len;
+			return 0;
+		}
+		unmap_extent_buffer(eb, eb->map_token, km);
+		eb->map_token = NULL;
+		save = 1;
+	}
+	err = __map_extent_buffer(eb, start, min_len, token, map,
+				   map_start, map_len, km);
+	if (!err && save) {
+		eb->map_token = *token;
+		eb->kaddr = *map;
+		eb->map_start = *map_start;
+		eb->map_len = *map_len;
+	}
+	return err;
+}
 EXPORT_SYMBOL(map_extent_buffer);
 
 void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km)
@@ -2574,7 +2617,6 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
 			((unsigned long)PAGE_CACHE_SIZE - 1);
 		src_off_in_page = src_end &
 			((unsigned long)PAGE_CACHE_SIZE - 1);
-
 		if (src_i == 0)
 			src_off_in_page += start_offset;
 		if (dst_i == 0)
@@ -2582,14 +2624,13 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
 
 		cur = min(len, src_off_in_page + 1);
 		cur = min(cur, dst_off_in_page + 1);
-
 		move_pages(extent_buffer_page(dst, dst_i),
 			   extent_buffer_page(dst, src_i),
 			   dst_off_in_page - cur + 1,
 			   src_off_in_page - cur + 1, cur);
 
-		dst_end -= cur - 1;
-		src_end -= cur - 1;
+		dst_end -= cur;
+		src_end -= cur;
 		len -= cur;
 	}
 }
diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h
index 75dc600dc0f9..4ef8bdd68348 100644
--- a/fs/btrfs/extent_map.h
+++ b/fs/btrfs/extent_map.h
@@ -70,7 +70,12 @@ struct extent_buffer {
 	struct list_head list;
 	struct list_head leak_list;
 	struct page *first_page;
+	struct page *last_page;
 	unsigned long alloc_addr;
+	char *map_token;
+	char *kaddr;
+	unsigned long map_start;
+	unsigned long map_len;
 };
 
 typedef struct extent_map *(get_extent_t)(struct inode *inode,
@@ -147,11 +152,6 @@ static inline void extent_buffer_get(struct extent_buffer *eb)
 	atomic_inc(&eb->refs);
 }
 
-static inline u64 extent_buffer_blocknr(struct extent_buffer *eb)
-{
-	return eb->start / 4096;
-}
-
 int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv,
 			  unsigned long start,
 			  unsigned long len);
diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c
index 226f6d028c3f..9a65e97a4e28 100644
--- a/fs/btrfs/file-item.c
+++ b/fs/btrfs/file-item.c
@@ -27,8 +27,8 @@
 int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
 			       struct btrfs_root *root,
 			       u64 objectid, u64 pos,
-			       u64 offset, u64 disk_num_blocks,
-			       u64 num_blocks)
+			       u64 offset, u64 disk_num_bytes,
+			       u64 num_bytes)
 {
 	int ret = 0;
 	struct btrfs_file_extent_item *item;
@@ -50,10 +50,10 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
 	leaf = path->nodes[0];
 	item = btrfs_item_ptr(leaf, path->slots[0],
 			      struct btrfs_file_extent_item);
-	btrfs_set_file_extent_disk_blocknr(leaf, item, offset);
-	btrfs_set_file_extent_disk_num_blocks(leaf, item, disk_num_blocks);
+	btrfs_set_file_extent_disk_bytenr(leaf, item, offset);
+	btrfs_set_file_extent_disk_num_bytes(leaf, item, disk_num_bytes);
 	btrfs_set_file_extent_offset(leaf, item, 0);
-	btrfs_set_file_extent_num_blocks(leaf, item, num_blocks);
+	btrfs_set_file_extent_num_bytes(leaf, item, num_bytes);
 	btrfs_set_file_extent_generation(leaf, item, trans->transid);
 	btrfs_set_file_extent_type(leaf, item, BTRFS_FILE_EXTENT_REG);
 	btrfs_mark_buffer_dirty(leaf);
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 1734ca695555..844d8807e44a 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -120,9 +120,9 @@ static int insert_inline_extent(struct btrfs_trans_handle *trans,
 	btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE);
 	ptr = btrfs_file_extent_inline_start(ei);
 
-	kaddr = kmap_atomic(page, KM_USER0);
+	kaddr = kmap_atomic(page, KM_USER1);
 	write_extent_buffer(leaf, kaddr + page_offset, ptr, size);
-	kunmap_atomic(kaddr, KM_USER0);
+	kunmap_atomic(kaddr, KM_USER1);
 	btrfs_mark_buffer_dirty(leaf);
 fail:
 	btrfs_free_path(path);
@@ -142,11 +142,12 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 	struct inode *inode = file->f_path.dentry->d_inode;
 	struct extent_map *em;
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
-	u64 hint_block;
-	u64 num_blocks;
+	u64 hint_byte;
+	u64 num_bytes;
 	u64 start_pos;
 	u64 end_of_last_block;
 	u64 end_pos = pos + write_bytes;
+	u32 inline_size;
 	loff_t isize = i_size_read(inode);
 
 	em = alloc_extent_map(GFP_NOFS);
@@ -156,11 +157,12 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 	em->bdev = inode->i_sb->s_bdev;
 
 	start_pos = pos & ~((u64)root->sectorsize - 1);
-	num_blocks = (write_bytes + pos - start_pos + root->sectorsize - 1) >>
-			inode->i_blkbits;
+	num_bytes = (write_bytes + pos - start_pos +
+		    root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
 
 	down_read(&BTRFS_I(inode)->root->snap_sem);
-	end_of_last_block = start_pos + (num_blocks << inode->i_blkbits) - 1;
+	end_of_last_block = start_pos + num_bytes - 1;
+
 	lock_extent(em_tree, start_pos, end_of_last_block, GFP_NOFS);
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
@@ -169,8 +171,8 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 		goto out_unlock;
 	}
 	btrfs_set_trans_block_group(trans, inode);
-	inode->i_blocks += num_blocks << 3;
-	hint_block = 0;
+	inode->i_blocks += num_bytes >> 9;
+	hint_byte = 0;
 
 	if ((end_of_last_block & 4095) == 0) {
 		printk("strange end of last %Lu %zu %Lu\n", start_pos, write_bytes, end_of_last_block);
@@ -191,11 +193,10 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 			err = btrfs_drop_extents(trans, root, inode,
 						 last_pos_in_file,
 						 last_pos_in_file + hole_size,
-						 &hint_block);
+						 &hint_byte);
 			if (err)
 				goto failed;
 
-			hole_size >>= inode->i_blkbits;
 			err = btrfs_insert_file_extent(trans, root,
 						       inode->i_ino,
 						       last_pos_in_file,
@@ -209,8 +210,10 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 	 * either allocate an extent for the new bytes or setup the key
 	 * to show we are doing inline data in the extent
 	 */
+	inline_size = end_pos - start_pos;
 	if (isize >= PAGE_CACHE_SIZE || pos + write_bytes < inode->i_size ||
-	    pos + write_bytes - start_pos > BTRFS_MAX_INLINE_DATA_SIZE(root)) {
+	    inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root) ||
+	    inline_size >= PAGE_CACHE_SIZE) {
 		u64 last_end;
 		for (i = 0; i < num_pages; i++) {
 			struct page *p = pages[i];
@@ -224,10 +227,9 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 	} else {
 		struct page *p = pages[0];
 		/* step one, delete the existing extents in this range */
-		/* FIXME blocksize != pagesize */
 		err = btrfs_drop_extents(trans, root, inode, start_pos,
 			 (pos + write_bytes + root->sectorsize -1) &
-			 ~((u64)root->sectorsize - 1), &hint_block);
+			 ~((u64)root->sectorsize - 1), &hint_byte);
 		if (err)
 			goto failed;
 
@@ -283,7 +285,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end)
  */
 int btrfs_drop_extents(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root, struct inode *inode,
-		       u64 start, u64 end, u64 *hint_block)
+		       u64 start, u64 end, u64 *hint_byte)
 {
 	int ret;
 	struct btrfs_key key;
@@ -346,8 +348,7 @@ next_slot:
 			found_type = btrfs_file_extent_type(leaf, extent);
 			if (found_type == BTRFS_FILE_EXTENT_REG) {
 				extent_end = key.offset +
-				 (btrfs_file_extent_num_blocks(leaf, extent) <<
-					 inode->i_blkbits);
+				     btrfs_file_extent_num_bytes(leaf, extent);
 				found_extent = 1;
 			} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
 				struct btrfs_item *item;
@@ -386,17 +387,17 @@ next_slot:
 
 		if (end < extent_end && end >= key.offset) {
 			if (found_extent) {
-				u64 disk_blocknr =
-				    btrfs_file_extent_disk_blocknr(leaf,extent);
-				u64 disk_num_blocks =
-				    btrfs_file_extent_disk_num_blocks(leaf,
+				u64 disk_bytenr =
+				    btrfs_file_extent_disk_bytenr(leaf, extent);
+				u64 disk_num_bytes =
+				    btrfs_file_extent_disk_num_bytes(leaf,
 								      extent);
 				read_extent_buffer(leaf, &old,
 						   (unsigned long)extent,
 						   sizeof(old));
-				if (disk_blocknr != 0) {
+				if (disk_bytenr != 0) {
 					ret = btrfs_inc_extent_ref(trans, root,
-					         disk_blocknr, disk_num_blocks);
+					         disk_bytenr, disk_num_bytes);
 					BUG_ON(ret);
 				}
 			}
@@ -410,21 +411,19 @@ next_slot:
 			keep = 1;
 			WARN_ON(start & (root->sectorsize - 1));
 			if (found_extent) {
-				new_num = (start - key.offset) >>
-					inode->i_blkbits;
-				old_num = btrfs_file_extent_num_blocks(leaf,
-								       extent);
-				*hint_block =
-					btrfs_file_extent_disk_blocknr(leaf,
-								       extent);
-				if (btrfs_file_extent_disk_blocknr(leaf,
-								   extent)) {
+				new_num = start - key.offset;
+				old_num = btrfs_file_extent_num_bytes(leaf,
+								      extent);
+				*hint_byte =
+					btrfs_file_extent_disk_bytenr(leaf,
+								      extent);
+				if (btrfs_file_extent_disk_bytenr(leaf,
+								  extent)) {
 					inode->i_blocks -=
-						(old_num - new_num) << 3;
+						(old_num - new_num) >> 9;
 				}
-				btrfs_set_file_extent_num_blocks(leaf,
-								 extent,
-								 new_num);
+				btrfs_set_file_extent_num_bytes(leaf, extent,
+								new_num);
 				btrfs_mark_buffer_dirty(leaf);
 			} else {
 				WARN_ON(1);
@@ -432,33 +431,32 @@ next_slot:
 		}
 		/* delete the entire extent */
 		if (!keep) {
-			u64 disk_blocknr = 0;
-			u64 disk_num_blocks = 0;
-			u64 extent_num_blocks = 0;
+			u64 disk_bytenr = 0;
+			u64 disk_num_bytes = 0;
+			u64 extent_num_bytes = 0;
 			if (found_extent) {
-				disk_blocknr =
-				      btrfs_file_extent_disk_blocknr(leaf,
+				disk_bytenr =
+				      btrfs_file_extent_disk_bytenr(leaf,
 								     extent);
-				disk_num_blocks =
-				      btrfs_file_extent_disk_num_blocks(leaf,
-									extent);
-				extent_num_blocks =
-				      btrfs_file_extent_num_blocks(leaf,
-								   extent);
-				*hint_block =
-					btrfs_file_extent_disk_blocknr(leaf,
+				disk_num_bytes =
+				      btrfs_file_extent_disk_num_bytes(leaf,
 								       extent);
+				extent_num_bytes =
+				      btrfs_file_extent_num_bytes(leaf, extent);
+				*hint_byte =
+					btrfs_file_extent_disk_bytenr(leaf,
+								      extent);
 			}
 			ret = btrfs_del_item(trans, root, path);
 			/* TODO update progress marker and return */
 			BUG_ON(ret);
 			btrfs_release_path(root, path);
 			extent = NULL;
-			if (found_extent && disk_blocknr != 0) {
-				inode->i_blocks -= extent_num_blocks << 3;
+			if (found_extent && disk_bytenr != 0) {
+				inode->i_blocks -= extent_num_bytes >> 9;
 				ret = btrfs_free_extent(trans, root,
-							disk_blocknr,
-							disk_num_blocks, 0);
+							disk_bytenr,
+							disk_num_bytes, 0);
 			}
 
 			BUG_ON(ret);
@@ -491,20 +489,19 @@ next_slot:
 					    (unsigned long)extent, sizeof(old));
 
 			btrfs_set_file_extent_offset(leaf, extent,
-				    le64_to_cpu(old.offset) +
-				    ((end - key.offset) >> inode->i_blkbits));
-			WARN_ON(le64_to_cpu(old.num_blocks) <
-				(extent_end - end) >> inode->i_blkbits);
-			btrfs_set_file_extent_num_blocks(leaf, extent,
-				    (extent_end - end) >> inode->i_blkbits);
-
+				    le64_to_cpu(old.offset) + end - key.offset);
+			WARN_ON(le64_to_cpu(old.num_bytes) <
+				(extent_end - end));
+			btrfs_set_file_extent_num_bytes(leaf, extent,
+							extent_end - end);
 			btrfs_set_file_extent_type(leaf, extent,
 						   BTRFS_FILE_EXTENT_REG);
+
 			btrfs_mark_buffer_dirty(path->nodes[0]);
-			if (le64_to_cpu(old.disk_blocknr) != 0) {
+			if (le64_to_cpu(old.disk_bytenr) != 0) {
 				inode->i_blocks +=
-				      btrfs_file_extent_num_blocks(leaf,
-								   extent) << 3;
+				      btrfs_file_extent_num_bytes(leaf,
+								  extent) >> 9;
 			}
 			ret = 0;
 			goto out;
@@ -531,12 +528,9 @@ static int prepare_pages(struct btrfs_root *root,
 	unsigned long index = pos >> PAGE_CACHE_SHIFT;
 	struct inode *inode = file->f_path.dentry->d_inode;
 	int err = 0;
-	u64 num_blocks;
 	u64 start_pos;
 
 	start_pos = pos & ~((u64)root->sectorsize - 1);
-	num_blocks = (write_bytes + pos - start_pos + root->sectorsize - 1) >>
-			inode->i_blkbits;
 
 	memset(pages, 0, num_pages * sizeof(struct page *));
 
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index fbe2836364e0..d6b3a55ed8e0 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -77,19 +77,19 @@ static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
 	struct btrfs_trans_handle *trans;
 	struct btrfs_key ins;
 	u64 alloc_hint = 0;
-	u64 num_blocks;
+	u64 num_bytes;
 	int ret;
-	u64 blocksize = 1 << inode->i_blkbits;
+	u64 blocksize = root->sectorsize;
 
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
 	btrfs_set_trans_block_group(trans, inode);
 	BUG_ON(!trans);
-	num_blocks = (end - start + blocksize) & ~(blocksize - 1);
+	num_bytes = (end - start + blocksize) & ~(blocksize - 1);
 	ret = btrfs_drop_extents(trans, root, inode,
-				 start, start + num_blocks, &alloc_hint);
-	num_blocks = num_blocks >> inode->i_blkbits;
-	ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_blocks, 0,
+				 start, start + num_bytes, &alloc_hint);
+
+	ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_bytes, 0,
 				 alloc_hint, (u64)-1, &ins, 1);
 	if (ret) {
 		WARN_ON(1);
@@ -186,7 +186,8 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end)
 zeroit:
 	printk("btrfs csum failed ino %lu off %llu\n",
 	       page->mapping->host->i_ino, (unsigned long long)start);
-	memset(kaddr + offset, 1, end - start + 1); flush_dcache_page(page);
+	memset(kaddr + offset, 1, end - start + 1);
+	flush_dcache_page(page);
 	kunmap_atomic(kaddr, KM_IRQ0);
 	return 0;
 }
@@ -547,7 +548,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 	struct extent_buffer *leaf;
 	struct btrfs_file_extent_item *fi;
 	u64 extent_start = 0;
-	u64 extent_num_blocks = 0;
+	u64 extent_num_bytes = 0;
 	u64 item_end = 0;
 	int found_extent;
 	int del_item;
@@ -593,8 +594,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 			if (btrfs_file_extent_type(leaf, fi) !=
 			    BTRFS_FILE_EXTENT_INLINE) {
 				item_end +=
-				    btrfs_file_extent_num_blocks(leaf, fi) <<
-				    inode->i_blkbits;
+				    btrfs_file_extent_num_bytes(leaf, fi);
 			}
 		}
 		if (found_type == BTRFS_CSUM_ITEM_KEY) {
@@ -626,28 +626,27 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 			   btrfs_file_extent_type(leaf, fi) !=
 			   BTRFS_FILE_EXTENT_INLINE) {
 			u64 num_dec;
-			extent_start = btrfs_file_extent_disk_blocknr(leaf, fi);
+			extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
 			if (!del_item) {
-				u64 orig_num_blocks =
-					btrfs_file_extent_num_blocks(leaf, fi);
-				extent_num_blocks = inode->i_size -
+				u64 orig_num_bytes =
+					btrfs_file_extent_num_bytes(leaf, fi);
+				extent_num_bytes = inode->i_size -
 					found_key.offset + root->sectorsize - 1;
-				extent_num_blocks >>= inode->i_blkbits;
-				btrfs_set_file_extent_num_blocks(leaf, fi,
-							 extent_num_blocks);
-				num_dec = (orig_num_blocks -
-					   extent_num_blocks) << 3;
+				btrfs_set_file_extent_num_bytes(leaf, fi,
+							 extent_num_bytes);
+				num_dec = (orig_num_bytes -
+					   extent_num_bytes) >> 9;
 				if (extent_start != 0) {
 					inode->i_blocks -= num_dec;
 				}
 				btrfs_mark_buffer_dirty(leaf);
 			} else {
-				extent_num_blocks =
-					btrfs_file_extent_disk_num_blocks(leaf,
-									  fi);
+				extent_num_bytes =
+					btrfs_file_extent_disk_num_bytes(leaf,
+									 fi);
 				/* FIXME blocksize != 4096 */
-				num_dec = btrfs_file_extent_num_blocks(leaf,
-								       fi) << 3;
+				num_dec = btrfs_file_extent_num_bytes(leaf,
+								       fi) >> 9;
 				if (extent_start != 0) {
 					found_extent = 1;
 					inode->i_blocks -= num_dec;
@@ -664,7 +663,7 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 		btrfs_release_path(root, path);
 		if (found_extent) {
 			ret = btrfs_free_extent(trans, root, extent_start,
-						extent_num_blocks, 0);
+						extent_num_bytes, 0);
 			BUG_ON(ret);
 		}
 	}
@@ -709,7 +708,8 @@ static int btrfs_cow_one_page(struct inode *inode, struct page *page,
 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
 {
 	struct inode *inode = mapping->host;
-	unsigned blocksize = 1 << inode->i_blkbits;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	u32 blocksize = root->sectorsize;
 	pgoff_t index = from >> PAGE_CACHE_SHIFT;
 	unsigned offset = from & (PAGE_CACHE_SIZE-1);
 	struct page *page;
@@ -719,7 +719,7 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
 	if ((offset & (blocksize - 1)) == 0)
 		goto out;
 
-	down_read(&BTRFS_I(inode)->root->snap_sem);
+	down_read(&root->snap_sem);
 	ret = -ENOMEM;
 	page = grab_cache_page(mapping, index);
 	if (!page)
@@ -778,8 +778,6 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
 		err = btrfs_drop_extents(trans, root, inode,
 					 pos, pos + hole_size, &alloc_hint);
 
-		hole_size >>= inode->i_blkbits;
-
 		err = btrfs_insert_file_extent(trans, root, inode->i_ino,
 					       pos, 0, 0, hole_size);
 		btrfs_end_transaction(trans, root);
@@ -1490,7 +1488,7 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
 {
 	int ret;
 	int err = 0;
-	u64 blocknr;
+	u64 bytenr;
 	u64 extent_start = 0;
 	u64 extent_end = 0;
 	u64 objectid = inode->i_ino;
@@ -1540,10 +1538,6 @@ again:
 	leaf = path->nodes[0];
 	item = btrfs_item_ptr(leaf, path->slots[0],
 			      struct btrfs_file_extent_item);
-
-	blocknr = btrfs_file_extent_disk_blocknr(leaf, item);
-	blocknr += btrfs_file_extent_offset(leaf, item);
-
 	/* are we inside the extent that was found? */
 	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
 	found_type = btrfs_key_type(&found_key);
@@ -1556,8 +1550,7 @@ again:
 	extent_start = found_key.offset;
 	if (found_type == BTRFS_FILE_EXTENT_REG) {
 		extent_end = extent_start +
-		       (btrfs_file_extent_num_blocks(leaf, item) <<
-			inode->i_blkbits);
+		       btrfs_file_extent_num_bytes(leaf, item);
 		err = 0;
 		if (start < extent_start || start >= extent_end) {
 			em->start = start;
@@ -1570,17 +1563,18 @@ again:
 			}
 			goto not_found_em;
 		}
-		if (btrfs_file_extent_disk_blocknr(leaf, item) == 0) {
+		bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
+		if (bytenr == 0) {
 			em->start = extent_start;
 			em->end = extent_end - 1;
 			em->block_start = EXTENT_MAP_HOLE;
 			em->block_end = EXTENT_MAP_HOLE;
 			goto insert;
 		}
-		em->block_start = blocknr << inode->i_blkbits;
+		bytenr += btrfs_file_extent_offset(leaf, item);
+		em->block_start = bytenr;
 		em->block_end = em->block_start +
-			(btrfs_file_extent_num_blocks(leaf, item) <<
-			 inode->i_blkbits) - 1;
+			btrfs_file_extent_num_bytes(leaf, item) - 1;
 		em->start = extent_start;
 		em->end = extent_end - 1;
 		goto insert;
@@ -1592,7 +1586,8 @@ again:
 		size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
 						    path->slots[0]));
 
-		extent_end = extent_start | ((u64)root->sectorsize - 1);
+		extent_end = (extent_start + size) |
+			((u64)root->sectorsize - 1);
 		if (start < extent_start || start >= extent_end) {
 			em->start = start;
 			if (start < extent_start) {
@@ -1617,8 +1612,10 @@ again:
 		ptr = btrfs_file_extent_inline_start(item);
 		map = kmap(page);
 		read_extent_buffer(leaf, map + page_offset, ptr, size);
+		/*
 		memset(map + page_offset + size, 0,
 		       root->sectorsize - (page_offset + size));
+		       */
 		flush_dcache_page(page);
 		kunmap(page);
 		set_extent_uptodate(em_tree, extent_start,
@@ -1836,13 +1833,13 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 	trans = btrfs_start_transaction(root, 1);
 	BUG_ON(!trans);
 
-	leaf = btrfs_alloc_free_block(trans, root, 0, 0);
+	leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, 0);
 	if (IS_ERR(leaf))
 		return PTR_ERR(leaf);
 
 	btrfs_set_header_nritems(leaf, 0);
 	btrfs_set_header_level(leaf, 0);
-	btrfs_set_header_blocknr(leaf, extent_buffer_blocknr(leaf));
+	btrfs_set_header_bytenr(leaf, leaf->start);
 	btrfs_set_header_generation(leaf, trans->transid);
 	btrfs_set_header_owner(leaf, root->root_key.objectid);
 	write_extent_buffer(leaf, root->fs_info->fsid,
@@ -1858,7 +1855,8 @@ static int create_subvol(struct btrfs_root *root, char *name, int namelen)
 	inode_item->nblocks = cpu_to_le64(1);
 	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
 
-	btrfs_set_root_blocknr(&root_item, extent_buffer_blocknr(leaf));
+	btrfs_set_root_bytenr(&root_item, leaf->start);
+	btrfs_set_root_level(&root_item, 0);
 	btrfs_set_root_refs(&root_item, 1);
 	btrfs_set_root_used(&root_item, 0);
 
@@ -1971,8 +1969,8 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
 
 	btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
-	btrfs_set_root_blocknr(&new_root_item,
-			       extent_buffer_blocknr(root->node));
+	btrfs_set_root_bytenr(&new_root_item, root->node->start);
+	btrfs_set_root_level(&new_root_item, btrfs_header_level(root->node));
 
 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
 				&new_root_item);
diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c
index a825ce078a54..9f8696c8a8e8 100644
--- a/fs/btrfs/print-tree.c
+++ b/fs/btrfs/print-tree.c
@@ -36,7 +36,7 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 	u32 type;
 
 	printk("leaf %llu total ptrs %d free space %d\n",
-		(unsigned long long)btrfs_header_blocknr(l), nr,
+		(unsigned long long)btrfs_header_bytenr(l), nr,
 		btrfs_leaf_free_space(root, l));
 	for (i = 0 ; i < nr ; i++) {
 		item = btrfs_item_nr(l, i);
@@ -65,8 +65,8 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 			break;
 		case BTRFS_ROOT_ITEM_KEY:
 			ri = btrfs_item_ptr(l, i, struct btrfs_root_item);
-			printk("\t\troot data blocknr %llu refs %u\n",
-				(unsigned long long)btrfs_disk_root_blocknr(l, ri),
+			printk("\t\troot data bytenr %llu refs %u\n",
+				(unsigned long long)btrfs_disk_root_bytenr(l, ri),
 				btrfs_disk_root_refs(l, ri));
 			break;
 		case BTRFS_EXTENT_ITEM_KEY:
@@ -84,12 +84,12 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 			           btrfs_file_extent_inline_len(l, item));
 				break;
 			}
-			printk("\t\textent data disk block %llu nr %llu\n",
-			       (unsigned long long)btrfs_file_extent_disk_blocknr(l, fi),
-			       (unsigned long long)btrfs_file_extent_disk_num_blocks(l, fi));
+			printk("\t\textent data disk bytenr %llu nr %llu\n",
+			       (unsigned long long)btrfs_file_extent_disk_bytenr(l, fi),
+			       (unsigned long long)btrfs_file_extent_disk_num_bytes(l, fi));
 			printk("\t\textent data offset %llu nr %llu\n",
 			  (unsigned long long)btrfs_file_extent_offset(l, fi),
-			  (unsigned long long)btrfs_file_extent_num_blocks(l, fi));
+			  (unsigned long long)btrfs_file_extent_num_bytes(l, fi));
 			break;
 		case BTRFS_BLOCK_GROUP_ITEM_KEY:
 			bi = btrfs_item_ptr(l, i,
@@ -106,16 +106,18 @@ void btrfs_print_tree(struct btrfs_root *root, struct extent_buffer *c)
 	int i;
 	u32 nr;
 	struct btrfs_key key;
+	int level;
 
 	if (!c)
 		return;
 	nr = btrfs_header_nritems(c);
-	if (btrfs_is_leaf(c)) {
+	level = btrfs_header_level(c);
+	if (level == 0) {
 		btrfs_print_leaf(root, c);
 		return;
 	}
 	printk("node %llu level %d total ptrs %d free spc %u\n",
-	       (unsigned long long)btrfs_header_blocknr(c),
+	       (unsigned long long)btrfs_header_bytenr(c),
 	       btrfs_header_level(c), nr,
 	       (u32)BTRFS_NODEPTRS_PER_BLOCK(root) - nr);
 	for (i = 0; i < nr; i++) {
@@ -129,7 +131,8 @@ void btrfs_print_tree(struct btrfs_root *root, struct extent_buffer *c)
 	}
 	for (i = 0; i < nr; i++) {
 		struct extent_buffer *next = read_tree_block(root,
-						btrfs_node_blockptr(c, i));
+					btrfs_node_blockptr(c, i),
+					btrfs_level_size(root, level - 1));
 		if (btrfs_is_leaf(next) &&
 		    btrfs_header_level(c) != 1)
 			BUG();
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 39a1435c68f1..5c4370f3a5b8 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -303,10 +303,12 @@ static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 {
 	struct btrfs_root *root = btrfs_sb(dentry->d_sb);
 	struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
+	int bits = dentry->d_sb->s_blocksize_bits;
 
 	buf->f_namelen = BTRFS_NAME_LEN;
-	buf->f_blocks = btrfs_super_total_blocks(disk_super);
-	buf->f_bfree = buf->f_blocks - btrfs_super_blocks_used(disk_super);
+	buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
+	buf->f_bfree = buf->f_blocks -
+		(btrfs_super_bytes_used(disk_super) >> bits);
 	buf->f_bavail = buf->f_bfree;
 	buf->f_bsize = dentry->d_sb->s_blocksize;
 	buf->f_type = BTRFS_SUPER_MAGIC;
diff --git a/fs/btrfs/sysfs.c b/fs/btrfs/sysfs.c
index 9654e90eec89..0bd1fd3d29df 100644
--- a/fs/btrfs/sysfs.c
+++ b/fs/btrfs/sysfs.c
@@ -42,14 +42,15 @@ static ssize_t root_block_limit_show(struct btrfs_root *root, char *buf)
 
 static ssize_t super_blocks_used_show(struct btrfs_fs_info *fs, char *buf)
 {
+
 	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_super_blocks_used(&fs->super_copy));
+		(unsigned long long)btrfs_super_bytes_used(&fs->super_copy));
 }
 
 static ssize_t super_total_blocks_show(struct btrfs_fs_info *fs, char *buf)
 {
 	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_super_total_blocks(&fs->super_copy));
+		(unsigned long long)btrfs_super_total_bytes(&fs->super_copy));
 }
 
 static ssize_t super_blocksize_show(struct btrfs_fs_info *fs, char *buf)
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 55289b71056e..60f61345a8d0 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -205,12 +205,13 @@ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
 
 	btrfs_write_dirty_block_groups(trans, extent_root);
 	while(1) {
-		old_extent_block = btrfs_root_blocknr(&extent_root->root_item);
-		if (old_extent_block ==
-		    extent_buffer_blocknr(extent_root->node))
+		old_extent_block = btrfs_root_bytenr(&extent_root->root_item);
+		if (old_extent_block == extent_root->node->start)
 			break;
-		btrfs_set_root_blocknr(&extent_root->root_item,
-			       extent_buffer_blocknr(extent_root->node));
+		btrfs_set_root_bytenr(&extent_root->root_item,
+				      extent_root->node->start);
+		btrfs_set_root_level(&extent_root->root_item,
+				     btrfs_header_level(extent_root->node));
 		ret = btrfs_update_root(trans, tree_root,
 					&extent_root->root_key,
 					&extent_root->root_item);
@@ -284,8 +285,8 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans,
 				     (unsigned long)root->root_key.objectid,
 				     BTRFS_ROOT_TRANS_TAG);
 			if (root->commit_root == root->node) {
-				WARN_ON(extent_buffer_blocknr(root->node) !=
-					btrfs_root_blocknr(&root->root_item));
+				WARN_ON(root->node->start !=
+					btrfs_root_bytenr(&root->root_item));
 				free_extent_buffer(root->commit_root);
 				root->commit_root = NULL;
 
@@ -314,8 +315,10 @@ static int add_dirty_roots(struct btrfs_trans_handle *trans,
 			root->commit_root = NULL;
 
 			root->root_key.offset = root->fs_info->generation;
-			btrfs_set_root_blocknr(&root->root_item,
-				       extent_buffer_blocknr(root->node));
+			btrfs_set_root_bytenr(&root->root_item,
+					      root->node->start);
+			btrfs_set_root_level(&root->root_item,
+					     btrfs_header_level(root->node));
 			err = btrfs_insert_root(trans, root->fs_info->tree_root,
 						&root->root_key,
 						&root->root_item);
@@ -407,8 +410,8 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 	struct dirty_root *dirty;
 	struct btrfs_trans_handle *trans;
 	unsigned long nr;
-	u64 num_blocks;
-	u64 blocks_used;
+	u64 num_bytes;
+	u64 bytes_used;
 	int ret = 0;
 	int err;
 
@@ -419,7 +422,7 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 		dirty = list_entry(list->next, struct dirty_root, list);
 		list_del_init(&dirty->list);
 
-		num_blocks = btrfs_root_used(&dirty->root->root_item);
+		num_bytes = btrfs_root_used(&dirty->root->root_item);
 		root = dirty->latest_root;
 
 		while(1) {
@@ -446,12 +449,12 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 		}
 		BUG_ON(ret);
 
-		num_blocks -= btrfs_root_used(&dirty->root->root_item);
-		blocks_used = btrfs_root_used(&root->root_item);
-		if (num_blocks) {
+		num_bytes -= btrfs_root_used(&dirty->root->root_item);
+		bytes_used = btrfs_root_used(&root->root_item);
+		if (num_bytes) {
 			record_root_in_trans(root);
 			btrfs_set_root_used(&root->root_item,
-						   blocks_used - num_blocks);
+					    bytes_used - num_bytes);
 		}
 		ret = btrfs_del_root(trans, tree_root, &dirty->root->root_key);
 		if (ret) {
@@ -560,7 +563,9 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	btrfs_set_super_generation(&root->fs_info->super_copy,
 				   cur_trans->transid);
 	btrfs_set_super_root(&root->fs_info->super_copy,
-		     extent_buffer_blocknr(root->fs_info->tree_root->node));
+			     root->fs_info->tree_root->node->start);
+	btrfs_set_super_root_level(&root->fs_info->super_copy,
+			   btrfs_header_level(root->fs_info->tree_root->node));
 
 	write_extent_buffer(root->fs_info->sb_buffer,
 			    &root->fs_info->super_copy, 0,
diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c
index daf019afa0a1..3feac2f28628 100644
--- a/fs/btrfs/tree-defrag.c
+++ b/fs/btrfs/tree-defrag.c
@@ -27,13 +27,15 @@ static void reada_defrag(struct btrfs_root *root,
 {
 	int i;
 	u32 nritems;
-	u64 blocknr;
+	u64 bytenr;
+	u32 blocksize;
 	int ret;
 
+	blocksize = btrfs_level_size(root, btrfs_header_level(node) - 1);
 	nritems = btrfs_header_nritems(node);
 	for (i = 0; i < nritems; i++) {
-		blocknr = btrfs_node_blockptr(node, i);
-		ret = readahead_tree_block(root, blocknr);
+		bytenr = btrfs_node_blockptr(node, i);
+		ret = readahead_tree_block(root, bytenr, blocksize);
 		if (ret)
 			break;
 	}
@@ -46,7 +48,7 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 {
 	struct extent_buffer *next;
 	struct extent_buffer *cur;
-	u64 blocknr;
+	u64 bytenr;
 	int ret = 0;
 	int is_extent = 0;
 
@@ -80,10 +82,11 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 
 			break;
 		}
-		blocknr = btrfs_node_blockptr(cur, path->slots[*level]);
+		bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
 
 		if (cache_only) {
-			next = btrfs_find_tree_block(root, blocknr);
+			next = btrfs_find_tree_block(root, bytenr,
+					   btrfs_level_size(root, *level - 1));
 			/* FIXME, test for defrag */
 			if (!next || !btrfs_buffer_uptodate(next)) {
 				free_extent_buffer(next);
@@ -91,7 +94,8 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 				continue;
 			}
 		} else {
-			next = read_tree_block(root, blocknr);
+			next = read_tree_block(root, bytenr,
+				       btrfs_level_size(root, *level - 1));
 		}
 		ret = btrfs_cow_block(trans, root, next, path->nodes[*level],
 				      path->slots[*level], &next);
-- 
cgit v1.2.3


From 6b80053d02be41886344b5007d04e345311ec0b5 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Mon, 15 Oct 2007 16:17:34 -0400
Subject: Btrfs: Add back the online defragging code

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.c       | 177 ++++++++++++++++++++++++-------------------------
 fs/btrfs/disk-io.c     |  60 +++++++++++++++++
 fs/btrfs/disk-io.h     |   7 ++
 fs/btrfs/extent-tree.c |   8 +--
 fs/btrfs/extent_map.h  |   2 +
 fs/btrfs/transaction.c |   4 +-
 fs/btrfs/tree-defrag.c |  14 ++--
 7 files changed, 166 insertions(+), 106 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 34de83630ae9..08ddf1873ab1 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -155,55 +155,49 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans,
 	return ret;
 }
 
-#if 0
-static int close_blocks(u64 blocknr, u64 other)
+static int close_blocks(u64 blocknr, u64 other, u32 blocksize)
 {
-	if (blocknr < other && other - blocknr < 8)
+	if (blocknr < other && other - (blocknr + blocksize) < 32768)
 		return 1;
-	if (blocknr > other && blocknr - other < 8)
+	if (blocknr > other && blocknr - (other + blocksize) < 32768)
 		return 1;
 	return 0;
 }
 
-static int should_defrag_leaf(struct extent_buffer *eb)
+static int should_defrag_leaf(struct extent_buffer *leaf)
 {
-	return 0;
-	struct btrfs_leaf *leaf = btrfs_buffer_leaf(eb);
-	struct btrfs_disk_key *key;
+	struct btrfs_key key;
 	u32 nritems;
 
-	if (buffer_defrag(bh))
+	if (btrfs_buffer_defrag(leaf))
 		return 1;
 
-	nritems = btrfs_header_nritems(&leaf->header);
+	nritems = btrfs_header_nritems(leaf);
 	if (nritems == 0)
 		return 0;
 
-	key = &leaf->items[0].key;
-	if (btrfs_disk_key_type(key) == BTRFS_DIR_ITEM_KEY)
+	btrfs_item_key_to_cpu(leaf, &key, 0);
+	if (key.type == BTRFS_DIR_ITEM_KEY)
 		return 1;
 
-	key = &leaf->items[nritems-1].key;
-	if (btrfs_disk_key_type(key) == BTRFS_DIR_ITEM_KEY)
+
+	btrfs_item_key_to_cpu(leaf, &key, nritems - 1);
+	if (key.type == BTRFS_DIR_ITEM_KEY)
 		return 1;
 	if (nritems > 4) {
-		key = &leaf->items[nritems/2].key;
-		if (btrfs_disk_key_type(key) == BTRFS_DIR_ITEM_KEY)
+		btrfs_item_key_to_cpu(leaf, &key, nritems / 2);
+		if (key.type == BTRFS_DIR_ITEM_KEY)
 			return 1;
 	}
 	return 0;
 }
-#endif
 
 int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root, struct extent_buffer *parent,
 		       int cache_only, u64 *last_ret)
 {
-	return 0;
-#if 0
-	struct btrfs_node *parent_node;
-	struct extent_buffer *cur_eb;
-	struct extent_buffer *tmp_eb;
+	struct extent_buffer *cur;
+	struct extent_buffer *tmp;
 	u64 blocknr;
 	u64 search_start = *last_ret;
 	u64 last_block = 0;
@@ -214,6 +208,8 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 	int i;
 	int err = 0;
 	int parent_level;
+	int uptodate;
+	u32 blocksize;
 
 	if (trans->transaction != root->fs_info->running_transaction) {
 		printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid,
@@ -225,12 +221,12 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 		       root->fs_info->generation);
 		WARN_ON(1);
 	}
-	if (buffer_defrag_done(parent))
+	if (btrfs_buffer_defrag_done(parent))
 		return 0;
 
-	parent_node = btrfs_buffer_node(parent);
-	parent_nritems = btrfs_header_nritems(&parent_node->header);
-	parent_level = btrfs_header_level(&parent_node->header);
+	parent_nritems = btrfs_header_nritems(parent);
+	parent_level = btrfs_header_level(parent);
+	blocksize = btrfs_level_size(root, parent_level - 1);
 
 	start_slot = 0;
 	end_slot = parent_nritems;
@@ -240,56 +236,60 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 
 	for (i = start_slot; i < end_slot; i++) {
 		int close = 1;
-		blocknr = btrfs_node_blockptr(parent_node, i);
+		blocknr = btrfs_node_blockptr(parent, i);
 		if (last_block == 0)
 			last_block = blocknr;
 		if (i > 0) {
-			other = btrfs_node_blockptr(parent_node, i - 1);
-			close = close_blocks(blocknr, other);
+			other = btrfs_node_blockptr(parent, i - 1);
+			close = close_blocks(blocknr, other, blocksize);
 		}
 		if (close && i < end_slot - 1) {
-			other = btrfs_node_blockptr(parent_node, i + 1);
-			close = close_blocks(blocknr, other);
+			other = btrfs_node_blockptr(parent, i + 1);
+			close = close_blocks(blocknr, other, blocksize);
 		}
 		if (close) {
 			last_block = blocknr;
 			continue;
 		}
 
-		cur_bh = btrfs_find_tree_block(root, blocknr);
-		if (!cur_bh || !buffer_uptodate(cur_bh) ||
-		    buffer_locked(cur_bh) ||
-		    (parent_level != 1 && !buffer_defrag(cur_bh)) ||
-		    (parent_level == 1 && !should_defrag_leaf(cur_bh))) {
+		cur = btrfs_find_tree_block(root, blocknr, blocksize);
+		if (cur)
+			uptodate = btrfs_buffer_uptodate(cur);
+		else
+			uptodate = 0;
+		if (!cur || !uptodate ||
+		    (parent_level != 1 && !btrfs_buffer_defrag(cur)) ||
+		    (parent_level == 1 && !should_defrag_leaf(cur))) {
 			if (cache_only) {
-				brelse(cur_bh);
+				free_extent_buffer(cur);
 				continue;
 			}
-			if (!cur_bh || !buffer_uptodate(cur_bh) ||
-			    buffer_locked(cur_bh)) {
-				brelse(cur_bh);
-				cur_bh = read_tree_block(root, blocknr);
+			if (!cur) {
+				cur = read_tree_block(root, blocknr,
+							 blocksize);
+			} else if (!uptodate) {
+				btrfs_read_buffer(cur);
 			}
 		}
 		if (search_start == 0)
-			search_start = last_block & ~((u64)65535);
+			search_start = last_block;
 
-		err = __btrfs_cow_block(trans, root, cur_bh, parent, i,
-					&tmp_bh, search_start,
-					min(8, end_slot - i));
+		err = __btrfs_cow_block(trans, root, cur, parent, i,
+					&tmp, search_start,
+					min(16 * blocksize,
+					    (end_slot - i) * blocksize));
 		if (err) {
-			brelse(cur_bh);
+			free_extent_buffer(cur);
 			break;
 		}
-		search_start = bh_blocknr(tmp_bh);
+		search_start = tmp->start;
 		*last_ret = search_start;
 		if (parent_level == 1)
-			clear_buffer_defrag(tmp_bh);
-		set_buffer_defrag_done(tmp_bh);
-		brelse(tmp_bh);
+			btrfs_clear_buffer_defrag(tmp);
+		btrfs_set_buffer_defrag_done(tmp);
+		free_extent_buffer(tmp);
 	}
 	return err;
-#endif
 }
 
 /*
@@ -892,22 +892,17 @@ static int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
 			     int level, int slot)
 {
-	return;
-#if 0
 	struct extent_buffer *node;
-	int i;
 	u32 nritems;
-	u64 bytenr;
 	u64 search;
-	u64 cluster_start;
-	int ret;
-	int nread = 0;
+	u64 lowest_read;
+	u64 highest_read;
+	u64 nread = 0;
 	int direction = path->reada;
-	int level;
-	struct radix_tree_root found;
-	unsigned long gang[8];
 	struct extent_buffer *eb;
-
+	u32 nr;
+	u32 blocksize;
+	u32 nscan = 0;
 
 	if (level == 0)
 		return;
@@ -917,42 +912,46 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
 
 	node = path->nodes[level];
 	search = btrfs_node_blockptr(node, slot);
-	eb = btrfs_find_tree_block(root, search);
+	blocksize = btrfs_level_size(root, level - 1);
+	eb = btrfs_find_tree_block(root, search, blocksize);
 	if (eb) {
 		free_extent_buffer(eb);
 		return;
 	}
 
-	init_bit_radix(&found);
+	highest_read = search;
+	lowest_read = search;
+
 	nritems = btrfs_header_nritems(node);
-	level = btrfs_header_level(node) - 1;
-	for (i = slot; i < nritems; i++) {
-		bytenr = btrfs_node_blockptr(node, i);
-		set_radix_bit(&found, blocknr);
-	}
-	if (direction > 0) {
-		cluster_start = search - 4;
-		if (cluster_start > search)
-			cluster_start = 0;
-	} else
-		cluster_start = search + 4;
+	nr = slot;
 	while(1) {
-		ret = find_first_radix_bit(&found, gang, 0, ARRAY_SIZE(gang));
-		if (!ret)
-			break;
-		for (i = 0; i < ret; i++) {
-			blocknr = gang[i];
-			clear_radix_bit(&found, blocknr);
-			if (path->reada == 1 && nread > 16)
-				continue;
-			if (close_blocks(cluster_start, blocknr)) {
-				readahead_tree_block(root, blocknr);
-				nread++;
-				cluster_start = blocknr;
-			}
+		if (direction < 0) {
+			if (nr == 0)
+				break;
+			nr--;
+		} else if (direction > 0) {
+			nr++;
+			if (nr >= nritems)
+				break;
 		}
+		search = btrfs_node_blockptr(node, nr);
+		if ((search >= lowest_read && search <= highest_read) ||
+		    (search < lowest_read && lowest_read - search <= 32768) ||
+		    (search > highest_read && search - highest_read <= 32768)) {
+			readahead_tree_block(root, search, blocksize);
+			nread += blocksize;
+		}
+		nscan++;
+		if (path->reada < 2 && (nread > (256 * 1024) || nscan > 32))
+			break;
+		if(nread > (1024 * 1024) || nscan > 128)
+			break;
+
+		if (search < lowest_read)
+			lowest_read = search;
+		if (search > highest_read)
+			highest_read = search;
 	}
-#endif
 }
 /*
  * look for key in the tree.  path is filled in with nodes along the way
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 0636f79672e9..8e606e6658aa 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -678,3 +678,63 @@ void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
 	balance_dirty_pages_ratelimited_nr(
 			root->fs_info->btree_inode->i_mapping, nr);
 }
+
+void btrfs_set_buffer_defrag(struct extent_buffer *buf)
+{
+	struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	set_extent_bits(&BTRFS_I(btree_inode)->extent_tree, buf->start,
+			buf->start + buf->len - 1, EXTENT_DEFRAG, GFP_NOFS);
+}
+
+void btrfs_set_buffer_defrag_done(struct extent_buffer *buf)
+{
+	struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	set_extent_bits(&BTRFS_I(btree_inode)->extent_tree, buf->start,
+			buf->start + buf->len - 1, EXTENT_DEFRAG_DONE,
+			GFP_NOFS);
+}
+
+int btrfs_buffer_defrag(struct extent_buffer *buf)
+{
+	struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	return test_range_bit(&BTRFS_I(btree_inode)->extent_tree,
+		     buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG, 0);
+}
+
+int btrfs_buffer_defrag_done(struct extent_buffer *buf)
+{
+	struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	return test_range_bit(&BTRFS_I(btree_inode)->extent_tree,
+		     buf->start, buf->start + buf->len - 1,
+		     EXTENT_DEFRAG_DONE, 0);
+}
+
+int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf)
+{
+	struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	return clear_extent_bits(&BTRFS_I(btree_inode)->extent_tree,
+		     buf->start, buf->start + buf->len - 1,
+		     EXTENT_DEFRAG_DONE, GFP_NOFS);
+}
+
+int btrfs_clear_buffer_defrag(struct extent_buffer *buf)
+{
+	struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	return clear_extent_bits(&BTRFS_I(btree_inode)->extent_tree,
+		     buf->start, buf->start + buf->len - 1,
+		     EXTENT_DEFRAG, GFP_NOFS);
+}
+
+int btrfs_read_buffer(struct extent_buffer *buf)
+{
+	struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	return read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
+					buf, 1);
+}
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index fd4db5f810cf..190b07bc642b 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -51,4 +51,11 @@ int btrfs_buffer_uptodate(struct extent_buffer *buf);
 int btrfs_set_buffer_uptodate(struct extent_buffer *buf);
 int wait_on_tree_block_writeback(struct btrfs_root *root,
 				 struct extent_buffer *buf);
+void btrfs_set_buffer_defrag(struct extent_buffer *buf);
+void btrfs_set_buffer_defrag_done(struct extent_buffer *buf);
+int btrfs_buffer_defrag(struct extent_buffer *buf);
+int btrfs_buffer_defrag_done(struct extent_buffer *buf);
+int btrfs_clear_buffer_defrag(struct extent_buffer *buf);
+int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf);
+int btrfs_read_buffer(struct extent_buffer *buf);
 #endif
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 1be8f9f04a13..0b0c947194ae 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1173,13 +1173,7 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 	buf->alloc_addr = (unsigned long)__builtin_return_address(0);
 	set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
 			 buf->start + buf->len - 1, GFP_NOFS);
-	/*
-	set_buffer_checked(buf);
-	set_buffer_defrag(buf);
-	*/
-	/* FIXME!!!!!!!!!!!!!!!!
-	set_radix_bit(&trans->transaction->dirty_pages, buf->pages[0]->index);
-	*/
+	btrfs_set_buffer_defrag(buf);
 	trans->blocks_used++;
 	return buf;
 }
diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h
index d74a2b3e3b5d..8409b5cbeda7 100644
--- a/fs/btrfs/extent_map.h
+++ b/fs/btrfs/extent_map.h
@@ -14,6 +14,8 @@
 #define EXTENT_LOCKED (1 << 3)
 #define EXTENT_NEW (1 << 4)
 #define EXTENT_DELALLOC (1 << 5)
+#define EXTENT_DEFRAG (1 << 6)
+#define EXTENT_DEFRAG_DONE (1 << 7)
 #define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK)
 
 
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 60f61345a8d0..87456ab74275 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -355,7 +355,7 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly)
 		return 0;
 
 	trans = btrfs_start_transaction(root, 1);
-	while (0) {
+	while (1) {
 		root->defrag_running = 1;
 		ret = btrfs_defrag_leaves(trans, root, cacheonly);
 		nr = trans->blocks_used;
@@ -400,7 +400,7 @@ int btrfs_defrag_dirty_roots(struct btrfs_fs_info *info)
 			btrfs_defrag_root(root, 1);
 		}
 	}
-	// btrfs_defrag_root(info->extent_root, 1);
+	btrfs_defrag_root(info->extent_root, 1);
 	return err;
 }
 
diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c
index 3feac2f28628..d23216aa4ab5 100644
--- a/fs/btrfs/tree-defrag.c
+++ b/fs/btrfs/tree-defrag.c
@@ -116,10 +116,10 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 	}
 	WARN_ON(*level < 0);
 	WARN_ON(*level >= BTRFS_MAX_LEVEL);
-#if 0
-	clear_buffer_defrag(path->nodes[*level]);
-	clear_buffer_defrag_done(path->nodes[*level]);
-#endif
+
+	btrfs_clear_buffer_defrag(path->nodes[*level]);
+	btrfs_clear_buffer_defrag_done(path->nodes[*level]);
+
 	free_extent_buffer(path->nodes[*level]);
 	path->nodes[*level] = NULL;
 	*level += 1;
@@ -148,10 +148,8 @@ static int defrag_walk_up(struct btrfs_trans_handle *trans,
 			root->defrag_level = i;
 			return 0;
 		} else {
-			/*
-			clear_buffer_defrag(path->nodes[*level]);
-			clear_buffer_defrag_done(path->nodes[*level]);
-			*/
+			btrfs_clear_buffer_defrag(path->nodes[*level]);
+			btrfs_clear_buffer_defrag_done(path->nodes[*level]);
 			free_extent_buffer(path->nodes[*level]);
 			path->nodes[*level] = NULL;
 			*level = i + 1;
-- 
cgit v1.2.3


From ff79f8190b6e955ff7a71faf804a3017d526e657 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Mon, 15 Oct 2007 16:22:25 -0400
Subject: Btrfs: Add back file data checksumming

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/disk-io.c    |  1 +
 fs/btrfs/disk-io.h    |  2 ++
 fs/btrfs/extent_map.c |  3 +++
 fs/btrfs/file-item.c  | 17 +++++++++--------
 fs/btrfs/file.c       |  1 +
 fs/btrfs/inode.c      | 21 ++++++++++-----------
 6 files changed, 26 insertions(+), 19 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 3878772fcffd..5accced2858a 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -594,6 +594,7 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	tree_root->nodesize = nodesize;
 	tree_root->leafsize = leafsize;
 	tree_root->sectorsize = sectorsize;
+	sb_set_blocksize(sb, sectorsize);
 
 	i_size_write(fs_info->btree_inode,
 		     btrfs_super_total_bytes(disk_super));
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 190b07bc642b..33d2ccfd74d8 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -58,4 +58,6 @@ int btrfs_buffer_defrag_done(struct extent_buffer *buf);
 int btrfs_clear_buffer_defrag(struct extent_buffer *buf);
 int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf);
 int btrfs_read_buffer(struct extent_buffer *buf);
+u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len);
+void btrfs_csum_final(u32 crc, char *result);
 #endif
diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c
index 2a8bc4bd43a9..e87e476dca92 100644
--- a/fs/btrfs/extent_map.c
+++ b/fs/btrfs/extent_map.c
@@ -2031,6 +2031,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree,
 		i = 1;
 		index++;
 		page_cache_get(page0);
+		mark_page_accessed(page0);
 		set_page_extent_mapped(page0);
 		set_page_private(page0, EXTENT_PAGE_PRIVATE_FIRST_PAGE |
 				 len << 2);
@@ -2049,6 +2050,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree,
 			goto fail;
 		}
 		set_page_extent_mapped(p);
+		mark_page_accessed(p);
 		if (i == 0) {
 			eb->first_page = p;
 			set_page_private(p, EXTENT_PAGE_PRIVATE_FIRST_PAGE |
@@ -2099,6 +2101,7 @@ struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree,
 			goto fail;
 		}
 		set_page_extent_mapped(p);
+		mark_page_accessed(p);
 
 		if (i == 0) {
 			eb->first_page = p;
diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c
index 9a65e97a4e28..10a4c7408600 100644
--- a/fs/btrfs/file-item.c
+++ b/fs/btrfs/file-item.c
@@ -136,27 +136,27 @@ int btrfs_csum_file_block(struct btrfs_trans_handle *trans,
 			  u64 objectid, u64 offset,
 			  char *data, size_t len)
 {
-	return 0;
-#if 0
 	int ret;
 	struct btrfs_key file_key;
 	struct btrfs_key found_key;
 	struct btrfs_path *path;
 	struct btrfs_csum_item *item;
-	struct extent_buffer *leaf;
+	struct extent_buffer *leaf = NULL;
 	u64 csum_offset;
+	u32 csum_result = ~(u32)0;
 
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
 
 	file_key.objectid = objectid;
 	file_key.offset = offset;
-	file_key.flags = 0;
 	btrfs_set_key_type(&file_key, BTRFS_CSUM_ITEM_KEY);
 
 	item = btrfs_lookup_csum(trans, root, path, objectid, offset, 1);
-	if (!IS_ERR(item))
+	if (!IS_ERR(item)) {
+		leaf = path->nodes[0];
 		goto found;
+	}
 	ret = PTR_ERR(item);
 	if (ret == -EFBIG) {
 		u32 item_size;
@@ -226,14 +226,15 @@ csum:
 	item = (struct btrfs_csum_item *)((unsigned char *)item +
 					  csum_offset * BTRFS_CRC32_SIZE);
 found:
-	/* FIXME!!!!!!!!!!!! */
-	ret = btrfs_csum_data(root, data, len, &item->csum);
+	csum_result = btrfs_csum_data(root, data, csum_result, len);
+	btrfs_csum_final(csum_result, (char *)&csum_result);
+	write_extent_buffer(leaf, &csum_result, (unsigned long)item,
+			    BTRFS_CRC32_SIZE);
 	btrfs_mark_buffer_dirty(path->nodes[0]);
 fail:
 	btrfs_release_path(root, path);
 	btrfs_free_path(path);
 	return ret;
-#endif
 }
 
 int btrfs_csum_truncate(struct btrfs_trans_handle *trans,
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index fe28404ae7f4..1af2b6534dad 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -289,6 +289,7 @@ static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 	 */
 	inline_size = end_pos;
 	if (isize >= BTRFS_MAX_INLINE_DATA_SIZE(root) ||
+	    inline_size > 16384 ||
 	    inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root)) {
 		u64 last_end;
 
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 80f84b48a70f..ac69f6926afc 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -135,7 +135,7 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 	struct btrfs_csum_item *item;
 	struct btrfs_path *path = NULL;
-	u64 private;
+	u32 csum;
 
 	mutex_lock(&root->fs_info->fs_mutex);
 	path = btrfs_alloc_path();
@@ -145,11 +145,12 @@ int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
 		/* a csum that isn't present is a preallocated region. */
 		if (ret == -ENOENT || ret == -EFBIG)
 			ret = 0;
-		private = 0;
+		csum = 0;
 		goto out;
 	}
-	memcpy((char *)&private, &item->csum, BTRFS_CRC32_SIZE);
-	set_state_private(em_tree, start, private);
+	read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
+			   BTRFS_CRC32_SIZE);
+	set_state_private(em_tree, start, csum);
 out:
 	if (path)
 		btrfs_free_path(path);
@@ -165,21 +166,19 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end)
 	char *kaddr;
 	u64 private;
 	int ret;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	u32 csum = ~(u32)0;
 
 	ret = get_state_private(em_tree, start, &private);
 	kaddr = kmap_atomic(page, KM_IRQ0);
 	if (ret) {
 		goto zeroit;
 	}
-	/*
-	struct btrfs_root *root = BTRFS_I(inode)->root;
-	char csum[BTRFS_CRC32_SIZE];
-	ret = btrfs_csum_data(root, kaddr + offset, end - start + 1, csum);
-	BUG_ON(ret);
-	if (memcmp(csum, &private, BTRFS_CRC32_SIZE)) {
+	csum = btrfs_csum_data(root, kaddr + offset, csum,  end - start + 1);
+	btrfs_csum_final(csum, (char *)&csum);
+	if (csum != private) {
 		goto zeroit;
 	}
-	*/
 	kunmap_atomic(kaddr, KM_IRQ0);
 	return 0;
 
-- 
cgit v1.2.3


From edbd8d4efe4ddaf29a175ae504e2c9a05a96ebee Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Fri, 21 Dec 2007 16:27:24 -0500
Subject: Btrfs: Support for online FS resize (grow and shrink)

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.h       |  10 ++
 fs/btrfs/disk-io.c     |  32 +++-
 fs/btrfs/disk-io.h     |   2 +
 fs/btrfs/extent-tree.c | 480 +++++++++++++++++++++++++++++++++++++++++++++++++
 fs/btrfs/inode.c       | 105 ++++++++++-
 fs/btrfs/ioctl.h       |   2 +
 fs/btrfs/root-tree.c   |   5 +
 fs/btrfs/super.c       |   7 +-
 8 files changed, 629 insertions(+), 14 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 10129cc6656f..b83a1bec346f 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -508,6 +508,8 @@ BTRFS_SETGET_STACK_FUNCS(block_group_used, struct btrfs_block_group_item,
 			 used, 64);
 BTRFS_SETGET_FUNCS(disk_block_group_used, struct btrfs_block_group_item,
 			 used, 64);
+BTRFS_SETGET_FUNCS(disk_block_group_flags, struct btrfs_block_group_item,
+		   flags, 8);
 
 /* struct btrfs_inode_ref */
 BTRFS_SETGET_FUNCS(inode_ref_name_len, struct btrfs_inode_ref, name_len, 16);
@@ -960,6 +962,9 @@ struct extent_buffer *__btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 					     int level,
 					     u64 hint,
 					     u64 empty_size);
+int btrfs_grow_extent_tree(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *root, u64 new_size);
+int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size);
 int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans,
 				 struct btrfs_root *root,
 				 struct btrfs_path *path, u64 bytenr,
@@ -1117,6 +1122,9 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root, struct btrfs_path *path,
 			u64 isize);
 /* inode.c */
+unsigned long btrfs_force_ra(struct address_space *mapping,
+			      struct file_ra_state *ra, struct file *file,
+			      pgoff_t offset, pgoff_t last_index);
 int btrfs_check_free_space(struct btrfs_root *root, u64 num_required,
 			   int for_del);
 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page);
@@ -1162,4 +1170,6 @@ void btrfs_sysfs_del_super(struct btrfs_fs_info *root);
 ssize_t btrfs_listxattr(struct dentry *dentry, char *buffer, size_t size);
 int btrfs_delete_xattrs(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root, struct inode *inode);
+/* super.c */
+u64 btrfs_parse_size(char *str);
 #endif
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index eebb4fb65c61..de026d9d9b2b 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -471,13 +471,17 @@ insert:
 	return root;
 }
 
-struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
-				      struct btrfs_key *location,
-				      const char *name, int namelen)
+struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
+					      struct btrfs_key *location)
 {
 	struct btrfs_root *root;
 	int ret;
 
+	if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
+		return fs_info->tree_root;
+	if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
+		return fs_info->extent_root;
+
 	root = radix_tree_lookup(&fs_info->fs_roots_radix,
 				 (unsigned long)location->objectid);
 	if (root)
@@ -494,6 +498,23 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 		kfree(root);
 		return ERR_PTR(ret);
 	}
+	ret = btrfs_find_dead_roots(fs_info->tree_root,
+				    root->root_key.objectid, root);
+	BUG_ON(ret);
+
+	return root;
+}
+
+struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
+				      struct btrfs_key *location,
+				      const char *name, int namelen)
+{
+	struct btrfs_root *root;
+	int ret;
+
+	root = btrfs_read_fs_root_no_name(fs_info, location);
+	if (!root)
+		return NULL;
 
 	ret = btrfs_set_root_name(root, name, namelen);
 	if (ret) {
@@ -509,11 +530,6 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 		kfree(root);
 		return ERR_PTR(ret);
 	}
-
-	ret = btrfs_find_dead_roots(fs_info->tree_root,
-				    root->root_key.objectid, root);
-	BUG_ON(ret);
-
 	return root;
 }
 #if 0
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 33d2ccfd74d8..8c3cfd02901f 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -39,6 +39,8 @@ struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 				      const char *name, int namelen);
 struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
 					       struct btrfs_key *location);
+struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
+					      struct btrfs_key *location);
 int btrfs_insert_dev_radix(struct btrfs_root *root,
 			   struct block_device *bdev,
 			   u64 device_id,
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 1412d556313f..de0fb0743cf9 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -18,6 +18,7 @@
 
 #include <linux/sched.h>
 #include <linux/crc32c.h>
+#include <linux/pagemap.h>
 #include "hash.h"
 #include "ctree.h"
 #include "disk-io.h"
@@ -1622,6 +1623,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 	int pending_ret;
 	u64 super_used, root_used;
 	u64 search_start = 0;
+	u64 new_hint;
 	struct btrfs_fs_info *info = root->fs_info;
 	struct btrfs_root *extent_root = info->extent_root;
 	struct btrfs_extent_item extent_item;
@@ -1629,6 +1631,10 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 
 	btrfs_set_stack_extent_refs(&extent_item, 1);
 
+	new_hint = max(hint_byte, 16ULL * 1024 * 1024 * 1024);
+	if (new_hint < btrfs_super_total_bytes(&info->super_copy))
+		hint_byte = new_hint;
+
 	WARN_ON(num_bytes < root->sectorsize);
 	ret = find_free_extent(trans, root, num_bytes, empty_size,
 			       search_start, search_end, hint_byte, ins,
@@ -2100,6 +2106,480 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info)
 	return 0;
 }
 
+static int relocate_inode_pages(struct inode *inode, u64 start, u64 len)
+{
+	u64 page_start;
+	u64 page_end;
+	u64 delalloc_start;
+	u64 existing_delalloc;
+	unsigned long last_index;
+	unsigned long first_index;
+	unsigned long i;
+	struct page *page;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+	struct file_ra_state ra;
+
+	mutex_lock(&inode->i_mutex);
+	first_index = start >> PAGE_CACHE_SHIFT;
+	last_index = (start + len - 1) >> PAGE_CACHE_SHIFT;
+
+	memset(&ra, 0, sizeof(ra));
+	file_ra_state_init(&ra, inode->i_mapping);
+	btrfs_force_ra(inode->i_mapping, &ra, NULL, first_index, last_index);
+
+	for (i = first_index; i <= last_index; i++) {
+		page = grab_cache_page(inode->i_mapping, i);
+		if (!page)
+			goto out_unlock;
+		if (!PageUptodate(page)) {
+			btrfs_readpage(NULL, page);
+			lock_page(page);
+			if (!PageUptodate(page)) {
+				unlock_page(page);
+				page_cache_release(page);
+				goto out_unlock;
+			}
+		}
+		page_start = (u64)page->index << PAGE_CACHE_SHIFT;
+		page_end = page_start + PAGE_CACHE_SIZE - 1;
+
+		lock_extent(em_tree, page_start, page_end, GFP_NOFS);
+
+		delalloc_start = page_start;
+		existing_delalloc =
+			count_range_bits(&BTRFS_I(inode)->extent_tree,
+					 &delalloc_start, page_end,
+					 PAGE_CACHE_SIZE, EXTENT_DELALLOC);
+
+		set_extent_delalloc(em_tree, page_start,
+				    page_end, GFP_NOFS);
+
+		spin_lock(&root->fs_info->delalloc_lock);
+		root->fs_info->delalloc_bytes += PAGE_CACHE_SIZE -
+						 existing_delalloc;
+		spin_unlock(&root->fs_info->delalloc_lock);
+
+		unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
+		set_page_dirty(page);
+		unlock_page(page);
+		page_cache_release(page);
+	}
+
+out_unlock:
+	mutex_unlock(&inode->i_mutex);
+	return 0;
+}
+
+static int relocate_one_reference(struct btrfs_root *extent_root,
+				  struct btrfs_path *path,
+				  struct btrfs_key *extent_key,
+				  u64 ref_root, u64 ref_gen, u64 ref_objectid,
+				  u64 ref_offset)
+{
+	struct inode *inode;
+	struct btrfs_root *found_root;
+	struct btrfs_key root_location;
+	int ret;
+
+	root_location.objectid = ref_root;
+	if (ref_gen == 0)
+		root_location.offset = 0;
+	else
+		root_location.offset = (u64)-1;
+	root_location.type = BTRFS_ROOT_ITEM_KEY;
+
+	found_root = btrfs_read_fs_root_no_name(extent_root->fs_info,
+						&root_location);
+	BUG_ON(!found_root);
+
+	if (ref_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
+		mutex_unlock(&extent_root->fs_info->fs_mutex);
+		inode = btrfs_iget_locked(extent_root->fs_info->sb,
+					  ref_objectid, found_root);
+		if (inode->i_state & I_NEW) {
+			/* the inode and parent dir are two different roots */
+			BTRFS_I(inode)->root = found_root;
+			BTRFS_I(inode)->location.objectid = ref_objectid;
+			BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
+			BTRFS_I(inode)->location.offset = 0;
+			btrfs_read_locked_inode(inode);
+			unlock_new_inode(inode);
+
+		}
+		/* this can happen if the reference is not against
+		 * the latest version of the tree root
+		 */
+		if (is_bad_inode(inode)) {
+			mutex_lock(&extent_root->fs_info->fs_mutex);
+			goto out;
+		}
+		relocate_inode_pages(inode, ref_offset, extent_key->offset);
+		/* FIXME, data=ordered will help get rid of this */
+		filemap_fdatawrite(inode->i_mapping);
+		iput(inode);
+		mutex_lock(&extent_root->fs_info->fs_mutex);
+	} else {
+		struct btrfs_trans_handle *trans;
+		struct btrfs_key found_key;
+		struct extent_buffer *eb;
+		int level;
+		int i;
+
+		trans = btrfs_start_transaction(found_root, 1);
+		eb = read_tree_block(found_root, extent_key->objectid,
+				     extent_key->offset);
+		level = btrfs_header_level(eb);
+
+		if (level == 0)
+			btrfs_item_key_to_cpu(eb, &found_key, 0);
+		else
+			btrfs_node_key_to_cpu(eb, &found_key, 0);
+
+		free_extent_buffer(eb);
+
+		path->lowest_level = level;
+		path->reada = 0;
+		ret = btrfs_search_slot(trans, found_root, &found_key, path,
+					0, 1);
+		path->lowest_level = 0;
+		for (i = level; i < BTRFS_MAX_LEVEL; i++) {
+			if (!path->nodes[i])
+				break;
+			free_extent_buffer(path->nodes[i]);
+			path->nodes[i] = NULL;
+		}
+		btrfs_release_path(found_root, path);
+		btrfs_end_transaction(trans, found_root);
+	}
+
+out:
+	return 0;
+}
+
+static int relocate_one_extent(struct btrfs_root *extent_root,
+			       struct btrfs_path *path,
+			       struct btrfs_key *extent_key)
+{
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	struct btrfs_extent_ref *ref;
+	struct extent_buffer *leaf;
+	u64 ref_root;
+	u64 ref_gen;
+	u64 ref_objectid;
+	u64 ref_offset;
+	u32 nritems;
+	u32 item_size;
+	int ret = 0;
+
+	key.objectid = extent_key->objectid;
+	key.type = BTRFS_EXTENT_REF_KEY;
+	key.offset = 0;
+
+	while(1) {
+		ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
+
+		BUG_ON(ret == 0);
+
+		if (ret < 0)
+			goto out;
+
+		ret = 0;
+		leaf = path->nodes[0];
+		nritems = btrfs_header_nritems(leaf);
+		if (path->slots[0] == nritems)
+			goto out;
+
+		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+		if (found_key.objectid != extent_key->objectid)
+			break;
+
+		if (found_key.type != BTRFS_EXTENT_REF_KEY)
+			break;
+
+		key.offset = found_key.offset + 1;
+		item_size = btrfs_item_size_nr(leaf, path->slots[0]);
+
+		ref = btrfs_item_ptr(leaf, path->slots[0],
+				     struct btrfs_extent_ref);
+		ref_root = btrfs_ref_root(leaf, ref);
+		ref_gen = btrfs_ref_generation(leaf, ref);
+		ref_objectid = btrfs_ref_objectid(leaf, ref);
+		ref_offset = btrfs_ref_offset(leaf, ref);
+		btrfs_release_path(extent_root, path);
+
+		ret = relocate_one_reference(extent_root, path,
+					     extent_key, ref_root, ref_gen,
+					     ref_objectid, ref_offset);
+		if (ret)
+			goto out;
+	}
+	ret = 0;
+out:
+	btrfs_release_path(extent_root, path);
+	return ret;
+}
+
+static int find_overlapping_extent(struct btrfs_root *root,
+				   struct btrfs_path *path, u64 new_size)
+{
+	struct btrfs_key found_key;
+	struct extent_buffer *leaf;
+	int ret;
+
+	while(1) {
+		if (path->slots[0] == 0) {
+			ret = btrfs_prev_leaf(root, path);
+			if (ret == 1) {
+				return 1;
+			}
+			if (ret < 0)
+				return ret;
+		} else {
+			path->slots[0]--;
+		}
+		leaf = path->nodes[0];
+		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+		if (found_key.type == BTRFS_EXTENT_ITEM_KEY) {
+			if (found_key.objectid + found_key.offset > new_size)
+				return 0;
+			else
+				return 1;
+		}
+	}
+	return 1;
+}
+
+int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size)
+{
+	struct btrfs_trans_handle *trans;
+	struct btrfs_root *tree_root = root->fs_info->tree_root;
+	struct btrfs_path *path;
+	u64 cur_byte;
+	u64 total_found;
+	u64 ptr;
+	struct btrfs_fs_info *info = root->fs_info;
+	struct extent_map_tree *block_group_cache;
+	struct btrfs_key key;
+	struct btrfs_key found_key = { 0, 0, 0 };
+	struct extent_buffer *leaf;
+	u32 nritems;
+	int ret;
+	int slot;
+
+	btrfs_set_super_total_bytes(&info->super_copy, new_size);
+	block_group_cache = &info->block_group_cache;
+	path = btrfs_alloc_path();
+	root = root->fs_info->extent_root;
+
+again:
+	total_found = 0;
+	key.objectid = new_size;
+	cur_byte = key.objectid;
+	key.offset = 0;
+	key.type = 0;
+	while(1) {
+		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+		if (ret < 0)
+			goto out;
+next:
+		leaf = path->nodes[0];
+		if (key.objectid == new_size - 1) {
+			ret = find_overlapping_extent(root, path, new_size);
+			if (ret != 0) {
+				btrfs_release_path(root, path);
+				ret = btrfs_search_slot(NULL, root, &key,
+							path, 0, 0);
+				if (ret < 0)
+					goto out;
+			}
+		}
+		nritems = btrfs_header_nritems(leaf);
+		ret = 0;
+		slot = path->slots[0];
+		if (slot < nritems)
+			btrfs_item_key_to_cpu(leaf, &found_key, slot);
+		if (slot == nritems ||
+		    btrfs_key_type(&found_key) != BTRFS_EXTENT_ITEM_KEY) {
+			path->slots[0]++;
+			if (path->slots[0] >= nritems) {
+				ret = btrfs_next_leaf(root, path);
+				if (ret < 0)
+					goto out;
+				if (ret == 1) {
+					ret = 0;
+					break;
+				}
+			}
+			goto next;
+		}
+		btrfs_item_key_to_cpu(leaf, &found_key, slot);
+		if (found_key.objectid + found_key.offset <= cur_byte)
+			continue;
+		total_found++;
+		cur_byte = found_key.objectid + found_key.offset;
+		key.objectid = cur_byte;
+		btrfs_release_path(root, path);
+		ret = relocate_one_extent(root, path, &found_key);
+	}
+
+	btrfs_release_path(root, path);
+
+	if (total_found > 0) {
+		trans = btrfs_start_transaction(tree_root, 1);
+		btrfs_commit_transaction(trans, tree_root);
+
+		mutex_unlock(&root->fs_info->fs_mutex);
+		btrfs_clean_old_snapshots(tree_root);
+		mutex_lock(&root->fs_info->fs_mutex);
+
+		trans = btrfs_start_transaction(tree_root, 1);
+		btrfs_commit_transaction(trans, tree_root);
+		goto again;
+	}
+
+	trans = btrfs_start_transaction(root, 1);
+	key.objectid = new_size;
+	key.offset = 0;
+	key.type = 0;
+	while(1) {
+		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
+		if (ret < 0)
+			goto out;
+bg_next:
+		leaf = path->nodes[0];
+		nritems = btrfs_header_nritems(leaf);
+		ret = 0;
+		slot = path->slots[0];
+		if (slot < nritems)
+			btrfs_item_key_to_cpu(leaf, &found_key, slot);
+		if (slot == nritems ||
+		    btrfs_key_type(&found_key) != BTRFS_BLOCK_GROUP_ITEM_KEY) {
+			if (slot < nritems) {
+				printk("shrinker found key %Lu %u %Lu\n",
+				       found_key.objectid, found_key.type,
+				       found_key.offset);
+				path->slots[0]++;
+			}
+			if (path->slots[0] >= nritems) {
+				ret = btrfs_next_leaf(root, path);
+				if (ret < 0)
+					break;
+				if (ret == 1) {
+					ret = 0;
+					break;
+				}
+			}
+			goto bg_next;
+		}
+		btrfs_item_key_to_cpu(leaf, &found_key, slot);
+		ret = get_state_private(&info->block_group_cache,
+					found_key.objectid, &ptr);
+		if (!ret)
+			kfree((void *)(unsigned long)ptr);
+
+		clear_extent_bits(&info->block_group_cache, found_key.objectid,
+				  found_key.objectid + found_key.offset - 1,
+				  (unsigned int)-1, GFP_NOFS);
+
+		key.objectid = found_key.objectid + 1;
+		btrfs_del_item(trans, root, path);
+		btrfs_release_path(root, path);
+	}
+	clear_extent_dirty(&info->free_space_cache, new_size, (u64)-1,
+			   GFP_NOFS);
+	btrfs_commit_transaction(trans, root);
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+int btrfs_grow_extent_tree(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *root, u64 new_size)
+{
+	struct btrfs_path *path;
+	u64 nr = 0;
+	u64 cur_byte;
+	u64 old_size;
+	struct btrfs_block_group_cache *cache;
+	struct btrfs_block_group_item *item;
+	struct btrfs_fs_info *info = root->fs_info;
+	struct extent_map_tree *block_group_cache;
+	struct btrfs_key key;
+	struct extent_buffer *leaf;
+	int ret;
+	int bit;
+
+	old_size = btrfs_super_total_bytes(&info->super_copy);
+	block_group_cache = &info->block_group_cache;
+
+	root = info->extent_root;
+
+	cache = btrfs_lookup_block_group(root->fs_info, old_size - 1);
+
+	cur_byte = cache->key.objectid + cache->key.offset;
+	if (cur_byte >= new_size)
+		goto set_size;
+
+	key.offset = BTRFS_BLOCK_GROUP_SIZE;
+	btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	while(cur_byte < new_size) {
+		key.objectid = cur_byte;
+		ret = btrfs_insert_empty_item(trans, root, path, &key,
+				        sizeof(struct btrfs_block_group_item));
+		BUG_ON(ret);
+		leaf = path->nodes[0];
+		item = btrfs_item_ptr(leaf, path->slots[0],
+				      struct btrfs_block_group_item);
+
+		btrfs_set_disk_block_group_used(leaf, item, 0);
+		if (nr % 3) {
+			btrfs_set_disk_block_group_flags(leaf, item,
+						 BTRFS_BLOCK_GROUP_DATA);
+		} else {
+			btrfs_set_disk_block_group_flags(leaf, item, 0);
+		}
+		nr++;
+
+		cache = kmalloc(sizeof(*cache), GFP_NOFS);
+		BUG_ON(!cache);
+
+		read_extent_buffer(leaf, &cache->item, (unsigned long)item,
+				   sizeof(cache->item));
+
+		memcpy(&cache->key, &key, sizeof(key));
+		cache->cached = 0;
+		cache->pinned = 0;
+		cur_byte = key.objectid + key.offset;
+		btrfs_release_path(root, path);
+
+		if (cache->item.flags & BTRFS_BLOCK_GROUP_DATA) {
+			bit = BLOCK_GROUP_DATA;
+			cache->data = BTRFS_BLOCK_GROUP_DATA;
+		} else {
+			bit = BLOCK_GROUP_METADATA;
+			cache->data = 0;
+		}
+
+		/* use EXTENT_LOCKED to prevent merging */
+		set_extent_bits(block_group_cache, key.objectid,
+				key.objectid + key.offset - 1,
+				bit | EXTENT_LOCKED, GFP_NOFS);
+		set_state_private(block_group_cache, key.objectid,
+				  (unsigned long)cache);
+	}
+	btrfs_free_path(path);
+set_size:
+	btrfs_set_super_total_bytes(&info->super_copy, new_size);
+	return 0;
+}
+
 int btrfs_read_block_groups(struct btrfs_root *root)
 {
 	struct btrfs_path *path;
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 5f5b7b89b144..f6a20112d9e7 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -2346,7 +2346,7 @@ fail_unlock:
 	return ret;
 }
 
-static unsigned long force_ra(struct address_space *mapping,
+unsigned long btrfs_force_ra(struct address_space *mapping,
 			      struct file_ra_state *ra, struct file *file,
 			      pgoff_t offset, pgoff_t last_index)
 {
@@ -2372,6 +2372,8 @@ int btrfs_defrag_file(struct file *file) {
 	unsigned long ra_index = 0;
 	u64 page_start;
 	u64 page_end;
+	u64 delalloc_start;
+	u64 existing_delalloc;
 	unsigned long i;
 	int ret;
 
@@ -2385,8 +2387,9 @@ int btrfs_defrag_file(struct file *file) {
 	last_index = inode->i_size >> PAGE_CACHE_SHIFT;
 	for (i = 0; i <= last_index; i++) {
 		if (i == ra_index) {
-			ra_index = force_ra(inode->i_mapping, &file->f_ra,
-					    file, ra_index, last_index);
+			ra_index = btrfs_force_ra(inode->i_mapping,
+						  &file->f_ra,
+						  file, ra_index, last_index);
 		}
 		page = grab_cache_page(inode->i_mapping, i);
 		if (!page)
@@ -2404,8 +2407,19 @@ int btrfs_defrag_file(struct file *file) {
 		page_end = page_start + PAGE_CACHE_SIZE - 1;
 
 		lock_extent(em_tree, page_start, page_end, GFP_NOFS);
+		delalloc_start = page_start;
+		existing_delalloc =
+			count_range_bits(&BTRFS_I(inode)->extent_tree,
+					 &delalloc_start, page_end,
+					 PAGE_CACHE_SIZE, EXTENT_DELALLOC);
 		set_extent_delalloc(em_tree, page_start,
 				    page_end, GFP_NOFS);
+
+		spin_lock(&root->fs_info->delalloc_lock);
+		root->fs_info->delalloc_bytes += PAGE_CACHE_SIZE -
+						 existing_delalloc;
+		spin_unlock(&root->fs_info->delalloc_lock);
+
 		unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
 		set_page_dirty(page);
 		unlock_page(page);
@@ -2418,6 +2432,89 @@ out_unlock:
 	return 0;
 }
 
+static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
+{
+	u64 new_size;
+	u64 old_size;
+	struct btrfs_ioctl_vol_args *vol_args;
+	struct btrfs_trans_handle *trans;
+	char *sizestr;
+	int ret = 0;
+	int namelen;
+	int mod = 0;
+
+	vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
+
+	if (!vol_args)
+		return -ENOMEM;
+
+	if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
+		ret = -EFAULT;
+		goto out;
+	}
+	namelen = strlen(vol_args->name);
+	if (namelen > BTRFS_VOL_NAME_MAX) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	sizestr = vol_args->name;
+	if (!strcmp(sizestr, "max"))
+		new_size = root->fs_info->sb->s_bdev->bd_inode->i_size;
+	else {
+		if (sizestr[0] == '-') {
+			mod = -1;
+			sizestr++;
+		} else if (sizestr[0] == '+') {
+			mod = 1;
+			sizestr++;
+		}
+		new_size = btrfs_parse_size(sizestr);
+		if (new_size == 0) {
+			ret = -EINVAL;
+			goto out;
+		}
+	}
+
+	mutex_lock(&root->fs_info->fs_mutex);
+	old_size = btrfs_super_total_bytes(&root->fs_info->super_copy);
+
+	if (mod < 0) {
+		if (new_size > old_size) {
+			ret = -EINVAL;
+			goto out_unlock;
+		}
+		new_size = old_size - new_size;
+	} else if (mod > 0) {
+		new_size = old_size + new_size;
+	}
+
+	if (new_size < 256 * 1024 * 1024) {
+		ret = -EINVAL;
+		goto out_unlock;
+	}
+	if (new_size > root->fs_info->sb->s_bdev->bd_inode->i_size) {
+		ret = -EFBIG;
+		goto out_unlock;
+	}
+	new_size = (new_size / root->sectorsize) * root->sectorsize;
+
+printk("new size is %Lu\n", new_size);
+	if (new_size > old_size) {
+		trans = btrfs_start_transaction(root, 1);
+		ret = btrfs_grow_extent_tree(trans, root, new_size);
+		btrfs_commit_transaction(trans, root);
+	} else {
+		ret = btrfs_shrink_extent_tree(root, new_size);
+	}
+
+out_unlock:
+	mutex_unlock(&root->fs_info->fs_mutex);
+out:
+	kfree(vol_args);
+	return ret;
+}
+
 static int btrfs_ioctl_snap_create(struct btrfs_root *root, void __user *arg)
 {
 	struct btrfs_ioctl_vol_args *vol_args;
@@ -2510,6 +2607,8 @@ long btrfs_ioctl(struct file *file, unsigned int
 		return btrfs_ioctl_snap_create(root, (void __user *)arg);
 	case BTRFS_IOC_DEFRAG:
 		return btrfs_ioctl_defrag(file);
+	case BTRFS_IOC_RESIZE:
+		return btrfs_ioctl_resize(root, (void __user *)arg);
 	}
 
 	return -ENOTTY;
diff --git a/fs/btrfs/ioctl.h b/fs/btrfs/ioctl.h
index 8bc47dec2865..8c6290665d49 100644
--- a/fs/btrfs/ioctl.h
+++ b/fs/btrfs/ioctl.h
@@ -30,4 +30,6 @@ struct btrfs_ioctl_vol_args {
 				   struct btrfs_ioctl_vol_args)
 #define BTRFS_IOC_DEFRAG _IOW(BTRFS_IOCTL_MAGIC, 2, \
 				   struct btrfs_ioctl_vol_args)
+#define BTRFS_IOC_RESIZE _IOW(BTRFS_IOCTL_MAGIC, 3, \
+				   struct btrfs_ioctl_vol_args)
 #endif
diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c
index 88bcdd33f56e..e2b9b86af696 100644
--- a/fs/btrfs/root-tree.c
+++ b/fs/btrfs/root-tree.c
@@ -179,6 +179,11 @@ int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	ret = btrfs_search_slot(trans, root, key, path, -1, 1);
 	if (ret < 0)
 		goto out;
+	if (ret) {
+btrfs_print_leaf(root, path->nodes[0]);
+printk("failed to del %Lu %u %Lu\n", key->objectid, key->type, key->offset);
+
+	}
 	BUG_ON(ret != 0);
 	leaf = path->nodes[0];
 	ri = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_item);
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 7cab680418a1..3bbd7f665694 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -74,9 +74,9 @@ static match_table_t tokens = {
 	{Opt_err, NULL}
 };
 
-static unsigned long parse_size(char *str)
+u64 btrfs_parse_size(char *str)
 {
-	unsigned long res;
+	u64 res;
 	int mult = 1;
 	char *end;
 	char last;
@@ -150,7 +150,8 @@ static int parse_options (char * options,
 			if (info) {
 				char *num = match_strdup(&args[0]);
 				if (num) {
-					info->max_extent = parse_size(num);
+					info->max_extent =
+						btrfs_parse_size(num);
 					kfree(num);
 
 					info->max_extent = max_t(u64,
-- 
cgit v1.2.3


From dc17ff8f11d129db9e83ab7244769e4eae05e14d Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Tue, 8 Jan 2008 15:46:30 -0500
Subject: Btrfs: Add data=ordered support

This forces file data extents down the disk along with the metadata that
references them.  The current implementation is fairly simple, and just
writes out all of the dirty pages in an inode before the commit.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/Makefile       |   2 +-
 fs/btrfs/btrfs_inode.h  |   1 +
 fs/btrfs/ctree.c        |   6 +-
 fs/btrfs/ctree.h        |   7 +-
 fs/btrfs/disk-io.c      |  16 +++-
 fs/btrfs/disk-io.h      |   2 +
 fs/btrfs/extent-tree.c  |   4 +-
 fs/btrfs/file.c         |   5 +-
 fs/btrfs/inode.c        |  46 ++++++----
 fs/btrfs/ordered-data.c | 221 ++++++++++++++++++++++++++++++++++++++++++++++++
 fs/btrfs/ordered-data.h |  39 +++++++++
 fs/btrfs/transaction.c  |  58 +++++++++++++
 fs/btrfs/transaction.h  |   8 +-
 13 files changed, 387 insertions(+), 28 deletions(-)
 create mode 100644 fs/btrfs/ordered-data.c
 create mode 100644 fs/btrfs/ordered-data.h

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index d5804c5ca075..ab9a9f8edbf8 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -5,7 +5,7 @@ obj-m  := btrfs.o
 btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
 	   hash.o file-item.o inode-item.o inode-map.o disk-io.o \
 	   transaction.o bit-radix.o inode.o file.o tree-defrag.o \
-	   extent_map.o sysfs.o struct-funcs.o xattr.o acl.o
+	   extent_map.o sysfs.o struct-funcs.o xattr.o acl.o ordered-data.o
 
 #btrfs-y := ctree.o disk-io.o radix-tree.o extent-tree.o print-tree.o \
 #	  root-tree.o dir-item.o hash.o file-item.o inode-item.o \
diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
index d1d5af471c3c..f27e633f1742 100644
--- a/fs/btrfs/btrfs_inode.h
+++ b/fs/btrfs/btrfs_inode.h
@@ -29,6 +29,7 @@ struct btrfs_inode {
 	struct extent_map_tree extent_tree;
 	struct inode vfs_inode;
 
+	u64 ordered_trans;
 	/*
 	 * transid of the trans_handle that last modified this inode
 	 */
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 35c57074a376..43d23148a4fe 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -221,7 +221,9 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		    struct extent_buffer **cow_ret)
 {
 	u64 search_start;
+	u64 header_trans;
 	int ret;
+
 	if (trans->transaction != root->fs_info->running_transaction) {
 		printk(KERN_CRIT "trans %Lu running %Lu\n", trans->transid,
 		       root->fs_info->running_transaction->transid);
@@ -232,7 +234,9 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		       root->fs_info->generation);
 		WARN_ON(1);
 	}
-	if (btrfs_header_generation(buf) == trans->transid) {
+
+	header_trans = btrfs_header_generation(buf);
+	if (header_trans == trans->transid) {
 		*cow_ret = buf;
 		return 0;
 	}
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 9873975ce0ee..b55dba58dfaa 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -16,8 +16,8 @@
  * Boston, MA 021110-1307, USA.
  */
 
-#ifndef __BTRFS__
-#define __BTRFS__
+#ifndef __BTRFS_CTREE__
+#define __BTRFS_CTREE__
 
 #include <linux/version.h>
 #include <linux/mm.h>
@@ -363,7 +363,6 @@ struct btrfs_root {
 	struct inode *inode;
 	struct kobject root_kobj;
 	struct completion kobj_unregister;
-	struct rw_semaphore snap_sem;
 	u64 objectid;
 	u64 last_trans;
 
@@ -1142,6 +1141,8 @@ void btrfs_destroy_cachep(void);
 long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
 				struct btrfs_root *root);
+struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
+			    u64 root_objectid);
 int btrfs_commit_write(struct file *file, struct page *page,
 		       unsigned from, unsigned to);
 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index a6170ff19e7f..34cf1f1f47be 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -406,7 +406,6 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 	memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
 	memset(&root->root_kobj, 0, sizeof(root->root_kobj));
 	init_completion(&root->kobj_unregister);
-	init_rwsem(&root->snap_sem);
 	root->defrag_running = 0;
 	root->defrag_level = 0;
 	root->root_key.objectid = objectid;
@@ -498,6 +497,21 @@ insert:
 	return root;
 }
 
+struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
+					u64 root_objectid)
+{
+	struct btrfs_root *root;
+
+	if (root_objectid == BTRFS_ROOT_TREE_OBJECTID)
+		return fs_info->tree_root;
+	if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID)
+		return fs_info->extent_root;
+
+	root = radix_tree_lookup(&fs_info->fs_roots_radix,
+				 (unsigned long)root_objectid);
+	return root;
+}
+
 struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
 					      struct btrfs_key *location)
 {
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 8c3cfd02901f..dae9fba8efcd 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -34,6 +34,8 @@ int write_ctree_super(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root);
 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
 					    u64 bytenr, u32 blocksize);
+struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
+					u64 root_objectid);
 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 				      struct btrfs_key *location,
 				      const char *name, int namelen);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index c906bb19b211..68137cd8506a 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1195,7 +1195,9 @@ static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes,
 			if (btrfs_buffer_uptodate(buf)) {
 				u64 transid =
 				    root->fs_info->running_transaction->transid;
-				if (btrfs_header_generation(buf) == transid) {
+				u64 header_transid =
+					btrfs_header_generation(buf);
+				if (header_transid == transid) {
 					free_extent_buffer(buf);
 					return 1;
 				}
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 94c93373cb7d..0a5f4defe59b 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -34,6 +34,7 @@
 #include "disk-io.h"
 #include "transaction.h"
 #include "btrfs_inode.h"
+#include "ordered-data.h"
 #include "ioctl.h"
 #include "print-tree.h"
 
@@ -329,6 +330,7 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans,
 		root->fs_info->delalloc_bytes += (end_of_last_block + 1 -
 					  start_pos) - existing_delalloc;
 		spin_unlock(&root->fs_info->delalloc_lock);
+		btrfs_add_ordered_inode(inode);
 	} else {
 		u64 aligned_end;
 		/* step one, delete the existing extents in this range */
@@ -724,8 +726,6 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 
 	pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL);
 
-	down_read(&BTRFS_I(inode)->root->snap_sem);
-
 	mutex_lock(&inode->i_mutex);
 	first_index = pos >> PAGE_CACHE_SHIFT;
 	last_index = (pos + count) >> PAGE_CACHE_SHIFT;
@@ -804,7 +804,6 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 	}
 out:
 	mutex_unlock(&inode->i_mutex);
-	up_read(&BTRFS_I(inode)->root->snap_sem);
 
 out_nolock:
 	kfree(pages);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 6d0cd9a7a615..6d6e1ac0a9a0 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -135,6 +135,7 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end)
 		alloc_hint = ins.objectid + ins.offset;
 		start += cur_alloc_size;
 	}
+	btrfs_add_ordered_inode(inode);
 out:
 	btrfs_end_transaction(trans, root);
 	return ret;
@@ -367,8 +368,8 @@ void btrfs_read_locked_inode(struct inode *inode)
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
 	mutex_lock(&root->fs_info->fs_mutex);
-
 	memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
+
 	ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
 	if (ret)
 		goto make_bad;
@@ -898,7 +899,6 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
 	if ((offset & (blocksize - 1)) == 0)
 		goto out;
 
-	down_read(&root->snap_sem);
 	ret = -ENOMEM;
 	page = grab_cache_page(mapping, index);
 	if (!page)
@@ -917,7 +917,6 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
 
 	unlock_page(page);
 	page_cache_release(page);
-	up_read(&BTRFS_I(inode)->root->snap_sem);
 out:
 	return ret;
 }
@@ -1146,6 +1145,19 @@ static int btrfs_find_actor(struct inode *inode, void *opaque)
 		args->root == BTRFS_I(inode)->root);
 }
 
+struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
+			    u64 root_objectid)
+{
+	struct btrfs_iget_args args;
+	args.ino = objectid;
+	args.root = btrfs_lookup_fs_root(btrfs_sb(s)->fs_info, root_objectid);
+
+	if (!args.root)
+		return NULL;
+
+	return ilookup5(s, objectid, btrfs_find_actor, (void *)&args);
+}
+
 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
 				struct btrfs_root *root)
 {
@@ -1336,7 +1348,6 @@ read_dir_items:
 
 			d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
 			btrfs_dir_item_key_to_cpu(leaf, di, &location);
-
 			over = filldir(dirent, name_ptr, name_len,
 				       found_key.offset,
 				       location.objectid,
@@ -2054,7 +2065,6 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
 
 	ret = -EINVAL;
 
-	down_read(&BTRFS_I(inode)->root->snap_sem);
 	lock_page(page);
 	wait_on_page_writeback(page);
 	size = i_size_read(inode);
@@ -2075,7 +2085,6 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
 	ret = btrfs_cow_one_page(inode, page, end);
 
 out_unlock:
-	up_read(&BTRFS_I(inode)->root->snap_sem);
 	unlock_page(page);
 out:
 	return ret;
@@ -2118,7 +2127,7 @@ static int noinline create_subvol(struct btrfs_root *root, char *name,
 	struct btrfs_root_item root_item;
 	struct btrfs_inode_item *inode_item;
 	struct extent_buffer *leaf;
-	struct btrfs_root *new_root;
+	struct btrfs_root *new_root = root;
 	struct inode *inode;
 	struct inode *dir;
 	int ret;
@@ -2230,7 +2239,7 @@ static int noinline create_subvol(struct btrfs_root *root, char *name,
 		goto fail;
 fail:
 	nr = trans->blocks_used;
-	err = btrfs_commit_transaction(trans, root);
+	err = btrfs_commit_transaction(trans, new_root);
 	if (err && !ret)
 		ret = err;
 fail_commit:
@@ -2253,10 +2262,6 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 	if (!root->ref_cows)
 		return -EINVAL;
 
-	down_write(&root->snap_sem);
-	freeze_bdev(root->fs_info->sb->s_bdev);
-	thaw_bdev(root->fs_info->sb->s_bdev, root->fs_info->sb);
-
 	mutex_lock(&root->fs_info->fs_mutex);
 	ret = btrfs_check_free_space(root, 1, 0);
 	if (ret)
@@ -2264,6 +2269,9 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 
 	trans = btrfs_start_transaction(root, 1);
 	BUG_ON(!trans);
+	err = btrfs_commit_transaction(trans, root);
+
+	trans = btrfs_start_transaction(root, 1);
 
 	ret = btrfs_update_inode(trans, root, root->inode);
 	if (ret)
@@ -2272,9 +2280,7 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 	ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
 				       0, &objectid);
 	if (ret)
-		goto fail;
-
-	memcpy(&new_root_item, &root->root_item,
+		goto fail; memcpy(&new_root_item, &root->root_item,
 	       sizeof(new_root_item));
 
 	key.objectid = objectid;
@@ -2285,12 +2291,20 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 	btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
 	free_extent_buffer(tmp);
 
+	/* write the ordered inodes to force all delayed allocations to
+	 * be filled.  Once this is done, we can copy the root
+	 */
+	mutex_lock(&root->fs_info->trans_mutex);
+	btrfs_write_ordered_inodes(trans, root);
+	mutex_unlock(&root->fs_info->trans_mutex);
+
 	btrfs_copy_root(trans, root, root->node, &tmp, objectid);
 
 	btrfs_set_root_bytenr(&new_root_item, tmp->start);
 	btrfs_set_root_level(&new_root_item, btrfs_header_level(tmp));
 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
 				&new_root_item);
+printk("new root %Lu node %Lu\n", objectid, tmp->start);
 	free_extent_buffer(tmp);
 	if (ret)
 		goto fail;
@@ -2321,7 +2335,6 @@ fail:
 		ret = err;
 fail_unlock:
 	mutex_unlock(&root->fs_info->fs_mutex);
-	up_write(&root->snap_sem);
 	btrfs_btree_balance_dirty(root, nr);
 	return ret;
 }
@@ -2608,6 +2621,7 @@ struct inode *btrfs_alloc_inode(struct super_block *sb)
 	if (!ei)
 		return NULL;
 	ei->last_trans = 0;
+	ei->ordered_trans = 0;
 	return &ei->vfs_inode;
 }
 
diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c
new file mode 100644
index 000000000000..411aba84d305
--- /dev/null
+++ b/fs/btrfs/ordered-data.c
@@ -0,0 +1,221 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/gfp.h>
+#include <linux/slab.h>
+#include "ctree.h"
+#include "transaction.h"
+#include "btrfs_inode.h"
+
+struct tree_entry {
+	u64 root_objectid;
+	u64 objectid;
+	struct rb_node rb_node;
+};
+
+/*
+ * returns > 0 if entry passed (root, objectid) is > entry,
+ * < 0 if (root, objectid) < entry and zero if they are equal
+ */
+static int comp_entry(struct tree_entry *entry, u64 root_objectid,
+		      u64 objectid)
+{
+	if (root_objectid < entry->root_objectid)
+		return -1;
+	if (root_objectid > entry->root_objectid)
+		return 1;
+	if (objectid < entry->objectid)
+		return -1;
+	if (objectid > entry->objectid)
+		return 1;
+	return 0;
+}
+
+static struct rb_node *tree_insert(struct rb_root *root, u64 root_objectid,
+				   u64 objectid, struct rb_node *node)
+{
+	struct rb_node ** p = &root->rb_node;
+	struct rb_node * parent = NULL;
+	struct tree_entry *entry;
+	int comp;
+
+	while(*p) {
+		parent = *p;
+		entry = rb_entry(parent, struct tree_entry, rb_node);
+
+		comp = comp_entry(entry, root_objectid, objectid);
+		if (comp < 0)
+			p = &(*p)->rb_left;
+		else if (comp > 0)
+			p = &(*p)->rb_right;
+		else
+			return parent;
+	}
+
+	rb_link_node(node, parent, p);
+	rb_insert_color(node, root);
+	return NULL;
+}
+
+static struct rb_node *__tree_search(struct rb_root *root, u64 root_objectid,
+				     u64 objectid, struct rb_node **prev_ret)
+{
+	struct rb_node * n = root->rb_node;
+	struct rb_node *prev = NULL;
+	struct tree_entry *entry;
+	struct tree_entry *prev_entry = NULL;
+	int comp;
+
+	while(n) {
+		entry = rb_entry(n, struct tree_entry, rb_node);
+		prev = n;
+		prev_entry = entry;
+		comp = comp_entry(entry, root_objectid, objectid);
+
+		if (comp < 0)
+			n = n->rb_left;
+		else if (comp > 0)
+			n = n->rb_right;
+		else
+			return n;
+	}
+	if (!prev_ret)
+		return NULL;
+
+	while(prev && comp_entry(prev_entry, root_objectid, objectid) >= 0) {
+		prev = rb_next(prev);
+		prev_entry = rb_entry(prev, struct tree_entry, rb_node);
+	}
+	*prev_ret = prev;
+	return NULL;
+}
+
+static inline struct rb_node *tree_search(struct rb_root *root,
+					  u64 root_objectid, u64 objectid)
+{
+	struct rb_node *prev;
+	struct rb_node *ret;
+	ret = __tree_search(root, root_objectid, objectid, &prev);
+	if (!ret)
+		return prev;
+	return ret;
+}
+
+int btrfs_add_ordered_inode(struct inode *inode)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	u64 root_objectid = root->root_key.objectid;
+	u64 transid = root->fs_info->running_transaction->transid;
+	struct tree_entry *entry;
+	struct rb_node *node;
+	struct btrfs_ordered_inode_tree *tree;
+
+	if (transid <= BTRFS_I(inode)->ordered_trans)
+		return 0;
+
+	tree = &root->fs_info->running_transaction->ordered_inode_tree;
+
+	read_lock(&tree->lock);
+	node = __tree_search(&tree->tree, root_objectid, inode->i_ino, NULL);
+	read_unlock(&tree->lock);
+	if (node) {
+		return 0;
+	}
+
+	entry = kmalloc(sizeof(*entry), GFP_NOFS);
+	if (!entry)
+		return -ENOMEM;
+
+	write_lock(&tree->lock);
+	entry->objectid = inode->i_ino;
+	entry->root_objectid = root_objectid;
+
+	node = tree_insert(&tree->tree, root_objectid,
+			   inode->i_ino, &entry->rb_node);
+
+	BTRFS_I(inode)->ordered_trans = transid;
+
+	write_unlock(&tree->lock);
+	if (node)
+		kfree(entry);
+	return 0;
+}
+
+int btrfs_find_first_ordered_inode(struct btrfs_ordered_inode_tree *tree,
+				       u64 *root_objectid, u64 *objectid)
+{
+	struct tree_entry *entry;
+	struct rb_node *node;
+
+	write_lock(&tree->lock);
+	node = tree_search(&tree->tree, *root_objectid, *objectid);
+	if (!node) {
+		write_unlock(&tree->lock);
+		return 0;
+	}
+	entry = rb_entry(node, struct tree_entry, rb_node);
+
+	while(comp_entry(entry, *root_objectid, *objectid) >= 0) {
+		node = rb_next(node);
+		if (!node)
+			break;
+		entry = rb_entry(node, struct tree_entry, rb_node);
+	}
+	if (!node) {
+		write_unlock(&tree->lock);
+		return 0;
+	}
+
+	*root_objectid = entry->root_objectid;
+	*objectid = entry->objectid;
+	write_unlock(&tree->lock);
+	return 1;
+}
+
+int btrfs_find_del_first_ordered_inode(struct btrfs_ordered_inode_tree *tree,
+				       u64 *root_objectid, u64 *objectid)
+{
+	struct tree_entry *entry;
+	struct rb_node *node;
+
+	write_lock(&tree->lock);
+	node = tree_search(&tree->tree, *root_objectid, *objectid);
+	if (!node) {
+		write_unlock(&tree->lock);
+		return 0;
+	}
+
+	entry = rb_entry(node, struct tree_entry, rb_node);
+	while(comp_entry(entry, *root_objectid, *objectid) >= 0) {
+		node = rb_next(node);
+		if (!node)
+			break;
+		entry = rb_entry(node, struct tree_entry, rb_node);
+	}
+	if (!node) {
+		write_unlock(&tree->lock);
+		return 0;
+	}
+
+	*root_objectid = entry->root_objectid;
+	*objectid = entry->objectid;
+	rb_erase(node, &tree->tree);
+	write_unlock(&tree->lock);
+	kfree(entry);
+	return 1;
+}
diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h
new file mode 100644
index 000000000000..aaf9eb142719
--- /dev/null
+++ b/fs/btrfs/ordered-data.h
@@ -0,0 +1,39 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef __BTRFS_ORDERED_DATA__
+#define __BTRFS_ORDERED_DATA__
+
+struct btrfs_ordered_inode_tree {
+	rwlock_t lock;
+	struct rb_root tree;
+};
+
+static inline void
+btrfs_ordered_inode_tree_init(struct btrfs_ordered_inode_tree *t)
+{
+	rwlock_init(&t->lock);
+	t->tree.rb_node = NULL;
+}
+
+int btrfs_add_ordered_inode(struct inode *inode);
+int btrfs_find_del_first_ordered_inode(struct btrfs_ordered_inode_tree *tree,
+				       u64 *root_objectid, u64 *objectid);
+int btrfs_find_first_ordered_inode(struct btrfs_ordered_inode_tree *tree,
+				       u64 *root_objectid, u64 *objectid);
+#endif
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 02721eea9a7a..3ed5868e7c0f 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -67,6 +67,7 @@ static int join_transaction(struct btrfs_root *root)
 		cur_trans->commit_done = 0;
 		cur_trans->start_time = get_seconds();
 		list_add_tail(&cur_trans->list, &root->fs_info->trans_list);
+		btrfs_ordered_inode_tree_init(&cur_trans->ordered_inode_tree);
 		extent_map_tree_init(&cur_trans->dirty_pages,
 				     root->fs_info->btree_inode->i_mapping,
 				     GFP_NOFS);
@@ -473,6 +474,60 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 	return ret;
 }
 
+int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans,
+				struct btrfs_root *root)
+{
+	struct btrfs_transaction *cur_trans = trans->transaction;
+	struct inode *inode;
+	u64 root_objectid = 0;
+	u64 objectid = 0;
+	u64 transid = trans->transid;
+	int ret;
+
+printk("write ordered trans %Lu\n", transid);
+	while(1) {
+		ret = btrfs_find_first_ordered_inode(
+				&cur_trans->ordered_inode_tree,
+				&root_objectid, &objectid);
+		if (!ret)
+			break;
+
+		mutex_unlock(&root->fs_info->trans_mutex);
+		mutex_unlock(&root->fs_info->fs_mutex);
+		inode = btrfs_ilookup(root->fs_info->sb, objectid,
+				      root_objectid);
+		if (inode) {
+			if (S_ISREG(inode->i_mode))
+				filemap_fdatawrite(inode->i_mapping);
+			iput(inode);
+		}
+		mutex_lock(&root->fs_info->fs_mutex);
+		mutex_lock(&root->fs_info->trans_mutex);
+	}
+	while(1) {
+		root_objectid = 0;
+		objectid = 0;
+		ret = btrfs_find_del_first_ordered_inode(
+				&cur_trans->ordered_inode_tree,
+				&root_objectid, &objectid);
+		if (!ret)
+			break;
+		mutex_unlock(&root->fs_info->trans_mutex);
+		mutex_unlock(&root->fs_info->fs_mutex);
+		inode = btrfs_ilookup(root->fs_info->sb, objectid,
+				      root_objectid);
+		if (inode) {
+			if (S_ISREG(inode->i_mode))
+				filemap_write_and_wait(inode->i_mapping);
+			iput(inode);
+		}
+		mutex_lock(&root->fs_info->fs_mutex);
+		mutex_lock(&root->fs_info->trans_mutex);
+	}
+printk("done write ordered trans %Lu\n", transid);
+	return 0;
+}
+
 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root)
 {
@@ -550,10 +605,13 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 		mutex_lock(&root->fs_info->fs_mutex);
 		mutex_lock(&root->fs_info->trans_mutex);
 		finish_wait(&cur_trans->writer_wait, &wait);
+		ret = btrfs_write_ordered_inodes(trans, root);
+
 	} while (cur_trans->num_writers > 1 ||
 		 (cur_trans->num_joined != joined));
 
 	WARN_ON(cur_trans != trans->transaction);
+
 	ret = add_dirty_roots(trans, &root->fs_info->fs_roots_radix,
 			      &dirty_fs_roots);
 	BUG_ON(ret);
diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h
index eef840bca91e..c157ddbe9d1e 100644
--- a/fs/btrfs/transaction.h
+++ b/fs/btrfs/transaction.h
@@ -16,9 +16,10 @@
  * Boston, MA 021110-1307, USA.
  */
 
-#ifndef __TRANSACTION__
-#define __TRANSACTION__
+#ifndef __BTRFS_TRANSACTION__
+#define __BTRFS_TRANSACTION__
 #include "btrfs_inode.h"
+#include "ordered-data.h"
 
 struct btrfs_transaction {
 	u64 transid;
@@ -30,6 +31,7 @@ struct btrfs_transaction {
 	struct list_head list;
 	struct extent_map_tree dirty_pages;
 	unsigned long start_time;
+	struct btrfs_ordered_inode_tree ordered_inode_tree;
 	wait_queue_head_t writer_wait;
 	wait_queue_head_t commit_wait;
 };
@@ -90,4 +92,6 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly);
 int btrfs_clean_old_snapshots(struct btrfs_root *root);
 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root);
+int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans,
+				struct btrfs_root *root);
 #endif
-- 
cgit v1.2.3


From e2008b61401ecb467a8ce1788fcd2116ae1cfbc1 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Tue, 8 Jan 2008 15:46:30 -0500
Subject: Btrfs: Add some simple throttling to wait for data=ordered and
 snapshot deletion

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.h       |  1 +
 fs/btrfs/disk-io.c     |  7 +++++++
 fs/btrfs/disk-io.h     |  1 +
 fs/btrfs/file.c        |  1 +
 fs/btrfs/inode.c       | 12 ++++++++++++
 fs/btrfs/transaction.c |  4 ++++
 6 files changed, 26 insertions(+)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index b55dba58dfaa..f0fb1978553c 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -345,6 +345,7 @@ struct btrfs_fs_info {
 	struct completion kobj_unregister;
 	int do_barriers;
 	int closing;
+	unsigned long throttles;
 
 	u64 total_pinned;
 	spinlock_t delalloc_lock;
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 34cf1f1f47be..e0940a39ff07 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -631,6 +631,7 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	fs_info->tree_root = tree_root;
 	fs_info->extent_root = extent_root;
 	fs_info->sb = sb;
+	fs_info->throttles = 0;
 	fs_info->mount_opt = 0;
 	fs_info->max_extent = (u64)-1;
 	fs_info->delalloc_bytes = 0;
@@ -889,6 +890,12 @@ void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
 	set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf);
 }
 
+void btrfs_throttle(struct btrfs_root *root)
+{
+	if (root->fs_info->throttles)
+		congestion_wait(WRITE, HZ/10);
+}
+
 void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
 {
 	balance_dirty_pages_ratelimited_nr(
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index dae9fba8efcd..828f3a2081b6 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -64,4 +64,5 @@ int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf);
 int btrfs_read_buffer(struct extent_buffer *buf);
 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len);
 void btrfs_csum_final(u32 crc, char *result);
+void btrfs_throttle(struct btrfs_root *root);
 #endif
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 0a5f4defe59b..897242e87fa7 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -800,6 +800,7 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 		balance_dirty_pages_ratelimited_nr(inode->i_mapping, num_pages);
 		if (num_pages < (root->leafsize >> PAGE_CACHE_SHIFT) + 1)
 			btrfs_btree_balance_dirty(root, 1);
+		btrfs_throttle(root);
 		cond_resched();
 	}
 out:
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 10cece11dbd8..2cb2dd32407e 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -610,6 +610,7 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
 fail:
 	mutex_unlock(&root->fs_info->fs_mutex);
 	btrfs_btree_balance_dirty(root, nr);
+	btrfs_throttle(root);
 	return ret;
 }
 
@@ -644,6 +645,7 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
 fail:
 	mutex_unlock(&root->fs_info->fs_mutex);
 	btrfs_btree_balance_dirty(root, nr);
+	btrfs_throttle(root);
 
 	if (ret && !err)
 		err = ret;
@@ -1010,6 +1012,7 @@ void btrfs_delete_inode(struct inode *inode)
 	btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
 	btrfs_btree_balance_dirty(root, nr);
+	btrfs_throttle(root);
 	return;
 
 no_delete_lock:
@@ -1017,6 +1020,7 @@ no_delete_lock:
 	btrfs_end_transaction(trans, root);
 	mutex_unlock(&root->fs_info->fs_mutex);
 	btrfs_btree_balance_dirty(root, nr);
+	btrfs_throttle(root);
 no_delete:
 	clear_inode(inode);
 }
@@ -1574,6 +1578,7 @@ fail:
 		iput(inode);
 	}
 	btrfs_btree_balance_dirty(root, nr);
+	btrfs_throttle(root);
 	return err;
 }
 
@@ -1633,6 +1638,7 @@ fail:
 		iput(inode);
 	}
 	btrfs_btree_balance_dirty(root, nr);
+	btrfs_throttle(root);
 	return err;
 }
 
@@ -1684,6 +1690,7 @@ fail:
 		iput(inode);
 	}
 	btrfs_btree_balance_dirty(root, nr);
+	btrfs_throttle(root);
 	return err;
 }
 
@@ -1752,6 +1759,7 @@ out_unlock:
 	if (drop_on_err)
 		iput(inode);
 	btrfs_btree_balance_dirty(root, nr);
+	btrfs_throttle(root);
 	return err;
 }
 
@@ -2117,6 +2125,7 @@ static void btrfs_truncate(struct inode *inode)
 	BUG_ON(ret);
 	mutex_unlock(&root->fs_info->fs_mutex);
 	btrfs_btree_balance_dirty(root, nr);
+	btrfs_throttle(root);
 }
 
 static int noinline create_subvol(struct btrfs_root *root, char *name,
@@ -2245,6 +2254,7 @@ fail:
 fail_commit:
 	mutex_unlock(&root->fs_info->fs_mutex);
 	btrfs_btree_balance_dirty(root, nr);
+	btrfs_throttle(root);
 	return ret;
 }
 
@@ -2287,6 +2297,7 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 fail_unlock:
 	mutex_unlock(&root->fs_info->fs_mutex);
 	btrfs_btree_balance_dirty(root, nr);
+	btrfs_throttle(root);
 	return ret;
 }
 
@@ -2827,6 +2838,7 @@ out_fail:
 		iput(inode);
 	}
 	btrfs_btree_balance_dirty(root, nr);
+	btrfs_throttle(root);
 	return err;
 }
 
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index dc9865323e38..614903f5c884 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -424,6 +424,7 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 
 		num_bytes = btrfs_root_used(&dirty->root->root_item);
 		root = dirty->latest_root;
+		root->fs_info->throttles++;
 
 		while(1) {
 			trans = btrfs_start_transaction(tree_root, 1);
@@ -447,6 +448,7 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
 			mutex_lock(&tree_root->fs_info->fs_mutex);
 		}
 		BUG_ON(ret);
+		root->fs_info->throttles--;
 
 		num_bytes -= btrfs_root_used(&dirty->root->root_item);
 		bytes_used = btrfs_root_used(&root->root_item);
@@ -484,6 +486,7 @@ int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans,
 	u64 objectid = 0;
 	int ret;
 
+	root->fs_info->throttles++;
 	while(1) {
 		ret = btrfs_find_first_ordered_inode(
 				&cur_trans->ordered_inode_tree,
@@ -523,6 +526,7 @@ int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans,
 		mutex_lock(&root->fs_info->fs_mutex);
 		mutex_lock(&root->fs_info->trans_mutex);
 	}
+	root->fs_info->throttles--;
 	return 0;
 }
 
-- 
cgit v1.2.3


From 0b86a832a1f38abec695864ec2eaedc9d2383f1b Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Mon, 24 Mar 2008 15:01:56 -0400
Subject: Btrfs: Add support for multiple devices per filesystem

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/Makefile      |   2 +-
 fs/btrfs/ctree.c       |  38 ++-
 fs/btrfs/ctree.h       | 313 ++++++++++++++++--
 fs/btrfs/disk-io.c     |  97 +++++-
 fs/btrfs/disk-io.h     |   2 +
 fs/btrfs/extent-tree.c | 429 +++++++------------------
 fs/btrfs/extent_io.c   |   6 +-
 fs/btrfs/extent_io.h   |   2 +-
 fs/btrfs/inode.c       |  26 +-
 fs/btrfs/print-tree.c  |  48 +++
 fs/btrfs/transaction.c |  50 ++-
 fs/btrfs/volumes.c     | 852 +++++++++++++++++++++++++++++++++++++++++++++++++
 fs/btrfs/volumes.h     |  78 +++++
 13 files changed, 1563 insertions(+), 380 deletions(-)
 create mode 100644 fs/btrfs/volumes.c
 create mode 100644 fs/btrfs/volumes.h

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 9c61609f3714..5ac6f2946c21 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -6,7 +6,7 @@ btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
 	   hash.o file-item.o inode-item.o inode-map.o disk-io.o \
 	   transaction.o bit-radix.o inode.o file.o tree-defrag.o \
 	   extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \
-	   extent_io.o
+	   extent_io.o volumes.o
 
 ifeq ($(CONFIG_FS_POSIX_ACL),y)
 btrfs-y += acl.o
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index ef8809c1e69c..70b6ddfe15a1 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -70,6 +70,14 @@ void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p)
 	memset(p, 0, sizeof(*p));
 }
 
+static void add_root_to_dirty_list(struct btrfs_root *root)
+{
+	if (root->track_dirty && list_empty(&root->dirty_list)) {
+		list_add(&root->dirty_list,
+			 &root->fs_info->dirty_cowonly_roots);
+	}
+}
+
 int btrfs_copy_root(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root,
 		      struct extent_buffer *buf,
@@ -196,6 +204,7 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 					  root_gen, 0, 0, 1);
 		}
 		free_extent_buffer(buf);
+		add_root_to_dirty_list(root);
 	} else {
 		root_gen = btrfs_header_generation(parent);
 		btrfs_set_node_blockptr(parent, parent_slot,
@@ -241,7 +250,7 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		return 0;
 	}
 
-	search_start = buf->start & ~((u64)BTRFS_BLOCK_GROUP_SIZE - 1);
+	search_start = buf->start & ~((u64)(1024 * 1024 * 1024) - 1);
 	ret = __btrfs_cow_block(trans, root, buf, parent,
 				 parent_slot, cow_ret, search_start, 0);
 	return ret;
@@ -724,6 +733,7 @@ static int balance_level(struct btrfs_trans_handle *trans,
 		BUG_ON(ret);
 
 		root->node = child;
+		add_root_to_dirty_list(root);
 		path->nodes[level] = NULL;
 		clean_tree_block(trans, root, mid);
 		wait_on_tree_block_writeback(root, mid);
@@ -1369,6 +1379,7 @@ static int noinline insert_new_root(struct btrfs_trans_handle *trans,
 	/* the super has an extra ref to root->node */
 	free_extent_buffer(root->node);
 	root->node = c;
+	add_root_to_dirty_list(root);
 	extent_buffer_get(c);
 	path->nodes[level] = c;
 	path->slots[level] = 0;
@@ -2777,3 +2788,28 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 	}
 	return 0;
 }
+
+int btrfs_previous_item(struct btrfs_root *root,
+			struct btrfs_path *path, u64 min_objectid,
+			int type)
+{
+	struct btrfs_key found_key;
+	struct extent_buffer *leaf;
+	int ret;
+
+	while(1) {
+		if (path->slots[0] == 0) {
+			ret = btrfs_prev_leaf(root, path);
+			if (ret != 0)
+				return ret;
+		} else {
+			path->slots[0]--;
+		}
+		leaf = path->nodes[0];
+		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+		if (found_key.type == type)
+			return 0;
+	}
+	return 1;
+}
+
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 92d892f92075..1453d995fef6 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -40,12 +40,44 @@ extern struct kmem_cache *btrfs_path_cachep;
 #define BTRFS_MAGIC "_B4RfS_M"
 
 #define BTRFS_MAX_LEVEL 8
+
+/* holds pointers to all of the tree roots */
 #define BTRFS_ROOT_TREE_OBJECTID 1ULL
+
+/* stores information about which extents are in use, and reference counts */
 #define BTRFS_EXTENT_TREE_OBJECTID 2ULL
+
+/* one per subvolume, storing files and directories */
 #define BTRFS_FS_TREE_OBJECTID 3ULL
+
+/* directory objectid inside the root tree */
 #define BTRFS_ROOT_TREE_DIR_OBJECTID 4ULL
+
+
+/*
+ * chunk tree stores translations from logical -> physical block numbering
+ * the super block points to the chunk tree
+ */
+#define BTRFS_CHUNK_TREE_OBJECTID 5ULL
+
+/*
+ * stores information about which areas of a given device are in use.
+ * one per device.  The tree of tree roots points to the device tree
+ */
+#define BTRFS_DEV_TREE_OBJECTID 6ULL
+
+/*
+ * All files have objectids higher than this.
+ */
 #define BTRFS_FIRST_FREE_OBJECTID 256ULL
 
+
+/*
+ * the device items go into the chunk tree.  The key is in the form
+ * [ 1 BTRFS_DEV_ITEM_KEY device_id ]
+ */
+#define BTRFS_DEV_ITEMS_OBJECTID 1ULL
+
 /*
  * we can actually store much bigger names, but lets not confuse the rest
  * of linux
@@ -95,6 +127,81 @@ struct btrfs_key {
 	u64 offset;
 } __attribute__ ((__packed__));
 
+struct btrfs_mapping_tree {
+	struct extent_map_tree map_tree;
+};
+
+#define BTRFS_DEV_UUID_SIZE 16
+struct btrfs_dev_item {
+	/* the internal btrfs device id */
+	__le64 devid;
+
+	/* size of the device */
+	__le64 total_bytes;
+
+	/* bytes used */
+	__le64 bytes_used;
+
+	/* optimal io alignment for this device */
+	__le32 io_align;
+
+	/* optimal io width for this device */
+	__le32 io_width;
+
+	/* minimal io size for this device */
+	__le32 sector_size;
+
+	/* the kernel device number */
+	__le64 rdev;
+
+	/* type and info about this device */
+	__le64 type;
+
+	/* partition number, 0 for whole dev */
+	__le32 partition;
+
+	/* length of the name data at the end of the item */
+	__le16 name_len;
+
+	/* physical drive uuid (or lvm uuid) */
+	u8 uuid[BTRFS_DEV_UUID_SIZE];
+	/* name goes here */
+} __attribute__ ((__packed__));
+
+struct btrfs_stripe {
+	__le64 devid;
+	__le64 offset;
+} __attribute__ ((__packed__));
+
+struct btrfs_chunk {
+	__le64 owner;
+	__le64 stripe_len;
+	__le64 type;
+
+	/* optimal io alignment for this chunk */
+	__le32 io_align;
+
+	/* optimal io width for this chunk */
+	__le32 io_width;
+
+	/* minimal io size for this chunk */
+	__le32 sector_size;
+
+	/* 2^16 stripes is quite a lot, a second limit is the size of a single
+	 * item in the btree
+	 */
+	__le16 num_stripes;
+	struct btrfs_stripe stripe;
+	/* additional stripes go here */
+} __attribute__ ((__packed__));
+
+static inline unsigned long btrfs_chunk_item_size(int num_stripes)
+{
+	BUG_ON(num_stripes == 0);
+	return sizeof(struct btrfs_chunk) +
+		sizeof(struct btrfs_stripe) * (num_stripes - 1);
+}
+
 #define BTRFS_FSID_SIZE 16
 /*
  * every tree block (leaf or node) starts with this header.
@@ -119,6 +226,13 @@ struct btrfs_header {
 					sizeof(struct btrfs_item) - \
 					sizeof(struct btrfs_file_extent_item))
 
+
+/*
+ * this is a very generous portion of the super block, giving us
+ * room to translate 14 chunks with 3 stripes each.
+ */
+#define BTRFS_SYSTEM_CHUNK_ARRAY_SIZE 2048
+
 /*
  * the super block basically lists the main trees of the FS
  * it currently lacks any block count etc etc
@@ -131,6 +245,7 @@ struct btrfs_super_block {
 	__le64 magic;
 	__le64 generation;
 	__le64 root;
+	__le64 chunk_root;
 	__le64 total_bytes;
 	__le64 bytes_used;
 	__le64 root_dir_objectid;
@@ -138,7 +253,10 @@ struct btrfs_super_block {
 	__le32 nodesize;
 	__le32 leafsize;
 	__le32 stripesize;
+	__le32 sys_chunk_array_size;
 	u8 root_level;
+	u8 chunk_root_level;
+	u8 sys_chunk_array[BTRFS_SYSTEM_CHUNK_ARRAY_SIZE];
 } __attribute__ ((__packed__));
 
 /*
@@ -208,12 +326,22 @@ struct btrfs_extent_ref {
 	__le64 offset;
 } __attribute__ ((__packed__));
 
+/* dev extents record free space on individual devices.  The owner
+ * field points back to the chunk allocation mapping tree that allocated
+ * the extent
+ */
+struct btrfs_dev_extent {
+	__le64 owner;
+	__le64 length;
+} __attribute__ ((__packed__));
+
+
 struct btrfs_inode_ref {
 	__le16 name_len;
 	/* name goes here */
 } __attribute__ ((__packed__));
 
-struct btrfs_inode_timespec {
+struct btrfs_timespec {
 	__le64 sec;
 	__le32 nsec;
 } __attribute__ ((__packed__));
@@ -231,13 +359,13 @@ struct btrfs_inode_item {
 	__le32 uid;
 	__le32 gid;
 	__le32 mode;
-	__le32 rdev;
+	__le64 rdev;
 	__le16 flags;
 	__le16 compat_flags;
-	struct btrfs_inode_timespec atime;
-	struct btrfs_inode_timespec ctime;
-	struct btrfs_inode_timespec mtime;
-	struct btrfs_inode_timespec otime;
+	struct btrfs_timespec atime;
+	struct btrfs_timespec ctime;
+	struct btrfs_timespec mtime;
+	struct btrfs_timespec otime;
 } __attribute__ ((__packed__));
 
 struct btrfs_dir_item {
@@ -290,29 +418,34 @@ struct btrfs_csum_item {
 	u8 csum;
 } __attribute__ ((__packed__));
 
-/* tag for the radix tree of block groups in ram */
-#define BTRFS_BLOCK_GROUP_SIZE (256 * 1024 * 1024)
-
+/* different types of block groups (and chunks) */
+#define BTRFS_BLOCK_GROUP_DATA     (1 << 0)
+#define BTRFS_BLOCK_GROUP_SYSTEM   (1 << 1)
+#define BTRFS_BLOCK_GROUP_METADATA (1 << 2)
 
-#define BTRFS_BLOCK_GROUP_DATA 1
-#define BTRFS_BLOCK_GROUP_MIXED 2
 
 struct btrfs_block_group_item {
 	__le64 used;
-	u8 flags;
+	__le64 chunk_tree;
+	__le64 chunk_objectid;
+	__le64 flags;
 } __attribute__ ((__packed__));
 
 struct btrfs_block_group_cache {
 	struct btrfs_key key;
 	struct btrfs_block_group_item item;
-	int data;
-	int cached;
 	u64 pinned;
+	u64 flags;
+	int cached;
 };
+
+struct btrfs_device;
 struct btrfs_fs_info {
 	u8 fsid[BTRFS_FSID_SIZE];
 	struct btrfs_root *extent_root;
 	struct btrfs_root *tree_root;
+	struct btrfs_root *chunk_root;
+	struct btrfs_root *dev_root;
 	struct radix_tree_root fs_roots_radix;
 
 	struct extent_io_tree free_space_cache;
@@ -321,6 +454,9 @@ struct btrfs_fs_info {
 	struct extent_io_tree pending_del;
 	struct extent_io_tree extent_ins;
 
+	/* logical->physical extent mapping */
+	struct btrfs_mapping_tree mapping_tree;
+
 	u64 generation;
 	u64 last_trans_committed;
 	unsigned long mount_opt;
@@ -330,6 +466,7 @@ struct btrfs_fs_info {
 	struct btrfs_transaction *running_transaction;
 	struct btrfs_super_block super_copy;
 	struct extent_buffer *sb_buffer;
+	struct block_device *__bdev;
 	struct super_block *sb;
 	struct inode *btree_inode;
 	spinlock_t hash_lock;
@@ -350,12 +487,17 @@ struct btrfs_fs_info {
 	unsigned long throttles;
 
 	u64 total_pinned;
+	struct list_head dirty_cowonly_roots;
+
+	struct list_head devices;
+	struct list_head *last_device;
 	spinlock_t delalloc_lock;
 	spinlock_t new_trans_lock;
 	u64 delalloc_bytes;
 	u64 last_alloc;
 	u64 last_data_alloc;
 };
+
 /*
  * in ram representation of the tree.  extent_root is used for all allocations
  * and for the extent tree extent_root root.
@@ -387,14 +529,19 @@ struct btrfs_root {
 	u64 highest_inode;
 	u64 last_inode_alloc;
 	int ref_cows;
+	int track_dirty;
 	struct btrfs_key defrag_progress;
 	int defrag_running;
 	int defrag_level;
 	char *name;
 	int in_sysfs;
+
+	/* the dirty list is only used by non-reference counted roots */
+	struct list_head dirty_list;
 };
 
 /*
+
  * inode items have the data typically returned from stat and store other
  * info about object characteristics.  There is one for every file and dir in
  * the FS
@@ -439,6 +586,10 @@ struct btrfs_root {
  */
 #define BTRFS_BLOCK_GROUP_ITEM_KEY 50
 
+#define BTRFS_DEV_EXTENT_KEY	75
+#define BTRFS_DEV_ITEM_KEY	76
+#define BTRFS_CHUNK_ITEM_KEY	77
+
 /*
  * string items are for debugging.  They just store a short string of
  * data in the FS
@@ -518,13 +669,104 @@ static inline void btrfs_set_##name(type *s, u##bits val)		\
 	s->member = cpu_to_le##bits(val);				\
 }
 
+BTRFS_SETGET_FUNCS(device_type, struct btrfs_dev_item, type, 64);
+BTRFS_SETGET_FUNCS(device_total_bytes, struct btrfs_dev_item, total_bytes, 64);
+BTRFS_SETGET_FUNCS(device_bytes_used, struct btrfs_dev_item, bytes_used, 64);
+BTRFS_SETGET_FUNCS(device_io_align, struct btrfs_dev_item, io_align, 32);
+BTRFS_SETGET_FUNCS(device_io_width, struct btrfs_dev_item, io_width, 32);
+BTRFS_SETGET_FUNCS(device_sector_size, struct btrfs_dev_item, sector_size, 32);
+BTRFS_SETGET_FUNCS(device_id, struct btrfs_dev_item, devid, 64);
+BTRFS_SETGET_FUNCS(device_rdev, struct btrfs_dev_item, rdev, 64);
+BTRFS_SETGET_FUNCS(device_partition, struct btrfs_dev_item, partition, 32);
+BTRFS_SETGET_FUNCS(device_name_len, struct btrfs_dev_item, name_len, 16);
+
+static inline char *btrfs_device_uuid(struct btrfs_dev_item *d)
+{
+	return (char *)d + offsetof(struct btrfs_dev_item, uuid);
+}
+
+static inline char *btrfs_device_name(struct btrfs_dev_item *d)
+{
+	return (char *)(d + 1);
+}
+
+BTRFS_SETGET_FUNCS(chunk_owner, struct btrfs_chunk, owner, 64);
+BTRFS_SETGET_FUNCS(chunk_stripe_len, struct btrfs_chunk, stripe_len, 64);
+BTRFS_SETGET_FUNCS(chunk_io_align, struct btrfs_chunk, io_align, 32);
+BTRFS_SETGET_FUNCS(chunk_io_width, struct btrfs_chunk, io_width, 32);
+BTRFS_SETGET_FUNCS(chunk_sector_size, struct btrfs_chunk, sector_size, 32);
+BTRFS_SETGET_FUNCS(chunk_type, struct btrfs_chunk, type, 64);
+BTRFS_SETGET_FUNCS(chunk_num_stripes, struct btrfs_chunk, num_stripes, 16);
+BTRFS_SETGET_FUNCS(stripe_devid, struct btrfs_stripe, devid, 64);
+BTRFS_SETGET_FUNCS(stripe_offset, struct btrfs_stripe, offset, 64);
+
+BTRFS_SETGET_STACK_FUNCS(stack_chunk_owner, struct btrfs_chunk, owner, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_chunk_stripe_len, struct btrfs_chunk,
+			 stripe_len, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_chunk_io_align, struct btrfs_chunk,
+			 io_align, 32);
+BTRFS_SETGET_STACK_FUNCS(stack_chunk_io_width, struct btrfs_chunk,
+			 io_width, 32);
+BTRFS_SETGET_STACK_FUNCS(stack_chunk_sector_size, struct btrfs_chunk,
+			 sector_size, 32);
+BTRFS_SETGET_STACK_FUNCS(stack_chunk_type, struct btrfs_chunk, type, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_chunk_num_stripes, struct btrfs_chunk,
+			 num_stripes, 16);
+BTRFS_SETGET_STACK_FUNCS(stack_stripe_devid, struct btrfs_stripe, devid, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_stripe_offset, struct btrfs_stripe, offset, 64);
+
+static inline struct btrfs_stripe *btrfs_stripe_nr(struct btrfs_chunk *c,
+						   int nr)
+{
+	unsigned long offset = (unsigned long)c;
+	offset += offsetof(struct btrfs_chunk, stripe);
+	offset += nr * sizeof(struct btrfs_stripe);
+	return (struct btrfs_stripe *)offset;
+}
+
+static inline u64 btrfs_stripe_offset_nr(struct extent_buffer *eb,
+					 struct btrfs_chunk *c, int nr)
+{
+	return btrfs_stripe_offset(eb, btrfs_stripe_nr(c, nr));
+}
+
+static inline void btrfs_set_stripe_offset_nr(struct extent_buffer *eb,
+					     struct btrfs_chunk *c, int nr,
+					     u64 val)
+{
+	btrfs_set_stripe_offset(eb, btrfs_stripe_nr(c, nr), val);
+}
+
+static inline u64 btrfs_stripe_devid_nr(struct extent_buffer *eb,
+					 struct btrfs_chunk *c, int nr)
+{
+	return btrfs_stripe_devid(eb, btrfs_stripe_nr(c, nr));
+}
+
+static inline void btrfs_set_stripe_devid_nr(struct extent_buffer *eb,
+					     struct btrfs_chunk *c, int nr,
+					     u64 val)
+{
+	btrfs_set_stripe_devid(eb, btrfs_stripe_nr(c, nr), val);
+}
+
 /* struct btrfs_block_group_item */
 BTRFS_SETGET_STACK_FUNCS(block_group_used, struct btrfs_block_group_item,
 			 used, 64);
 BTRFS_SETGET_FUNCS(disk_block_group_used, struct btrfs_block_group_item,
 			 used, 64);
-BTRFS_SETGET_FUNCS(disk_block_group_flags, struct btrfs_block_group_item,
-		   flags, 8);
+BTRFS_SETGET_STACK_FUNCS(block_group_chunk_tree, struct btrfs_block_group_item,
+			 chunk_tree, 64);
+BTRFS_SETGET_FUNCS(disk_block_group_chunk_tree, struct btrfs_block_group_item,
+			 chunk_tree, 64);
+BTRFS_SETGET_STACK_FUNCS(block_group_chunk_objectid,
+			struct btrfs_block_group_item, chunk_objectid, 64);
+BTRFS_SETGET_FUNCS(disk_block_group_chunk_objecitd,
+		   struct btrfs_block_group_item, chunk_objectid, 64);
+BTRFS_SETGET_FUNCS(disk_block_group_flags,
+		   struct btrfs_block_group_item, flags, 64);
+BTRFS_SETGET_STACK_FUNCS(block_group_flags,
+			struct btrfs_block_group_item, flags, 64);
 
 /* struct btrfs_inode_ref */
 BTRFS_SETGET_FUNCS(inode_ref_name_len, struct btrfs_inode_ref, name_len, 16);
@@ -538,49 +780,53 @@ BTRFS_SETGET_FUNCS(inode_nlink, struct btrfs_inode_item, nlink, 32);
 BTRFS_SETGET_FUNCS(inode_uid, struct btrfs_inode_item, uid, 32);
 BTRFS_SETGET_FUNCS(inode_gid, struct btrfs_inode_item, gid, 32);
 BTRFS_SETGET_FUNCS(inode_mode, struct btrfs_inode_item, mode, 32);
-BTRFS_SETGET_FUNCS(inode_rdev, struct btrfs_inode_item, rdev, 32);
+BTRFS_SETGET_FUNCS(inode_rdev, struct btrfs_inode_item, rdev, 64);
 BTRFS_SETGET_FUNCS(inode_flags, struct btrfs_inode_item, flags, 16);
 BTRFS_SETGET_FUNCS(inode_compat_flags, struct btrfs_inode_item,
 		   compat_flags, 16);
 
-static inline struct btrfs_inode_timespec *
+static inline struct btrfs_timespec *
 btrfs_inode_atime(struct btrfs_inode_item *inode_item)
 {
 	unsigned long ptr = (unsigned long)inode_item;
 	ptr += offsetof(struct btrfs_inode_item, atime);
-	return (struct btrfs_inode_timespec *)ptr;
+	return (struct btrfs_timespec *)ptr;
 }
 
-static inline struct btrfs_inode_timespec *
+static inline struct btrfs_timespec *
 btrfs_inode_mtime(struct btrfs_inode_item *inode_item)
 {
 	unsigned long ptr = (unsigned long)inode_item;
 	ptr += offsetof(struct btrfs_inode_item, mtime);
-	return (struct btrfs_inode_timespec *)ptr;
+	return (struct btrfs_timespec *)ptr;
 }
 
-static inline struct btrfs_inode_timespec *
+static inline struct btrfs_timespec *
 btrfs_inode_ctime(struct btrfs_inode_item *inode_item)
 {
 	unsigned long ptr = (unsigned long)inode_item;
 	ptr += offsetof(struct btrfs_inode_item, ctime);
-	return (struct btrfs_inode_timespec *)ptr;
+	return (struct btrfs_timespec *)ptr;
 }
 
-static inline struct btrfs_inode_timespec *
+static inline struct btrfs_timespec *
 btrfs_inode_otime(struct btrfs_inode_item *inode_item)
 {
 	unsigned long ptr = (unsigned long)inode_item;
 	ptr += offsetof(struct btrfs_inode_item, otime);
-	return (struct btrfs_inode_timespec *)ptr;
+	return (struct btrfs_timespec *)ptr;
 }
 
-BTRFS_SETGET_FUNCS(timespec_sec, struct btrfs_inode_timespec, sec, 64);
-BTRFS_SETGET_FUNCS(timespec_nsec, struct btrfs_inode_timespec, nsec, 32);
+BTRFS_SETGET_FUNCS(timespec_sec, struct btrfs_timespec, sec, 64);
+BTRFS_SETGET_FUNCS(timespec_nsec, struct btrfs_timespec, nsec, 32);
 
 /* struct btrfs_extent_item */
 BTRFS_SETGET_FUNCS(extent_refs, struct btrfs_extent_item, refs, 32);
 
+/* struct btrfs_dev_extent */
+BTRFS_SETGET_FUNCS(dev_extent_owner, struct btrfs_dev_extent, owner, 64);
+BTRFS_SETGET_FUNCS(dev_extent_length, struct btrfs_dev_extent, length, 64);
+
 /* struct btrfs_extent_ref */
 BTRFS_SETGET_FUNCS(ref_root, struct btrfs_extent_ref, root, 64);
 BTRFS_SETGET_FUNCS(ref_generation, struct btrfs_extent_ref, generation, 64);
@@ -846,8 +1092,14 @@ BTRFS_SETGET_STACK_FUNCS(super_bytenr, struct btrfs_super_block, bytenr, 64);
 BTRFS_SETGET_STACK_FUNCS(super_generation, struct btrfs_super_block,
 			 generation, 64);
 BTRFS_SETGET_STACK_FUNCS(super_root, struct btrfs_super_block, root, 64);
+BTRFS_SETGET_STACK_FUNCS(super_sys_array_size,
+			 struct btrfs_super_block, sys_chunk_array_size, 32);
 BTRFS_SETGET_STACK_FUNCS(super_root_level, struct btrfs_super_block,
 			 root_level, 8);
+BTRFS_SETGET_STACK_FUNCS(super_chunk_root, struct btrfs_super_block,
+			 chunk_root, 64);
+BTRFS_SETGET_STACK_FUNCS(super_chunk_root_level, struct btrfs_super_block,
+			 chunk_root_level, 64);
 BTRFS_SETGET_STACK_FUNCS(super_total_bytes, struct btrfs_super_block,
 			 total_bytes, 64);
 BTRFS_SETGET_STACK_FUNCS(super_bytes_used, struct btrfs_super_block,
@@ -1009,7 +1261,14 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
 				    struct btrfs_root *root);
 int btrfs_free_block_groups(struct btrfs_fs_info *info);
 int btrfs_read_block_groups(struct btrfs_root *root);
+int btrfs_make_block_group(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *root, u64 bytes_used,
+			   u64 type, u64 chunk_tree, u64 chunk_objectid,
+			   u64 size);
 /* ctree.c */
+int btrfs_previous_item(struct btrfs_root *root,
+			struct btrfs_path *path, u64 min_objectid,
+			int type);
 int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		    struct btrfs_root *root, struct extent_buffer *buf,
 		    struct extent_buffer *parent, int parent_slot,
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 88e21bdbc478..8e37fa120cc8 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -28,6 +28,7 @@
 #include "disk-io.h"
 #include "transaction.h"
 #include "btrfs_inode.h"
+#include "volumes.h"
 #include "print-tree.h"
 
 #if 0
@@ -234,6 +235,19 @@ static int btree_writepage_io_hook(struct page *page, u64 start, u64 end)
 	return 0;
 }
 
+static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	u64 offset;
+	offset = bio->bi_sector << 9;
+	if (offset == BTRFS_SUPER_INFO_OFFSET) {
+		bio->bi_bdev = root->fs_info->sb->s_bdev;
+		submit_bio(rw, bio);
+		return 0;
+	}
+	return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio);
+}
+
 static int btree_writepage(struct page *page, struct writeback_control *wbc)
 {
 	struct extent_io_tree *tree;
@@ -345,6 +359,23 @@ int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize)
 	return ret;
 }
 
+static int close_all_devices(struct btrfs_fs_info *fs_info)
+{
+	struct list_head *list;
+	struct list_head *next;
+	struct btrfs_device *device;
+
+	list = &fs_info->devices;
+	while(!list_empty(list)) {
+		next = list->next;
+		list_del(next);
+		device = list_entry(next, struct btrfs_device, dev_list);
+		kfree(device->name);
+		kfree(device);
+	}
+	return 0;
+}
+
 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
 				      u32 blocksize)
 {
@@ -420,6 +451,8 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 	root->leafsize = leafsize;
 	root->stripesize = stripesize;
 	root->ref_cows = 0;
+	root->track_dirty = 0;
+
 	root->fs_info = fs_info;
 	root->objectid = objectid;
 	root->last_trans = 0;
@@ -427,6 +460,8 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 	root->last_inode_alloc = 0;
 	root->name = NULL;
 	root->in_sysfs = 0;
+
+	INIT_LIST_HEAD(&root->dirty_list);
 	memset(&root->root_key, 0, sizeof(root->root_key));
 	memset(&root->root_item, 0, sizeof(root->root_item));
 	memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
@@ -634,6 +669,10 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 					       GFP_NOFS);
 	struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
 						GFP_NOFS);
+	struct btrfs_root *chunk_root = kmalloc(sizeof(struct btrfs_root),
+						GFP_NOFS);
+	struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root),
+					      GFP_NOFS);
 	int ret;
 	int err = -EIO;
 	struct btrfs_super_block *disk_super;
@@ -657,6 +696,12 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	fs_info->last_trans_committed = 0;
 	fs_info->tree_root = tree_root;
 	fs_info->extent_root = extent_root;
+	fs_info->chunk_root = chunk_root;
+	fs_info->dev_root = dev_root;
+	INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
+	INIT_LIST_HEAD(&fs_info->devices);
+	btrfs_mapping_init(&fs_info->mapping_tree);
+	fs_info->last_device = &fs_info->devices;
 	fs_info->sb = sb;
 	fs_info->throttles = 0;
 	fs_info->mount_opt = 0;
@@ -714,12 +759,12 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 		goto fail_iput;
 	}
 #endif
-	__setup_root(512, 512, 512, 512, tree_root,
+	__setup_root(4096, 4096, 4096, 4096, tree_root,
 		     fs_info, BTRFS_ROOT_TREE_OBJECTID);
 
 	fs_info->sb_buffer = read_tree_block(tree_root,
 					     BTRFS_SUPER_INFO_OFFSET,
-					     512);
+					     4096);
 
 	if (!fs_info->sb_buffer)
 		goto fail_iput;
@@ -730,6 +775,7 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	read_extent_buffer(fs_info->sb_buffer, fs_info->fsid,
 			   (unsigned long)btrfs_super_fsid(fs_info->sb_buffer),
 			   BTRFS_FSID_SIZE);
+
 	disk_super = &fs_info->super_copy;
 	if (!btrfs_super_root(disk_super))
 		goto fail_sb_buffer;
@@ -753,23 +799,47 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 		goto fail_sb_buffer;
 	}
 
+	mutex_lock(&fs_info->fs_mutex);
+	ret = btrfs_read_sys_array(tree_root);
+	BUG_ON(ret);
+
+	blocksize = btrfs_level_size(tree_root,
+				     btrfs_super_chunk_root_level(disk_super));
+
+	__setup_root(nodesize, leafsize, sectorsize, stripesize,
+		     chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
+
+	chunk_root->node = read_tree_block(chunk_root,
+					   btrfs_super_chunk_root(disk_super),
+					   blocksize);
+	BUG_ON(!chunk_root->node);
+
+	ret = btrfs_read_chunk_tree(chunk_root);
+	BUG_ON(ret);
+
 	blocksize = btrfs_level_size(tree_root,
 				     btrfs_super_root_level(disk_super));
 
+
 	tree_root->node = read_tree_block(tree_root,
 					  btrfs_super_root(disk_super),
 					  blocksize);
 	if (!tree_root->node)
 		goto fail_sb_buffer;
 
-	mutex_lock(&fs_info->fs_mutex);
 
 	ret = find_and_setup_root(tree_root, fs_info,
 				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
-	if (ret) {
-		mutex_unlock(&fs_info->fs_mutex);
+	if (ret)
 		goto fail_tree_root;
-	}
+	extent_root->track_dirty = 1;
+
+	ret = find_and_setup_root(tree_root, fs_info,
+				  BTRFS_DEV_TREE_OBJECTID, dev_root);
+	dev_root->track_dirty = 1;
+
+	if (ret)
+		goto fail_extent_root;
 
 	btrfs_read_block_groups(extent_root);
 
@@ -777,7 +847,10 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	mutex_unlock(&fs_info->fs_mutex);
 	return tree_root;
 
+fail_extent_root:
+	free_extent_buffer(extent_root->node);
 fail_tree_root:
+	mutex_unlock(&fs_info->fs_mutex);
 	free_extent_buffer(tree_root->node);
 fail_sb_buffer:
 	free_extent_buffer(fs_info->sb_buffer);
@@ -874,6 +947,12 @@ int close_ctree(struct btrfs_root *root)
 	if (fs_info->tree_root->node)
 		free_extent_buffer(fs_info->tree_root->node);
 
+	if (root->fs_info->chunk_root->node);
+		free_extent_buffer(root->fs_info->chunk_root->node);
+
+	if (root->fs_info->dev_root->node);
+		free_extent_buffer(root->fs_info->dev_root->node);
+
 	free_extent_buffer(fs_info->sb_buffer);
 
 	btrfs_free_block_groups(root->fs_info);
@@ -901,8 +980,13 @@ int close_ctree(struct btrfs_root *root)
 		kfree(hasher);
 	}
 #endif
+	close_all_devices(fs_info);
+	btrfs_mapping_tree_free(&fs_info->mapping_tree);
+
 	kfree(fs_info->extent_root);
 	kfree(fs_info->tree_root);
+	kfree(fs_info->chunk_root);
+	kfree(fs_info->dev_root);
 	return 0;
 }
 
@@ -1016,4 +1100,5 @@ int btrfs_read_buffer(struct extent_buffer *buf)
 
 static struct extent_io_ops btree_extent_io_ops = {
 	.writepage_io_hook = btree_writepage_io_hook,
+	.submit_bio_hook = btree_submit_bio_hook,
 };
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 828f3a2081b6..206cb48638f7 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -20,6 +20,7 @@
 #define __DISKIO__
 
 #define BTRFS_SUPER_INFO_OFFSET (16 * 1024)
+struct btrfs_device;
 
 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
 				      u32 blocksize);
@@ -65,4 +66,5 @@ int btrfs_read_buffer(struct extent_buffer *buf);
 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len);
 void btrfs_csum_final(u32 crc, char *result);
 void btrfs_throttle(struct btrfs_root *root);
+int btrfs_open_device(struct btrfs_device *dev);
 #endif
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index ebfd304138ca..2cd957d6e8d8 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -24,37 +24,19 @@
 #include "disk-io.h"
 #include "print-tree.h"
 #include "transaction.h"
+#include "volumes.h"
 
-#define BLOCK_GROUP_DATA EXTENT_WRITEBACK
+#define BLOCK_GROUP_DATA     EXTENT_WRITEBACK
 #define BLOCK_GROUP_METADATA EXTENT_UPTODATE
+#define BLOCK_GROUP_SYSTEM   EXTENT_NEW
+
 #define BLOCK_GROUP_DIRTY EXTENT_DIRTY
 
 static int finish_current_insert(struct btrfs_trans_handle *trans, struct
 				 btrfs_root *extent_root);
 static int del_pending_extents(struct btrfs_trans_handle *trans, struct
 			       btrfs_root *extent_root);
-static int find_previous_extent(struct btrfs_root *root,
-				struct btrfs_path *path)
-{
-	struct btrfs_key found_key;
-	struct extent_buffer *leaf;
-	int ret;
 
-	while(1) {
-		if (path->slots[0] == 0) {
-			ret = btrfs_prev_leaf(root, path);
-			if (ret != 0)
-				return ret;
-		} else {
-			path->slots[0]--;
-		}
-		leaf = path->nodes[0];
-		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
-		if (found_key.type == BTRFS_EXTENT_ITEM_KEY)
-			return 0;
-	}
-	return 1;
-}
 
 static int cache_block_group(struct btrfs_root *root,
 			     struct btrfs_block_group_cache *block_group)
@@ -91,7 +73,7 @@ static int cache_block_group(struct btrfs_root *root,
 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 	if (ret < 0)
 		return ret;
-	ret = find_previous_extent(root, path);
+	ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY);
 	if (ret < 0)
 		return ret;
 	if (ret == 0) {
@@ -168,7 +150,8 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
 	block_group_cache = &info->block_group_cache;
 	ret = find_first_extent_bit(block_group_cache,
 				    bytenr, &start, &end,
-				    BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA);
+				    BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA |
+				    BLOCK_GROUP_SYSTEM);
 	if (ret) {
 		return NULL;
 	}
@@ -182,23 +165,38 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
 		return block_group;
 	return NULL;
 }
-static u64 noinline find_search_start(struct btrfs_root *root,
+
+static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
+{
+	if ((bits & BLOCK_GROUP_DATA) &&
+	    (cache->flags & BTRFS_BLOCK_GROUP_DATA))
+		return 1;
+	if ((bits & BLOCK_GROUP_METADATA) &&
+	     (cache->flags & BTRFS_BLOCK_GROUP_METADATA))
+		return 1;
+	if ((bits & BLOCK_GROUP_SYSTEM) &&
+	     (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM))
+		return 1;
+	return 0;
+}
+
+static int noinline find_search_start(struct btrfs_root *root,
 			      struct btrfs_block_group_cache **cache_ret,
-			      u64 search_start, int num, int data)
+			      u64 *start_ret, int num, int data)
 {
 	int ret;
 	struct btrfs_block_group_cache *cache = *cache_ret;
 	struct extent_io_tree *free_space_cache;
-	struct extent_state *state;
 	u64 last;
 	u64 start = 0;
+	u64 end = 0;
 	u64 cache_miss = 0;
 	u64 total_fs_bytes;
+	u64 search_start = *start_ret;
 	int wrapped = 0;
 
-	if (!cache) {
+	if (!cache)
 		goto out;
-	}
 	total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
 	free_space_cache = &root->fs_info->free_space_cache;
 
@@ -208,6 +206,9 @@ again:
 		goto out;
 
 	last = max(search_start, cache->key.objectid);
+	if (!block_group_bits(cache, data)) {
+		goto new_group;
+	}
 
 	while(1) {
 		ret = find_first_extent_bit(&root->fs_info->free_space_cache,
@@ -225,22 +226,20 @@ again:
 				cache_miss = start;
 			continue;
 		}
-		if (data != BTRFS_BLOCK_GROUP_MIXED &&
-		    start + num > cache->key.objectid + cache->key.offset)
+		if (start + num > cache->key.objectid + cache->key.offset)
 			goto new_group;
 		if (start + num  > total_fs_bytes)
 			goto new_group;
-		return start;
+		*start_ret = start;
+		return 0;
 	}
 out:
 	cache = btrfs_lookup_block_group(root->fs_info, search_start);
 	if (!cache) {
-		printk("Unable to find block group for %Lu\n",
-		       search_start);
+		printk("Unable to find block group for %Lu\n", search_start);
 		WARN_ON(1);
-		return search_start;
 	}
-	return search_start;
+	return -ENOSPC;
 
 new_group:
 	last = cache->key.objectid + cache->key.offset;
@@ -251,7 +250,6 @@ no_cache:
 		if (!wrapped) {
 			wrapped = 1;
 			last = search_start;
-			data = BTRFS_BLOCK_GROUP_MIXED;
 			goto wrapped;
 		}
 		goto out;
@@ -299,7 +297,6 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
 	int ret;
 	int full_search = 0;
 	int factor = 8;
-	int data_swap = 0;
 
 	block_group_cache = &info->block_group_cache;
 	total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
@@ -307,19 +304,12 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
 	if (!owner)
 		factor = 8;
 
-	if (data == BTRFS_BLOCK_GROUP_MIXED) {
-		bit = BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA;
-		factor = 10;
-	} else if (data)
-		bit = BLOCK_GROUP_DATA;
-	else
-		bit = BLOCK_GROUP_METADATA;
+	bit = data;
 
 	if (search_start && search_start < total_fs_bytes) {
 		struct btrfs_block_group_cache *shint;
 		shint = btrfs_lookup_block_group(info, search_start);
-		if (shint && (shint->data == data ||
-			      shint->data == BTRFS_BLOCK_GROUP_MIXED)) {
+		if (shint && block_group_bits(shint, data)) {
 			used = btrfs_block_group_used(&shint->item);
 			if (used + shint->pinned <
 			    div_factor(shint->key.offset, factor)) {
@@ -327,8 +317,8 @@ struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
 			}
 		}
 	}
-	if (hint && hint->key.objectid < total_fs_bytes &&
-	    (hint->data == data || hint->data == BTRFS_BLOCK_GROUP_MIXED)) {
+	if (hint && block_group_bits(hint, data) &&
+	    hint->key.objectid < total_fs_bytes) {
 		used = btrfs_block_group_used(&hint->item);
 		if (used + hint->pinned <
 		    div_factor(hint->key.offset, factor)) {
@@ -379,12 +369,6 @@ again:
 		full_search = 1;
 		goto again;
 	}
-	if (!data_swap) {
-		data_swap = 1;
-		bit = BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA;
-		last = search_start;
-		goto again;
-	}
 found:
 	return found_group;
 }
@@ -1002,7 +986,7 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
 static int update_block_group(struct btrfs_trans_handle *trans,
 			      struct btrfs_root *root,
 			      u64 bytenr, u64 num_bytes, int alloc,
-			      int mark_free, int data)
+			      int mark_free)
 {
 	struct btrfs_block_group_cache *cache;
 	struct btrfs_fs_info *info = root->fs_info;
@@ -1027,41 +1011,6 @@ static int update_block_group(struct btrfs_trans_handle *trans,
 		old_val = btrfs_block_group_used(&cache->item);
 		num_bytes = min(total, cache->key.offset - byte_in_group);
 		if (alloc) {
-			if (cache->data != data &&
-			    old_val < (cache->key.offset >> 1)) {
-				int bit_to_clear;
-				int bit_to_set;
-				cache->data = data;
-				if (data) {
-					bit_to_clear = BLOCK_GROUP_METADATA;
-					bit_to_set = BLOCK_GROUP_DATA;
-					cache->item.flags &=
-						~BTRFS_BLOCK_GROUP_MIXED;
-					cache->item.flags |=
-						BTRFS_BLOCK_GROUP_DATA;
-				} else {
-					bit_to_clear = BLOCK_GROUP_DATA;
-					bit_to_set = BLOCK_GROUP_METADATA;
-					cache->item.flags &=
-						~BTRFS_BLOCK_GROUP_MIXED;
-					cache->item.flags &=
-						~BTRFS_BLOCK_GROUP_DATA;
-				}
-				clear_extent_bits(&info->block_group_cache,
-						  start, end, bit_to_clear,
-						  GFP_NOFS);
-				set_extent_bits(&info->block_group_cache,
-						start, end, bit_to_set,
-						GFP_NOFS);
-			} else if (cache->data != data &&
-				   cache->data != BTRFS_BLOCK_GROUP_MIXED) {
-				cache->data = BTRFS_BLOCK_GROUP_MIXED;
-				set_extent_bits(&info->block_group_cache,
-						start, end,
-						BLOCK_GROUP_DATA |
-						BLOCK_GROUP_METADATA,
-						GFP_NOFS);
-			}
 			old_val += num_bytes;
 		} else {
 			old_val -= num_bytes;
@@ -1357,7 +1306,7 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 			return ret;
 		}
 		ret = update_block_group(trans, root, bytenr, num_bytes, 0,
-					 mark_free, 0);
+					 mark_free);
 		BUG_ON(ret);
 	}
 	btrfs_free_path(path);
@@ -1450,38 +1399,21 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans,
 				     u64 exclude_start, u64 exclude_nr,
 				     int data)
 {
-	struct btrfs_path *path;
-	struct btrfs_key key;
-	u64 hole_size = 0;
-	u64 aligned;
 	int ret;
-	int slot = 0;
-	u64 last_byte = 0;
-	u64 *last_ptr = NULL;
 	u64 orig_search_start = search_start;
-	int start_found;
-	struct extent_buffer *l;
 	struct btrfs_root * root = orig_root->fs_info->extent_root;
 	struct btrfs_fs_info *info = root->fs_info;
 	u64 total_needed = num_bytes;
-	int level;
 	struct btrfs_block_group_cache *block_group;
 	int full_scan = 0;
 	int wrapped = 0;
-	int empty_cluster;
-	u64 cached_start;
 
 	WARN_ON(num_bytes < root->sectorsize);
 	btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
 
-	level = btrfs_header_level(root->node);
-
-	if (num_bytes >= 32 * 1024 * 1024 && hint_byte) {
-		data = BTRFS_BLOCK_GROUP_MIXED;
-	}
-
 	if (search_end == (u64)-1)
 		search_end = btrfs_super_total_bytes(&info->super_copy);
+
 	if (hint_byte) {
 		block_group = btrfs_lookup_block_group(info, hint_byte);
 		if (!block_group)
@@ -1495,7 +1427,7 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans,
 	}
 
 	total_needed += empty_size;
-	path = btrfs_alloc_path();
+
 check_failed:
 	if (!block_group) {
 		block_group = btrfs_lookup_block_group(info, search_start);
@@ -1503,135 +1435,49 @@ check_failed:
 			block_group = btrfs_lookup_block_group(info,
 						       orig_search_start);
 	}
-	search_start = find_search_start(root, &block_group, search_start,
-					 total_needed, data);
-	search_start = stripe_align(root, search_start);
-	cached_start = search_start;
-	btrfs_init_path(path);
-	ins->objectid = search_start;
-	ins->offset = 0;
-	start_found = 0;
-	path->reada = 2;
-
-	ret = btrfs_search_slot(trans, root, ins, path, 0, 0);
-	if (ret < 0)
-		goto error;
-	ret = find_previous_extent(root, path);
-	if (ret < 0)
+	ret = find_search_start(root, &block_group, &search_start,
+				total_needed, data);
+	if (ret)
 		goto error;
-	l = path->nodes[0];
-	btrfs_item_key_to_cpu(l, &key, path->slots[0]);
-	while (1) {
-		l = path->nodes[0];
-		slot = path->slots[0];
-		if (slot >= btrfs_header_nritems(l)) {
-			ret = btrfs_next_leaf(root, path);
-			if (ret == 0)
-				continue;
-			if (ret < 0)
-				goto error;
 
-			search_start = max(search_start,
-					   block_group->key.objectid);
-			if (!start_found) {
-				aligned = stripe_align(root, search_start);
-				ins->objectid = aligned;
-				if (aligned >= search_end) {
-					ret = -ENOSPC;
-					goto error;
-				}
-				ins->offset = search_end - aligned;
-				start_found = 1;
-				goto check_pending;
-			}
-			ins->objectid = stripe_align(root,
-						     last_byte > search_start ?
-						     last_byte : search_start);
-			if (search_end <= ins->objectid) {
-				ret = -ENOSPC;
-				goto error;
-			}
-			ins->offset = search_end - ins->objectid;
-			BUG_ON(ins->objectid >= search_end);
-			goto check_pending;
-		}
-		btrfs_item_key_to_cpu(l, &key, slot);
-
-		if (key.objectid >= search_start && key.objectid > last_byte &&
-		    start_found) {
-			if (last_byte < search_start)
-				last_byte = search_start;
-			aligned = stripe_align(root, last_byte);
-			hole_size = key.objectid - aligned;
-			if (key.objectid > aligned && hole_size >= num_bytes) {
-				ins->objectid = aligned;
-				ins->offset = hole_size;
-				goto check_pending;
-			}
-		}
-		if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY) {
-			if (!start_found && btrfs_key_type(&key) ==
-			    BTRFS_BLOCK_GROUP_ITEM_KEY) {
-				last_byte = key.objectid;
-				start_found = 1;
-			}
-			goto next;
-		}
-
-
-		start_found = 1;
-		last_byte = key.objectid + key.offset;
-
-		if (!full_scan && data != BTRFS_BLOCK_GROUP_MIXED &&
-		    last_byte >= block_group->key.objectid +
-		    block_group->key.offset) {
-			btrfs_release_path(root, path);
-			search_start = block_group->key.objectid +
-				block_group->key.offset;
-			goto new_group;
-		}
-next:
-		path->slots[0]++;
-		cond_resched();
-	}
-check_pending:
-	/* we have to make sure we didn't find an extent that has already
-	 * been allocated by the map tree or the original allocation
-	 */
-	btrfs_release_path(root, path);
-	BUG_ON(ins->objectid < search_start);
+	search_start = stripe_align(root, search_start);
+	ins->objectid = search_start;
+	ins->offset = num_bytes;
 
 	if (ins->objectid + num_bytes >= search_end)
 		goto enospc;
-	if (!full_scan && data != BTRFS_BLOCK_GROUP_MIXED &&
-	    ins->objectid + num_bytes > block_group->
-	    key.objectid + block_group->key.offset) {
+
+	if (ins->objectid + num_bytes >
+	    block_group->key.objectid + block_group->key.offset) {
 		search_start = block_group->key.objectid +
 			block_group->key.offset;
 		goto new_group;
 	}
+
 	if (test_range_bit(&info->extent_ins, ins->objectid,
 			   ins->objectid + num_bytes -1, EXTENT_LOCKED, 0)) {
 		search_start = ins->objectid + num_bytes;
 		goto new_group;
 	}
+
 	if (test_range_bit(&info->pinned_extents, ins->objectid,
 			   ins->objectid + num_bytes -1, EXTENT_DIRTY, 0)) {
 		search_start = ins->objectid + num_bytes;
 		goto new_group;
 	}
+
 	if (exclude_nr > 0 && (ins->objectid + num_bytes > exclude_start &&
 	    ins->objectid < exclude_start + exclude_nr)) {
 		search_start = exclude_start + exclude_nr;
 		goto new_group;
 	}
-	if (!data) {
+
+	if (!(data & BLOCK_GROUP_DATA)) {
 		block_group = btrfs_lookup_block_group(info, ins->objectid);
 		if (block_group)
 			trans->block_group = block_group;
 	}
 	ins->offset = num_bytes;
-	btrfs_free_path(path);
 	return 0;
 
 new_group:
@@ -1646,7 +1492,6 @@ enospc:
 			if (!full_scan)
 				total_needed -= empty_size;
 			full_scan = 1;
-			data = BTRFS_BLOCK_GROUP_MIXED;
 		} else
 			wrapped = 1;
 	}
@@ -1657,8 +1502,6 @@ enospc:
 	goto check_failed;
 
 error:
-	btrfs_release_path(root, path);
-	btrfs_free_path(path);
 	return ret;
 }
 /*
@@ -1689,6 +1532,13 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 	struct btrfs_path *path;
 	struct btrfs_key keys[2];
 
+	if (data)
+		data = BLOCK_GROUP_DATA;
+	else if (root == root->fs_info->chunk_root)
+		data = BLOCK_GROUP_SYSTEM;
+	else
+		data = BLOCK_GROUP_METADATA;
+
 	new_hint = max(hint_byte, root->fs_info->alloc_start);
 	if (new_hint < btrfs_super_total_bytes(&info->super_copy))
 		hint_byte = new_hint;
@@ -1718,7 +1568,6 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 		set_extent_bits(&root->fs_info->extent_ins, ins->objectid,
 				ins->objectid + ins->offset - 1,
 				EXTENT_LOCKED, GFP_NOFS);
-		WARN_ON(data == 1);
 		goto update_block;
 	}
 
@@ -1768,8 +1617,7 @@ int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
 	}
 
 update_block:
-	ret = update_block_group(trans, root, ins->objectid, ins->offset, 1, 0,
-				 data);
+	ret = update_block_group(trans, root, ins->objectid, ins->offset, 1, 0);
 	if (ret) {
 		printk("update block group failed for %Lu %Lu\n",
 		       ins->objectid, ins->offset);
@@ -2457,7 +2305,7 @@ again:
 	if (ret < 0)
 		goto out;
 
-	ret = find_previous_extent(root, path);
+	ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY);
 	if (ret < 0)
 		goto out;
 	if (ret == 0) {
@@ -2604,95 +2452,48 @@ out:
 int btrfs_grow_extent_tree(struct btrfs_trans_handle *trans,
 			   struct btrfs_root *root, u64 new_size)
 {
-	struct btrfs_path *path;
-	u64 nr = 0;
-	u64 cur_byte;
-	u64 old_size;
-	unsigned long rem;
-	struct btrfs_block_group_cache *cache;
-	struct btrfs_block_group_item *item;
-	struct btrfs_fs_info *info = root->fs_info;
-	struct extent_io_tree *block_group_cache;
-	struct btrfs_key key;
-	struct extent_buffer *leaf;
-	int ret;
-	int bit;
-
-	old_size = btrfs_super_total_bytes(&info->super_copy);
-	block_group_cache = &info->block_group_cache;
-
-	root = info->extent_root;
-
-	cache = btrfs_lookup_block_group(root->fs_info, old_size - 1);
-
-	cur_byte = cache->key.objectid + cache->key.offset;
-	if (cur_byte >= new_size)
-		goto set_size;
-
-	key.offset = BTRFS_BLOCK_GROUP_SIZE;
-	btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
+	btrfs_set_super_total_bytes(&root->fs_info->super_copy, new_size);
+	return 0;
+}
 
-	path = btrfs_alloc_path();
-	if (!path)
-		return -ENOMEM;
+int find_first_block_group(struct btrfs_root *root, struct btrfs_path *path,
+			   struct btrfs_key *key)
+{
+	int ret;
+	struct btrfs_key found_key;
+	struct extent_buffer *leaf;
+	int slot;
 
-	while(cur_byte < new_size) {
-		key.objectid = cur_byte;
-		ret = btrfs_insert_empty_item(trans, root, path, &key,
-				        sizeof(struct btrfs_block_group_item));
-		BUG_ON(ret);
+	ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
+	if (ret < 0)
+		return ret;
+	while(1) {
+		slot = path->slots[0];
 		leaf = path->nodes[0];
-		item = btrfs_item_ptr(leaf, path->slots[0],
-				      struct btrfs_block_group_item);
-
-		btrfs_set_disk_block_group_used(leaf, item, 0);
-		div_long_long_rem(nr, 3, &rem);
-		if (rem) {
-			btrfs_set_disk_block_group_flags(leaf, item,
-						 BTRFS_BLOCK_GROUP_DATA);
-		} else {
-			btrfs_set_disk_block_group_flags(leaf, item, 0);
-		}
-		nr++;
-
-		cache = kmalloc(sizeof(*cache), GFP_NOFS);
-		BUG_ON(!cache);
-
-		read_extent_buffer(leaf, &cache->item, (unsigned long)item,
-				   sizeof(cache->item));
-
-		memcpy(&cache->key, &key, sizeof(key));
-		cache->cached = 0;
-		cache->pinned = 0;
-		cur_byte = key.objectid + key.offset;
-		btrfs_release_path(root, path);
-
-		if (cache->item.flags & BTRFS_BLOCK_GROUP_DATA) {
-			bit = BLOCK_GROUP_DATA;
-			cache->data = BTRFS_BLOCK_GROUP_DATA;
-		} else {
-			bit = BLOCK_GROUP_METADATA;
-			cache->data = 0;
+		if (slot >= btrfs_header_nritems(leaf)) {
+			ret = btrfs_next_leaf(root, path);
+			if (ret == 0)
+				continue;
+			if (ret < 0)
+				goto error;
+			break;
 		}
+		btrfs_item_key_to_cpu(leaf, &found_key, slot);
 
-		/* use EXTENT_LOCKED to prevent merging */
-		set_extent_bits(block_group_cache, key.objectid,
-				key.objectid + key.offset - 1,
-				bit | EXTENT_LOCKED, GFP_NOFS);
-		set_state_private(block_group_cache, key.objectid,
-				  (unsigned long)cache);
+		if (found_key.objectid >= key->objectid &&
+		    found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY)
+			return 0;
+		path->slots[0]++;
 	}
-	btrfs_free_path(path);
-set_size:
-	btrfs_set_super_total_bytes(&info->super_copy, new_size);
-	return 0;
+	ret = -ENOENT;
+error:
+	return ret;
 }
 
 int btrfs_read_block_groups(struct btrfs_root *root)
 {
 	struct btrfs_path *path;
 	int ret;
-	int err = 0;
 	int bit;
 	struct btrfs_block_group_cache *cache;
 	struct btrfs_fs_info *info = root->fs_info;
@@ -2702,28 +2503,28 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 	struct extent_buffer *leaf;
 
 	block_group_cache = &info->block_group_cache;
-
 	root = info->extent_root;
 	key.objectid = 0;
-	key.offset = BTRFS_BLOCK_GROUP_SIZE;
+	key.offset = 0;
 	btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
-
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
 
 	while(1) {
-		ret = btrfs_search_slot(NULL, info->extent_root,
-					&key, path, 0, 0);
-		if (ret != 0) {
-			err = ret;
-			break;
+		ret = find_first_block_group(root, path, &key);
+		if (ret > 0) {
+			ret = 0;
+			goto error;
 		}
+		if (ret != 0)
+			goto error;
+
 		leaf = path->nodes[0];
 		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
 		cache = kmalloc(sizeof(*cache), GFP_NOFS);
 		if (!cache) {
-			err = -1;
+			ret = -ENOMEM;
 			break;
 		}
 
@@ -2733,18 +2534,17 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 		memcpy(&cache->key, &found_key, sizeof(found_key));
 		cache->cached = 0;
 		cache->pinned = 0;
+
 		key.objectid = found_key.objectid + found_key.offset;
 		btrfs_release_path(root, path);
-
-		if (cache->item.flags & BTRFS_BLOCK_GROUP_MIXED) {
-			bit = BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA;
-			cache->data = BTRFS_BLOCK_GROUP_MIXED;
-		} else if (cache->item.flags & BTRFS_BLOCK_GROUP_DATA) {
+		cache->flags = btrfs_block_group_flags(&cache->item);
+		bit = 0;
+		if (cache->flags & BTRFS_BLOCK_GROUP_DATA) {
 			bit = BLOCK_GROUP_DATA;
-			cache->data = BTRFS_BLOCK_GROUP_DATA;
-		} else {
+		} else if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM) {
+			bit = BLOCK_GROUP_SYSTEM;
+		} else if (cache->flags & BTRFS_BLOCK_GROUP_METADATA) {
 			bit = BLOCK_GROUP_METADATA;
-			cache->data = 0;
 		}
 
 		/* use EXTENT_LOCKED to prevent merging */
@@ -2758,7 +2558,8 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 		    btrfs_super_total_bytes(&info->super_copy))
 			break;
 	}
-
+	ret = 0;
+error:
 	btrfs_free_path(path);
-	return 0;
+	return ret;
 }
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index e8130c876330..7e3a1ebde9fc 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -1706,9 +1706,9 @@ static int submit_one_bio(int rw, struct bio *bio)
 		WARN_ON(1);
 	}
 	if (tree->ops && tree->ops->submit_bio_hook)
-		tree->ops->submit_bio_hook(rw, bio);
-
-	submit_bio(rw, bio);
+		tree->ops->submit_bio_hook(page->mapping->host, rw, bio);
+	else
+		submit_bio(rw, bio);
 	if (bio_flagged(bio, BIO_EOPNOTSUPP))
 		ret = -EOPNOTSUPP;
 	bio_put(bio);
diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h
index 9d6654667089..0dca89328f98 100644
--- a/fs/btrfs/extent_io.h
+++ b/fs/btrfs/extent_io.h
@@ -28,7 +28,7 @@ struct extent_state;
 struct extent_io_ops {
 	int (*fill_delalloc)(struct inode *inode, u64 start, u64 end);
 	int (*writepage_io_hook)(struct page *page, u64 start, u64 end);
-	int (*submit_bio_hook)(int rw, struct bio *bio);
+	int (*submit_bio_hook)(struct inode *inode, int rw, struct bio *bio);
 	int (*readpage_io_hook)(struct page *page, u64 start, u64 end);
 	int (*readpage_end_io_hook)(struct page *page, u64 start, u64 end,
 				    struct extent_state *state);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 315dd550d560..17063cd2cb73 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -40,6 +40,7 @@
 #include "btrfs_inode.h"
 #include "ioctl.h"
 #include "print-tree.h"
+#include "volumes.h"
 
 struct btrfs_iget_args {
 	u64 ino;
@@ -295,20 +296,20 @@ int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end,
 	return 0;
 }
 
-int btrfs_submit_bio_hook(int rw, struct bio *bio)
+int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio)
 {
-	struct bio_vec *bvec = bio->bi_io_vec;
-	struct inode *inode = bvec->bv_page->mapping->host;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_trans_handle *trans;
 	int ret = 0;
 
-	if (rw != WRITE)
-		return 0;
+	if (rw != WRITE) {
+		goto mapit;
+	}
 
 	if (btrfs_test_opt(root, NODATASUM) ||
-	    btrfs_test_flag(inode, NODATASUM))
-		return 0;
+	    btrfs_test_flag(inode, NODATASUM)) {
+		goto mapit;
+	}
 
 	mutex_lock(&root->fs_info->fs_mutex);
 	trans = btrfs_start_transaction(root, 1);
@@ -317,7 +318,8 @@ int btrfs_submit_bio_hook(int rw, struct bio *bio)
 	ret = btrfs_end_transaction(trans, root);
 	BUG_ON(ret);
 	mutex_unlock(&root->fs_info->fs_mutex);
-	return ret;
+mapit:
+	return btrfs_map_bio(root, rw, bio);
 }
 
 int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
@@ -406,7 +408,7 @@ void btrfs_read_locked_inode(struct inode *inode)
 	struct btrfs_path *path;
 	struct extent_buffer *leaf;
 	struct btrfs_inode_item *inode_item;
-	struct btrfs_inode_timespec *tspec;
+	struct btrfs_timespec *tspec;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_key location;
 	u64 alloc_group_block;
@@ -455,7 +457,8 @@ void btrfs_read_locked_inode(struct inode *inode)
 	BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
 	if (!BTRFS_I(inode)->block_group) {
 		BTRFS_I(inode)->block_group = btrfs_find_block_group(root,
-						         NULL, 0, 0, 0);
+						 NULL, 0,
+						 BTRFS_BLOCK_GROUP_METADATA, 0);
 	}
 	btrfs_free_path(path);
 	inode_item = NULL;
@@ -1550,7 +1553,8 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 		owner = 0;
 	else
 		owner = 1;
-	group = btrfs_find_block_group(root, group, 0, 0, owner);
+	group = btrfs_find_block_group(root, group, 0,
+				       BTRFS_BLOCK_GROUP_METADATA, owner);
 	BTRFS_I(inode)->block_group = group;
 	BTRFS_I(inode)->flags = 0;
 
diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c
index da0b4dcf3617..9c1335dad40c 100644
--- a/fs/btrfs/print-tree.c
+++ b/fs/btrfs/print-tree.c
@@ -20,6 +20,40 @@
 #include "disk-io.h"
 #include "print-tree.h"
 
+static void print_chunk(struct extent_buffer *eb, struct btrfs_chunk *chunk)
+{
+	int num_stripes = btrfs_chunk_num_stripes(eb, chunk);
+	int i;
+	printk("\t\tchunk owner %llu type %llu num_stripes %d\n",
+	       (unsigned long long)btrfs_chunk_owner(eb, chunk),
+	       (unsigned long long)btrfs_chunk_type(eb, chunk),
+	       num_stripes);
+	for (i = 0 ; i < num_stripes ; i++) {
+		printk("\t\t\tstripe %d devid %llu offset %llu\n", i,
+		      (unsigned long long)btrfs_stripe_devid_nr(eb, chunk, i),
+		      (unsigned long long)btrfs_stripe_offset_nr(eb, chunk, i));
+	}
+}
+static void print_dev_item(struct extent_buffer *eb,
+			   struct btrfs_dev_item *dev_item)
+{
+	char *name;
+	int name_len;
+
+	name_len = btrfs_device_name_len(eb, dev_item);
+	name = kmalloc(name_len, GFP_NOFS);
+	if (name) {
+		read_extent_buffer(eb, name,
+				   (unsigned long)btrfs_device_name(dev_item),
+				   name_len);
+	}
+	printk("\t\tdev item name %.*s devid %llu "
+	       "total_bytes %llu bytes used %Lu\n", name_len, name,
+	       (unsigned long long)btrfs_device_id(eb, dev_item),
+	       (unsigned long long)btrfs_device_total_bytes(eb, dev_item),
+	       (unsigned long long)btrfs_device_bytes_used(eb, dev_item));
+	kfree(name);
+}
 void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 {
 	int i;
@@ -34,6 +68,7 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 	struct btrfs_key key;
 	struct btrfs_key found_key;
 	struct btrfs_extent_ref *ref;
+	struct btrfs_dev_extent *dev_extent;
 	u32 type;
 
 	printk("leaf %llu total ptrs %d free space %d\n",
@@ -106,6 +141,19 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 			printk("\t\tblock group used %llu\n",
 			       (unsigned long long)btrfs_disk_block_group_used(l, bi));
 			break;
+		case BTRFS_CHUNK_ITEM_KEY:
+			print_chunk(l, btrfs_item_ptr(l, i, struct btrfs_chunk));
+			break;
+		case BTRFS_DEV_ITEM_KEY:
+			print_dev_item(l, btrfs_item_ptr(l, i,
+					struct btrfs_dev_item));
+			break;
+		case BTRFS_DEV_EXTENT_KEY:
+			dev_extent = btrfs_item_ptr(l, i,
+						    struct btrfs_dev_extent);
+			printk("\t\tdev extent owner %llu length %llu\n",
+			       (unsigned long long)btrfs_dev_extent_owner(l, dev_extent),
+			       (unsigned long long)btrfs_dev_extent_length(l, dev_extent));
 		};
 	}
 }
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index e9a0983897f3..5e9f69244f9f 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -198,29 +198,42 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
 	return werr;
 }
 
-int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
-			    struct btrfs_root *root)
+static int update_cowonly_root(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root)
 {
 	int ret;
-	u64 old_extent_block;
-	struct btrfs_fs_info *fs_info = root->fs_info;
-	struct btrfs_root *tree_root = fs_info->tree_root;
-	struct btrfs_root *extent_root = fs_info->extent_root;
+	u64 old_root_bytenr;
+	struct btrfs_root *tree_root = root->fs_info->tree_root;
 
-	btrfs_write_dirty_block_groups(trans, extent_root);
+	btrfs_write_dirty_block_groups(trans, root);
 	while(1) {
-		old_extent_block = btrfs_root_bytenr(&extent_root->root_item);
-		if (old_extent_block == extent_root->node->start)
+		old_root_bytenr = btrfs_root_bytenr(&root->root_item);
+		if (old_root_bytenr == root->node->start)
 			break;
-		btrfs_set_root_bytenr(&extent_root->root_item,
-				      extent_root->node->start);
-		btrfs_set_root_level(&extent_root->root_item,
-				     btrfs_header_level(extent_root->node));
+		btrfs_set_root_bytenr(&root->root_item,
+				       root->node->start);
+		btrfs_set_root_level(&root->root_item,
+				     btrfs_header_level(root->node));
 		ret = btrfs_update_root(trans, tree_root,
-					&extent_root->root_key,
-					&extent_root->root_item);
+					&root->root_key,
+					&root->root_item);
 		BUG_ON(ret);
-		btrfs_write_dirty_block_groups(trans, extent_root);
+		btrfs_write_dirty_block_groups(trans, root);
+	}
+	return 0;
+}
+
+int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
+			    struct btrfs_root *root)
+{
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	struct list_head *next;
+
+	while(!list_empty(&fs_info->dirty_cowonly_roots)) {
+		next = fs_info->dirty_cowonly_roots.next;
+		list_del_init(next);
+		root = list_entry(next, struct btrfs_root, dirty_list);
+		update_cowonly_root(trans, root);
 	}
 	return 0;
 }
@@ -616,6 +629,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	unsigned long timeout = 1;
 	struct btrfs_transaction *cur_trans;
 	struct btrfs_transaction *prev_trans = NULL;
+	struct btrfs_root *chunk_root = root->fs_info->chunk_root;
 	struct list_head dirty_fs_roots;
 	struct extent_io_tree *pinned_copy;
 	DEFINE_WAIT(wait);
@@ -714,6 +728,10 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	btrfs_set_super_root_level(&root->fs_info->super_copy,
 			   btrfs_header_level(root->fs_info->tree_root->node));
 
+	btrfs_set_super_chunk_root(&root->fs_info->super_copy,
+				   chunk_root->node->start);
+	btrfs_set_super_chunk_root_level(&root->fs_info->super_copy,
+					 btrfs_header_level(chunk_root->node));
 	write_extent_buffer(root->fs_info->sb_buffer,
 			    &root->fs_info->super_copy, 0,
 			    sizeof(root->fs_info->super_copy));
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
new file mode 100644
index 000000000000..90a8d45dc6d7
--- /dev/null
+++ b/fs/btrfs/volumes.c
@@ -0,0 +1,852 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+#include <linux/sched.h>
+#include <linux/bio.h>
+#include "ctree.h"
+#include "extent_map.h"
+#include "disk-io.h"
+#include "transaction.h"
+#include "print-tree.h"
+#include "volumes.h"
+
+struct map_lookup {
+	struct btrfs_device *dev;
+	u64 physical;
+};
+
+/*
+ * this uses a pretty simple search, the expectation is that it is
+ * called very infrequently and that a given device has a small number
+ * of extents
+ */
+static int find_free_dev_extent(struct btrfs_trans_handle *trans,
+				struct btrfs_device *device,
+				struct btrfs_path *path,
+				u64 num_bytes, u64 *start)
+{
+	struct btrfs_key key;
+	struct btrfs_root *root = device->dev_root;
+	struct btrfs_dev_extent *dev_extent = NULL;
+	u64 hole_size = 0;
+	u64 last_byte = 0;
+	u64 search_start = 0;
+	u64 search_end = device->total_bytes;
+	int ret;
+	int slot = 0;
+	int start_found;
+	struct extent_buffer *l;
+
+	start_found = 0;
+	path->reada = 2;
+
+	/* FIXME use last free of some kind */
+
+	key.objectid = device->devid;
+	key.offset = search_start;
+	key.type = BTRFS_DEV_EXTENT_KEY;
+	ret = btrfs_search_slot(trans, root, &key, path, 0, 0);
+	if (ret < 0)
+		goto error;
+	ret = btrfs_previous_item(root, path, 0, key.type);
+	if (ret < 0)
+		goto error;
+	l = path->nodes[0];
+	btrfs_item_key_to_cpu(l, &key, path->slots[0]);
+	while (1) {
+		l = path->nodes[0];
+		slot = path->slots[0];
+		if (slot >= btrfs_header_nritems(l)) {
+			ret = btrfs_next_leaf(root, path);
+			if (ret == 0)
+				continue;
+			if (ret < 0)
+				goto error;
+no_more_items:
+			if (!start_found) {
+				if (search_start >= search_end) {
+					ret = -ENOSPC;
+					goto error;
+				}
+				*start = search_start;
+				start_found = 1;
+				goto check_pending;
+			}
+			*start = last_byte > search_start ?
+				last_byte : search_start;
+			if (search_end <= *start) {
+				ret = -ENOSPC;
+				goto error;
+			}
+			goto check_pending;
+		}
+		btrfs_item_key_to_cpu(l, &key, slot);
+
+		if (key.objectid < device->devid)
+			goto next;
+
+		if (key.objectid > device->devid)
+			goto no_more_items;
+
+		if (key.offset >= search_start && key.offset > last_byte &&
+		    start_found) {
+			if (last_byte < search_start)
+				last_byte = search_start;
+			hole_size = key.offset - last_byte;
+			if (key.offset > last_byte &&
+			    hole_size >= num_bytes) {
+				*start = last_byte;
+				goto check_pending;
+			}
+		}
+		if (btrfs_key_type(&key) != BTRFS_DEV_EXTENT_KEY) {
+			goto next;
+		}
+
+		start_found = 1;
+		dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent);
+		last_byte = key.offset + btrfs_dev_extent_length(l, dev_extent);
+next:
+		path->slots[0]++;
+		cond_resched();
+	}
+check_pending:
+	/* we have to make sure we didn't find an extent that has already
+	 * been allocated by the map tree or the original allocation
+	 */
+	btrfs_release_path(root, path);
+	BUG_ON(*start < search_start);
+
+	if (*start + num_bytes >= search_end) {
+		ret = -ENOSPC;
+		goto error;
+	}
+	/* check for pending inserts here */
+	return 0;
+
+error:
+	btrfs_release_path(root, path);
+	return ret;
+}
+
+int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
+			   struct btrfs_device *device,
+			   u64 owner, u64 num_bytes, u64 *start)
+{
+	int ret;
+	struct btrfs_path *path;
+	struct btrfs_root *root = device->dev_root;
+	struct btrfs_dev_extent *extent;
+	struct extent_buffer *leaf;
+	struct btrfs_key key;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	ret = find_free_dev_extent(trans, device, path, num_bytes, start);
+	if (ret)
+		goto err;
+
+	key.objectid = device->devid;
+	key.offset = *start;
+	key.type = BTRFS_DEV_EXTENT_KEY;
+	ret = btrfs_insert_empty_item(trans, root, path, &key,
+				      sizeof(*extent));
+	BUG_ON(ret);
+
+	leaf = path->nodes[0];
+	extent = btrfs_item_ptr(leaf, path->slots[0],
+				struct btrfs_dev_extent);
+	btrfs_set_dev_extent_owner(leaf, extent, owner);
+	btrfs_set_dev_extent_length(leaf, extent, num_bytes);
+	btrfs_mark_buffer_dirty(leaf);
+err:
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int find_next_chunk(struct btrfs_root *root, u64 *objectid)
+{
+	struct btrfs_path *path;
+	int ret;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+
+	path = btrfs_alloc_path();
+	BUG_ON(!path);
+
+	key.objectid = (u64)-1;
+	key.offset = (u64)-1;
+	key.type = BTRFS_CHUNK_ITEM_KEY;
+
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		goto error;
+
+	BUG_ON(ret == 0);
+
+	ret = btrfs_previous_item(root, path, 0, BTRFS_CHUNK_ITEM_KEY);
+	if (ret) {
+		*objectid = 0;
+	} else {
+		btrfs_item_key_to_cpu(path->nodes[0], &found_key,
+				      path->slots[0]);
+		*objectid = found_key.objectid + found_key.offset;
+	}
+	ret = 0;
+error:
+	btrfs_free_path(path);
+	return ret;
+}
+
+static struct btrfs_device *next_device(struct list_head *head,
+					struct list_head *last)
+{
+	struct list_head *next = last->next;
+	struct btrfs_device *dev;
+
+	if (list_empty(head))
+		return NULL;
+
+	if (next == head)
+		next = next->next;
+
+	dev = list_entry(next, struct btrfs_device, dev_list);
+	return dev;
+}
+
+static int find_next_devid(struct btrfs_root *root, struct btrfs_path *path,
+			   u64 *objectid)
+{
+	int ret;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+
+	key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
+	key.type = BTRFS_DEV_ITEM_KEY;
+	key.offset = (u64)-1;
+
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		goto error;
+
+	BUG_ON(ret == 0);
+
+	ret = btrfs_previous_item(root, path, BTRFS_DEV_ITEMS_OBJECTID,
+				  BTRFS_DEV_ITEM_KEY);
+	if (ret) {
+		*objectid = 1;
+	} else {
+		btrfs_item_key_to_cpu(path->nodes[0], &found_key,
+				      path->slots[0]);
+		*objectid = found_key.offset + 1;
+	}
+	ret = 0;
+error:
+	btrfs_release_path(root, path);
+	return ret;
+}
+
+/*
+ * the device information is stored in the chunk root
+ * the btrfs_device struct should be fully filled in
+ */
+int btrfs_add_device(struct btrfs_trans_handle *trans,
+		     struct btrfs_root *root,
+		     struct btrfs_device *device)
+{
+	int ret;
+	struct btrfs_path *path;
+	struct btrfs_dev_item *dev_item;
+	struct extent_buffer *leaf;
+	struct btrfs_key key;
+	unsigned long ptr;
+	u64 free_devid;
+
+	root = root->fs_info->chunk_root;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	ret = find_next_devid(root, path, &free_devid);
+	if (ret)
+		goto out;
+
+	key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
+	key.type = BTRFS_DEV_ITEM_KEY;
+	key.offset = free_devid;
+
+	ret = btrfs_insert_empty_item(trans, root, path, &key,
+				      sizeof(*dev_item) + device->name_len);
+	if (ret)
+		goto out;
+
+	leaf = path->nodes[0];
+	dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item);
+
+	btrfs_set_device_id(leaf, dev_item, device->devid);
+	btrfs_set_device_type(leaf, dev_item, device->type);
+	btrfs_set_device_io_align(leaf, dev_item, device->io_align);
+	btrfs_set_device_io_width(leaf, dev_item, device->io_width);
+	btrfs_set_device_sector_size(leaf, dev_item, device->sector_size);
+	btrfs_set_device_rdev(leaf, dev_item, device->rdev);
+	btrfs_set_device_partition(leaf, dev_item, device->partition);
+	btrfs_set_device_name_len(leaf, dev_item, device->name_len);
+	btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes);
+	btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used);
+
+	ptr = (unsigned long)btrfs_device_name(dev_item);
+	write_extent_buffer(leaf, device->name, ptr, device->name_len);
+
+	ptr = (unsigned long)btrfs_device_uuid(dev_item);
+	write_extent_buffer(leaf, device->uuid, ptr, BTRFS_DEV_UUID_SIZE);
+	btrfs_mark_buffer_dirty(leaf);
+	ret = 0;
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+int btrfs_update_device(struct btrfs_trans_handle *trans,
+			struct btrfs_device *device)
+{
+	int ret;
+	struct btrfs_path *path;
+	struct btrfs_root *root;
+	struct btrfs_dev_item *dev_item;
+	struct extent_buffer *leaf;
+	struct btrfs_key key;
+
+	root = device->dev_root->fs_info->chunk_root;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
+	key.type = BTRFS_DEV_ITEM_KEY;
+	key.offset = device->devid;
+
+	ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
+	if (ret < 0)
+		goto out;
+
+	if (ret > 0) {
+		ret = -ENOENT;
+		goto out;
+	}
+
+	leaf = path->nodes[0];
+	dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item);
+
+	btrfs_set_device_id(leaf, dev_item, device->devid);
+	btrfs_set_device_type(leaf, dev_item, device->type);
+	btrfs_set_device_io_align(leaf, dev_item, device->io_align);
+	btrfs_set_device_io_width(leaf, dev_item, device->io_width);
+	btrfs_set_device_sector_size(leaf, dev_item, device->sector_size);
+	btrfs_set_device_rdev(leaf, dev_item, device->rdev);
+	btrfs_set_device_partition(leaf, dev_item, device->partition);
+	btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes);
+	btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used);
+	btrfs_mark_buffer_dirty(leaf);
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+int btrfs_add_system_chunk(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *root,
+			   struct btrfs_key *key,
+			   struct btrfs_chunk *chunk, int item_size)
+{
+	struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
+	struct btrfs_disk_key disk_key;
+	u32 array_size;
+	u8 *ptr;
+
+	array_size = btrfs_super_sys_array_size(super_copy);
+	if (array_size + item_size > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE)
+		return -EFBIG;
+
+	ptr = super_copy->sys_chunk_array + array_size;
+	btrfs_cpu_key_to_disk(&disk_key, key);
+	memcpy(ptr, &disk_key, sizeof(disk_key));
+	ptr += sizeof(disk_key);
+	memcpy(ptr, chunk, item_size);
+	item_size += sizeof(disk_key);
+	btrfs_set_super_sys_array_size(super_copy, array_size + item_size);
+	return 0;
+}
+
+int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
+		      struct btrfs_root *extent_root, u64 *start,
+		      u64 *num_bytes, u32 type)
+{
+	u64 dev_offset;
+	struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root;
+	struct btrfs_stripe *stripes;
+	struct btrfs_device *device = NULL;
+	struct btrfs_chunk *chunk;
+	struct list_head *dev_list = &extent_root->fs_info->devices;
+	struct list_head *last_dev = extent_root->fs_info->last_device;
+	struct extent_map_tree *em_tree;
+	struct map_lookup *map;
+	struct extent_map *em;
+	u64 physical;
+	u64 calc_size = 1024 * 1024 * 1024;
+	int num_stripes;
+	int ret;
+	int index = 0;
+	struct btrfs_key key;
+
+
+	ret = find_next_chunk(chunk_root, &key.objectid);
+	if (ret)
+		return ret;
+
+	num_stripes = 1;
+	chunk = kmalloc(btrfs_chunk_item_size(num_stripes), GFP_NOFS);
+	if (!chunk)
+		return -ENOMEM;
+
+	stripes = &chunk->stripe;
+
+	*num_bytes = calc_size;
+	while(index < num_stripes) {
+		device = next_device(dev_list, last_dev);
+		BUG_ON(!device);
+		last_dev = &device->dev_list;
+		extent_root->fs_info->last_device = last_dev;
+
+		ret = btrfs_alloc_dev_extent(trans, device,
+					     key.objectid,
+					     calc_size, &dev_offset);
+		BUG_ON(ret);
+
+		device->bytes_used += calc_size;
+		ret = btrfs_update_device(trans, device);
+		BUG_ON(ret);
+
+		btrfs_set_stack_stripe_devid(stripes + index, device->devid);
+		btrfs_set_stack_stripe_offset(stripes + index, dev_offset);
+		physical = dev_offset;
+		index++;
+	}
+
+	/* key.objectid was set above */
+	key.offset = *num_bytes;
+	key.type = BTRFS_CHUNK_ITEM_KEY;
+	btrfs_set_stack_chunk_owner(chunk, extent_root->root_key.objectid);
+	btrfs_set_stack_chunk_stripe_len(chunk, 64 * 1024);
+	btrfs_set_stack_chunk_type(chunk, type);
+	btrfs_set_stack_chunk_num_stripes(chunk, num_stripes);
+	btrfs_set_stack_chunk_io_align(chunk, extent_root->sectorsize);
+	btrfs_set_stack_chunk_io_width(chunk, extent_root->sectorsize);
+	btrfs_set_stack_chunk_sector_size(chunk, extent_root->sectorsize);
+
+	ret = btrfs_insert_item(trans, chunk_root, &key, chunk,
+				btrfs_chunk_item_size(num_stripes));
+	BUG_ON(ret);
+	*start = key.objectid;
+
+	em = alloc_extent_map(GFP_NOFS);
+	if (!em)
+		return -ENOMEM;
+	map = kmalloc(sizeof(*map), GFP_NOFS);
+	if (!map) {
+		free_extent_map(em);
+		return -ENOMEM;
+	}
+
+	em->bdev = (struct block_device *)map;
+	em->start = key.objectid;
+	em->len = key.offset;
+	em->block_start = 0;
+
+	map->physical = physical;
+	map->dev = device;
+
+	if (!map->dev) {
+		kfree(map);
+		free_extent_map(em);
+		return -EIO;
+	}
+	kfree(chunk);
+
+	em_tree = &extent_root->fs_info->mapping_tree.map_tree;
+	spin_lock(&em_tree->lock);
+	ret = add_extent_mapping(em_tree, em);
+	BUG_ON(ret);
+	spin_unlock(&em_tree->lock);
+	free_extent_map(em);
+	return ret;
+}
+
+void btrfs_mapping_init(struct btrfs_mapping_tree *tree)
+{
+	extent_map_tree_init(&tree->map_tree, GFP_NOFS);
+}
+
+void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree)
+{
+	struct extent_map *em;
+
+	while(1) {
+		spin_lock(&tree->map_tree.lock);
+		em = lookup_extent_mapping(&tree->map_tree, 0, (u64)-1);
+		if (em)
+			remove_extent_mapping(&tree->map_tree, em);
+		spin_unlock(&tree->map_tree.lock);
+		if (!em)
+			break;
+		kfree(em->bdev);
+		/* once for us */
+		free_extent_map(em);
+		/* once for the tree */
+		free_extent_map(em);
+	}
+}
+
+int btrfs_map_block(struct btrfs_mapping_tree *map_tree,
+		    u64 logical, u64 *phys, u64 *length,
+		    struct btrfs_device **dev)
+{
+	struct extent_map *em;
+	struct map_lookup *map;
+	struct extent_map_tree *em_tree = &map_tree->map_tree;
+	u64 offset;
+
+
+	spin_lock(&em_tree->lock);
+	em = lookup_extent_mapping(em_tree, logical, *length);
+	BUG_ON(!em);
+
+	BUG_ON(em->start > logical || em->start + em->len < logical);
+	map = (struct map_lookup *)em->bdev;
+	offset = logical - em->start;
+	*phys = map->physical + offset;
+	*length = em->len - offset;
+	*dev = map->dev;
+	free_extent_map(em);
+	spin_unlock(&em_tree->lock);
+	return 0;
+}
+
+int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio)
+{
+	struct btrfs_mapping_tree *map_tree;
+	struct btrfs_device *dev;
+	u64 logical = bio->bi_sector << 9;
+	u64 physical;
+	u64 length = 0;
+	u64 map_length;
+	struct bio_vec *bvec;
+	int i;
+	int ret;
+
+	bio_for_each_segment(bvec, bio, i) {
+		length += bvec->bv_len;
+	}
+	map_tree = &root->fs_info->mapping_tree;
+	map_length = length;
+	ret = btrfs_map_block(map_tree, logical, &physical, &map_length, &dev);
+	BUG_ON(map_length < length);
+	bio->bi_sector = physical >> 9;
+	bio->bi_bdev = dev->bdev;
+	submit_bio(rw, bio);
+	return 0;
+}
+
+struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid)
+{
+	struct btrfs_device *dev;
+	struct list_head *cur = root->fs_info->devices.next;
+	struct list_head *head = &root->fs_info->devices;
+
+	while(cur != head) {
+		dev = list_entry(cur, struct btrfs_device, dev_list);
+		if (dev->devid == devid)
+			return dev;
+		cur = cur->next;
+	}
+	return NULL;
+}
+
+static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
+			  struct extent_buffer *leaf,
+			  struct btrfs_chunk *chunk)
+{
+	struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree;
+	struct map_lookup *map;
+	struct extent_map *em;
+	u64 logical;
+	u64 length;
+	u64 devid;
+	int ret;
+
+	logical = key->objectid;
+	length = key->offset;
+	spin_lock(&map_tree->map_tree.lock);
+	em = lookup_extent_mapping(&map_tree->map_tree, logical, 1);
+
+	/* already mapped? */
+	if (em && em->start <= logical && em->start + em->len > logical) {
+		free_extent_map(em);
+		spin_unlock(&map_tree->map_tree.lock);
+		return 0;
+	} else if (em) {
+		free_extent_map(em);
+	}
+	spin_unlock(&map_tree->map_tree.lock);
+
+	map = kzalloc(sizeof(*map), GFP_NOFS);
+	if (!map)
+		return -ENOMEM;
+
+	em = alloc_extent_map(GFP_NOFS);
+	if (!em)
+		return -ENOMEM;
+	map = kmalloc(sizeof(*map), GFP_NOFS);
+	if (!map) {
+		free_extent_map(em);
+		return -ENOMEM;
+	}
+
+	em->bdev = (struct block_device *)map;
+	em->start = logical;
+	em->len = length;
+	em->block_start = 0;
+
+	map->physical = btrfs_stripe_offset_nr(leaf, chunk, 0);
+	devid = btrfs_stripe_devid_nr(leaf, chunk, 0);
+	map->dev = btrfs_find_device(root, devid);
+	if (!map->dev) {
+		kfree(map);
+		free_extent_map(em);
+		return -EIO;
+	}
+
+	spin_lock(&map_tree->map_tree.lock);
+	ret = add_extent_mapping(&map_tree->map_tree, em);
+	BUG_ON(ret);
+	spin_unlock(&map_tree->map_tree.lock);
+	free_extent_map(em);
+
+	return 0;
+}
+
+static int fill_device_from_item(struct extent_buffer *leaf,
+				 struct btrfs_dev_item *dev_item,
+				 struct btrfs_device *device)
+{
+	unsigned long ptr;
+	char *name;
+
+	device->devid = btrfs_device_id(leaf, dev_item);
+	device->total_bytes = btrfs_device_total_bytes(leaf, dev_item);
+	device->bytes_used = btrfs_device_bytes_used(leaf, dev_item);
+	device->type = btrfs_device_type(leaf, dev_item);
+	device->io_align = btrfs_device_io_align(leaf, dev_item);
+	device->io_width = btrfs_device_io_width(leaf, dev_item);
+	device->sector_size = btrfs_device_sector_size(leaf, dev_item);
+	device->rdev = btrfs_device_rdev(leaf, dev_item);
+	device->partition = btrfs_device_partition(leaf, dev_item);
+	device->name_len = btrfs_device_name_len(leaf, dev_item);
+
+	ptr = (unsigned long)btrfs_device_uuid(dev_item);
+	read_extent_buffer(leaf, device->uuid, ptr, BTRFS_DEV_UUID_SIZE);
+
+	name = kmalloc(device->name_len + 1, GFP_NOFS);
+	if (!name)
+		return -ENOMEM;
+	device->name = name;
+	ptr = (unsigned long)btrfs_device_name(dev_item);
+	read_extent_buffer(leaf, name, ptr, device->name_len);
+	name[device->name_len] = '\0';
+	return 0;
+}
+
+static int read_one_dev(struct btrfs_root *root, struct btrfs_key *key,
+			struct extent_buffer *leaf,
+			struct btrfs_dev_item *dev_item)
+{
+	struct btrfs_device *device;
+	u64 devid;
+	int ret;
+
+	devid = btrfs_device_id(leaf, dev_item);
+	if (btrfs_find_device(root, devid))
+		return 0;
+
+	device = kmalloc(sizeof(*device), GFP_NOFS);
+	if (!device)
+		return -ENOMEM;
+
+	fill_device_from_item(leaf, dev_item, device);
+	device->dev_root = root->fs_info->dev_root;
+	device->bdev = root->fs_info->sb->s_bdev;
+	list_add(&device->dev_list, &root->fs_info->devices);
+	memcpy(&device->dev_key, key, sizeof(*key));
+	ret = 0;
+#if 0
+	ret = btrfs_open_device(device);
+	if (ret) {
+		kfree(device);
+	}
+#endif
+	return ret;
+}
+
+int btrfs_read_sys_array(struct btrfs_root *root)
+{
+	struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
+	struct extent_buffer *sb = root->fs_info->sb_buffer;
+	struct btrfs_disk_key *disk_key;
+	struct btrfs_dev_item *dev_item;
+	struct btrfs_chunk *chunk;
+	struct btrfs_key key;
+	u32 num_stripes;
+	u32 array_size;
+	u32 len = 0;
+	u8 *ptr;
+	unsigned long sb_ptr;
+	u32 cur;
+	int ret;
+	int dev_only = 1;
+
+	array_size = btrfs_super_sys_array_size(super_copy);
+
+	/*
+	 * we do this loop twice, once for the device items and
+	 * once for all of the chunks.  This way there are device
+	 * structs filled in for every chunk
+	 */
+again:
+	ptr = super_copy->sys_chunk_array;
+	sb_ptr = offsetof(struct btrfs_super_block, sys_chunk_array);
+	cur = 0;
+
+	while (cur < array_size) {
+		disk_key = (struct btrfs_disk_key *)ptr;
+		btrfs_disk_key_to_cpu(&key, disk_key);
+
+		len = sizeof(*disk_key);
+		ptr += len;
+		sb_ptr += len;
+		cur += len;
+
+		if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID &&
+		    key.type == BTRFS_DEV_ITEM_KEY) {
+			dev_item = (struct btrfs_dev_item *)sb_ptr;
+			if (dev_only) {
+				ret = read_one_dev(root, &key, sb, dev_item);
+				BUG_ON(ret);
+			}
+			len = sizeof(*dev_item);
+			len += btrfs_device_name_len(sb, dev_item);
+		} else if (key.type == BTRFS_CHUNK_ITEM_KEY) {
+
+			chunk = (struct btrfs_chunk *)sb_ptr;
+			if (!dev_only) {
+				ret = read_one_chunk(root, &key, sb, chunk);
+				BUG_ON(ret);
+			}
+			num_stripes = btrfs_chunk_num_stripes(sb, chunk);
+			len = btrfs_chunk_item_size(num_stripes);
+		} else {
+			BUG();
+		}
+		ptr += len;
+		sb_ptr += len;
+		cur += len;
+	}
+	if (dev_only == 1) {
+		dev_only = 0;
+		goto again;
+	}
+	return 0;
+}
+
+int btrfs_read_chunk_tree(struct btrfs_root *root)
+{
+	struct btrfs_path *path;
+	struct extent_buffer *leaf;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	int ret;
+	int slot;
+
+	root = root->fs_info->chunk_root;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	/* first we search for all of the device items, and then we
+	 * read in all of the chunk items.  This way we can create chunk
+	 * mappings that reference all of the devices that are afound
+	 */
+	key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
+	key.offset = 0;
+	key.type = 0;
+again:
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	while(1) {
+		leaf = path->nodes[0];
+		slot = path->slots[0];
+		if (slot >= btrfs_header_nritems(leaf)) {
+			ret = btrfs_next_leaf(root, path);
+			if (ret == 0)
+				continue;
+			if (ret < 0)
+				goto error;
+			break;
+		}
+		btrfs_item_key_to_cpu(leaf, &found_key, slot);
+		if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) {
+			if (found_key.objectid != BTRFS_DEV_ITEMS_OBJECTID)
+				break;
+			if (found_key.type == BTRFS_DEV_ITEM_KEY) {
+				struct btrfs_dev_item *dev_item;
+				dev_item = btrfs_item_ptr(leaf, slot,
+						  struct btrfs_dev_item);
+				ret = read_one_dev(root, &found_key, leaf,
+						   dev_item);
+				BUG_ON(ret);
+			}
+		} else if (found_key.type == BTRFS_CHUNK_ITEM_KEY) {
+			struct btrfs_chunk *chunk;
+			chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk);
+			ret = read_one_chunk(root, &found_key, leaf, chunk);
+		}
+		path->slots[0]++;
+	}
+	if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) {
+		key.objectid = 0;
+		btrfs_release_path(root, path);
+		goto again;
+	}
+
+	btrfs_free_path(path);
+	ret = 0;
+error:
+	return ret;
+}
+
diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h
new file mode 100644
index 000000000000..4a47dcb00a73
--- /dev/null
+++ b/fs/btrfs/volumes.h
@@ -0,0 +1,78 @@
+/*
+ * Copyright (C) 2007 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef __BTRFS_VOLUMES_
+#define __BTRFS_VOLUMES_
+struct btrfs_device {
+	struct list_head dev_list;
+	struct btrfs_root *dev_root;
+	struct btrfs_key dev_key;
+
+	struct block_device *bdev;
+
+	/* the internal btrfs device id */
+	u64 devid;
+
+	/* size of the device */
+	u64 total_bytes;
+
+	/* bytes used */
+	u64 bytes_used;
+
+	/* optimal io alignment for this device */
+	u32 io_align;
+
+	/* optimal io width for this device */
+	u32 io_width;
+
+	/* minimal io size for this device */
+	u32 sector_size;
+
+	/* the kernel device number */
+	u64 rdev;
+
+	/* type and info about this device */
+	u64 type;
+
+	/* partition number, 0 for whole dev */
+	int partition;
+
+	/* length of the name data at the end of the item */
+	int name_len;
+
+	/* physical drive uuid (or lvm uuid) */
+	u8 uuid[BTRFS_DEV_UUID_SIZE];
+
+	char *name;
+};
+
+int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
+			   struct btrfs_device *device,
+			   u64 owner, u64 num_bytes, u64 *start);
+int btrfs_map_block(struct btrfs_mapping_tree *map_tree,
+		    u64 logical, u64 *phys, u64 *length,
+		    struct btrfs_device **dev);
+int btrfs_read_sys_array(struct btrfs_root *root);
+int btrfs_read_chunk_tree(struct btrfs_root *root);
+int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
+		      struct btrfs_root *extent_root, u64 *start,
+		      u64 *num_bytes, u32 type);
+void btrfs_mapping_init(struct btrfs_mapping_tree *tree);
+void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree);
+int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio);
+#endif
-- 
cgit v1.2.3


From 8a4b83cc8bd75fca29ac68615896d9e92820e7c2 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Mon, 24 Mar 2008 15:02:07 -0400
Subject: Btrfs: Add support for device scanning and detection ioctls

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.h   |  21 ++++-
 fs/btrfs/disk-io.c |  24 ++++--
 fs/btrfs/disk-io.h |   4 +-
 fs/btrfs/ioctl.h   |   6 +-
 fs/btrfs/super.c   |  61 ++++++++++----
 fs/btrfs/volumes.c | 236 +++++++++++++++++++++++++++++++++++++++++++++++++----
 fs/btrfs/volumes.h |  25 ++++++
 7 files changed, 333 insertions(+), 44 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index acf22ad6115c..7556f8319c60 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -37,7 +37,7 @@ extern struct kmem_cache *btrfs_transaction_cachep;
 extern struct kmem_cache *btrfs_bit_radix_cachep;
 extern struct kmem_cache *btrfs_path_cachep;
 
-#define BTRFS_MAGIC "_B4RfS_M"
+#define BTRFS_MAGIC "_B5RfS_M"
 
 #define BTRFS_MAX_LEVEL 8
 
@@ -238,6 +238,7 @@ struct btrfs_super_block {
 	__le64 total_bytes;
 	__le64 bytes_used;
 	__le64 root_dir_objectid;
+	__le64 num_devices;
 	__le32 sectorsize;
 	__le32 nodesize;
 	__le32 leafsize;
@@ -440,6 +441,7 @@ struct btrfs_block_group_cache {
 };
 
 struct btrfs_device;
+struct btrfs_fs_devices;
 struct btrfs_fs_info {
 	u8 fsid[BTRFS_FSID_SIZE];
 	struct btrfs_root *extent_root;
@@ -489,7 +491,7 @@ struct btrfs_fs_info {
 	u64 total_pinned;
 	struct list_head dirty_cowonly_roots;
 
-	struct list_head devices;
+	struct btrfs_fs_devices *fs_devices;
 	struct list_head space_info;
 	spinlock_t delalloc_lock;
 	spinlock_t new_trans_lock;
@@ -677,6 +679,19 @@ BTRFS_SETGET_FUNCS(device_io_width, struct btrfs_dev_item, io_width, 32);
 BTRFS_SETGET_FUNCS(device_sector_size, struct btrfs_dev_item, sector_size, 32);
 BTRFS_SETGET_FUNCS(device_id, struct btrfs_dev_item, devid, 64);
 
+BTRFS_SETGET_STACK_FUNCS(stack_device_type, struct btrfs_dev_item, type, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_device_total_bytes, struct btrfs_dev_item,
+			 total_bytes, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_device_bytes_used, struct btrfs_dev_item,
+			 bytes_used, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_device_io_align, struct btrfs_dev_item,
+			 io_align, 32);
+BTRFS_SETGET_STACK_FUNCS(stack_device_io_width, struct btrfs_dev_item,
+			 io_width, 32);
+BTRFS_SETGET_STACK_FUNCS(stack_device_sector_size, struct btrfs_dev_item,
+			 sector_size, 32);
+BTRFS_SETGET_STACK_FUNCS(stack_device_id, struct btrfs_dev_item, devid, 64);
+
 static inline char *btrfs_device_uuid(struct btrfs_dev_item *d)
 {
 	return (char *)d + offsetof(struct btrfs_dev_item, uuid);
@@ -1106,6 +1121,8 @@ BTRFS_SETGET_STACK_FUNCS(super_stripesize, struct btrfs_super_block,
 			 stripesize, 32);
 BTRFS_SETGET_STACK_FUNCS(super_root_dir, struct btrfs_super_block,
 			 root_dir_objectid, 64);
+BTRFS_SETGET_STACK_FUNCS(super_num_devices, struct btrfs_super_block,
+			 num_devices, 64);
 
 static inline unsigned long btrfs_leaf_data(struct extent_buffer *l)
 {
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 4890151cd68d..f971a29e4f20 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -365,12 +365,12 @@ static int close_all_devices(struct btrfs_fs_info *fs_info)
 	struct list_head *next;
 	struct btrfs_device *device;
 
-	list = &fs_info->devices;
-	while(!list_empty(list)) {
-		next = list->next;
-		list_del(next);
+	list = &fs_info->fs_devices->devices;
+	list_for_each(next, list) {
 		device = list_entry(next, struct btrfs_device, dev_list);
-		kfree(device);
+		if (device->bdev && device->bdev != fs_info->sb->s_bdev)
+			close_bdev_excl(device->bdev);
+		device->bdev = NULL;
 	}
 	return 0;
 }
@@ -655,7 +655,8 @@ static int add_hasher(struct btrfs_fs_info *info, char *type) {
 	return 0;
 }
 #endif
-struct btrfs_root *open_ctree(struct super_block *sb)
+struct btrfs_root *open_ctree(struct super_block *sb,
+			      struct btrfs_fs_devices *fs_devices)
 {
 	u32 sectorsize;
 	u32 nodesize;
@@ -697,8 +698,8 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	fs_info->extent_root = extent_root;
 	fs_info->chunk_root = chunk_root;
 	fs_info->dev_root = dev_root;
+	fs_info->fs_devices = fs_devices;
 	INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
-	INIT_LIST_HEAD(&fs_info->devices);
 	INIT_LIST_HEAD(&fs_info->space_info);
 	btrfs_mapping_init(&fs_info->mapping_tree);
 	fs_info->sb = sb;
@@ -779,6 +780,12 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	if (!btrfs_super_root(disk_super))
 		goto fail_sb_buffer;
 
+	if (btrfs_super_num_devices(disk_super) != fs_devices->num_devices) {
+		printk("Btrfs: wanted %llu devices, but found %llu\n",
+		       (unsigned long long)btrfs_super_num_devices(disk_super),
+		       (unsigned long long)fs_devices->num_devices);
+		goto fail_sb_buffer;
+	}
 	nodesize = btrfs_super_nodesize(disk_super);
 	leafsize = btrfs_super_leafsize(disk_super);
 	sectorsize = btrfs_super_sectorsize(disk_super);
@@ -799,8 +806,6 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	}
 
 	mutex_lock(&fs_info->fs_mutex);
-	ret = btrfs_read_super_device(tree_root, fs_info->sb_buffer);
-	BUG_ON(ret);
 
 	ret = btrfs_read_sys_array(tree_root);
 	BUG_ON(ret);
@@ -859,6 +864,7 @@ fail_sb_buffer:
 fail_iput:
 	iput(fs_info->btree_inode);
 fail:
+	close_all_devices(fs_info);
 	kfree(extent_root);
 	kfree(tree_root);
 	kfree(fs_info);
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 206cb48638f7..b7cbc58a5553 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -21,6 +21,7 @@
 
 #define BTRFS_SUPER_INFO_OFFSET (16 * 1024)
 struct btrfs_device;
+struct btrfs_fs_devices;
 
 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
 				      u32 blocksize);
@@ -29,7 +30,8 @@ struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
 						   u64 bytenr, u32 blocksize);
 int clean_tree_block(struct btrfs_trans_handle *trans,
 		     struct btrfs_root *root, struct extent_buffer *buf);
-struct btrfs_root *open_ctree(struct super_block *sb);
+struct btrfs_root *open_ctree(struct super_block *sb,
+			      struct btrfs_fs_devices *fs_devices);
 int close_ctree(struct btrfs_root *root);
 int write_ctree_super(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root);
diff --git a/fs/btrfs/ioctl.h b/fs/btrfs/ioctl.h
index 8c6290665d49..4551e82013c8 100644
--- a/fs/btrfs/ioctl.h
+++ b/fs/btrfs/ioctl.h
@@ -22,8 +22,10 @@
 
 #define BTRFS_IOCTL_MAGIC 0x94
 #define BTRFS_VOL_NAME_MAX 255
+#define BTRFS_PATH_NAME_MAX 4095
+
 struct btrfs_ioctl_vol_args {
-	char name[BTRFS_VOL_NAME_MAX + 1];
+	char name[BTRFS_PATH_NAME_MAX + 1];
 };
 
 #define BTRFS_IOC_SNAP_CREATE _IOW(BTRFS_IOCTL_MAGIC, 1, \
@@ -32,4 +34,6 @@ struct btrfs_ioctl_vol_args {
 				   struct btrfs_ioctl_vol_args)
 #define BTRFS_IOC_RESIZE _IOW(BTRFS_IOCTL_MAGIC, 3, \
 				   struct btrfs_ioctl_vol_args)
+#define BTRFS_IOC_SCAN_DEV _IOW(BTRFS_IOCTL_MAGIC, 4, \
+				   struct btrfs_ioctl_vol_args)
 #endif
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 67ed216df475..9624923a33dc 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -44,6 +44,7 @@
 #include "ioctl.h"
 #include "print-tree.h"
 #include "xattr.h"
+#include "volumes.h"
 
 #define BTRFS_SUPER_MAGIC 0x9123683E
 
@@ -216,7 +217,9 @@ static int parse_options (char * options,
 	return 1;
 }
 
-static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
+static int btrfs_fill_super(struct super_block * sb,
+			    struct btrfs_fs_devices *fs_devices,
+			    void * data, int silent)
 {
 	struct inode * inode;
 	struct dentry * root_dentry;
@@ -231,7 +234,7 @@ static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
 	sb->s_xattr = btrfs_xattr_handlers;
 	sb->s_time_gran = 1;
 
-	tree_root = open_ctree(sb);
+	tree_root = open_ctree(sb, fs_devices);
 
 	if (!tree_root || IS_ERR(tree_root)) {
 		printk("btrfs: open_ctree failed\n");
@@ -334,18 +337,23 @@ static int test_bdev_super(struct super_block *s, void *data)
 
 int btrfs_get_sb_bdev(struct file_system_type *fs_type,
 	int flags, const char *dev_name, void *data,
-	int (*fill_super)(struct super_block *, void *, int),
 	struct vfsmount *mnt, const char *subvol)
 {
 	struct block_device *bdev = NULL;
 	struct super_block *s;
 	struct dentry *root;
+	struct btrfs_fs_devices *fs_devices = NULL;
 	int error = 0;
 
-	bdev = open_bdev_excl(dev_name, flags, fs_type);
-	if (IS_ERR(bdev))
-		return PTR_ERR(bdev);
+	error = btrfs_scan_one_device(dev_name, flags, fs_type, &fs_devices);
+	if (error)
+		return error;
 
+	error = btrfs_open_devices(fs_devices, flags, fs_type);
+	if (error)
+		return error;
+
+	bdev = fs_devices->lowest_bdev;
 	/*
 	 * once the super is inserted into the list by sget, s_umount
 	 * will protect the lockfs code from trying to start a snapshot
@@ -372,7 +380,8 @@ int btrfs_get_sb_bdev(struct file_system_type *fs_type,
 		s->s_flags = flags;
 		strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
 		sb_set_blocksize(s, block_size(bdev));
-		error = fill_super(s, data, flags & MS_SILENT ? 1 : 0);
+		error = btrfs_fill_super(s, fs_devices, data,
+					 flags & MS_SILENT ? 1 : 0);
 		if (error) {
 			up_write(&s->s_umount);
 			deactivate_super(s);
@@ -408,7 +417,7 @@ int btrfs_get_sb_bdev(struct file_system_type *fs_type,
 error_s:
 	error = PTR_ERR(s);
 error_bdev:
-	close_bdev_excl(bdev);
+	btrfs_close_devices(fs_devices);
 error:
 	return error;
 }
@@ -421,8 +430,7 @@ static int btrfs_get_sb(struct file_system_type *fs_type,
 	char *subvol_name = NULL;
 
 	parse_options((char *)data, NULL, &subvol_name);
-	ret = btrfs_get_sb_bdev(fs_type, flags, dev_name, data,
-			btrfs_fill_super, mnt,
+	ret = btrfs_get_sb_bdev(fs_type, flags, dev_name, data, mnt,
 			subvol_name ? subvol_name : "default");
 	if (subvol_name)
 		kfree(subvol_name);
@@ -445,13 +453,6 @@ static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 	return 0;
 }
 
-static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
-				unsigned long arg)
-{
-	printk("btrfs control ioctl %d\n", cmd);
-	return 0;
-}
-
 static struct file_system_type btrfs_fs_type = {
 	.owner		= THIS_MODULE,
 	.name		= "btrfs",
@@ -460,6 +461,31 @@ static struct file_system_type btrfs_fs_type = {
 	.fs_flags	= FS_REQUIRES_DEV,
 };
 
+static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
+				unsigned long arg)
+{
+	struct btrfs_ioctl_vol_args *vol;
+	struct btrfs_fs_devices *fs_devices;
+	int ret;
+	int len;
+
+	vol = kmalloc(sizeof(*vol), GFP_KERNEL);
+	if (copy_from_user(vol, (void __user *)arg, sizeof(*vol))) {
+		ret = -EFAULT;
+		goto out;
+	}
+	len = strnlen(vol->name, BTRFS_PATH_NAME_MAX);
+	switch (cmd) {
+	case BTRFS_IOC_SCAN_DEV:
+		ret = btrfs_scan_one_device(vol->name, MS_RDONLY,
+					    &btrfs_fs_type, &fs_devices);
+		break;
+	}
+out:
+	kfree(vol);
+	return 0;
+}
+
 static void btrfs_write_super_lockfs(struct super_block *sb)
 {
 	struct btrfs_root *root = btrfs_sb(sb);
@@ -567,6 +593,7 @@ static void __exit exit_btrfs_fs(void)
 	btrfs_interface_exit();
 	unregister_filesystem(&btrfs_fs_type);
 	btrfs_exit_sysfs();
+	btrfs_cleanup_fs_uuids();
 }
 
 module_init(init_btrfs_fs)
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 16fb6bbe6e28..263f01cc3db4 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -17,6 +17,7 @@
  */
 #include <linux/sched.h>
 #include <linux/bio.h>
+#include <linux/buffer_head.h>
 #include "ctree.h"
 #include "extent_map.h"
 #include "disk-io.h"
@@ -28,6 +29,215 @@ struct map_lookup {
 	struct btrfs_device *dev;
 	u64 physical;
 };
+static DEFINE_MUTEX(uuid_mutex);
+static LIST_HEAD(fs_uuids);
+
+int btrfs_cleanup_fs_uuids(void)
+{
+	struct btrfs_fs_devices *fs_devices;
+	struct list_head *uuid_cur;
+	struct list_head *devices_cur;
+	struct btrfs_device *dev;
+
+	list_for_each(uuid_cur, &fs_uuids) {
+		fs_devices = list_entry(uuid_cur, struct btrfs_fs_devices,
+					list);
+		while(!list_empty(&fs_devices->devices)) {
+			devices_cur = fs_devices->devices.next;
+			dev = list_entry(devices_cur, struct btrfs_device,
+					 dev_list);
+			printk("uuid cleanup finds %s\n", dev->name);
+			if (dev->bdev) {
+				printk("closing\n");
+				close_bdev_excl(dev->bdev);
+			}
+			list_del(&dev->dev_list);
+			kfree(dev);
+		}
+	}
+	return 0;
+}
+
+static struct btrfs_device *__find_device(struct list_head *head, u64 devid)
+{
+	struct btrfs_device *dev;
+	struct list_head *cur;
+
+	list_for_each(cur, head) {
+		dev = list_entry(cur, struct btrfs_device, dev_list);
+		if (dev->devid == devid)
+			return dev;
+	}
+	return NULL;
+}
+
+static struct btrfs_fs_devices *find_fsid(u8 *fsid)
+{
+	struct list_head *cur;
+	struct btrfs_fs_devices *fs_devices;
+
+	list_for_each(cur, &fs_uuids) {
+		fs_devices = list_entry(cur, struct btrfs_fs_devices, list);
+		if (memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE) == 0)
+			return fs_devices;
+	}
+	return NULL;
+}
+
+static int device_list_add(const char *path,
+			   struct btrfs_super_block *disk_super,
+			   u64 devid, struct btrfs_fs_devices **fs_devices_ret)
+{
+	struct btrfs_device *device;
+	struct btrfs_fs_devices *fs_devices;
+	u64 found_transid = btrfs_super_generation(disk_super);
+
+	fs_devices = find_fsid(disk_super->fsid);
+	if (!fs_devices) {
+		fs_devices = kmalloc(sizeof(*fs_devices), GFP_NOFS);
+		if (!fs_devices)
+			return -ENOMEM;
+		INIT_LIST_HEAD(&fs_devices->devices);
+		list_add(&fs_devices->list, &fs_uuids);
+		memcpy(fs_devices->fsid, disk_super->fsid, BTRFS_FSID_SIZE);
+		fs_devices->latest_devid = devid;
+		fs_devices->latest_trans = found_transid;
+		fs_devices->lowest_devid = (u64)-1;
+		fs_devices->num_devices = 0;
+		device = NULL;
+	} else {
+		device = __find_device(&fs_devices->devices, devid);
+	}
+	if (!device) {
+		device = kzalloc(sizeof(*device), GFP_NOFS);
+		if (!device) {
+			/* we can safely leave the fs_devices entry around */
+			return -ENOMEM;
+		}
+		device->devid = devid;
+		device->name = kstrdup(path, GFP_NOFS);
+		if (!device->name) {
+			kfree(device);
+			return -ENOMEM;
+		}
+		list_add(&device->dev_list, &fs_devices->devices);
+		fs_devices->num_devices++;
+	}
+
+	if (found_transid > fs_devices->latest_trans) {
+		fs_devices->latest_devid = devid;
+		fs_devices->latest_trans = found_transid;
+	}
+	if (fs_devices->lowest_devid > devid) {
+		fs_devices->lowest_devid = devid;
+		printk("lowest devid now %Lu\n", devid);
+	}
+	*fs_devices_ret = fs_devices;
+	return 0;
+}
+
+int btrfs_close_devices(struct btrfs_fs_devices *fs_devices)
+{
+	struct list_head *head = &fs_devices->devices;
+	struct list_head *cur;
+	struct btrfs_device *device;
+
+	mutex_lock(&uuid_mutex);
+	list_for_each(cur, head) {
+		device = list_entry(cur, struct btrfs_device, dev_list);
+		if (device->bdev) {
+			close_bdev_excl(device->bdev);
+			printk("close devices closes %s\n", device->name);
+		}
+		device->bdev = NULL;
+	}
+	mutex_unlock(&uuid_mutex);
+	return 0;
+}
+
+int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
+		       int flags, void *holder)
+{
+	struct block_device *bdev;
+	struct list_head *head = &fs_devices->devices;
+	struct list_head *cur;
+	struct btrfs_device *device;
+	int ret;
+
+	mutex_lock(&uuid_mutex);
+	list_for_each(cur, head) {
+		device = list_entry(cur, struct btrfs_device, dev_list);
+		bdev = open_bdev_excl(device->name, flags, holder);
+printk("opening %s devid %Lu\n", device->name, device->devid);
+		if (IS_ERR(bdev)) {
+			printk("open %s failed\n", device->name);
+			ret = PTR_ERR(bdev);
+			goto fail;
+		}
+		if (device->devid == fs_devices->latest_devid)
+			fs_devices->latest_bdev = bdev;
+		if (device->devid == fs_devices->lowest_devid) {
+			fs_devices->lowest_bdev = bdev;
+printk("lowest bdev %s\n", device->name);
+		}
+		device->bdev = bdev;
+	}
+	mutex_unlock(&uuid_mutex);
+	return 0;
+fail:
+	mutex_unlock(&uuid_mutex);
+	btrfs_close_devices(fs_devices);
+	return ret;
+}
+
+int btrfs_scan_one_device(const char *path, int flags, void *holder,
+			  struct btrfs_fs_devices **fs_devices_ret)
+{
+	struct btrfs_super_block *disk_super;
+	struct block_device *bdev;
+	struct buffer_head *bh;
+	int ret;
+	u64 devid;
+
+	mutex_lock(&uuid_mutex);
+
+	printk("scan one opens %s\n", path);
+	bdev = open_bdev_excl(path, flags, holder);
+
+	if (IS_ERR(bdev)) {
+		printk("open failed\n");
+		ret = PTR_ERR(bdev);
+		goto error;
+	}
+
+	ret = set_blocksize(bdev, 4096);
+	if (ret)
+		goto error_close;
+	bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096);
+	if (!bh) {
+		ret = -EIO;
+		goto error_close;
+	}
+	disk_super = (struct btrfs_super_block *)bh->b_data;
+	if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
+	    sizeof(disk_super->magic))) {
+		printk("no btrfs found on %s\n", path);
+		ret = -ENOENT;
+		goto error_brelse;
+	}
+	devid = le64_to_cpu(disk_super->dev_item.devid);
+	printk("found device %Lu on %s\n", devid, path);
+	ret = device_list_add(path, disk_super, devid, fs_devices_ret);
+
+error_brelse:
+	brelse(bh);
+error_close:
+	close_bdev_excl(bdev);
+	printk("scan one closes bdev %s\n", path);
+error:
+	mutex_unlock(&uuid_mutex);
+	return ret;
+}
 
 /*
  * this uses a pretty simple search, the expectation is that it is
@@ -56,6 +266,10 @@ static int find_free_dev_extent(struct btrfs_trans_handle *trans,
 
 	/* FIXME use last free of some kind */
 
+	/* we don't want to overwrite the superblock on the drive,
+	 * so we make sure to start at an offset of at least 1MB
+	 */
+	search_start = max((u64)1024 * 1024, search_start);
 	key.objectid = device->devid;
 	key.offset = search_start;
 	key.type = BTRFS_DEV_EXTENT_KEY;
@@ -285,6 +499,7 @@ int btrfs_add_device(struct btrfs_trans_handle *trans,
 	leaf = path->nodes[0];
 	dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item);
 
+	device->devid = free_devid;
 	btrfs_set_device_id(leaf, dev_item, device->devid);
 	btrfs_set_device_type(leaf, dev_item, device->type);
 	btrfs_set_device_io_align(leaf, dev_item, device->io_align);
@@ -382,7 +597,7 @@ int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
 	struct btrfs_device *device = NULL;
 	struct btrfs_chunk *chunk;
 	struct list_head private_devs;
-	struct list_head *dev_list = &extent_root->fs_info->devices;
+	struct list_head *dev_list = &extent_root->fs_info->fs_devices->devices;
 	struct list_head *cur;
 	struct extent_map_tree *em_tree;
 	struct map_lookup *map;
@@ -449,7 +664,7 @@ again:
 					     key.objectid,
 					     calc_size, &dev_offset);
 		BUG_ON(ret);
-
+printk("alloc chunk size %Lu from dev %Lu\n", calc_size, device->devid);
 		device->bytes_used += calc_size;
 		ret = btrfs_update_device(trans, device);
 		BUG_ON(ret);
@@ -592,17 +807,9 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio)
 
 struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid)
 {
-	struct btrfs_device *dev;
-	struct list_head *cur = root->fs_info->devices.next;
-	struct list_head *head = &root->fs_info->devices;
+	struct list_head *head = &root->fs_info->fs_devices->devices;
 
-	while(cur != head) {
-		dev = list_entry(cur, struct btrfs_device, dev_list);
-		if (dev->devid == devid)
-			return dev;
-		cur = cur->next;
-	}
-	return NULL;
+	return __find_device(head, devid);
 }
 
 static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
@@ -699,15 +906,16 @@ static int read_one_dev(struct btrfs_root *root,
 	devid = btrfs_device_id(leaf, dev_item);
 	device = btrfs_find_device(root, devid);
 	if (!device) {
+		printk("warning devid %Lu not found already\n", devid);
 		device = kmalloc(sizeof(*device), GFP_NOFS);
 		if (!device)
 			return -ENOMEM;
-		list_add(&device->dev_list, &root->fs_info->devices);
+		list_add(&device->dev_list,
+			 &root->fs_info->fs_devices->devices);
 	}
 
 	fill_device_from_item(leaf, dev_item, device);
 	device->dev_root = root->fs_info->dev_root;
-	device->bdev = root->fs_info->sb->s_bdev;
 	ret = 0;
 #if 0
 	ret = btrfs_open_device(device);
diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h
index 20259128152e..12f297eb0559 100644
--- a/fs/btrfs/volumes.h
+++ b/fs/btrfs/volumes.h
@@ -24,6 +24,8 @@ struct btrfs_device {
 
 	struct block_device *bdev;
 
+	char *name;
+
 	/* the internal btrfs device id */
 	u64 devid;
 
@@ -49,6 +51,20 @@ struct btrfs_device {
 	u8 uuid[BTRFS_DEV_UUID_SIZE];
 };
 
+struct btrfs_fs_devices {
+	u8 fsid[BTRFS_FSID_SIZE]; /* FS specific uuid */
+
+	/* the device with this id has the most recent coyp of the super */
+	u64 latest_devid;
+	u64 latest_trans;
+	u64 lowest_devid;
+	u64 num_devices;
+	struct block_device *latest_bdev;
+	struct block_device *lowest_bdev;
+	struct list_head devices;
+	struct list_head list;
+};
+
 int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
 			   struct btrfs_device *device,
 			   u64 owner, u64 num_bytes, u64 *start);
@@ -67,4 +83,13 @@ int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf);
 int btrfs_map_block(struct btrfs_mapping_tree *map_tree,
 		    u64 logical, u64 *phys, u64 *length,
 		    struct btrfs_device **dev);
+int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
+		       int flags, void *holder);
+int btrfs_scan_one_device(const char *path, int flags, void *holder,
+			  struct btrfs_fs_devices **fs_devices_ret);
+int btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
+int btrfs_add_device(struct btrfs_trans_handle *trans,
+		     struct btrfs_root *root,
+		     struct btrfs_device *device);
+int btrfs_cleanup_fs_uuids(void);
 #endif
-- 
cgit v1.2.3


From 0999df54f850fe1aba29b10d5c869493af107478 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Tue, 1 Apr 2008 13:48:14 -0400
Subject: Btrfs: Verify checksums on tree blocks found without read_tree_block

Checksums were only verified by btrfs_read_tree_block, which meant the
functions to probe the page cache for blocks were not validating checksums.
Normally this is fine because the buffers will only be in cache if they
have already been validated.

But, there is a window while the buffer is being read from disk where
it could be up to date in the cache but not yet verified.  This patch
makes sure all buffers go through checksum verification before they
are used.

This is safer, and it prevents modification of buffers before they go
through the csum code.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.c       |  5 +++
 fs/btrfs/disk-io.c     | 86 +++++++++++++++++++++++++++++---------------------
 fs/btrfs/disk-io.h     |  2 ++
 fs/btrfs/extent-tree.c |  2 ++
 fs/btrfs/tree-defrag.c |  1 +
 5 files changed, 60 insertions(+), 36 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 6ba5394834be..df090bf2eec0 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -158,6 +158,8 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 	} else {
 		root_gen = 0;
 	}
+	if (!(buf->flags & EXTENT_CSUM))
+		WARN_ON(1);
 
 	WARN_ON(root->ref_cows && trans->transid !=
 		root->fs_info->running_transaction->transid);
@@ -245,6 +247,8 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		       root->fs_info->generation);
 		WARN_ON(1);
 	}
+	if (!(buf->flags & EXTENT_CSUM))
+		WARN_ON(1);
 
 	header_trans = btrfs_header_generation(buf);
 	spin_lock(&root->fs_info->hash_lock);
@@ -396,6 +400,7 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 		if (search_start == 0)
 			search_start = last_block;
 
+		btrfs_verify_block_csum(root, cur);
 		err = __btrfs_cow_block(trans, root, cur, parent, i,
 					&tmp, search_start,
 					min(16 * blocksize,
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 5547607681f4..e40fb318ad99 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -46,27 +46,6 @@ static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
 
 static struct extent_io_ops btree_extent_io_ops;
 
-struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
-					    u64 bytenr, u32 blocksize)
-{
-	struct inode *btree_inode = root->fs_info->btree_inode;
-	struct extent_buffer *eb;
-	eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
-				bytenr, blocksize, GFP_NOFS);
-	return eb;
-}
-
-struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
-						 u64 bytenr, u32 blocksize)
-{
-	struct inode *btree_inode = root->fs_info->btree_inode;
-	struct extent_buffer *eb;
-
-	eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
-				 bytenr, blocksize, NULL, GFP_NOFS);
-	return eb;
-}
-
 struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
 				    size_t page_offset, u64 start, u64 len,
 				    int create)
@@ -380,36 +359,29 @@ static int close_all_devices(struct btrfs_fs_info *fs_info)
 	return 0;
 }
 
-struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
-				      u32 blocksize)
+int btrfs_verify_block_csum(struct btrfs_root *root,
+			    struct extent_buffer *buf)
 {
-	struct extent_buffer *buf = NULL;
-	struct inode *btree_inode = root->fs_info->btree_inode;
 	struct extent_io_tree *io_tree;
 	u64 end;
 	int ret;
 
-	io_tree = &BTRFS_I(btree_inode)->io_tree;
-
-	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
-	if (!buf)
-		return NULL;
-	read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, buf, 0, 1,
-				 btree_get_extent);
-
+	io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
 	if (buf->flags & EXTENT_CSUM)
-		return buf;
+		return 0;
 
-	end = buf->start + PAGE_CACHE_SIZE - 1;
+	end = min_t(u64, buf->len, PAGE_CACHE_SIZE);
+	end = buf->start + end - 1;
 	if (test_range_bit(io_tree, buf->start, end, EXTENT_CSUM, 1)) {
 		buf->flags |= EXTENT_CSUM;
-		return buf;
+		return 0;
 	}
 
 	lock_extent(io_tree, buf->start, end, GFP_NOFS);
 
 	if (test_range_bit(io_tree, buf->start, end, EXTENT_CSUM, 1)) {
 		buf->flags |= EXTENT_CSUM;
+		ret = 0;
 		goto out_unlock;
 	}
 
@@ -419,6 +391,48 @@ struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
 
 out_unlock:
 	unlock_extent(io_tree, buf->start, end, GFP_NOFS);
+	return ret;
+}
+
+struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
+					    u64 bytenr, u32 blocksize)
+{
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	struct extent_buffer *eb;
+	eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
+				bytenr, blocksize, GFP_NOFS);
+	return eb;
+}
+
+struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
+						 u64 bytenr, u32 blocksize)
+{
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	struct extent_buffer *eb;
+
+	eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
+				 bytenr, blocksize, NULL, GFP_NOFS);
+	return eb;
+}
+
+
+struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
+				      u32 blocksize)
+{
+	struct extent_buffer *buf = NULL;
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	struct extent_io_tree *io_tree;
+	int ret;
+
+	io_tree = &BTRFS_I(btree_inode)->io_tree;
+
+	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
+	if (!buf)
+		return NULL;
+	read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, buf, 0, 1,
+				 btree_get_extent);
+
+	ret = btrfs_verify_block_csum(root, buf);
 	return buf;
 }
 
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index b7cbc58a5553..05b88d0e75eb 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -69,4 +69,6 @@ u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len);
 void btrfs_csum_final(u32 crc, char *result);
 void btrfs_throttle(struct btrfs_root *root);
 int btrfs_open_device(struct btrfs_device *dev);
+int btrfs_verify_block_csum(struct btrfs_root *root,
+			    struct extent_buffer *buf);
 #endif
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index cf283b0271ac..a34c289aec21 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -2069,6 +2069,8 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans,
 				BUG_ON(ret);
 				continue;
 			}
+		} else if (next) {
+			btrfs_verify_block_csum(root, next);
 		}
 		WARN_ON(*level <= 0);
 		if (path->nodes[*level-1])
diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c
index 5935cbd8f2b8..256af1870eef 100644
--- a/fs/btrfs/tree-defrag.c
+++ b/fs/btrfs/tree-defrag.c
@@ -101,6 +101,7 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 				path->slots[*level]++;
 				continue;
 			}
+			btrfs_verify_block_csum(root, next);
 		} else {
 			next = read_tree_block(root, bytenr,
 				       btrfs_level_size(root, *level - 1));
-- 
cgit v1.2.3


From 22c599485b1fdd95e4476a4752596a6cf6c6629a Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Wed, 9 Apr 2008 16:28:12 -0400
Subject: Btrfs: Handle data block end_io through the async work queue

Before it was done by the bio end_io routine, the work queue code is able
to scale much better with faster IO subsystems.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/disk-io.c | 34 +++++++++++++++++++++++-----------
 fs/btrfs/disk-io.h |  2 ++
 fs/btrfs/inode.c   |  4 +++-
 3 files changed, 28 insertions(+), 12 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 82109204788c..94ff87d0eae4 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -54,6 +54,7 @@ struct end_io_wq {
 	void *private;
 	struct btrfs_fs_info *info;
 	int error;
+	int metadata;
 	struct list_head list;
 };
 
@@ -308,29 +309,40 @@ static int end_workqueue_bio(struct bio *bio,
 #endif
 }
 
-static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio)
+int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
+			int metadata)
 {
-	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct end_io_wq *end_io_wq;
-	u64 offset;
-	offset = bio->bi_sector << 9;
-
-	if (rw & (1 << BIO_RW)) {
-		return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio);
-	}
-
 	end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
 	if (!end_io_wq)
 		return -ENOMEM;
 
 	end_io_wq->private = bio->bi_private;
 	end_io_wq->end_io = bio->bi_end_io;
-	end_io_wq->info = root->fs_info;
+	end_io_wq->info = info;
 	end_io_wq->error = 0;
 	end_io_wq->bio = bio;
+	end_io_wq->metadata = metadata;
 
 	bio->bi_private = end_io_wq;
 	bio->bi_end_io = end_workqueue_bio;
+	return 0;
+}
+
+static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	u64 offset;
+	int ret;
+
+	offset = bio->bi_sector << 9;
+
+	if (rw & (1 << BIO_RW)) {
+		return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio);
+	}
+
+	ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1);
+	BUG_ON(ret);
 
 	if (offset == BTRFS_SUPER_INFO_OFFSET) {
 		bio->bi_bdev = root->fs_info->sb->s_bdev;
@@ -880,7 +892,7 @@ void btrfs_end_io_csum(struct work_struct *work)
 		end_io_wq = list_entry(next, struct end_io_wq, list);
 
 		bio = end_io_wq->bio;
-		if (!bio_ready_for_csum(bio)) {
+		if (end_io_wq->metadata && !bio_ready_for_csum(bio)) {
 			spin_lock_irqsave(&fs_info->end_io_work_lock, flags);
 			was_empty = list_empty(&fs_info->end_io_work_list);
 			list_add_tail(&end_io_wq->list,
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 05b88d0e75eb..4fac0ccbf8f8 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -71,4 +71,6 @@ void btrfs_throttle(struct btrfs_root *root);
 int btrfs_open_device(struct btrfs_device *dev);
 int btrfs_verify_block_csum(struct btrfs_root *root,
 			    struct extent_buffer *buf);
+int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
+			int metadata);
 #endif
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 7ae677d8a6de..e1ef1acdb350 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -328,7 +328,9 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio)
 	struct btrfs_trans_handle *trans;
 	int ret = 0;
 
-	if (rw != WRITE) {
+	if (!(rw & (1 << BIO_RW))) {
+		ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
+		BUG_ON(ret);
 		goto mapit;
 	}
 
-- 
cgit v1.2.3


From f29844623de29a12358d7fba35d0959465b64adf Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Thu, 10 Apr 2008 16:19:33 -0400
Subject: Btrfs: Write out all super blocks on commit, and bring back proper
 barrier support

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/disk-io.c | 118 ++++++++++++++++++++++++++++++++++++++++++++++++++---
 fs/btrfs/disk-io.h |   1 +
 fs/btrfs/volumes.c |   8 ++--
 fs/btrfs/volumes.h |   3 ++
 4 files changed, 122 insertions(+), 8 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 59bdf0474be3..cf1de75f088a 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -382,7 +382,7 @@ static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
 	BUG_ON(ret);
 
 	if (offset == BTRFS_SUPER_INFO_OFFSET) {
-		bio->bi_bdev = root->fs_info->sb->s_bdev;
+		bio->bi_bdev = root->fs_info->fs_devices->latest_bdev;
 		submit_bio(rw, bio);
 		return 0;
 	}
@@ -988,7 +988,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	spin_lock_init(&fs_info->new_trans_lock);
 
 	init_completion(&fs_info->kobj_unregister);
-	sb_set_blocksize(sb, 4096);
+	sb_set_blocksize(sb, BTRFS_SUPER_INFO_SIZE);
 	fs_info->tree_root = tree_root;
 	fs_info->extent_root = extent_root;
 	fs_info->chunk_root = chunk_root;
@@ -1169,14 +1169,121 @@ fail:
 	return ERR_PTR(err);
 }
 
+static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
+{
+	char b[BDEVNAME_SIZE];
+
+	if (uptodate) {
+		set_buffer_uptodate(bh);
+	} else {
+		if (!buffer_eopnotsupp(bh) && printk_ratelimit()) {
+			printk(KERN_WARNING "lost page write due to "
+					"I/O error on %s\n",
+				       bdevname(bh->b_bdev, b));
+		}
+		set_buffer_write_io_error(bh);
+		clear_buffer_uptodate(bh);
+	}
+	unlock_buffer(bh);
+	put_bh(bh);
+}
+
+int write_all_supers(struct btrfs_root *root)
+{
+	struct list_head *cur;
+	struct list_head *head = &root->fs_info->fs_devices->devices;
+	struct btrfs_device *dev;
+	struct extent_buffer *sb;
+	struct btrfs_dev_item *dev_item;
+	struct buffer_head *bh;
+	int ret;
+	int do_barriers;
+
+	do_barriers = !btrfs_test_opt(root, NOBARRIER);
+
+	sb = root->fs_info->sb_buffer;
+	dev_item = (struct btrfs_dev_item *)offsetof(struct btrfs_super_block,
+						      dev_item);
+	list_for_each(cur, head) {
+		dev = list_entry(cur, struct btrfs_device, dev_list);
+		btrfs_set_device_type(sb, dev_item, dev->type);
+		btrfs_set_device_id(sb, dev_item, dev->devid);
+		btrfs_set_device_total_bytes(sb, dev_item, dev->total_bytes);
+		btrfs_set_device_bytes_used(sb, dev_item, dev->bytes_used);
+		btrfs_set_device_io_align(sb, dev_item, dev->io_align);
+		btrfs_set_device_io_width(sb, dev_item, dev->io_width);
+		btrfs_set_device_sector_size(sb, dev_item, dev->sector_size);
+		write_extent_buffer(sb, dev->uuid,
+				    (unsigned long)btrfs_device_uuid(dev_item),
+				    BTRFS_DEV_UUID_SIZE);
+
+		btrfs_set_header_flag(sb, BTRFS_HEADER_FLAG_WRITTEN);
+		csum_tree_block(root, sb, 0);
+
+		bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET /
+			      root->fs_info->sb->s_blocksize,
+			      BTRFS_SUPER_INFO_SIZE);
+
+		read_extent_buffer(sb, bh->b_data, 0, BTRFS_SUPER_INFO_SIZE);
+		dev->pending_io = bh;
+
+		get_bh(bh);
+		set_buffer_uptodate(bh);
+		lock_buffer(bh);
+		bh->b_end_io = btrfs_end_buffer_write_sync;
+
+		if (do_barriers && dev->barriers) {
+			ret = submit_bh(WRITE_BARRIER, bh);
+			if (ret == -EOPNOTSUPP) {
+				printk("btrfs: disabling barriers on dev %s\n",
+				       dev->name);
+				set_buffer_uptodate(bh);
+				dev->barriers = 0;
+				get_bh(bh);
+				lock_buffer(bh);
+				ret = submit_bh(WRITE, bh);
+			}
+		} else {
+			ret = submit_bh(WRITE, bh);
+		}
+		BUG_ON(ret);
+	}
+
+	list_for_each(cur, head) {
+		dev = list_entry(cur, struct btrfs_device, dev_list);
+		BUG_ON(!dev->pending_io);
+		bh = dev->pending_io;
+		wait_on_buffer(bh);
+		if (!buffer_uptodate(dev->pending_io)) {
+			if (do_barriers && dev->barriers) {
+				printk("btrfs: disabling barriers on dev %s\n",
+				       dev->name);
+				set_buffer_uptodate(bh);
+				get_bh(bh);
+				lock_buffer(bh);
+				dev->barriers = 0;
+				ret = submit_bh(WRITE, bh);
+				BUG_ON(ret);
+				wait_on_buffer(bh);
+				BUG_ON(!buffer_uptodate(bh));
+			} else {
+				BUG();
+			}
+
+		}
+		dev->pending_io = NULL;
+		brelse(bh);
+	}
+	return 0;
+}
+
 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root)
 {
 	int ret;
-	struct extent_buffer *super = root->fs_info->sb_buffer;
-	struct inode *btree_inode = root->fs_info->btree_inode;
-	struct super_block *sb = root->fs_info->sb;
 
+	ret = write_all_supers(root);
+#if 0
 	if (!btrfs_test_opt(root, NOBARRIER))
 		blkdev_issue_flush(sb->s_bdev, NULL);
 	set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, super);
@@ -1184,6 +1291,7 @@ int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
 				     super->start, super->len);
 	if (!btrfs_test_opt(root, NOBARRIER))
 		blkdev_issue_flush(sb->s_bdev, NULL);
+#endif
 	return ret;
 }
 
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 4fac0ccbf8f8..60b01902db79 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -20,6 +20,7 @@
 #define __DISKIO__
 
 #define BTRFS_SUPER_INFO_OFFSET (16 * 1024)
+#define BTRFS_SUPER_INFO_SIZE 4096
 struct btrfs_device;
 struct btrfs_fs_devices;
 
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 3b927f698320..07d43553141c 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -125,6 +125,7 @@ static int device_list_add(const char *path,
 			return -ENOMEM;
 		}
 		device->devid = devid;
+		device->barriers = 1;
 		device->name = kstrdup(path, GFP_NOFS);
 		if (!device->name) {
 			kfree(device);
@@ -208,6 +209,7 @@ int btrfs_scan_one_device(const char *path, int flags, void *holder,
 	struct buffer_head *bh;
 	int ret;
 	u64 devid;
+	u64 transid;
 
 	mutex_lock(&uuid_mutex);
 
@@ -236,14 +238,14 @@ int btrfs_scan_one_device(const char *path, int flags, void *holder,
 		goto error_brelse;
 	}
 	devid = le64_to_cpu(disk_super->dev_item.devid);
-	printk("found device %Lu on %s\n", devid, path);
+	transid = btrfs_super_generation(disk_super);
+	printk("found device %Lu transid %Lu on %s\n", devid, transid, path);
 	ret = device_list_add(path, disk_super, devid, fs_devices_ret);
 
 error_brelse:
 	brelse(bh);
 error_close:
 	close_bdev_excl(bdev);
-	printk("scan one closes bdev %s\n", path);
 error:
 	mutex_unlock(&uuid_mutex);
 	return ret;
@@ -1143,7 +1145,7 @@ static int read_one_dev(struct btrfs_root *root,
 	device = btrfs_find_device(root, devid);
 	if (!device) {
 		printk("warning devid %Lu not found already\n", devid);
-		device = kmalloc(sizeof(*device), GFP_NOFS);
+		device = kzalloc(sizeof(*device), GFP_NOFS);
 		if (!device)
 			return -ENOMEM;
 		list_add(&device->dev_list,
diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h
index 3d5d0a9cb827..89548837a1cc 100644
--- a/fs/btrfs/volumes.h
+++ b/fs/btrfs/volumes.h
@@ -21,9 +21,12 @@
 
 #include <linux/bio.h>
 
+struct buffer_head;
 struct btrfs_device {
 	struct list_head dev_list;
 	struct btrfs_root *dev_root;
+	struct buffer_head *pending_io;
+	int barriers;
 	spinlock_t io_lock;
 
 	struct block_device *bdev;
-- 
cgit v1.2.3


From 44b8bd7edda4f63de180d0f7325c9fb704b3806b Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Wed, 16 Apr 2008 11:14:51 -0400
Subject: Btrfs: Create a work queue for bio writes

This allows checksumming to happen in parallel among many cpus, and
keeps us from bogging down pdflush with the checksumming code.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.h     |  3 ++
 fs/btrfs/disk-io.c   | 93 ++++++++++++++++++++++++++++++++++++++++++++++++++--
 fs/btrfs/disk-io.h   |  3 ++
 fs/btrfs/extent_io.h |  5 +--
 fs/btrfs/inode.c     | 28 +++++++++++-----
 5 files changed, 119 insertions(+), 13 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index a22edcf49174..ff15b8513f90 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -514,8 +514,11 @@ struct btrfs_fs_info {
 	struct list_head hashers;
 	struct list_head dead_roots;
 	struct list_head end_io_work_list;
+	struct list_head async_submit_work_list;
 	struct work_struct end_io_work;
+	struct work_struct async_submit_work;
 	spinlock_t end_io_work_lock;
+	spinlock_t async_submit_work_lock;
 
 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
 	struct work_struct trans_work;
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 79c284c87286..9e41ea93ebce 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -47,6 +47,7 @@ static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
 
 static struct extent_io_ops btree_extent_io_ops;
 static struct workqueue_struct *end_io_workqueue;
+static struct workqueue_struct *async_submit_workqueue;
 
 struct end_io_wq {
 	struct bio *bio;
@@ -58,6 +59,15 @@ struct end_io_wq {
 	struct list_head list;
 };
 
+struct async_submit_bio {
+	struct inode *inode;
+	struct bio *bio;
+	struct list_head list;
+	extent_submit_bio_hook_t *submit_bio_hook;
+	int rw;
+	int mirror_num;
+};
+
 struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
 				    size_t page_offset, u64 start, u64 len,
 				    int create)
@@ -365,7 +375,31 @@ int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
 	return 0;
 }
 
-static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
+int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
+			int rw, struct bio *bio, int mirror_num,
+			extent_submit_bio_hook_t *submit_bio_hook)
+{
+	struct async_submit_bio *async;
+
+	async = kmalloc(sizeof(*async), GFP_NOFS);
+	if (!async)
+		return -ENOMEM;
+
+	async->inode = inode;
+	async->rw = rw;
+	async->bio = bio;
+	async->mirror_num = mirror_num;
+	async->submit_bio_hook = submit_bio_hook;
+
+	spin_lock(&fs_info->async_submit_work_lock);
+	list_add_tail(&async->list, &fs_info->async_submit_work_list);
+	spin_unlock(&fs_info->async_submit_work_lock);
+
+	queue_work(async_submit_workqueue, &fs_info->async_submit_work);
+	return 0;
+}
+
+static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
 				 int mirror_num)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
@@ -389,6 +423,17 @@ static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
 	return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num);
 }
 
+static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
+				 int mirror_num)
+{
+	if (!(rw & (1 << BIO_RW))) {
+		return __btree_submit_bio_hook(inode, rw, bio, mirror_num);
+	}
+	return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
+				   inode, rw, bio, mirror_num,
+				   __btree_submit_bio_hook);
+}
+
 static int btree_writepage(struct page *page, struct writeback_control *wbc)
 {
 	struct extent_io_tree *tree;
@@ -903,9 +948,9 @@ static int bio_ready_for_csum(struct bio *bio)
 }
 
 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
-void btrfs_end_io_csum(void *p)
+static void btrfs_end_io_csum(void *p)
 #else
-void btrfs_end_io_csum(struct work_struct *work)
+static void btrfs_end_io_csum(struct work_struct *work)
 #endif
 {
 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
@@ -959,6 +1004,39 @@ void btrfs_end_io_csum(struct work_struct *work)
 	}
 }
 
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
+static void btrfs_async_submit_work(void *p)
+#else
+static void btrfs_async_submit_work(struct work_struct *work)
+#endif
+{
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
+	struct btrfs_fs_info *fs_info = p;
+#else
+	struct btrfs_fs_info *fs_info = container_of(work,
+						     struct btrfs_fs_info,
+						     async_submit_work);
+#endif
+	struct async_submit_bio *async;
+	struct list_head *next;
+
+	while(1) {
+		spin_lock(&fs_info->async_submit_work_lock);
+		if (list_empty(&fs_info->async_submit_work_list)) {
+			spin_unlock(&fs_info->async_submit_work_lock);
+			return;
+		}
+		next = fs_info->async_submit_work_list.next;
+		list_del(next);
+		spin_unlock(&fs_info->async_submit_work_lock);
+
+		async = list_entry(next, struct async_submit_bio, list);
+		async->submit_bio_hook(async->inode, async->rw, async->bio,
+				       async->mirror_num);
+		kfree(async);
+	}
+}
+
 struct btrfs_root *open_ctree(struct super_block *sb,
 			      struct btrfs_fs_devices *fs_devices)
 {
@@ -987,14 +1065,17 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	}
 	end_io_workqueue = create_workqueue("btrfs-end-io");
 	BUG_ON(!end_io_workqueue);
+	async_submit_workqueue = create_workqueue("btrfs-async-submit");
 
 	INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
 	INIT_LIST_HEAD(&fs_info->trans_list);
 	INIT_LIST_HEAD(&fs_info->dead_roots);
 	INIT_LIST_HEAD(&fs_info->hashers);
 	INIT_LIST_HEAD(&fs_info->end_io_work_list);
+	INIT_LIST_HEAD(&fs_info->async_submit_work_list);
 	spin_lock_init(&fs_info->hash_lock);
 	spin_lock_init(&fs_info->end_io_work_lock);
+	spin_lock_init(&fs_info->async_submit_work_lock);
 	spin_lock_init(&fs_info->delalloc_lock);
 	spin_lock_init(&fs_info->new_trans_lock);
 
@@ -1041,9 +1122,12 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 
 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
 	INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum, fs_info);
+	INIT_WORK(&fs_info->async_submit_work, btrfs_async_submit_work,
+		  fs_info);
 	INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info);
 #else
 	INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum);
+	INIT_WORK(&fs_info->async_submit_work, btrfs_async_submit_work);
 	INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
 #endif
 	BTRFS_I(fs_info->btree_inode)->root = tree_root;
@@ -1403,6 +1487,9 @@ int close_ctree(struct btrfs_root *root)
 	flush_workqueue(end_io_workqueue);
 	destroy_workqueue(end_io_workqueue);
 
+	flush_workqueue(async_submit_workqueue);
+	destroy_workqueue(async_submit_workqueue);
+
 	iput(fs_info->btree_inode);
 #if 0
 	while(!list_empty(&fs_info->hashers)) {
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 60b01902db79..5d5f6979c41c 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -74,4 +74,7 @@ int btrfs_verify_block_csum(struct btrfs_root *root,
 			    struct extent_buffer *buf);
 int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
 			int metadata);
+int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
+			int rw, struct bio *bio, int mirror_num,
+			extent_submit_bio_hook_t *submit_bio_hook);
 #endif
diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h
index b47859ccd78a..232300d13be2 100644
--- a/fs/btrfs/extent_io.h
+++ b/fs/btrfs/extent_io.h
@@ -24,11 +24,12 @@
 
 struct extent_state;
 
+typedef	int (extent_submit_bio_hook_t)(struct inode *inode, int rw,
+				       struct bio *bio, int mirror_num);
 struct extent_io_ops {
 	int (*fill_delalloc)(struct inode *inode, u64 start, u64 end);
 	int (*writepage_io_hook)(struct page *page, u64 start, u64 end);
-	int (*submit_bio_hook)(struct inode *inode, int rw, struct bio *bio,
-			       int mirror_num);
+	extent_submit_bio_hook_t *submit_bio_hook;
 	int (*merge_bio_hook)(struct page *page, unsigned long offset,
 			      size_t size, struct bio *bio);
 	int (*readpage_io_hook)(struct page *page, u64 start, u64 end);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 7869c0157a1f..457daf725d29 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -324,13 +324,29 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
 	return 0;
 }
 
-int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
+int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
 			  int mirror_num)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_trans_handle *trans;
 	int ret = 0;
 
+	mutex_lock(&root->fs_info->fs_mutex);
+	trans = btrfs_start_transaction(root, 1);
+	btrfs_set_trans_block_group(trans, inode);
+	btrfs_csum_file_blocks(trans, root, inode, bio);
+	ret = btrfs_end_transaction(trans, root);
+	BUG_ON(ret);
+	mutex_unlock(&root->fs_info->fs_mutex);
+	return btrfs_map_bio(root, rw, bio, mirror_num);
+}
+
+int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
+			  int mirror_num)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	int ret = 0;
+
 	if (!(rw & (1 << BIO_RW))) {
 		ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
 		BUG_ON(ret);
@@ -342,13 +358,9 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
 		goto mapit;
 	}
 
-	mutex_lock(&root->fs_info->fs_mutex);
-	trans = btrfs_start_transaction(root, 1);
-	btrfs_set_trans_block_group(trans, inode);
-	btrfs_csum_file_blocks(trans, root, inode, bio);
-	ret = btrfs_end_transaction(trans, root);
-	BUG_ON(ret);
-	mutex_unlock(&root->fs_info->fs_mutex);
+	return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
+				   inode, rw, bio, mirror_num,
+				   __btrfs_submit_bio_hook);
 mapit:
 	return btrfs_map_bio(root, rw, bio, mirror_num);
 }
-- 
cgit v1.2.3


From ca7a79ad8dbe24669a8ced01f9fc0126b872745d Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Mon, 12 May 2008 12:59:19 -0400
Subject: Btrfs: Pass down the expected generation number when reading tree
 blocks

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.c       | 55 +++++++++++++++++++-------------------------------
 fs/btrfs/disk-io.c     | 30 ++++++++++++---------------
 fs/btrfs/disk-io.h     |  7 ++++---
 fs/btrfs/extent-tree.c | 15 ++++++++------
 fs/btrfs/print-tree.c  |  3 ++-
 fs/btrfs/tree-defrag.c |  9 ++++++---
 6 files changed, 55 insertions(+), 64 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 29e5674b043d..289d71d8653a 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -302,6 +302,7 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 	struct extent_buffer *cur;
 	struct extent_buffer *tmp;
 	u64 blocknr;
+	u64 gen;
 	u64 search_start = *last_ret;
 	u64 last_block = 0;
 	u64 other;
@@ -354,6 +355,7 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 
 		progress_passed = 1;
 		blocknr = btrfs_node_blockptr(parent, i);
+		gen = btrfs_node_ptr_generation(parent, i);
 		if (last_block == 0)
 			last_block = blocknr;
 
@@ -387,15 +389,14 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 			}
 			if (!cur) {
 				cur = read_tree_block(root, blocknr,
-							 blocksize);
+							 blocksize, gen);
 			} else if (!uptodate) {
-				btrfs_read_buffer(cur);
+				btrfs_read_buffer(cur, gen);
 			}
 		}
 		if (search_start == 0)
 			search_start = last_block;
 
-		btrfs_verify_block_csum(root, cur);
 		err = __btrfs_cow_block(trans, root, cur, parent, i,
 					&tmp, search_start,
 					min(16 * blocksize,
@@ -696,12 +697,17 @@ static int bin_search(struct extent_buffer *eb, struct btrfs_key *key,
 static struct extent_buffer *read_node_slot(struct btrfs_root *root,
 				   struct extent_buffer *parent, int slot)
 {
+	int level = btrfs_header_level(parent);
 	if (slot < 0)
 		return NULL;
 	if (slot >= btrfs_header_nritems(parent))
 		return NULL;
+
+	BUG_ON(level == 0);
+
 	return read_tree_block(root, btrfs_node_blockptr(parent, slot),
-		       btrfs_level_size(root, btrfs_header_level(parent) - 1));
+		       btrfs_level_size(root, level - 1),
+		       btrfs_node_ptr_generation(parent, slot));
 }
 
 static int balance_level(struct btrfs_trans_handle *trans,
@@ -1076,7 +1082,8 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
 		if ((search >= lowest_read && search <= highest_read) ||
 		    (search < lowest_read && lowest_read - search <= 32768) ||
 		    (search > highest_read && search - highest_read <= 32768)) {
-			readahead_tree_block(root, search, blocksize);
+			readahead_tree_block(root, search, blocksize,
+				     btrfs_node_ptr_generation(node, nr));
 			nread += blocksize;
 		}
 		nscan++;
@@ -1109,8 +1116,6 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
 		      ins_len, int cow)
 {
 	struct extent_buffer *b;
-	u64 bytenr;
-	u64 ptr_gen;
 	int slot;
 	int ret;
 	int level;
@@ -1174,20 +1179,12 @@ again:
 			/* this is only true while dropping a snapshot */
 			if (level == lowest_level)
 				break;
-			bytenr = btrfs_node_blockptr(b, slot);
-			ptr_gen = btrfs_node_ptr_generation(b, slot);
+
 			if (should_reada)
 				reada_for_search(root, p, level, slot,
 						 key->objectid);
-			b = read_tree_block(root, bytenr,
-					    btrfs_level_size(root, level - 1));
-			if (ptr_gen != btrfs_header_generation(b)) {
-				printk("block %llu bad gen wanted %llu "
-				       "found %llu\n",
-			        (unsigned long long)b->start,
-				(unsigned long long)ptr_gen,
-			        (unsigned long long)btrfs_header_generation(b));
-			}
+
+			b = read_node_slot(root, b, slot);
 		} else {
 			p->slots[level] = slot;
 			if (ins_len > 0 && btrfs_leaf_free_space(root, b) <
@@ -1650,8 +1647,7 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 	if (slot >= btrfs_header_nritems(upper) - 1)
 		return 1;
 
-	right = read_tree_block(root, btrfs_node_blockptr(upper, slot + 1),
-				root->leafsize);
+	right = read_node_slot(root, upper, slot + 1);
 	free_space = btrfs_leaf_free_space(root, right);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
 		free_extent_buffer(right);
@@ -1826,8 +1822,7 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 		return 1;
 	}
 
-	left = read_tree_block(root, btrfs_node_blockptr(path->nodes[1],
-			       slot - 1), root->leafsize);
+	left = read_node_slot(root, path->nodes[1], slot - 1);
 	free_space = btrfs_leaf_free_space(root, left);
 	if (free_space < data_size + sizeof(struct btrfs_item)) {
 		free_extent_buffer(left);
@@ -2742,7 +2737,6 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root,
  */
 int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path)
 {
-	u64 bytenr;
 	int slot;
 	int level = 1;
 	struct extent_buffer *c;
@@ -2762,12 +2756,10 @@ int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path)
 		}
 		slot--;
 
-		bytenr = btrfs_node_blockptr(c, slot);
 		if (next)
 			free_extent_buffer(next);
 
-		next = read_tree_block(root, bytenr,
-				       btrfs_level_size(root, level - 1));
+		next = read_node_slot(root, c, slot);
 		break;
 	}
 	path->slots[level] = slot;
@@ -2782,8 +2774,7 @@ int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path)
 		path->slots[level] = slot;
 		if (!level)
 			break;
-		next = read_tree_block(root, btrfs_node_blockptr(next, slot),
-				       btrfs_level_size(root, level - 1));
+		next = read_node_slot(root, next, slot);
 	}
 	return 0;
 }
@@ -2797,7 +2788,6 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 {
 	int slot;
 	int level = 1;
-	u64 bytenr;
 	struct extent_buffer *c;
 	struct extent_buffer *next = NULL;
 
@@ -2814,15 +2804,13 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 			continue;
 		}
 
-		bytenr = btrfs_node_blockptr(c, slot);
 		if (next)
 			free_extent_buffer(next);
 
 		if (path->reada)
 			reada_for_search(root, path, level, slot, 0);
 
-		next = read_tree_block(root, bytenr,
-				       btrfs_level_size(root, level -1));
+		next = read_node_slot(root, c, slot);
 		break;
 	}
 	path->slots[level] = slot;
@@ -2836,8 +2824,7 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 			break;
 		if (path->reada)
 			reada_for_search(root, path, level, 0, 0);
-		next = read_tree_block(root, btrfs_node_blockptr(next, 0),
-				       btrfs_level_size(root, level - 1));
+		next = read_node_slot(root, next, 0);
 	}
 	return 0;
 }
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index c3083165b172..edee7a44f861 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -207,7 +207,7 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
 
 static int btree_read_extent_buffer_pages(struct btrfs_root *root,
 					  struct extent_buffer *eb,
-					  u64 start)
+					  u64 start, u64 parent_transid)
 {
 	struct extent_io_tree *io_tree;
 	int ret;
@@ -254,7 +254,8 @@ int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
 		WARN_ON(1);
 	}
 	eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
-	ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE);
+	ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE,
+					     btrfs_header_generation(eb));
 	BUG_ON(ret);
 	btrfs_clear_buffer_defrag(eb);
 	found_start = btrfs_header_bytenr(eb);
@@ -562,7 +563,8 @@ static struct address_space_operations btree_aops = {
 	.sync_page	= block_sync_page,
 };
 
-int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize)
+int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
+			 u64 parent_transid)
 {
 	struct extent_buffer *buf = NULL;
 	struct inode *btree_inode = root->fs_info->btree_inode;
@@ -592,12 +594,6 @@ static int close_all_devices(struct btrfs_fs_info *fs_info)
 	return 0;
 }
 
-int btrfs_verify_block_csum(struct btrfs_root *root,
-			    struct extent_buffer *buf)
-{
-	return btrfs_buffer_uptodate(buf);
-}
-
 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
 					    u64 bytenr, u32 blocksize)
 {
@@ -621,7 +617,7 @@ struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
 
 
 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
-				      u32 blocksize)
+				      u32 blocksize, u64 parent_transid)
 {
 	struct extent_buffer *buf = NULL;
 	struct inode *btree_inode = root->fs_info->btree_inode;
@@ -634,7 +630,7 @@ struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
 	if (!buf)
 		return NULL;
 
-	ret = btree_read_extent_buffer_pages(root, buf, 0);
+	ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
 
 	if (ret == 0) {
 		buf->flags |= EXTENT_UPTODATE;
@@ -715,7 +711,7 @@ static int find_and_setup_root(struct btrfs_root *tree_root,
 
 	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
 	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
-				     blocksize);
+				     blocksize, 0);
 	BUG_ON(!root->node);
 	return 0;
 }
@@ -771,7 +767,7 @@ out:
 	}
 	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
 	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
-				     blocksize);
+				     blocksize, 0);
 	BUG_ON(!root->node);
 insert:
 	root->ref_cows = 1;
@@ -1288,7 +1284,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 
 	chunk_root->node = read_tree_block(chunk_root,
 					   btrfs_super_chunk_root(disk_super),
-					   blocksize);
+					   blocksize, 0);
 	BUG_ON(!chunk_root->node);
 
 	read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
@@ -1304,7 +1300,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 
 	tree_root->node = read_tree_block(tree_root,
 					  btrfs_super_root(disk_super),
-					  blocksize);
+					  blocksize, 0);
 	if (!tree_root->node)
 		goto fail_sb_buffer;
 
@@ -1732,11 +1728,11 @@ int btrfs_clear_buffer_defrag(struct extent_buffer *buf)
 		     EXTENT_DEFRAG, GFP_NOFS);
 }
 
-int btrfs_read_buffer(struct extent_buffer *buf)
+int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
 {
 	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
 	int ret;
-	ret = btree_read_extent_buffer_pages(root, buf, 0);
+	ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
 	if (ret == 0) {
 		buf->flags |= EXTENT_UPTODATE;
 	}
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 5d5f6979c41c..e29c895d5230 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -25,8 +25,9 @@ struct btrfs_device;
 struct btrfs_fs_devices;
 
 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
-				      u32 blocksize);
-int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize);
+				      u32 blocksize, u64 parent_transid);
+int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
+			 u64 parent_transid);
 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
 						   u64 bytenr, u32 blocksize);
 int clean_tree_block(struct btrfs_trans_handle *trans,
@@ -65,7 +66,7 @@ int btrfs_buffer_defrag(struct extent_buffer *buf);
 int btrfs_buffer_defrag_done(struct extent_buffer *buf);
 int btrfs_clear_buffer_defrag(struct extent_buffer *buf);
 int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf);
-int btrfs_read_buffer(struct extent_buffer *buf);
+int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid);
 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len);
 void btrfs_csum_final(u32 crc, char *result);
 void btrfs_throttle(struct btrfs_root *root);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index df95fb660d49..db07dde4a870 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1338,7 +1338,8 @@ static int finish_current_insert(struct btrfs_trans_handle *trans,
 					&extent_item, sizeof(extent_item));
 		clear_extent_bits(&info->extent_ins, start, end, EXTENT_LOCKED,
 				  GFP_NOFS);
-		eb = read_tree_block(extent_root, ins.objectid, ins.offset);
+		eb = read_tree_block(extent_root, ins.objectid, ins.offset,
+				     trans->transid);
 		level = btrfs_header_level(eb);
 		if (level == 0) {
 			btrfs_item_key(eb, &first, 0);
@@ -2076,7 +2077,8 @@ static void noinline reada_walk_down(struct btrfs_root *root,
 			}
 		}
 		mutex_unlock(&root->fs_info->fs_mutex);
-		ret = readahead_tree_block(root, bytenr, blocksize);
+		ret = readahead_tree_block(root, bytenr, blocksize,
+					   btrfs_node_ptr_generation(node, i));
 		last = bytenr + blocksize;
 		cond_resched();
 		mutex_lock(&root->fs_info->fs_mutex);
@@ -2096,6 +2098,7 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans,
 	u64 root_owner;
 	u64 root_gen;
 	u64 bytenr;
+	u64 ptr_gen;
 	struct extent_buffer *next;
 	struct extent_buffer *cur;
 	struct extent_buffer *parent;
@@ -2132,6 +2135,7 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans,
 			break;
 		}
 		bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
+		ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
 		blocksize = btrfs_level_size(root, *level - 1);
 		ret = lookup_extent_ref(trans, root, bytenr, blocksize, &refs);
 		BUG_ON(ret);
@@ -2152,7 +2156,8 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans,
 			reada_walk_down(root, cur, path->slots[*level]);
 
 			mutex_unlock(&root->fs_info->fs_mutex);
-			next = read_tree_block(root, bytenr, blocksize);
+			next = read_tree_block(root, bytenr, blocksize,
+					       ptr_gen);
 			mutex_lock(&root->fs_info->fs_mutex);
 
 			/* we've dropped the lock, double check */
@@ -2173,8 +2178,6 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans,
 				BUG_ON(ret);
 				continue;
 			}
-		} else if (next) {
-			btrfs_verify_block_csum(root, next);
 		}
 		WARN_ON(*level <= 0);
 		if (path->nodes[*level-1])
@@ -2609,7 +2612,7 @@ static int noinline relocate_one_reference(struct btrfs_root *extent_root,
 		int i;
 
 		eb = read_tree_block(found_root, extent_key->objectid,
-				     extent_key->offset);
+				     extent_key->offset, 0);
 		level = btrfs_header_level(eb);
 
 		if (level == 0)
diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c
index e99f3249d05a..14d863720302 100644
--- a/fs/btrfs/print-tree.c
+++ b/fs/btrfs/print-tree.c
@@ -186,7 +186,8 @@ void btrfs_print_tree(struct btrfs_root *root, struct extent_buffer *c)
 	for (i = 0; i < nr; i++) {
 		struct extent_buffer *next = read_tree_block(root,
 					btrfs_node_blockptr(c, i),
-					btrfs_level_size(root, level - 1));
+					btrfs_level_size(root, level - 1),
+					btrfs_node_ptr_generation(c, i));
 		if (btrfs_is_leaf(next) &&
 		    btrfs_header_level(c) != 1)
 			BUG();
diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c
index 256af1870eef..5085e9e693b9 100644
--- a/fs/btrfs/tree-defrag.c
+++ b/fs/btrfs/tree-defrag.c
@@ -28,6 +28,7 @@ static void reada_defrag(struct btrfs_root *root,
 	int i;
 	u32 nritems;
 	u64 bytenr;
+	u64 gen;
 	u32 blocksize;
 	int ret;
 
@@ -35,7 +36,8 @@ static void reada_defrag(struct btrfs_root *root,
 	nritems = btrfs_header_nritems(node);
 	for (i = 0; i < nritems; i++) {
 		bytenr = btrfs_node_blockptr(node, i);
-		ret = readahead_tree_block(root, bytenr, blocksize);
+		gen = btrfs_node_ptr_generation(node, i);
+		ret = readahead_tree_block(root, bytenr, blocksize, gen);
 		if (ret)
 			break;
 	}
@@ -101,10 +103,11 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 				path->slots[*level]++;
 				continue;
 			}
-			btrfs_verify_block_csum(root, next);
 		} else {
 			next = read_tree_block(root, bytenr,
-				       btrfs_level_size(root, *level - 1));
+				       btrfs_level_size(root, *level - 1),
+				       btrfs_node_ptr_generation(cur,
+							 path->slots[*level]));
 		}
 		ret = btrfs_cow_block(trans, root, next, path->nodes[*level],
 				      path->slots[*level], &next);
-- 
cgit v1.2.3


From 1259ab75c62462b8ffad90067b5e1f6312786a18 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Mon, 12 May 2008 13:39:03 -0400
Subject: Btrfs: Handle write errors on raid1 and raid10

When duplicate copies exist, writes are allowed to fail to one of those
copies.  This changeset includes a few changes that allow the FS to
continue even when some IOs fail.

It also adds verification of the parent generation number for btree blocks.
This generation is stored in the pointer to a block, and it ensures
that missed writes to are detected.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.c       |  2 +-
 fs/btrfs/disk-io.c     | 56 ++++++++++++++++++++++++++++++++++++++----
 fs/btrfs/disk-io.h     |  2 +-
 fs/btrfs/extent-tree.c |  4 +--
 fs/btrfs/extent_io.c   | 49 +++++++++++++++++++++++++++++++------
 fs/btrfs/extent_io.h   |  7 +++++-
 fs/btrfs/inode.c       | 66 +++++++++++++++++++++++++++++++-------------------
 fs/btrfs/tree-defrag.c |  7 +++---
 fs/btrfs/volumes.c     | 11 ++++++---
 9 files changed, 155 insertions(+), 49 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 289d71d8653a..02e571e6ee62 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -379,7 +379,7 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 
 		cur = btrfs_find_tree_block(root, blocknr, blocksize);
 		if (cur)
-			uptodate = btrfs_buffer_uptodate(cur);
+			uptodate = btrfs_buffer_uptodate(cur, gen);
 		else
 			uptodate = 0;
 		if (!cur || !uptodate) {
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index edee7a44f861..574b1245964e 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -205,6 +205,33 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
 	return 0;
 }
 
+static int verify_parent_transid(struct extent_io_tree *io_tree,
+				 struct extent_buffer *eb, u64 parent_transid)
+{
+	int ret;
+
+	if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
+		return 0;
+
+	lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS);
+	if (extent_buffer_uptodate(io_tree, eb) &&
+	    btrfs_header_generation(eb) == parent_transid) {
+		ret = 0;
+		goto out;
+	}
+	printk("parent transid verify failed on %llu wanted %llu found %llu\n",
+	       (unsigned long long)eb->start,
+	       (unsigned long long)parent_transid,
+	       (unsigned long long)btrfs_header_generation(eb));
+	ret = 1;
+out:
+	clear_extent_buffer_uptodate(io_tree, eb);
+	unlock_extent(io_tree, eb->start, eb->start + eb->len - 1,
+		      GFP_NOFS);
+	return ret;
+
+}
+
 static int btree_read_extent_buffer_pages(struct btrfs_root *root,
 					  struct extent_buffer *eb,
 					  u64 start, u64 parent_transid)
@@ -218,7 +245,8 @@ static int btree_read_extent_buffer_pages(struct btrfs_root *root,
 	while (1) {
 		ret = read_extent_buffer_pages(io_tree, eb, start, 1,
 					       btree_get_extent, mirror_num);
-		if (!ret)
+		if (!ret &&
+		    !verify_parent_transid(io_tree, eb, parent_transid))
 			return ret;
 
 		num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
@@ -330,6 +358,13 @@ int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
 		ret = -EIO;
 		goto err;
 	}
+	if (memcmp_extent_buffer(eb, root->fs_info->fsid,
+				 (unsigned long)btrfs_header_fsid(eb),
+				 BTRFS_FSID_SIZE)) {
+		printk("bad fsid on block %Lu\n", eb->start);
+		ret = -EIO;
+		goto err;
+	}
 	found_level = btrfs_header_level(eb);
 
 	ret = csum_tree_block(root, eb, 1);
@@ -1363,7 +1398,9 @@ static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
 					"I/O error on %s\n",
 				       bdevname(bh->b_bdev, b));
 		}
-		set_buffer_write_io_error(bh);
+		/* note, we dont' set_buffer_write_io_error because we have
+		 * our own ways of dealing with the IO errors
+		 */
 		clear_buffer_uptodate(bh);
 	}
 	unlock_buffer(bh);
@@ -1459,7 +1496,8 @@ int write_all_supers(struct btrfs_root *root)
 				ret = submit_bh(WRITE, bh);
 				BUG_ON(ret);
 				wait_on_buffer(bh);
-				BUG_ON(!buffer_uptodate(bh));
+				if (!buffer_uptodate(bh))
+					total_errors++;
 			} else {
 				total_errors++;
 			}
@@ -1607,10 +1645,18 @@ int close_ctree(struct btrfs_root *root)
 	return 0;
 }
 
-int btrfs_buffer_uptodate(struct extent_buffer *buf)
+int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
 {
+	int ret;
 	struct inode *btree_inode = buf->first_page->mapping->host;
-	return extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf);
+
+	ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf);
+	if (!ret)
+		return ret;
+
+	ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
+				    parent_transid);
+	return !ret;
 }
 
 int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index e29c895d5230..30d1ed293c25 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -56,7 +56,7 @@ int btrfs_insert_dev_radix(struct btrfs_root *root,
 void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr);
 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root);
 void btrfs_mark_buffer_dirty(struct extent_buffer *buf);
-int btrfs_buffer_uptodate(struct extent_buffer *buf);
+int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid);
 int btrfs_set_buffer_uptodate(struct extent_buffer *buf);
 int wait_on_tree_block_writeback(struct btrfs_root *root,
 				 struct extent_buffer *buf);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index db07dde4a870..605018c6045c 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1366,7 +1366,7 @@ static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes,
 	if (!pending) {
 		buf = btrfs_find_tree_block(root, bytenr, num_bytes);
 		if (buf) {
-			if (btrfs_buffer_uptodate(buf)) {
+			if (btrfs_buffer_uptodate(buf, 0)) {
 				u64 transid =
 				    root->fs_info->running_transaction->transid;
 				u64 header_transid =
@@ -2151,7 +2151,7 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans,
 			continue;
 		}
 		next = btrfs_find_tree_block(root, bytenr, blocksize);
-		if (!next || !btrfs_buffer_uptodate(next)) {
+		if (!next || !btrfs_buffer_uptodate(next, ptr_gen)) {
 			free_extent_buffer(next);
 			reada_walk_down(root, cur, path->slots[*level]);
 
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index dd403b426ff5..2a3624adc0cf 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -1366,7 +1366,7 @@ static int end_bio_extent_writepage(struct bio *bio,
 				   unsigned int bytes_done, int err)
 #endif
 {
-	const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
+	int uptodate = err == 0;
 	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
 	struct extent_state *state = bio->bi_private;
 	struct extent_io_tree *tree = state->tree;
@@ -1375,6 +1375,7 @@ static int end_bio_extent_writepage(struct bio *bio,
 	u64 end;
 	u64 cur;
 	int whole_page;
+	int ret;
 	unsigned long flags;
 
 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
@@ -1395,17 +1396,30 @@ static int end_bio_extent_writepage(struct bio *bio,
 		if (--bvec >= bio->bi_io_vec)
 			prefetchw(&bvec->bv_page->flags);
 
+		if (tree->ops && tree->ops->writepage_end_io_hook) {
+			ret = tree->ops->writepage_end_io_hook(page, start,
+						       end, state);
+			if (ret)
+				uptodate = 0;
+		}
+
+		if (!uptodate && tree->ops &&
+		    tree->ops->writepage_io_failed_hook) {
+			ret = tree->ops->writepage_io_failed_hook(bio, page,
+							 start, end, state);
+			if (ret == 0) {
+				state = NULL;
+				uptodate = (err == 0);
+				continue;
+			}
+		}
+
 		if (!uptodate) {
 			clear_extent_uptodate(tree, start, end, GFP_ATOMIC);
 			ClearPageUptodate(page);
 			SetPageError(page);
 		}
 
-		if (tree->ops && tree->ops->writepage_end_io_hook) {
-			tree->ops->writepage_end_io_hook(page, start, end,
-							 state);
-		}
-
 		/*
 		 * bios can get merged in funny ways, and so we need to
 		 * be careful with the state variable.  We know the
@@ -2073,9 +2087,9 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
 		} else {
 			ret = 0;
 		}
-		if (ret)
+		if (ret) {
 			SetPageError(page);
-		else {
+		} else {
 			unsigned long max_nr = end_index + 1;
 			set_range_writeback(tree, cur, cur + iosize - 1);
 			if (!PageWriteback(page)) {
@@ -2948,6 +2962,25 @@ int set_extent_buffer_dirty(struct extent_io_tree *tree,
 }
 EXPORT_SYMBOL(set_extent_buffer_dirty);
 
+int clear_extent_buffer_uptodate(struct extent_io_tree *tree,
+				struct extent_buffer *eb)
+{
+	unsigned long i;
+	struct page *page;
+	unsigned long num_pages;
+
+	num_pages = num_extent_pages(eb->start, eb->len);
+	eb->flags &= ~EXTENT_UPTODATE;
+
+	clear_extent_uptodate(tree, eb->start, eb->start + eb->len - 1,
+			      GFP_NOFS);
+	for (i = 0; i < num_pages; i++) {
+		page = extent_buffer_page(eb, i);
+		ClearPageUptodate(page);
+	}
+	return 0;
+}
+
 int set_extent_buffer_uptodate(struct extent_io_tree *tree,
 				struct extent_buffer *eb)
 {
diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h
index e48346147514..f1960dafaa19 100644
--- a/fs/btrfs/extent_io.h
+++ b/fs/btrfs/extent_io.h
@@ -36,9 +36,12 @@ struct extent_io_ops {
 	int (*readpage_io_failed_hook)(struct bio *bio, struct page *page,
 				       u64 start, u64 end,
 				       struct extent_state *state);
+	int (*writepage_io_failed_hook)(struct bio *bio, struct page *page,
+					u64 start, u64 end,
+				       struct extent_state *state);
 	int (*readpage_end_io_hook)(struct page *page, u64 start, u64 end,
 				    struct extent_state *state);
-	void (*writepage_end_io_hook)(struct page *page, u64 start, u64 end,
+	int (*writepage_end_io_hook)(struct page *page, u64 start, u64 end,
 				      struct extent_state *state);
 	int (*set_bit_hook)(struct inode *inode, u64 start, u64 end,
 			    unsigned long old, unsigned long bits);
@@ -212,6 +215,8 @@ int set_extent_buffer_dirty(struct extent_io_tree *tree,
 			     struct extent_buffer *eb);
 int set_extent_buffer_uptodate(struct extent_io_tree *tree,
 			       struct extent_buffer *eb);
+int clear_extent_buffer_uptodate(struct extent_io_tree *tree,
+				struct extent_buffer *eb);
 int extent_buffer_uptodate(struct extent_io_tree *tree,
 			   struct extent_buffer *eb);
 int map_extent_buffer(struct extent_buffer *eb, unsigned long offset,
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index a492fd238c88..08760ff9bab7 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -430,9 +430,9 @@ struct io_failure_record {
 	int last_mirror;
 };
 
-int btrfs_readpage_io_failed_hook(struct bio *failed_bio,
-				  struct page *page, u64 start, u64 end,
-				  struct extent_state *state)
+int btrfs_io_failed_hook(struct bio *failed_bio,
+			 struct page *page, u64 start, u64 end,
+			 struct extent_state *state)
 {
 	struct io_failure_record *failrec = NULL;
 	u64 private;
@@ -443,6 +443,7 @@ int btrfs_readpage_io_failed_hook(struct bio *failed_bio,
 	struct bio *bio;
 	int num_copies;
 	int ret;
+	int rw;
 	u64 logical;
 
 	ret = get_state_private(failure_tree, start, &private);
@@ -505,7 +506,41 @@ int btrfs_readpage_io_failed_hook(struct bio *failed_bio,
 	bio->bi_bdev = failed_bio->bi_bdev;
 	bio->bi_size = 0;
 	bio_add_page(bio, page, failrec->len, start - page_offset(page));
-	btrfs_submit_bio_hook(inode, READ, bio, failrec->last_mirror);
+	if (failed_bio->bi_rw & (1 << BIO_RW))
+		rw = WRITE;
+	else
+		rw = READ;
+
+	BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio,
+						      failrec->last_mirror);
+	return 0;
+}
+
+int btrfs_clean_io_failures(struct inode *inode, u64 start)
+{
+	u64 private;
+	u64 private_failure;
+	struct io_failure_record *failure;
+	int ret;
+
+	private = 0;
+	if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
+			     (u64)-1, 1, EXTENT_DIRTY)) {
+		ret = get_state_private(&BTRFS_I(inode)->io_failure_tree,
+					start, &private_failure);
+		if (ret == 0) {
+			failure = (struct io_failure_record *)(unsigned long)
+				   private_failure;
+			set_state_private(&BTRFS_I(inode)->io_failure_tree,
+					  failure->start, 0);
+			clear_extent_bits(&BTRFS_I(inode)->io_failure_tree,
+					  failure->start,
+					  failure->start + failure->len - 1,
+					  EXTENT_DIRTY | EXTENT_LOCKED,
+					  GFP_NOFS);
+			kfree(failure);
+		}
+	}
 	return 0;
 }
 
@@ -547,26 +582,7 @@ int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
 	/* if the io failure tree for this inode is non-empty,
 	 * check to see if we've recovered from a failed IO
 	 */
-	private = 0;
-	if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
-			     (u64)-1, 1, EXTENT_DIRTY)) {
-		u64 private_failure;
-		struct io_failure_record *failure;
-		ret = get_state_private(&BTRFS_I(inode)->io_failure_tree,
-					start, &private_failure);
-		if (ret == 0) {
-			failure = (struct io_failure_record *)(unsigned long)
-				   private_failure;
-			set_state_private(&BTRFS_I(inode)->io_failure_tree,
-					  failure->start, 0);
-			clear_extent_bits(&BTRFS_I(inode)->io_failure_tree,
-					  failure->start,
-					  failure->start + failure->len - 1,
-					  EXTENT_DIRTY | EXTENT_LOCKED,
-					  GFP_NOFS);
-			kfree(failure);
-		}
-	}
+	btrfs_clean_io_failures(inode, start);
 	return 0;
 
 zeroit:
@@ -3657,7 +3673,7 @@ static struct extent_io_ops btrfs_extent_io_ops = {
 	.merge_bio_hook = btrfs_merge_bio_hook,
 	.readpage_io_hook = btrfs_readpage_io_hook,
 	.readpage_end_io_hook = btrfs_readpage_end_io_hook,
-	.readpage_io_failed_hook = btrfs_readpage_io_failed_hook,
+	.readpage_io_failed_hook = btrfs_io_failed_hook,
 	.set_bit_hook = btrfs_set_bit_hook,
 	.clear_bit_hook = btrfs_clear_bit_hook,
 };
diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c
index 5085e9e693b9..c02e2bf2f028 100644
--- a/fs/btrfs/tree-defrag.c
+++ b/fs/btrfs/tree-defrag.c
@@ -51,6 +51,7 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 	struct extent_buffer *next;
 	struct extent_buffer *cur;
 	u64 bytenr;
+	u64 ptr_gen;
 	int ret = 0;
 	int is_extent = 0;
 
@@ -93,11 +94,12 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 			break;
 		}
 		bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
+		ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
 
 		if (cache_only) {
 			next = btrfs_find_tree_block(root, bytenr,
 					   btrfs_level_size(root, *level - 1));
-			if (!next || !btrfs_buffer_uptodate(next) ||
+			if (!next || !btrfs_buffer_uptodate(next, ptr_gen) ||
 			    !btrfs_buffer_defrag(next)) {
 				free_extent_buffer(next);
 				path->slots[*level]++;
@@ -106,8 +108,7 @@ static int defrag_walk_down(struct btrfs_trans_handle *trans,
 		} else {
 			next = read_tree_block(root, bytenr,
 				       btrfs_level_size(root, *level - 1),
-				       btrfs_node_ptr_generation(cur,
-							 path->slots[*level]));
+				       ptr_gen);
 		}
 		ret = btrfs_cow_block(trans, root, next, path->nodes[*level],
 				      path->slots[*level], &next);
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index b5d7bd1915b4..5fc7fb481474 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -1807,14 +1807,19 @@ static int end_bio_multi_stripe(struct bio *bio,
 	if (atomic_dec_and_test(&multi->stripes_pending)) {
 		bio->bi_private = multi->private;
 		bio->bi_end_io = multi->end_io;
-
 		/* only send an error to the higher layers if it is
 		 * beyond the tolerance of the multi-bio
 		 */
-		if (atomic_read(&multi->error) > multi->max_errors)
+		if (atomic_read(&multi->error) > multi->max_errors) {
 			err = -EIO;
-		else
+		} else if (err) {
+			/*
+			 * this bio is actually up to date, we didn't
+			 * go over the max number of errors
+			 */
+			set_bit(BIO_UPTODATE, &bio->bi_flags);
 			err = 0;
+		}
 		kfree(multi);
 
 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
-- 
cgit v1.2.3


From dfe25020689bb2d318782d2c9c7141203583fc70 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Tue, 13 May 2008 13:46:40 -0400
Subject: Btrfs: Add mount -o degraded to allow mounts to continue with missing
 devices

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.h   |   3 +
 fs/btrfs/disk-io.c |  49 ++++++----
 fs/btrfs/disk-io.h |   3 +-
 fs/btrfs/super.c   |  25 +++--
 fs/btrfs/volumes.c | 279 ++++++++++++++++++++++++++++++++++++++---------------
 fs/btrfs/volumes.h |   3 +
 6 files changed, 253 insertions(+), 109 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 6c2c2c4e4d2a..a9377fb16c37 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -657,6 +657,7 @@ struct btrfs_root {
 #define BTRFS_MOUNT_NODATACOW		(1 << 1)
 #define BTRFS_MOUNT_NOBARRIER		(1 << 2)
 #define BTRFS_MOUNT_SSD			(1 << 3)
+#define BTRFS_MOUNT_DEGRADED		(1 << 4)
 
 #define btrfs_clear_opt(o, opt)		((o) &= ~BTRFS_MOUNT_##opt)
 #define btrfs_set_opt(o, opt)		((o) |= BTRFS_MOUNT_##opt)
@@ -1606,4 +1607,6 @@ int btrfs_delete_xattrs(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root, struct inode *inode);
 /* super.c */
 u64 btrfs_parse_size(char *str);
+int btrfs_parse_options(char *options, struct btrfs_root *root,
+			char **subvol_name);
 #endif
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 574b1245964e..38b0d9ecda6a 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -614,21 +614,6 @@ int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
 	return ret;
 }
 
-static int close_all_devices(struct btrfs_fs_info *fs_info)
-{
-	struct list_head *list;
-	struct list_head *next;
-	struct btrfs_device *device;
-
-	list = &fs_info->fs_devices->devices;
-	list_for_each(next, list) {
-		device = list_entry(next, struct btrfs_device, dev_list);
-		close_bdev_excl(device->bdev);
-		device->bdev = NULL;
-	}
-	return 0;
-}
-
 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
 					    u64 bytenr, u32 blocksize)
 {
@@ -927,6 +912,8 @@ static int btrfs_congested_fn(void *congested_data, int bdi_bits)
 
 	list_for_each(cur, &info->fs_devices->devices) {
 		device = list_entry(cur, struct btrfs_device, dev_list);
+		if (!device->bdev)
+			continue;
 		bdi = blk_get_backing_dev_info(device->bdev);
 		if (bdi && bdi_congested(bdi, bdi_bits)) {
 			ret = 1;
@@ -1140,7 +1127,8 @@ static void btrfs_async_submit_work(struct work_struct *work)
 }
 
 struct btrfs_root *open_ctree(struct super_block *sb,
-			      struct btrfs_fs_devices *fs_devices)
+			      struct btrfs_fs_devices *fs_devices,
+			      char *options)
 {
 	u32 sectorsize;
 	u32 nodesize;
@@ -1276,12 +1264,19 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	if (!btrfs_super_root(disk_super))
 		goto fail_sb_buffer;
 
-	if (btrfs_super_num_devices(disk_super) != fs_devices->num_devices) {
+	btrfs_parse_options(options, tree_root, NULL);
+
+	if (btrfs_super_num_devices(disk_super) > fs_devices->num_devices) {
 		printk("Btrfs: wanted %llu devices, but found %llu\n",
 		       (unsigned long long)btrfs_super_num_devices(disk_super),
 		       (unsigned long long)fs_devices->num_devices);
-		goto fail_sb_buffer;
+		if (btrfs_test_opt(tree_root, DEGRADED))
+			printk("continuing in degraded mode\n");
+		else {
+			goto fail_sb_buffer;
+		}
 	}
+
 	fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
 
 	nodesize = btrfs_super_nodesize(disk_super);
@@ -1329,6 +1324,8 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	ret = btrfs_read_chunk_tree(chunk_root);
 	BUG_ON(ret);
 
+	btrfs_close_extra_devices(fs_devices);
+
 	blocksize = btrfs_level_size(tree_root,
 				     btrfs_super_root_level(disk_super));
 
@@ -1374,7 +1371,7 @@ fail_sb_buffer:
 fail_iput:
 	iput(fs_info->btree_inode);
 fail:
-	close_all_devices(fs_info);
+	btrfs_close_devices(fs_info->fs_devices);
 	btrfs_mapping_tree_free(&fs_info->mapping_tree);
 
 	kfree(extent_root);
@@ -1429,6 +1426,13 @@ int write_all_supers(struct btrfs_root *root)
 	dev_item = &sb->dev_item;
 	list_for_each(cur, head) {
 		dev = list_entry(cur, struct btrfs_device, dev_list);
+		if (!dev->bdev) {
+			total_errors++;
+			continue;
+		}
+		if (!dev->in_fs_metadata)
+			continue;
+
 		btrfs_set_stack_device_type(dev_item, dev->type);
 		btrfs_set_stack_device_id(dev_item, dev->devid);
 		btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
@@ -1482,6 +1486,11 @@ int write_all_supers(struct btrfs_root *root)
 
 	list_for_each(cur, head) {
 		dev = list_entry(cur, struct btrfs_device, dev_list);
+		if (!dev->bdev)
+			continue;
+		if (!dev->in_fs_metadata)
+			continue;
+
 		BUG_ON(!dev->pending_io);
 		bh = dev->pending_io;
 		wait_on_buffer(bh);
@@ -1631,7 +1640,7 @@ int close_ctree(struct btrfs_root *root)
 		kfree(hasher);
 	}
 #endif
-	close_all_devices(fs_info);
+	btrfs_close_devices(fs_info->fs_devices);
 	btrfs_mapping_tree_free(&fs_info->mapping_tree);
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 30d1ed293c25..2bc64fefe6ea 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -33,7 +33,8 @@ struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
 int clean_tree_block(struct btrfs_trans_handle *trans,
 		     struct btrfs_root *root, struct extent_buffer *buf);
 struct btrfs_root *open_ctree(struct super_block *sb,
-			      struct btrfs_fs_devices *fs_devices);
+			      struct btrfs_fs_devices *fs_devices,
+			      char *options);
 int close_ctree(struct btrfs_root *root);
 int write_ctree_super(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root);
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 020e5a83e31f..273a5b511f50 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -65,11 +65,13 @@ static void btrfs_put_super (struct super_block * sb)
 }
 
 enum {
-	Opt_subvol, Opt_nodatasum, Opt_nodatacow, Opt_max_extent,
-	Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd, Opt_err,
+	Opt_degraded, Opt_subvol, Opt_nodatasum, Opt_nodatacow,
+	Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier,
+	Opt_ssd, Opt_err,
 };
 
 static match_table_t tokens = {
+	{Opt_degraded, "degraded"},
 	{Opt_subvol, "subvol=%s"},
 	{Opt_nodatasum, "nodatasum"},
 	{Opt_nodatacow, "nodatacow"},
@@ -106,9 +108,8 @@ u64 btrfs_parse_size(char *str)
 	return res;
 }
 
-static int parse_options (char * options,
-			  struct btrfs_root *root,
-			  char **subvol_name)
+int btrfs_parse_options(char *options, struct btrfs_root *root,
+			char **subvol_name)
 {
 	char * p;
 	struct btrfs_fs_info *info = NULL;
@@ -135,6 +136,12 @@ static int parse_options (char * options,
 
 		token = match_token(p, tokens, args);
 		switch (token) {
+		case Opt_degraded:
+			if (info) {
+				printk("btrfs: allowing degraded mounts\n");
+				btrfs_set_opt(info->mount_opt, DEGRADED);
+			}
+			break;
 		case Opt_subvol:
 			if (subvol_name) {
 				*subvol_name = match_strdup(&args[0]);
@@ -234,7 +241,7 @@ static int btrfs_fill_super(struct super_block * sb,
 	sb->s_xattr = btrfs_xattr_handlers;
 	sb->s_time_gran = 1;
 
-	tree_root = open_ctree(sb, fs_devices);
+	tree_root = open_ctree(sb, fs_devices, (char *)data);
 
 	if (IS_ERR(tree_root)) {
 		printk("btrfs: open_ctree failed\n");
@@ -267,8 +274,6 @@ static int btrfs_fill_super(struct super_block * sb,
 		goto fail_close;
 	}
 
-	parse_options((char *)data, tree_root, NULL);
-
 	/* this does the super kobj at the same time */
 	err = btrfs_sysfs_add_super(tree_root->fs_info);
 	if (err)
@@ -341,7 +346,7 @@ int btrfs_get_sb_bdev(struct file_system_type *fs_type,
 	if (error)
 		return error;
 
-	bdev = fs_devices->lowest_bdev;
+	bdev = fs_devices->latest_bdev;
 	btrfs_lock_volumes();
 	s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
 	btrfs_unlock_volumes();
@@ -411,7 +416,7 @@ static int btrfs_get_sb(struct file_system_type *fs_type,
 	int ret;
 	char *subvol_name = NULL;
 
-	parse_options((char *)data, NULL, &subvol_name);
+	btrfs_parse_options((char *)data, NULL, &subvol_name);
 	ret = btrfs_get_sb_bdev(fs_type, flags, dev_name, data, mnt,
 			subvol_name ? subvol_name : "default");
 	if (subvol_name)
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 5fc7fb481474..43f74d17bcea 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -73,6 +73,7 @@ int btrfs_cleanup_fs_uuids(void)
 				close_bdev_excl(dev->bdev);
 			}
 			list_del(&dev->dev_list);
+			kfree(dev->name);
 			kfree(dev);
 		}
 	}
@@ -127,7 +128,6 @@ static int device_list_add(const char *path,
 		memcpy(fs_devices->fsid, disk_super->fsid, BTRFS_FSID_SIZE);
 		fs_devices->latest_devid = devid;
 		fs_devices->latest_trans = found_transid;
-		fs_devices->lowest_devid = (u64)-1;
 		fs_devices->num_devices = 0;
 		device = NULL;
 	} else {
@@ -159,13 +159,35 @@ static int device_list_add(const char *path,
 		fs_devices->latest_devid = devid;
 		fs_devices->latest_trans = found_transid;
 	}
-	if (fs_devices->lowest_devid > devid) {
-		fs_devices->lowest_devid = devid;
-	}
 	*fs_devices_ret = fs_devices;
 	return 0;
 }
 
+int btrfs_close_extra_devices(struct btrfs_fs_devices *fs_devices)
+{
+	struct list_head *head = &fs_devices->devices;
+	struct list_head *cur;
+	struct btrfs_device *device;
+
+	mutex_lock(&uuid_mutex);
+again:
+	list_for_each(cur, head) {
+		device = list_entry(cur, struct btrfs_device, dev_list);
+		if (!device->in_fs_metadata) {
+printk("getting rid of extra dev %s\n", device->name);
+			if (device->bdev)
+				close_bdev_excl(device->bdev);
+			list_del(&device->dev_list);
+			list_del(&device->dev_alloc_list);
+			fs_devices->num_devices--;
+			kfree(device->name);
+			kfree(device);
+			goto again;
+		}
+	}
+	mutex_unlock(&uuid_mutex);
+	return 0;
+}
 int btrfs_close_devices(struct btrfs_fs_devices *fs_devices)
 {
 	struct list_head *head = &fs_devices->devices;
@@ -179,6 +201,7 @@ int btrfs_close_devices(struct btrfs_fs_devices *fs_devices)
 			close_bdev_excl(device->bdev);
 		}
 		device->bdev = NULL;
+		device->in_fs_metadata = 0;
 	}
 	mutex_unlock(&uuid_mutex);
 	return 0;
@@ -199,6 +222,9 @@ int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 		if (device->bdev)
 			continue;
 
+		if (!device->name)
+			continue;
+
 		bdev = open_bdev_excl(device->name, flags, holder);
 
 		if (IS_ERR(bdev)) {
@@ -209,10 +235,8 @@ int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 		set_blocksize(bdev, 4096);
 		if (device->devid == fs_devices->latest_devid)
 			fs_devices->latest_bdev = bdev;
-		if (device->devid == fs_devices->lowest_devid) {
-			fs_devices->lowest_bdev = bdev;
-		}
 		device->bdev = bdev;
+		device->in_fs_metadata = 0;
 
 	}
 	mutex_unlock(&uuid_mutex);
@@ -439,7 +463,8 @@ int btrfs_free_dev_extent(struct btrfs_trans_handle *trans,
 	}
 	BUG_ON(ret);
 
-	device->bytes_used -= btrfs_dev_extent_length(leaf, extent);
+	if (device->bytes_used > 0)
+		device->bytes_used -= btrfs_dev_extent_length(leaf, extent);
 	ret = btrfs_del_item(trans, root, path);
 	BUG_ON(ret);
 
@@ -460,6 +485,7 @@ int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
 	struct extent_buffer *leaf;
 	struct btrfs_key key;
 
+	WARN_ON(!device->in_fs_metadata);
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
@@ -674,8 +700,6 @@ static int btrfs_rm_dev_item(struct btrfs_root *root,
 
 	next_dev = list_entry(fs_devices->devices.next, struct btrfs_device,
 			      dev_list);
-	if (bdev == fs_devices->lowest_bdev)
-		fs_devices->lowest_bdev = next_dev->bdev;
 	if (bdev == root->fs_info->sb->s_bdev)
 		root->fs_info->sb->s_bdev = next_dev->bdev;
 	if (bdev == fs_devices->latest_bdev)
@@ -698,7 +722,7 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
 {
 	struct btrfs_device *device;
 	struct block_device *bdev;
-	struct buffer_head *bh;
+	struct buffer_head *bh = NULL;
 	struct btrfs_super_block *disk_super;
 	u64 all_avail;
 	u64 devid;
@@ -712,47 +736,73 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
 		root->fs_info->avail_metadata_alloc_bits;
 
 	if ((all_avail & BTRFS_BLOCK_GROUP_RAID10) &&
-	    root->fs_info->fs_devices->num_devices <= 4) {
+	    btrfs_super_num_devices(&root->fs_info->super_copy) <= 4) {
 		printk("btrfs: unable to go below four devices on raid10\n");
 		ret = -EINVAL;
 		goto out;
 	}
 
 	if ((all_avail & BTRFS_BLOCK_GROUP_RAID1) &&
-	    root->fs_info->fs_devices->num_devices <= 2) {
+	    btrfs_super_num_devices(&root->fs_info->super_copy) <= 2) {
 		printk("btrfs: unable to go below two devices on raid1\n");
 		ret = -EINVAL;
 		goto out;
 	}
 
-	bdev = open_bdev_excl(device_path, 0, root->fs_info->bdev_holder);
-	if (IS_ERR(bdev)) {
-		ret = PTR_ERR(bdev);
-		goto out;
-	}
+	if (strcmp(device_path, "missing") == 0) {
+		struct list_head *cur;
+		struct list_head *devices;
+		struct btrfs_device *tmp;
 
-	bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096);
-	if (!bh) {
-		ret = -EIO;
-		goto error_close;
-	}
-	disk_super = (struct btrfs_super_block *)bh->b_data;
-	if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
-	    sizeof(disk_super->magic))) {
-		ret = -ENOENT;
-		goto error_brelse;
-	}
-	if (memcmp(disk_super->fsid, root->fs_info->fsid, BTRFS_FSID_SIZE)) {
-		ret = -ENOENT;
-		goto error_brelse;
-	}
-	devid = le64_to_cpu(disk_super->dev_item.devid);
-	device = btrfs_find_device(root, devid, NULL);
-	if (!device) {
-		ret = -ENOENT;
-		goto error_brelse;
-	}
+		device = NULL;
+		devices = &root->fs_info->fs_devices->devices;
+		list_for_each(cur, devices) {
+			tmp = list_entry(cur, struct btrfs_device, dev_list);
+			if (tmp->in_fs_metadata && !tmp->bdev) {
+				device = tmp;
+				break;
+			}
+		}
+		bdev = NULL;
+		bh = NULL;
+		disk_super = NULL;
+		if (!device) {
+			printk("btrfs: no missing devices found to remove\n");
+			goto out;
+		}
+
+	} else {
+		bdev = open_bdev_excl(device_path, 0,
+				      root->fs_info->bdev_holder);
+		if (IS_ERR(bdev)) {
+			ret = PTR_ERR(bdev);
+			goto out;
+		}
+
+		bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096);
+		if (!bh) {
+			ret = -EIO;
+			goto error_close;
+		}
+		disk_super = (struct btrfs_super_block *)bh->b_data;
+		if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
+		    sizeof(disk_super->magic))) {
+			ret = -ENOENT;
+			goto error_brelse;
+		}
+		if (memcmp(disk_super->fsid, root->fs_info->fsid,
+			   BTRFS_FSID_SIZE)) {
+			ret = -ENOENT;
+			goto error_brelse;
+		}
+		devid = le64_to_cpu(disk_super->dev_item.devid);
+		device = btrfs_find_device(root, devid, NULL);
+		if (!device) {
+			ret = -ENOENT;
+			goto error_brelse;
+		}
 
+	}
 	root->fs_info->fs_devices->num_devices--;
 
 	ret = btrfs_shrink_device(device, 0);
@@ -764,19 +814,25 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
 	if (ret)
 		goto error_brelse;
 
-	/* make sure this device isn't detected as part of the FS anymore */
-	memset(&disk_super->magic, 0, sizeof(disk_super->magic));
-	set_buffer_dirty(bh);
-	sync_dirty_buffer(bh);
-
-	brelse(bh);
-
-	/* one close for the device struct or super_block */
-	close_bdev_excl(device->bdev);
+	if (bh) {
+		/* make sure this device isn't detected as part of
+		 * the FS anymore
+		 */
+		memset(&disk_super->magic, 0, sizeof(disk_super->magic));
+		set_buffer_dirty(bh);
+		sync_dirty_buffer(bh);
 
-	/* one close for us */
-	close_bdev_excl(device->bdev);
+		brelse(bh);
+	}
 
+	if (device->bdev) {
+		/* one close for the device struct or super_block */
+		close_bdev_excl(device->bdev);
+	}
+	if (bdev) {
+		/* one close for us */
+		close_bdev_excl(bdev);
+	}
 	kfree(device->name);
 	kfree(device);
 	ret = 0;
@@ -785,7 +841,8 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
 error_brelse:
 	brelse(bh);
 error_close:
-	close_bdev_excl(bdev);
+	if (bdev)
+		close_bdev_excl(bdev);
 out:
 	mutex_unlock(&uuid_mutex);
 	mutex_unlock(&root->fs_info->fs_mutex);
@@ -839,6 +896,7 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path)
 	device->total_bytes = i_size_read(bdev->bd_inode);
 	device->dev_root = root->fs_info->dev_root;
 	device->bdev = bdev;
+	device->in_fs_metadata = 1;
 
 	ret = btrfs_add_device(trans, root, device);
 	if (ret)
@@ -1041,8 +1099,10 @@ int btrfs_relocate_chunk(struct btrfs_root *root,
 					    map->stripes[i].physical);
 		BUG_ON(ret);
 
-		ret = btrfs_update_device(trans, map->stripes[i].dev);
-		BUG_ON(ret);
+		if (map->stripes[i].dev) {
+			ret = btrfs_update_device(trans, map->stripes[i].dev);
+			BUG_ON(ret);
+		}
 	}
 	ret = btrfs_free_chunk(trans, root, chunk_tree, chunk_objectid,
 			       chunk_offset);
@@ -1415,10 +1475,13 @@ again:
 	while(index < num_stripes) {
 		device = list_entry(cur, struct btrfs_device, dev_alloc_list);
 
-		avail = device->total_bytes - device->bytes_used;
+		if (device->total_bytes > device->bytes_used)
+			avail = device->total_bytes - device->bytes_used;
+		else
+			avail = 0;
 		cur = cur->next;
 
-		if (avail >= min_free) {
+		if (device->in_fs_metadata && avail >= min_free) {
 			u64 ignored_start = 0;
 			ret = find_free_dev_extent(trans, device, path,
 						   min_free,
@@ -1430,7 +1493,7 @@ again:
 				if (type & BTRFS_BLOCK_GROUP_DUP)
 					index++;
 			}
-		} else if (avail > max_avail)
+		} else if (device->in_fs_metadata && avail > max_avail)
 			max_avail = avail;
 		if (cur == dev_list)
 			break;
@@ -1610,6 +1673,22 @@ int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len)
 	return ret;
 }
 
+static int find_live_mirror(struct map_lookup *map, int first, int num,
+			    int optimal)
+{
+	int i;
+	if (map->stripes[optimal].dev->bdev)
+		return optimal;
+	for (i = first; i < first + num; i++) {
+		if (map->stripes[i].dev->bdev)
+			return i;
+	}
+	/* we couldn't find one that doesn't fail.  Just return something
+	 * and the io error handling code will clean up eventually
+	 */
+	return optimal;
+}
+
 static int __btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
 			     u64 logical, u64 *length,
 			     struct btrfs_multi_bio **multi_ret,
@@ -1712,8 +1791,11 @@ again:
 			num_stripes = map->num_stripes;
 		else if (mirror_num)
 			stripe_index = mirror_num - 1;
-		else
-			stripe_index = current->pid % map->num_stripes;
+		else {
+			stripe_index = find_live_mirror(map, 0,
+					    map->num_stripes,
+					    current->pid % map->num_stripes);
+		}
 
 	} else if (map->type & BTRFS_BLOCK_GROUP_DUP) {
 		if (rw & (1 << BIO_RW))
@@ -1731,8 +1813,11 @@ again:
 			num_stripes = map->sub_stripes;
 		else if (mirror_num)
 			stripe_index += mirror_num - 1;
-		else
-			stripe_index += current->pid % map->sub_stripes;
+		else {
+			stripe_index = find_live_mirror(map, stripe_index,
+					      map->sub_stripes, stripe_index +
+					      current->pid % map->sub_stripes);
+		}
 	} else {
 		/*
 		 * after this do_div call, stripe_nr is the number of stripes
@@ -1749,9 +1834,11 @@ again:
 			struct backing_dev_info *bdi;
 
 			device = map->stripes[stripe_index].dev;
-			bdi = blk_get_backing_dev_info(device->bdev);
-			if (bdi->unplug_io_fn) {
-				bdi->unplug_io_fn(bdi, unplug_page);
+			if (device->bdev) {
+				bdi = blk_get_backing_dev_info(device->bdev);
+				if (bdi->unplug_io_fn) {
+					bdi->unplug_io_fn(bdi, unplug_page);
+				}
 			}
 		} else {
 			multi->stripes[i].physical =
@@ -1880,12 +1967,21 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio,
 		}
 		bio->bi_sector = multi->stripes[dev_nr].physical >> 9;
 		dev = multi->stripes[dev_nr].dev;
-
-		bio->bi_bdev = dev->bdev;
-		spin_lock(&dev->io_lock);
-		dev->total_ios++;
-		spin_unlock(&dev->io_lock);
-		submit_bio(rw, bio);
+		if (dev && dev->bdev) {
+			bio->bi_bdev = dev->bdev;
+			spin_lock(&dev->io_lock);
+			dev->total_ios++;
+			spin_unlock(&dev->io_lock);
+			submit_bio(rw, bio);
+		} else {
+			bio->bi_bdev = root->fs_info->fs_devices->latest_bdev;
+			bio->bi_sector = logical >> 9;
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
+			bio_endio(bio, bio->bi_size, -EIO);
+#else
+			bio_endio(bio, -EIO);
+#endif
+		}
 		dev_nr++;
 	}
 	if (total_devs == 1)
@@ -1901,6 +1997,27 @@ struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid,
 	return __find_device(head, devid, uuid);
 }
 
+static struct btrfs_device *add_missing_dev(struct btrfs_root *root,
+					    u64 devid, u8 *dev_uuid)
+{
+	struct btrfs_device *device;
+	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
+
+	device = kzalloc(sizeof(*device), GFP_NOFS);
+	list_add(&device->dev_list,
+		 &fs_devices->devices);
+	list_add(&device->dev_alloc_list,
+		 &fs_devices->alloc_list);
+	device->barriers = 1;
+	device->dev_root = root->fs_info->dev_root;
+	device->devid = devid;
+	fs_devices->num_devices++;
+	spin_lock_init(&device->io_lock);
+	memcpy(device->uuid, dev_uuid, BTRFS_UUID_SIZE);
+	return device;
+}
+
+
 static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
 			  struct extent_buffer *leaf,
 			  struct btrfs_chunk *chunk)
@@ -1965,11 +2082,22 @@ static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
 				   btrfs_stripe_dev_uuid_nr(chunk, i),
 				   BTRFS_UUID_SIZE);
 		map->stripes[i].dev = btrfs_find_device(root, devid, uuid);
-		if (!map->stripes[i].dev) {
+
+		if (!map->stripes[i].dev && !btrfs_test_opt(root, DEGRADED)) {
 			kfree(map);
 			free_extent_map(em);
 			return -EIO;
 		}
+		if (!map->stripes[i].dev) {
+			map->stripes[i].dev =
+				add_missing_dev(root, devid, uuid);
+			if (!map->stripes[i].dev) {
+				kfree(map);
+				free_extent_map(em);
+				return -EIO;
+			}
+		}
+		map->stripes[i].dev->in_fs_metadata = 1;
 	}
 
 	spin_lock(&map_tree->map_tree.lock);
@@ -2016,20 +2144,15 @@ static int read_one_dev(struct btrfs_root *root,
 			   BTRFS_UUID_SIZE);
 	device = btrfs_find_device(root, devid, dev_uuid);
 	if (!device) {
-		printk("warning devid %Lu not found already\n", devid);
-		device = kzalloc(sizeof(*device), GFP_NOFS);
+		printk("warning devid %Lu missing\n", devid);
+		device = add_missing_dev(root, devid, dev_uuid);
 		if (!device)
 			return -ENOMEM;
-		list_add(&device->dev_list,
-			 &root->fs_info->fs_devices->devices);
-		list_add(&device->dev_alloc_list,
-			 &root->fs_info->fs_devices->alloc_list);
-		device->barriers = 1;
-		spin_lock_init(&device->io_lock);
 	}
 
 	fill_device_from_item(leaf, dev_item, device);
 	device->dev_root = root->fs_info->dev_root;
+	device->in_fs_metadata = 1;
 	ret = 0;
 #if 0
 	ret = btrfs_open_device(device);
diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h
index 0f94a69e6eb6..454fe8103329 100644
--- a/fs/btrfs/volumes.h
+++ b/fs/btrfs/volumes.h
@@ -27,8 +27,10 @@ struct btrfs_device {
 	struct list_head dev_alloc_list;
 	struct btrfs_root *dev_root;
 	struct buffer_head *pending_io;
+	u64 generation;
 
 	int barriers;
+	int in_fs_metadata;
 
 	spinlock_t io_lock;
 
@@ -122,6 +124,7 @@ int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 int btrfs_scan_one_device(const char *path, int flags, void *holder,
 			  struct btrfs_fs_devices **fs_devices_ret);
 int btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
+int btrfs_close_extra_devices(struct btrfs_fs_devices *fs_devices);
 int btrfs_add_device(struct btrfs_trans_handle *trans,
 		     struct btrfs_root *root,
 		     struct btrfs_device *device);
-- 
cgit v1.2.3


From 89ce8a63d0c761fbb02089850605360f389477d8 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Wed, 25 Jun 2008 16:01:31 -0400
Subject: Add btrfs_end_transaction_throttle to force writers to wait for
 pending commits

The existing throttle mechanism was often not sufficient to prevent
new writers from coming in and making a given transaction run forever.
This adds an explicit wait at the end of most operations so they will
allow the current transaction to close.

There is no wait inside file_write, inode updates, or cow filling, all which
have different deadlock possibilities.

This is a temporary measure until better asynchronous commit support is
added.  This code leads to stalls as it waits for data=ordered
writeback, and it really needs to be fixed.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/disk-io.c     | 18 --------------
 fs/btrfs/disk-io.h     |  1 -
 fs/btrfs/file.c        |  1 -
 fs/btrfs/inode.c       | 24 ++++++------------
 fs/btrfs/ioctl.c       |  2 --
 fs/btrfs/transaction.c | 66 ++++++++++++++++++++++++++++++++++----------------
 fs/btrfs/transaction.h |  2 ++
 7 files changed, 55 insertions(+), 59 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 3cc480b83819..52569b57692d 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1672,24 +1672,6 @@ void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
 	set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf);
 }
 
-void btrfs_throttle(struct btrfs_root *root)
-{
-#if 0
-	struct backing_dev_info *bdi;
-
-	bdi = &root->fs_info->bdi;
-	if (atomic_read(&root->fs_info->throttles) &&
-	    bdi_write_congested(bdi)) {
-#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18)
-		congestion_wait(WRITE, HZ/20);
-#else
-		blk_congestion_wait(WRITE, HZ/20);
-#endif
-
-	}
-#endif
-}
-
 void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
 {
 	/*
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 2bc64fefe6ea..deff6b4815a7 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -70,7 +70,6 @@ int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf);
 int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid);
 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len);
 void btrfs_csum_final(u32 crc, char *result);
-void btrfs_throttle(struct btrfs_root *root);
 int btrfs_open_device(struct btrfs_device *dev);
 int btrfs_verify_block_csum(struct btrfs_root *root,
 			    struct extent_buffer *buf);
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index b7f8f92daf8a..ece221cba90c 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -934,7 +934,6 @@ static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
 		balance_dirty_pages_ratelimited_nr(inode->i_mapping, num_pages);
 		if (num_pages < (root->leafsize >> PAGE_CACHE_SHIFT) + 1)
 			btrfs_btree_balance_dirty(root, 1);
-		btrfs_throttle(root);
 		cond_resched();
 	}
 out:
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index cf27b5984627..bbba3350d023 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -855,10 +855,9 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
 		btrfs_del_ordered_inode(inode, 1);
 	}
 
-	btrfs_end_transaction(trans, root);
+	btrfs_end_transaction_throttle(trans, root);
 fail:
 	btrfs_btree_balance_dirty(root, nr);
-	btrfs_throttle(root);
 	return ret;
 }
 
@@ -889,10 +888,9 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
 	}
 
 	nr = trans->blocks_used;
-	ret = btrfs_end_transaction(trans, root);
+	ret = btrfs_end_transaction_throttle(trans, root);
 fail:
 	btrfs_btree_balance_dirty(root, nr);
-	btrfs_throttle(root);
 
 	if (ret && !err)
 		err = ret;
@@ -1871,14 +1869,13 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
 	btrfs_update_inode_block_group(trans, dir);
 out_unlock:
 	nr = trans->blocks_used;
-	btrfs_end_transaction(trans, root);
+	btrfs_end_transaction_throttle(trans, root);
 fail:
 	if (drop_inode) {
 		inode_dec_link_count(inode);
 		iput(inode);
 	}
 	btrfs_btree_balance_dirty(root, nr);
-	btrfs_throttle(root);
 	return err;
 }
 
@@ -1936,14 +1933,13 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry,
 	btrfs_update_inode_block_group(trans, dir);
 out_unlock:
 	nr = trans->blocks_used;
-	btrfs_end_transaction(trans, root);
+	btrfs_end_transaction_throttle(trans, root);
 fail:
 	if (drop_inode) {
 		inode_dec_link_count(inode);
 		iput(inode);
 	}
 	btrfs_btree_balance_dirty(root, nr);
-	btrfs_throttle(root);
 	return err;
 }
 
@@ -1985,14 +1981,13 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
 		drop_inode = 1;
 
 	nr = trans->blocks_used;
-	btrfs_end_transaction(trans, root);
+	btrfs_end_transaction_throttle(trans, root);
 fail:
 	if (drop_inode) {
 		inode_dec_link_count(inode);
 		iput(inode);
 	}
 	btrfs_btree_balance_dirty(root, nr);
-	btrfs_throttle(root);
 	return err;
 }
 
@@ -2055,13 +2050,12 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 
 out_fail:
 	nr = trans->blocks_used;
-	btrfs_end_transaction(trans, root);
+	btrfs_end_transaction_throttle(trans, root);
 
 out_unlock:
 	if (drop_on_err)
 		iput(inode);
 	btrfs_btree_balance_dirty(root, nr);
-	btrfs_throttle(root);
 	return err;
 }
 
@@ -2587,10 +2581,9 @@ static void btrfs_truncate(struct inode *inode)
 	btrfs_update_inode(trans, root, inode);
 	nr = trans->blocks_used;
 
-	ret = btrfs_end_transaction(trans, root);
+	ret = btrfs_end_transaction_throttle(trans, root);
 	BUG_ON(ret);
 	btrfs_btree_balance_dirty(root, nr);
-	btrfs_throttle(root);
 }
 
 /*
@@ -2912,14 +2905,13 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
 
 out_unlock:
 	nr = trans->blocks_used;
-	btrfs_end_transaction(trans, root);
+	btrfs_end_transaction_throttle(trans, root);
 out_fail:
 	if (drop_inode) {
 		inode_dec_link_count(inode);
 		iput(inode);
 	}
 	btrfs_btree_balance_dirty(root, nr);
-	btrfs_throttle(root);
 	return err;
 }
 
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 6002eb64daf9..026039a2ac58 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -164,7 +164,6 @@ fail:
 		ret = err;
 fail_commit:
 	btrfs_btree_balance_dirty(root, nr);
-	btrfs_throttle(root);
 	return ret;
 }
 
@@ -206,7 +205,6 @@ static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
 
 fail_unlock:
 	btrfs_btree_balance_dirty(root, nr);
-	btrfs_throttle(root);
 	return ret;
 }
 
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 5a1ee0665ae8..69ed5f85a387 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -130,8 +130,27 @@ struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
 	return h;
 }
 
-int btrfs_end_transaction(struct btrfs_trans_handle *trans,
-			  struct btrfs_root *root)
+static noinline int wait_for_commit(struct btrfs_root *root,
+				    struct btrfs_transaction *commit)
+{
+	DEFINE_WAIT(wait);
+	mutex_lock(&root->fs_info->trans_mutex);
+	while(!commit->commit_done) {
+		prepare_to_wait(&commit->commit_wait, &wait,
+				TASK_UNINTERRUPTIBLE);
+		if (commit->commit_done)
+			break;
+		mutex_unlock(&root->fs_info->trans_mutex);
+		schedule();
+		mutex_lock(&root->fs_info->trans_mutex);
+	}
+	mutex_unlock(&root->fs_info->trans_mutex);
+	finish_wait(&commit->commit_wait, &wait);
+	return 0;
+}
+
+static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root, int throttle)
 {
 	struct btrfs_transaction *cur_trans;
 
@@ -140,8 +159,18 @@ int btrfs_end_transaction(struct btrfs_trans_handle *trans,
 	WARN_ON(cur_trans != trans->transaction);
 	WARN_ON(cur_trans->num_writers < 1);
 	cur_trans->num_writers--;
+
 	if (waitqueue_active(&cur_trans->writer_wait))
 		wake_up(&cur_trans->writer_wait);
+
+	if (cur_trans->in_commit && throttle) {
+		int ret;
+		mutex_unlock(&root->fs_info->trans_mutex);
+		ret = wait_for_commit(root, cur_trans);
+		BUG_ON(ret);
+		mutex_lock(&root->fs_info->trans_mutex);
+	}
+
 	put_transaction(cur_trans);
 	mutex_unlock(&root->fs_info->trans_mutex);
 	memset(trans, 0, sizeof(*trans));
@@ -149,6 +178,18 @@ int btrfs_end_transaction(struct btrfs_trans_handle *trans,
 	return 0;
 }
 
+int btrfs_end_transaction(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root)
+{
+	return __btrfs_end_transaction(trans, root, 0);
+}
+
+int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans,
+				   struct btrfs_root *root)
+{
+	return __btrfs_end_transaction(trans, root, 1);
+}
+
 
 int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
 				     struct btrfs_root *root)
@@ -240,25 +281,6 @@ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
 	return 0;
 }
 
-static noinline int wait_for_commit(struct btrfs_root *root,
-				    struct btrfs_transaction *commit)
-{
-	DEFINE_WAIT(wait);
-	mutex_lock(&root->fs_info->trans_mutex);
-	while(!commit->commit_done) {
-		prepare_to_wait(&commit->commit_wait, &wait,
-				TASK_UNINTERRUPTIBLE);
-		if (commit->commit_done)
-			break;
-		mutex_unlock(&root->fs_info->trans_mutex);
-		schedule();
-		mutex_lock(&root->fs_info->trans_mutex);
-	}
-	mutex_unlock(&root->fs_info->trans_mutex);
-	finish_wait(&commit->commit_wait, &wait);
-	return 0;
-}
-
 struct dirty_root {
 	struct list_head list;
 	struct btrfs_root *root;
@@ -680,6 +702,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 			     root->fs_info->btree_inode->i_mapping, GFP_NOFS);
 
 	trans->transaction->in_commit = 1;
+printk("trans %Lu in commit\n", trans->transid);
 	cur_trans = trans->transaction;
 	if (cur_trans->list.prev != &root->fs_info->trans_list) {
 		prev_trans = list_entry(cur_trans->list.prev,
@@ -760,6 +783,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	kfree(pinned_copy);
 
 	cur_trans->commit_done = 1;
+printk("trans %Lu done in commit\n", cur_trans->transid);
 	root->fs_info->last_trans_committed = cur_trans->transid;
 	wake_up(&cur_trans->commit_wait);
 	put_transaction(cur_trans);
diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h
index c3172ddb3321..52559b51b181 100644
--- a/fs/btrfs/transaction.h
+++ b/fs/btrfs/transaction.h
@@ -101,4 +101,6 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root);
 int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root);
+int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans,
+				   struct btrfs_root *root);
 #endif
-- 
cgit v1.2.3


From 3f157a2fd2ad731e1ed9964fecdc5f459f04a4a4 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Wed, 25 Jun 2008 16:01:31 -0400
Subject: Btrfs: Online btree defragmentation fixes

The btree defragger wasn't making forward progress because the new key wasn't
being saved by the btrfs_search_forward function.

This also disables the automatic btree defrag, it wasn't scaling well to
huge filesystems.  The auto-defrag needs to be done differently.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/ctree.c       | 170 ++++++++++++++++++++++++++++++++++++++++++++++---
 fs/btrfs/ctree.h       |   7 +-
 fs/btrfs/disk-io.c     |  61 +-----------------
 fs/btrfs/disk-io.h     |   6 --
 fs/btrfs/extent-tree.c |   2 -
 fs/btrfs/super.c       |   1 -
 fs/btrfs/transaction.c |  35 +---------
 fs/btrfs/transaction.h |   1 -
 fs/btrfs/tree-defrag.c |  36 ++++++-----
 9 files changed, 190 insertions(+), 129 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 7f4cc2b88d09..0cb80f32a9c7 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -63,10 +63,9 @@ void btrfs_free_path(struct btrfs_path *p)
 void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p)
 {
 	int i;
-	int keep = p->keep_locks;
-	int skip = p->skip_locking;
 
 	for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
+		p->slots[i] = 0;
 		if (!p->nodes[i])
 			continue;
 		if (p->locks[i]) {
@@ -74,10 +73,8 @@ void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p)
 			p->locks[i] = 0;
 		}
 		free_extent_buffer(p->nodes[i]);
+		p->nodes[i] = NULL;
 	}
-	memset(p, 0, sizeof(*p));
-	p->keep_locks = keep;
-	p->skip_locking = skip;
 }
 
 struct extent_buffer *btrfs_root_node(struct btrfs_root *root)
@@ -463,8 +460,6 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 		search_start = cur->start;
 		last_block = cur->start;
 		*last_ret = search_start;
-		if (parent_level == 1)
-			btrfs_clear_buffer_defrag(cur);
 		btrfs_tree_unlock(cur);
 		free_extent_buffer(cur);
 	}
@@ -2969,8 +2964,138 @@ int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path)
 	return 1;
 }
 
+/*
+ * A helper function to walk down the tree starting at min_key, and looking
+ * for nodes or leaves that are either in cache or have a minimum
+ * transaction id.  This is used by the btree defrag code, but could
+ * also be used to search for blocks that have changed since a given
+ * transaction id.
+ *
+ * This does not cow, but it does stuff the starting key it finds back
+ * into min_key, so you can call btrfs_search_slot with cow=1 on the
+ * key and get a writable path.
+ *
+ * This does lock as it descends, and path->keep_locks should be set
+ * to 1 by the caller.
+ *
+ * This honors path->lowest_level to prevent descent past a given level
+ * of the tree.
+ *
+ * returns zero if something useful was found, < 0 on error and 1 if there
+ * was nothing in the tree that matched the search criteria.
+ */
+int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key,
+			 struct btrfs_path *path, int cache_only,
+			 u64 min_trans)
+{
+	struct extent_buffer *cur;
+	struct btrfs_key found_key;
+	int slot;
+	u32 nritems;
+	int level;
+	int ret = 1;
+
+again:
+	cur = btrfs_lock_root_node(root);
+	level = btrfs_header_level(cur);
+	path->nodes[level] = cur;
+	path->locks[level] = 1;
+
+	if (btrfs_header_generation(cur) < min_trans) {
+		ret = 1;
+		goto out;
+	}
+	while(1) {
+		nritems = btrfs_header_nritems(cur);
+		level = btrfs_header_level(cur);
+		bin_search(cur, min_key, level, &slot);
+
+		/* at level = 0, we're done, setup the path and exit */
+		if (level == 0) {
+			ret = 0;
+			path->slots[level] = slot;
+			btrfs_item_key_to_cpu(cur, &found_key, slot);
+			goto out;
+		}
+		/*
+		 * check this node pointer against the cache_only and
+		 * min_trans parameters.  If it isn't in cache or is too
+		 * old, skip to the next one.
+		 */
+		while(slot < nritems) {
+			u64 blockptr;
+			u64 gen;
+			struct extent_buffer *tmp;
+			blockptr = btrfs_node_blockptr(cur, slot);
+			gen = btrfs_node_ptr_generation(cur, slot);
+			if (gen < min_trans) {
+				slot++;
+				continue;
+			}
+			if (!cache_only)
+				break;
+
+			tmp = btrfs_find_tree_block(root, blockptr,
+					    btrfs_level_size(root, level - 1));
+
+			if (tmp && btrfs_buffer_uptodate(tmp, gen)) {
+				free_extent_buffer(tmp);
+				break;
+			}
+			if (tmp)
+				free_extent_buffer(tmp);
+			slot++;
+		}
+		/*
+		 * we didn't find a candidate key in this node, walk forward
+		 * and find another one
+		 */
+		if (slot >= nritems) {
+			ret = btrfs_find_next_key(root, path, min_key, level,
+						  cache_only, min_trans);
+			if (ret == 0) {
+				btrfs_release_path(root, path);
+				goto again;
+			} else {
+				goto out;
+			}
+		}
+		/* save our key for returning back */
+		btrfs_node_key_to_cpu(cur, &found_key, slot);
+		path->slots[level] = slot;
+		if (level == path->lowest_level) {
+			ret = 0;
+			unlock_up(path, level, 1);
+			goto out;
+		}
+		cur = read_node_slot(root, cur, slot);
+
+		btrfs_tree_lock(cur);
+		path->locks[level - 1] = 1;
+		path->nodes[level - 1] = cur;
+		unlock_up(path, level, 1);
+	}
+out:
+	if (ret == 0)
+		memcpy(min_key, &found_key, sizeof(found_key));
+	return ret;
+}
+
+/*
+ * this is similar to btrfs_next_leaf, but does not try to preserve
+ * and fixup the path.  It looks for and returns the next key in the
+ * tree based on the current path and the cache_only and min_trans
+ * parameters.
+ *
+ * 0 is returned if another key is found, < 0 if there are any errors
+ * and 1 is returned if there are no higher keys in the tree
+ *
+ * path->keep_locks should be set to 1 on the search made before
+ * calling this function.
+ */
 int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path,
-			struct btrfs_key *key, int lowest_level)
+			struct btrfs_key *key, int lowest_level,
+			int cache_only, u64 min_trans)
 {
 	int level = lowest_level;
 	int slot;
@@ -2982,6 +3107,7 @@ int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path,
 
 		slot = path->slots[level] + 1;
 		c = path->nodes[level];
+next:
 		if (slot >= btrfs_header_nritems(c)) {
 			level++;
 			if (level == BTRFS_MAX_LEVEL) {
@@ -2991,8 +3117,28 @@ int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path,
 		}
 		if (level == 0)
 			btrfs_item_key_to_cpu(c, key, slot);
-		else
+		else {
+			u64 blockptr = btrfs_node_blockptr(c, slot);
+			u64 gen = btrfs_node_ptr_generation(c, slot);
+
+			if (cache_only) {
+				struct extent_buffer *cur;
+				cur = btrfs_find_tree_block(root, blockptr,
+					    btrfs_level_size(root, level - 1));
+				if (!cur || !btrfs_buffer_uptodate(cur, gen)) {
+					slot++;
+					if (cur)
+						free_extent_buffer(cur);
+					goto next;
+				}
+				free_extent_buffer(cur);
+			}
+			if (gen < min_trans) {
+				slot++;
+				goto next;
+			}
 			btrfs_node_key_to_cpu(c, key, slot);
+		}
 		return 0;
 	}
 	return 1;
@@ -3095,6 +3241,12 @@ done:
 	return 0;
 }
 
+/*
+ * this uses btrfs_prev_leaf to walk backwards in the tree, and keeps
+ * searching until it gets past min_objectid or finds an item of 'type'
+ *
+ * returns 0 if something is found, 1 if nothing was found and < 0 on error
+ */
 int btrfs_previous_item(struct btrfs_root *root,
 			struct btrfs_path *path, u64 min_objectid,
 			int type)
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index ca8e6f15859e..a28796482b4a 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -609,6 +609,7 @@ struct btrfs_root {
 	u64 last_inode_alloc;
 	int ref_cows;
 	int track_dirty;
+	u64 defrag_trans_start;
 	struct btrfs_key defrag_progress;
 	struct btrfs_key defrag_max;
 	int defrag_running;
@@ -1412,7 +1413,11 @@ int btrfs_previous_item(struct btrfs_root *root,
 struct extent_buffer *btrfs_root_node(struct btrfs_root *root);
 struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root);
 int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path,
-			struct btrfs_key *key, int lowest_level);
+			struct btrfs_key *key, int lowest_level,
+			int cache_only, u64 min_trans);
+int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key,
+			 struct btrfs_path *path, int cache_only,
+			 u64 min_trans);
 
 int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		    struct btrfs_root *root, struct extent_buffer *buf,
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 31ca9f89388d..4cdc0b6a2672 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -295,7 +295,6 @@ int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
 	ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE,
 					     btrfs_header_generation(eb));
 	BUG_ON(ret);
-	btrfs_clear_buffer_defrag(eb);
 	found_start = btrfs_header_bytenr(eb);
 	if (found_start != start) {
 		printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
@@ -355,7 +354,6 @@ int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
 	}
 	eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
 
-	btrfs_clear_buffer_defrag(eb);
 	found_start = btrfs_header_bytenr(eb);
 	if (found_start != start) {
 		ret = -EIO;
@@ -736,6 +734,7 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 	memset(&root->root_item, 0, sizeof(root->root_item));
 	memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
 	memset(&root->root_kobj, 0, sizeof(root->root_kobj));
+	root->defrag_trans_start = fs_info->generation;
 	init_completion(&root->kobj_unregister);
 	root->defrag_running = 0;
 	root->defrag_level = 0;
@@ -1168,7 +1167,6 @@ static int transaction_kthread(void *arg)
 			goto sleep;
 		}
 		mutex_unlock(&root->fs_info->trans_mutex);
-		btrfs_defrag_dirty_roots(root->fs_info);
 		trans = btrfs_start_transaction(root, 1);
 		ret = btrfs_commit_transaction(trans, root);
 sleep:
@@ -1434,12 +1432,12 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 						   tree_root,
 						   "btrfs-transaction");
 	if (!fs_info->transaction_kthread)
-		goto fail_trans_kthread;
+		goto fail_cleaner;
 
 
 	return tree_root;
 
-fail_trans_kthread:
+fail_cleaner:
 	kthread_stop(fs_info->cleaner_kthread);
 fail_extent_root:
 	free_extent_buffer(extent_root->node);
@@ -1662,7 +1660,6 @@ int close_ctree(struct btrfs_root *root)
 	kthread_stop(root->fs_info->transaction_kthread);
 	kthread_stop(root->fs_info->cleaner_kthread);
 
-	btrfs_defrag_dirty_roots(root->fs_info);
 	btrfs_clean_old_snapshots(root);
 	trans = btrfs_start_transaction(root, 1);
 	ret = btrfs_commit_transaction(trans, root);
@@ -1794,58 +1791,6 @@ void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
 	return;
 }
 
-void btrfs_set_buffer_defrag(struct extent_buffer *buf)
-{
-	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
-	struct inode *btree_inode = root->fs_info->btree_inode;
-	set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start,
-			buf->start + buf->len - 1, EXTENT_DEFRAG, GFP_NOFS);
-}
-
-void btrfs_set_buffer_defrag_done(struct extent_buffer *buf)
-{
-	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
-	struct inode *btree_inode = root->fs_info->btree_inode;
-	set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start,
-			buf->start + buf->len - 1, EXTENT_DEFRAG_DONE,
-			GFP_NOFS);
-}
-
-int btrfs_buffer_defrag(struct extent_buffer *buf)
-{
-	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
-	struct inode *btree_inode = root->fs_info->btree_inode;
-	return test_range_bit(&BTRFS_I(btree_inode)->io_tree,
-		     buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG, 0);
-}
-
-int btrfs_buffer_defrag_done(struct extent_buffer *buf)
-{
-	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
-	struct inode *btree_inode = root->fs_info->btree_inode;
-	return test_range_bit(&BTRFS_I(btree_inode)->io_tree,
-		     buf->start, buf->start + buf->len - 1,
-		     EXTENT_DEFRAG_DONE, 0);
-}
-
-int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf)
-{
-	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
-	struct inode *btree_inode = root->fs_info->btree_inode;
-	return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree,
-		     buf->start, buf->start + buf->len - 1,
-		     EXTENT_DEFRAG_DONE, GFP_NOFS);
-}
-
-int btrfs_clear_buffer_defrag(struct extent_buffer *buf)
-{
-	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
-	struct inode *btree_inode = root->fs_info->btree_inode;
-	return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree,
-		     buf->start, buf->start + buf->len - 1,
-		     EXTENT_DEFRAG, GFP_NOFS);
-}
-
 int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
 {
 	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index deff6b4815a7..353c3c50c957 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -61,12 +61,6 @@ int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid);
 int btrfs_set_buffer_uptodate(struct extent_buffer *buf);
 int wait_on_tree_block_writeback(struct btrfs_root *root,
 				 struct extent_buffer *buf);
-void btrfs_set_buffer_defrag(struct extent_buffer *buf);
-void btrfs_set_buffer_defrag_done(struct extent_buffer *buf);
-int btrfs_buffer_defrag(struct extent_buffer *buf);
-int btrfs_buffer_defrag_done(struct extent_buffer *buf);
-int btrfs_clear_buffer_defrag(struct extent_buffer *buf);
-int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf);
 int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid);
 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len);
 void btrfs_csum_final(u32 crc, char *result);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index dc3c03c6612d..5e0857ffbc35 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -2095,8 +2095,6 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 
 	set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
 			 buf->start + buf->len - 1, GFP_NOFS);
-	if (!btrfs_test_opt(root, SSD))
-		btrfs_set_buffer_defrag(buf);
 	trans->blocks_used++;
 	return buf;
 }
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 726d6871fa13..5e28cf5c2e85 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -365,7 +365,6 @@ int btrfs_sync_fs(struct super_block *sb, int wait)
 		return 0;
 	}
 	btrfs_clean_old_snapshots(root);
-	btrfs_defrag_dirty_roots(root->fs_info);
 	trans = btrfs_start_transaction(root, 1);
 	ret = btrfs_commit_transaction(trans, root);
 	sb->s_dirt = 0;
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 8e909cb97c6d..98f422d9ab07 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -30,7 +30,6 @@ extern struct kmem_cache *btrfs_trans_handle_cachep;
 extern struct kmem_cache *btrfs_transaction_cachep;
 
 #define BTRFS_ROOT_TRANS_TAG 0
-#define BTRFS_ROOT_DEFRAG_TAG 1
 
 static noinline void put_transaction(struct btrfs_transaction *transaction)
 {
@@ -92,9 +91,6 @@ static noinline int record_root_in_trans(struct btrfs_root *root)
 			radix_tree_tag_set(&root->fs_info->fs_roots_radix,
 				   (unsigned long)root->root_key.objectid,
 				   BTRFS_ROOT_TRANS_TAG);
-			radix_tree_tag_set(&root->fs_info->fs_roots_radix,
-				   (unsigned long)root->root_key.objectid,
-				   BTRFS_ROOT_DEFRAG_TAG);
 			root->commit_root = btrfs_root_node(root);
 		} else {
 			WARN_ON(1);
@@ -403,44 +399,15 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly)
 		cond_resched();
 
 		trans = btrfs_start_transaction(root, 1);
-		if (ret != -EAGAIN)
+		if (root->fs_info->closing || ret != -EAGAIN)
 			break;
 	}
 	root->defrag_running = 0;
 	smp_mb();
-	radix_tree_tag_clear(&info->fs_roots_radix,
-		     (unsigned long)root->root_key.objectid,
-		     BTRFS_ROOT_DEFRAG_TAG);
 	btrfs_end_transaction(trans, root);
 	return 0;
 }
 
-int btrfs_defrag_dirty_roots(struct btrfs_fs_info *info)
-{
-	struct btrfs_root *gang[1];
-	struct btrfs_root *root;
-	int i;
-	int ret;
-	int err = 0;
-	u64 last = 0;
-
-	while(1) {
-		ret = radix_tree_gang_lookup_tag(&info->fs_roots_radix,
-						 (void **)gang, last,
-						 ARRAY_SIZE(gang),
-						 BTRFS_ROOT_DEFRAG_TAG);
-		if (ret == 0)
-			break;
-		for (i = 0; i < ret; i++) {
-			root = gang[i];
-			last = root->root_key.objectid + 1;
-			btrfs_defrag_root(root, 1);
-		}
-	}
-	btrfs_defrag_root(info->extent_root, 1);
-	return err;
-}
-
 static noinline int drop_dirty_roots(struct btrfs_root *tree_root,
 				     struct list_head *list)
 {
diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h
index e1e5a06b65f4..9ccd5a5b170f 100644
--- a/fs/btrfs/transaction.h
+++ b/fs/btrfs/transaction.h
@@ -84,7 +84,6 @@ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
 
 int btrfs_add_dead_root(struct btrfs_root *root, struct btrfs_root *latest,
 			struct list_head *dead_list);
-int btrfs_defrag_dirty_roots(struct btrfs_fs_info *info);
 int btrfs_defrag_root(struct btrfs_root *root, int cacheonly);
 int btrfs_clean_old_snapshots(struct btrfs_root *root);
 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c
index b17693f61fbc..cc2650b06952 100644
--- a/fs/btrfs/tree-defrag.c
+++ b/fs/btrfs/tree-defrag.c
@@ -32,10 +32,13 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 	int wret;
 	int level;
 	int orig_level;
-	int i;
 	int is_extent = 0;
 	int next_key_ret = 0;
 	u64 last_ret = 0;
+	u64 min_trans = 0;
+
+	if (cache_only)
+		goto out;
 
 	if (root->fs_info->extent_root == root) {
 		/*
@@ -43,10 +46,6 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 		 * we can't defrag the extent root without deadlock
 		 */
 		goto out;
-#if 0
-		mutex_lock(&root->fs_info->alloc_mutex);
-		is_extent = 1;
-#endif
 	}
 
 	if (root->ref_cows == 0 && !is_extent)
@@ -84,6 +83,17 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 
 	path->lowest_level = 1;
 	path->keep_locks = 1;
+	if (cache_only)
+		min_trans = root->defrag_trans_start;
+
+	ret = btrfs_search_forward(root, &key, path, cache_only, min_trans);
+	if (ret < 0)
+		goto out;
+	if (ret > 0) {
+		ret = 0;
+		goto out;
+	}
+	btrfs_release_path(root, path);
 	wret = btrfs_search_slot(trans, root, &key, path, 0, 1);
 
 	if (wret < 0) {
@@ -95,7 +105,8 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 		goto out;
 	}
 	path->slots[1] = btrfs_header_nritems(path->nodes[1]);
-	next_key_ret = btrfs_find_next_key(root, path, &key, 1);
+	next_key_ret = btrfs_find_next_key(root, path, &key, 1, cache_only,
+					   min_trans);
 	ret = btrfs_realloc_node(trans, root,
 				 path->nodes[1], 0,
 				 cache_only, &last_ret,
@@ -106,19 +117,9 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 		ret = -EAGAIN;
 	}
 
-	for (i = 1; i < BTRFS_MAX_LEVEL; i++) {
-		if (path->locks[i]) {
-			btrfs_tree_unlock(path->nodes[i]);
-			path->locks[i] = 0;
-		}
-		if (path->nodes[i]) {
-			free_extent_buffer(path->nodes[i]);
-			path->nodes[i] = NULL;
-		}
-	}
+	btrfs_release_path(root, path);
 	if (is_extent)
 		btrfs_extent_post_op(trans, root);
-
 out:
 	if (is_extent)
 		mutex_unlock(&root->fs_info->alloc_mutex);
@@ -138,6 +139,7 @@ done:
 	if (ret != -EAGAIN) {
 		memset(&root->defrag_progress, 0,
 		       sizeof(root->defrag_progress));
+		root->defrag_trans_start = trans->transid;
 	}
 	return ret;
 }
-- 
cgit v1.2.3


From 777e6bd706ee40897545463871de5b456fbc46dc Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Fri, 15 Aug 2008 15:34:15 -0400
Subject: Btrfs: Transaction commit: don't use filemap_fdatawait

After writing out all the remaining btree blocks in the transaction,
the commit code would use filemap_fdatawait to make sure it was all
on disk.  This means it would wait for blocks written by other procs
as well.

The new code walks the list of blocks for this transaction again
and waits only for those required by this transaction.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/disk-io.c     | 11 ++++-------
 fs/btrfs/disk-io.h     |  1 +
 fs/btrfs/transaction.c | 36 +++++++++++++++++++++++++++++++-----
 3 files changed, 36 insertions(+), 12 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 9902d29abd06..9601b13c7d7a 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -429,7 +429,7 @@ int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
 	return 0;
 }
 
-static int congested_async(struct btrfs_fs_info *info, int iodone)
+int btrfs_congested_async(struct btrfs_fs_info *info, int iodone)
 {
 	int limit = 256 * info->fs_devices->open_devices;
 
@@ -438,9 +438,6 @@ static int congested_async(struct btrfs_fs_info *info, int iodone)
 	if (atomic_read(&info->nr_async_submits) > limit)
 		return 1;
 
-	limit = 8192 * info->fs_devices->open_devices;
-	if (iodone)
-		limit = (limit * 3) / 2;
 	return atomic_read(&info->nr_async_bios) > limit;
 }
 
@@ -454,7 +451,7 @@ static void run_one_async_submit(struct btrfs_work *work)
 	atomic_dec(&fs_info->nr_async_submits);
 
 	if ((async->bio->bi_rw & (1 << BIO_RW)) &&
-	    !congested_async(fs_info, 1)) {
+	    !btrfs_congested_async(fs_info, 1)) {
 		clear_bdi_congested(&fs_info->bdi, WRITE);
 	}
 	async->submit_bio_hook(async->inode, async->rw, async->bio,
@@ -963,7 +960,7 @@ static int btrfs_congested_fn(void *congested_data, int bdi_bits)
 	struct backing_dev_info *bdi;
 
 	if ((bdi_bits & (1 << BDI_write_congested)) &&
-	    congested_async(info, 0))
+	    btrfs_congested_async(info, 0))
 		return 1;
 
 	list_for_each(cur, &info->fs_devices->devices) {
@@ -1844,7 +1841,7 @@ void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
 	struct extent_io_tree *tree;
 	u64 num_dirty;
 	u64 start = 0;
-	unsigned long thresh = 16 * 1024 * 1024;
+	unsigned long thresh = 2 * 1024 * 1024;
 	tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
 
 	if (current_is_pdflush())
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 353c3c50c957..e904a69347a4 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -72,4 +72,5 @@ int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
 int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
 			int rw, struct bio *bio, int mirror_num,
 			extent_submit_bio_hook_t *submit_bio_hook);
+int btrfs_congested_async(struct btrfs_fs_info *info, int iodone);
 #endif
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 9d3d08e9f8d1..6bcb0876f9bb 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -303,12 +303,12 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
 				     struct btrfs_root *root)
 {
 	int ret;
-	int err;
+	int err = 0;
 	int werr = 0;
 	struct extent_io_tree *dirty_pages;
 	struct page *page;
 	struct inode *btree_inode = root->fs_info->btree_inode;
-	u64 start;
+	u64 start = 0;
 	u64 end;
 	unsigned long index;
 
@@ -317,12 +317,15 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
 	}
 	dirty_pages = &trans->transaction->dirty_pages;
 	while(1) {
-		ret = find_first_extent_bit(dirty_pages, 0, &start, &end,
+		ret = find_first_extent_bit(dirty_pages, start, &start, &end,
 					    EXTENT_DIRTY);
 		if (ret)
 			break;
-		clear_extent_dirty(dirty_pages, start, end, GFP_NOFS);
 		while(start <= end) {
+			if (btrfs_congested_async(root->fs_info, 0))
+				congestion_wait(WRITE, HZ/10);
+			cond_resched();
+
 			index = start >> PAGE_CACHE_SHIFT;
 			start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
 			page = find_lock_page(btree_inode->i_mapping, index);
@@ -343,7 +346,30 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
 			page_cache_release(page);
 		}
 	}
-	err = filemap_fdatawait(btree_inode->i_mapping);
+	while(1) {
+		ret = find_first_extent_bit(dirty_pages, 0, &start, &end,
+					    EXTENT_DIRTY);
+		if (ret)
+			break;
+
+		clear_extent_dirty(dirty_pages, start, end, GFP_NOFS);
+		while(start <= end) {
+			index = start >> PAGE_CACHE_SHIFT;
+			start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
+			page = find_get_page(btree_inode->i_mapping, index);
+			if (!page)
+				continue;
+			if (PageDirty(page)) {
+				lock_page(page);
+				err = write_one_page(page, 0);
+				if (err)
+					werr = err;
+			}
+			wait_on_page_writeback(page);
+			page_cache_release(page);
+			cond_resched();
+		}
+	}
 	if (err)
 		werr = err;
 	return werr;
-- 
cgit v1.2.3


From b64a2851ba25b3410a87d3d1b751155612105c8e Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Wed, 20 Aug 2008 13:39:41 -0400
Subject: Btrfs: Wait for async bio submissions to make some progress at queue
 time

Before, the btrfs bdi congestion function was used to test for too many
async bios.  This keeps that check to throttle pdflush, but also
adds a check while queuing bios.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/disk-io.c     | 16 +++++++++-------
 fs/btrfs/disk-io.h     |  1 +
 fs/btrfs/transaction.c |  2 --
 fs/btrfs/volumes.c     | 18 +++++++++++++++++-
 4 files changed, 27 insertions(+), 10 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 92e14dd9bddb..bbba14b629d2 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -429,7 +429,7 @@ int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
 	return 0;
 }
 
-static unsigned long async_submit_limit(struct btrfs_fs_info *info)
+unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
 {
 	unsigned long limit = min_t(unsigned long,
 				    info->workers.max_workers,
@@ -439,7 +439,8 @@ static unsigned long async_submit_limit(struct btrfs_fs_info *info)
 
 int btrfs_congested_async(struct btrfs_fs_info *info, int iodone)
 {
-	return atomic_read(&info->nr_async_bios) > async_submit_limit(info);
+	return atomic_read(&info->nr_async_bios) >
+		btrfs_async_submit_limit(info);
 }
 
 static void run_one_async_submit(struct btrfs_work *work)
@@ -451,12 +452,13 @@ static void run_one_async_submit(struct btrfs_work *work)
 	async = container_of(work, struct  async_submit_bio, work);
 	fs_info = BTRFS_I(async->inode)->root->fs_info;
 
-	limit = async_submit_limit(fs_info);
+	limit = btrfs_async_submit_limit(fs_info);
 	limit = limit * 2 / 3;
 
 	atomic_dec(&fs_info->nr_async_submits);
 
-	if (atomic_read(&fs_info->nr_async_submits) < limit)
+	if (atomic_read(&fs_info->nr_async_submits) < limit &&
+	    waitqueue_active(&fs_info->async_submit_wait))
 		wake_up(&fs_info->async_submit_wait);
 
 	async->submit_bio_hook(async->inode, async->rw, async->bio,
@@ -469,7 +471,7 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
 			extent_submit_bio_hook_t *submit_bio_hook)
 {
 	struct async_submit_bio *async;
-	int limit = async_submit_limit(fs_info);
+	int limit = btrfs_async_submit_limit(fs_info);
 
 	async = kmalloc(sizeof(*async), GFP_NOFS);
 	if (!async)
@@ -1863,10 +1865,10 @@ void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
 	struct extent_io_tree *tree;
 	u64 num_dirty;
 	u64 start = 0;
-	unsigned long thresh = 12 * 1024 * 1024;
+	unsigned long thresh = 96 * 1024 * 1024;
 	tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
 
-	if (current_is_pdflush())
+	if (current_is_pdflush() || current->flags & PF_MEMALLOC)
 		return;
 
 	num_dirty = count_range_bits(tree, &start, (u64)-1,
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index e904a69347a4..2562a273ae18 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -73,4 +73,5 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
 			int rw, struct bio *bio, int mirror_num,
 			extent_submit_bio_hook_t *submit_bio_hook);
 int btrfs_congested_async(struct btrfs_fs_info *info, int iodone);
+unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info);
 #endif
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 6bcb0876f9bb..eff3ad72991b 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -322,8 +322,6 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
 		if (ret)
 			break;
 		while(start <= end) {
-			if (btrfs_congested_async(root->fs_info, 0))
-				congestion_wait(WRITE, HZ/10);
 			cond_resched();
 
 			index = start >> PAGE_CACHE_SHIFT;
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 23a5b0aba00a..2652660e6079 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -138,12 +138,18 @@ int run_scheduled_bios(struct btrfs_device *device)
 {
 	struct bio *pending;
 	struct backing_dev_info *bdi;
+	struct btrfs_fs_info *fs_info;
 	struct bio *tail;
 	struct bio *cur;
 	int again = 0;
 	unsigned long num_run = 0;
+	unsigned long limit;
 
 	bdi = device->bdev->bd_inode->i_mapping->backing_dev_info;
+	fs_info = device->dev_root->fs_info;
+	limit = btrfs_async_submit_limit(fs_info);
+	limit = limit * 2 / 3;
+
 loop:
 	spin_lock(&device->io_lock);
 
@@ -179,7 +185,11 @@ loop:
 		cur = pending;
 		pending = pending->bi_next;
 		cur->bi_next = NULL;
-		atomic_dec(&device->dev_root->fs_info->nr_async_bios);
+		atomic_dec(&fs_info->nr_async_bios);
+
+		if (atomic_read(&fs_info->nr_async_bios) < limit &&
+		    waitqueue_active(&fs_info->async_submit_wait))
+			wake_up(&fs_info->async_submit_wait);
 
 		BUG_ON(atomic_read(&cur->bi_cnt) == 0);
 		bio_get(cur);
@@ -2135,6 +2145,7 @@ int schedule_bio(struct btrfs_root *root, struct btrfs_device *device,
 		 int rw, struct bio *bio)
 {
 	int should_queue = 1;
+	unsigned long limit;
 
 	/* don't bother with additional async steps for reads, right now */
 	if (!(rw & (1 << BIO_RW))) {
@@ -2171,6 +2182,11 @@ int schedule_bio(struct btrfs_root *root, struct btrfs_device *device,
 	if (should_queue)
 		btrfs_queue_worker(&root->fs_info->submit_workers,
 				   &device->work);
+
+	limit = btrfs_async_submit_limit(root->fs_info);
+	wait_event_timeout(root->fs_info->async_submit_wait,
+			   (atomic_read(&root->fs_info->nr_async_bios) < limit),
+			   HZ/10);
 	return 0;
 }
 
-- 
cgit v1.2.3


From e02119d5a7b4396c5a872582fddc8bd6d305a70a Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Fri, 5 Sep 2008 16:13:11 -0400
Subject: Btrfs: Add a write ahead tree log to optimize synchronous operations

File syncs and directory syncs are optimized by copying their
items into a special (copy-on-write) log tree.  There is one log tree per
subvolume and the btrfs super block points to a tree of log tree roots.

After a crash, items are copied out of the log tree and back into the
subvolume.  See tree-log.c for all the details.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/Makefile      |    3 +-
 fs/btrfs/btrfs_inode.h |    8 +
 fs/btrfs/compat.h      |   15 +
 fs/btrfs/ctree.c       |   73 +-
 fs/btrfs/ctree.h       |   72 +-
 fs/btrfs/dir-item.c    |    3 +
 fs/btrfs/disk-io.c     |  138 ++-
 fs/btrfs/disk-io.h     |    8 +-
 fs/btrfs/extent-tree.c |   93 +-
 fs/btrfs/file.c        |   39 +-
 fs/btrfs/inode.c       |  261 ++---
 fs/btrfs/root-tree.c   |    5 +-
 fs/btrfs/transaction.c |   45 +-
 fs/btrfs/transaction.h |    1 +
 fs/btrfs/tree-defrag.c |    4 +-
 fs/btrfs/tree-log.c    | 2804 ++++++++++++++++++++++++++++++++++++++++++++++++
 fs/btrfs/tree-log.h    |   41 +
 17 files changed, 3408 insertions(+), 205 deletions(-)
 create mode 100644 fs/btrfs/tree-log.c
 create mode 100644 fs/btrfs/tree-log.h

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 3a01065d4ef1..b7addbfd8c22 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -7,8 +7,7 @@ btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
 	   transaction.o bit-radix.o inode.o file.o tree-defrag.o \
 	   extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \
 	   extent_io.o volumes.o async-thread.o ioctl.o locking.o orphan.o \
-	   ref-cache.o acl.o export.o
-
+	   ref-cache.o export.o tree-log.o acl.o
 else
 
 # Normal Makefile
diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
index 111f90524ae1..fcc8cf27e906 100644
--- a/fs/btrfs/btrfs_inode.h
+++ b/fs/btrfs/btrfs_inode.h
@@ -33,6 +33,7 @@ struct btrfs_inode {
 	struct extent_io_tree io_failure_tree;
 	struct mutex csum_mutex;
 	struct mutex extent_mutex;
+	struct mutex log_mutex;
 	struct inode vfs_inode;
 	struct btrfs_ordered_inode_tree ordered_tree;
 
@@ -44,10 +45,17 @@ struct btrfs_inode {
 
 	struct list_head delalloc_inodes;
 
+	/* full 64 bit generation number */
+	u64 generation;
+
 	/*
 	 * transid of the trans_handle that last modified this inode
 	 */
 	u64 last_trans;
+	/*
+	 * transid that last logged this inode
+	 */
+	u64 logged_trans;
 	u64 delalloc_bytes;
 	u64 disk_i_size;
 	u32 flags;
diff --git a/fs/btrfs/compat.h b/fs/btrfs/compat.h
index d45fb37887bc..b0ed1887d9b1 100644
--- a/fs/btrfs/compat.h
+++ b/fs/btrfs/compat.h
@@ -22,6 +22,21 @@ static inline struct dentry *d_obtain_alias(struct inode *inode)
 }
 #endif
 
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
+static inline void btrfs_drop_nlink(struct inode *inode)
+{
+	inode->i_nlink--;
+}
+
+static inline void btrfs_inc_nlink(struct inode *inode)
+{
+	inode->i_nlink++;
+}
+#else
+# define btrfs_drop_nlink(inode) drop_nlink(inode)
+# define btrfs_inc_nlink(inode)	inc_nlink(inode)
+#endif
+
 /*
  * Even if AppArmor isn't enabled, it still has different prototypes.
  * Add more distro/version pairs here to declare which has AppArmor applied.
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 7114faafa9d4..579124043d9b 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -60,7 +60,7 @@ void btrfs_free_path(struct btrfs_path *p)
 	kmem_cache_free(btrfs_path_cachep, p);
 }
 
-void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p)
+void noinline btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p)
 {
 	int i;
 
@@ -176,7 +176,7 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans,
 	return 0;
 }
 
-int __btrfs_cow_block(struct btrfs_trans_handle *trans,
+int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root,
 			     struct extent_buffer *buf,
 			     struct extent_buffer *parent, int parent_slot,
@@ -294,7 +294,7 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 	return 0;
 }
 
-int btrfs_cow_block(struct btrfs_trans_handle *trans,
+int noinline btrfs_cow_block(struct btrfs_trans_handle *trans,
 		    struct btrfs_root *root, struct extent_buffer *buf,
 		    struct extent_buffer *parent, int parent_slot,
 		    struct extent_buffer **cow_ret, u64 prealloc_dest)
@@ -677,9 +677,10 @@ static int noinline check_block(struct btrfs_root *root,
  *
  * slot may point to max if the key is bigger than all of the keys
  */
-static int generic_bin_search(struct extent_buffer *eb, unsigned long p,
-			      int item_size, struct btrfs_key *key,
-			      int max, int *slot)
+static noinline int generic_bin_search(struct extent_buffer *eb,
+				       unsigned long p,
+				       int item_size, struct btrfs_key *key,
+				       int max, int *slot)
 {
 	int low = 0;
 	int high = max;
@@ -765,7 +766,7 @@ static int bin_search(struct extent_buffer *eb, struct btrfs_key *key,
 	return -1;
 }
 
-static struct extent_buffer *read_node_slot(struct btrfs_root *root,
+static noinline struct extent_buffer *read_node_slot(struct btrfs_root *root,
 				   struct extent_buffer *parent, int slot)
 {
 	int level = btrfs_header_level(parent);
@@ -781,7 +782,7 @@ static struct extent_buffer *read_node_slot(struct btrfs_root *root,
 		       btrfs_node_ptr_generation(parent, slot));
 }
 
-static int balance_level(struct btrfs_trans_handle *trans,
+static noinline int balance_level(struct btrfs_trans_handle *trans,
 			 struct btrfs_root *root,
 			 struct btrfs_path *path, int level)
 {
@@ -1128,8 +1129,9 @@ static int noinline push_nodes_for_insert(struct btrfs_trans_handle *trans,
 /*
  * readahead one full node of leaves
  */
-static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
-			     int level, int slot, u64 objectid)
+static noinline void reada_for_search(struct btrfs_root *root,
+				      struct btrfs_path *path,
+				      int level, int slot, u64 objectid)
 {
 	struct extent_buffer *node;
 	struct btrfs_disk_key disk_key;
@@ -1201,7 +1203,8 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
 	}
 }
 
-static void unlock_up(struct btrfs_path *path, int level, int lowest_unlock)
+static noinline void unlock_up(struct btrfs_path *path, int level,
+			       int lowest_unlock)
 {
 	int i;
 	int skip_level = level;
@@ -1759,8 +1762,9 @@ static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root
  *
  * returns 0 on success and < 0 on failure
  */
-static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, struct btrfs_path *path, int level)
+static noinline int split_node(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root,
+			       struct btrfs_path *path, int level)
 {
 	u64 root_gen;
 	struct extent_buffer *c;
@@ -1874,7 +1878,8 @@ static int leaf_space_used(struct extent_buffer *l, int start, int nr)
  * the start of the leaf data.  IOW, how much room
  * the leaf has left for both items and data
  */
-int btrfs_leaf_free_space(struct btrfs_root *root, struct extent_buffer *leaf)
+int noinline btrfs_leaf_free_space(struct btrfs_root *root,
+				   struct extent_buffer *leaf)
 {
 	int nritems = btrfs_header_nritems(leaf);
 	int ret;
@@ -2283,9 +2288,11 @@ out:
  *
  * returns 0 if all went well and < 0 on failure.
  */
-static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, struct btrfs_key *ins_key,
-		      struct btrfs_path *path, int data_size, int extend)
+static noinline int split_leaf(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root,
+			       struct btrfs_key *ins_key,
+			       struct btrfs_path *path, int data_size,
+			       int extend)
 {
 	u64 root_gen;
 	struct extent_buffer *l;
@@ -3079,6 +3086,7 @@ int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path)
  * was nothing in the tree that matched the search criteria.
  */
 int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key,
+			 struct btrfs_key *max_key,
 			 struct btrfs_path *path, int cache_only,
 			 u64 min_trans)
 {
@@ -3093,6 +3101,7 @@ int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key,
 again:
 	cur = btrfs_lock_root_node(root);
 	level = btrfs_header_level(cur);
+	WARN_ON(path->nodes[level]);
 	path->nodes[level] = cur;
 	path->locks[level] = 1;
 
@@ -3107,6 +3116,8 @@ again:
 
 		/* at level = 0, we're done, setup the path and exit */
 		if (level == 0) {
+			if (slot >= nritems)
+				goto find_next_key;
 			ret = 0;
 			path->slots[level] = slot;
 			btrfs_item_key_to_cpu(cur, &found_key, slot);
@@ -3123,6 +3134,8 @@ again:
 			u64 blockptr;
 			u64 gen;
 			struct extent_buffer *tmp;
+			struct btrfs_disk_key disk_key;
+
 			blockptr = btrfs_node_blockptr(cur, slot);
 			gen = btrfs_node_ptr_generation(cur, slot);
 			if (gen < min_trans) {
@@ -3132,6 +3145,14 @@ again:
 			if (!cache_only)
 				break;
 
+			if (max_key) {
+				btrfs_node_key(cur, &disk_key, slot);
+				if (comp_keys(&disk_key, max_key) >= 0) {
+					ret = 1;
+					goto out;
+				}
+			}
+
 			tmp = btrfs_find_tree_block(root, blockptr,
 					    btrfs_level_size(root, level - 1));
 
@@ -3143,14 +3164,16 @@ again:
 				free_extent_buffer(tmp);
 			slot++;
 		}
+find_next_key:
 		/*
 		 * we didn't find a candidate key in this node, walk forward
 		 * and find another one
 		 */
 		if (slot >= nritems) {
-			ret = btrfs_find_next_key(root, path, min_key, level,
+			path->slots[level] = slot;
+			sret = btrfs_find_next_key(root, path, min_key, level,
 						  cache_only, min_trans);
-			if (ret == 0) {
+			if (sret == 0) {
 				btrfs_release_path(root, path);
 				goto again;
 			} else {
@@ -3351,6 +3374,7 @@ int btrfs_previous_item(struct btrfs_root *root,
 {
 	struct btrfs_key found_key;
 	struct extent_buffer *leaf;
+	u32 nritems;
 	int ret;
 
 	while(1) {
@@ -3362,9 +3386,20 @@ int btrfs_previous_item(struct btrfs_root *root,
 			path->slots[0]--;
 		}
 		leaf = path->nodes[0];
+		nritems = btrfs_header_nritems(leaf);
+		if (nritems == 0)
+			return 1;
+		if (path->slots[0] == nritems)
+			path->slots[0]--;
+
 		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
 		if (found_key.type == type)
 			return 0;
+		if (found_key.objectid < min_objectid)
+			break;
+		if (found_key.objectid == min_objectid &&
+		    found_key.type < type)
+			break;
 	}
 	return 1;
 }
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index b305ae7e10b0..6532b60683ef 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -77,6 +77,10 @@ struct btrfs_ordered_sum;
 /* orhpan objectid for tracking unlinked/truncated files */
 #define BTRFS_ORPHAN_OBJECTID -5ULL
 
+/* does write ahead logging to speed up fsyncs */
+#define BTRFS_TREE_LOG_OBJECTID -6ULL
+#define BTRFS_TREE_LOG_FIXUP_OBJECTID -7ULL
+
 /*
  * All files have objectids higher than this.
  */
@@ -276,6 +280,7 @@ struct btrfs_super_block {
 	__le64 generation;
 	__le64 root;
 	__le64 chunk_root;
+	__le64 log_root;
 	__le64 total_bytes;
 	__le64 bytes_used;
 	__le64 root_dir_objectid;
@@ -287,6 +292,7 @@ struct btrfs_super_block {
 	__le32 sys_chunk_array_size;
 	u8 root_level;
 	u8 chunk_root_level;
+	u8 log_root_level;
 	struct btrfs_dev_item dev_item;
 	char label[BTRFS_LABEL_SIZE];
 	u8 sys_chunk_array[BTRFS_SYSTEM_CHUNK_ARRAY_SIZE];
@@ -392,7 +398,10 @@ struct btrfs_timespec {
  * make a new item type
  */
 struct btrfs_inode_item {
+	/* nfs style generation number */
 	__le64 generation;
+	/* transid that last touched this inode */
+	__le64 transid;
 	__le64 size;
 	__le64 nblocks;
 	__le64 block_group;
@@ -409,8 +418,13 @@ struct btrfs_inode_item {
 	struct btrfs_timespec otime;
 } __attribute__ ((__packed__));
 
+struct btrfs_dir_log_item {
+	__le64 end;
+} __attribute__ ((__packed__));
+
 struct btrfs_dir_item {
 	struct btrfs_disk_key location;
+	__le64 transid;
 	__le16 data_len;
 	__le16 name_len;
 	u8 type;
@@ -505,6 +519,9 @@ struct btrfs_fs_info {
 	struct btrfs_root *tree_root;
 	struct btrfs_root *chunk_root;
 	struct btrfs_root *dev_root;
+
+	/* the log root tree is a directory of all the other log roots */
+	struct btrfs_root *log_root_tree;
 	struct radix_tree_root fs_roots_radix;
 
 	struct extent_io_tree free_space_cache;
@@ -518,6 +535,7 @@ struct btrfs_fs_info {
 
 	u64 generation;
 	u64 last_trans_committed;
+	u64 last_trans_new_blockgroup;
 	u64 open_ioctl_trans;
 	unsigned long mount_opt;
 	u64 max_extent;
@@ -527,6 +545,9 @@ struct btrfs_fs_info {
 	wait_queue_head_t transaction_throttle;
 	wait_queue_head_t transaction_wait;
 	wait_queue_head_t async_submit_wait;
+
+	wait_queue_head_t tree_log_wait;
+
 	struct btrfs_super_block super_copy;
 	struct btrfs_super_block super_for_commit;
 	struct block_device *__bdev;
@@ -535,6 +556,7 @@ struct btrfs_fs_info {
 	struct backing_dev_info bdi;
 	spinlock_t hash_lock;
 	struct mutex trans_mutex;
+	struct mutex tree_log_mutex;
 	struct mutex transaction_kthread_mutex;
 	struct mutex cleaner_mutex;
 	struct mutex alloc_mutex;
@@ -544,8 +566,13 @@ struct btrfs_fs_info {
 	struct list_head trans_list;
 	struct list_head hashers;
 	struct list_head dead_roots;
+
 	atomic_t nr_async_submits;
 	atomic_t nr_async_bios;
+	atomic_t tree_log_writers;
+	atomic_t tree_log_commit;
+	unsigned long tree_log_batch;
+	u64 tree_log_transid;
 
 	/*
 	 * this is used by the balancing code to wait for all the pending
@@ -583,6 +610,7 @@ struct btrfs_fs_info {
 	struct completion kobj_unregister;
 	int do_barriers;
 	int closing;
+	int log_root_recovering;
 	atomic_t throttles;
 	atomic_t throttle_gen;
 
@@ -596,6 +624,7 @@ struct btrfs_fs_info {
 	u64 delalloc_bytes;
 	u64 last_alloc;
 	u64 last_data_alloc;
+	u64 last_log_alloc;
 
 	spinlock_t ref_cache_lock;
 	u64 total_ref_cache_size;
@@ -632,6 +661,7 @@ struct btrfs_root {
 	struct btrfs_leaf_ref_tree *ref_tree;
 	struct btrfs_leaf_ref_tree ref_tree_struct;
 	struct btrfs_dirty_root *dirty_root;
+	struct btrfs_root *log_root;
 
 	struct btrfs_root_item root_item;
 	struct btrfs_key root_key;
@@ -640,6 +670,7 @@ struct btrfs_root {
 	struct kobject root_kobj;
 	struct completion kobj_unregister;
 	struct mutex objectid_mutex;
+	struct mutex log_mutex;
 
 	u64 objectid;
 	u64 last_trans;
@@ -692,6 +723,8 @@ struct btrfs_root {
  * dir items are the name -> inode pointers in a directory.  There is one
  * for every name in a directory.
  */
+#define BTRFS_DIR_LOG_ITEM_KEY  14
+#define BTRFS_DIR_LOG_INDEX_KEY 15
 #define BTRFS_DIR_ITEM_KEY	16
 #define BTRFS_DIR_INDEX_KEY	17
 /*
@@ -703,7 +736,8 @@ struct btrfs_root {
  */
 #define BTRFS_CSUM_ITEM_KEY	19
 
-/* reserve 20-31 for other file stuff */
+
+/* reserve 21-31 for other file/dir stuff */
 
 /*
  * root items point to tree roots.  There are typically in the root
@@ -938,6 +972,7 @@ BTRFS_SETGET_FUNCS(inode_ref_index, struct btrfs_inode_ref, index, 64);
 
 /* struct btrfs_inode_item */
 BTRFS_SETGET_FUNCS(inode_generation, struct btrfs_inode_item, generation, 64);
+BTRFS_SETGET_FUNCS(inode_transid, struct btrfs_inode_item, transid, 64);
 BTRFS_SETGET_FUNCS(inode_size, struct btrfs_inode_item, size, 64);
 BTRFS_SETGET_FUNCS(inode_nblocks, struct btrfs_inode_item, nblocks, 64);
 BTRFS_SETGET_FUNCS(inode_block_group, struct btrfs_inode_item, block_group, 64);
@@ -1126,10 +1161,13 @@ static inline void btrfs_set_item_key(struct extent_buffer *eb,
 	write_eb_member(eb, item, struct btrfs_item, key, disk_key);
 }
 
+BTRFS_SETGET_FUNCS(dir_log_end, struct btrfs_dir_log_item, end, 64);
+
 /* struct btrfs_dir_item */
 BTRFS_SETGET_FUNCS(dir_data_len, struct btrfs_dir_item, data_len, 16);
 BTRFS_SETGET_FUNCS(dir_type, struct btrfs_dir_item, type, 8);
 BTRFS_SETGET_FUNCS(dir_name_len, struct btrfs_dir_item, name_len, 16);
+BTRFS_SETGET_FUNCS(dir_transid, struct btrfs_dir_item, transid, 64);
 
 static inline void btrfs_dir_item_key(struct extent_buffer *eb,
 				      struct btrfs_dir_item *item,
@@ -1301,7 +1339,11 @@ BTRFS_SETGET_STACK_FUNCS(super_root_level, struct btrfs_super_block,
 BTRFS_SETGET_STACK_FUNCS(super_chunk_root, struct btrfs_super_block,
 			 chunk_root, 64);
 BTRFS_SETGET_STACK_FUNCS(super_chunk_root_level, struct btrfs_super_block,
-			 chunk_root_level, 64);
+			 chunk_root_level, 8);
+BTRFS_SETGET_STACK_FUNCS(super_log_root, struct btrfs_super_block,
+			 log_root, 64);
+BTRFS_SETGET_STACK_FUNCS(super_log_root_level, struct btrfs_super_block,
+			 log_root_level, 8);
 BTRFS_SETGET_STACK_FUNCS(super_total_bytes, struct btrfs_super_block,
 			 total_bytes, 64);
 BTRFS_SETGET_STACK_FUNCS(super_bytes_used, struct btrfs_super_block,
@@ -1405,6 +1447,12 @@ static inline struct dentry *fdentry(struct file *file) {
 }
 
 /* extent-tree.c */
+int btrfs_lookup_extent(struct btrfs_root *root, struct btrfs_path *path,
+			u64 start, u64 len);
+int btrfs_update_pinned_extents(struct btrfs_root *root,
+				u64 bytenr, u64 num, int pin);
+int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans,
+			struct btrfs_root *root, struct extent_buffer *leaf);
 int btrfs_cross_ref_exists(struct btrfs_trans_handle *trans,
 			   struct btrfs_root *root,
 			   struct btrfs_key *key, u64 bytenr);
@@ -1448,6 +1496,11 @@ int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans,
 				u64 root_objectid, u64 ref_generation,
 				u64 owner, u64 owner_offset,
 				struct btrfs_key *ins);
+int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans,
+				struct btrfs_root *root,
+				u64 root_objectid, u64 ref_generation,
+				u64 owner, u64 owner_offset,
+				struct btrfs_key *ins);
 int btrfs_reserve_extent(struct btrfs_trans_handle *trans,
 				  struct btrfs_root *root,
 				  u64 num_bytes, u64 min_alloc_size,
@@ -1488,9 +1541,9 @@ int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path,
 			struct btrfs_key *key, int lowest_level,
 			int cache_only, u64 min_trans);
 int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key,
+			 struct btrfs_key *max_key,
 			 struct btrfs_path *path, int cache_only,
 			 u64 min_trans);
-
 int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		    struct btrfs_root *root, struct extent_buffer *buf,
 		    struct extent_buffer *parent, int parent_slot,
@@ -1656,6 +1709,18 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans,
 #define PageChecked PageFsMisc
 #endif
 
+int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
+		       struct btrfs_root *root,
+		       struct inode *dir, struct inode *inode,
+		       const char *name, int name_len);
+int btrfs_add_link(struct btrfs_trans_handle *trans,
+		   struct inode *parent_inode, struct inode *inode,
+		   const char *name, int name_len, int add_backref, u64 index);
+int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root,
+			       struct inode *inode, u64 new_size,
+			       u32 min_type);
+
 int btrfs_start_delalloc_inodes(struct btrfs_root *root);
 int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end);
 int btrfs_writepages(struct address_space *mapping,
@@ -1715,6 +1780,7 @@ int btrfs_update_inode(struct btrfs_trans_handle *trans,
 long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 
 /* file.c */
+int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync);
 int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end);
 int btrfs_check_file(struct btrfs_root *root, struct inode *inode);
 extern struct file_operations btrfs_file_operations;
diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c
index 125094617fe8..e4f30090d640 100644
--- a/fs/btrfs/dir-item.c
+++ b/fs/btrfs/dir-item.c
@@ -96,6 +96,7 @@ int btrfs_insert_xattr_item(struct btrfs_trans_handle *trans,
 	btrfs_set_dir_item_key(leaf, dir_item, &disk_key);
 	btrfs_set_dir_type(leaf, dir_item, BTRFS_FT_XATTR);
 	btrfs_set_dir_name_len(leaf, dir_item, name_len);
+	btrfs_set_dir_transid(leaf, dir_item, trans->transid);
 	btrfs_set_dir_data_len(leaf, dir_item, data_len);
 	name_ptr = (unsigned long)(dir_item + 1);
 	data_ptr = (unsigned long)((char *)name_ptr + name_len);
@@ -142,6 +143,7 @@ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_set_dir_type(leaf, dir_item, type);
 	btrfs_set_dir_data_len(leaf, dir_item, 0);
 	btrfs_set_dir_name_len(leaf, dir_item, name_len);
+	btrfs_set_dir_transid(leaf, dir_item, trans->transid);
 	name_ptr = (unsigned long)(dir_item + 1);
 
 	write_extent_buffer(leaf, name, name_ptr, name_len);
@@ -169,6 +171,7 @@ second_insert:
 	btrfs_set_dir_type(leaf, dir_item, type);
 	btrfs_set_dir_data_len(leaf, dir_item, 0);
 	btrfs_set_dir_name_len(leaf, dir_item, name_len);
+	btrfs_set_dir_transid(leaf, dir_item, trans->transid);
 	name_ptr = (unsigned long)(dir_item + 1);
 	write_extent_buffer(leaf, name, name_ptr, name_len);
 	btrfs_mark_buffer_dirty(leaf);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 8e7a938bfbc7..a4373db5967a 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -41,6 +41,7 @@
 #include "async-thread.h"
 #include "locking.h"
 #include "ref-cache.h"
+#include "tree-log.h"
 
 #if 0
 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
@@ -694,6 +695,18 @@ struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
 }
 
 
+int btrfs_write_tree_block(struct extent_buffer *buf)
+{
+	return btrfs_fdatawrite_range(buf->first_page->mapping, buf->start,
+				      buf->start + buf->len - 1, WB_SYNC_NONE);
+}
+
+int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
+{
+	return btrfs_wait_on_page_writeback_range(buf->first_page->mapping,
+				  buf->start, buf->start + buf->len -1);
+}
+
 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
 				      u32 blocksize, u64 parent_transid)
 {
@@ -732,15 +745,6 @@ int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	return 0;
 }
 
-int wait_on_tree_block_writeback(struct btrfs_root *root,
-				 struct extent_buffer *buf)
-{
-	struct inode *btree_inode = root->fs_info->btree_inode;
-	wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->io_tree,
-					buf);
-	return 0;
-}
-
 static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 			u32 stripesize, struct btrfs_root *root,
 			struct btrfs_fs_info *fs_info,
@@ -771,6 +775,7 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 	spin_lock_init(&root->node_lock);
 	spin_lock_init(&root->list_lock);
 	mutex_init(&root->objectid_mutex);
+	mutex_init(&root->log_mutex);
 
 	btrfs_leaf_ref_tree_init(&root->ref_tree_struct);
 	root->ref_tree = &root->ref_tree_struct;
@@ -809,11 +814,74 @@ static int find_and_setup_root(struct btrfs_root *tree_root,
 	return 0;
 }
 
-struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
-					       struct btrfs_key *location)
+int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans,
+			     struct btrfs_fs_info *fs_info)
+{
+	struct extent_buffer *eb;
+	int ret;
+
+	if (!fs_info->log_root_tree)
+		return 0;
+
+	eb = fs_info->log_root_tree->node;
+
+	WARN_ON(btrfs_header_level(eb) != 0);
+	WARN_ON(btrfs_header_nritems(eb) != 0);
+
+	ret = btrfs_free_extent(trans, fs_info->tree_root,
+				eb->start, eb->len,
+				BTRFS_TREE_LOG_OBJECTID, 0, 0, 0, 1);
+	BUG_ON(ret);
+
+	free_extent_buffer(eb);
+	kfree(fs_info->log_root_tree);
+	fs_info->log_root_tree = NULL;
+	return 0;
+}
+
+int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
+			     struct btrfs_fs_info *fs_info)
 {
 	struct btrfs_root *root;
 	struct btrfs_root *tree_root = fs_info->tree_root;
+
+	root = kzalloc(sizeof(*root), GFP_NOFS);
+	if (!root)
+		return -ENOMEM;
+
+	__setup_root(tree_root->nodesize, tree_root->leafsize,
+		     tree_root->sectorsize, tree_root->stripesize,
+		     root, fs_info, BTRFS_TREE_LOG_OBJECTID);
+
+	root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
+	root->root_key.type = BTRFS_ROOT_ITEM_KEY;
+	root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
+	root->ref_cows = 0;
+
+	root->node = btrfs_alloc_free_block(trans, root, root->leafsize,
+					    BTRFS_TREE_LOG_OBJECTID,
+					    0, 0, 0, 0, 0);
+
+	btrfs_set_header_nritems(root->node, 0);
+	btrfs_set_header_level(root->node, 0);
+	btrfs_set_header_bytenr(root->node, root->node->start);
+	btrfs_set_header_generation(root->node, trans->transid);
+	btrfs_set_header_owner(root->node, BTRFS_TREE_LOG_OBJECTID);
+
+	write_extent_buffer(root->node, root->fs_info->fsid,
+			    (unsigned long)btrfs_header_fsid(root->node),
+			    BTRFS_FSID_SIZE);
+	btrfs_mark_buffer_dirty(root->node);
+	btrfs_tree_unlock(root->node);
+	fs_info->log_root_tree = root;
+	return 0;
+}
+
+struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
+					       struct btrfs_key *location)
+{
+	struct btrfs_root *root;
+	struct btrfs_fs_info *fs_info = tree_root->fs_info;
 	struct btrfs_path *path;
 	struct extent_buffer *l;
 	u64 highest_inode;
@@ -863,11 +931,13 @@ out:
 				     blocksize, 0);
 	BUG_ON(!root->node);
 insert:
-	root->ref_cows = 1;
-	ret = btrfs_find_highest_inode(root, &highest_inode);
-	if (ret == 0) {
-		root->highest_inode = highest_inode;
-		root->last_inode_alloc = highest_inode;
+	if (location->objectid != BTRFS_TREE_LOG_OBJECTID) {
+		root->ref_cows = 1;
+		ret = btrfs_find_highest_inode(root, &highest_inode);
+		if (ret == 0) {
+			root->highest_inode = highest_inode;
+			root->last_inode_alloc = highest_inode;
+		}
 	}
 	return root;
 }
@@ -907,7 +977,7 @@ struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
 	if (root)
 		return root;
 
-	root = btrfs_read_fs_root_no_radix(fs_info, location);
+	root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
 	if (IS_ERR(root))
 		return root;
 	ret = radix_tree_insert(&fs_info->fs_roots_radix,
@@ -1250,16 +1320,18 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	u32 blocksize;
 	u32 stripesize;
 	struct buffer_head *bh;
-	struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
+	struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root),
 						 GFP_NOFS);
-	struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
+	struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root),
 					       GFP_NOFS);
 	struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info),
 						GFP_NOFS);
-	struct btrfs_root *chunk_root = kmalloc(sizeof(struct btrfs_root),
+	struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root),
 						GFP_NOFS);
-	struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root),
+	struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root),
 					      GFP_NOFS);
+	struct btrfs_root *log_tree_root;
+
 	int ret;
 	int err = -EINVAL;
 
@@ -1343,6 +1415,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
 
 	mutex_init(&fs_info->trans_mutex);
+	mutex_init(&fs_info->tree_log_mutex);
 	mutex_init(&fs_info->drop_mutex);
 	mutex_init(&fs_info->alloc_mutex);
 	mutex_init(&fs_info->chunk_mutex);
@@ -1352,6 +1425,10 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	init_waitqueue_head(&fs_info->transaction_throttle);
 	init_waitqueue_head(&fs_info->transaction_wait);
 	init_waitqueue_head(&fs_info->async_submit_wait);
+	init_waitqueue_head(&fs_info->tree_log_wait);
+	atomic_set(&fs_info->tree_log_commit, 0);
+	atomic_set(&fs_info->tree_log_writers, 0);
+	fs_info->tree_log_transid = 0;
 
 #if 0
 	ret = add_hasher(fs_info, "crc32c");
@@ -1532,7 +1609,26 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	if (!fs_info->transaction_kthread)
 		goto fail_cleaner;
 
+	if (btrfs_super_log_root(disk_super) != 0) {
+		u32 blocksize;
+		u64 bytenr = btrfs_super_log_root(disk_super);
+
+		blocksize =
+		     btrfs_level_size(tree_root,
+				      btrfs_super_log_root_level(disk_super));
 
+		log_tree_root = kzalloc(sizeof(struct btrfs_root),
+						      GFP_NOFS);
+
+		__setup_root(nodesize, leafsize, sectorsize, stripesize,
+			     log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
+
+		log_tree_root->node = read_tree_block(tree_root, bytenr,
+						      blocksize, 0);
+		ret = btrfs_recover_log_trees(log_tree_root);
+		BUG_ON(ret);
+	}
+	fs_info->last_trans_committed = btrfs_super_generation(disk_super);
 	return tree_root;
 
 fail_cleaner:
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 2562a273ae18..6b6fdc697f31 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -45,7 +45,7 @@ struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
 				      struct btrfs_key *location,
 				      const char *name, int namelen);
-struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
+struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
 					       struct btrfs_key *location);
 struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
 					      struct btrfs_key *location);
@@ -74,4 +74,10 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
 			extent_submit_bio_hook_t *submit_bio_hook);
 int btrfs_congested_async(struct btrfs_fs_info *info, int iodone);
 unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info);
+int btrfs_write_tree_block(struct extent_buffer *buf);
+int btrfs_wait_tree_block_writeback(struct extent_buffer *buf);
+int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans,
+			     struct btrfs_fs_info *fs_info);
+int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
+			     struct btrfs_fs_info *fs_info);
 #endif
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index e63b3b4bed7c..646b9148ca21 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -496,6 +496,23 @@ static int match_extent_ref(struct extent_buffer *leaf,
 	return ret == 0;
 }
 
+/* simple helper to search for an existing extent at a given offset */
+int btrfs_lookup_extent(struct btrfs_root *root, struct btrfs_path *path,
+			u64 start, u64 len)
+{
+	int ret;
+	struct btrfs_key key;
+
+	maybe_lock_mutex(root);
+	key.objectid = start;
+	key.offset = len;
+	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
+	ret = btrfs_search_slot(NULL, root->fs_info->extent_root, &key, path,
+				0, 0);
+	maybe_unlock_mutex(root);
+	return ret;
+}
+
 static int noinline lookup_extent_backref(struct btrfs_trans_handle *trans,
 					  struct btrfs_root *root,
 					  struct btrfs_path *path, u64 bytenr,
@@ -1409,7 +1426,7 @@ static u64 first_logical_byte(struct btrfs_root *root, u64 search_start)
 }
 
 
-static int update_pinned_extents(struct btrfs_root *root,
+int btrfs_update_pinned_extents(struct btrfs_root *root,
 				u64 bytenr, u64 num, int pin)
 {
 	u64 len;
@@ -1492,7 +1509,7 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
 					    EXTENT_DIRTY);
 		if (ret)
 			break;
-		update_pinned_extents(root, start, end + 1 - start, 0);
+		btrfs_update_pinned_extents(root, start, end + 1 - start, 0);
 		clear_extent_dirty(unpin, start, end, GFP_NOFS);
 		set_extent_dirty(free_space_cache, start, end, GFP_NOFS);
 		if (need_resched()) {
@@ -1538,14 +1555,11 @@ static int finish_current_insert(struct btrfs_trans_handle *trans,
 		clear_extent_bits(&info->extent_ins, start, end, EXTENT_LOCKED,
 				  GFP_NOFS);
 
-		eb = btrfs_find_tree_block(extent_root, ins.objectid,
+		eb = btrfs_find_create_tree_block(extent_root, ins.objectid,
 					   ins.offset);
 
-		if (!btrfs_buffer_uptodate(eb, trans->transid)) {
-			mutex_unlock(&extent_root->fs_info->alloc_mutex);
+		if (!btrfs_buffer_uptodate(eb, trans->transid))
 			btrfs_read_buffer(eb, trans->transid);
-			mutex_lock(&extent_root->fs_info->alloc_mutex);
-		}
 
 		btrfs_tree_lock(eb);
 		level = btrfs_header_level(eb);
@@ -1585,13 +1599,20 @@ static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes,
 		struct extent_buffer *buf;
 		buf = btrfs_find_tree_block(root, bytenr, num_bytes);
 		if (buf) {
+			/* we can reuse a block if it hasn't been written
+			 * and it is from this transaction.  We can't
+			 * reuse anything from the tree log root because
+			 * it has tiny sub-transactions.
+			 */
 			if (btrfs_buffer_uptodate(buf, 0) &&
 			    btrfs_try_tree_lock(buf)) {
 				u64 transid =
 				    root->fs_info->running_transaction->transid;
 				u64 header_transid =
 					btrfs_header_generation(buf);
-				if (header_transid == transid &&
+				if (btrfs_header_owner(buf) !=
+				    BTRFS_TREE_LOG_OBJECTID &&
+				    header_transid == transid &&
 				    !btrfs_header_flag(buf,
 					       BTRFS_HEADER_FLAG_WRITTEN)) {
 					clean_tree_block(NULL, root, buf);
@@ -1603,7 +1624,7 @@ static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes,
 			}
 			free_extent_buffer(buf);
 		}
-		update_pinned_extents(root, bytenr, num_bytes, 1);
+		btrfs_update_pinned_extents(root, bytenr, num_bytes, 1);
 	} else {
 		set_extent_bits(&root->fs_info->pending_del,
 				bytenr, bytenr + num_bytes - 1,
@@ -1801,7 +1822,7 @@ static int del_pending_extents(struct btrfs_trans_handle *trans, struct
 				  GFP_NOFS);
 		if (!test_range_bit(&extent_root->fs_info->extent_ins,
 				    start, end, EXTENT_LOCKED, 0)) {
-			update_pinned_extents(extent_root, start,
+			btrfs_update_pinned_extents(extent_root, start,
 					      end + 1 - start, 1);
 			ret = __free_extent(trans, extent_root,
 					     start, end + 1 - start,
@@ -1919,6 +1940,12 @@ static int noinline find_free_extent(struct btrfs_trans_handle *trans,
 	if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) {
 		last_ptr = &root->fs_info->last_data_alloc;
 	}
+	if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
+		last_ptr = &root->fs_info->last_log_alloc;
+		if (!last_ptr == 0 && root->fs_info->last_alloc) {
+			*last_ptr = root->fs_info->last_alloc + empty_cluster;
+		}
+	}
 
 	if (last_ptr) {
 		if (*last_ptr)
@@ -2268,6 +2295,35 @@ int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans,
 	maybe_unlock_mutex(root);
 	return ret;
 }
+
+/*
+ * this is used by the tree logging recovery code.  It records that
+ * an extent has been allocated and makes sure to clear the free
+ * space cache bits as well
+ */
+int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans,
+				struct btrfs_root *root,
+				u64 root_objectid, u64 ref_generation,
+				u64 owner, u64 owner_offset,
+				struct btrfs_key *ins)
+{
+	int ret;
+	struct btrfs_block_group_cache *block_group;
+
+	maybe_lock_mutex(root);
+	block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid);
+	cache_block_group(root, block_group);
+
+	clear_extent_dirty(&root->fs_info->free_space_cache,
+			   ins->objectid, ins->objectid + ins->offset - 1,
+			   GFP_NOFS);
+	ret = __btrfs_alloc_reserved_extent(trans, root, root_objectid,
+					    ref_generation, owner,
+					    owner_offset, ins);
+	maybe_unlock_mutex(root);
+	return ret;
+}
+
 /*
  * finds a free extent and does all the dirty work required for allocation
  * returns the key for the extent through ins, and a tree buffer for
@@ -2350,9 +2406,8 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 	return buf;
 }
 
-static int noinline drop_leaf_ref_no_cache(struct btrfs_trans_handle *trans,
-					   struct btrfs_root *root,
-					   struct extent_buffer *leaf)
+int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans,
+			struct btrfs_root *root, struct extent_buffer *leaf)
 {
 	u64 leaf_owner;
 	u64 leaf_generation;
@@ -2402,9 +2457,9 @@ static int noinline drop_leaf_ref_no_cache(struct btrfs_trans_handle *trans,
 	return 0;
 }
 
-static int noinline drop_leaf_ref(struct btrfs_trans_handle *trans,
-					 struct btrfs_root *root,
-					 struct btrfs_leaf_ref *ref)
+static int noinline cache_drop_leaf_ref(struct btrfs_trans_handle *trans,
+					struct btrfs_root *root,
+					struct btrfs_leaf_ref *ref)
 {
 	int i;
 	int ret;
@@ -2512,7 +2567,7 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans,
 		    btrfs_header_nritems(cur))
 			break;
 		if (*level == 0) {
-			ret = drop_leaf_ref_no_cache(trans, root, cur);
+			ret = btrfs_drop_leaf_ref(trans, root, cur);
 			BUG_ON(ret);
 			break;
 		}
@@ -2552,7 +2607,7 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans,
 			btrfs_node_key_to_cpu(cur, &key, path->slots[*level]);
 			ref = btrfs_lookup_leaf_ref(root, bytenr);
 			if (ref) {
-				ret = drop_leaf_ref(trans, root, ref);
+				ret = cache_drop_leaf_ref(trans, root, ref);
 				BUG_ON(ret);
 				btrfs_remove_leaf_ref(root, ref);
 				btrfs_free_leaf_ref(root, ref);
@@ -3628,6 +3683,8 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans,
 	extent_root = root->fs_info->extent_root;
 	block_group_cache = &root->fs_info->block_group_cache;
 
+	root->fs_info->last_trans_new_blockgroup = trans->transid;
+
 	cache = kzalloc(sizeof(*cache), GFP_NOFS);
 	BUG_ON(!cache);
 	cache->key.objectid = chunk_offset;
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index e9e86fbaa243..84ecf3ab8511 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -36,6 +36,8 @@
 #include "btrfs_inode.h"
 #include "ioctl.h"
 #include "print-tree.h"
+#include "tree-log.h"
+#include "locking.h"
 #include "compat.h"
 
 
@@ -988,10 +990,27 @@ out_nolock:
 	*ppos = pos;
 
 	if (num_written > 0 && ((file->f_flags & O_SYNC) || IS_SYNC(inode))) {
-		err = sync_page_range(inode, inode->i_mapping,
-				      start_pos, num_written);
+		struct btrfs_trans_handle *trans;
+
+		err = btrfs_fdatawrite_range(inode->i_mapping, start_pos,
+					     start_pos + num_written -1,
+					     WB_SYNC_NONE);
+		if (err < 0)
+			num_written = err;
+
+		err = btrfs_wait_on_page_writeback_range(inode->i_mapping,
+				 start_pos, start_pos + num_written - 1);
 		if (err < 0)
 			num_written = err;
+
+		trans = btrfs_start_transaction(root, 1);
+		ret = btrfs_log_dentry_safe(trans, root, file->f_dentry);
+		if (ret == 0) {
+			btrfs_sync_log(trans, root);
+			btrfs_end_transaction(trans, root);
+		} else {
+			btrfs_commit_transaction(trans, root);
+		}
 	} else if (num_written > 0 && (file->f_flags & O_DIRECT)) {
 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22)
 		do_sync_file_range(file, start_pos,
@@ -1019,8 +1038,7 @@ int btrfs_release_file(struct inode * inode, struct file * filp)
 	return 0;
 }
 
-static int btrfs_sync_file(struct file *file,
-			   struct dentry *dentry, int datasync)
+int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync)
 {
 	struct inode *inode = dentry->d_inode;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
@@ -1043,6 +1061,8 @@ static int btrfs_sync_file(struct file *file,
 	}
 	mutex_unlock(&root->fs_info->trans_mutex);
 
+	filemap_fdatawait(inode->i_mapping);
+
 	/*
 	 * ok we haven't committed the transaction yet, lets do a commit
 	 */
@@ -1054,7 +1074,16 @@ static int btrfs_sync_file(struct file *file,
 		ret = -ENOMEM;
 		goto out;
 	}
-	ret = btrfs_commit_transaction(trans, root);
+
+	ret = btrfs_log_dentry_safe(trans, root, file->f_dentry);
+	if (ret < 0)
+		goto out;
+	if (ret > 0) {
+		ret = btrfs_commit_transaction(trans, root);
+	} else {
+		btrfs_sync_log(trans, root);
+		ret = btrfs_end_transaction(trans, root);
+	}
 out:
 	return ret > 0 ? EIO : ret;
 }
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 43d3f2649ca3..65df9d830230 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -46,6 +46,8 @@
 #include "volumes.h"
 #include "ordered-data.h"
 #include "xattr.h"
+#include "compat.h"
+#include "tree-log.h"
 
 struct btrfs_iget_args {
 	u64 ino;
@@ -586,6 +588,7 @@ nocow:
 			  &ordered_extent->list);
 
 	btrfs_ordered_update_i_size(inode, ordered_extent);
+	btrfs_update_inode(trans, root, inode);
 	btrfs_remove_ordered_extent(inode, ordered_extent);
 
 	/* once for us */
@@ -593,7 +596,6 @@ nocow:
 	/* once for the tree */
 	btrfs_put_ordered_extent(ordered_extent);
 
-	btrfs_update_inode(trans, root, inode);
 	btrfs_end_transaction(trans, root);
 	return 0;
 }
@@ -1007,7 +1009,8 @@ void btrfs_read_locked_inode(struct inode *inode)
 	inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
 
 	inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
-	inode->i_generation = btrfs_inode_generation(leaf, inode_item);
+	BTRFS_I(inode)->generation = btrfs_inode_generation(leaf, inode_item);
+	inode->i_generation = BTRFS_I(inode)->generation;
 	inode->i_rdev = 0;
 	rdev = btrfs_inode_rdev(leaf, inode_item);
 
@@ -1056,7 +1059,8 @@ make_bad:
 	make_bad_inode(inode);
 }
 
-static void fill_inode_item(struct extent_buffer *leaf,
+static void fill_inode_item(struct btrfs_trans_handle *trans,
+			    struct extent_buffer *leaf,
 			    struct btrfs_inode_item *item,
 			    struct inode *inode)
 {
@@ -1082,7 +1086,8 @@ static void fill_inode_item(struct extent_buffer *leaf,
 				inode->i_ctime.tv_nsec);
 
 	btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
-	btrfs_set_inode_generation(leaf, item, inode->i_generation);
+	btrfs_set_inode_generation(leaf, item, BTRFS_I(inode)->generation);
+	btrfs_set_inode_transid(leaf, item, trans->transid);
 	btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
 	btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
 	btrfs_set_inode_block_group(leaf, item,
@@ -1112,7 +1117,7 @@ int noinline btrfs_update_inode(struct btrfs_trans_handle *trans,
 	inode_item = btrfs_item_ptr(leaf, path->slots[0],
 				  struct btrfs_inode_item);
 
-	fill_inode_item(leaf, inode_item, inode);
+	fill_inode_item(trans, leaf, inode_item, inode);
 	btrfs_mark_buffer_dirty(leaf);
 	btrfs_set_inode_last_trans(trans, inode);
 	ret = 0;
@@ -1122,14 +1127,12 @@ failed:
 }
 
 
-static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
-			      struct btrfs_root *root,
-			      struct inode *dir,
-			      struct dentry *dentry)
+int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
+		       struct btrfs_root *root,
+		       struct inode *dir, struct inode *inode,
+		       const char *name, int name_len)
 {
 	struct btrfs_path *path;
-	const char *name = dentry->d_name.name;
-	int name_len = dentry->d_name.len;
 	int ret = 0;
 	struct extent_buffer *leaf;
 	struct btrfs_dir_item *di;
@@ -1160,13 +1163,12 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
 	btrfs_release_path(root, path);
 
 	ret = btrfs_del_inode_ref(trans, root, name, name_len,
-				  dentry->d_inode->i_ino,
-				  dentry->d_parent->d_inode->i_ino, &index);
+				  inode->i_ino,
+				  dir->i_ino, &index);
 	if (ret) {
 		printk("failed to delete reference to %.*s, "
 		       "inode %lu parent %lu\n", name_len, name,
-		       dentry->d_inode->i_ino,
-		       dentry->d_parent->d_inode->i_ino);
+		       inode->i_ino, dir->i_ino);
 		goto err;
 	}
 
@@ -1183,21 +1185,25 @@ static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
 	ret = btrfs_delete_one_dir_name(trans, root, path, di);
 	btrfs_release_path(root, path);
 
-	dentry->d_inode->i_ctime = dir->i_ctime;
+	ret = btrfs_del_inode_ref_in_log(trans, root, name, name_len,
+					 inode, dir->i_ino);
+	BUG_ON(ret);
+
+	ret = btrfs_del_dir_entries_in_log(trans, root, name, name_len,
+					   dir, index);
+	BUG_ON(ret);
 err:
 	btrfs_free_path(path);
-	if (!ret) {
-		btrfs_i_size_write(dir, dir->i_size - name_len * 2);
-		dir->i_mtime = dir->i_ctime = CURRENT_TIME;
-		btrfs_update_inode(trans, root, dir);
-#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
-		dentry->d_inode->i_nlink--;
-#else
-		drop_nlink(dentry->d_inode);
-#endif
-		ret = btrfs_update_inode(trans, root, dentry->d_inode);
-		dir->i_sb->s_dirt = 1;
-	}
+	if (ret)
+		goto out;
+
+	btrfs_i_size_write(dir, dir->i_size - name_len * 2);
+	inode->i_ctime = dir->i_mtime = dir->i_ctime = CURRENT_TIME;
+	btrfs_update_inode(trans, root, dir);
+	btrfs_drop_nlink(inode);
+	ret = btrfs_update_inode(trans, root, inode);
+	dir->i_sb->s_dirt = 1;
+out:
 	return ret;
 }
 
@@ -1218,7 +1224,8 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
 	trans = btrfs_start_transaction(root, 1);
 
 	btrfs_set_trans_block_group(trans, dir);
-	ret = btrfs_unlink_trans(trans, root, dir, dentry);
+	ret = btrfs_unlink_inode(trans, root, dir, dentry->d_inode,
+				 dentry->d_name.name, dentry->d_name.len);
 
 	if (inode->i_nlink == 0)
 		ret = btrfs_orphan_add(trans, inode);
@@ -1256,7 +1263,8 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
 		goto fail_trans;
 
 	/* now the directory is empty */
-	err = btrfs_unlink_trans(trans, root, dir, dentry);
+	err = btrfs_unlink_inode(trans, root, dir, dentry->d_inode,
+				 dentry->d_name.name, dentry->d_name.len);
 	if (!err) {
 		btrfs_i_size_write(inode, 0);
 	}
@@ -1283,10 +1291,10 @@ fail:
  * min_type is the minimum key type to truncate down to.  If set to 0, this
  * will kill all the items on this inode, including the INODE_ITEM_KEY.
  */
-static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
-				   struct btrfs_root *root,
-				   struct inode *inode,
-				   u32 min_type)
+noinline int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
+					struct btrfs_root *root,
+					struct inode *inode,
+					u64 new_size, u32 min_type)
 {
 	int ret;
 	struct btrfs_path *path;
@@ -1307,7 +1315,9 @@ static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
 	int extent_type = -1;
 	u64 mask = root->sectorsize - 1;
 
-	btrfs_drop_extent_cache(inode, inode->i_size & (~mask), (u64)-1);
+	if (root->ref_cows)
+		btrfs_drop_extent_cache(inode,
+					new_size & (~mask), (u64)-1);
 	path = btrfs_alloc_path();
 	path->reada = -1;
 	BUG_ON(!path);
@@ -1324,7 +1334,13 @@ search_again:
 		goto error;
 	}
 	if (ret > 0) {
-		BUG_ON(path->slots[0] == 0);
+		/* there are no items in the tree for us to truncate, we're
+		 * done
+		 */
+		if (path->slots[0] == 0) {
+			ret = 0;
+			goto error;
+		}
 		path->slots[0]--;
 	}
 
@@ -1358,10 +1374,10 @@ search_again:
 		}
 		if (found_type == BTRFS_CSUM_ITEM_KEY) {
 			ret = btrfs_csum_truncate(trans, root, path,
-						  inode->i_size);
+						  new_size);
 			BUG_ON(ret);
 		}
-		if (item_end < inode->i_size) {
+		if (item_end < new_size) {
 			if (found_type == BTRFS_DIR_ITEM_KEY) {
 				found_type = BTRFS_INODE_ITEM_KEY;
 			} else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
@@ -1378,7 +1394,7 @@ search_again:
 			btrfs_set_key_type(&key, found_type);
 			goto next;
 		}
-		if (found_key.offset >= inode->i_size)
+		if (found_key.offset >= new_size)
 			del_item = 1;
 		else
 			del_item = 0;
@@ -1394,7 +1410,7 @@ search_again:
 			if (!del_item) {
 				u64 orig_num_bytes =
 					btrfs_file_extent_num_bytes(leaf, fi);
-				extent_num_bytes = inode->i_size -
+				extent_num_bytes = new_size -
 					found_key.offset + root->sectorsize - 1;
 				extent_num_bytes = extent_num_bytes &
 					~((u64)root->sectorsize - 1);
@@ -1402,7 +1418,7 @@ search_again:
 							 extent_num_bytes);
 				num_dec = (orig_num_bytes -
 					   extent_num_bytes);
-				if (extent_start != 0)
+				if (root->ref_cows && extent_start != 0)
 					dec_i_blocks(inode, num_dec);
 				btrfs_mark_buffer_dirty(leaf);
 			} else {
@@ -1413,22 +1429,29 @@ search_again:
 				num_dec = btrfs_file_extent_num_bytes(leaf, fi);
 				if (extent_start != 0) {
 					found_extent = 1;
-					dec_i_blocks(inode, num_dec);
+					if (root->ref_cows)
+						dec_i_blocks(inode, num_dec);
+				}
+				if (root->ref_cows) {
+					root_gen =
+						btrfs_header_generation(leaf);
 				}
-				root_gen = btrfs_header_generation(leaf);
 				root_owner = btrfs_header_owner(leaf);
 			}
 		} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
 			if (!del_item) {
-				u32 newsize = inode->i_size - found_key.offset;
-				dec_i_blocks(inode, item_end + 1 -
-					    found_key.offset - newsize);
-				newsize =
-				    btrfs_file_extent_calc_inline_size(newsize);
+				u32 size = new_size - found_key.offset;
+
+				if (root->ref_cows) {
+					dec_i_blocks(inode, item_end + 1 -
+						    found_key.offset - size);
+				}
+				size =
+				    btrfs_file_extent_calc_inline_size(size);
 				ret = btrfs_truncate_item(trans, root, path,
-							  newsize, 1);
+							  size, 1);
 				BUG_ON(ret);
-			} else {
+			} else if (root->ref_cows) {
 				dec_i_blocks(inode, item_end + 1 -
 					     found_key.offset);
 			}
@@ -1666,7 +1689,7 @@ void btrfs_delete_inode(struct inode *inode)
 	trans = btrfs_start_transaction(root, 1);
 
 	btrfs_set_trans_block_group(trans, inode);
-	ret = btrfs_truncate_in_trans(trans, root, inode, 0);
+	ret = btrfs_truncate_inode_items(trans, root, inode, inode->i_size, 0);
 	if (ret) {
 		btrfs_orphan_del(NULL, inode);
 		goto no_delete_lock;
@@ -1753,15 +1776,20 @@ static int fixup_tree_root_location(struct btrfs_root *root,
 	return 0;
 }
 
-static int btrfs_init_locked_inode(struct inode *inode, void *p)
+static noinline void init_btrfs_i(struct inode *inode)
 {
-	struct btrfs_iget_args *args = p;
-	inode->i_ino = args->ino;
-	BTRFS_I(inode)->root = args->root;
-	BTRFS_I(inode)->delalloc_bytes = 0;
-	inode->i_mapping->writeback_index = 0;
-	BTRFS_I(inode)->disk_i_size = 0;
-	BTRFS_I(inode)->index_cnt = (u64)-1;
+	struct btrfs_inode *bi = BTRFS_I(inode);
+
+	bi->i_acl = NULL;
+	bi->i_default_acl = NULL;
+
+	bi->generation = 0;
+	bi->last_trans = 0;
+	bi->logged_trans = 0;
+	bi->delalloc_bytes = 0;
+	bi->disk_i_size = 0;
+	bi->flags = 0;
+	bi->index_cnt = (u64)-1;
 	extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
 	extent_io_tree_init(&BTRFS_I(inode)->io_tree,
 			     inode->i_mapping, GFP_NOFS);
@@ -1771,6 +1799,15 @@ static int btrfs_init_locked_inode(struct inode *inode, void *p)
 	btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
 	mutex_init(&BTRFS_I(inode)->csum_mutex);
 	mutex_init(&BTRFS_I(inode)->extent_mutex);
+	mutex_init(&BTRFS_I(inode)->log_mutex);
+}
+
+static int btrfs_init_locked_inode(struct inode *inode, void *p)
+{
+	struct btrfs_iget_args *args = p;
+	inode->i_ino = args->ino;
+	init_btrfs_i(inode);
+	BTRFS_I(inode)->root = args->root;
 	return 0;
 }
 
@@ -2263,21 +2300,10 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 	 * btrfs_get_inode_index_count has an explanation for the magic
 	 * number
 	 */
+	init_btrfs_i(inode);
 	BTRFS_I(inode)->index_cnt = 2;
-
-	extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
-	extent_io_tree_init(&BTRFS_I(inode)->io_tree,
-			     inode->i_mapping, GFP_NOFS);
-	extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
-			     inode->i_mapping, GFP_NOFS);
-	btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
-	INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes);
-	mutex_init(&BTRFS_I(inode)->csum_mutex);
-	mutex_init(&BTRFS_I(inode)->extent_mutex);
-	BTRFS_I(inode)->delalloc_bytes = 0;
-	inode->i_mapping->writeback_index = 0;
-	BTRFS_I(inode)->disk_i_size = 0;
 	BTRFS_I(inode)->root = root;
+	BTRFS_I(inode)->generation = trans->transid;
 
 	if (mode & S_IFDIR)
 		owner = 0;
@@ -2290,7 +2316,6 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 		new_inode_group = group;
 	}
 	BTRFS_I(inode)->block_group = new_inode_group;
-	BTRFS_I(inode)->flags = 0;
 
 	key[0].objectid = objectid;
 	btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY);
@@ -2318,7 +2343,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
 	inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
 				  struct btrfs_inode_item);
-	fill_inode_item(path->nodes[0], inode_item, inode);
+	fill_inode_item(trans, path->nodes[0], inode_item, inode);
 
 	ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
 			     struct btrfs_inode_ref);
@@ -2349,38 +2374,34 @@ static inline u8 btrfs_inode_type(struct inode *inode)
 	return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
 }
 
-static int btrfs_add_link(struct btrfs_trans_handle *trans,
-			    struct dentry *dentry, struct inode *inode,
-			    int add_backref, u64 index)
+int btrfs_add_link(struct btrfs_trans_handle *trans,
+		   struct inode *parent_inode, struct inode *inode,
+		   const char *name, int name_len, int add_backref, u64 index)
 {
 	int ret;
 	struct btrfs_key key;
-	struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
-	struct inode *parent_inode = dentry->d_parent->d_inode;
+	struct btrfs_root *root = BTRFS_I(parent_inode)->root;
 
 	key.objectid = inode->i_ino;
 	btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
 	key.offset = 0;
 
-	ret = btrfs_insert_dir_item(trans, root,
-				    dentry->d_name.name, dentry->d_name.len,
-				    dentry->d_parent->d_inode->i_ino,
+	ret = btrfs_insert_dir_item(trans, root, name, name_len,
+				    parent_inode->i_ino,
 				    &key, btrfs_inode_type(inode),
 				    index);
 	if (ret == 0) {
 		if (add_backref) {
 			ret = btrfs_insert_inode_ref(trans, root,
-					     dentry->d_name.name,
-					     dentry->d_name.len,
-					     inode->i_ino,
-					     parent_inode->i_ino,
-					     index);
+						     name, name_len,
+						     inode->i_ino,
+						     parent_inode->i_ino,
+						     index);
 		}
 		btrfs_i_size_write(parent_inode, parent_inode->i_size +
-				   dentry->d_name.len * 2);
+				   name_len * 2);
 		parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
-		ret = btrfs_update_inode(trans, root,
-					 dentry->d_parent->d_inode);
+		ret = btrfs_update_inode(trans, root, parent_inode);
 	}
 	return ret;
 }
@@ -2389,7 +2410,9 @@ static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
 			    struct dentry *dentry, struct inode *inode,
 			    int backref, u64 index)
 {
-	int err = btrfs_add_link(trans, dentry, inode, backref, index);
+	int err = btrfs_add_link(trans, dentry->d_parent->d_inode,
+				 inode, dentry->d_name.name,
+				 dentry->d_name.len, backref, index);
 	if (!err) {
 		d_instantiate(dentry, inode);
 		return 0;
@@ -2513,19 +2536,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry,
 		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
 		inode->i_fop = &btrfs_file_operations;
 		inode->i_op = &btrfs_file_inode_operations;
-		extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
-		extent_io_tree_init(&BTRFS_I(inode)->io_tree,
-				     inode->i_mapping, GFP_NOFS);
-		extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
-				     inode->i_mapping, GFP_NOFS);
-		INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes);
-		mutex_init(&BTRFS_I(inode)->csum_mutex);
-		mutex_init(&BTRFS_I(inode)->extent_mutex);
-		BTRFS_I(inode)->delalloc_bytes = 0;
-		BTRFS_I(inode)->disk_i_size = 0;
-		inode->i_mapping->writeback_index = 0;
 		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
-		btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
 	}
 	dir->i_sb->s_dirt = 1;
 	btrfs_update_inode_block_group(trans, inode);
@@ -2556,11 +2567,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
 	if (inode->i_nlink == 0)
 		return -ENOENT;
 
-#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
-	inode->i_nlink++;
-#else
-	inc_nlink(inode);
-#endif
+	btrfs_inc_nlink(inode);
 	err = btrfs_check_free_space(root, 1, 0);
 	if (err)
 		goto fail;
@@ -2650,7 +2657,9 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 	if (err)
 		goto out_fail;
 
-	err = btrfs_add_link(trans, dentry, inode, 0, index);
+	err = btrfs_add_link(trans, dentry->d_parent->d_inode,
+				 inode, dentry->d_name.name,
+				 dentry->d_name.len, 0, index);
 	if (err)
 		goto out_fail;
 
@@ -3221,7 +3230,7 @@ static void btrfs_truncate(struct inode *inode)
 	if (ret)
 		goto out;
 	/* FIXME, add redo link to tree so we don't leak on crash */
-	ret = btrfs_truncate_in_trans(trans, root, inode,
+	ret = btrfs_truncate_inode_items(trans, root, inode, inode->i_size,
 				      BTRFS_EXTENT_DATA_KEY);
 	btrfs_update_inode(trans, root, inode);
 
@@ -3304,6 +3313,7 @@ struct inode *btrfs_alloc_inode(struct super_block *sb)
 	if (!ei)
 		return NULL;
 	ei->last_trans = 0;
+	ei->logged_trans = 0;
 	btrfs_ordered_inode_tree_init(&ei->ordered_tree);
 	ei->i_acl = BTRFS_ACL_NOT_CACHED;
 	ei->i_default_acl = BTRFS_ACL_NOT_CACHED;
@@ -3463,31 +3473,39 @@ static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
 
 	btrfs_set_trans_block_group(trans, new_dir);
 
-	old_dentry->d_inode->i_nlink++;
+	btrfs_inc_nlink(old_dentry->d_inode);
 	old_dir->i_ctime = old_dir->i_mtime = ctime;
 	new_dir->i_ctime = new_dir->i_mtime = ctime;
 	old_inode->i_ctime = ctime;
 
-	ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
+	ret = btrfs_unlink_inode(trans, root, old_dir, old_dentry->d_inode,
+				 old_dentry->d_name.name,
+				 old_dentry->d_name.len);
 	if (ret)
 		goto out_fail;
 
 	if (new_inode) {
 		new_inode->i_ctime = CURRENT_TIME;
-		ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
+		ret = btrfs_unlink_inode(trans, root, new_dir,
+					 new_dentry->d_inode,
+					 new_dentry->d_name.name,
+					 new_dentry->d_name.len);
 		if (ret)
 			goto out_fail;
 		if (new_inode->i_nlink == 0) {
-			ret = btrfs_orphan_add(trans, new_inode);
+			ret = btrfs_orphan_add(trans, new_dentry->d_inode);
 			if (ret)
 				goto out_fail;
 		}
+
 	}
 	ret = btrfs_set_inode_index(new_dir, old_inode, &index);
 	if (ret)
 		goto out_fail;
 
-	ret = btrfs_add_link(trans, new_dentry, old_inode, 1, index);
+	ret = btrfs_add_link(trans, new_dentry->d_parent->d_inode,
+			     old_inode, new_dentry->d_name.name,
+			     new_dentry->d_name.len, 1, index);
 	if (ret)
 		goto out_fail;
 
@@ -3577,19 +3595,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
 		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
 		inode->i_fop = &btrfs_file_operations;
 		inode->i_op = &btrfs_file_inode_operations;
-		extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
-		extent_io_tree_init(&BTRFS_I(inode)->io_tree,
-				     inode->i_mapping, GFP_NOFS);
-		extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
-				     inode->i_mapping, GFP_NOFS);
-		INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes);
-		mutex_init(&BTRFS_I(inode)->csum_mutex);
-		mutex_init(&BTRFS_I(inode)->extent_mutex);
-		BTRFS_I(inode)->delalloc_bytes = 0;
-		BTRFS_I(inode)->disk_i_size = 0;
-		inode->i_mapping->writeback_index = 0;
 		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
-		btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
 	}
 	dir->i_sb->s_dirt = 1;
 	btrfs_update_inode_block_group(trans, inode);
@@ -3691,6 +3697,7 @@ static struct file_operations btrfs_dir_file_operations = {
 	.compat_ioctl	= btrfs_ioctl,
 #endif
 	.release        = btrfs_release_file,
+	.fsync		= btrfs_sync_file,
 };
 
 static struct extent_io_ops btrfs_extent_io_ops = {
diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c
index 36726696e58b..e3984f902e71 100644
--- a/fs/btrfs/root-tree.c
+++ b/fs/btrfs/root-tree.c
@@ -202,8 +202,9 @@ again:
 		memcpy(&found_key, &key, sizeof(key));
 		key.offset++;
 		btrfs_release_path(root, path);
-		dead_root = btrfs_read_fs_root_no_radix(root->fs_info,
-							&found_key);
+		dead_root =
+			btrfs_read_fs_root_no_radix(root->fs_info->tree_root,
+						    &found_key);
 		if (IS_ERR(dead_root)) {
 			ret = PTR_ERR(dead_root);
 			goto err;
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index eff3ad72991b..49c4f5b40ed6 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -25,6 +25,7 @@
 #include "transaction.h"
 #include "locking.h"
 #include "ref-cache.h"
+#include "tree-log.h"
 
 static int total_trans = 0;
 extern struct kmem_cache *btrfs_trans_handle_cachep;
@@ -57,6 +58,7 @@ static noinline int join_transaction(struct btrfs_root *root)
 		root->fs_info->generation++;
 		root->fs_info->last_alloc = 0;
 		root->fs_info->last_data_alloc = 0;
+		root->fs_info->last_log_alloc = 0;
 		cur_trans->num_writers = 1;
 		cur_trans->num_joined = 0;
 		cur_trans->transid = root->fs_info->generation;
@@ -83,7 +85,7 @@ static noinline int join_transaction(struct btrfs_root *root)
 	return 0;
 }
 
-static noinline int record_root_in_trans(struct btrfs_root *root)
+noinline int btrfs_record_root_in_trans(struct btrfs_root *root)
 {
 	struct btrfs_dirty_root *dirty;
 	u64 running_trans_id = root->fs_info->running_transaction->transid;
@@ -151,7 +153,7 @@ static void wait_current_trans(struct btrfs_root *root)
 	}
 }
 
-struct btrfs_trans_handle *start_transaction(struct btrfs_root *root,
+static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root,
 					     int num_blocks, int wait)
 {
 	struct btrfs_trans_handle *h =
@@ -164,7 +166,7 @@ struct btrfs_trans_handle *start_transaction(struct btrfs_root *root,
 	ret = join_transaction(root);
 	BUG_ON(ret);
 
-	record_root_in_trans(root);
+	btrfs_record_root_in_trans(root);
 	h->transid = root->fs_info->running_transaction->transid;
 	h->transaction = root->fs_info->running_transaction;
 	h->blocks_reserved = num_blocks;
@@ -456,6 +458,8 @@ static noinline int add_dirty_roots(struct btrfs_trans_handle *trans,
 			BUG_ON(!root->ref_tree);
 			dirty = root->dirty_root;
 
+			btrfs_free_log(trans, root);
+
 			if (root->commit_root == root->node) {
 				WARN_ON(root->node->start !=
 					btrfs_root_bytenr(&root->root_item));
@@ -600,7 +604,7 @@ static noinline int drop_dirty_roots(struct btrfs_root *tree_root,
 		num_bytes -= btrfs_root_used(&dirty->root->root_item);
 		bytes_used = btrfs_root_used(&root->root_item);
 		if (num_bytes) {
-			record_root_in_trans(root);
+			btrfs_record_root_in_trans(root);
 			btrfs_set_root_used(&root->root_item,
 					    bytes_used - num_bytes);
 		}
@@ -745,7 +749,6 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	int ret;
 
 	INIT_LIST_HEAD(&dirty_fs_roots);
-
 	mutex_lock(&root->fs_info->trans_mutex);
 	if (trans->transaction->in_commit) {
 		cur_trans = trans->transaction;
@@ -821,10 +824,30 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 
 	WARN_ON(cur_trans != trans->transaction);
 
+	/* btrfs_commit_tree_roots is responsible for getting the
+	 * various roots consistent with each other.  Every pointer
+	 * in the tree of tree roots has to point to the most up to date
+	 * root for every subvolume and other tree.  So, we have to keep
+	 * the tree logging code from jumping in and changing any
+	 * of the trees.
+	 *
+	 * At this point in the commit, there can't be any tree-log
+	 * writers, but a little lower down we drop the trans mutex
+	 * and let new people in.  By holding the tree_log_mutex
+	 * from now until after the super is written, we avoid races
+	 * with the tree-log code.
+	 */
+	mutex_lock(&root->fs_info->tree_log_mutex);
+
 	ret = add_dirty_roots(trans, &root->fs_info->fs_roots_radix,
 			      &dirty_fs_roots);
 	BUG_ON(ret);
 
+	/* add_dirty_roots gets rid of all the tree log roots, it is now
+	 * safe to free the root of tree log roots
+	 */
+	btrfs_free_log_root_tree(trans, root->fs_info);
+
 	ret = btrfs_commit_tree_roots(trans, root);
 	BUG_ON(ret);
 
@@ -843,6 +866,12 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 				   chunk_root->node->start);
 	btrfs_set_super_chunk_root_level(&root->fs_info->super_copy,
 					 btrfs_header_level(chunk_root->node));
+
+	if (!root->fs_info->log_root_recovering) {
+		btrfs_set_super_log_root(&root->fs_info->super_copy, 0);
+		btrfs_set_super_log_root_level(&root->fs_info->super_copy, 0);
+	}
+
 	memcpy(&root->fs_info->super_for_commit, &root->fs_info->super_copy,
 	       sizeof(root->fs_info->super_copy));
 
@@ -857,6 +886,12 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	BUG_ON(ret);
 	write_ctree_super(trans, root);
 
+	/*
+	 * the super is written, we can safely allow the tree-loggers
+	 * to go about their business
+	 */
+	mutex_unlock(&root->fs_info->tree_log_mutex);
+
 	btrfs_finish_extent_commit(trans, root, pinned_copy);
 	mutex_lock(&root->fs_info->trans_mutex);
 
diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h
index 598baa312417..cc63650d60d6 100644
--- a/fs/btrfs/transaction.h
+++ b/fs/btrfs/transaction.h
@@ -98,4 +98,5 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans,
 				   struct btrfs_root *root);
 void btrfs_throttle(struct btrfs_root *root);
+int btrfs_record_root_in_trans(struct btrfs_root *root);
 #endif
diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c
index cc2650b06952..b3bb5bbad76e 100644
--- a/fs/btrfs/tree-defrag.c
+++ b/fs/btrfs/tree-defrag.c
@@ -81,12 +81,12 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 		memcpy(&key, &root->defrag_progress, sizeof(key));
 	}
 
-	path->lowest_level = 1;
 	path->keep_locks = 1;
 	if (cache_only)
 		min_trans = root->defrag_trans_start;
 
-	ret = btrfs_search_forward(root, &key, path, cache_only, min_trans);
+	ret = btrfs_search_forward(root, &key, NULL, path,
+				   cache_only, min_trans);
 	if (ret < 0)
 		goto out;
 	if (ret > 0) {
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
new file mode 100644
index 000000000000..d1ce8314b948
--- /dev/null
+++ b/fs/btrfs/tree-log.c
@@ -0,0 +1,2804 @@
+/*
+ * Copyright (C) 2008 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/sched.h>
+#include "ctree.h"
+#include "transaction.h"
+#include "disk-io.h"
+#include "locking.h"
+#include "print-tree.h"
+#include "compat.h"
+
+/* magic values for the inode_only field in btrfs_log_inode:
+ *
+ * LOG_INODE_ALL means to log everything
+ * LOG_INODE_EXISTS means to log just enough to recreate the inode
+ * during log replay
+ */
+#define LOG_INODE_ALL 0
+#define LOG_INODE_EXISTS 1
+
+/*
+ * stages for the tree walking.  The first
+ * stage (0) is to only pin down the blocks we find
+ * the second stage (1) is to make sure that all the inodes
+ * we find in the log are created in the subvolume.
+ *
+ * The last stage is to deal with directories and links and extents
+ * and all the other fun semantics
+ */
+#define LOG_WALK_PIN_ONLY 0
+#define LOG_WALK_REPLAY_INODES 1
+#define LOG_WALK_REPLAY_ALL 2
+
+static int __btrfs_log_inode(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root, struct inode *inode,
+			     int inode_only);
+
+/*
+ * tree logging is a special write ahead log used to make sure that
+ * fsyncs and O_SYNCs can happen without doing full tree commits.
+ *
+ * Full tree commits are expensive because they require commonly
+ * modified blocks to be recowed, creating many dirty pages in the
+ * extent tree an 4x-6x higher write load than ext3.
+ *
+ * Instead of doing a tree commit on every fsync, we use the
+ * key ranges and transaction ids to find items for a given file or directory
+ * that have changed in this transaction.  Those items are copied into
+ * a special tree (one per subvolume root), that tree is written to disk
+ * and then the fsync is considered complete.
+ *
+ * After a crash, items are copied out of the log-tree back into the
+ * subvolume tree.  Any file data extents found are recorded in the extent
+ * allocation tree, and the log-tree freed.
+ *
+ * The log tree is read three times, once to pin down all the extents it is
+ * using in ram and once, once to create all the inodes logged in the tree
+ * and once to do all the other items.
+ */
+
+/*
+ * btrfs_add_log_tree adds a new per-subvolume log tree into the
+ * tree of log tree roots.  This must be called with a tree log transaction
+ * running (see start_log_trans).
+ */
+int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
+		      struct btrfs_root *root)
+{
+	struct btrfs_key key;
+	struct btrfs_root_item root_item;
+	struct btrfs_inode_item *inode_item;
+	struct extent_buffer *leaf;
+	struct btrfs_root *new_root = root;
+	int ret;
+	u64 objectid = root->root_key.objectid;
+
+	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
+				      BTRFS_TREE_LOG_OBJECTID,
+				      0, 0, 0, 0, 0);
+	if (IS_ERR(leaf)) {
+		ret = PTR_ERR(leaf);
+		return ret;
+	}
+
+	btrfs_set_header_nritems(leaf, 0);
+	btrfs_set_header_level(leaf, 0);
+	btrfs_set_header_bytenr(leaf, leaf->start);
+	btrfs_set_header_generation(leaf, trans->transid);
+	btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
+
+	write_extent_buffer(leaf, root->fs_info->fsid,
+			    (unsigned long)btrfs_header_fsid(leaf),
+			    BTRFS_FSID_SIZE);
+	btrfs_mark_buffer_dirty(leaf);
+
+	inode_item = &root_item.inode;
+	memset(inode_item, 0, sizeof(*inode_item));
+	inode_item->generation = cpu_to_le64(1);
+	inode_item->size = cpu_to_le64(3);
+	inode_item->nlink = cpu_to_le32(1);
+	inode_item->nblocks = cpu_to_le64(1);
+	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
+
+	btrfs_set_root_bytenr(&root_item, leaf->start);
+	btrfs_set_root_level(&root_item, 0);
+	btrfs_set_root_refs(&root_item, 0);
+	btrfs_set_root_used(&root_item, 0);
+
+	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
+	root_item.drop_level = 0;
+
+	btrfs_tree_unlock(leaf);
+	free_extent_buffer(leaf);
+	leaf = NULL;
+
+	btrfs_set_root_dirid(&root_item, 0);
+
+	key.objectid = BTRFS_TREE_LOG_OBJECTID;
+	key.offset = objectid;
+	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
+	ret = btrfs_insert_root(trans, root->fs_info->log_root_tree, &key,
+				&root_item);
+	if (ret)
+		goto fail;
+
+	new_root = btrfs_read_fs_root_no_radix(root->fs_info->log_root_tree,
+					       &key);
+	BUG_ON(!new_root);
+
+	WARN_ON(root->log_root);
+	root->log_root = new_root;
+
+	/*
+	 * log trees do not get reference counted because they go away
+	 * before a real commit is actually done.  They do store pointers
+	 * to file data extents, and those reference counts still get
+	 * updated (along with back refs to the log tree).
+	 */
+	new_root->ref_cows = 0;
+	new_root->last_trans = trans->transid;
+fail:
+	return ret;
+}
+
+/*
+ * start a sub transaction and setup the log tree
+ * this increments the log tree writer count to make the people
+ * syncing the tree wait for us to finish
+ */
+static int start_log_trans(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *root)
+{
+	int ret;
+	mutex_lock(&root->fs_info->tree_log_mutex);
+	if (!root->fs_info->log_root_tree) {
+		ret = btrfs_init_log_root_tree(trans, root->fs_info);
+		BUG_ON(ret);
+	}
+	if (!root->log_root) {
+		ret = btrfs_add_log_tree(trans, root);
+		BUG_ON(ret);
+	}
+	atomic_inc(&root->fs_info->tree_log_writers);
+	root->fs_info->tree_log_batch++;
+	mutex_unlock(&root->fs_info->tree_log_mutex);
+	return 0;
+}
+
+/*
+ * returns 0 if there was a log transaction running and we were able
+ * to join, or returns -ENOENT if there were not transactions
+ * in progress
+ */
+static int join_running_log_trans(struct btrfs_root *root)
+{
+	int ret = -ENOENT;
+
+	smp_mb();
+	if (!root->log_root)
+		return -ENOENT;
+
+	mutex_lock(&root->fs_info->tree_log_mutex);
+	if (root->log_root) {
+		ret = 0;
+		atomic_inc(&root->fs_info->tree_log_writers);
+		root->fs_info->tree_log_batch++;
+	}
+	mutex_unlock(&root->fs_info->tree_log_mutex);
+	return ret;
+}
+
+/*
+ * indicate we're done making changes to the log tree
+ * and wake up anyone waiting to do a sync
+ */
+static int end_log_trans(struct btrfs_root *root)
+{
+	atomic_dec(&root->fs_info->tree_log_writers);
+	smp_mb();
+	if (waitqueue_active(&root->fs_info->tree_log_wait))
+		wake_up(&root->fs_info->tree_log_wait);
+	return 0;
+}
+
+
+/*
+ * the walk control struct is used to pass state down the chain when
+ * processing the log tree.  The stage field tells us which part
+ * of the log tree processing we are currently doing.  The others
+ * are state fields used for that specific part
+ */
+struct walk_control {
+	/* should we free the extent on disk when done?  This is used
+	 * at transaction commit time while freeing a log tree
+	 */
+	int free;
+
+	/* should we write out the extent buffer?  This is used
+	 * while flushing the log tree to disk during a sync
+	 */
+	int write;
+
+	/* should we wait for the extent buffer io to finish?  Also used
+	 * while flushing the log tree to disk for a sync
+	 */
+	int wait;
+
+	/* pin only walk, we record which extents on disk belong to the
+	 * log trees
+	 */
+	int pin;
+
+	/* what stage of the replay code we're currently in */
+	int stage;
+
+	/* the root we are currently replaying */
+	struct btrfs_root *replay_dest;
+
+	/* the trans handle for the current replay */
+	struct btrfs_trans_handle *trans;
+
+	/* the function that gets used to process blocks we find in the
+	 * tree.  Note the extent_buffer might not be up to date when it is
+	 * passed in, and it must be checked or read if you need the data
+	 * inside it
+	 */
+	int (*process_func)(struct btrfs_root *log, struct extent_buffer *eb,
+			    struct walk_control *wc, u64 gen);
+};
+
+/*
+ * process_func used to pin down extents, write them or wait on them
+ */
+static int process_one_buffer(struct btrfs_root *log,
+			      struct extent_buffer *eb,
+			      struct walk_control *wc, u64 gen)
+{
+	if (wc->pin) {
+		mutex_lock(&log->fs_info->alloc_mutex);
+		btrfs_update_pinned_extents(log->fs_info->extent_root,
+					    eb->start, eb->len, 1);
+		mutex_unlock(&log->fs_info->alloc_mutex);
+	}
+
+	if (btrfs_buffer_uptodate(eb, gen)) {
+		if (wc->write)
+			btrfs_write_tree_block(eb);
+		if (wc->wait)
+			btrfs_wait_tree_block_writeback(eb);
+	}
+	return 0;
+}
+
+/*
+ * Item overwrite used by replay and tree logging.  eb, slot and key all refer
+ * to the src data we are copying out.
+ *
+ * root is the tree we are copying into, and path is a scratch
+ * path for use in this function (it should be released on entry and
+ * will be released on exit).
+ *
+ * If the key is already in the destination tree the existing item is
+ * overwritten.  If the existing item isn't big enough, it is extended.
+ * If it is too large, it is truncated.
+ *
+ * If the key isn't in the destination yet, a new item is inserted.
+ */
+static noinline int overwrite_item(struct btrfs_trans_handle *trans,
+				   struct btrfs_root *root,
+				   struct btrfs_path *path,
+				   struct extent_buffer *eb, int slot,
+				   struct btrfs_key *key)
+{
+	int ret;
+	u32 item_size;
+	u64 saved_i_size = 0;
+	int save_old_i_size = 0;
+	unsigned long src_ptr;
+	unsigned long dst_ptr;
+	int overwrite_root = 0;
+
+	if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID)
+		overwrite_root = 1;
+
+	item_size = btrfs_item_size_nr(eb, slot);
+	src_ptr = btrfs_item_ptr_offset(eb, slot);
+
+	/* look for the key in the destination tree */
+	ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
+	if (ret == 0) {
+		char *src_copy;
+		char *dst_copy;
+		u32 dst_size = btrfs_item_size_nr(path->nodes[0],
+						  path->slots[0]);
+		if (dst_size != item_size)
+			goto insert;
+
+		if (item_size == 0) {
+			btrfs_release_path(root, path);
+			return 0;
+		}
+		dst_copy = kmalloc(item_size, GFP_NOFS);
+		src_copy = kmalloc(item_size, GFP_NOFS);
+
+		read_extent_buffer(eb, src_copy, src_ptr, item_size);
+
+		dst_ptr = btrfs_item_ptr_offset(path->nodes[0], path->slots[0]);
+		read_extent_buffer(path->nodes[0], dst_copy, dst_ptr,
+				   item_size);
+		ret = memcmp(dst_copy, src_copy, item_size);
+
+		kfree(dst_copy);
+		kfree(src_copy);
+		/*
+		 * they have the same contents, just return, this saves
+		 * us from cowing blocks in the destination tree and doing
+		 * extra writes that may not have been done by a previous
+		 * sync
+		 */
+		if (ret == 0) {
+			btrfs_release_path(root, path);
+			return 0;
+		}
+
+	}
+insert:
+	btrfs_release_path(root, path);
+	/* try to insert the key into the destination tree */
+	ret = btrfs_insert_empty_item(trans, root, path,
+				      key, item_size);
+
+	/* make sure any existing item is the correct size */
+	if (ret == -EEXIST) {
+		u32 found_size;
+		found_size = btrfs_item_size_nr(path->nodes[0],
+						path->slots[0]);
+		if (found_size > item_size) {
+			btrfs_truncate_item(trans, root, path, item_size, 1);
+		} else if (found_size < item_size) {
+			ret = btrfs_del_item(trans, root,
+					     path);
+			BUG_ON(ret);
+
+			btrfs_release_path(root, path);
+			ret = btrfs_insert_empty_item(trans,
+				  root, path, key, item_size);
+			BUG_ON(ret);
+		}
+	} else if (ret) {
+		BUG();
+	}
+	dst_ptr = btrfs_item_ptr_offset(path->nodes[0],
+					path->slots[0]);
+
+	/* don't overwrite an existing inode if the generation number
+	 * was logged as zero.  This is done when the tree logging code
+	 * is just logging an inode to make sure it exists after recovery.
+	 *
+	 * Also, don't overwrite i_size on directories during replay.
+	 * log replay inserts and removes directory items based on the
+	 * state of the tree found in the subvolume, and i_size is modified
+	 * as it goes
+	 */
+	if (key->type == BTRFS_INODE_ITEM_KEY && ret == -EEXIST) {
+		struct btrfs_inode_item *src_item;
+		struct btrfs_inode_item *dst_item;
+
+		src_item = (struct btrfs_inode_item *)src_ptr;
+		dst_item = (struct btrfs_inode_item *)dst_ptr;
+
+		if (btrfs_inode_generation(eb, src_item) == 0)
+			goto no_copy;
+
+		if (overwrite_root &&
+		    S_ISDIR(btrfs_inode_mode(eb, src_item)) &&
+		    S_ISDIR(btrfs_inode_mode(path->nodes[0], dst_item))) {
+			save_old_i_size = 1;
+			saved_i_size = btrfs_inode_size(path->nodes[0],
+							dst_item);
+		}
+	}
+
+	copy_extent_buffer(path->nodes[0], eb, dst_ptr,
+			   src_ptr, item_size);
+
+	if (save_old_i_size) {
+		struct btrfs_inode_item *dst_item;
+		dst_item = (struct btrfs_inode_item *)dst_ptr;
+		btrfs_set_inode_size(path->nodes[0], dst_item, saved_i_size);
+	}
+
+	/* make sure the generation is filled in */
+	if (key->type == BTRFS_INODE_ITEM_KEY) {
+		struct btrfs_inode_item *dst_item;
+		dst_item = (struct btrfs_inode_item *)dst_ptr;
+		if (btrfs_inode_generation(path->nodes[0], dst_item) == 0) {
+			btrfs_set_inode_generation(path->nodes[0], dst_item,
+						   trans->transid);
+		}
+	}
+no_copy:
+	btrfs_mark_buffer_dirty(path->nodes[0]);
+	btrfs_release_path(root, path);
+	return 0;
+}
+
+/*
+ * simple helper to read an inode off the disk from a given root
+ * This can only be called for subvolume roots and not for the log
+ */
+static noinline struct inode *read_one_inode(struct btrfs_root *root,
+					     u64 objectid)
+{
+	struct inode *inode;
+	inode = btrfs_iget_locked(root->fs_info->sb, objectid, root);
+	if (inode->i_state & I_NEW) {
+		BTRFS_I(inode)->root = root;
+		BTRFS_I(inode)->location.objectid = objectid;
+		BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
+		BTRFS_I(inode)->location.offset = 0;
+		btrfs_read_locked_inode(inode);
+		unlock_new_inode(inode);
+
+	}
+	if (is_bad_inode(inode)) {
+		iput(inode);
+		inode = NULL;
+	}
+	return inode;
+}
+
+/* replays a single extent in 'eb' at 'slot' with 'key' into the
+ * subvolume 'root'.  path is released on entry and should be released
+ * on exit.
+ *
+ * extents in the log tree have not been allocated out of the extent
+ * tree yet.  So, this completes the allocation, taking a reference
+ * as required if the extent already exists or creating a new extent
+ * if it isn't in the extent allocation tree yet.
+ *
+ * The extent is inserted into the file, dropping any existing extents
+ * from the file that overlap the new one.
+ */
+static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
+				      struct btrfs_root *root,
+				      struct btrfs_path *path,
+				      struct extent_buffer *eb, int slot,
+				      struct btrfs_key *key)
+{
+	int found_type;
+	u64 mask = root->sectorsize - 1;
+	u64 extent_end;
+	u64 alloc_hint;
+	u64 start = key->offset;
+	struct btrfs_file_extent_item *item;
+	struct inode *inode = NULL;
+	unsigned long size;
+	int ret = 0;
+
+	item = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
+	found_type = btrfs_file_extent_type(eb, item);
+
+	if (found_type == BTRFS_FILE_EXTENT_REG)
+		extent_end = start + btrfs_file_extent_num_bytes(eb, item);
+	else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
+		size = btrfs_file_extent_inline_len(eb,
+						    btrfs_item_nr(eb, slot));
+		extent_end = (start + size + mask) & ~mask;
+	} else {
+		ret = 0;
+		goto out;
+	}
+
+	inode = read_one_inode(root, key->objectid);
+	if (!inode) {
+		ret = -EIO;
+		goto out;
+	}
+
+	/*
+	 * first check to see if we already have this extent in the
+	 * file.  This must be done before the btrfs_drop_extents run
+	 * so we don't try to drop this extent.
+	 */
+	ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino,
+				       start, 0);
+
+	if (ret == 0 && found_type == BTRFS_FILE_EXTENT_REG) {
+		struct btrfs_file_extent_item cmp1;
+		struct btrfs_file_extent_item cmp2;
+		struct btrfs_file_extent_item *existing;
+		struct extent_buffer *leaf;
+
+		leaf = path->nodes[0];
+		existing = btrfs_item_ptr(leaf, path->slots[0],
+					  struct btrfs_file_extent_item);
+
+		read_extent_buffer(eb, &cmp1, (unsigned long)item,
+				   sizeof(cmp1));
+		read_extent_buffer(leaf, &cmp2, (unsigned long)existing,
+				   sizeof(cmp2));
+
+		/*
+		 * we already have a pointer to this exact extent,
+		 * we don't have to do anything
+		 */
+		if (memcmp(&cmp1, &cmp2, sizeof(cmp1)) == 0) {
+			btrfs_release_path(root, path);
+			goto out;
+		}
+	}
+	btrfs_release_path(root, path);
+
+	/* drop any overlapping extents */
+	ret = btrfs_drop_extents(trans, root, inode,
+			 start, extent_end, start, &alloc_hint);
+	BUG_ON(ret);
+
+	BUG_ON(ret);
+	if (found_type == BTRFS_FILE_EXTENT_REG) {
+		struct btrfs_key ins;
+
+		ins.objectid = btrfs_file_extent_disk_bytenr(eb, item);
+		ins.offset = btrfs_file_extent_disk_num_bytes(eb, item);
+		ins.type = BTRFS_EXTENT_ITEM_KEY;
+
+		/* insert the extent pointer in the file */
+		ret = overwrite_item(trans, root, path, eb, slot, key);
+		BUG_ON(ret);
+
+		/*
+		 * is this extent already allocated in the extent
+		 * allocation tree?  If so, just add a reference
+		 */
+		ret = btrfs_lookup_extent(root, path, ins.objectid, ins.offset);
+		btrfs_release_path(root, path);
+		if (ret == 0) {
+			ret = btrfs_inc_extent_ref(trans, root,
+				   ins.objectid, ins.offset,
+				   root->root_key.objectid,
+				   trans->transid, key->objectid, start);
+		} else {
+			/*
+			 * insert the extent pointer in the extent
+			 * allocation tree
+			 */
+			ret = btrfs_alloc_logged_extent(trans, root,
+						root->root_key.objectid,
+						trans->transid, key->objectid,
+						start, &ins);
+			BUG_ON(ret);
+		}
+	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
+		/* inline extents are easy, we just overwrite them */
+		ret = overwrite_item(trans, root, path, eb, slot, key);
+		BUG_ON(ret);
+	}
+	/* btrfs_drop_extents changes i_blocks, update it here */
+	inode->i_blocks += (extent_end - start) >> 9;
+	btrfs_update_inode(trans, root, inode);
+out:
+	if (inode)
+		iput(inode);
+	return ret;
+}
+
+/*
+ * when cleaning up conflicts between the directory names in the
+ * subvolume, directory names in the log and directory names in the
+ * inode back references, we may have to unlink inodes from directories.
+ *
+ * This is a helper function to do the unlink of a specific directory
+ * item
+ */
+static noinline int drop_one_dir_item(struct btrfs_trans_handle *trans,
+				      struct btrfs_root *root,
+				      struct btrfs_path *path,
+				      struct inode *dir,
+				      struct btrfs_dir_item *di)
+{
+	struct inode *inode;
+	char *name;
+	int name_len;
+	struct extent_buffer *leaf;
+	struct btrfs_key location;
+	int ret;
+
+	leaf = path->nodes[0];
+
+	btrfs_dir_item_key_to_cpu(leaf, di, &location);
+	name_len = btrfs_dir_name_len(leaf, di);
+	name = kmalloc(name_len, GFP_NOFS);
+	read_extent_buffer(leaf, name, (unsigned long)(di + 1), name_len);
+	btrfs_release_path(root, path);
+
+	inode = read_one_inode(root, location.objectid);
+	BUG_ON(!inode);
+
+	btrfs_inc_nlink(inode);
+	ret = btrfs_unlink_inode(trans, root, dir, inode, name, name_len);
+	kfree(name);
+
+	iput(inode);
+	return ret;
+}
+
+/*
+ * helper function to see if a given name and sequence number found
+ * in an inode back reference are already in a directory and correctly
+ * point to this inode
+ */
+static noinline int inode_in_dir(struct btrfs_root *root,
+				 struct btrfs_path *path,
+				 u64 dirid, u64 objectid, u64 index,
+				 const char *name, int name_len)
+{
+	struct btrfs_dir_item *di;
+	struct btrfs_key location;
+	int match = 0;
+
+	di = btrfs_lookup_dir_index_item(NULL, root, path, dirid,
+					 index, name, name_len, 0);
+	if (di && !IS_ERR(di)) {
+		btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
+		if (location.objectid != objectid)
+			goto out;
+	} else
+		goto out;
+	btrfs_release_path(root, path);
+
+	di = btrfs_lookup_dir_item(NULL, root, path, dirid, name, name_len, 0);
+	if (di && !IS_ERR(di)) {
+		btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
+		if (location.objectid != objectid)
+			goto out;
+	} else
+		goto out;
+	match = 1;
+out:
+	btrfs_release_path(root, path);
+	return match;
+}
+
+/*
+ * helper function to check a log tree for a named back reference in
+ * an inode.  This is used to decide if a back reference that is
+ * found in the subvolume conflicts with what we find in the log.
+ *
+ * inode backreferences may have multiple refs in a single item,
+ * during replay we process one reference at a time, and we don't
+ * want to delete valid links to a file from the subvolume if that
+ * link is also in the log.
+ */
+static noinline int backref_in_log(struct btrfs_root *log,
+				   struct btrfs_key *key,
+				   char *name, int namelen)
+{
+	struct btrfs_path *path;
+	struct btrfs_inode_ref *ref;
+	unsigned long ptr;
+	unsigned long ptr_end;
+	unsigned long name_ptr;
+	int found_name_len;
+	int item_size;
+	int ret;
+	int match = 0;
+
+	path = btrfs_alloc_path();
+	ret = btrfs_search_slot(NULL, log, key, path, 0, 0);
+	if (ret != 0)
+		goto out;
+
+	item_size = btrfs_item_size_nr(path->nodes[0], path->slots[0]);
+	ptr = btrfs_item_ptr_offset(path->nodes[0], path->slots[0]);
+	ptr_end = ptr + item_size;
+	while (ptr < ptr_end) {
+		ref = (struct btrfs_inode_ref *)ptr;
+		found_name_len = btrfs_inode_ref_name_len(path->nodes[0], ref);
+		if (found_name_len == namelen) {
+			name_ptr = (unsigned long)(ref + 1);
+			ret = memcmp_extent_buffer(path->nodes[0], name,
+						   name_ptr, namelen);
+			if (ret == 0) {
+				match = 1;
+				goto out;
+			}
+		}
+		ptr = (unsigned long)(ref + 1) + found_name_len;
+	}
+out:
+	btrfs_free_path(path);
+	return match;
+}
+
+
+/*
+ * replay one inode back reference item found in the log tree.
+ * eb, slot and key refer to the buffer and key found in the log tree.
+ * root is the destination we are replaying into, and path is for temp
+ * use by this function.  (it should be released on return).
+ */
+static noinline int add_inode_ref(struct btrfs_trans_handle *trans,
+				  struct btrfs_root *root,
+				  struct btrfs_root *log,
+				  struct btrfs_path *path,
+				  struct extent_buffer *eb, int slot,
+				  struct btrfs_key *key)
+{
+	struct inode *dir;
+	int ret;
+	struct btrfs_key location;
+	struct btrfs_inode_ref *ref;
+	struct btrfs_dir_item *di;
+	struct inode *inode;
+	char *name;
+	int namelen;
+	unsigned long ref_ptr;
+	unsigned long ref_end;
+
+	location.objectid = key->objectid;
+	location.type = BTRFS_INODE_ITEM_KEY;
+	location.offset = 0;
+
+	/*
+	 * it is possible that we didn't log all the parent directories
+	 * for a given inode.  If we don't find the dir, just don't
+	 * copy the back ref in.  The link count fixup code will take
+	 * care of the rest
+	 */
+	dir = read_one_inode(root, key->offset);
+	if (!dir)
+		return -ENOENT;
+
+	inode = read_one_inode(root, key->objectid);
+	BUG_ON(!dir);
+
+	ref_ptr = btrfs_item_ptr_offset(eb, slot);
+	ref_end = ref_ptr + btrfs_item_size_nr(eb, slot);
+
+again:
+	ref = (struct btrfs_inode_ref *)ref_ptr;
+
+	namelen = btrfs_inode_ref_name_len(eb, ref);
+	name = kmalloc(namelen, GFP_NOFS);
+	BUG_ON(!name);
+
+	read_extent_buffer(eb, name, (unsigned long)(ref + 1), namelen);
+
+	/* if we already have a perfect match, we're done */
+	if (inode_in_dir(root, path, dir->i_ino, inode->i_ino,
+			 btrfs_inode_ref_index(eb, ref),
+			 name, namelen)) {
+		goto out;
+	}
+
+	/*
+	 * look for a conflicting back reference in the metadata.
+	 * if we find one we have to unlink that name of the file
+	 * before we add our new link.  Later on, we overwrite any
+	 * existing back reference, and we don't want to create
+	 * dangling pointers in the directory.
+	 */
+conflict_again:
+	ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
+	if (ret == 0) {
+		char *victim_name;
+		int victim_name_len;
+		struct btrfs_inode_ref *victim_ref;
+		unsigned long ptr;
+		unsigned long ptr_end;
+		struct extent_buffer *leaf = path->nodes[0];
+
+		/* are we trying to overwrite a back ref for the root directory
+		 * if so, just jump out, we're done
+		 */
+		if (key->objectid == key->offset)
+			goto out_nowrite;
+
+		/* check all the names in this back reference to see
+		 * if they are in the log.  if so, we allow them to stay
+		 * otherwise they must be unlinked as a conflict
+		 */
+		ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
+		ptr_end = ptr + btrfs_item_size_nr(leaf, path->slots[0]);
+		while(ptr < ptr_end) {
+			victim_ref = (struct btrfs_inode_ref *)ptr;
+			victim_name_len = btrfs_inode_ref_name_len(leaf,
+								   victim_ref);
+			victim_name = kmalloc(victim_name_len, GFP_NOFS);
+			BUG_ON(!victim_name);
+
+			read_extent_buffer(leaf, victim_name,
+					   (unsigned long)(victim_ref + 1),
+					   victim_name_len);
+
+			if (!backref_in_log(log, key, victim_name,
+					    victim_name_len)) {
+				btrfs_inc_nlink(inode);
+				btrfs_release_path(root, path);
+				ret = btrfs_unlink_inode(trans, root, dir,
+							 inode, victim_name,
+							 victim_name_len);
+				kfree(victim_name);
+				btrfs_release_path(root, path);
+				goto conflict_again;
+			}
+			kfree(victim_name);
+			ptr = (unsigned long)(victim_ref + 1) + victim_name_len;
+		}
+		BUG_ON(ret);
+	}
+	btrfs_release_path(root, path);
+
+	/* look for a conflicting sequence number */
+	di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
+					 btrfs_inode_ref_index(eb, ref),
+					 name, namelen, 0);
+	if (di && !IS_ERR(di)) {
+		ret = drop_one_dir_item(trans, root, path, dir, di);
+		BUG_ON(ret);
+	}
+	btrfs_release_path(root, path);
+
+
+	/* look for a conflicting name */
+	di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
+				   name, namelen, 0);
+	if (di && !IS_ERR(di)) {
+		ret = drop_one_dir_item(trans, root, path, dir, di);
+		BUG_ON(ret);
+	}
+	btrfs_release_path(root, path);
+
+	/* insert our name */
+	ret = btrfs_add_link(trans, dir, inode, name, namelen, 0,
+			     btrfs_inode_ref_index(eb, ref));
+	BUG_ON(ret);
+
+	btrfs_update_inode(trans, root, inode);
+
+out:
+	ref_ptr = (unsigned long)(ref + 1) + namelen;
+	kfree(name);
+	if (ref_ptr < ref_end)
+		goto again;
+
+	/* finally write the back reference in the inode */
+	ret = overwrite_item(trans, root, path, eb, slot, key);
+	BUG_ON(ret);
+
+out_nowrite:
+	btrfs_release_path(root, path);
+	iput(dir);
+	iput(inode);
+	return 0;
+}
+
+/*
+ * replay one csum item from the log tree into the subvolume 'root'
+ * eb, slot and key all refer to the log tree
+ * path is for temp use by this function and should be released on return
+ *
+ * This copies the checksums out of the log tree and inserts them into
+ * the subvolume.  Any existing checksums for this range in the file
+ * are overwritten, and new items are added where required.
+ *
+ * We keep this simple by reusing the btrfs_ordered_sum code from
+ * the data=ordered mode.  This basically means making a copy
+ * of all the checksums in ram, which we have to do anyway for kmap
+ * rules.
+ *
+ * The copy is then sent down to btrfs_csum_file_blocks, which
+ * does all the hard work of finding existing items in the file
+ * or adding new ones.
+ */
+static noinline int replay_one_csum(struct btrfs_trans_handle *trans,
+				      struct btrfs_root *root,
+				      struct btrfs_path *path,
+				      struct extent_buffer *eb, int slot,
+				      struct btrfs_key *key)
+{
+	int ret;
+	u32 item_size = btrfs_item_size_nr(eb, slot);
+	u64 cur_offset;
+	unsigned long file_bytes;
+	struct btrfs_ordered_sum *sums;
+	struct btrfs_sector_sum *sector_sum;
+	struct inode *inode;
+	unsigned long ptr;
+
+	file_bytes = (item_size / BTRFS_CRC32_SIZE) * root->sectorsize;
+	inode = read_one_inode(root, key->objectid);
+	if (!inode) {
+		return -EIO;
+	}
+
+	sums = kzalloc(btrfs_ordered_sum_size(root, file_bytes), GFP_NOFS);
+	if (!sums) {
+		iput(inode);
+		return -ENOMEM;
+	}
+
+	INIT_LIST_HEAD(&sums->list);
+	sums->len = file_bytes;
+	sums->file_offset = key->offset;
+
+	/*
+	 * copy all the sums into the ordered sum struct
+	 */
+	sector_sum = sums->sums;
+	cur_offset = key->offset;
+	ptr = btrfs_item_ptr_offset(eb, slot);
+	while(item_size > 0) {
+		sector_sum->offset = cur_offset;
+		read_extent_buffer(eb, &sector_sum->sum, ptr, BTRFS_CRC32_SIZE);
+		sector_sum++;
+		item_size -= BTRFS_CRC32_SIZE;
+		ptr += BTRFS_CRC32_SIZE;
+		cur_offset += root->sectorsize;
+	}
+
+	/* let btrfs_csum_file_blocks add them into the file */
+	ret = btrfs_csum_file_blocks(trans, root, inode, sums);
+	BUG_ON(ret);
+	kfree(sums);
+	iput(inode);
+
+	return 0;
+}
+/*
+ * There are a few corners where the link count of the file can't
+ * be properly maintained during replay.  So, instead of adding
+ * lots of complexity to the log code, we just scan the backrefs
+ * for any file that has been through replay.
+ *
+ * The scan will update the link count on the inode to reflect the
+ * number of back refs found.  If it goes down to zero, the iput
+ * will free the inode.
+ */
+static noinline int fixup_inode_link_count(struct btrfs_trans_handle *trans,
+					   struct btrfs_root *root,
+					   struct inode *inode)
+{
+	struct btrfs_path *path;
+	int ret;
+	struct btrfs_key key;
+	u64 nlink = 0;
+	unsigned long ptr;
+	unsigned long ptr_end;
+	int name_len;
+
+	key.objectid = inode->i_ino;
+	key.type = BTRFS_INODE_REF_KEY;
+	key.offset = (u64)-1;
+
+	path = btrfs_alloc_path();
+
+	while(1) {
+		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+		if (ret < 0)
+			break;
+		if (ret > 0) {
+			if (path->slots[0] == 0)
+				break;
+			path->slots[0]--;
+		}
+		btrfs_item_key_to_cpu(path->nodes[0], &key,
+				      path->slots[0]);
+		if (key.objectid != inode->i_ino ||
+		    key.type != BTRFS_INODE_REF_KEY)
+			break;
+		ptr = btrfs_item_ptr_offset(path->nodes[0], path->slots[0]);
+		ptr_end = ptr + btrfs_item_size_nr(path->nodes[0],
+						   path->slots[0]);
+		while(ptr < ptr_end) {
+			struct btrfs_inode_ref *ref;
+
+			ref = (struct btrfs_inode_ref *)ptr;
+			name_len = btrfs_inode_ref_name_len(path->nodes[0],
+							    ref);
+			ptr = (unsigned long)(ref + 1) + name_len;
+			nlink++;
+		}
+
+		if (key.offset == 0)
+			break;
+		key.offset--;
+		btrfs_release_path(root, path);
+	}
+	btrfs_free_path(path);
+	if (nlink != inode->i_nlink) {
+		inode->i_nlink = nlink;
+		btrfs_update_inode(trans, root, inode);
+	}
+
+	return 0;
+}
+
+static noinline int fixup_inode_link_counts(struct btrfs_trans_handle *trans,
+					    struct btrfs_root *root,
+					    struct btrfs_path *path)
+{
+	int ret;
+	struct btrfs_key key;
+	struct inode *inode;
+
+	key.objectid = BTRFS_TREE_LOG_FIXUP_OBJECTID;
+	key.type = BTRFS_ORPHAN_ITEM_KEY;
+	key.offset = (u64)-1;
+	while(1) {
+		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
+		if (ret < 0)
+			break;
+
+		if (ret == 1) {
+			if (path->slots[0] == 0)
+				break;
+			path->slots[0]--;
+		}
+
+		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
+		if (key.objectid != BTRFS_TREE_LOG_FIXUP_OBJECTID ||
+		    key.type != BTRFS_ORPHAN_ITEM_KEY)
+			break;
+
+		ret = btrfs_del_item(trans, root, path);
+		BUG_ON(ret);
+
+		btrfs_release_path(root, path);
+		inode = read_one_inode(root, key.offset);
+		BUG_ON(!inode);
+
+		ret = fixup_inode_link_count(trans, root, inode);
+		BUG_ON(ret);
+
+		iput(inode);
+
+		if (key.offset == 0)
+			break;
+		key.offset--;
+	}
+	btrfs_release_path(root, path);
+	return 0;
+}
+
+
+/*
+ * record a given inode in the fixup dir so we can check its link
+ * count when replay is done.  The link count is incremented here
+ * so the inode won't go away until we check it
+ */
+static noinline int link_to_fixup_dir(struct btrfs_trans_handle *trans,
+				      struct btrfs_root *root,
+				      struct btrfs_path *path,
+				      u64 objectid)
+{
+	struct btrfs_key key;
+	int ret = 0;
+	struct inode *inode;
+
+	inode = read_one_inode(root, objectid);
+	BUG_ON(!inode);
+
+	key.objectid = BTRFS_TREE_LOG_FIXUP_OBJECTID;
+	btrfs_set_key_type(&key, BTRFS_ORPHAN_ITEM_KEY);
+	key.offset = objectid;
+
+	ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
+
+	btrfs_release_path(root, path);
+	if (ret == 0) {
+		btrfs_inc_nlink(inode);
+		btrfs_update_inode(trans, root, inode);
+	} else if (ret == -EEXIST) {
+		ret = 0;
+	} else {
+		BUG();
+	}
+	iput(inode);
+
+	return ret;
+}
+
+/*
+ * when replaying the log for a directory, we only insert names
+ * for inodes that actually exist.  This means an fsync on a directory
+ * does not implicitly fsync all the new files in it
+ */
+static noinline int insert_one_name(struct btrfs_trans_handle *trans,
+				    struct btrfs_root *root,
+				    struct btrfs_path *path,
+				    u64 dirid, u64 index,
+				    char *name, int name_len, u8 type,
+				    struct btrfs_key *location)
+{
+	struct inode *inode;
+	struct inode *dir;
+	int ret;
+
+	inode = read_one_inode(root, location->objectid);
+	if (!inode)
+		return -ENOENT;
+
+	dir = read_one_inode(root, dirid);
+	if (!dir) {
+		iput(inode);
+		return -EIO;
+	}
+	ret = btrfs_add_link(trans, dir, inode, name, name_len, 1, index);
+
+	/* FIXME, put inode into FIXUP list */
+
+	iput(inode);
+	iput(dir);
+	return ret;
+}
+
+/*
+ * take a single entry in a log directory item and replay it into
+ * the subvolume.
+ *
+ * if a conflicting item exists in the subdirectory already,
+ * the inode it points to is unlinked and put into the link count
+ * fix up tree.
+ *
+ * If a name from the log points to a file or directory that does
+ * not exist in the FS, it is skipped.  fsyncs on directories
+ * do not force down inodes inside that directory, just changes to the
+ * names or unlinks in a directory.
+ */
+static noinline int replay_one_name(struct btrfs_trans_handle *trans,
+				    struct btrfs_root *root,
+				    struct btrfs_path *path,
+				    struct extent_buffer *eb,
+				    struct btrfs_dir_item *di,
+				    struct btrfs_key *key)
+{
+	char *name;
+	int name_len;
+	struct btrfs_dir_item *dst_di;
+	struct btrfs_key found_key;
+	struct btrfs_key log_key;
+	struct inode *dir;
+	struct inode *inode;
+	u8 log_type;
+	int ret;
+
+	dir = read_one_inode(root, key->objectid);
+	BUG_ON(!dir);
+
+	name_len = btrfs_dir_name_len(eb, di);
+	name = kmalloc(name_len, GFP_NOFS);
+	log_type = btrfs_dir_type(eb, di);
+	read_extent_buffer(eb, name, (unsigned long)(di + 1),
+		   name_len);
+
+	btrfs_dir_item_key_to_cpu(eb, di, &log_key);
+	if (key->type == BTRFS_DIR_ITEM_KEY) {
+		dst_di = btrfs_lookup_dir_item(trans, root, path, key->objectid,
+				       name, name_len, 1);
+	}
+	else if (key->type == BTRFS_DIR_INDEX_KEY) {
+		dst_di = btrfs_lookup_dir_index_item(trans, root, path,
+						     key->objectid,
+						     key->offset, name,
+						     name_len, 1);
+	} else {
+		BUG();
+	}
+	if (!dst_di || IS_ERR(dst_di)) {
+		/* we need a sequence number to insert, so we only
+		 * do inserts for the BTRFS_DIR_INDEX_KEY types
+		 */
+		if (key->type != BTRFS_DIR_INDEX_KEY)
+			goto out;
+		goto insert;
+	}
+
+	btrfs_dir_item_key_to_cpu(path->nodes[0], dst_di, &found_key);
+	/* the existing item matches the logged item */
+	if (found_key.objectid == log_key.objectid &&
+	    found_key.type == log_key.type &&
+	    found_key.offset == log_key.offset &&
+	    btrfs_dir_type(path->nodes[0], dst_di) == log_type) {
+		goto out;
+	}
+
+	/*
+	 * don't drop the conflicting directory entry if the inode
+	 * for the new entry doesn't exist
+	 */
+	inode = read_one_inode(root, log_key.objectid);
+	if (!inode)
+		goto out;
+
+	iput(inode);
+	ret = drop_one_dir_item(trans, root, path, dir, dst_di);
+	BUG_ON(ret);
+
+	if (key->type == BTRFS_DIR_INDEX_KEY)
+		goto insert;
+out:
+	btrfs_release_path(root, path);
+	kfree(name);
+	iput(dir);
+	return 0;
+
+insert:
+	btrfs_release_path(root, path);
+	ret = insert_one_name(trans, root, path, key->objectid, key->offset,
+			      name, name_len, log_type, &log_key);
+
+	if (ret && ret != -ENOENT)
+		BUG();
+	goto out;
+}
+
+/*
+ * find all the names in a directory item and reconcile them into
+ * the subvolume.  Only BTRFS_DIR_ITEM_KEY types will have more than
+ * one name in a directory item, but the same code gets used for
+ * both directory index types
+ */
+static noinline int replay_one_dir_item(struct btrfs_trans_handle *trans,
+					struct btrfs_root *root,
+					struct btrfs_path *path,
+					struct extent_buffer *eb, int slot,
+					struct btrfs_key *key)
+{
+	int ret;
+	u32 item_size = btrfs_item_size_nr(eb, slot);
+	struct btrfs_dir_item *di;
+	int name_len;
+	unsigned long ptr;
+	unsigned long ptr_end;
+
+	ptr = btrfs_item_ptr_offset(eb, slot);
+	ptr_end = ptr + item_size;
+	while(ptr < ptr_end) {
+		di = (struct btrfs_dir_item *)ptr;
+		name_len = btrfs_dir_name_len(eb, di);
+		ret = replay_one_name(trans, root, path, eb, di, key);
+		BUG_ON(ret);
+		ptr = (unsigned long)(di + 1);
+		ptr += name_len;
+	}
+	return 0;
+}
+
+/*
+ * directory replay has two parts.  There are the standard directory
+ * items in the log copied from the subvolume, and range items
+ * created in the log while the subvolume was logged.
+ *
+ * The range items tell us which parts of the key space the log
+ * is authoritative for.  During replay, if a key in the subvolume
+ * directory is in a logged range item, but not actually in the log
+ * that means it was deleted from the directory before the fsync
+ * and should be removed.
+ */
+static noinline int find_dir_range(struct btrfs_root *root,
+				   struct btrfs_path *path,
+				   u64 dirid, int key_type,
+				   u64 *start_ret, u64 *end_ret)
+{
+	struct btrfs_key key;
+	u64 found_end;
+	struct btrfs_dir_log_item *item;
+	int ret;
+	int nritems;
+
+	if (*start_ret == (u64)-1)
+		return 1;
+
+	key.objectid = dirid;
+	key.type = key_type;
+	key.offset = *start_ret;
+
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		goto out;
+	if (ret > 0) {
+		if (path->slots[0] == 0)
+			goto out;
+		path->slots[0]--;
+	}
+	if (ret != 0)
+		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
+
+	if (key.type != key_type || key.objectid != dirid) {
+		ret = 1;
+		goto next;
+	}
+	item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+			      struct btrfs_dir_log_item);
+	found_end = btrfs_dir_log_end(path->nodes[0], item);
+
+	if (*start_ret >= key.offset && *start_ret <= found_end) {
+		ret = 0;
+		*start_ret = key.offset;
+		*end_ret = found_end;
+		goto out;
+	}
+	ret = 1;
+next:
+	/* check the next slot in the tree to see if it is a valid item */
+	nritems = btrfs_header_nritems(path->nodes[0]);
+	if (path->slots[0] >= nritems) {
+		ret = btrfs_next_leaf(root, path);
+		if (ret)
+			goto out;
+	} else {
+		path->slots[0]++;
+	}
+
+	btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
+
+	if (key.type != key_type || key.objectid != dirid) {
+		ret = 1;
+		goto out;
+	}
+	item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+			      struct btrfs_dir_log_item);
+	found_end = btrfs_dir_log_end(path->nodes[0], item);
+	*start_ret = key.offset;
+	*end_ret = found_end;
+	ret = 0;
+out:
+	btrfs_release_path(root, path);
+	return ret;
+}
+
+/*
+ * this looks for a given directory item in the log.  If the directory
+ * item is not in the log, the item is removed and the inode it points
+ * to is unlinked
+ */
+static noinline int check_item_in_log(struct btrfs_trans_handle *trans,
+				      struct btrfs_root *root,
+				      struct btrfs_root *log,
+				      struct btrfs_path *path,
+				      struct btrfs_path *log_path,
+				      struct inode *dir,
+				      struct btrfs_key *dir_key)
+{
+	int ret;
+	struct extent_buffer *eb;
+	int slot;
+	u32 item_size;
+	struct btrfs_dir_item *di;
+	struct btrfs_dir_item *log_di;
+	int name_len;
+	unsigned long ptr;
+	unsigned long ptr_end;
+	char *name;
+	struct inode *inode;
+	struct btrfs_key location;
+
+again:
+	eb = path->nodes[0];
+	slot = path->slots[0];
+	item_size = btrfs_item_size_nr(eb, slot);
+	ptr = btrfs_item_ptr_offset(eb, slot);
+	ptr_end = ptr + item_size;
+	while(ptr < ptr_end) {
+		di = (struct btrfs_dir_item *)ptr;
+		name_len = btrfs_dir_name_len(eb, di);
+		name = kmalloc(name_len, GFP_NOFS);
+		if (!name) {
+			ret = -ENOMEM;
+			goto out;
+		}
+		read_extent_buffer(eb, name, (unsigned long)(di + 1),
+				  name_len);
+		log_di = NULL;
+		if (dir_key->type == BTRFS_DIR_ITEM_KEY) {
+			log_di = btrfs_lookup_dir_item(trans, log, log_path,
+						       dir_key->objectid,
+						       name, name_len, 0);
+		} else if (dir_key->type == BTRFS_DIR_INDEX_KEY) {
+			log_di = btrfs_lookup_dir_index_item(trans, log,
+						     log_path,
+						     dir_key->objectid,
+						     dir_key->offset,
+						     name, name_len, 0);
+		}
+		if (!log_di || IS_ERR(log_di)) {
+			btrfs_dir_item_key_to_cpu(eb, di, &location);
+			btrfs_release_path(root, path);
+			btrfs_release_path(log, log_path);
+			inode = read_one_inode(root, location.objectid);
+			BUG_ON(!inode);
+
+			ret = link_to_fixup_dir(trans, root,
+						path, location.objectid);
+			BUG_ON(ret);
+			btrfs_inc_nlink(inode);
+			ret = btrfs_unlink_inode(trans, root, dir, inode,
+						 name, name_len);
+			BUG_ON(ret);
+			kfree(name);
+			iput(inode);
+
+			/* there might still be more names under this key
+			 * check and repeat if required
+			 */
+			ret = btrfs_search_slot(NULL, root, dir_key, path,
+						0, 0);
+			if (ret == 0)
+				goto again;
+			ret = 0;
+			goto out;
+		}
+		btrfs_release_path(log, log_path);
+		kfree(name);
+
+		ptr = (unsigned long)(di + 1);
+		ptr += name_len;
+	}
+	ret = 0;
+out:
+	btrfs_release_path(root, path);
+	btrfs_release_path(log, log_path);
+	return ret;
+}
+
+/*
+ * deletion replay happens before we copy any new directory items
+ * out of the log or out of backreferences from inodes.  It
+ * scans the log to find ranges of keys that log is authoritative for,
+ * and then scans the directory to find items in those ranges that are
+ * not present in the log.
+ *
+ * Anything we don't find in the log is unlinked and removed from the
+ * directory.
+ */
+static noinline int replay_dir_deletes(struct btrfs_trans_handle *trans,
+				       struct btrfs_root *root,
+				       struct btrfs_root *log,
+				       struct btrfs_path *path,
+				       u64 dirid)
+{
+	u64 range_start;
+	u64 range_end;
+	int key_type = BTRFS_DIR_LOG_ITEM_KEY;
+	int ret = 0;
+	struct btrfs_key dir_key;
+	struct btrfs_key found_key;
+	struct btrfs_path *log_path;
+	struct inode *dir;
+
+	dir_key.objectid = dirid;
+	dir_key.type = BTRFS_DIR_ITEM_KEY;
+	log_path = btrfs_alloc_path();
+	if (!log_path)
+		return -ENOMEM;
+
+	dir = read_one_inode(root, dirid);
+	/* it isn't an error if the inode isn't there, that can happen
+	 * because we replay the deletes before we copy in the inode item
+	 * from the log
+	 */
+	if (!dir) {
+		btrfs_free_path(log_path);
+		return 0;
+	}
+again:
+	range_start = 0;
+	range_end = 0;
+	while(1) {
+		ret = find_dir_range(log, path, dirid, key_type,
+				     &range_start, &range_end);
+		if (ret != 0)
+			break;
+
+		dir_key.offset = range_start;
+		while(1) {
+			int nritems;
+			ret = btrfs_search_slot(NULL, root, &dir_key, path,
+						0, 0);
+			if (ret < 0)
+				goto out;
+
+			nritems = btrfs_header_nritems(path->nodes[0]);
+			if (path->slots[0] >= nritems) {
+				ret = btrfs_next_leaf(root, path);
+				if (ret)
+					break;
+			}
+			btrfs_item_key_to_cpu(path->nodes[0], &found_key,
+					      path->slots[0]);
+			if (found_key.objectid != dirid ||
+			    found_key.type != dir_key.type)
+				goto next_type;
+
+			if (found_key.offset > range_end)
+				break;
+
+			ret = check_item_in_log(trans, root, log, path,
+						log_path, dir, &found_key);
+			BUG_ON(ret);
+			if (found_key.offset == (u64)-1)
+				break;
+			dir_key.offset = found_key.offset + 1;
+		}
+		btrfs_release_path(root, path);
+		if (range_end == (u64)-1)
+			break;
+		range_start = range_end + 1;
+	}
+
+next_type:
+	ret = 0;
+	if (key_type == BTRFS_DIR_LOG_ITEM_KEY) {
+		key_type = BTRFS_DIR_LOG_INDEX_KEY;
+		dir_key.type = BTRFS_DIR_INDEX_KEY;
+		btrfs_release_path(root, path);
+		goto again;
+	}
+out:
+	btrfs_release_path(root, path);
+	btrfs_free_path(log_path);
+	iput(dir);
+	return ret;
+}
+
+/*
+ * the process_func used to replay items from the log tree.  This
+ * gets called in two different stages.  The first stage just looks
+ * for inodes and makes sure they are all copied into the subvolume.
+ *
+ * The second stage copies all the other item types from the log into
+ * the subvolume.  The two stage approach is slower, but gets rid of
+ * lots of complexity around inodes referencing other inodes that exist
+ * only in the log (references come from either directory items or inode
+ * back refs).
+ */
+static int replay_one_buffer(struct btrfs_root *log, struct extent_buffer *eb,
+			     struct walk_control *wc, u64 gen)
+{
+	int nritems;
+	struct btrfs_path *path;
+	struct btrfs_root *root = wc->replay_dest;
+	struct btrfs_key key;
+	u32 item_size;
+	int level;
+	int i;
+	int ret;
+
+	btrfs_read_buffer(eb, gen);
+
+	level = btrfs_header_level(eb);
+
+	if (level != 0)
+		return 0;
+
+	path = btrfs_alloc_path();
+	BUG_ON(!path);
+
+	nritems = btrfs_header_nritems(eb);
+	for (i = 0; i < nritems; i++) {
+		btrfs_item_key_to_cpu(eb, &key, i);
+		item_size = btrfs_item_size_nr(eb, i);
+
+		/* inode keys are done during the first stage */
+		if (key.type == BTRFS_INODE_ITEM_KEY &&
+		    wc->stage == LOG_WALK_REPLAY_INODES) {
+			struct inode *inode;
+			struct btrfs_inode_item *inode_item;
+			u32 mode;
+
+			inode_item = btrfs_item_ptr(eb, i,
+					    struct btrfs_inode_item);
+			mode = btrfs_inode_mode(eb, inode_item);
+			if (S_ISDIR(mode)) {
+				ret = replay_dir_deletes(wc->trans,
+					 root, log, path, key.objectid);
+				BUG_ON(ret);
+			}
+			ret = overwrite_item(wc->trans, root, path,
+					     eb, i, &key);
+			BUG_ON(ret);
+
+			/* for regular files, truncate away
+			 * extents past the new EOF
+			 */
+			if (S_ISREG(mode)) {
+				inode = read_one_inode(root,
+						       key.objectid);
+				BUG_ON(!inode);
+
+				ret = btrfs_truncate_inode_items(wc->trans,
+					root, inode, inode->i_size,
+					BTRFS_EXTENT_DATA_KEY);
+				BUG_ON(ret);
+				iput(inode);
+			}
+			ret = link_to_fixup_dir(wc->trans, root,
+						path, key.objectid);
+			BUG_ON(ret);
+		}
+		if (wc->stage < LOG_WALK_REPLAY_ALL)
+			continue;
+
+		/* these keys are simply copied */
+		if (key.type == BTRFS_XATTR_ITEM_KEY) {
+			ret = overwrite_item(wc->trans, root, path,
+					     eb, i, &key);
+			BUG_ON(ret);
+		} else if (key.type == BTRFS_INODE_REF_KEY) {
+			ret = add_inode_ref(wc->trans, root, log, path,
+					    eb, i, &key);
+			BUG_ON(ret && ret != -ENOENT);
+		} else if (key.type == BTRFS_EXTENT_DATA_KEY) {
+			ret = replay_one_extent(wc->trans, root, path,
+						eb, i, &key);
+			BUG_ON(ret);
+		} else if (key.type == BTRFS_CSUM_ITEM_KEY) {
+			ret = replay_one_csum(wc->trans, root, path,
+					      eb, i, &key);
+			BUG_ON(ret);
+		} else if (key.type == BTRFS_DIR_ITEM_KEY ||
+			   key.type == BTRFS_DIR_INDEX_KEY) {
+			ret = replay_one_dir_item(wc->trans, root, path,
+						  eb, i, &key);
+			BUG_ON(ret);
+		}
+	}
+	btrfs_free_path(path);
+	return 0;
+}
+
+static int noinline walk_down_log_tree(struct btrfs_trans_handle *trans,
+				   struct btrfs_root *root,
+				   struct btrfs_path *path, int *level,
+				   struct walk_control *wc)
+{
+	u64 root_owner;
+	u64 root_gen;
+	u64 bytenr;
+	u64 ptr_gen;
+	struct extent_buffer *next;
+	struct extent_buffer *cur;
+	struct extent_buffer *parent;
+	u32 blocksize;
+	int ret = 0;
+
+	WARN_ON(*level < 0);
+	WARN_ON(*level >= BTRFS_MAX_LEVEL);
+
+	while(*level > 0) {
+		WARN_ON(*level < 0);
+		WARN_ON(*level >= BTRFS_MAX_LEVEL);
+		cur = path->nodes[*level];
+
+		if (btrfs_header_level(cur) != *level)
+			WARN_ON(1);
+
+		if (path->slots[*level] >=
+		    btrfs_header_nritems(cur))
+			break;
+
+		bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
+		ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
+		blocksize = btrfs_level_size(root, *level - 1);
+
+		parent = path->nodes[*level];
+		root_owner = btrfs_header_owner(parent);
+		root_gen = btrfs_header_generation(parent);
+
+		next = btrfs_find_create_tree_block(root, bytenr, blocksize);
+
+		wc->process_func(root, next, wc, ptr_gen);
+
+		if (*level == 1) {
+			path->slots[*level]++;
+			if (wc->free) {
+				btrfs_read_buffer(next, ptr_gen);
+
+				btrfs_tree_lock(next);
+				clean_tree_block(trans, root, next);
+				btrfs_wait_tree_block_writeback(next);
+				btrfs_tree_unlock(next);
+
+				ret = btrfs_drop_leaf_ref(trans, root, next);
+				BUG_ON(ret);
+
+				WARN_ON(root_owner !=
+					BTRFS_TREE_LOG_OBJECTID);
+				ret = btrfs_free_extent(trans, root, bytenr,
+							blocksize, root_owner,
+							root_gen, 0, 0, 1);
+				BUG_ON(ret);
+			}
+			free_extent_buffer(next);
+			continue;
+		}
+		btrfs_read_buffer(next, ptr_gen);
+
+		WARN_ON(*level <= 0);
+		if (path->nodes[*level-1])
+			free_extent_buffer(path->nodes[*level-1]);
+		path->nodes[*level-1] = next;
+		*level = btrfs_header_level(next);
+		path->slots[*level] = 0;
+		cond_resched();
+	}
+	WARN_ON(*level < 0);
+	WARN_ON(*level >= BTRFS_MAX_LEVEL);
+
+	if (path->nodes[*level] == root->node) {
+		parent = path->nodes[*level];
+	} else {
+		parent = path->nodes[*level + 1];
+	}
+	bytenr = path->nodes[*level]->start;
+
+	blocksize = btrfs_level_size(root, *level);
+	root_owner = btrfs_header_owner(parent);
+	root_gen = btrfs_header_generation(parent);
+
+	wc->process_func(root, path->nodes[*level], wc,
+			 btrfs_header_generation(path->nodes[*level]));
+
+	if (wc->free) {
+		next = path->nodes[*level];
+		btrfs_tree_lock(next);
+		clean_tree_block(trans, root, next);
+		btrfs_wait_tree_block_writeback(next);
+		btrfs_tree_unlock(next);
+
+		if (*level == 0) {
+			ret = btrfs_drop_leaf_ref(trans, root, next);
+			BUG_ON(ret);
+		}
+		WARN_ON(root_owner != BTRFS_TREE_LOG_OBJECTID);
+		ret = btrfs_free_extent(trans, root, bytenr, blocksize,
+					  root_owner, root_gen, 0, 0, 1);
+		BUG_ON(ret);
+	}
+	free_extent_buffer(path->nodes[*level]);
+	path->nodes[*level] = NULL;
+	*level += 1;
+
+	cond_resched();
+	return 0;
+}
+
+static int noinline walk_up_log_tree(struct btrfs_trans_handle *trans,
+				 struct btrfs_root *root,
+				 struct btrfs_path *path, int *level,
+				 struct walk_control *wc)
+{
+	u64 root_owner;
+	u64 root_gen;
+	int i;
+	int slot;
+	int ret;
+
+	for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
+		slot = path->slots[i];
+		if (slot < btrfs_header_nritems(path->nodes[i]) - 1) {
+			struct extent_buffer *node;
+			node = path->nodes[i];
+			path->slots[i]++;
+			*level = i;
+			WARN_ON(*level == 0);
+			return 0;
+		} else {
+			if (path->nodes[*level] == root->node) {
+				root_owner = root->root_key.objectid;
+				root_gen =
+				   btrfs_header_generation(path->nodes[*level]);
+			} else {
+				struct extent_buffer *node;
+				node = path->nodes[*level + 1];
+				root_owner = btrfs_header_owner(node);
+				root_gen = btrfs_header_generation(node);
+			}
+			wc->process_func(root, path->nodes[*level], wc,
+				 btrfs_header_generation(path->nodes[*level]));
+			if (wc->free) {
+				struct extent_buffer *next;
+
+				next = path->nodes[*level];
+
+				btrfs_tree_lock(next);
+				clean_tree_block(trans, root, next);
+				btrfs_wait_tree_block_writeback(next);
+				btrfs_tree_unlock(next);
+
+				if (*level == 0) {
+					ret = btrfs_drop_leaf_ref(trans, root,
+								  next);
+					BUG_ON(ret);
+				}
+
+				WARN_ON(root_owner != BTRFS_TREE_LOG_OBJECTID);
+				ret = btrfs_free_extent(trans, root,
+						path->nodes[*level]->start,
+						path->nodes[*level]->len,
+						root_owner, root_gen, 0, 0, 1);
+				BUG_ON(ret);
+			}
+			free_extent_buffer(path->nodes[*level]);
+			path->nodes[*level] = NULL;
+			*level = i + 1;
+		}
+	}
+	return 1;
+}
+
+/*
+ * drop the reference count on the tree rooted at 'snap'.  This traverses
+ * the tree freeing any blocks that have a ref count of zero after being
+ * decremented.
+ */
+static int walk_log_tree(struct btrfs_trans_handle *trans,
+			 struct btrfs_root *log, struct walk_control *wc)
+{
+	int ret = 0;
+	int wret;
+	int level;
+	struct btrfs_path *path;
+	int i;
+	int orig_level;
+
+	path = btrfs_alloc_path();
+	BUG_ON(!path);
+
+	level = btrfs_header_level(log->node);
+	orig_level = level;
+	path->nodes[level] = log->node;
+	extent_buffer_get(log->node);
+	path->slots[level] = 0;
+
+	while(1) {
+		wret = walk_down_log_tree(trans, log, path, &level, wc);
+		if (wret > 0)
+			break;
+		if (wret < 0)
+			ret = wret;
+
+		wret = walk_up_log_tree(trans, log, path, &level, wc);
+		if (wret > 0)
+			break;
+		if (wret < 0)
+			ret = wret;
+	}
+
+	/* was the root node processed? if not, catch it here */
+	if (path->nodes[orig_level]) {
+		wc->process_func(log, path->nodes[orig_level], wc,
+			 btrfs_header_generation(path->nodes[orig_level]));
+		if (wc->free) {
+			struct extent_buffer *next;
+
+			next = path->nodes[orig_level];
+
+			btrfs_tree_lock(next);
+			clean_tree_block(trans, log, next);
+			btrfs_wait_tree_block_writeback(next);
+			btrfs_tree_unlock(next);
+
+			if (orig_level == 0) {
+				ret = btrfs_drop_leaf_ref(trans, log,
+							  next);
+				BUG_ON(ret);
+			}
+			WARN_ON(log->root_key.objectid !=
+				BTRFS_TREE_LOG_OBJECTID);
+			ret = btrfs_free_extent(trans, log,
+						next->start, next->len,
+						log->root_key.objectid,
+						btrfs_header_generation(next),
+						0, 0, 1);
+			BUG_ON(ret);
+		}
+	}
+
+	for (i = 0; i <= orig_level; i++) {
+		if (path->nodes[i]) {
+			free_extent_buffer(path->nodes[i]);
+			path->nodes[i] = NULL;
+		}
+	}
+	btrfs_free_path(path);
+	if (wc->free)
+		free_extent_buffer(log->node);
+	return ret;
+}
+
+int wait_log_commit(struct btrfs_root *log)
+{
+	DEFINE_WAIT(wait);
+	u64 transid = log->fs_info->tree_log_transid;
+
+	do {
+		prepare_to_wait(&log->fs_info->tree_log_wait, &wait,
+				TASK_UNINTERRUPTIBLE);
+		mutex_unlock(&log->fs_info->tree_log_mutex);
+		if (atomic_read(&log->fs_info->tree_log_commit))
+			schedule();
+		finish_wait(&log->fs_info->tree_log_wait, &wait);
+		mutex_lock(&log->fs_info->tree_log_mutex);
+	} while(transid == log->fs_info->tree_log_transid &&
+		atomic_read(&log->fs_info->tree_log_commit));
+	return 0;
+}
+
+/*
+ * btrfs_sync_log does sends a given tree log down to the disk and
+ * updates the super blocks to record it.  When this call is done,
+ * you know that any inodes previously logged are safely on disk
+ */
+int btrfs_sync_log(struct btrfs_trans_handle *trans,
+		   struct btrfs_root *root)
+{
+	int ret;
+	unsigned long batch;
+	struct btrfs_root *log = root->log_root;
+	struct walk_control wc = {
+		.write = 1,
+		.process_func = process_one_buffer
+	};
+
+	mutex_lock(&log->fs_info->tree_log_mutex);
+	if (atomic_read(&log->fs_info->tree_log_commit)) {
+		wait_log_commit(log);
+		goto out;
+	}
+	atomic_set(&log->fs_info->tree_log_commit, 1);
+
+	while(1) {
+		mutex_unlock(&log->fs_info->tree_log_mutex);
+		schedule_timeout_uninterruptible(1);
+		mutex_lock(&log->fs_info->tree_log_mutex);
+		batch = log->fs_info->tree_log_batch;
+
+		while(atomic_read(&log->fs_info->tree_log_writers)) {
+			DEFINE_WAIT(wait);
+			prepare_to_wait(&log->fs_info->tree_log_wait, &wait,
+					TASK_UNINTERRUPTIBLE);
+			batch = log->fs_info->tree_log_batch;
+			mutex_unlock(&log->fs_info->tree_log_mutex);
+			if (atomic_read(&log->fs_info->tree_log_writers))
+				schedule();
+			mutex_lock(&log->fs_info->tree_log_mutex);
+			finish_wait(&log->fs_info->tree_log_wait, &wait);
+		}
+		if (batch == log->fs_info->tree_log_batch)
+			break;
+	}
+	ret = walk_log_tree(trans, log, &wc);
+	BUG_ON(ret);
+
+	ret = walk_log_tree(trans, log->fs_info->log_root_tree, &wc);
+	BUG_ON(ret);
+
+	wc.wait = 1;
+
+	ret = walk_log_tree(trans, log, &wc);
+	BUG_ON(ret);
+
+	ret = walk_log_tree(trans, log->fs_info->log_root_tree, &wc);
+	BUG_ON(ret);
+
+	btrfs_set_super_log_root(&root->fs_info->super_for_commit,
+				 log->fs_info->log_root_tree->node->start);
+	btrfs_set_super_log_root_level(&root->fs_info->super_for_commit,
+		       btrfs_header_level(log->fs_info->log_root_tree->node));
+
+	write_ctree_super(trans, log->fs_info->tree_root);
+	log->fs_info->tree_log_transid++;
+	log->fs_info->tree_log_batch = 0;
+	atomic_set(&log->fs_info->tree_log_commit, 0);
+	smp_mb();
+	if (waitqueue_active(&log->fs_info->tree_log_wait))
+		wake_up(&log->fs_info->tree_log_wait);
+out:
+	mutex_unlock(&log->fs_info->tree_log_mutex);
+	return 0;
+
+}
+
+/*
+ * free all the extents used by the tree log.  This should be called
+ * at commit time of the full transaction
+ */
+int btrfs_free_log(struct btrfs_trans_handle *trans, struct btrfs_root *root)
+{
+	int ret;
+	struct btrfs_root *log;
+	struct key;
+	struct walk_control wc = {
+		.free = 1,
+		.process_func = process_one_buffer
+	};
+
+	if (!root->log_root)
+		return 0;
+
+	log = root->log_root;
+	ret = walk_log_tree(trans, log, &wc);
+	BUG_ON(ret);
+
+	log = root->log_root;
+	ret = btrfs_del_root(trans, root->fs_info->log_root_tree,
+			     &log->root_key);
+	BUG_ON(ret);
+	root->log_root = NULL;
+	kfree(root->log_root);
+	return 0;
+}
+
+/*
+ * helper function to update the item for a given subvolumes log root
+ * in the tree of log roots
+ */
+static int update_log_root(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *log)
+{
+	u64 bytenr = btrfs_root_bytenr(&log->root_item);
+	int ret;
+
+	if (log->node->start == bytenr)
+		return 0;
+
+	btrfs_set_root_bytenr(&log->root_item, log->node->start);
+	btrfs_set_root_level(&log->root_item, btrfs_header_level(log->node));
+	ret = btrfs_update_root(trans, log->fs_info->log_root_tree,
+				&log->root_key, &log->root_item);
+	BUG_ON(ret);
+	return ret;
+}
+
+/*
+ * If both a file and directory are logged, and unlinks or renames are
+ * mixed in, we have a few interesting corners:
+ *
+ * create file X in dir Y
+ * link file X to X.link in dir Y
+ * fsync file X
+ * unlink file X but leave X.link
+ * fsync dir Y
+ *
+ * After a crash we would expect only X.link to exist.  But file X
+ * didn't get fsync'd again so the log has back refs for X and X.link.
+ *
+ * We solve this by removing directory entries and inode backrefs from the
+ * log when a file that was logged in the current transaction is
+ * unlinked.  Any later fsync will include the updated log entries, and
+ * we'll be able to reconstruct the proper directory items from backrefs.
+ *
+ * This optimizations allows us to avoid relogging the entire inode
+ * or the entire directory.
+ */
+int btrfs_del_dir_entries_in_log(struct btrfs_trans_handle *trans,
+				 struct btrfs_root *root,
+				 const char *name, int name_len,
+				 struct inode *dir, u64 index)
+{
+	struct btrfs_root *log;
+	struct btrfs_dir_item *di;
+	struct btrfs_path *path;
+	int ret;
+	int bytes_del = 0;
+
+	ret = join_running_log_trans(root);
+	if (ret)
+		return 0;
+
+	mutex_lock(&BTRFS_I(dir)->log_mutex);
+
+	log = root->log_root;
+	path = btrfs_alloc_path();
+	di = btrfs_lookup_dir_item(trans, log, path, dir->i_ino,
+				   name, name_len, -1);
+	if (di && !IS_ERR(di)) {
+		ret = btrfs_delete_one_dir_name(trans, log, path, di);
+		bytes_del += name_len;
+		BUG_ON(ret);
+	}
+	btrfs_release_path(log, path);
+	di = btrfs_lookup_dir_index_item(trans, log, path, dir->i_ino,
+					 index, name, name_len, -1);
+	if (di && !IS_ERR(di)) {
+		ret = btrfs_delete_one_dir_name(trans, log, path, di);
+		bytes_del += name_len;
+		BUG_ON(ret);
+	}
+
+	/* update the directory size in the log to reflect the names
+	 * we have removed
+	 */
+	if (bytes_del) {
+		struct btrfs_key key;
+
+		key.objectid = dir->i_ino;
+		key.offset = 0;
+		key.type = BTRFS_INODE_ITEM_KEY;
+		btrfs_release_path(log, path);
+
+		ret = btrfs_search_slot(trans, log, &key, path, 0, 1);
+		if (ret == 0) {
+			struct btrfs_inode_item *item;
+			u64 i_size;
+
+			item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+					      struct btrfs_inode_item);
+			i_size = btrfs_inode_size(path->nodes[0], item);
+			if (i_size > bytes_del)
+				i_size -= bytes_del;
+			else
+				i_size = 0;
+			btrfs_set_inode_size(path->nodes[0], item, i_size);
+			btrfs_mark_buffer_dirty(path->nodes[0]);
+		} else
+			ret = 0;
+		btrfs_release_path(log, path);
+	}
+
+	btrfs_free_path(path);
+	mutex_unlock(&BTRFS_I(dir)->log_mutex);
+	end_log_trans(root);
+
+	return 0;
+}
+
+/* see comments for btrfs_del_dir_entries_in_log */
+int btrfs_del_inode_ref_in_log(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root,
+			       const char *name, int name_len,
+			       struct inode *inode, u64 dirid)
+{
+	struct btrfs_root *log;
+	u64 index;
+	int ret;
+
+	ret = join_running_log_trans(root);
+	if (ret)
+		return 0;
+	log = root->log_root;
+	mutex_lock(&BTRFS_I(inode)->log_mutex);
+
+	ret = btrfs_del_inode_ref(trans, log, name, name_len, inode->i_ino,
+				  dirid, &index);
+	mutex_unlock(&BTRFS_I(inode)->log_mutex);
+	end_log_trans(root);
+
+	if (ret == 0 || ret == -ENOENT)
+		return 0;
+	return ret;
+}
+
+/*
+ * creates a range item in the log for 'dirid'.  first_offset and
+ * last_offset tell us which parts of the key space the log should
+ * be considered authoritative for.
+ */
+static noinline int insert_dir_log_key(struct btrfs_trans_handle *trans,
+				       struct btrfs_root *log,
+				       struct btrfs_path *path,
+				       int key_type, u64 dirid,
+				       u64 first_offset, u64 last_offset)
+{
+	int ret;
+	struct btrfs_key key;
+	struct btrfs_dir_log_item *item;
+
+	key.objectid = dirid;
+	key.offset = first_offset;
+	if (key_type == BTRFS_DIR_ITEM_KEY)
+		key.type = BTRFS_DIR_LOG_ITEM_KEY;
+	else
+		key.type = BTRFS_DIR_LOG_INDEX_KEY;
+	ret = btrfs_insert_empty_item(trans, log, path, &key, sizeof(*item));
+	BUG_ON(ret);
+
+	item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+			      struct btrfs_dir_log_item);
+	btrfs_set_dir_log_end(path->nodes[0], item, last_offset);
+	btrfs_mark_buffer_dirty(path->nodes[0]);
+	btrfs_release_path(log, path);
+	return 0;
+}
+
+/*
+ * log all the items included in the current transaction for a given
+ * directory.  This also creates the range items in the log tree required
+ * to replay anything deleted before the fsync
+ */
+static noinline int log_dir_items(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root, struct inode *inode,
+			  struct btrfs_path *path,
+			  struct btrfs_path *dst_path, int key_type,
+			  u64 min_offset, u64 *last_offset_ret)
+{
+	struct btrfs_key min_key;
+	struct btrfs_key max_key;
+	struct btrfs_root *log = root->log_root;
+	struct extent_buffer *src;
+	int ret;
+	int i;
+	int nritems;
+	u64 first_offset = min_offset;
+	u64 last_offset = (u64)-1;
+
+	log = root->log_root;
+	max_key.objectid = inode->i_ino;
+	max_key.offset = (u64)-1;
+	max_key.type = key_type;
+
+	min_key.objectid = inode->i_ino;
+	min_key.type = key_type;
+	min_key.offset = min_offset;
+
+	path->keep_locks = 1;
+
+	ret = btrfs_search_forward(root, &min_key, &max_key,
+				   path, 0, trans->transid);
+
+	/*
+	 * we didn't find anything from this transaction, see if there
+	 * is anything at all
+	 */
+	if (ret != 0 || min_key.objectid != inode->i_ino ||
+	    min_key.type != key_type) {
+		min_key.objectid = inode->i_ino;
+		min_key.type = key_type;
+		min_key.offset = (u64)-1;
+		btrfs_release_path(root, path);
+		ret = btrfs_search_slot(NULL, root, &min_key, path, 0, 0);
+		if (ret < 0) {
+			btrfs_release_path(root, path);
+			return ret;
+		}
+		ret = btrfs_previous_item(root, path, inode->i_ino, key_type);
+
+		/* if ret == 0 there are items for this type,
+		 * create a range to tell us the last key of this type.
+		 * otherwise, there are no items in this directory after
+		 * *min_offset, and we create a range to indicate that.
+		 */
+		if (ret == 0) {
+			struct btrfs_key tmp;
+			btrfs_item_key_to_cpu(path->nodes[0], &tmp,
+					      path->slots[0]);
+			if (key_type == tmp.type) {
+				first_offset = max(min_offset, tmp.offset) + 1;
+			}
+		}
+		goto done;
+	}
+
+	/* go backward to find any previous key */
+	ret = btrfs_previous_item(root, path, inode->i_ino, key_type);
+	if (ret == 0) {
+		struct btrfs_key tmp;
+		btrfs_item_key_to_cpu(path->nodes[0], &tmp, path->slots[0]);
+		if (key_type == tmp.type) {
+			first_offset = tmp.offset;
+			ret = overwrite_item(trans, log, dst_path,
+					     path->nodes[0], path->slots[0],
+					     &tmp);
+		}
+	}
+	btrfs_release_path(root, path);
+
+	/* find the first key from this transaction again */
+	ret = btrfs_search_slot(NULL, root, &min_key, path, 0, 0);
+	if (ret != 0) {
+		WARN_ON(1);
+		goto done;
+	}
+
+	/*
+	 * we have a block from this transaction, log every item in it
+	 * from our directory
+	 */
+	while(1) {
+		struct btrfs_key tmp;
+		src = path->nodes[0];
+		nritems = btrfs_header_nritems(src);
+		for (i = path->slots[0]; i < nritems; i++) {
+			btrfs_item_key_to_cpu(src, &min_key, i);
+
+			if (min_key.objectid != inode->i_ino ||
+			    min_key.type != key_type)
+				goto done;
+			ret = overwrite_item(trans, log, dst_path, src, i,
+					     &min_key);
+			BUG_ON(ret);
+		}
+		path->slots[0] = nritems;
+
+		/*
+		 * look ahead to the next item and see if it is also
+		 * from this directory and from this transaction
+		 */
+		ret = btrfs_next_leaf(root, path);
+		if (ret == 1) {
+			last_offset = (u64)-1;
+			goto done;
+		}
+		btrfs_item_key_to_cpu(path->nodes[0], &tmp, path->slots[0]);
+		if (tmp.objectid != inode->i_ino || tmp.type != key_type) {
+			last_offset = (u64)-1;
+			goto done;
+		}
+		if (btrfs_header_generation(path->nodes[0]) != trans->transid) {
+			ret = overwrite_item(trans, log, dst_path,
+					     path->nodes[0], path->slots[0],
+					     &tmp);
+
+			BUG_ON(ret);
+			last_offset = tmp.offset;
+			goto done;
+		}
+	}
+done:
+	*last_offset_ret = last_offset;
+	btrfs_release_path(root, path);
+	btrfs_release_path(log, dst_path);
+
+	/* insert the log range keys to indicate where the log is valid */
+	ret = insert_dir_log_key(trans, log, path, key_type, inode->i_ino,
+				 first_offset, last_offset);
+	BUG_ON(ret);
+	return 0;
+}
+
+/*
+ * logging directories is very similar to logging inodes, We find all the items
+ * from the current transaction and write them to the log.
+ *
+ * The recovery code scans the directory in the subvolume, and if it finds a
+ * key in the range logged that is not present in the log tree, then it means
+ * that dir entry was unlinked during the transaction.
+ *
+ * In order for that scan to work, we must include one key smaller than
+ * the smallest logged by this transaction and one key larger than the largest
+ * key logged by this transaction.
+ */
+static noinline int log_directory_changes(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root, struct inode *inode,
+			  struct btrfs_path *path,
+			  struct btrfs_path *dst_path)
+{
+	u64 min_key;
+	u64 max_key;
+	int ret;
+	int key_type = BTRFS_DIR_ITEM_KEY;
+
+again:
+	min_key = 0;
+	max_key = 0;
+	while(1) {
+		ret = log_dir_items(trans, root, inode, path,
+				    dst_path, key_type, min_key,
+				    &max_key);
+		BUG_ON(ret);
+		if (max_key == (u64)-1)
+			break;
+		min_key = max_key + 1;
+	}
+
+	if (key_type == BTRFS_DIR_ITEM_KEY) {
+		key_type = BTRFS_DIR_INDEX_KEY;
+		goto again;
+	}
+	return 0;
+}
+
+/*
+ * a helper function to drop items from the log before we relog an
+ * inode.  max_key_type indicates the highest item type to remove.
+ * This cannot be run for file data extents because it does not
+ * free the extents they point to.
+ */
+static int drop_objectid_items(struct btrfs_trans_handle *trans,
+				  struct btrfs_root *log,
+				  struct btrfs_path *path,
+				  u64 objectid, int max_key_type)
+{
+	int ret;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+
+	key.objectid = objectid;
+	key.type = max_key_type;
+	key.offset = (u64)-1;
+
+	while(1) {
+		ret = btrfs_search_slot(trans, log, &key, path, -1, 1);
+
+		if (ret != 1)
+			break;
+
+		if (path->slots[0] == 0)
+			break;
+
+		path->slots[0]--;
+		btrfs_item_key_to_cpu(path->nodes[0], &found_key,
+				      path->slots[0]);
+
+		if (found_key.objectid != objectid)
+			break;
+
+		ret = btrfs_del_item(trans, log, path);
+		BUG_ON(ret);
+		btrfs_release_path(log, path);
+	}
+	btrfs_release_path(log, path);
+	return 0;
+}
+
+/* log a single inode in the tree log.
+ * At least one parent directory for this inode must exist in the tree
+ * or be logged already.
+ *
+ * Any items from this inode changed by the current transaction are copied
+ * to the log tree.  An extra reference is taken on any extents in this
+ * file, allowing us to avoid a whole pile of corner cases around logging
+ * blocks that have been removed from the tree.
+ *
+ * See LOG_INODE_ALL and related defines for a description of what inode_only
+ * does.
+ *
+ * This handles both files and directories.
+ */
+static int __btrfs_log_inode(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root, struct inode *inode,
+			     int inode_only)
+{
+	struct btrfs_path *path;
+	struct btrfs_path *dst_path;
+	struct btrfs_key min_key;
+	struct btrfs_key max_key;
+	struct btrfs_root *log = root->log_root;
+	unsigned long src_offset;
+	unsigned long dst_offset;
+	struct extent_buffer *src;
+	struct btrfs_file_extent_item *extent;
+	struct btrfs_inode_item *inode_item;
+	u32 size;
+	int ret;
+
+	log = root->log_root;
+
+	path = btrfs_alloc_path();
+	dst_path = btrfs_alloc_path();
+
+	min_key.objectid = inode->i_ino;
+	min_key.type = BTRFS_INODE_ITEM_KEY;
+	min_key.offset = 0;
+
+	max_key.objectid = inode->i_ino;
+	if (inode_only == LOG_INODE_EXISTS || S_ISDIR(inode->i_mode))
+		max_key.type = BTRFS_XATTR_ITEM_KEY;
+	else
+		max_key.type = (u8)-1;
+	max_key.offset = (u64)-1;
+
+	/*
+	 * if this inode has already been logged and we're in inode_only
+	 * mode, we don't want to delete the things that have already
+	 * been written to the log.
+	 *
+	 * But, if the inode has been through an inode_only log,
+	 * the logged_trans field is not set.  This allows us to catch
+	 * any new names for this inode in the backrefs by logging it
+	 * again
+	 */
+	if (inode_only == LOG_INODE_EXISTS &&
+	    BTRFS_I(inode)->logged_trans == trans->transid) {
+		btrfs_free_path(path);
+		btrfs_free_path(dst_path);
+		goto out;
+	}
+	mutex_lock(&BTRFS_I(inode)->log_mutex);
+
+	/*
+	 * a brute force approach to making sure we get the most uptodate
+	 * copies of everything.
+	 */
+	if (S_ISDIR(inode->i_mode)) {
+		int max_key_type = BTRFS_DIR_LOG_INDEX_KEY;
+
+		if (inode_only == LOG_INODE_EXISTS)
+			max_key_type = BTRFS_XATTR_ITEM_KEY;
+		ret = drop_objectid_items(trans, log, path,
+					  inode->i_ino, max_key_type);
+	} else {
+		ret = btrfs_truncate_inode_items(trans, log, inode, 0, 0);
+	}
+	BUG_ON(ret);
+	path->keep_locks = 1;
+
+	while(1) {
+		ret = btrfs_search_forward(root, &min_key, &max_key,
+					   path, 0, trans->transid);
+		if (ret != 0)
+			break;
+
+		if (min_key.objectid != inode->i_ino)
+			break;
+		if (min_key.type > max_key.type)
+			break;
+
+		src = path->nodes[0];
+		size = btrfs_item_size_nr(src, path->slots[0]);
+		ret = btrfs_insert_empty_item(trans, log, dst_path, &min_key,
+					      size);
+		if (ret)
+			BUG();
+
+		dst_offset = btrfs_item_ptr_offset(dst_path->nodes[0],
+						   dst_path->slots[0]);
+
+		src_offset = btrfs_item_ptr_offset(src, path->slots[0]);
+
+		copy_extent_buffer(dst_path->nodes[0], src, dst_offset,
+				   src_offset, size);
+
+		if (inode_only == LOG_INODE_EXISTS &&
+		    min_key.type == BTRFS_INODE_ITEM_KEY) {
+			inode_item = btrfs_item_ptr(dst_path->nodes[0],
+						    dst_path->slots[0],
+						    struct btrfs_inode_item);
+			btrfs_set_inode_size(dst_path->nodes[0], inode_item, 0);
+
+			/* set the generation to zero so the recover code
+			 * can tell the difference between an logging
+			 * just to say 'this inode exists' and a logging
+			 * to say 'update this inode with these values'
+			 */
+			btrfs_set_inode_generation(dst_path->nodes[0],
+						   inode_item, 0);
+		}
+		/* take a reference on file data extents so that truncates
+		 * or deletes of this inode don't have to relog the inode
+		 * again
+		 */
+		if (btrfs_key_type(&min_key) == BTRFS_EXTENT_DATA_KEY) {
+			int found_type;
+			extent = btrfs_item_ptr(src, path->slots[0],
+						struct btrfs_file_extent_item);
+
+			found_type = btrfs_file_extent_type(src, extent);
+			if (found_type == BTRFS_FILE_EXTENT_REG) {
+				u64 ds = btrfs_file_extent_disk_bytenr(src,
+								   extent);
+				u64 dl = btrfs_file_extent_disk_num_bytes(src,
+								      extent);
+				/* ds == 0 is a hole */
+				if (ds != 0) {
+					ret = btrfs_inc_extent_ref(trans, log,
+						   ds, dl,
+						   log->root_key.objectid,
+						   0,
+						   inode->i_ino,
+						   min_key.offset);
+					BUG_ON(ret);
+				}
+			}
+		}
+
+		btrfs_mark_buffer_dirty(dst_path->nodes[0]);
+		btrfs_release_path(root, path);
+		btrfs_release_path(log, dst_path);
+
+		if (min_key.offset < (u64)-1)
+			min_key.offset++;
+		else if (min_key.type < (u8)-1)
+			min_key.type++;
+		else if (min_key.objectid < (u64)-1)
+			min_key.objectid++;
+		else
+			break;
+	}
+	if (inode_only == LOG_INODE_ALL && S_ISDIR(inode->i_mode)) {
+		btrfs_release_path(root, path);
+		btrfs_release_path(log, dst_path);
+		ret = log_directory_changes(trans, root, inode, path, dst_path);
+		BUG_ON(ret);
+	}
+	mutex_unlock(&BTRFS_I(inode)->log_mutex);
+
+	btrfs_free_path(path);
+	btrfs_free_path(dst_path);
+
+	mutex_lock(&root->fs_info->tree_log_mutex);
+	ret = update_log_root(trans, log);
+	BUG_ON(ret);
+	mutex_unlock(&root->fs_info->tree_log_mutex);
+out:
+	return 0;
+}
+
+int btrfs_log_inode(struct btrfs_trans_handle *trans,
+		    struct btrfs_root *root, struct inode *inode,
+		    int inode_only)
+{
+	int ret;
+
+	start_log_trans(trans, root);
+	ret = __btrfs_log_inode(trans, root, inode, inode_only);
+	end_log_trans(root);
+	return ret;
+}
+
+/*
+ * helper function around btrfs_log_inode to make sure newly created
+ * parent directories also end up in the log.  A minimal inode and backref
+ * only logging is done of any parent directories that are older than
+ * the last committed transaction
+ */
+int btrfs_log_dentry(struct btrfs_trans_handle *trans,
+		    struct btrfs_root *root, struct dentry *dentry)
+{
+	int inode_only = LOG_INODE_ALL;
+	struct super_block *sb;
+	int ret;
+
+	start_log_trans(trans, root);
+	sb = dentry->d_inode->i_sb;
+	while(1) {
+		ret = __btrfs_log_inode(trans, root, dentry->d_inode,
+					inode_only);
+		BUG_ON(ret);
+		inode_only = LOG_INODE_EXISTS;
+
+		dentry = dentry->d_parent;
+		if (!dentry || !dentry->d_inode || sb != dentry->d_inode->i_sb)
+			break;
+
+		if (BTRFS_I(dentry->d_inode)->generation <=
+		    root->fs_info->last_trans_committed)
+			break;
+	}
+	end_log_trans(root);
+	return 0;
+}
+
+/*
+ * it is not safe to log dentry if the chunk root has added new
+ * chunks.  This returns 0 if the dentry was logged, and 1 otherwise.
+ * If this returns 1, you must commit the transaction to safely get your
+ * data on disk.
+ */
+int btrfs_log_dentry_safe(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root, struct dentry *dentry)
+{
+	u64 gen;
+	gen = root->fs_info->last_trans_new_blockgroup;
+	if (gen > root->fs_info->last_trans_committed)
+		return 1;
+	else
+		return btrfs_log_dentry(trans, root, dentry);
+}
+
+/*
+ * should be called during mount to recover any replay any log trees
+ * from the FS
+ */
+int btrfs_recover_log_trees(struct btrfs_root *log_root_tree)
+{
+	int ret;
+	struct btrfs_path *path;
+	struct btrfs_trans_handle *trans;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	struct btrfs_key tmp_key;
+	struct btrfs_root *log;
+	struct btrfs_fs_info *fs_info = log_root_tree->fs_info;
+	struct walk_control wc = {
+		.process_func = process_one_buffer,
+		.stage = 0,
+	};
+
+	fs_info->log_root_recovering = 1;
+	path = btrfs_alloc_path();
+	BUG_ON(!path);
+
+	trans = btrfs_start_transaction(fs_info->tree_root, 1);
+
+	wc.trans = trans;
+	wc.pin = 1;
+
+	walk_log_tree(trans, log_root_tree, &wc);
+
+again:
+	key.objectid = BTRFS_TREE_LOG_OBJECTID;
+	key.offset = (u64)-1;
+	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
+
+	while(1) {
+		ret = btrfs_search_slot(NULL, log_root_tree, &key, path, 0, 0);
+		if (ret < 0)
+			break;
+		if (ret > 0) {
+			if (path->slots[0] == 0)
+				break;
+			path->slots[0]--;
+		}
+		btrfs_item_key_to_cpu(path->nodes[0], &found_key,
+				      path->slots[0]);
+		btrfs_release_path(log_root_tree, path);
+		if (found_key.objectid != BTRFS_TREE_LOG_OBJECTID)
+			break;
+
+		log = btrfs_read_fs_root_no_radix(log_root_tree,
+						  &found_key);
+		BUG_ON(!log);
+
+
+		tmp_key.objectid = found_key.offset;
+		tmp_key.type = BTRFS_ROOT_ITEM_KEY;
+		tmp_key.offset = (u64)-1;
+
+		wc.replay_dest = btrfs_read_fs_root_no_name(fs_info, &tmp_key);
+
+		BUG_ON(!wc.replay_dest);
+
+		btrfs_record_root_in_trans(wc.replay_dest);
+		ret = walk_log_tree(trans, log, &wc);
+		BUG_ON(ret);
+
+		if (wc.stage == LOG_WALK_REPLAY_ALL) {
+			ret = fixup_inode_link_counts(trans, wc.replay_dest,
+						      path);
+			BUG_ON(ret);
+		}
+
+		key.offset = found_key.offset - 1;
+		free_extent_buffer(log->node);
+		kfree(log);
+
+		if (found_key.offset == 0)
+			break;
+	}
+	btrfs_release_path(log_root_tree, path);
+
+	/* step one is to pin it all, step two is to replay just inodes */
+	if (wc.pin) {
+		wc.pin = 0;
+		wc.process_func = replay_one_buffer;
+		wc.stage = LOG_WALK_REPLAY_INODES;
+		goto again;
+	}
+	/* step three is to replay everything */
+	if (wc.stage < LOG_WALK_REPLAY_ALL) {
+		wc.stage++;
+		goto again;
+	}
+
+	btrfs_free_path(path);
+
+	free_extent_buffer(log_root_tree->node);
+	log_root_tree->log_root = NULL;
+	fs_info->log_root_recovering = 0;
+
+	/* step 4: commit the transaction, which also unpins the blocks */
+	btrfs_commit_transaction(trans, fs_info->tree_root);
+
+	kfree(log_root_tree);
+	return 0;
+}
diff --git a/fs/btrfs/tree-log.h b/fs/btrfs/tree-log.h
new file mode 100644
index 000000000000..b9409b32ed02
--- /dev/null
+++ b/fs/btrfs/tree-log.h
@@ -0,0 +1,41 @@
+/*
+ * Copyright (C) 2008 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef __TREE_LOG_
+#define __TREE_LOG_
+
+int btrfs_sync_log(struct btrfs_trans_handle *trans,
+		   struct btrfs_root *root);
+int btrfs_free_log(struct btrfs_trans_handle *trans, struct btrfs_root *root);
+int btrfs_log_dentry(struct btrfs_trans_handle *trans,
+		    struct btrfs_root *root, struct dentry *dentry);
+int btrfs_recover_log_trees(struct btrfs_root *tree_root);
+int btrfs_log_dentry_safe(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root, struct dentry *dentry);
+int btrfs_log_inode(struct btrfs_trans_handle *trans,
+		    struct btrfs_root *root, struct inode *inode,
+		    int inode_only);
+int btrfs_del_dir_entries_in_log(struct btrfs_trans_handle *trans,
+				 struct btrfs_root *root,
+				 const char *name, int name_len,
+				 struct inode *dir, u64 index);
+int btrfs_del_inode_ref_in_log(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root,
+			       const char *name, int name_len,
+			       struct inode *inode, u64 dirid);
+#endif
-- 
cgit v1.2.3


From 4bef084857ab8fe71cf49eae349c25e440a49150 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Mon, 8 Sep 2008 11:18:08 -0400
Subject: Btrfs: Tree logging fixes

* Pin down data blocks to prevent them from being reallocated like so:

trans 1: allocate file extent
trans 2: free file extent
trans 3: free file extent during old snapshot deletion
trans 3: allocate file extent to new file
trans 3: fsync new file

Before the tree logging code, this was legal because the fsync
would commit the transation that did the final data extent free
and the transaction that allocated the extent to the new file
at the same time.

With the tree logging code, the tree log subtransaction can commit
before the transaction that freed the extent.  If we crash,
we're left with two different files using the extent.

* Don't wait in start_transaction if log replay is going on.  This
avoids deadlocks from iput while we're cleaning up link counts in the
replay code.

* Don't deadlock in replay_one_name by trying to read an inode off
the disk while holding paths for the directory

* Hold the buffer lock while we mark a buffer as written.  This
closes a race where someone is changing a buffer while we write it.
They are supposed to mark it dirty again after they change it, but
this violates the cow rules.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/disk-io.c     | 33 ++++++++++++++++++++++++++++++---
 fs/btrfs/disk-io.h     |  1 +
 fs/btrfs/extent-tree.c | 20 +++++++++++++++++---
 fs/btrfs/extent_io.c   | 47 ++++++++++++++++++++++++++++++++++++-----------
 fs/btrfs/extent_io.h   |  1 +
 fs/btrfs/transaction.c | 16 +++++++++++++---
 fs/btrfs/tree-log.c    | 13 +++++++++----
 7 files changed, 107 insertions(+), 24 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index a4373db5967a..42bf99168056 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -307,9 +307,7 @@ int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
 		goto err;
 	}
 	found_level = btrfs_header_level(eb);
-	spin_lock(&root->fs_info->hash_lock);
-	btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
-	spin_unlock(&root->fs_info->hash_lock);
+
 	csum_tree_block(root, eb, 0);
 err:
 	free_extent_buffer(eb);
@@ -1998,7 +1996,36 @@ int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
 	return ret;
 }
 
+int btree_lock_page_hook(struct page *page)
+{
+	struct inode *inode = page->mapping->host;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
+	struct extent_buffer *eb;
+	unsigned long len;
+	u64 bytenr = page_offset(page);
+
+	if (page->private == EXTENT_PAGE_PRIVATE)
+		goto out;
+
+	len = page->private >> 2;
+	eb = find_extent_buffer(io_tree, bytenr, len, GFP_NOFS);
+	if (!eb)
+		goto out;
+
+	btrfs_tree_lock(eb);
+	spin_lock(&root->fs_info->hash_lock);
+	btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
+	spin_unlock(&root->fs_info->hash_lock);
+	btrfs_tree_unlock(eb);
+	free_extent_buffer(eb);
+out:
+	lock_page(page);
+	return 0;
+}
+
 static struct extent_io_ops btree_extent_io_ops = {
+	.write_cache_pages_lock_hook = btree_lock_page_hook,
 	.writepage_io_hook = btree_writepage_io_hook,
 	.readpage_end_io_hook = btree_readpage_end_io_hook,
 	.submit_bio_hook = btree_submit_bio_hook,
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 6b6fdc697f31..f84f5058dbbb 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -80,4 +80,5 @@ int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans,
 			     struct btrfs_fs_info *fs_info);
 int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
 			     struct btrfs_fs_info *fs_info);
+int btree_lock_page_hook(struct page *page);
 #endif
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 646b9148ca21..3181759da1cf 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1590,13 +1590,17 @@ static int finish_current_insert(struct btrfs_trans_handle *trans,
 }
 
 static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes,
-			  int pending)
+			  int is_data, int pending)
 {
 	int err = 0;
 
 	WARN_ON(!mutex_is_locked(&root->fs_info->alloc_mutex));
 	if (!pending) {
 		struct extent_buffer *buf;
+
+		if (is_data)
+			goto pinit;
+
 		buf = btrfs_find_tree_block(root, bytenr, num_bytes);
 		if (buf) {
 			/* we can reuse a block if it hasn't been written
@@ -1624,6 +1628,7 @@ static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes,
 			}
 			free_extent_buffer(buf);
 		}
+pinit:
 		btrfs_update_pinned_extents(root, bytenr, num_bytes, 1);
 	} else {
 		set_extent_bits(&root->fs_info->pending_del,
@@ -1744,7 +1749,8 @@ static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 #endif
 
 		if (pin) {
-			ret = pin_down_bytes(root, bytenr, num_bytes, 0);
+			ret = pin_down_bytes(root, bytenr, num_bytes,
+			     owner_objectid >= BTRFS_FIRST_FREE_OBJECTID, 0);
 			if (ret > 0)
 				mark_free = 1;
 			BUG_ON(ret < 0);
@@ -1862,9 +1868,17 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
 		ref_generation = 0;
 
 	if (root == extent_root) {
-		pin_down_bytes(root, bytenr, num_bytes, 1);
+		pin_down_bytes(root, bytenr, num_bytes, 0, 1);
 		return 0;
 	}
+	/* if metadata always pin */
+	if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
+		pin = 1;
+
+	/* if data pin when any transaction has committed this */
+	if (ref_generation != trans->transid)
+		pin = 1;
+
 	ret = __free_extent(trans, root, bytenr, num_bytes, root_objectid,
 			    ref_generation, owner_objectid, owner_offset,
 			    pin, pin == 0);
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 239e7c908abf..319a0c7a4a58 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -29,7 +29,10 @@ static struct kmem_cache *extent_buffer_cache;
 
 static LIST_HEAD(buffers);
 static LIST_HEAD(states);
+
+#ifdef LEAK_DEBUG
 static spinlock_t leak_lock = SPIN_LOCK_UNLOCKED;
+#endif
 
 #define BUFFER_LRU_MAX 64
 
@@ -106,7 +109,9 @@ EXPORT_SYMBOL(extent_io_tree_init);
 struct extent_state *alloc_extent_state(gfp_t mask)
 {
 	struct extent_state *state;
+#ifdef LEAK_DEBUG
 	unsigned long flags;
+#endif
 
 	state = kmem_cache_alloc(extent_state_cache, mask);
 	if (!state)
@@ -114,10 +119,11 @@ struct extent_state *alloc_extent_state(gfp_t mask)
 	state->state = 0;
 	state->private = 0;
 	state->tree = NULL;
+#ifdef LEAK_DEBUG
 	spin_lock_irqsave(&leak_lock, flags);
 	list_add(&state->leak_list, &states);
 	spin_unlock_irqrestore(&leak_lock, flags);
-
+#endif
 	atomic_set(&state->refs, 1);
 	init_waitqueue_head(&state->wq);
 	return state;
@@ -129,11 +135,15 @@ void free_extent_state(struct extent_state *state)
 	if (!state)
 		return;
 	if (atomic_dec_and_test(&state->refs)) {
+#ifdef LEAK_DEBUG
 		unsigned long flags;
+#endif
 		WARN_ON(state->tree);
+#ifdef LEAK_DEBUG
 		spin_lock_irqsave(&leak_lock, flags);
 		list_del(&state->leak_list);
 		spin_unlock_irqrestore(&leak_lock, flags);
+#endif
 		kmem_cache_free(extent_state_cache, state);
 	}
 }
@@ -2070,13 +2080,13 @@ done:
 }
 
 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,22)
-/* Taken directly from 2.6.23 for 2.6.18 back port */
+/* Taken directly from 2.6.23 with a mod for a lockpage hook */
 typedef int (*writepage_t)(struct page *page, struct writeback_control *wbc,
                                 void *data);
+#endif
 
 /**
- * write_cache_pages - walk the list of dirty pages of the given address space
- * and write all of them.
+ * write_cache_pages - walk the list of dirty pages of the given address space and write all of them.
  * @mapping: address space structure to write
  * @wbc: subtract the number of written pages from *@wbc->nr_to_write
  * @writepage: function called for each page
@@ -2090,9 +2100,10 @@ typedef int (*writepage_t)(struct page *page, struct writeback_control *wbc,
  * WB_SYNC_ALL then we were called for data integrity and we must wait for
  * existing IO to complete.
  */
-static int write_cache_pages(struct address_space *mapping,
-		      struct writeback_control *wbc, writepage_t writepage,
-		      void *data)
+int extent_write_cache_pages(struct extent_io_tree *tree,
+			     struct address_space *mapping,
+			     struct writeback_control *wbc,
+			     writepage_t writepage, void *data)
 {
 	struct backing_dev_info *bdi = mapping->backing_dev_info;
 	int ret = 0;
@@ -2138,7 +2149,10 @@ retry:
 			 * swizzled back from swapper_space to tmpfs file
 			 * mapping
 			 */
-			lock_page(page);
+			if (tree->ops && tree->ops->write_cache_pages_lock_hook)
+				tree->ops->write_cache_pages_lock_hook(page);
+			else
+				lock_page(page);
 
 			if (unlikely(page->mapping != mapping)) {
 				unlock_page(page);
@@ -2187,9 +2201,12 @@ retry:
 	}
 	if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
 		mapping->writeback_index = index;
+
+	if (wbc->range_cont)
+		wbc->range_start = index << PAGE_CACHE_SHIFT;
 	return ret;
 }
-#endif
+EXPORT_SYMBOL(extent_write_cache_pages);
 
 int extent_write_full_page(struct extent_io_tree *tree, struct page *page,
 			  get_extent_t *get_extent,
@@ -2214,7 +2231,8 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page,
 
 	ret = __extent_writepage(page, wbc, &epd);
 
-	write_cache_pages(mapping, &wbc_writepages, __extent_writepage, &epd);
+	extent_write_cache_pages(tree, mapping, &wbc_writepages,
+				 __extent_writepage, &epd);
 	if (epd.bio) {
 		submit_one_bio(WRITE, epd.bio, 0);
 	}
@@ -2235,7 +2253,8 @@ int extent_writepages(struct extent_io_tree *tree,
 		.get_extent = get_extent,
 	};
 
-	ret = write_cache_pages(mapping, wbc, __extent_writepage, &epd);
+	ret = extent_write_cache_pages(tree, mapping, wbc,
+				       __extent_writepage, &epd);
 	if (epd.bio) {
 		submit_one_bio(WRITE, epd.bio, 0);
 	}
@@ -2567,15 +2586,19 @@ static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree,
 						   gfp_t mask)
 {
 	struct extent_buffer *eb = NULL;
+#ifdef LEAK_DEBUG
 	unsigned long flags;
+#endif
 
 	eb = kmem_cache_zalloc(extent_buffer_cache, mask);
 	eb->start = start;
 	eb->len = len;
 	mutex_init(&eb->mutex);
+#ifdef LEAK_DEBUG
 	spin_lock_irqsave(&leak_lock, flags);
 	list_add(&eb->leak_list, &buffers);
 	spin_unlock_irqrestore(&leak_lock, flags);
+#endif
 	atomic_set(&eb->refs, 1);
 
 	return eb;
@@ -2583,10 +2606,12 @@ static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree,
 
 static void __free_extent_buffer(struct extent_buffer *eb)
 {
+#ifdef LEAK_DEBUG
 	unsigned long flags;
 	spin_lock_irqsave(&leak_lock, flags);
 	list_del(&eb->leak_list);
 	spin_unlock_irqrestore(&leak_lock, flags);
+#endif
 	kmem_cache_free(extent_buffer_cache, eb);
 }
 
diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h
index 315cfceae312..3cb411a5f4d3 100644
--- a/fs/btrfs/extent_io.h
+++ b/fs/btrfs/extent_io.h
@@ -50,6 +50,7 @@ struct extent_io_ops {
 			    unsigned long old, unsigned long bits);
 	int (*clear_bit_hook)(struct inode *inode, u64 start, u64 end,
 			    unsigned long old, unsigned long bits);
+	int (*write_cache_pages_lock_hook)(struct page *page);
 };
 
 struct extent_io_tree {
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 49c4f5b40ed6..61a377bcb2fb 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -161,7 +161,8 @@ static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root,
 	int ret;
 
 	mutex_lock(&root->fs_info->trans_mutex);
-	if ((wait == 1 && !root->fs_info->open_ioctl_trans) || wait == 2)
+	if (!root->fs_info->log_root_recovering &&
+	    ((wait == 1 && !root->fs_info->open_ioctl_trans) || wait == 2))
 		wait_current_trans(root);
 	ret = join_transaction(root);
 	BUG_ON(ret);
@@ -328,9 +329,17 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
 
 			index = start >> PAGE_CACHE_SHIFT;
 			start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
-			page = find_lock_page(btree_inode->i_mapping, index);
+			page = find_get_page(btree_inode->i_mapping, index);
 			if (!page)
 				continue;
+
+			btree_lock_page_hook(page);
+			if (!page->mapping) {
+				unlock_page(page);
+				page_cache_release(page);
+				continue;
+			}
+
 			if (PageWriteback(page)) {
 				if (PageDirty(page))
 					wait_on_page_writeback(page);
@@ -360,7 +369,8 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
 			if (!page)
 				continue;
 			if (PageDirty(page)) {
-				lock_page(page);
+				btree_lock_page_hook(page);
+				wait_on_page_writeback(page);
 				err = write_one_page(page, 0);
 				if (err)
 					werr = err;
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index d1ce8314b948..13d7ee8e0c52 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -1176,8 +1176,8 @@ static noinline int replay_one_name(struct btrfs_trans_handle *trans,
 	struct btrfs_key found_key;
 	struct btrfs_key log_key;
 	struct inode *dir;
-	struct inode *inode;
 	u8 log_type;
+	int exists;
 	int ret;
 
 	dir = read_one_inode(root, key->objectid);
@@ -1190,6 +1190,13 @@ static noinline int replay_one_name(struct btrfs_trans_handle *trans,
 		   name_len);
 
 	btrfs_dir_item_key_to_cpu(eb, di, &log_key);
+	exists = btrfs_lookup_inode(trans, root, path, &log_key, 0);
+	if (exists == 0)
+		exists = 1;
+	else
+		exists = 0;
+	btrfs_release_path(root, path);
+
 	if (key->type == BTRFS_DIR_ITEM_KEY) {
 		dst_di = btrfs_lookup_dir_item(trans, root, path, key->objectid,
 				       name, name_len, 1);
@@ -1224,11 +1231,9 @@ static noinline int replay_one_name(struct btrfs_trans_handle *trans,
 	 * don't drop the conflicting directory entry if the inode
 	 * for the new entry doesn't exist
 	 */
-	inode = read_one_inode(root, log_key.objectid);
-	if (!inode)
+	if (!exists)
 		goto out;
 
-	iput(inode);
 	ret = drop_one_dir_item(trans, root, path, dir, dst_di);
 	BUG_ON(ret);
 
-- 
cgit v1.2.3


From c8b978188c9a0fd3d535c13debd19d522b726f1f Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Wed, 29 Oct 2008 14:49:59 -0400
Subject: Btrfs: Add zlib compression support

This is a large change for adding compression on reading and writing,
both for inline and regular extents.  It does some fairly large
surgery to the writeback paths.

Compression is off by default and enabled by mount -o compress.  Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.

If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.

* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler.  This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.

* Inline extents are inserted at delalloc time now.  This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.

* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.

From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field.  Neither the encryption or the
'other' field are currently used.

In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k.  This is a
software only limit, the disk format supports u64 sized compressed extents.

In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k.  This is a software only limit
and will be subject to tuning later.

Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data.  This way additional encodings can be
layered on without having to figure out which encoding to checksum.

Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread.  This makes it tricky to
spread the compression load across all the cpus on the box.  We'll have to
look at parallel pdflush walks of dirty inodes at a later time.

Decompression is hooked into readpages and it does spread across CPUs nicely.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/Kconfig              |   2 +
 fs/btrfs/Makefile       |   3 +-
 fs/btrfs/compression.c  | 454 ++++++++++++++++++++++++++++++++++
 fs/btrfs/compression.h  |  47 ++++
 fs/btrfs/ctree.h        |  99 ++++++--
 fs/btrfs/disk-io.c      |  18 +-
 fs/btrfs/disk-io.h      |   1 +
 fs/btrfs/extent-tree.c  |  27 +-
 fs/btrfs/extent_io.c    | 411 ++++++++++++++++++++++++-------
 fs/btrfs/extent_io.h    |  17 +-
 fs/btrfs/extent_map.c   |   9 +
 fs/btrfs/extent_map.h   |   6 +-
 fs/btrfs/file-item.c    |  75 +++++-
 fs/btrfs/file.c         | 263 +++++---------------
 fs/btrfs/inode.c        | 584 ++++++++++++++++++++++++++++++++++++++++----
 fs/btrfs/ordered-data.c |   9 +-
 fs/btrfs/ordered-data.h |  10 +-
 fs/btrfs/print-tree.c   |   7 +-
 fs/btrfs/super.c        |  10 +-
 fs/btrfs/tree-log.c     |   3 +-
 fs/btrfs/volumes.c      |   2 +
 fs/btrfs/zlib.c         | 637 ++++++++++++++++++++++++++++++++++++++++++++++++
 22 files changed, 2315 insertions(+), 379 deletions(-)
 create mode 100644 fs/btrfs/compression.c
 create mode 100644 fs/btrfs/compression.h
 create mode 100644 fs/btrfs/zlib.c

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/Kconfig b/fs/Kconfig
index 18f5a85b47c6..31cce5d88b1a 100644
--- a/fs/Kconfig
+++ b/fs/Kconfig
@@ -501,6 +501,8 @@ config BTRFS_FS
 	tristate "Btrfs filesystem (EXPERIMENTAL) Unstable disk format"
 	depends on EXPERIMENTAL
 	select LIBCRC32C
+	select ZLIB_INFLATE
+	select ZLIB_DEFLATE
 	help
 	  Btrfs is a new filesystem with extents, writable snapshotting,
 	  support for multiple devices and many more features.
diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 7125716e142b..d2cf5a54a4b8 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -7,7 +7,8 @@ btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
 	   transaction.o inode.o file.o tree-defrag.o \
 	   extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \
 	   extent_io.o volumes.o async-thread.o ioctl.o locking.o orphan.o \
-	   ref-cache.o export.o tree-log.o acl.o free-space-cache.o
+	   ref-cache.o export.o tree-log.o acl.o free-space-cache.o zlib.o \
+	   compression.o
 else
 
 # Normal Makefile
diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c
new file mode 100644
index 000000000000..c5470367ca5c
--- /dev/null
+++ b/fs/btrfs/compression.c
@@ -0,0 +1,454 @@
+/*
+ * Copyright (C) 2008 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/kernel.h>
+#include <linux/bio.h>
+#include <linux/buffer_head.h>
+#include <linux/file.h>
+#include <linux/fs.h>
+#include <linux/pagemap.h>
+#include <linux/highmem.h>
+#include <linux/time.h>
+#include <linux/init.h>
+#include <linux/string.h>
+#include <linux/smp_lock.h>
+#include <linux/backing-dev.h>
+#include <linux/mpage.h>
+#include <linux/swap.h>
+#include <linux/writeback.h>
+#include <linux/bit_spinlock.h>
+#include <linux/version.h>
+#include "ctree.h"
+#include "disk-io.h"
+#include "transaction.h"
+#include "btrfs_inode.h"
+#include "volumes.h"
+#include "ordered-data.h"
+#include "compat.h"
+#include "compression.h"
+#include "extent_io.h"
+#include "extent_map.h"
+
+struct compressed_bio {
+	/* number of bios pending for this compressed extent */
+	atomic_t pending_bios;
+
+	/* the pages with the compressed data on them */
+	struct page **compressed_pages;
+
+	/* inode that owns this data */
+	struct inode *inode;
+
+	/* starting offset in the inode for our pages */
+	u64 start;
+
+	/* number of bytes in the inode we're working on */
+	unsigned long len;
+
+	/* number of bytes on disk */
+	unsigned long compressed_len;
+
+	/* number of compressed pages in the array */
+	unsigned long nr_pages;
+
+	/* IO errors */
+	int errors;
+
+	/* for reads, this is the bio we are copying the data into */
+	struct bio *orig_bio;
+};
+
+static struct bio *compressed_bio_alloc(struct block_device *bdev,
+					u64 first_byte, gfp_t gfp_flags)
+{
+	struct bio *bio;
+	int nr_vecs;
+
+	nr_vecs = bio_get_nr_vecs(bdev);
+	bio = bio_alloc(gfp_flags, nr_vecs);
+
+	if (bio == NULL && (current->flags & PF_MEMALLOC)) {
+		while (!bio && (nr_vecs /= 2))
+			bio = bio_alloc(gfp_flags, nr_vecs);
+	}
+
+	if (bio) {
+		bio->bi_size = 0;
+		bio->bi_bdev = bdev;
+		bio->bi_sector = first_byte >> 9;
+	}
+	return bio;
+}
+
+/* when we finish reading compressed pages from the disk, we
+ * decompress them and then run the bio end_io routines on the
+ * decompressed pages (in the inode address space).
+ *
+ * This allows the checksumming and other IO error handling routines
+ * to work normally
+ *
+ * The compressed pages are freed here, and it must be run
+ * in process context
+ */
+static void end_compressed_bio_read(struct bio *bio, int err)
+{
+	struct extent_io_tree *tree;
+	struct compressed_bio *cb = bio->bi_private;
+	struct inode *inode;
+	struct page *page;
+	unsigned long index;
+	int ret;
+
+	if (err)
+		cb->errors = 1;
+
+	/* if there are more bios still pending for this compressed
+	 * extent, just exit
+	 */
+	if (!atomic_dec_and_test(&cb->pending_bios))
+		goto out;
+
+	/* ok, we're the last bio for this extent, lets start
+	 * the decompression.
+	 */
+	inode = cb->inode;
+	tree = &BTRFS_I(inode)->io_tree;
+	ret = btrfs_zlib_decompress_biovec(cb->compressed_pages,
+					cb->start,
+					cb->orig_bio->bi_io_vec,
+					cb->orig_bio->bi_vcnt,
+					cb->compressed_len);
+	if (ret)
+		cb->errors = 1;
+
+	/* release the compressed pages */
+	index = 0;
+	for (index = 0; index < cb->nr_pages; index++) {
+		page = cb->compressed_pages[index];
+		page->mapping = NULL;
+		page_cache_release(page);
+	}
+
+	/* do io completion on the original bio */
+	if (cb->errors)
+		bio_io_error(cb->orig_bio);
+	else
+		bio_endio(cb->orig_bio, 0);
+
+	/* finally free the cb struct */
+	kfree(cb->compressed_pages);
+	kfree(cb);
+out:
+	bio_put(bio);
+}
+
+/*
+ * Clear the writeback bits on all of the file
+ * pages for a compressed write
+ */
+static noinline int end_compressed_writeback(struct inode *inode, u64 start,
+					     unsigned long ram_size)
+{
+	unsigned long index = start >> PAGE_CACHE_SHIFT;
+	unsigned long end_index = (start + ram_size - 1) >> PAGE_CACHE_SHIFT;
+	struct page *pages[16];
+	unsigned long nr_pages = end_index - index + 1;
+	int i;
+	int ret;
+
+	while(nr_pages > 0) {
+		ret = find_get_pages_contig(inode->i_mapping, index,
+				     min(nr_pages, ARRAY_SIZE(pages)), pages);
+		if (ret == 0) {
+			nr_pages -= 1;
+			index += 1;
+			continue;
+		}
+		for (i = 0; i < ret; i++) {
+			end_page_writeback(pages[i]);
+			page_cache_release(pages[i]);
+		}
+		nr_pages -= ret;
+		index += ret;
+	}
+	/* the inode may be gone now */
+	return 0;
+}
+
+/*
+ * do the cleanup once all the compressed pages hit the disk.
+ * This will clear writeback on the file pages and free the compressed
+ * pages.
+ *
+ * This also calls the writeback end hooks for the file pages so that
+ * metadata and checksums can be updated in the file.
+ */
+static void end_compressed_bio_write(struct bio *bio, int err)
+{
+	struct extent_io_tree *tree;
+	struct compressed_bio *cb = bio->bi_private;
+	struct inode *inode;
+	struct page *page;
+	unsigned long index;
+
+	if (err)
+		cb->errors = 1;
+
+	/* if there are more bios still pending for this compressed
+	 * extent, just exit
+	 */
+	if (!atomic_dec_and_test(&cb->pending_bios))
+		goto out;
+
+	/* ok, we're the last bio for this extent, step one is to
+	 * call back into the FS and do all the end_io operations
+	 */
+	inode = cb->inode;
+	tree = &BTRFS_I(inode)->io_tree;
+	tree->ops->writepage_end_io_hook(cb->compressed_pages[0],
+					 cb->start,
+					 cb->start + cb->len - 1,
+					 NULL, 1);
+
+	end_compressed_writeback(inode, cb->start, cb->len);
+	/* note, our inode could be gone now */
+
+	/*
+	 * release the compressed pages, these came from alloc_page and
+	 * are not attached to the inode at all
+	 */
+	index = 0;
+	for (index = 0; index < cb->nr_pages; index++) {
+		page = cb->compressed_pages[index];
+		page->mapping = NULL;
+		page_cache_release(page);
+	}
+
+	/* finally free the cb struct */
+	kfree(cb->compressed_pages);
+	kfree(cb);
+out:
+	bio_put(bio);
+}
+
+/*
+ * worker function to build and submit bios for previously compressed pages.
+ * The corresponding pages in the inode should be marked for writeback
+ * and the compressed pages should have a reference on them for dropping
+ * when the IO is complete.
+ *
+ * This also checksums the file bytes and gets things ready for
+ * the end io hooks.
+ */
+int btrfs_submit_compressed_write(struct inode *inode, u64 start,
+				 unsigned long len, u64 disk_start,
+				 unsigned long compressed_len,
+				 struct page **compressed_pages,
+				 unsigned long nr_pages)
+{
+	struct bio *bio = NULL;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct compressed_bio *cb;
+	unsigned long bytes_left;
+	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
+	int page_index = 0;
+	struct page *page;
+	u64 first_byte = disk_start;
+	struct block_device *bdev;
+	int ret;
+
+	WARN_ON(start & ((u64)PAGE_CACHE_SIZE - 1));
+	cb = kmalloc(sizeof(*cb), GFP_NOFS);
+	atomic_set(&cb->pending_bios, 0);
+	cb->errors = 0;
+	cb->inode = inode;
+	cb->start = start;
+	cb->len = len;
+	cb->compressed_pages = compressed_pages;
+	cb->compressed_len = compressed_len;
+	cb->orig_bio = NULL;
+	cb->nr_pages = nr_pages;
+
+	bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
+
+	ret = btrfs_csum_file_bytes(root, inode, start, len);
+	BUG_ON(ret);
+
+	bio = compressed_bio_alloc(bdev, first_byte, GFP_NOFS);
+	bio->bi_private = cb;
+	bio->bi_end_io = end_compressed_bio_write;
+	atomic_inc(&cb->pending_bios);
+
+	/* create and submit bios for the compressed pages */
+	bytes_left = compressed_len;
+	while(bytes_left > 0) {
+		page = compressed_pages[page_index];
+		page->mapping = inode->i_mapping;
+		if (bio->bi_size)
+			ret = io_tree->ops->merge_bio_hook(page, 0,
+							   PAGE_CACHE_SIZE,
+							   bio, 0);
+		else
+			ret = 0;
+
+		if (ret || bio_add_page(bio, page, PAGE_CACHE_SIZE, 0) <
+		    PAGE_CACHE_SIZE) {
+			bio_get(bio);
+
+			ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
+			BUG_ON(ret);
+
+			ret = btrfs_map_bio(root, WRITE, bio, 0, 1);
+			BUG_ON(ret);
+
+			bio_put(bio);
+
+			bio = compressed_bio_alloc(bdev, first_byte, GFP_NOFS);
+			atomic_inc(&cb->pending_bios);
+			bio->bi_private = cb;
+			bio->bi_end_io = end_compressed_bio_write;
+			bio_add_page(bio, page, PAGE_CACHE_SIZE, 0);
+		}
+		page_index++;
+		bytes_left -= PAGE_CACHE_SIZE;
+		first_byte += PAGE_CACHE_SIZE;
+	}
+	bio_get(bio);
+
+	ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
+	BUG_ON(ret);
+
+	ret = btrfs_map_bio(root, WRITE, bio, 0, 1);
+	BUG_ON(ret);
+
+	bio_put(bio);
+	return 0;
+}
+
+/*
+ * for a compressed read, the bio we get passed has all the inode pages
+ * in it.  We don't actually do IO on those pages but allocate new ones
+ * to hold the compressed pages on disk.
+ *
+ * bio->bi_sector points to the compressed extent on disk
+ * bio->bi_io_vec points to all of the inode pages
+ * bio->bi_vcnt is a count of pages
+ *
+ * After the compressed pages are read, we copy the bytes into the
+ * bio we were passed and then call the bio end_io calls
+ */
+int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
+				 int mirror_num, unsigned long bio_flags)
+{
+	struct extent_io_tree *tree;
+	struct extent_map_tree *em_tree;
+	struct compressed_bio *cb;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	unsigned long uncompressed_len = bio->bi_vcnt * PAGE_CACHE_SIZE;
+	unsigned long compressed_len;
+	unsigned long nr_pages;
+	unsigned long page_index;
+	struct page *page;
+	struct block_device *bdev;
+	struct bio *comp_bio;
+	u64 cur_disk_byte = (u64)bio->bi_sector << 9;
+	struct extent_map *em;
+	int ret;
+
+	tree = &BTRFS_I(inode)->io_tree;
+	em_tree = &BTRFS_I(inode)->extent_tree;
+
+	/* we need the actual starting offset of this extent in the file */
+	spin_lock(&em_tree->lock);
+	em = lookup_extent_mapping(em_tree,
+				   page_offset(bio->bi_io_vec->bv_page),
+				   PAGE_CACHE_SIZE);
+	spin_unlock(&em_tree->lock);
+
+	cb = kmalloc(sizeof(*cb), GFP_NOFS);
+	atomic_set(&cb->pending_bios, 0);
+	cb->errors = 0;
+	cb->inode = inode;
+
+	cb->start = em->start;
+	compressed_len = em->block_len;
+	free_extent_map(em);
+
+	cb->len = uncompressed_len;
+	cb->compressed_len = compressed_len;
+	cb->orig_bio = bio;
+
+	nr_pages = (compressed_len + PAGE_CACHE_SIZE - 1) /
+				 PAGE_CACHE_SIZE;
+	cb->compressed_pages = kmalloc(sizeof(struct page *) * nr_pages,
+				       GFP_NOFS);
+	bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
+
+	for (page_index = 0; page_index < nr_pages; page_index++) {
+		cb->compressed_pages[page_index] = alloc_page(GFP_NOFS |
+							      __GFP_HIGHMEM);
+	}
+	cb->nr_pages = nr_pages;
+
+	comp_bio = compressed_bio_alloc(bdev, cur_disk_byte, GFP_NOFS);
+	comp_bio->bi_private = cb;
+	comp_bio->bi_end_io = end_compressed_bio_read;
+	atomic_inc(&cb->pending_bios);
+
+	for (page_index = 0; page_index < nr_pages; page_index++) {
+		page = cb->compressed_pages[page_index];
+		page->mapping = inode->i_mapping;
+		if (comp_bio->bi_size)
+			ret = tree->ops->merge_bio_hook(page, 0,
+							PAGE_CACHE_SIZE,
+							comp_bio, 0);
+		else
+			ret = 0;
+
+		if (ret || bio_add_page(comp_bio, page, PAGE_CACHE_SIZE, 0) <
+		    PAGE_CACHE_SIZE) {
+			bio_get(comp_bio);
+
+			ret = btrfs_bio_wq_end_io(root->fs_info, comp_bio, 0);
+			BUG_ON(ret);
+
+			ret = btrfs_map_bio(root, READ, comp_bio, 0, 0);
+			BUG_ON(ret);
+
+			bio_put(comp_bio);
+
+			comp_bio = compressed_bio_alloc(bdev, cur_disk_byte,
+							GFP_NOFS);
+			atomic_inc(&cb->pending_bios);
+			bio->bi_private = cb;
+			bio->bi_end_io = end_compressed_bio_write;
+			bio_add_page(bio, page, PAGE_CACHE_SIZE, 0);
+		}
+		cur_disk_byte += PAGE_CACHE_SIZE;
+	}
+	bio_get(comp_bio);
+
+	ret = btrfs_bio_wq_end_io(root->fs_info, comp_bio, 0);
+	BUG_ON(ret);
+
+	ret = btrfs_map_bio(root, READ, comp_bio, 0, 0);
+	BUG_ON(ret);
+
+	bio_put(comp_bio);
+	return 0;
+}
diff --git a/fs/btrfs/compression.h b/fs/btrfs/compression.h
new file mode 100644
index 000000000000..421f5b4aa715
--- /dev/null
+++ b/fs/btrfs/compression.h
@@ -0,0 +1,47 @@
+/*
+ * Copyright (C) 2008 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef __BTRFS_COMPRESSION_
+#define __BTRFS_COMPRESSION_
+
+int btrfs_zlib_decompress(unsigned char *data_in,
+			  struct page *dest_page,
+			  unsigned long start_byte,
+			  size_t srclen, size_t destlen);
+int btrfs_zlib_compress_pages(struct address_space *mapping,
+			      u64 start, unsigned long len,
+			      struct page **pages,
+			      unsigned long nr_dest_pages,
+			      unsigned long *out_pages,
+			      unsigned long *total_in,
+			      unsigned long *total_out,
+			      unsigned long max_out);
+int btrfs_zlib_decompress_biovec(struct page **pages_in,
+			      u64 disk_start,
+			      struct bio_vec *bvec,
+			      int vcnt,
+			      size_t srclen);
+void btrfs_zlib_exit(void);
+int btrfs_submit_compressed_write(struct inode *inode, u64 start,
+				  unsigned long len, u64 disk_start,
+				  unsigned long compressed_len,
+				  struct page **compressed_pages,
+				  unsigned long nr_pages);
+int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
+				 int mirror_num, unsigned long bio_flags);
+#endif
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 8559f39fd47f..793d8fdda244 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -400,10 +400,18 @@ struct btrfs_timespec {
 	__le32 nsec;
 } __attribute__ ((__packed__));
 
-/*
- * there is no padding here on purpose.  If you want to extent the inode,
- * make a new item type
- */
+typedef enum {
+	BTRFS_COMPRESS_NONE = 0,
+	BTRFS_COMPRESS_ZLIB = 1,
+	BTRFS_COMPRESS_LAST = 2,
+} btrfs_compression_type;
+
+/* we don't understand any encryption methods right now */
+typedef enum {
+	BTRFS_ENCRYPTION_NONE = 0,
+	BTRFS_ENCRYPTION_LAST = 1,
+} btrfs_encryption_type;
+
 struct btrfs_inode_item {
 	/* nfs style generation number */
 	__le64 generation;
@@ -419,6 +427,7 @@ struct btrfs_inode_item {
 	__le64 rdev;
 	__le16 flags;
 	__le16 compat_flags;
+
 	struct btrfs_timespec atime;
 	struct btrfs_timespec ctime;
 	struct btrfs_timespec mtime;
@@ -454,8 +463,33 @@ struct btrfs_root_item {
 #define BTRFS_FILE_EXTENT_INLINE 1
 
 struct btrfs_file_extent_item {
+	/*
+	 * transaction id that created this extent
+	 */
 	__le64 generation;
+	/*
+	 * max number of bytes to hold this extent in ram
+	 * when we split a compressed extent we can't know how big
+	 * each of the resulting pieces will be.  So, this is
+	 * an upper limit on the size of the extent in ram instead of
+	 * an exact limit.
+	 */
+	__le64 ram_bytes;
+
+	/*
+	 * 32 bits for the various ways we might encode the data,
+	 * including compression and encryption.  If any of these
+	 * are set to something a given disk format doesn't understand
+	 * it is treated like an incompat flag for reading and writing,
+	 * but not for stat.
+	 */
+	u8 compression;
+	u8 encryption;
+	__le16 other_encoding; /* spare for later use */
+
+	/* are we inline data or a real extent? */
 	u8 type;
+
 	/*
 	 * disk space consumed by the extent, checksum blocks are included
 	 * in these numbers
@@ -471,9 +505,11 @@ struct btrfs_file_extent_item {
 	 */
 	__le64 offset;
 	/*
-	 * the logical number of file blocks (no csums included)
+	 * the logical number of file blocks (no csums included).  This
+	 * always reflects the size uncompressed and without encoding.
 	 */
 	__le64 num_bytes;
+
 } __attribute__ ((__packed__));
 
 struct btrfs_csum_item {
@@ -814,6 +850,7 @@ struct btrfs_root {
 #define BTRFS_MOUNT_NOBARRIER		(1 << 2)
 #define BTRFS_MOUNT_SSD			(1 << 3)
 #define BTRFS_MOUNT_DEGRADED		(1 << 4)
+#define BTRFS_MOUNT_COMPRESS		(1 << 5)
 
 #define btrfs_clear_opt(o, opt)		((o) &= ~BTRFS_MOUNT_##opt)
 #define btrfs_set_opt(o, opt)		((o) |= BTRFS_MOUNT_##opt)
@@ -825,6 +862,7 @@ struct btrfs_root {
 #define BTRFS_INODE_NODATASUM		(1 << 0)
 #define BTRFS_INODE_NODATACOW		(1 << 1)
 #define BTRFS_INODE_READONLY		(1 << 2)
+#define BTRFS_INODE_NOCOMPRESS		(1 << 3)
 #define btrfs_clear_flag(inode, flag)	(BTRFS_I(inode)->flags &= \
 					 ~BTRFS_INODE_##flag)
 #define btrfs_set_flag(inode, flag)	(BTRFS_I(inode)->flags |= \
@@ -1424,14 +1462,6 @@ static inline u32 btrfs_file_extent_calc_inline_size(u32 datasize)
 	return offsetof(struct btrfs_file_extent_item, disk_bytenr) + datasize;
 }
 
-static inline u32 btrfs_file_extent_inline_len(struct extent_buffer *eb,
-					       struct btrfs_item *e)
-{
-	unsigned long offset;
-	offset = offsetof(struct btrfs_file_extent_item, disk_bytenr);
-	return btrfs_item_size(eb, e) - offset;
-}
-
 BTRFS_SETGET_FUNCS(file_extent_disk_bytenr, struct btrfs_file_extent_item,
 		   disk_bytenr, 64);
 BTRFS_SETGET_FUNCS(file_extent_generation, struct btrfs_file_extent_item,
@@ -1442,6 +1472,36 @@ BTRFS_SETGET_FUNCS(file_extent_offset, struct btrfs_file_extent_item,
 		  offset, 64);
 BTRFS_SETGET_FUNCS(file_extent_num_bytes, struct btrfs_file_extent_item,
 		   num_bytes, 64);
+BTRFS_SETGET_FUNCS(file_extent_ram_bytes, struct btrfs_file_extent_item,
+		   ram_bytes, 64);
+BTRFS_SETGET_FUNCS(file_extent_compression, struct btrfs_file_extent_item,
+		   compression, 8);
+BTRFS_SETGET_FUNCS(file_extent_encryption, struct btrfs_file_extent_item,
+		   encryption, 8);
+BTRFS_SETGET_FUNCS(file_extent_other_encoding, struct btrfs_file_extent_item,
+		   other_encoding, 16);
+
+/* this returns the number of file bytes represented by the inline item.
+ * If an item is compressed, this is the uncompressed size
+ */
+static inline u32 btrfs_file_extent_inline_len(struct extent_buffer *eb,
+					       struct btrfs_file_extent_item *e)
+{
+	return btrfs_file_extent_ram_bytes(eb, e);
+}
+
+/*
+ * this returns the number of bytes used by the item on disk, minus the
+ * size of any extent headers.  If a file is compressed on disk, this is
+ * the compressed size
+ */
+static inline u32 btrfs_file_extent_inline_item_len(struct extent_buffer *eb,
+						    struct btrfs_item *e)
+{
+	unsigned long offset;
+	offset = offsetof(struct btrfs_file_extent_item, disk_bytenr);
+	return btrfs_item_size(eb, e) - offset;
+}
 
 static inline struct btrfs_root *btrfs_sb(struct super_block *sb)
 {
@@ -1745,10 +1805,11 @@ int btrfs_lookup_inode(struct btrfs_trans_handle *trans, struct btrfs_root
 int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode,
 			  struct bio *bio);
 int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
-			       struct btrfs_root *root,
-			       u64 objectid, u64 pos, u64 disk_offset,
-			       u64 disk_num_bytes,
-			     u64 num_bytes, u64 offset);
+			     struct btrfs_root *root,
+			     u64 objectid, u64 pos,
+			     u64 disk_offset, u64 disk_num_bytes,
+			     u64 num_bytes, u64 offset, u64 ram_bytes,
+			     u8 compression, u8 encryption, u16 other_encoding);
 int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root,
 			     struct btrfs_path *path, u64 objectid,
@@ -1758,6 +1819,8 @@ int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans,
 			   struct btrfs_ordered_sum *sums);
 int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
 		       struct bio *bio);
+int btrfs_csum_file_bytes(struct btrfs_root *root, struct inode *inode,
+			  u64 start, unsigned long len);
 struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans,
 					  struct btrfs_root *root,
 					  struct btrfs_path *path,
@@ -1799,7 +1862,7 @@ void btrfs_invalidate_dcache_root(struct btrfs_root *root, char *name,
 				  int namelen);
 
 int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
-			 size_t size, struct bio *bio);
+			 size_t size, struct bio *bio, unsigned long bio_flags);
 
 unsigned long btrfs_force_ra(struct address_space *mapping,
 			      struct file_ra_state *ra, struct file *file,
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 0be044bb6194..dc95f636a11b 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -83,6 +83,7 @@ struct async_submit_bio {
 	extent_submit_bio_hook_t *submit_bio_hook;
 	int rw;
 	int mirror_num;
+	unsigned long bio_flags;
 	struct btrfs_work work;
 };
 
@@ -115,6 +116,7 @@ struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
 	}
 	em->start = 0;
 	em->len = (u64)-1;
+	em->block_len = (u64)-1;
 	em->block_start = 0;
 	em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
 
@@ -469,12 +471,13 @@ static void run_one_async_submit(struct btrfs_work *work)
 		wake_up(&fs_info->async_submit_wait);
 
 	async->submit_bio_hook(async->inode, async->rw, async->bio,
-			       async->mirror_num);
+			       async->mirror_num, async->bio_flags);
 	kfree(async);
 }
 
 int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
 			int rw, struct bio *bio, int mirror_num,
+			unsigned long bio_flags,
 			extent_submit_bio_hook_t *submit_bio_hook)
 {
 	struct async_submit_bio *async;
@@ -491,6 +494,7 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
 	async->submit_bio_hook = submit_bio_hook;
 	async->work.func = run_one_async_submit;
 	async->work.flags = 0;
+	async->bio_flags = bio_flags;
 
 	while(atomic_read(&fs_info->async_submit_draining) &&
 	      atomic_read(&fs_info->nr_async_submits)) {
@@ -530,7 +534,7 @@ static int btree_csum_one_bio(struct bio *bio)
 }
 
 static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
-				 int mirror_num)
+				 int mirror_num, unsigned long bio_flags)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	int ret;
@@ -556,17 +560,17 @@ static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
 }
 
 static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
-				 int mirror_num)
+				 int mirror_num, unsigned long bio_flags)
 {
 	/*
 	 * kthread helpers are used to submit writes so that checksumming
 	 * can happen in parallel across all CPUs
 	 */
 	if (!(rw & (1 << BIO_RW))) {
-		return __btree_submit_bio_hook(inode, rw, bio, mirror_num);
+		return __btree_submit_bio_hook(inode, rw, bio, mirror_num, 0);
 	}
 	return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
-				   inode, rw, bio, mirror_num,
+				   inode, rw, bio, mirror_num, 0,
 				   __btree_submit_bio_hook);
 }
 
@@ -1407,6 +1411,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	fs_info->btree_inode = new_inode(sb);
 	fs_info->btree_inode->i_ino = 1;
 	fs_info->btree_inode->i_nlink = 1;
+
 	fs_info->thread_pool_size = min(num_online_cpus() + 2, 8);
 
 	INIT_LIST_HEAD(&fs_info->ordered_extents);
@@ -1508,6 +1513,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	 */
 	btrfs_init_workers(&fs_info->workers, "worker",
 			   fs_info->thread_pool_size);
+
 	btrfs_init_workers(&fs_info->submit_workers, "submit",
 			   min_t(u64, fs_devices->num_devices,
 			   fs_info->thread_pool_size));
@@ -1559,6 +1565,8 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	}
 
 	fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
+	fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
+				    4 * 1024 * 1024 / PAGE_CACHE_SIZE);
 
 	nodesize = btrfs_super_nodesize(disk_super);
 	leafsize = btrfs_super_leafsize(disk_super);
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index f84f5058dbbb..4eb1f1408d21 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -71,6 +71,7 @@ int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
 			int metadata);
 int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
 			int rw, struct bio *bio, int mirror_num,
+			unsigned long bio_flags,
 			extent_submit_bio_hook_t *submit_bio_hook);
 int btrfs_congested_async(struct btrfs_fs_info *info, int iodone);
 unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 280ac1aa9b6d..bbf04e80a1a3 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -3278,6 +3278,7 @@ static int noinline relocate_data_extent(struct inode *reloc_inode,
 
 	em->start = extent_key->objectid - offset;
 	em->len = extent_key->offset;
+	em->block_len = extent_key->offset;
 	em->block_start = extent_key->objectid;
 	em->bdev = root->fs_info->fs_devices->latest_bdev;
 	set_bit(EXTENT_FLAG_PINNED, &em->flags);
@@ -3314,10 +3315,14 @@ struct btrfs_ref_path {
 };
 
 struct disk_extent {
+	u64 ram_bytes;
 	u64 disk_bytenr;
 	u64 disk_num_bytes;
 	u64 offset;
 	u64 num_bytes;
+	u8 compression;
+	u8 encryption;
+	u16 other_encoding;
 };
 
 static int is_cowonly_root(u64 root_objectid)
@@ -3631,6 +3636,11 @@ static int noinline get_new_locations(struct inode *reloc_inode,
 			btrfs_file_extent_disk_num_bytes(leaf, fi);
 		exts[nr].offset = btrfs_file_extent_offset(leaf, fi);
 		exts[nr].num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
+		exts[nr].ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
+		exts[nr].compression = btrfs_file_extent_compression(leaf, fi);
+		exts[nr].encryption = btrfs_file_extent_encryption(leaf, fi);
+		exts[nr].other_encoding = btrfs_file_extent_other_encoding(leaf,
+									   fi);
 		WARN_ON(exts[nr].offset > 0);
 		WARN_ON(exts[nr].num_bytes != exts[nr].disk_num_bytes);
 
@@ -3846,6 +3856,8 @@ next:
 						new_extents[0].disk_bytenr);
 			btrfs_set_file_extent_disk_num_bytes(leaf, fi,
 						new_extents[0].disk_num_bytes);
+			btrfs_set_file_extent_ram_bytes(leaf, fi,
+						new_extents[0].ram_bytes);
 			ext_offset += new_extents[0].offset;
 			btrfs_set_file_extent_offset(leaf, fi, ext_offset);
 			btrfs_mark_buffer_dirty(leaf);
@@ -3911,6 +3923,16 @@ next:
 						new_extents[i].disk_bytenr);
 				btrfs_set_file_extent_disk_num_bytes(leaf, fi,
 						new_extents[i].disk_num_bytes);
+				btrfs_set_file_extent_ram_bytes(leaf, fi,
+						new_extents[i].ram_bytes);
+
+				btrfs_set_file_extent_compression(leaf, fi,
+						new_extents[i].compression);
+				btrfs_set_file_extent_encryption(leaf, fi,
+						new_extents[i].encryption);
+				btrfs_set_file_extent_other_encoding(leaf, fi,
+						new_extents[i].other_encoding);
+
 				btrfs_set_file_extent_num_bytes(leaf, fi,
 							extent_len);
 				ext_offset += new_extents[i].offset;
@@ -4169,6 +4191,8 @@ static int noinline replace_extents_in_leaf(struct btrfs_trans_handle *trans,
 		ref->extents[ext_index].num_bytes = new_extent->disk_num_bytes;
 
 		btrfs_set_file_extent_generation(leaf, fi, trans->transid);
+		btrfs_set_file_extent_ram_bytes(leaf, fi,
+						new_extent->ram_bytes);
 		btrfs_set_file_extent_disk_bytenr(leaf, fi,
 						new_extent->disk_bytenr);
 		btrfs_set_file_extent_disk_num_bytes(leaf, fi,
@@ -4847,7 +4871,8 @@ static struct inode noinline *create_reloc_inode(struct btrfs_fs_info *fs_info,
 	BUG_ON(err);
 
 	err = btrfs_insert_file_extent(trans, root, objectid, 0, 0, 0,
-				       group->key.offset, 0);
+				       group->key.offset, 0, group->key.offset,
+				       0, 0, 0);
 	BUG_ON(err);
 
 	inode = btrfs_iget_locked(root->fs_info->sb, objectid, root);
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 563b2d12f4f2..314041fdfa43 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -30,6 +30,7 @@ static struct kmem_cache *extent_buffer_cache;
 static LIST_HEAD(buffers);
 static LIST_HEAD(states);
 
+#define LEAK_DEBUG 1
 #ifdef LEAK_DEBUG
 static spinlock_t leak_lock = SPIN_LOCK_UNLOCKED;
 #endif
@@ -1067,8 +1068,8 @@ EXPORT_SYMBOL(find_first_extent_bit_state);
  *
  * 1 is returned if we find something, 0 if nothing was in the tree
  */
-static noinline u64 find_lock_delalloc_range(struct extent_io_tree *tree,
-					     u64 *start, u64 *end, u64 max_bytes)
+static noinline u64 find_delalloc_range(struct extent_io_tree *tree,
+					u64 *start, u64 *end, u64 max_bytes)
 {
 	struct rb_node *node;
 	struct extent_state *state;
@@ -1077,11 +1078,11 @@ static noinline u64 find_lock_delalloc_range(struct extent_io_tree *tree,
 	u64 total_bytes = 0;
 
 	spin_lock_irq(&tree->lock);
+
 	/*
 	 * this search will find all the extents that end after
 	 * our range starts.
 	 */
-search_again:
 	node = tree_search(tree, cur_start);
 	if (!node) {
 		if (!found)
@@ -1100,40 +1101,6 @@ search_again:
 				*end = state->end;
 			goto out;
 		}
-		if (!found && !(state->state & EXTENT_BOUNDARY)) {
-			struct extent_state *prev_state;
-			struct rb_node *prev_node = node;
-			while(1) {
-				prev_node = rb_prev(prev_node);
-				if (!prev_node)
-					break;
-				prev_state = rb_entry(prev_node,
-						      struct extent_state,
-						      rb_node);
-				if ((prev_state->end + 1 != state->start) ||
-				    !(prev_state->state & EXTENT_DELALLOC))
-					break;
-				if ((cur_start - prev_state->start) * 2 >
-				     max_bytes)
-					break;
-				state = prev_state;
-				node = prev_node;
-			}
-		}
-		if (state->state & EXTENT_LOCKED) {
-			DEFINE_WAIT(wait);
-			atomic_inc(&state->refs);
-			prepare_to_wait(&state->wq, &wait,
-					TASK_UNINTERRUPTIBLE);
-			spin_unlock_irq(&tree->lock);
-			schedule();
-			spin_lock_irq(&tree->lock);
-			finish_wait(&state->wq, &wait);
-			free_extent_state(state);
-			goto search_again;
-		}
-		set_state_cb(tree, state, EXTENT_LOCKED);
-		state->state |= EXTENT_LOCKED;
 		if (!found)
 			*start = state->start;
 		found++;
@@ -1151,6 +1118,208 @@ out:
 	return found;
 }
 
+static noinline int __unlock_for_delalloc(struct inode *inode,
+					  struct page *locked_page,
+					  u64 start, u64 end)
+{
+	int ret;
+	struct page *pages[16];
+	unsigned long index = start >> PAGE_CACHE_SHIFT;
+	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
+	unsigned long nr_pages = end_index - index + 1;
+	int i;
+
+	if (index == locked_page->index && end_index == index)
+		return 0;
+
+	while(nr_pages > 0) {
+		ret = find_get_pages_contig(inode->i_mapping, index,
+				     min(nr_pages, ARRAY_SIZE(pages)), pages);
+		for (i = 0; i < ret; i++) {
+			if (pages[i] != locked_page)
+				unlock_page(pages[i]);
+			page_cache_release(pages[i]);
+		}
+		nr_pages -= ret;
+		index += ret;
+		cond_resched();
+	}
+	return 0;
+}
+
+static noinline int lock_delalloc_pages(struct inode *inode,
+					struct page *locked_page,
+					u64 delalloc_start,
+					u64 delalloc_end)
+{
+	unsigned long index = delalloc_start >> PAGE_CACHE_SHIFT;
+	unsigned long start_index = index;
+	unsigned long end_index = delalloc_end >> PAGE_CACHE_SHIFT;
+	unsigned long pages_locked = 0;
+	struct page *pages[16];
+	unsigned long nrpages;
+	int ret;
+	int i;
+
+	/* the caller is responsible for locking the start index */
+	if (index == locked_page->index && index == end_index)
+		return 0;
+
+	/* skip the page at the start index */
+	nrpages = end_index - index + 1;
+	while(nrpages > 0) {
+		ret = find_get_pages_contig(inode->i_mapping, index,
+				     min(nrpages, ARRAY_SIZE(pages)), pages);
+		if (ret == 0) {
+			ret = -EAGAIN;
+			goto done;
+		}
+		/* now we have an array of pages, lock them all */
+		for (i = 0; i < ret; i++) {
+			/*
+			 * the caller is taking responsibility for
+			 * locked_page
+			 */
+			if (pages[i] != locked_page)
+				lock_page(pages[i]);
+			page_cache_release(pages[i]);
+		}
+		pages_locked += ret;
+		nrpages -= ret;
+		index += ret;
+		cond_resched();
+	}
+	ret = 0;
+done:
+	if (ret && pages_locked) {
+		__unlock_for_delalloc(inode, locked_page,
+			      delalloc_start,
+			      ((u64)(start_index + pages_locked - 1)) <<
+			      PAGE_CACHE_SHIFT);
+	}
+	return ret;
+}
+
+/*
+ * find a contiguous range of bytes in the file marked as delalloc, not
+ * more than 'max_bytes'.  start and end are used to return the range,
+ *
+ * 1 is returned if we find something, 0 if nothing was in the tree
+ */
+static noinline u64 find_lock_delalloc_range(struct inode *inode,
+					     struct extent_io_tree *tree,
+					     struct page *locked_page,
+					     u64 *start, u64 *end,
+					     u64 max_bytes)
+{
+	u64 delalloc_start;
+	u64 delalloc_end;
+	u64 found;
+	int ret;
+	int loops = 0;
+
+again:
+	/* step one, find a bunch of delalloc bytes starting at start */
+	delalloc_start = *start;
+	delalloc_end = 0;
+	found = find_delalloc_range(tree, &delalloc_start, &delalloc_end,
+				    max_bytes);
+	if (!found) {
+		*start = delalloc_start;
+		*end = delalloc_end;
+		return found;
+	}
+
+	/*
+	 * make sure to limit the number of pages we try to lock down
+	 * if we're looping.
+	 */
+	if (delalloc_end + 1 - delalloc_start > max_bytes && loops) {
+		delalloc_end = (delalloc_start + PAGE_CACHE_SIZE - 1) &
+			~((u64)PAGE_CACHE_SIZE - 1);
+	}
+	/* step two, lock all the pages after the page that has start */
+	ret = lock_delalloc_pages(inode, locked_page,
+				  delalloc_start, delalloc_end);
+	if (ret == -EAGAIN) {
+		/* some of the pages are gone, lets avoid looping by
+		 * shortening the size of the delalloc range we're searching
+		 */
+		if (!loops) {
+			unsigned long offset = (*start) & (PAGE_CACHE_SIZE - 1);
+			max_bytes = PAGE_CACHE_SIZE - offset;
+			loops = 1;
+			goto again;
+		} else {
+			found = 0;
+			goto out_failed;
+		}
+	}
+	BUG_ON(ret);
+
+	/* step three, lock the state bits for the whole range */
+	lock_extent(tree, delalloc_start, delalloc_end, GFP_NOFS);
+
+	/* then test to make sure it is all still delalloc */
+	ret = test_range_bit(tree, delalloc_start, delalloc_end,
+			     EXTENT_DELALLOC, 1);
+	if (!ret) {
+		unlock_extent(tree, delalloc_start, delalloc_end, GFP_NOFS);
+		__unlock_for_delalloc(inode, locked_page,
+			      delalloc_start, delalloc_end);
+		cond_resched();
+		goto again;
+	}
+	*start = delalloc_start;
+	*end = delalloc_end;
+out_failed:
+	return found;
+}
+
+int extent_clear_unlock_delalloc(struct inode *inode,
+				struct extent_io_tree *tree,
+				u64 start, u64 end, struct page *locked_page,
+				int clear_dirty, int set_writeback,
+				int end_writeback)
+{
+	int ret;
+	struct page *pages[16];
+	unsigned long index = start >> PAGE_CACHE_SHIFT;
+	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
+	unsigned long nr_pages = end_index - index + 1;
+	int i;
+	int clear_bits = EXTENT_LOCKED | EXTENT_DELALLOC;
+
+	if (clear_dirty)
+		clear_bits |= EXTENT_DIRTY;
+
+	clear_extent_bit(tree, start, end, clear_bits, 1, 0, GFP_NOFS);
+
+	while(nr_pages > 0) {
+		ret = find_get_pages_contig(inode->i_mapping, index,
+				     min(nr_pages, ARRAY_SIZE(pages)), pages);
+		for (i = 0; i < ret; i++) {
+			if (pages[i] == locked_page) {
+				page_cache_release(pages[i]);
+				continue;
+			}
+			if (clear_dirty)
+				clear_page_dirty_for_io(pages[i]);
+			if (set_writeback)
+				set_page_writeback(pages[i]);
+			if (end_writeback)
+				end_page_writeback(pages[i]);
+			unlock_page(pages[i]);
+			page_cache_release(pages[i]);
+		}
+		nr_pages -= ret;
+		index += ret;
+		cond_resched();
+	}
+	return 0;
+}
+EXPORT_SYMBOL(extent_clear_unlock_delalloc);
+
 /*
  * count the number of bytes in the tree that have a given bit(s)
  * set.  This can be fairly slow, except for EXTENT_DIRTY which is
@@ -1631,38 +1800,26 @@ extent_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
 	return bio;
 }
 
-static int submit_one_bio(int rw, struct bio *bio, int mirror_num)
+static int submit_one_bio(int rw, struct bio *bio, int mirror_num,
+			  unsigned long bio_flags)
 {
 	int ret = 0;
 	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
 	struct page *page = bvec->bv_page;
 	struct extent_io_tree *tree = bio->bi_private;
-	struct rb_node *node;
-	struct extent_state *state;
 	u64 start;
 	u64 end;
 
 	start = ((u64)page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset;
 	end = start + bvec->bv_len - 1;
 
-	spin_lock_irq(&tree->lock);
-	node = __etree_search(tree, start, NULL, NULL);
-	BUG_ON(!node);
-	state = rb_entry(node, struct extent_state, rb_node);
-	while(state->end < end) {
-		node = rb_next(node);
-		state = rb_entry(node, struct extent_state, rb_node);
-	}
-	BUG_ON(state->end != end);
-	spin_unlock_irq(&tree->lock);
-
 	bio->bi_private = NULL;
 
 	bio_get(bio);
 
 	if (tree->ops && tree->ops->submit_bio_hook)
 		tree->ops->submit_bio_hook(page->mapping->host, rw, bio,
-					   mirror_num);
+					   mirror_num, bio_flags);
 	else
 		submit_bio(rw, bio);
 	if (bio_flagged(bio, BIO_EOPNOTSUPP))
@@ -1678,39 +1835,56 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree,
 			      struct bio **bio_ret,
 			      unsigned long max_pages,
 			      bio_end_io_t end_io_func,
-			      int mirror_num)
+			      int mirror_num,
+			      unsigned long prev_bio_flags,
+			      unsigned long bio_flags)
 {
 	int ret = 0;
 	struct bio *bio;
 	int nr;
+	int contig = 0;
+	int this_compressed = bio_flags & EXTENT_BIO_COMPRESSED;
+	int old_compressed = prev_bio_flags & EXTENT_BIO_COMPRESSED;
+	size_t page_size = min(size, PAGE_CACHE_SIZE);
 
 	if (bio_ret && *bio_ret) {
 		bio = *bio_ret;
-		if (bio->bi_sector + (bio->bi_size >> 9) != sector ||
+		if (old_compressed)
+			contig = bio->bi_sector == sector;
+		else
+			contig = bio->bi_sector + (bio->bi_size >> 9) ==
+				sector;
+
+		if (prev_bio_flags != bio_flags || !contig ||
 		    (tree->ops && tree->ops->merge_bio_hook &&
-		     tree->ops->merge_bio_hook(page, offset, size, bio)) ||
-		    bio_add_page(bio, page, size, offset) < size) {
-			ret = submit_one_bio(rw, bio, mirror_num);
+		     tree->ops->merge_bio_hook(page, offset, page_size, bio,
+					       bio_flags)) ||
+		    bio_add_page(bio, page, page_size, offset) < page_size) {
+			ret = submit_one_bio(rw, bio, mirror_num,
+					     prev_bio_flags);
 			bio = NULL;
 		} else {
 			return 0;
 		}
 	}
-	nr = bio_get_nr_vecs(bdev);
+	if (this_compressed)
+		nr = BIO_MAX_PAGES;
+	else
+		nr = bio_get_nr_vecs(bdev);
+
 	bio = extent_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH);
 	if (!bio) {
 		printk("failed to allocate bio nr %d\n", nr);
 	}
 
-
-	bio_add_page(bio, page, size, offset);
+	bio_add_page(bio, page, page_size, offset);
 	bio->bi_end_io = end_io_func;
 	bio->bi_private = tree;
 
 	if (bio_ret) {
 		*bio_ret = bio;
 	} else {
-		ret = submit_one_bio(rw, bio, mirror_num);
+		ret = submit_one_bio(rw, bio, mirror_num, bio_flags);
 	}
 
 	return ret;
@@ -1738,7 +1912,8 @@ void set_page_extent_head(struct page *page, unsigned long len)
 static int __extent_read_full_page(struct extent_io_tree *tree,
 				   struct page *page,
 				   get_extent_t *get_extent,
-				   struct bio **bio, int mirror_num)
+				   struct bio **bio, int mirror_num,
+				   unsigned long *bio_flags)
 {
 	struct inode *inode = page->mapping->host;
 	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
@@ -1756,13 +1931,27 @@ static int __extent_read_full_page(struct extent_io_tree *tree,
 	int nr = 0;
 	size_t page_offset = 0;
 	size_t iosize;
+	size_t disk_io_size;
 	size_t blocksize = inode->i_sb->s_blocksize;
+	unsigned long this_bio_flag = 0;
 
 	set_page_extent_mapped(page);
 
 	end = page_end;
 	lock_extent(tree, start, end, GFP_NOFS);
 
+	if (page->index == last_byte >> PAGE_CACHE_SHIFT) {
+		char *userpage;
+		size_t zero_offset = last_byte & (PAGE_CACHE_SIZE - 1);
+
+		if (zero_offset) {
+			iosize = PAGE_CACHE_SIZE - zero_offset;
+			userpage = kmap_atomic(page, KM_USER0);
+			memset(userpage + zero_offset, 0, iosize);
+			flush_dcache_page(page);
+			kunmap_atomic(userpage, KM_USER0);
+		}
+	}
 	while (cur <= end) {
 		if (cur >= last_byte) {
 			char *userpage;
@@ -1793,10 +1982,19 @@ printk("2bad mapping end %Lu cur %Lu\n", end, cur);
 		}
 		BUG_ON(end < cur);
 
+		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
+			this_bio_flag = EXTENT_BIO_COMPRESSED;
+
 		iosize = min(extent_map_end(em) - cur, end - cur + 1);
 		cur_end = min(extent_map_end(em) - 1, end);
 		iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1);
-		sector = (em->block_start + extent_offset) >> 9;
+		if (this_bio_flag & EXTENT_BIO_COMPRESSED) {
+			disk_io_size = em->block_len;
+			sector = em->block_start >> 9;
+		} else {
+			sector = (em->block_start + extent_offset) >> 9;
+			disk_io_size = iosize;
+		}
 		bdev = em->bdev;
 		block_start = em->block_start;
 		free_extent_map(em);
@@ -1845,10 +2043,13 @@ printk("2bad mapping end %Lu cur %Lu\n", end, cur);
 			unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1;
 			pnr -= page->index;
 			ret = submit_extent_page(READ, tree, page,
-					 sector, iosize, page_offset,
+					 sector, disk_io_size, page_offset,
 					 bdev, bio, pnr,
-					 end_bio_extent_readpage, mirror_num);
+					 end_bio_extent_readpage, mirror_num,
+					 *bio_flags,
+					 this_bio_flag);
 			nr++;
+			*bio_flags = this_bio_flag;
 		}
 		if (ret)
 			SetPageError(page);
@@ -1867,11 +2068,13 @@ int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
 			    get_extent_t *get_extent)
 {
 	struct bio *bio = NULL;
+	unsigned long bio_flags = 0;
 	int ret;
 
-	ret = __extent_read_full_page(tree, page, get_extent, &bio, 0);
+	ret = __extent_read_full_page(tree, page, get_extent, &bio, 0,
+				      &bio_flags);
 	if (bio)
-		submit_one_bio(READ, bio, 0);
+		submit_one_bio(READ, bio, 0, bio_flags);
 	return ret;
 }
 EXPORT_SYMBOL(extent_read_full_page);
@@ -1909,6 +2112,8 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
 	unsigned long end_index = i_size >> PAGE_CACHE_SHIFT;
 	u64 nr_delalloc;
 	u64 delalloc_end;
+	int page_started;
+	int compressed;
 
 	WARN_ON(!PageLocked(page));
 	pg_offset = i_size & (PAGE_CACHE_SIZE - 1);
@@ -1934,27 +2139,33 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
 
 	delalloc_start = start;
 	delalloc_end = 0;
+	page_started = 0;
 	while(delalloc_end < page_end) {
-		nr_delalloc = find_lock_delalloc_range(tree, &delalloc_start,
+		nr_delalloc = find_lock_delalloc_range(inode, tree,
+						       page,
+						       &delalloc_start,
 						       &delalloc_end,
 						       128 * 1024 * 1024);
 		if (nr_delalloc == 0) {
 			delalloc_start = delalloc_end + 1;
 			continue;
 		}
-		tree->ops->fill_delalloc(inode, delalloc_start,
-					 delalloc_end);
-		clear_extent_bit(tree, delalloc_start,
-				 delalloc_end,
-				 EXTENT_LOCKED | EXTENT_DELALLOC,
-				 1, 0, GFP_NOFS);
+		tree->ops->fill_delalloc(inode, page, delalloc_start,
+					 delalloc_end, &page_started);
 		delalloc_start = delalloc_end + 1;
 	}
+
+	/* did the fill delalloc function already unlock and start the IO? */
+	if (page_started) {
+		return 0;
+	}
+
 	lock_extent(tree, start, page_end, GFP_NOFS);
 	unlock_start = start;
 
 	if (tree->ops && tree->ops->writepage_start_hook) {
-		ret = tree->ops->writepage_start_hook(page, start, page_end);
+		ret = tree->ops->writepage_start_hook(page, start,
+						      page_end);
 		if (ret == -EAGAIN) {
 			unlock_extent(tree, start, page_end, GFP_NOFS);
 			redirty_page_for_writepage(wbc, page);
@@ -2006,10 +2217,15 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
 		sector = (em->block_start + extent_offset) >> 9;
 		bdev = em->bdev;
 		block_start = em->block_start;
+		compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
 		free_extent_map(em);
 		em = NULL;
 
-		if (block_start == EXTENT_MAP_HOLE ||
+		/*
+		 * compressed and inline extents are written through other
+		 * paths in the FS
+		 */
+		if (compressed || block_start == EXTENT_MAP_HOLE ||
 		    block_start == EXTENT_MAP_INLINE) {
 			clear_extent_dirty(tree, cur,
 					   cur + iosize - 1, GFP_NOFS);
@@ -2017,16 +2233,28 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
 			unlock_extent(tree, unlock_start, cur + iosize -1,
 				      GFP_NOFS);
 
-			if (tree->ops && tree->ops->writepage_end_io_hook)
+			/*
+			 * end_io notification does not happen here for
+			 * compressed extents
+			 */
+			if (!compressed && tree->ops &&
+			    tree->ops->writepage_end_io_hook)
 				tree->ops->writepage_end_io_hook(page, cur,
 							 cur + iosize - 1,
 							 NULL, 1);
-			cur = cur + iosize;
+			else if (compressed) {
+				/* we don't want to end_page_writeback on
+				 * a compressed extent.  this happens
+				 * elsewhere
+				 */
+				nr++;
+			}
+
+			cur += iosize;
 			pg_offset += iosize;
 			unlock_start = cur;
 			continue;
 		}
-
 		/* leave this out until we have a page_mkwrite call */
 		if (0 && !test_range_bit(tree, cur, cur + iosize - 1,
 				   EXTENT_DIRTY, 0)) {
@@ -2034,6 +2262,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
 			pg_offset += iosize;
 			continue;
 		}
+
 		clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS);
 		if (tree->ops && tree->ops->writepage_io_hook) {
 			ret = tree->ops->writepage_io_hook(page, cur,
@@ -2057,7 +2286,8 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
 			ret = submit_extent_page(WRITE, tree, page, sector,
 						 iosize, pg_offset, bdev,
 						 &epd->bio, max_nr,
-						 end_bio_extent_writepage, 0);
+						 end_bio_extent_writepage,
+						 0, 0, 0);
 			if (ret)
 				SetPageError(page);
 		}
@@ -2226,7 +2456,7 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page,
 	extent_write_cache_pages(tree, mapping, &wbc_writepages,
 				 __extent_writepage, &epd);
 	if (epd.bio) {
-		submit_one_bio(WRITE, epd.bio, 0);
+		submit_one_bio(WRITE, epd.bio, 0, 0);
 	}
 	return ret;
 }
@@ -2248,7 +2478,7 @@ int extent_writepages(struct extent_io_tree *tree,
 	ret = extent_write_cache_pages(tree, mapping, wbc,
 				       __extent_writepage, &epd);
 	if (epd.bio) {
-		submit_one_bio(WRITE, epd.bio, 0);
+		submit_one_bio(WRITE, epd.bio, 0, 0);
 	}
 	return ret;
 }
@@ -2262,6 +2492,7 @@ int extent_readpages(struct extent_io_tree *tree,
 	struct bio *bio = NULL;
 	unsigned page_idx;
 	struct pagevec pvec;
+	unsigned long bio_flags = 0;
 
 	pagevec_init(&pvec, 0);
 	for (page_idx = 0; page_idx < nr_pages; page_idx++) {
@@ -2281,7 +2512,7 @@ int extent_readpages(struct extent_io_tree *tree,
 			if (!pagevec_add(&pvec, page))
 				__pagevec_lru_add(&pvec);
 			__extent_read_full_page(tree, page, get_extent,
-						&bio, 0);
+						&bio, 0, &bio_flags);
 		}
 		page_cache_release(page);
 	}
@@ -2289,7 +2520,7 @@ int extent_readpages(struct extent_io_tree *tree,
 		__pagevec_lru_add(&pvec);
 	BUG_ON(!list_empty(pages));
 	if (bio)
-		submit_one_bio(READ, bio, 0);
+		submit_one_bio(READ, bio, 0, bio_flags);
 	return 0;
 }
 EXPORT_SYMBOL(extent_readpages);
@@ -2414,7 +2645,8 @@ int extent_prepare_write(struct extent_io_tree *tree,
 			ret = submit_extent_page(READ, tree, page,
 					 sector, iosize, page_offset, em->bdev,
 					 NULL, 1,
-					 end_bio_extent_preparewrite, 0);
+					 end_bio_extent_preparewrite, 0,
+					 0, 0);
 			iocount++;
 			block_start = block_start + iosize;
 		} else {
@@ -2495,7 +2727,9 @@ int try_release_extent_mapping(struct extent_map_tree *map,
 			}
 			if (!test_range_bit(tree, em->start,
 					    extent_map_end(em) - 1,
-					    EXTENT_LOCKED, 0)) {
+					    EXTENT_LOCKED | EXTENT_WRITEBACK |
+					    EXTENT_ORDERED,
+					    0)) {
 				remove_extent_mapping(map, em);
 				/* once for the rb tree */
 				free_extent_map(em);
@@ -2923,6 +3157,7 @@ int read_extent_buffer_pages(struct extent_io_tree *tree,
 	int inc_all_pages = 0;
 	unsigned long num_pages;
 	struct bio *bio = NULL;
+	unsigned long bio_flags = 0;
 
 	if (eb->flags & EXTENT_UPTODATE)
 		return 0;
@@ -2973,7 +3208,7 @@ int read_extent_buffer_pages(struct extent_io_tree *tree,
 			ClearPageError(page);
 			err = __extent_read_full_page(tree, page,
 						      get_extent, &bio,
-						      mirror_num);
+						      mirror_num, &bio_flags);
 			if (err) {
 				ret = err;
 				printk("err %d from __extent_read_full_page\n", ret);
@@ -2984,7 +3219,7 @@ int read_extent_buffer_pages(struct extent_io_tree *tree,
 	}
 
 	if (bio)
-		submit_one_bio(READ, bio, mirror_num);
+		submit_one_bio(READ, bio, mirror_num, bio_flags);
 
 	if (ret || !wait) {
 		if (ret)
diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h
index c9d1908a1ae3..86f859b87a6e 100644
--- a/fs/btrfs/extent_io.h
+++ b/fs/btrfs/extent_io.h
@@ -18,6 +18,9 @@
 #define EXTENT_BOUNDARY (1 << 11)
 #define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK)
 
+/* flags for bio submission */
+#define EXTENT_BIO_COMPRESSED 1
+
 /*
  * page->private values.  Every page that is controlled by the extent
  * map has page->private set to one.
@@ -28,14 +31,17 @@
 struct extent_state;
 
 typedef	int (extent_submit_bio_hook_t)(struct inode *inode, int rw,
-				       struct bio *bio, int mirror_num);
+				       struct bio *bio, int mirror_num,
+				       unsigned long bio_flags);
 struct extent_io_ops {
-	int (*fill_delalloc)(struct inode *inode, u64 start, u64 end);
+	int (*fill_delalloc)(struct inode *inode, struct page *locked_page,
+			     u64 start, u64 end, int *page_started);
 	int (*writepage_start_hook)(struct page *page, u64 start, u64 end);
 	int (*writepage_io_hook)(struct page *page, u64 start, u64 end);
 	extent_submit_bio_hook_t *submit_bio_hook;
 	int (*merge_bio_hook)(struct page *page, unsigned long offset,
-			      size_t size, struct bio *bio);
+			      size_t size, struct bio *bio,
+			      unsigned long bio_flags);
 	int (*readpage_io_hook)(struct page *page, u64 start, u64 end);
 	int (*readpage_io_failed_hook)(struct bio *bio, struct page *page,
 				       u64 start, u64 end,
@@ -245,4 +251,9 @@ void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km);
 int release_extent_buffer_tail_pages(struct extent_buffer *eb);
 int extent_range_uptodate(struct extent_io_tree *tree,
 			  u64 start, u64 end);
+int extent_clear_unlock_delalloc(struct inode *inode,
+				struct extent_io_tree *tree,
+				u64 start, u64 end, struct page *locked_page,
+				int clear_dirty, int set_writeback,
+				int clear_writeback);
 #endif
diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c
index 74b2a29880d3..fd3ebfb8c3c5 100644
--- a/fs/btrfs/extent_map.c
+++ b/fs/btrfs/extent_map.c
@@ -184,6 +184,13 @@ static int mergable_maps(struct extent_map *prev, struct extent_map *next)
 	if (test_bit(EXTENT_FLAG_PINNED, &prev->flags))
 		return 0;
 
+	/*
+	 * don't merge compressed extents, we need to know their
+	 * actual size
+	 */
+	if (test_bit(EXTENT_FLAG_COMPRESSED, &prev->flags))
+		return 0;
+
 	if (extent_map_end(prev) == next->start &&
 	    prev->flags == next->flags &&
 	    prev->bdev == next->bdev &&
@@ -239,6 +246,7 @@ int add_extent_mapping(struct extent_map_tree *tree,
 		if (rb && mergable_maps(merge, em)) {
 			em->start = merge->start;
 			em->len += merge->len;
+			em->block_len += merge->block_len;
 			em->block_start = merge->block_start;
 			merge->in_tree = 0;
 			rb_erase(&merge->rb_node, &tree->map);
@@ -250,6 +258,7 @@ int add_extent_mapping(struct extent_map_tree *tree,
 		merge = rb_entry(rb, struct extent_map, rb_node);
 	if (rb && mergable_maps(em, merge)) {
 		em->len += merge->len;
+		em->block_len += merge->len;
 		rb_erase(&merge->rb_node, &tree->map);
 		merge->in_tree = 0;
 		free_extent_map(merge);
diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h
index 26ac6fe0b268..abbcbeb28c79 100644
--- a/fs/btrfs/extent_map.h
+++ b/fs/btrfs/extent_map.h
@@ -10,6 +10,7 @@
 
 /* bits for the flags field */
 #define EXTENT_FLAG_PINNED 0 /* this entry not yet on disk, don't free it */
+#define EXTENT_FLAG_COMPRESSED 1
 
 struct extent_map {
 	struct rb_node rb_node;
@@ -18,6 +19,7 @@ struct extent_map {
 	u64 start;
 	u64 len;
 	u64 block_start;
+	u64 block_len;
 	unsigned long flags;
 	struct block_device *bdev;
 	atomic_t refs;
@@ -38,9 +40,9 @@ static inline u64 extent_map_end(struct extent_map *em)
 
 static inline u64 extent_map_block_end(struct extent_map *em)
 {
-	if (em->block_start + em->len < em->block_start)
+	if (em->block_start + em->block_len < em->block_start)
 		return (u64)-1;
-	return em->block_start + em->len;
+	return em->block_start + em->block_len;
 }
 
 void extent_map_tree_init(struct extent_map_tree *tree, gfp_t mask);
diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c
index 6dbe88b9d7d4..f4d3fa71bc41 100644
--- a/fs/btrfs/file-item.c
+++ b/fs/btrfs/file-item.c
@@ -31,7 +31,8 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root,
 			     u64 objectid, u64 pos,
 			     u64 disk_offset, u64 disk_num_bytes,
-			     u64 num_bytes, u64 offset)
+			     u64 num_bytes, u64 offset, u64 ram_bytes,
+			     u8 compression, u8 encryption, u16 other_encoding)
 {
 	int ret = 0;
 	struct btrfs_file_extent_item *item;
@@ -57,8 +58,13 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
 	btrfs_set_file_extent_disk_num_bytes(leaf, item, disk_num_bytes);
 	btrfs_set_file_extent_offset(leaf, item, offset);
 	btrfs_set_file_extent_num_bytes(leaf, item, num_bytes);
+	btrfs_set_file_extent_ram_bytes(leaf, item, ram_bytes);
 	btrfs_set_file_extent_generation(leaf, item, trans->transid);
 	btrfs_set_file_extent_type(leaf, item, BTRFS_FILE_EXTENT_REG);
+	btrfs_set_file_extent_compression(leaf, item, compression);
+	btrfs_set_file_extent_encryption(leaf, item, encryption);
+	btrfs_set_file_extent_other_encoding(leaf, item, other_encoding);
+
 	btrfs_mark_buffer_dirty(leaf);
 out:
 	btrfs_free_path(path);
@@ -213,6 +219,73 @@ found:
 	return 0;
 }
 
+int btrfs_csum_file_bytes(struct btrfs_root *root, struct inode *inode,
+			  u64 start, unsigned long len)
+{
+	struct btrfs_ordered_sum *sums;
+	struct btrfs_sector_sum *sector_sum;
+	struct btrfs_ordered_extent *ordered;
+	char *data;
+	struct page *page;
+	unsigned long total_bytes = 0;
+	unsigned long this_sum_bytes = 0;
+
+	sums = kzalloc(btrfs_ordered_sum_size(root, len), GFP_NOFS);
+	if (!sums)
+		return -ENOMEM;
+
+	sector_sum = sums->sums;
+	sums->file_offset = start;
+	sums->len = len;
+	INIT_LIST_HEAD(&sums->list);
+	ordered = btrfs_lookup_ordered_extent(inode, sums->file_offset);
+	BUG_ON(!ordered);
+
+	while(len > 0) {
+		if (start >= ordered->file_offset + ordered->len ||
+		    start < ordered->file_offset) {
+			sums->len = this_sum_bytes;
+			this_sum_bytes = 0;
+			btrfs_add_ordered_sum(inode, ordered, sums);
+			btrfs_put_ordered_extent(ordered);
+
+			sums = kzalloc(btrfs_ordered_sum_size(root, len),
+				       GFP_NOFS);
+			BUG_ON(!sums);
+			sector_sum = sums->sums;
+			sums->len = len;
+			sums->file_offset = start;
+			ordered = btrfs_lookup_ordered_extent(inode,
+						      sums->file_offset);
+			BUG_ON(!ordered);
+		}
+
+		page = find_get_page(inode->i_mapping,
+				     start >> PAGE_CACHE_SHIFT);
+
+		data = kmap_atomic(page, KM_USER0);
+		sector_sum->sum = ~(u32)0;
+		sector_sum->sum = btrfs_csum_data(root, data, sector_sum->sum,
+						  PAGE_CACHE_SIZE);
+		kunmap_atomic(data, KM_USER0);
+		btrfs_csum_final(sector_sum->sum,
+				 (char *)&sector_sum->sum);
+		sector_sum->offset = page_offset(page);
+		page_cache_release(page);
+
+		sector_sum++;
+		total_bytes += PAGE_CACHE_SIZE;
+		this_sum_bytes += PAGE_CACHE_SIZE;
+		start += PAGE_CACHE_SIZE;
+
+		WARN_ON(len < PAGE_CACHE_SIZE);
+		len -= PAGE_CACHE_SIZE;
+	}
+	btrfs_add_ordered_sum(inode, ordered, sums);
+	btrfs_put_ordered_extent(ordered);
+	return 0;
+}
+
 int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
 		       struct bio *bio)
 {
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 69abbe19add2..0aa15436590e 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -95,153 +95,6 @@ static void noinline btrfs_drop_pages(struct page **pages, size_t num_pages)
 	}
 }
 
-/* this does all the hard work for inserting an inline extent into
- * the btree.  Any existing inline extent is extended as required to make room,
- * otherwise things are inserted as required into the btree
- */
-static int noinline insert_inline_extent(struct btrfs_trans_handle *trans,
-				struct btrfs_root *root, struct inode *inode,
-				u64 offset, size_t size,
-				struct page **pages, size_t page_offset,
-				int num_pages)
-{
-	struct btrfs_key key;
-	struct btrfs_path *path;
-	struct extent_buffer *leaf;
-	char *kaddr;
-	unsigned long ptr;
-	struct btrfs_file_extent_item *ei;
-	struct page *page;
-	u32 datasize;
-	int err = 0;
-	int ret;
-	int i;
-	ssize_t cur_size;
-
-	path = btrfs_alloc_path();
-	if (!path)
-		return -ENOMEM;
-
-	btrfs_set_trans_block_group(trans, inode);
-
-	key.objectid = inode->i_ino;
-	key.offset = offset;
-	btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
-
-	ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
-	if (ret < 0) {
-		err = ret;
-		goto fail;
-	}
-	if (ret == 1) {
-		struct btrfs_key found_key;
-
-		if (path->slots[0] == 0)
-			goto insert;
-
-		path->slots[0]--;
-		leaf = path->nodes[0];
-		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
-
-		if (found_key.objectid != inode->i_ino)
-			goto insert;
-
-		if (found_key.type != BTRFS_EXTENT_DATA_KEY)
-			goto insert;
-		ei = btrfs_item_ptr(leaf, path->slots[0],
-				    struct btrfs_file_extent_item);
-
-		if (btrfs_file_extent_type(leaf, ei) !=
-		    BTRFS_FILE_EXTENT_INLINE) {
-			goto insert;
-		}
-		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
-		ret = 0;
-	}
-	if (ret == 0) {
-		u32 found_size;
-		u64 found_end;
-
-		leaf = path->nodes[0];
-		ei = btrfs_item_ptr(leaf, path->slots[0],
-				    struct btrfs_file_extent_item);
-
-		if (btrfs_file_extent_type(leaf, ei) !=
-		    BTRFS_FILE_EXTENT_INLINE) {
-			err = ret;
-			btrfs_print_leaf(root, leaf);
-			printk("found wasn't inline offset %Lu inode %lu\n",
-			       offset, inode->i_ino);
-			goto fail;
-		}
-		found_size = btrfs_file_extent_inline_len(leaf,
-					  btrfs_item_nr(leaf, path->slots[0]));
-		found_end = key.offset + found_size;
-
-		if (found_end < offset + size) {
-			btrfs_release_path(root, path);
-			ret = btrfs_search_slot(trans, root, &key, path,
-						offset + size - found_end, 1);
-			BUG_ON(ret != 0);
-
-			ret = btrfs_extend_item(trans, root, path,
-						offset + size - found_end);
-			if (ret) {
-				err = ret;
-				goto fail;
-			}
-			leaf = path->nodes[0];
-			ei = btrfs_item_ptr(leaf, path->slots[0],
-					    struct btrfs_file_extent_item);
-			inode_add_bytes(inode, offset + size - found_end);
-		}
-		if (found_end < offset) {
-			ptr = btrfs_file_extent_inline_start(ei) + found_size;
-			memset_extent_buffer(leaf, 0, ptr, offset - found_end);
-		}
-	} else {
-insert:
-		btrfs_release_path(root, path);
-		datasize = offset + size - key.offset;
-		inode_add_bytes(inode, datasize);
-		datasize = btrfs_file_extent_calc_inline_size(datasize);
-		ret = btrfs_insert_empty_item(trans, root, path, &key,
-					      datasize);
-		if (ret) {
-			err = ret;
-			printk("got bad ret %d\n", ret);
-			goto fail;
-		}
-		leaf = path->nodes[0];
-		ei = btrfs_item_ptr(leaf, path->slots[0],
-				    struct btrfs_file_extent_item);
-		btrfs_set_file_extent_generation(leaf, ei, trans->transid);
-		btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE);
-	}
-	ptr = btrfs_file_extent_inline_start(ei) + offset - key.offset;
-
-	cur_size = size;
-	i = 0;
-	while (size > 0) {
-		page = pages[i];
-		kaddr = kmap_atomic(page, KM_USER0);
-		cur_size = min_t(size_t, PAGE_CACHE_SIZE - page_offset, size);
-		write_extent_buffer(leaf, kaddr + page_offset, ptr, cur_size);
-		kunmap_atomic(kaddr, KM_USER0);
-		page_offset = 0;
-		ptr += cur_size;
-		size -= cur_size;
-		if (i >= num_pages) {
-			printk("i %d num_pages %d\n", i, num_pages);
-		}
-		i++;
-	}
-	btrfs_mark_buffer_dirty(leaf);
-fail:
-	btrfs_free_path(path);
-	return err;
-}
-
 /*
  * after copy_from_user, pages need to be dirtied and we need to make
  * sure holes are created between the current EOF and the start of
@@ -267,8 +120,6 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans,
 	u64 start_pos;
 	u64 end_of_last_block;
 	u64 end_pos = pos + write_bytes;
-	u64 inline_size;
-	int did_inline = 0;
 	loff_t isize = i_size_read(inode);
 
 	start_pos = pos & ~((u64)root->sectorsize - 1);
@@ -314,7 +165,8 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans,
 			err = btrfs_insert_file_extent(trans, root,
 						       inode->i_ino,
 						       last_pos_in_file,
-						       0, 0, hole_size, 0);
+						       0, 0, hole_size, 0,
+						       hole_size, 0, 0, 0);
 			btrfs_drop_extent_cache(inode, last_pos_in_file,
 					last_pos_in_file + hole_size - 1, 0);
 			mutex_unlock(&BTRFS_I(inode)->extent_mutex);
@@ -324,57 +176,19 @@ static int noinline dirty_and_release_pages(struct btrfs_trans_handle *trans,
 			goto failed;
 	}
 
-	/*
-	 * either allocate an extent for the new bytes or setup the key
-	 * to show we are doing inline data in the extent
+	/* check for reserved extents on each page, we don't want
+	 * to reset the delalloc bit on things that already have
+	 * extents reserved.
 	 */
-	inline_size = end_pos;
-	if (isize >= BTRFS_MAX_INLINE_DATA_SIZE(root) ||
-	    inline_size > root->fs_info->max_inline ||
-	    (inline_size & (root->sectorsize -1)) == 0 ||
-	    inline_size >= BTRFS_MAX_INLINE_DATA_SIZE(root)) {
-		/* check for reserved extents on each page, we don't want
-		 * to reset the delalloc bit on things that already have
-		 * extents reserved.
-		 */
-		btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block);
-		for (i = 0; i < num_pages; i++) {
-			struct page *p = pages[i];
-			SetPageUptodate(p);
-			ClearPageChecked(p);
-			set_page_dirty(p);
-		}
-	} else {
-		u64 aligned_end;
-		/* step one, delete the existing extents in this range */
-		aligned_end = (pos + write_bytes + root->sectorsize - 1) &
-			~((u64)root->sectorsize - 1);
-		mutex_lock(&BTRFS_I(inode)->extent_mutex);
-		err = btrfs_drop_extents(trans, root, inode, start_pos,
-					 aligned_end, aligned_end, &hint_byte);
-		if (err)
-			goto failed;
-		if (isize > inline_size)
-			inline_size = min_t(u64, isize, aligned_end);
-		inline_size -= start_pos;
-		err = insert_inline_extent(trans, root, inode, start_pos,
-					   inline_size, pages, 0, num_pages);
-		btrfs_drop_extent_cache(inode, start_pos, aligned_end - 1, 0);
-		BUG_ON(err);
-		mutex_unlock(&BTRFS_I(inode)->extent_mutex);
-
-		/*
-		 * an ugly way to do all the prop accounting around
-		 * the page bits and mapping tags
-		 */
-		set_page_writeback(pages[0]);
-		end_page_writeback(pages[0]);
-		did_inline = 1;
+	btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block);
+	for (i = 0; i < num_pages; i++) {
+		struct page *p = pages[i];
+		SetPageUptodate(p);
+		ClearPageChecked(p);
+		set_page_dirty(p);
 	}
 	if (end_pos > isize) {
 		i_size_write(inode, end_pos);
-		if (did_inline)
-			BTRFS_I(inode)->disk_i_size = end_pos;
 		btrfs_update_inode(trans, root, inode);
 	}
 failed:
@@ -399,6 +213,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
 	int ret;
 	int testend = 1;
 	unsigned long flags;
+	int compressed = 0;
 
 	WARN_ON(end < start);
 	if (end == (u64)-1) {
@@ -434,6 +249,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
 			free_extent_map(em);
 			continue;
 		}
+		compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
 		clear_bit(EXTENT_FLAG_PINNED, &em->flags);
 		remove_extent_mapping(em_tree, em);
 
@@ -442,6 +258,12 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
 			split->start = em->start;
 			split->len = start - em->start;
 			split->block_start = em->block_start;
+
+			if (compressed)
+				split->block_len = em->block_len;
+			else
+				split->block_len = split->len;
+
 			split->bdev = em->bdev;
 			split->flags = flags;
 			ret = add_extent_mapping(em_tree, split);
@@ -459,7 +281,13 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
 			split->bdev = em->bdev;
 			split->flags = flags;
 
-			split->block_start = em->block_start + diff;
+			if (compressed) {
+				split->block_len = em->block_len;
+				split->block_start = em->block_start;
+			} else {
+				split->block_len = split->len;
+				split->block_start = em->block_start + diff;
+			}
 
 			ret = add_extent_mapping(em_tree, split);
 			BUG_ON(ret);
@@ -533,7 +361,7 @@ int btrfs_check_file(struct btrfs_root *root, struct inode *inode)
 			struct btrfs_item *item;
 			item = btrfs_item_nr(leaf, slot);
 			extent_end = found_key.offset +
-			     btrfs_file_extent_inline_len(leaf, item);
+			     btrfs_file_extent_inline_len(leaf, extent);
 			extent_end = (extent_end + root->sectorsize - 1) &
 				~((u64)root->sectorsize -1 );
 		}
@@ -573,6 +401,10 @@ int noinline btrfs_drop_extents(struct btrfs_trans_handle *trans,
 	u64 extent_end = 0;
 	u64 search_start = start;
 	u64 leaf_start;
+	u64 ram_bytes = 0;
+	u8 compression = 0;
+	u8 encryption = 0;
+	u16 other_encoding = 0;
 	u64 root_gen;
 	u64 root_owner;
 	struct extent_buffer *leaf;
@@ -589,6 +421,7 @@ int noinline btrfs_drop_extents(struct btrfs_trans_handle *trans,
 	int recow;
 	int ret;
 
+	inline_limit = 0;
 	btrfs_drop_extent_cache(inode, start, end - 1, 0);
 
 	path = btrfs_alloc_path();
@@ -637,6 +470,12 @@ next_slot:
 			extent = btrfs_item_ptr(leaf, slot,
 						struct btrfs_file_extent_item);
 			found_type = btrfs_file_extent_type(leaf, extent);
+			compression = btrfs_file_extent_compression(leaf,
+								    extent);
+			encryption = btrfs_file_extent_encryption(leaf,
+								  extent);
+			other_encoding = btrfs_file_extent_other_encoding(leaf,
+								  extent);
 			if (found_type == BTRFS_FILE_EXTENT_REG) {
 				extent_end =
 				     btrfs_file_extent_disk_bytenr(leaf,
@@ -646,13 +485,13 @@ next_slot:
 
 				extent_end = key.offset +
 				     btrfs_file_extent_num_bytes(leaf, extent);
+				ram_bytes = btrfs_file_extent_ram_bytes(leaf,
+								extent);
 				found_extent = 1;
 			} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
-				struct btrfs_item *item;
-				item = btrfs_item_nr(leaf, slot);
 				found_inline = 1;
 				extent_end = key.offset +
-				     btrfs_file_extent_inline_len(leaf, item);
+				     btrfs_file_extent_inline_len(leaf, extent);
 			}
 		} else {
 			extent_end = search_start;
@@ -680,10 +519,9 @@ next_slot:
 			search_start = (extent_end + mask) & ~mask;
 		} else
 			search_start = extent_end;
-		if (end <= extent_end && start >= key.offset && found_inline) {
+
+		if (end <= extent_end && start >= key.offset && found_inline)
 			*hint_byte = EXTENT_MAP_INLINE;
-			goto out;
-		}
 
 		if (found_extent) {
 			read_extent_buffer(leaf, &old, (unsigned long)extent,
@@ -770,12 +608,27 @@ next_slot:
 			write_extent_buffer(leaf, &old,
 					    (unsigned long)extent, sizeof(old));
 
+			btrfs_set_file_extent_compression(leaf, extent,
+							  compression);
+			btrfs_set_file_extent_encryption(leaf, extent,
+							 encryption);
+			btrfs_set_file_extent_other_encoding(leaf, extent,
+							     other_encoding);
 			btrfs_set_file_extent_offset(leaf, extent,
 				    le64_to_cpu(old.offset) + end - key.offset);
 			WARN_ON(le64_to_cpu(old.num_bytes) <
 				(extent_end - end));
 			btrfs_set_file_extent_num_bytes(leaf, extent,
 							extent_end - end);
+
+			/*
+			 * set the ram bytes to the size of the full extent
+			 * before splitting.  This is a worst case flag,
+			 * but its the best we can do because we don't know
+			 * how splitting affects compression
+			 */
+			btrfs_set_file_extent_ram_bytes(leaf, extent,
+							ram_bytes);
 			btrfs_set_file_extent_type(leaf, extent,
 						   BTRFS_FILE_EXTENT_REG);
 
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index bf4bed6ca4d6..9797592dc86b 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -49,6 +49,7 @@
 #include "compat.h"
 #include "tree-log.h"
 #include "ref-cache.h"
+#include "compression.h"
 
 struct btrfs_iget_args {
 	u64 ino;
@@ -83,6 +84,7 @@ static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
 };
 
 static void btrfs_truncate(struct inode *inode);
+static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end);
 
 /*
  * a very lame attempt at stopping writes when the FS is 85% full.  There
@@ -113,58 +115,375 @@ int btrfs_check_free_space(struct btrfs_root *root, u64 num_required,
 	return ret;
 }
 
+/*
+ * this does all the hard work for inserting an inline extent into
+ * the btree.  The caller should have done a btrfs_drop_extents so that
+ * no overlapping inline items exist in the btree
+ */
+static int noinline insert_inline_extent(struct btrfs_trans_handle *trans,
+				struct btrfs_root *root, struct inode *inode,
+				u64 start, size_t size, size_t compressed_size,
+				struct page **compressed_pages)
+{
+	struct btrfs_key key;
+	struct btrfs_path *path;
+	struct extent_buffer *leaf;
+	struct page *page = NULL;
+	char *kaddr;
+	unsigned long ptr;
+	struct btrfs_file_extent_item *ei;
+	int err = 0;
+	int ret;
+	size_t cur_size = size;
+	size_t datasize;
+	unsigned long offset;
+	int use_compress = 0;
+
+	if (compressed_size && compressed_pages) {
+		use_compress = 1;
+		cur_size = compressed_size;
+	}
+
+	path = btrfs_alloc_path(); if (!path)
+		return -ENOMEM;
+
+	btrfs_set_trans_block_group(trans, inode);
+
+	key.objectid = inode->i_ino;
+	key.offset = start;
+	btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
+	inode_add_bytes(inode, size);
+	datasize = btrfs_file_extent_calc_inline_size(cur_size);
+
+	inode_add_bytes(inode, size);
+	ret = btrfs_insert_empty_item(trans, root, path, &key,
+				      datasize);
+	BUG_ON(ret);
+	if (ret) {
+		err = ret;
+		printk("got bad ret %d\n", ret);
+		goto fail;
+	}
+	leaf = path->nodes[0];
+	ei = btrfs_item_ptr(leaf, path->slots[0],
+			    struct btrfs_file_extent_item);
+	btrfs_set_file_extent_generation(leaf, ei, trans->transid);
+	btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE);
+	btrfs_set_file_extent_encryption(leaf, ei, 0);
+	btrfs_set_file_extent_other_encoding(leaf, ei, 0);
+	btrfs_set_file_extent_ram_bytes(leaf, ei, size);
+	ptr = btrfs_file_extent_inline_start(ei);
+
+	if (use_compress) {
+		struct page *cpage;
+		int i = 0;
+		while(compressed_size > 0) {
+			cpage = compressed_pages[i];
+			cur_size = min(compressed_size,
+				       PAGE_CACHE_SIZE);
+
+			kaddr = kmap(cpage);
+			write_extent_buffer(leaf, kaddr, ptr, cur_size);
+			kunmap(cpage);
+
+			i++;
+			ptr += cur_size;
+			compressed_size -= cur_size;
+		}
+		btrfs_set_file_extent_compression(leaf, ei,
+						  BTRFS_COMPRESS_ZLIB);
+	} else {
+		page = find_get_page(inode->i_mapping,
+				     start >> PAGE_CACHE_SHIFT);
+		btrfs_set_file_extent_compression(leaf, ei, 0);
+		kaddr = kmap_atomic(page, KM_USER0);
+		offset = start & (PAGE_CACHE_SIZE - 1);
+		write_extent_buffer(leaf, kaddr + offset, ptr, size);
+		kunmap_atomic(kaddr, KM_USER0);
+		page_cache_release(page);
+	}
+	btrfs_mark_buffer_dirty(leaf);
+	btrfs_free_path(path);
+
+	BTRFS_I(inode)->disk_i_size = inode->i_size;
+	btrfs_update_inode(trans, root, inode);
+	return 0;
+fail:
+	btrfs_free_path(path);
+	return err;
+}
+
+
+/*
+ * conditionally insert an inline extent into the file.  This
+ * does the checks required to make sure the data is small enough
+ * to fit as an inline extent.
+ */
+static int cow_file_range_inline(struct btrfs_trans_handle *trans,
+				 struct btrfs_root *root,
+				 struct inode *inode, u64 start, u64 end,
+				 size_t compressed_size,
+				 struct page **compressed_pages)
+{
+	u64 isize = i_size_read(inode);
+	u64 actual_end = min(end + 1, isize);
+	u64 inline_len = actual_end - start;
+	u64 aligned_end = (end + root->sectorsize - 1) &
+			~((u64)root->sectorsize - 1);
+	u64 hint_byte;
+	u64 data_len = inline_len;
+	int ret;
+
+	if (compressed_size)
+		data_len = compressed_size;
+
+	if (start > 0 ||
+	    data_len >= BTRFS_MAX_INLINE_DATA_SIZE(root) ||
+	    (!compressed_size &&
+	    (actual_end & (root->sectorsize - 1)) == 0) ||
+	    end + 1 < isize ||
+	    data_len > root->fs_info->max_inline) {
+		return 1;
+	}
+
+	mutex_lock(&BTRFS_I(inode)->extent_mutex);
+	ret = btrfs_drop_extents(trans, root, inode, start,
+				 aligned_end, aligned_end, &hint_byte);
+	BUG_ON(ret);
+
+	if (isize > actual_end)
+		inline_len = min_t(u64, isize, actual_end);
+	ret = insert_inline_extent(trans, root, inode, start,
+				   inline_len, compressed_size,
+				   compressed_pages);
+	BUG_ON(ret);
+	btrfs_drop_extent_cache(inode, start, aligned_end, 0);
+	mutex_unlock(&BTRFS_I(inode)->extent_mutex);
+	return 0;
+}
+
 /*
  * when extent_io.c finds a delayed allocation range in the file,
  * the call backs end up in this code.  The basic idea is to
  * allocate extents on disk for the range, and create ordered data structs
  * in ram to track those extents.
+ *
+ * locked_page is the page that writepage had locked already.  We use
+ * it to make sure we don't do extra locks or unlocks.
+ *
+ * *page_started is set to one if we unlock locked_page and do everything
+ * required to start IO on it.  It may be clean and already done with
+ * IO when we return.
  */
-static int cow_file_range(struct inode *inode, u64 start, u64 end)
+static int cow_file_range(struct inode *inode, struct page *locked_page,
+			  u64 start, u64 end, int *page_started)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_trans_handle *trans;
 	u64 alloc_hint = 0;
 	u64 num_bytes;
+	unsigned long ram_size;
+	u64 orig_start;
+	u64 disk_num_bytes;
 	u64 cur_alloc_size;
 	u64 blocksize = root->sectorsize;
-	u64 orig_num_bytes;
+	u64 actual_end;
 	struct btrfs_key ins;
 	struct extent_map *em;
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 	int ret = 0;
+	struct page **pages = NULL;
+	unsigned long nr_pages;
+	unsigned long nr_pages_ret = 0;
+	unsigned long total_compressed = 0;
+	unsigned long total_in = 0;
+	unsigned long max_compressed = 128 * 1024;
+	unsigned long max_uncompressed = 256 * 1024;
+	int i;
+	int will_compress;
 
 	trans = btrfs_join_transaction(root, 1);
 	BUG_ON(!trans);
 	btrfs_set_trans_block_group(trans, inode);
+	orig_start = start;
+
+	/*
+	 * compression made this loop a bit ugly, but the basic idea is to
+	 * compress some pages but keep the total size of the compressed
+	 * extent relatively small.  If compression is off, this goto target
+	 * is never used.
+	 */
+again:
+	will_compress = 0;
+	nr_pages = (end >> PAGE_CACHE_SHIFT) - (start >> PAGE_CACHE_SHIFT) + 1;
+	nr_pages = min(nr_pages, (128 * 1024UL) / PAGE_CACHE_SIZE);
 
+	actual_end = min_t(u64, i_size_read(inode), end + 1);
+	total_compressed = actual_end - start;
+
+	/* we want to make sure that amount of ram required to uncompress
+	 * an extent is reasonable, so we limit the total size in ram
+	 * of a compressed extent to 256k
+	 */
+	total_compressed = min(total_compressed, max_uncompressed);
 	num_bytes = (end - start + blocksize) & ~(blocksize - 1);
 	num_bytes = max(blocksize,  num_bytes);
-	orig_num_bytes = num_bytes;
+	disk_num_bytes = num_bytes;
+	total_in = 0;
+	ret = 0;
 
-	if (alloc_hint == EXTENT_MAP_INLINE)
-		goto out;
+	/* we do compression for mount -o compress and when the
+	 * inode has not been flagged as nocompress
+	 */
+	if (!btrfs_test_flag(inode, NOCOMPRESS) &&
+	    btrfs_test_opt(root, COMPRESS)) {
+		WARN_ON(pages);
+		pages = kmalloc(sizeof(struct page *) * nr_pages, GFP_NOFS);
+
+		/* we want to make sure the amount of IO required to satisfy
+		 * a random read is reasonably small, so we limit the size
+		 * of a compressed extent to 128k
+		 */
+		ret = btrfs_zlib_compress_pages(inode->i_mapping, start,
+						total_compressed, pages,
+						nr_pages, &nr_pages_ret,
+						&total_in,
+						&total_compressed,
+						max_compressed);
+
+		if (!ret) {
+			unsigned long offset = total_compressed &
+				(PAGE_CACHE_SIZE - 1);
+			struct page *page = pages[nr_pages_ret - 1];
+			char *kaddr;
+
+			/* zero the tail end of the last page, we might be
+			 * sending it down to disk
+			 */
+			if (offset) {
+				kaddr = kmap_atomic(page, KM_USER0);
+				memset(kaddr + offset, 0,
+				       PAGE_CACHE_SIZE - offset);
+				kunmap_atomic(kaddr, KM_USER0);
+			}
+			will_compress = 1;
+		}
+	}
+	if (start == 0) {
+		/* lets try to make an inline extent */
+		if (ret || total_in < (end - start + 1)) {
+			/* we didn't compress the entire range, try
+			 * to make an uncompressed inline extent.  This
+			 * is almost sure to fail, but maybe inline sizes
+			 * will get bigger later
+			 */
+			ret = cow_file_range_inline(trans, root, inode,
+						    start, end, 0, NULL);
+		} else {
+			ret = cow_file_range_inline(trans, root, inode,
+						    start, end,
+						    total_compressed, pages);
+		}
+		if (ret == 0) {
+			extent_clear_unlock_delalloc(inode,
+						     &BTRFS_I(inode)->io_tree,
+						     start, end, NULL,
+						     1, 1, 1);
+			*page_started = 1;
+			ret = 0;
+			goto free_pages_out;
+		}
+	}
+
+	if (will_compress) {
+		/*
+		 * we aren't doing an inline extent round the compressed size
+		 * up to a block size boundary so the allocator does sane
+		 * things
+		 */
+		total_compressed = (total_compressed + blocksize - 1) &
+			~(blocksize - 1);
+
+		/*
+		 * one last check to make sure the compression is really a
+		 * win, compare the page count read with the blocks on disk
+		 */
+		total_in = (total_in + PAGE_CACHE_SIZE - 1) &
+			~(PAGE_CACHE_SIZE - 1);
+		if (total_compressed >= total_in) {
+			will_compress = 0;
+		} else {
+			disk_num_bytes = total_compressed;
+			num_bytes = total_in;
+		}
+	}
+	if (!will_compress && pages) {
+		/*
+		 * the compression code ran but failed to make things smaller,
+		 * free any pages it allocated and our page pointer array
+		 */
+		for (i = 0; i < nr_pages_ret; i++) {
+			page_cache_release(pages[i]);
+		}
+		kfree(pages);
+		pages = NULL;
+		total_compressed = 0;
+		nr_pages_ret = 0;
+
+		/* flag the file so we don't compress in the future */
+		btrfs_set_flag(inode, NOCOMPRESS);
+	}
+
+	BUG_ON(disk_num_bytes >
+	       btrfs_super_total_bytes(&root->fs_info->super_copy));
 
-	BUG_ON(num_bytes > btrfs_super_total_bytes(&root->fs_info->super_copy));
 	mutex_lock(&BTRFS_I(inode)->extent_mutex);
 	btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0);
 	mutex_unlock(&BTRFS_I(inode)->extent_mutex);
 
-	while(num_bytes > 0) {
-		cur_alloc_size = min(num_bytes, root->fs_info->max_extent);
+	while(disk_num_bytes > 0) {
+		unsigned long min_bytes;
+
+		/*
+		 * the max size of a compressed extent is pretty small,
+		 * make the code a little less complex by forcing
+		 * the allocator to find a whole compressed extent at once
+		 */
+		if (will_compress)
+			min_bytes = disk_num_bytes;
+		else
+			min_bytes = root->sectorsize;
+
+		cur_alloc_size = min(disk_num_bytes, root->fs_info->max_extent);
 		ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
-					   root->sectorsize, 0, alloc_hint,
+					   min_bytes, 0, alloc_hint,
 					   (u64)-1, &ins, 1);
 		if (ret) {
 			WARN_ON(1);
-			goto out;
+			goto free_pages_out_fail;
 		}
 		em = alloc_extent_map(GFP_NOFS);
 		em->start = start;
-		em->len = ins.offset;
+
+		if (will_compress) {
+			ram_size = num_bytes;
+			em->len = num_bytes;
+		} else {
+			/* ramsize == disk size */
+			ram_size = ins.offset;
+			em->len = ins.offset;
+		}
+
 		em->block_start = ins.objectid;
+		em->block_len = ins.offset;
 		em->bdev = root->fs_info->fs_devices->latest_bdev;
+
 		mutex_lock(&BTRFS_I(inode)->extent_mutex);
 		set_bit(EXTENT_FLAG_PINNED, &em->flags);
+
+		if (will_compress)
+			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
+
 		while(1) {
 			spin_lock(&em_tree->lock);
 			ret = add_extent_mapping(em_tree, em);
@@ -174,26 +493,95 @@ static int cow_file_range(struct inode *inode, u64 start, u64 end)
 				break;
 			}
 			btrfs_drop_extent_cache(inode, start,
-						start + ins.offset - 1, 0);
+						start + ram_size - 1, 0);
 		}
 		mutex_unlock(&BTRFS_I(inode)->extent_mutex);
 
 		cur_alloc_size = ins.offset;
 		ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
-					       ins.offset, 0);
+					       ram_size, cur_alloc_size, 0,
+					       will_compress);
 		BUG_ON(ret);
-		if (num_bytes < cur_alloc_size) {
-			printk("num_bytes %Lu cur_alloc %Lu\n", num_bytes,
+
+		if (disk_num_bytes < cur_alloc_size) {
+			printk("num_bytes %Lu cur_alloc %Lu\n", disk_num_bytes,
 			       cur_alloc_size);
 			break;
 		}
+
+		if (will_compress) {
+			/*
+			 * we're doing compression, we and we need to
+			 * submit the compressed extents down to the device.
+			 *
+			 * We lock down all the file pages, clearing their
+			 * dirty bits and setting them writeback.  Everyone
+			 * that wants to modify the page will wait on the
+			 * ordered extent above.
+			 *
+			 * The writeback bits on the file pages are
+			 * cleared when the compressed pages are on disk
+			 */
+			btrfs_end_transaction(trans, root);
+
+			if (start <= page_offset(locked_page) &&
+			    page_offset(locked_page) < start + ram_size) {
+				*page_started = 1;
+			}
+
+			extent_clear_unlock_delalloc(inode,
+						     &BTRFS_I(inode)->io_tree,
+						     start,
+						     start + ram_size - 1,
+						     NULL, 1, 1, 0);
+
+			ret = btrfs_submit_compressed_write(inode, start,
+						 ram_size, ins.objectid,
+						 cur_alloc_size, pages,
+						 nr_pages_ret);
+
+			BUG_ON(ret);
+			trans = btrfs_join_transaction(root, 1);
+			if (start + ram_size < end) {
+				start += ram_size;
+				alloc_hint = ins.objectid + ins.offset;
+				/* pages will be freed at end_bio time */
+				pages = NULL;
+				goto again;
+			} else {
+				/* we've written everything, time to go */
+				break;
+			}
+		}
+		/* we're not doing compressed IO, don't unlock the first
+		 * page (which the caller expects to stay locked), don't
+		 * clear any dirty bits and don't set any writeback bits
+		 */
+		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
+					     start, start + ram_size - 1,
+					     locked_page, 0, 0, 0);
+		disk_num_bytes -= cur_alloc_size;
 		num_bytes -= cur_alloc_size;
 		alloc_hint = ins.objectid + ins.offset;
 		start += cur_alloc_size;
 	}
+
+	ret = 0;
 out:
 	btrfs_end_transaction(trans, root);
+
 	return ret;
+
+free_pages_out_fail:
+	extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
+				     start, end, locked_page, 0, 0, 0);
+free_pages_out:
+	for (i = 0; i < nr_pages_ret; i++)
+		page_cache_release(pages[i]);
+	if (pages)
+		kfree(pages);
+
+	goto out;
 }
 
 /*
@@ -203,7 +591,8 @@ out:
  * If no cow copies or snapshots exist, we write directly to the existing
  * blocks on disk
  */
-static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end)
+static int run_delalloc_nocow(struct inode *inode, struct page *locked_page,
+			      u64 start, u64 end, int *page_started)
 {
 	u64 extent_start;
 	u64 extent_end;
@@ -260,6 +649,11 @@ again:
 		extent_end = extent_start + extent_num_bytes;
 		err = 0;
 
+		if (btrfs_file_extent_compression(leaf, item) ||
+		    btrfs_file_extent_encryption(leaf,item) ||
+		    btrfs_file_extent_other_encoding(leaf, item))
+			goto not_found;
+
 		if (loops && start != extent_start)
 			goto not_found;
 
@@ -284,7 +678,8 @@ again:
 		bytenr += btrfs_file_extent_offset(leaf, item);
 		extent_num_bytes = min(end + 1, extent_end) - start;
 		ret = btrfs_add_ordered_extent(inode, start, bytenr,
-						extent_num_bytes, 1);
+						extent_num_bytes,
+						extent_num_bytes, 1, 0);
 		if (ret) {
 			err = ret;
 			goto out;
@@ -300,7 +695,8 @@ again:
 not_found:
 		btrfs_end_transaction(trans, root);
 		btrfs_free_path(path);
-		return cow_file_range(inode, start, end);
+		return cow_file_range(inode, locked_page, start, end,
+				      page_started);
 	}
 out:
 	WARN_ON(err);
@@ -312,16 +708,19 @@ out:
 /*
  * extent_io.c call back to do delayed allocation processing
  */
-static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
+static int run_delalloc_range(struct inode *inode, struct page *locked_page,
+			      u64 start, u64 end, int *page_started)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	int ret;
 
 	if (btrfs_test_opt(root, NODATACOW) ||
 	    btrfs_test_flag(inode, NODATACOW))
-		ret = run_delalloc_nocow(inode, start, end);
+		ret = run_delalloc_nocow(inode, locked_page, start, end,
+					 page_started);
 	else
-		ret = cow_file_range(inode, start, end);
+		ret = cow_file_range(inode, locked_page, start, end,
+				     page_started);
 
 	return ret;
 }
@@ -383,7 +782,8 @@ int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end,
  * we don't create bios that span stripes or chunks
  */
 int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
-			 size_t size, struct bio *bio)
+			 size_t size, struct bio *bio,
+			 unsigned long bio_flags)
 {
 	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
 	struct btrfs_mapping_tree *map_tree;
@@ -413,7 +813,7 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
  * are inserted into the btree
  */
 int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
-			  int mirror_num)
+			  int mirror_num, unsigned long bio_flags)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	int ret = 0;
@@ -429,7 +829,7 @@ int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
  * or reading the csums from the tree before a read
  */
 int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
-			  int mirror_num)
+			  int mirror_num, unsigned long bio_flags)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	int ret = 0;
@@ -444,11 +844,17 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
 
 	if (!(rw & (1 << BIO_RW))) {
 		btrfs_lookup_bio_sums(root, inode, bio);
+
+		if (bio_flags & EXTENT_BIO_COMPRESSED) {
+			return btrfs_submit_compressed_read(inode, bio,
+						    mirror_num, bio_flags);
+		}
+
 		goto mapit;
 	}
 	return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
 				   inode, rw, bio, mirror_num,
-				   __btrfs_submit_bio_hook);
+				   bio_flags, __btrfs_submit_bio_hook);
 mapit:
 	return btrfs_map_bio(root, rw, bio, mirror_num, 0);
 }
@@ -539,7 +945,7 @@ out_page:
  * good idea.  This causes problems because we want to make sure COW
  * properly happens and the data=ordered rules are followed.
  *
- * In our case any range that doesn't have the EXTENT_ORDERED bit set
+ * In our case any range that doesn't have the ORDERED bit set
  * hasn't been properly setup for IO.  We kick off an async process
  * to fix it up.  The async helper will wait for ordered extents, set
  * the delalloc bit and make it safe to write the page.
@@ -632,10 +1038,21 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
 	btrfs_set_file_extent_disk_bytenr(leaf, extent_item,
 					  ordered_extent->start);
 	btrfs_set_file_extent_disk_num_bytes(leaf, extent_item,
-					     ordered_extent->len);
+					     ordered_extent->disk_len);
 	btrfs_set_file_extent_offset(leaf, extent_item, 0);
+
+	if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered_extent->flags))
+		btrfs_set_file_extent_compression(leaf, extent_item, 1);
+	else
+		btrfs_set_file_extent_compression(leaf, extent_item, 0);
+	btrfs_set_file_extent_encryption(leaf, extent_item, 0);
+	btrfs_set_file_extent_other_encoding(leaf, extent_item, 0);
+
+	/* ram bytes = extent_num_bytes for now */
 	btrfs_set_file_extent_num_bytes(leaf, extent_item,
 					ordered_extent->len);
+	btrfs_set_file_extent_ram_bytes(leaf, extent_item,
+					ordered_extent->len);
 	btrfs_mark_buffer_dirty(leaf);
 
 	btrfs_drop_extent_cache(inode, ordered_extent->file_offset,
@@ -644,7 +1061,7 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
 	mutex_unlock(&BTRFS_I(inode)->extent_mutex);
 
 	ins.objectid = ordered_extent->start;
-	ins.offset = ordered_extent->len;
+	ins.offset = ordered_extent->disk_len;
 	ins.type = BTRFS_EXTENT_ITEM_KEY;
 	ret = btrfs_alloc_reserved_extent(trans, root, leaf->start,
 					  root->root_key.objectid,
@@ -714,6 +1131,7 @@ int btrfs_io_failed_hook(struct bio *failed_bio,
 	int ret;
 	int rw;
 	u64 logical;
+	unsigned long bio_flags = 0;
 
 	ret = get_state_private(failure_tree, start, &private);
 	if (ret) {
@@ -738,6 +1156,8 @@ int btrfs_io_failed_hook(struct bio *failed_bio,
 		}
 		logical = start - em->start;
 		logical = em->block_start + logical;
+		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
+			bio_flags = EXTENT_BIO_COMPRESSED;
 		failrec->logical = logical;
 		free_extent_map(em);
 		set_extent_bits(failure_tree, start, end, EXTENT_LOCKED |
@@ -781,7 +1201,8 @@ int btrfs_io_failed_hook(struct bio *failed_bio,
 		rw = READ;
 
 	BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio,
-						      failrec->last_mirror);
+						      failrec->last_mirror,
+						      bio_flags);
 	return 0;
 }
 
@@ -1644,10 +2065,8 @@ search_again:
 				item_end +=
 				    btrfs_file_extent_num_bytes(leaf, fi);
 			} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
-				struct btrfs_item *item = btrfs_item_nr(leaf,
-							        path->slots[0]);
 				item_end += btrfs_file_extent_inline_len(leaf,
-									 item);
+									 fi);
 			}
 			item_end--;
 		}
@@ -1715,7 +2134,14 @@ search_again:
 				root_owner = btrfs_header_owner(leaf);
 			}
 		} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
-			if (!del_item) {
+			/*
+			 * we can't truncate inline items that have had
+			 * special encodings
+			 */
+			if (!del_item &&
+			    btrfs_file_extent_compression(leaf, fi) == 0 &&
+			    btrfs_file_extent_encryption(leaf, fi) == 0 &&
+			    btrfs_file_extent_other_encoding(leaf, fi) == 0) {
 				u32 size = new_size - found_key.offset;
 
 				if (root->ref_cows) {
@@ -1926,7 +2352,8 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
 			err = btrfs_insert_file_extent(trans, root,
 						       inode->i_ino,
 						       hole_start, 0, 0,
-						       hole_size, 0);
+						       hole_size, 0, hole_size,
+						       0, 0, 0);
 			btrfs_drop_extent_cache(inode, hole_start,
 						(u64)-1, 0);
 			btrfs_check_file(root, inode);
@@ -2894,11 +3321,50 @@ static int merge_extent_mapping(struct extent_map_tree *em_tree,
 	start_diff = map_start - em->start;
 	em->start = map_start;
 	em->len = map_len;
-	if (em->block_start < EXTENT_MAP_LAST_BYTE)
+	if (em->block_start < EXTENT_MAP_LAST_BYTE &&
+	    !test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
 		em->block_start += start_diff;
+		em->block_len -= start_diff;
+	}
 	return add_extent_mapping(em_tree, em);
 }
 
+static noinline int uncompress_inline(struct btrfs_path *path,
+				      struct inode *inode, struct page *page,
+				      size_t pg_offset, u64 extent_offset,
+				      struct btrfs_file_extent_item *item)
+{
+	int ret;
+	struct extent_buffer *leaf = path->nodes[0];
+	char *tmp;
+	size_t max_size;
+	unsigned long inline_size;
+	unsigned long ptr;
+
+	WARN_ON(pg_offset != 0);
+	max_size = btrfs_file_extent_ram_bytes(leaf, item);
+	inline_size = btrfs_file_extent_inline_item_len(leaf,
+					btrfs_item_nr(leaf, path->slots[0]));
+	tmp = kmalloc(inline_size, GFP_NOFS);
+	ptr = btrfs_file_extent_inline_start(item);
+
+	read_extent_buffer(leaf, tmp, ptr, inline_size);
+
+	max_size = min(PAGE_CACHE_SIZE, max_size);
+	ret = btrfs_zlib_decompress(tmp, page, extent_offset,
+				    inline_size, max_size);
+	if (ret) {
+		char *kaddr = kmap_atomic(page, KM_USER0);
+		unsigned long copy_size = min_t(u64,
+				  PAGE_CACHE_SIZE - pg_offset,
+				  max_size - extent_offset);
+		memset(kaddr + pg_offset, 0, copy_size);
+		kunmap_atomic(kaddr, KM_USER0);
+	}
+	kfree(tmp);
+	return 0;
+}
+
 /*
  * a bit scary, this does extent mapping from logical file offset to the disk.
  * the ugly parts come from merging extents from the disk with the
@@ -2927,6 +3393,7 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 	struct btrfs_trans_handle *trans = NULL;
+	int compressed;
 
 again:
 	spin_lock(&em_tree->lock);
@@ -2951,6 +3418,7 @@ again:
 	em->bdev = root->fs_info->fs_devices->latest_bdev;
 	em->start = EXTENT_MAP_HOLE;
 	em->len = (u64)-1;
+	em->block_len = (u64)-1;
 
 	if (!path) {
 		path = btrfs_alloc_path();
@@ -2983,6 +3451,7 @@ again:
 
 	found_type = btrfs_file_extent_type(leaf, item);
 	extent_start = found_key.offset;
+	compressed = btrfs_file_extent_compression(leaf, item);
 	if (found_type == BTRFS_FILE_EXTENT_REG) {
 		extent_end = extent_start +
 		       btrfs_file_extent_num_bytes(leaf, item);
@@ -3005,10 +3474,18 @@ again:
 			em->block_start = EXTENT_MAP_HOLE;
 			goto insert;
 		}
-		bytenr += btrfs_file_extent_offset(leaf, item);
-		em->block_start = bytenr;
 		em->start = extent_start;
 		em->len = extent_end - extent_start;
+		if (compressed) {
+			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
+			em->block_start = bytenr;
+			em->block_len = btrfs_file_extent_disk_num_bytes(leaf,
+									 item);
+		} else {
+			bytenr += btrfs_file_extent_offset(leaf, item);
+			em->block_start = bytenr;
+			em->block_len = em->len;
+		}
 		goto insert;
 	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
 		u64 page_start;
@@ -3018,8 +3495,7 @@ again:
 		size_t extent_offset;
 		size_t copy_size;
 
-		size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
-						    path->slots[0]));
+		size = btrfs_file_extent_inline_len(leaf, item);
 		extent_end = (extent_start + size + root->sectorsize - 1) &
 			~((u64)root->sectorsize - 1);
 		if (start < extent_start || start >= extent_end) {
@@ -3035,9 +3511,10 @@ again:
 		}
 		em->block_start = EXTENT_MAP_INLINE;
 
-		if (!page) {
+		if (!page || create) {
 			em->start = extent_start;
-			em->len = size;
+			em->len = (size + root->sectorsize - 1) &
+			~((u64)root->sectorsize - 1);
 			goto out;
 		}
 
@@ -3048,11 +3525,22 @@ again:
 		em->start = extent_start + extent_offset;
 		em->len = (copy_size + root->sectorsize - 1) &
 			~((u64)root->sectorsize - 1);
-		map = kmap(page);
+		if (compressed)
+			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
 		ptr = btrfs_file_extent_inline_start(item) + extent_offset;
 		if (create == 0 && !PageUptodate(page)) {
-			read_extent_buffer(leaf, map + pg_offset, ptr,
-					   copy_size);
+			if (btrfs_file_extent_compression(leaf, item) ==
+			    BTRFS_COMPRESS_ZLIB) {
+				ret = uncompress_inline(path, inode, page,
+							pg_offset,
+							extent_offset, item);
+				BUG_ON(ret);
+			} else {
+				map = kmap(page);
+				read_extent_buffer(leaf, map + pg_offset, ptr,
+						   copy_size);
+				kunmap(page);
+			}
 			flush_dcache_page(page);
 		} else if (create && PageUptodate(page)) {
 			if (!trans) {
@@ -3063,11 +3551,12 @@ again:
 				trans = btrfs_join_transaction(root, 1);
 				goto again;
 			}
+			map = kmap(page);
 			write_extent_buffer(leaf, map + pg_offset, ptr,
 					    copy_size);
+			kunmap(page);
 			btrfs_mark_buffer_dirty(leaf);
 		}
-		kunmap(page);
 		set_extent_uptodate(io_tree, em->start,
 				    extent_map_end(em) - 1, GFP_NOFS);
 		goto insert;
@@ -3779,6 +4268,11 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
 	btrfs_set_file_extent_generation(leaf, ei, trans->transid);
 	btrfs_set_file_extent_type(leaf, ei,
 				   BTRFS_FILE_EXTENT_INLINE);
+	btrfs_set_file_extent_encryption(leaf, ei, 0);
+	btrfs_set_file_extent_compression(leaf, ei, 0);
+	btrfs_set_file_extent_other_encoding(leaf, ei, 0);
+	btrfs_set_file_extent_ram_bytes(leaf, ei, name_len);
+
 	ptr = btrfs_file_extent_inline_start(ei);
 	write_extent_buffer(leaf, symname, ptr, name_len);
 	btrfs_mark_buffer_dirty(leaf);
diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c
index 2eb6caba57c2..b5745bb96d40 100644
--- a/fs/btrfs/ordered-data.c
+++ b/fs/btrfs/ordered-data.c
@@ -165,7 +165,8 @@ static inline struct rb_node *tree_search(struct btrfs_ordered_inode_tree *tree,
  * inserted.
  */
 int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset,
-			     u64 start, u64 len, int nocow)
+			     u64 start, u64 len, u64 disk_len, int nocow,
+			     int compressed)
 {
 	struct btrfs_ordered_inode_tree *tree;
 	struct rb_node *node;
@@ -180,9 +181,12 @@ int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset,
 	entry->file_offset = file_offset;
 	entry->start = start;
 	entry->len = len;
+	entry->disk_len = disk_len;
 	entry->inode = inode;
 	if (nocow)
 		set_bit(BTRFS_ORDERED_NOCOW, &entry->flags);
+	if (compressed)
+		set_bit(BTRFS_ORDERED_COMPRESSED, &entry->flags);
 
 	/* one ref for the tree */
 	atomic_set(&entry->refs, 1);
@@ -389,9 +393,10 @@ void btrfs_start_ordered_extent(struct inode *inode,
 	 * for pdflush to find them
 	 */
 	btrfs_fdatawrite_range(inode->i_mapping, start, end, WB_SYNC_NONE);
-	if (wait)
+	if (wait) {
 		wait_event(entry->wait, test_bit(BTRFS_ORDERED_COMPLETE,
 						 &entry->flags));
+	}
 }
 
 /*
diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h
index f50f8870a144..1ef464145d22 100644
--- a/fs/btrfs/ordered-data.h
+++ b/fs/btrfs/ordered-data.h
@@ -66,6 +66,8 @@ struct btrfs_ordered_sum {
 
 #define BTRFS_ORDERED_NOCOW 2 /* set when we want to write in place */
 
+#define BTRFS_ORDERED_COMPRESSED 3 /* writing a compressed extent */
+
 struct btrfs_ordered_extent {
 	/* logical offset in the file */
 	u64 file_offset;
@@ -73,9 +75,12 @@ struct btrfs_ordered_extent {
 	/* disk byte number */
 	u64 start;
 
-	/* length of the extent in bytes */
+	/* ram length of the extent in bytes */
 	u64 len;
 
+	/* extent length on disk */
+	u64 disk_len;
+
 	/* flags (described above) */
 	unsigned long flags;
 
@@ -127,7 +132,8 @@ int btrfs_remove_ordered_extent(struct inode *inode,
 int btrfs_dec_test_ordered_pending(struct inode *inode,
 				       u64 file_offset, u64 io_size);
 int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset,
-			     u64 start, u64 len, int nocow);
+			     u64 start, u64 len, u64 disk_len, int nocow,
+			     int compressed);
 int btrfs_add_ordered_sum(struct inode *inode,
 			  struct btrfs_ordered_extent *entry,
 			  struct btrfs_ordered_sum *sum);
diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c
index bd9ab3e9a7f2..64725c13aa11 100644
--- a/fs/btrfs/print-tree.c
+++ b/fs/btrfs/print-tree.c
@@ -115,15 +115,16 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 			if (btrfs_file_extent_type(l, fi) ==
 			    BTRFS_FILE_EXTENT_INLINE) {
 				printk("\t\tinline extent data size %u\n",
-			           btrfs_file_extent_inline_len(l, item));
+			           btrfs_file_extent_inline_len(l, fi));
 				break;
 			}
 			printk("\t\textent data disk bytenr %llu nr %llu\n",
 			       (unsigned long long)btrfs_file_extent_disk_bytenr(l, fi),
 			       (unsigned long long)btrfs_file_extent_disk_num_bytes(l, fi));
-			printk("\t\textent data offset %llu nr %llu\n",
+			printk("\t\textent data offset %llu nr %llu ram %llu\n",
 			  (unsigned long long)btrfs_file_extent_offset(l, fi),
-			  (unsigned long long)btrfs_file_extent_num_bytes(l, fi));
+			  (unsigned long long)btrfs_file_extent_num_bytes(l, fi),
+			  (unsigned long long)btrfs_file_extent_ram_bytes(l, fi));
 			break;
 		case BTRFS_BLOCK_GROUP_ITEM_KEY:
 			bi = btrfs_item_ptr(l, i,
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 2e6039825b7b..431fdf144b58 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -47,6 +47,7 @@
 #include "volumes.h"
 #include "version.h"
 #include "export.h"
+#include "compression.h"
 
 #define BTRFS_SUPER_MAGIC 0x9123683E
 
@@ -69,7 +70,7 @@ static void btrfs_put_super (struct super_block * sb)
 enum {
 	Opt_degraded, Opt_subvol, Opt_device, Opt_nodatasum, Opt_nodatacow,
 	Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier,
-	Opt_ssd, Opt_thread_pool, Opt_noacl,  Opt_err,
+	Opt_ssd, Opt_thread_pool, Opt_noacl,  Opt_compress, Opt_err,
 };
 
 static match_table_t tokens = {
@@ -83,6 +84,7 @@ static match_table_t tokens = {
 	{Opt_max_inline, "max_inline=%s"},
 	{Opt_alloc_start, "alloc_start=%s"},
 	{Opt_thread_pool, "thread_pool=%d"},
+	{Opt_compress, "compress"},
 	{Opt_ssd, "ssd"},
 	{Opt_noacl, "noacl"},
 	{Opt_err, NULL},
@@ -163,6 +165,10 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 			btrfs_set_opt(info->mount_opt, NODATACOW);
 			btrfs_set_opt(info->mount_opt, NODATASUM);
 			break;
+		case Opt_compress:
+			printk(KERN_INFO "btrfs: use compression\n");
+			btrfs_set_opt(info->mount_opt, COMPRESS);
+			break;
 		case Opt_ssd:
 			printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
 			btrfs_set_opt(info->mount_opt, SSD);
@@ -622,6 +628,7 @@ static int __init init_btrfs_fs(void)
 	err = btrfs_interface_init();
 	if (err)
 		goto free_extent_map;
+
 	err = register_filesystem(&btrfs_fs_type);
 	if (err)
 		goto unregister_ioctl;
@@ -651,6 +658,7 @@ static void __exit exit_btrfs_fs(void)
 	unregister_filesystem(&btrfs_fs_type);
 	btrfs_exit_sysfs();
 	btrfs_cleanup_fs_uuids();
+	btrfs_zlib_exit();
 }
 
 module_init(init_btrfs_fs)
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index cf618cc8b34a..e6d579053a47 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -540,8 +540,7 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
 	if (found_type == BTRFS_FILE_EXTENT_REG)
 		extent_end = start + btrfs_file_extent_num_bytes(eb, item);
 	else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
-		size = btrfs_file_extent_inline_len(eb,
-						    btrfs_item_nr(eb, slot));
+		size = btrfs_file_extent_inline_len(eb, item);
 		extent_end = (start + size + mask) & ~mask;
 	} else {
 		ret = 0;
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 2eed7f91f51a..7db4cfd03a98 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -1816,6 +1816,7 @@ again:
 	em->start = key.offset;
 	em->len = *num_bytes;
 	em->block_start = 0;
+	em->block_len = em->len;
 
 	if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
 		ret = btrfs_add_system_chunk(trans, chunk_root, &key,
@@ -2323,6 +2324,7 @@ static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
 	em->start = logical;
 	em->len = length;
 	em->block_start = 0;
+	em->block_len = em->len;
 
 	map->num_stripes = num_stripes;
 	map->io_width = btrfs_chunk_io_width(leaf, chunk);
diff --git a/fs/btrfs/zlib.c b/fs/btrfs/zlib.c
new file mode 100644
index 000000000000..e99309180a11
--- /dev/null
+++ b/fs/btrfs/zlib.c
@@ -0,0 +1,637 @@
+/*
+ * Copyright (C) 2008 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ * Based on jffs2 zlib code:
+ * Copyright © 2001-2007 Red Hat, Inc.
+ * Created by David Woodhouse <dwmw2@infradead.org>
+ */
+
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/zlib.h>
+#include <linux/zutil.h>
+#include <linux/vmalloc.h>
+#include <linux/init.h>
+#include <linux/err.h>
+#include <linux/sched.h>
+#include <linux/pagemap.h>
+#include <linux/bio.h>
+
+/* Plan: call deflate() with avail_in == *sourcelen,
+	avail_out = *dstlen - 12 and flush == Z_FINISH.
+	If it doesn't manage to finish,	call it again with
+	avail_in == 0 and avail_out set to the remaining 12
+	bytes for it to clean up.
+   Q: Is 12 bytes sufficient?
+*/
+#define STREAM_END_SPACE 12
+
+struct workspace {
+	z_stream inf_strm;
+	z_stream def_strm;
+	char *buf;
+	struct list_head list;
+};
+
+static LIST_HEAD(idle_workspace);
+static DEFINE_SPINLOCK(workspace_lock);
+static unsigned long num_workspace;
+static atomic_t alloc_workspace = ATOMIC_INIT(0);
+static DECLARE_WAIT_QUEUE_HEAD(workspace_wait);
+
+/*
+ * this finds an available zlib workspace or allocates a new one
+ * NULL or an ERR_PTR is returned if things go bad.
+ */
+static struct workspace *find_zlib_workspace(void)
+{
+	struct workspace *workspace;
+	int ret;
+	int cpus = num_online_cpus();
+
+again:
+	spin_lock(&workspace_lock);
+	if (!list_empty(&idle_workspace)) {
+		workspace = list_entry(idle_workspace.next, struct workspace,
+				       list);
+		list_del(&workspace->list);
+		num_workspace--;
+		spin_unlock(&workspace_lock);
+		return workspace;
+
+	}
+	spin_unlock(&workspace_lock);
+	if (atomic_read(&alloc_workspace) > cpus) {
+		DEFINE_WAIT(wait);
+		prepare_to_wait(&workspace_wait, &wait, TASK_UNINTERRUPTIBLE);
+		if (atomic_read(&alloc_workspace) > cpus)
+			schedule();
+		finish_wait(&workspace_wait, &wait);
+		goto again;
+	}
+	atomic_inc(&alloc_workspace);
+	workspace = kzalloc(sizeof(*workspace), GFP_NOFS);
+	if (!workspace) {
+		ret = -ENOMEM;
+		goto fail;
+	}
+
+	workspace->def_strm.workspace = vmalloc(zlib_deflate_workspacesize());
+	if (!workspace->def_strm.workspace) {
+		ret = -ENOMEM;
+		goto fail;
+	}
+	workspace->inf_strm.workspace = vmalloc(zlib_inflate_workspacesize());
+	if (!workspace->inf_strm.workspace) {
+		ret = -ENOMEM;
+		goto fail_inflate;
+	}
+	workspace->buf = kmalloc(PAGE_CACHE_SIZE, GFP_NOFS);
+	if (!workspace->buf) {
+		ret = -ENOMEM;
+		goto fail_kmalloc;
+	}
+	return workspace;
+
+fail_kmalloc:
+	vfree(workspace->inf_strm.workspace);
+fail_inflate:
+	vfree(workspace->def_strm.workspace);
+fail:
+	kfree(workspace);
+	atomic_dec(&alloc_workspace);
+	wake_up(&workspace_wait);
+	return ERR_PTR(ret);
+}
+
+/*
+ * put a workspace struct back on the list or free it if we have enough
+ * idle ones sitting around
+ */
+static int free_workspace(struct workspace *workspace)
+{
+	spin_lock(&workspace_lock);
+	if (num_workspace < num_online_cpus()) {
+		list_add_tail(&workspace->list, &idle_workspace);
+		num_workspace++;
+		spin_unlock(&workspace_lock);
+		if (waitqueue_active(&workspace_wait))
+			wake_up(&workspace_wait);
+		return 0;
+	}
+	spin_unlock(&workspace_lock);
+	vfree(workspace->def_strm.workspace);
+	vfree(workspace->inf_strm.workspace);
+	kfree(workspace->buf);
+	kfree(workspace);
+
+	atomic_dec(&alloc_workspace);
+	if (waitqueue_active(&workspace_wait))
+		wake_up(&workspace_wait);
+	return 0;
+}
+
+/*
+ * cleanup function for module exit
+ */
+static void free_workspaces(void)
+{
+	struct workspace *workspace;
+	while(!list_empty(&idle_workspace)) {
+		workspace = list_entry(idle_workspace.next, struct workspace,
+				       list);
+		list_del(&workspace->list);
+		vfree(workspace->def_strm.workspace);
+		vfree(workspace->inf_strm.workspace);
+		kfree(workspace->buf);
+		kfree(workspace);
+		atomic_dec(&alloc_workspace);
+	}
+}
+
+/*
+ * given an address space and start/len, compress the bytes.
+ *
+ * pages are allocated to hold the compressed result and stored
+ * in 'pages'
+ *
+ * out_pages is used to return the number of pages allocated.  There
+ * may be pages allocated even if we return an error
+ *
+ * total_in is used to return the number of bytes actually read.  It
+ * may be smaller then len if we had to exit early because we
+ * ran out of room in the pages array or because we cross the
+ * max_out threshold.
+ *
+ * total_out is used to return the total number of compressed bytes
+ *
+ * max_out tells us the max number of bytes that we're allowed to
+ * stuff into pages
+ */
+int btrfs_zlib_compress_pages(struct address_space *mapping,
+			      u64 start, unsigned long len,
+			      struct page **pages,
+			      unsigned long nr_dest_pages,
+			      unsigned long *out_pages,
+			      unsigned long *total_in,
+			      unsigned long *total_out,
+			      unsigned long max_out)
+{
+	int ret;
+	struct workspace *workspace;
+	char *data_in;
+	char *cpage_out;
+	int nr_pages = 0;
+	struct page *in_page = NULL;
+	struct page *out_page = NULL;
+	int out_written = 0;
+	int in_read = 0;
+	unsigned long bytes_left;
+
+	*out_pages = 0;
+	*total_out = 0;
+	*total_in = 0;
+
+	workspace = find_zlib_workspace();
+	if (!workspace)
+		return -1;
+
+	if (Z_OK != zlib_deflateInit(&workspace->def_strm, 3)) {
+		printk(KERN_WARNING "deflateInit failed\n");
+		ret = -1;
+		goto out;
+	}
+
+	workspace->def_strm.total_in = 0;
+	workspace->def_strm.total_out = 0;
+
+	in_page = find_get_page(mapping, start >> PAGE_CACHE_SHIFT);
+	data_in = kmap(in_page);
+
+	out_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
+	cpage_out = kmap(out_page);
+	pages[0] = out_page;
+	nr_pages = 1;
+
+	workspace->def_strm.next_in = data_in;
+	workspace->def_strm.next_out = cpage_out;
+	workspace->def_strm.avail_out = PAGE_CACHE_SIZE;
+	workspace->def_strm.avail_in = min(len, PAGE_CACHE_SIZE);
+
+	out_written = 0;
+	in_read = 0;
+
+	while (workspace->def_strm.total_in < len) {
+		ret = zlib_deflate(&workspace->def_strm, Z_SYNC_FLUSH);
+		if (ret != Z_OK) {
+			printk(KERN_DEBUG "btrfs deflate in loop returned %d\n",
+			       ret);
+			zlib_deflateEnd(&workspace->def_strm);
+			ret = -1;
+			goto out;
+		}
+
+		/* we're making it bigger, give up */
+		if (workspace->def_strm.total_in > 8192 &&
+		    workspace->def_strm.total_in <
+		    workspace->def_strm.total_out) {
+			ret = -1;
+			goto out;
+		}
+		/* we need another page for writing out.  Test this
+		 * before the total_in so we will pull in a new page for
+		 * the stream end if required
+		 */
+		if (workspace->def_strm.avail_out == 0) {
+			kunmap(out_page);
+			if (nr_pages == nr_dest_pages) {
+				out_page = NULL;
+				ret = -1;
+				goto out;
+			}
+			out_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
+			cpage_out = kmap(out_page);
+			pages[nr_pages] = out_page;
+			nr_pages++;
+			workspace->def_strm.avail_out = PAGE_CACHE_SIZE;
+			workspace->def_strm.next_out = cpage_out;
+		}
+		/* we're all done */
+		if (workspace->def_strm.total_in >= len)
+			break;
+
+		/* we've read in a full page, get a new one */
+		if (workspace->def_strm.avail_in == 0) {
+			if (workspace->def_strm.total_out > max_out)
+				break;
+
+			bytes_left = len - workspace->def_strm.total_in;
+			kunmap(in_page);
+			page_cache_release(in_page);
+
+			start += PAGE_CACHE_SIZE;
+			in_page = find_get_page(mapping,
+						start >> PAGE_CACHE_SHIFT);
+			data_in = kmap(in_page);
+			workspace->def_strm.avail_in = min(bytes_left,
+							   PAGE_CACHE_SIZE);
+			workspace->def_strm.next_in = data_in;
+		}
+	}
+	workspace->def_strm.avail_in = 0;
+	ret = zlib_deflate(&workspace->def_strm, Z_FINISH);
+	zlib_deflateEnd(&workspace->def_strm);
+
+	if (ret != Z_STREAM_END) {
+		ret = -1;
+		goto out;
+	}
+
+	if (workspace->def_strm.total_out >= workspace->def_strm.total_in) {
+		ret = -1;
+		goto out;
+	}
+
+	ret = 0;
+	*total_out = workspace->def_strm.total_out;
+	*total_in = workspace->def_strm.total_in;
+out:
+	*out_pages = nr_pages;
+	if (out_page)
+		kunmap(out_page);
+
+	if (in_page) {
+		kunmap(in_page);
+		page_cache_release(in_page);
+	}
+	free_workspace(workspace);
+	return ret;
+}
+
+/*
+ * pages_in is an array of pages with compressed data.
+ *
+ * disk_start is the starting logical offset of this array in the file
+ *
+ * bvec is a bio_vec of pages from the file that we want to decompress into
+ *
+ * vcnt is the count of pages in the biovec
+ *
+ * srclen is the number of bytes in pages_in
+ *
+ * The basic idea is that we have a bio that was created by readpages.
+ * The pages in the bio are for the uncompressed data, and they may not
+ * be contiguous.  They all correspond to the range of bytes covered by
+ * the compressed extent.
+ */
+int btrfs_zlib_decompress_biovec(struct page **pages_in,
+			      u64 disk_start,
+			      struct bio_vec *bvec,
+			      int vcnt,
+			      size_t srclen)
+{
+	int ret = 0;
+	int wbits = MAX_WBITS;
+	struct workspace *workspace;
+	char *data_in;
+	size_t total_out = 0;
+	unsigned long page_bytes_left;
+	unsigned long page_in_index = 0;
+	unsigned long page_out_index = 0;
+	struct page *page_out;
+	unsigned long total_pages_in = (srclen + PAGE_CACHE_SIZE - 1) /
+					PAGE_CACHE_SIZE;
+	unsigned long buf_start;
+	unsigned long buf_offset;
+	unsigned long bytes;
+	unsigned long working_bytes;
+	unsigned long pg_offset;
+	unsigned long start_byte;
+	unsigned long current_buf_start;
+	char *kaddr;
+
+	workspace = find_zlib_workspace();
+	if (!workspace)
+		return -ENOMEM;
+
+	data_in = kmap(pages_in[page_in_index]);
+	workspace->inf_strm.next_in = data_in;
+	workspace->inf_strm.avail_in = min(srclen, PAGE_CACHE_SIZE);
+	workspace->inf_strm.total_in = 0;
+
+	workspace->inf_strm.total_out = 0;
+	workspace->inf_strm.next_out = workspace->buf;
+	workspace->inf_strm.avail_out = PAGE_CACHE_SIZE;
+	page_out = bvec[page_out_index].bv_page;
+	page_bytes_left = PAGE_CACHE_SIZE;
+	pg_offset = 0;
+
+	/* If it's deflate, and it's got no preset dictionary, then
+	   we can tell zlib to skip the adler32 check. */
+	if (srclen > 2 && !(data_in[1] & PRESET_DICT) &&
+	    ((data_in[0] & 0x0f) == Z_DEFLATED) &&
+	    !(((data_in[0]<<8) + data_in[1]) % 31)) {
+
+		wbits = -((data_in[0] >> 4) + 8);
+		workspace->inf_strm.next_in += 2;
+		workspace->inf_strm.avail_in -= 2;
+	}
+
+	if (Z_OK != zlib_inflateInit2(&workspace->inf_strm, wbits)) {
+		printk(KERN_WARNING "inflateInit failed\n");
+		ret = -1;
+		goto out;
+	}
+	while(workspace->inf_strm.total_in < srclen) {
+		ret = zlib_inflate(&workspace->inf_strm, Z_NO_FLUSH);
+		if (ret != Z_OK && ret != Z_STREAM_END) {
+			break;
+		}
+
+		/*
+		 * buf start is the byte offset we're of the start of
+		 * our workspace buffer
+		 */
+		buf_start = total_out;
+
+		/* total_out is the last byte of the workspace buffer */
+		total_out = workspace->inf_strm.total_out;
+
+		working_bytes = total_out - buf_start;
+
+		/*
+		 * start byte is the first byte of the page we're currently
+		 * copying into relative to the start of the compressed data.
+		 */
+		start_byte = page_offset(page_out) - disk_start;
+
+		if (working_bytes == 0) {
+			/* we didn't make progress in this inflate
+			 * call, we're done
+			 */
+			if (ret != Z_STREAM_END)
+				ret = -1;
+			break;
+		}
+
+		/* we haven't yet hit data corresponding to this page */
+		if (total_out <= start_byte) {
+			goto next;
+		}
+
+		/*
+		 * the start of the data we care about is offset into
+		 * the middle of our working buffer
+		 */
+		if (total_out > start_byte && buf_start < start_byte) {
+			buf_offset = start_byte - buf_start;
+			working_bytes -= buf_offset;
+		} else {
+			buf_offset = 0;
+		}
+		current_buf_start = buf_start;
+
+		/* copy bytes from the working buffer into the pages */
+		while(working_bytes > 0) {
+			bytes = min(PAGE_CACHE_SIZE - pg_offset,
+				    PAGE_CACHE_SIZE - buf_offset);
+			bytes = min(bytes, working_bytes);
+			kaddr = kmap_atomic(page_out, KM_USER0);
+			memcpy(kaddr + pg_offset, workspace->buf + buf_offset,
+			       bytes);
+			kunmap_atomic(kaddr, KM_USER0);
+			flush_dcache_page(page_out);
+
+			pg_offset += bytes;
+			page_bytes_left -= bytes;
+			buf_offset += bytes;
+			working_bytes -= bytes;
+			current_buf_start += bytes;
+
+			/* check if we need to pick another page */
+			if (page_bytes_left == 0) {
+				page_out_index++;
+				if (page_out_index >= vcnt) {
+					ret = 0;
+					goto done;
+				}
+				page_out = bvec[page_out_index].bv_page;
+				pg_offset = 0;
+				page_bytes_left = PAGE_CACHE_SIZE;
+				start_byte = page_offset(page_out) - disk_start;
+
+				/*
+				 * make sure our new page is covered by this
+				 * working buffer
+				 */
+				if (total_out <= start_byte) {
+					goto next;
+				}
+
+				/* the next page in the biovec might not
+				 * be adjacent to the last page, but it
+				 * might still be found inside this working
+				 * buffer.  bump our offset pointer
+				 */
+				if (total_out > start_byte &&
+				    current_buf_start < start_byte) {
+					buf_offset = start_byte - buf_start;
+					working_bytes = total_out - start_byte;
+					current_buf_start = buf_start +
+						buf_offset;
+				}
+			}
+		}
+next:
+		workspace->inf_strm.next_out = workspace->buf;
+		workspace->inf_strm.avail_out = PAGE_CACHE_SIZE;
+
+		if (workspace->inf_strm.avail_in == 0) {
+			unsigned long tmp;
+			kunmap(pages_in[page_in_index]);
+			page_in_index++;
+			if (page_in_index >= total_pages_in) {
+				data_in = NULL;
+				break;
+			}
+			data_in = kmap(pages_in[page_in_index]);
+			workspace->inf_strm.next_in = data_in;
+			tmp = srclen - workspace->inf_strm.total_in;
+			workspace->inf_strm.avail_in = min(tmp,
+							   PAGE_CACHE_SIZE);
+		}
+	}
+	if (ret != Z_STREAM_END) {
+		ret = -1;
+	} else {
+		ret = 0;
+	}
+done:
+	zlib_inflateEnd(&workspace->inf_strm);
+	if (data_in)
+		kunmap(pages_in[page_in_index]);
+out:
+	free_workspace(workspace);
+	return ret;
+}
+
+/*
+ * a less complex decompression routine.  Our compressed data fits in a
+ * single page, and we want to read a single page out of it.
+ * start_byte tells us the offset into the compressed data we're interested in
+ */
+int btrfs_zlib_decompress(unsigned char *data_in,
+			  struct page *dest_page,
+			  unsigned long start_byte,
+			  size_t srclen, size_t destlen)
+{
+	int ret = 0;
+	int wbits = MAX_WBITS;
+	struct workspace *workspace;
+	unsigned long bytes_left = destlen;
+	unsigned long total_out = 0;
+	char *kaddr;
+
+	if (destlen > PAGE_CACHE_SIZE)
+		return -ENOMEM;
+
+	workspace = find_zlib_workspace();
+	if (!workspace)
+		return -ENOMEM;
+
+	workspace->inf_strm.next_in = data_in;
+	workspace->inf_strm.avail_in = srclen;
+	workspace->inf_strm.total_in = 0;
+
+	workspace->inf_strm.next_out = workspace->buf;
+	workspace->inf_strm.avail_out = PAGE_CACHE_SIZE;
+	workspace->inf_strm.total_out = 0;
+	/* If it's deflate, and it's got no preset dictionary, then
+	   we can tell zlib to skip the adler32 check. */
+	if (srclen > 2 && !(data_in[1] & PRESET_DICT) &&
+	    ((data_in[0] & 0x0f) == Z_DEFLATED) &&
+	    !(((data_in[0]<<8) + data_in[1]) % 31)) {
+
+		wbits = -((data_in[0] >> 4) + 8);
+		workspace->inf_strm.next_in += 2;
+		workspace->inf_strm.avail_in -= 2;
+	}
+
+	if (Z_OK != zlib_inflateInit2(&workspace->inf_strm, wbits)) {
+		printk(KERN_WARNING "inflateInit failed\n");
+		ret = -1;
+		goto out;
+	}
+
+	while(bytes_left > 0) {
+		unsigned long buf_start;
+		unsigned long buf_offset;
+		unsigned long bytes;
+		unsigned long pg_offset = 0;
+
+		ret = zlib_inflate(&workspace->inf_strm, Z_NO_FLUSH);
+		if (ret != Z_OK && ret != Z_STREAM_END) {
+			break;
+		}
+
+		buf_start = total_out;
+		total_out = workspace->inf_strm.total_out;
+
+		if (total_out == buf_start) {
+			ret = -1;
+			break;
+		}
+
+		if (total_out <= start_byte) {
+			goto next;
+		}
+
+		if (total_out > start_byte && buf_start < start_byte) {
+			buf_offset = start_byte - buf_start;
+		} else {
+			buf_offset = 0;
+		}
+
+		bytes = min(PAGE_CACHE_SIZE - pg_offset,
+			    PAGE_CACHE_SIZE - buf_offset);
+		bytes = min(bytes, bytes_left);
+
+		kaddr = kmap_atomic(dest_page, KM_USER0);
+		memcpy(kaddr + pg_offset, workspace->buf + buf_offset, bytes);
+		kunmap_atomic(kaddr, KM_USER0);
+
+		pg_offset += bytes;
+		bytes_left -= bytes;
+next:
+		workspace->inf_strm.next_out = workspace->buf;
+		workspace->inf_strm.avail_out = PAGE_CACHE_SIZE;
+	}
+	if (ret != Z_STREAM_END && bytes_left != 0) {
+		ret = -1;
+	} else {
+		ret = 0;
+	}
+	zlib_inflateEnd(&workspace->inf_strm);
+out:
+	free_workspace(workspace);
+	return ret;
+}
+
+void btrfs_zlib_exit(void)
+{
+    free_workspaces();
+}
-- 
cgit v1.2.3


From 4a69a41009c4ac691f7d9c289f5f37fabeddce46 Mon Sep 17 00:00:00 2001
From: Chris Mason <chris.mason@oracle.com>
Date: Thu, 6 Nov 2008 22:03:00 -0500
Subject: Btrfs: Add ordered async work queues

Btrfs uses kernel threads to create async work queues for cpu intensive
operations such as checksumming and decompression.  These work well,
but they make it difficult to keep IO order intact.

A single writepages call from pdflush or fsync will turn into a number
of bios, and each bio is checksummed in parallel.  Once the checksum is
computed, the bio is sent down to the disk, and since we don't control
the order in which the parallel operations happen, they might go down to
the disk in almost any order.

The code deals with this somewhat by having deep work queues for a single
kernel thread, making it very likely that a single thread will process all
the bios for a single inode.

This patch introduces an explicitly ordered work queue.  As work structs
are placed into the queue they are put onto the tail of a list.  They have
three callbacks:

->func (cpu intensive processing here)
->ordered_func (order sensitive processing here)
->ordered_free (free the work struct, all processing is done)

The work struct has three callbacks.  The func callback does the cpu intensive
work, and when it completes the work struct is marked as done.

Every time a work struct completes, the list is checked to see if the head
is marked as done.  If so the ordered_func callback is used to do the
order sensitive processing and the ordered_free callback is used to do
any cleanup.  Then we loop back and check the head of the list again.

This patch also changes the checksumming code to use the ordered workqueues.
One a 4 drive array, it increases streaming writes from 280MB/s to 350MB/s.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
---
 fs/btrfs/async-thread.c | 67 ++++++++++++++++++++++++++++++++++++++--
 fs/btrfs/async-thread.h | 18 ++++++++++-
 fs/btrfs/disk-io.c      | 81 ++++++++++++++++++++++++++++++++++---------------
 fs/btrfs/disk-io.h      |  4 ++-
 fs/btrfs/inode.c        | 19 ++++++++++--
 5 files changed, 158 insertions(+), 31 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/async-thread.c b/fs/btrfs/async-thread.c
index d82efd722a48..e1e49715459e 100644
--- a/fs/btrfs/async-thread.c
+++ b/fs/btrfs/async-thread.c
@@ -23,6 +23,10 @@
 # include <linux/freezer.h>
 #include "async-thread.h"
 
+#define WORK_QUEUED_BIT 0
+#define WORK_DONE_BIT 1
+#define WORK_ORDER_DONE_BIT 2
+
 /*
  * container for the kthread task pointer and the list of pending work
  * One of these is allocated per thread.
@@ -88,6 +92,47 @@ static void check_busy_worker(struct btrfs_worker_thread *worker)
 	}
 }
 
+static noinline int run_ordered_completions(struct btrfs_workers *workers,
+					    struct btrfs_work *work)
+{
+	unsigned long flags;
+
+	if (!workers->ordered)
+		return 0;
+
+	set_bit(WORK_DONE_BIT, &work->flags);
+
+	spin_lock_irqsave(&workers->lock, flags);
+
+	while(!list_empty(&workers->order_list)) {
+		work = list_entry(workers->order_list.next,
+				  struct btrfs_work, order_list);
+
+		if (!test_bit(WORK_DONE_BIT, &work->flags))
+			break;
+
+		/* we are going to call the ordered done function, but
+		 * we leave the work item on the list as a barrier so
+		 * that later work items that are done don't have their
+		 * functions called before this one returns
+		 */
+		if (test_and_set_bit(WORK_ORDER_DONE_BIT, &work->flags))
+			break;
+
+		spin_unlock_irqrestore(&workers->lock, flags);
+
+		work->ordered_func(work);
+
+		/* now take the lock again and call the freeing code */
+		spin_lock_irqsave(&workers->lock, flags);
+		list_del(&work->order_list);
+		work->ordered_free(work);
+	}
+
+	spin_unlock_irqrestore(&workers->lock, flags);
+	return 0;
+}
+
 /*
  * main loop for servicing work items
  */
@@ -102,7 +147,7 @@ static int worker_loop(void *arg)
 			cur = worker->pending.next;
 			work = list_entry(cur, struct btrfs_work, list);
 			list_del(&work->list);
-			clear_bit(0, &work->flags);
+			clear_bit(WORK_QUEUED_BIT, &work->flags);
 
 			work->worker = worker;
 			spin_unlock_irq(&worker->lock);
@@ -110,8 +155,15 @@ static int worker_loop(void *arg)
 			work->func(work);
 
 			atomic_dec(&worker->num_pending);
+			/*
+			 * unless this is an ordered work queue,
+			 * 'work' was probably freed by func above.
+			 */
+			run_ordered_completions(worker->workers, work);
+
 			spin_lock_irq(&worker->lock);
 			check_idle_worker(worker);
+
 		}
 		worker->working = 0;
 		if (freezing(current)) {
@@ -154,10 +206,12 @@ void btrfs_init_workers(struct btrfs_workers *workers, char *name, int max)
 	workers->num_workers = 0;
 	INIT_LIST_HEAD(&workers->worker_list);
 	INIT_LIST_HEAD(&workers->idle_list);
+	INIT_LIST_HEAD(&workers->order_list);
 	spin_lock_init(&workers->lock);
 	workers->max_workers = max;
 	workers->idle_thresh = 32;
 	workers->name = name;
+	workers->ordered = 0;
 }
 
 /*
@@ -296,7 +350,7 @@ int btrfs_requeue_work(struct btrfs_work *work)
 	struct btrfs_worker_thread *worker = work->worker;
 	unsigned long flags;
 
-	if (test_and_set_bit(0, &work->flags))
+	if (test_and_set_bit(WORK_QUEUED_BIT, &work->flags))
 		goto out;
 
 	spin_lock_irqsave(&worker->lock, flags);
@@ -330,10 +384,17 @@ int btrfs_queue_worker(struct btrfs_workers *workers, struct btrfs_work *work)
 	int wake = 0;
 
 	/* don't requeue something already on a list */
-	if (test_and_set_bit(0, &work->flags))
+	if (test_and_set_bit(WORK_QUEUED_BIT, &work->flags))
 		goto out;
 
 	worker = find_worker(workers);
+	if (workers->ordered) {
+		spin_lock_irqsave(&workers->lock, flags);
+		list_add_tail(&work->order_list, &workers->order_list);
+		spin_unlock_irqrestore(&workers->lock, flags);
+	} else {
+		INIT_LIST_HEAD(&work->order_list);
+	}
 
 	spin_lock_irqsave(&worker->lock, flags);
 	atomic_inc(&worker->num_pending);
diff --git a/fs/btrfs/async-thread.h b/fs/btrfs/async-thread.h
index 4ec9a2ee0f9d..31be4ed8b63e 100644
--- a/fs/btrfs/async-thread.h
+++ b/fs/btrfs/async-thread.h
@@ -37,10 +37,16 @@ struct btrfs_worker_thread;
  */
 struct btrfs_work {
 	/*
-	 * only func should be set to the function you want called
+	 * func should be set to the function you want called
 	 * your work struct is passed as the only arg
+	 *
+	 * ordered_func must be set for work sent to an ordered work queue,
+	 * and it is called to complete a given work item in the same
+	 * order they were sent to the queue.
 	 */
 	void (*func)(struct btrfs_work *work);
+	void (*ordered_func)(struct btrfs_work *work);
+	void (*ordered_free)(struct btrfs_work *work);
 
 	/*
 	 * flags should be set to zero.  It is used to make sure the
@@ -51,6 +57,7 @@ struct btrfs_work {
 	/* don't touch these */
 	struct btrfs_worker_thread *worker;
 	struct list_head list;
+	struct list_head order_list;
 };
 
 struct btrfs_workers {
@@ -63,6 +70,9 @@ struct btrfs_workers {
 	/* once a worker has this many requests or fewer, it is idle */
 	int idle_thresh;
 
+	/* force completions in the order they were queued */
+	int ordered;
+
 	/* list with all the work threads.  The workers on the idle thread
 	 * may be actively servicing jobs, but they haven't yet hit the
 	 * idle thresh limit above.
@@ -70,6 +80,12 @@ struct btrfs_workers {
 	struct list_head worker_list;
 	struct list_head idle_list;
 
+	/*
+	 * when operating in ordered mode, this maintains the list
+	 * of work items waiting for completion
+	 */
+	struct list_head order_list;
+
 	/* lock for finding the next worker thread to queue on */
 	spinlock_t lock;
 
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 94b4e50f6b2c..e0a28f705a64 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -80,7 +80,8 @@ struct async_submit_bio {
 	struct inode *inode;
 	struct bio *bio;
 	struct list_head list;
-	extent_submit_bio_hook_t *submit_bio_hook;
+	extent_submit_bio_hook_t *submit_bio_start;
+	extent_submit_bio_hook_t *submit_bio_done;
 	int rw;
 	int mirror_num;
 	unsigned long bio_flags;
@@ -452,7 +453,18 @@ int btrfs_congested_async(struct btrfs_fs_info *info, int iodone)
 		btrfs_async_submit_limit(info);
 }
 
-static void run_one_async_submit(struct btrfs_work *work)
+static void run_one_async_start(struct btrfs_work *work)
+{
+	struct btrfs_fs_info *fs_info;
+	struct async_submit_bio *async;
+
+	async = container_of(work, struct  async_submit_bio, work);
+	fs_info = BTRFS_I(async->inode)->root->fs_info;
+	async->submit_bio_start(async->inode, async->rw, async->bio,
+			       async->mirror_num, async->bio_flags);
+}
+
+static void run_one_async_done(struct btrfs_work *work)
 {
 	struct btrfs_fs_info *fs_info;
 	struct async_submit_bio *async;
@@ -470,15 +482,23 @@ static void run_one_async_submit(struct btrfs_work *work)
 	    waitqueue_active(&fs_info->async_submit_wait))
 		wake_up(&fs_info->async_submit_wait);
 
-	async->submit_bio_hook(async->inode, async->rw, async->bio,
+	async->submit_bio_done(async->inode, async->rw, async->bio,
 			       async->mirror_num, async->bio_flags);
+}
+
+static void run_one_async_free(struct btrfs_work *work)
+{
+	struct async_submit_bio *async;
+
+	async = container_of(work, struct  async_submit_bio, work);
 	kfree(async);
 }
 
 int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
 			int rw, struct bio *bio, int mirror_num,
 			unsigned long bio_flags,
-			extent_submit_bio_hook_t *submit_bio_hook)
+			extent_submit_bio_hook_t *submit_bio_start,
+			extent_submit_bio_hook_t *submit_bio_done)
 {
 	struct async_submit_bio *async;
 	int limit = btrfs_async_submit_limit(fs_info);
@@ -491,8 +511,13 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
 	async->rw = rw;
 	async->bio = bio;
 	async->mirror_num = mirror_num;
-	async->submit_bio_hook = submit_bio_hook;
-	async->work.func = run_one_async_submit;
+	async->submit_bio_start = submit_bio_start;
+	async->submit_bio_done = submit_bio_done;
+
+	async->work.func = run_one_async_start;
+	async->work.ordered_func = run_one_async_done;
+	async->work.ordered_free = run_one_async_free;
+
 	async->work.flags = 0;
 	async->bio_flags = bio_flags;
 
@@ -533,29 +558,25 @@ static int btree_csum_one_bio(struct bio *bio)
 	return 0;
 }
 
-static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
-				 int mirror_num, unsigned long bio_flags)
+static int __btree_submit_bio_start(struct inode *inode, int rw,
+				    struct bio *bio, int mirror_num,
+				    unsigned long bio_flags)
 {
-	struct btrfs_root *root = BTRFS_I(inode)->root;
-	int ret;
-
 	/*
 	 * when we're called for a write, we're already in the async
 	 * submission context.  Just jump into btrfs_map_bio
 	 */
-	if (rw & (1 << BIO_RW)) {
-		btree_csum_one_bio(bio);
-		return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
-				     mirror_num, 1);
-	}
+	btree_csum_one_bio(bio);
+	return 0;
+}
 
+static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
+				 int mirror_num, unsigned long bio_flags)
+{
 	/*
-	 * called for a read, do the setup so that checksum validation
-	 * can happen in the async kernel threads
+	 * when we're called for a write, we're already in the async
+	 * submission context.  Just jump into btrfs_map_bio
 	 */
-	ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1);
-	BUG_ON(ret);
-
 	return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
 }
 
@@ -567,11 +588,22 @@ static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
 	 * can happen in parallel across all CPUs
 	 */
 	if (!(rw & (1 << BIO_RW))) {
-		return __btree_submit_bio_hook(inode, rw, bio, mirror_num, 0);
+		int ret;
+		/*
+		 * called for a read, do the setup so that checksum validation
+		 * can happen in the async kernel threads
+		 */
+		ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
+					  bio, 1);
+		BUG_ON(ret);
+
+		return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
+				     mirror_num, 1);
 	}
 	return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
 				   inode, rw, bio, mirror_num, 0,
-				   __btree_submit_bio_hook);
+				   __btree_submit_bio_start,
+				   __btree_submit_bio_done);
 }
 
 static int btree_writepage(struct page *page, struct writeback_control *wbc)
@@ -1534,7 +1566,8 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	 * were sent by the writeback daemons, improving overall locality
 	 * of the IO going down the pipe.
 	 */
-	fs_info->workers.idle_thresh = 128;
+	fs_info->workers.idle_thresh = 8;
+	fs_info->workers.ordered = 1;
 
 	btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1);
 	btrfs_init_workers(&fs_info->endio_workers, "endio",
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 4eb1f1408d21..b8d5948fa279 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -72,7 +72,9 @@ int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
 int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
 			int rw, struct bio *bio, int mirror_num,
 			unsigned long bio_flags,
-			extent_submit_bio_hook_t *submit_bio_hook);
+			extent_submit_bio_hook_t *submit_bio_start,
+			extent_submit_bio_hook_t *submit_bio_done);
+
 int btrfs_congested_async(struct btrfs_fs_info *info, int iodone);
 unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info);
 int btrfs_write_tree_block(struct extent_buffer *buf);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 806708dd7e38..3df0ffad976e 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -881,7 +881,7 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
  * At IO completion time the cums attached on the ordered extent record
  * are inserted into the btree
  */
-int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
+int __btrfs_submit_bio_start(struct inode *inode, int rw, struct bio *bio,
 			  int mirror_num, unsigned long bio_flags)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
@@ -889,7 +889,21 @@ int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
 
 	ret = btrfs_csum_one_bio(root, inode, bio);
 	BUG_ON(ret);
+	return 0;
+}
 
+/*
+ * in order to insert checksums into the metadata in large chunks,
+ * we wait until bio submission time.   All the pages in the bio are
+ * checksummed and sums are attached onto the ordered extent record.
+ *
+ * At IO completion time the cums attached on the ordered extent record
+ * are inserted into the btree
+ */
+int __btrfs_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
+			  int mirror_num, unsigned long bio_flags)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
 	return btrfs_map_bio(root, rw, bio, mirror_num, 1);
 }
 
@@ -922,7 +936,8 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
 		/* we're doing a write, do the async checksumming */
 		return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
 				   inode, rw, bio, mirror_num,
-				   bio_flags, __btrfs_submit_bio_hook);
+				   bio_flags, __btrfs_submit_bio_start,
+				   __btrfs_submit_bio_done);
 	}
 
 mapit:
-- 
cgit v1.2.3


From c146afad2c7fea6a366d4945c1bab9b03880f526 Mon Sep 17 00:00:00 2001
From: Yan Zheng <zheng.yan@oracle.com>
Date: Wed, 12 Nov 2008 14:34:12 -0500
Subject: Btrfs: mount ro and remount support

This patch adds mount ro and remount support. The main
changes in patch are: adding btrfs_remount and related
helper function; splitting the transaction related code
out of close_ctree into btrfs_commit_super; updating
allocator to properly handle read only block group.

Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
---
 fs/btrfs/ctree.h       |  1 +
 fs/btrfs/disk-io.c     | 83 ++++++++++++++++++++++++++++++++++++++------------
 fs/btrfs/disk-io.h     |  2 ++
 fs/btrfs/extent-tree.c | 61 +++++++++++++++++++++----------------
 fs/btrfs/inode.c       | 15 +++------
 fs/btrfs/ioctl.c       | 22 +++++++++++++
 fs/btrfs/super.c       | 37 ++++++++++++++++++++--
 7 files changed, 163 insertions(+), 58 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index f575939e0258..c4c6c127323b 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -541,6 +541,7 @@ struct btrfs_space_info {
 	u64 bytes_used;
 	u64 bytes_pinned;
 	u64 bytes_reserved;
+	u64 bytes_readonly;
 	int full;
 	int force_alloc;
 	struct list_head list;
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 3b0e974a9e9c..c599f0ee997a 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1075,10 +1075,12 @@ struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
 		kfree(root);
 		return ERR_PTR(ret);
 	}
-	ret = btrfs_find_dead_roots(fs_info->tree_root,
-				    root->root_key.objectid, root);
-	BUG_ON(ret);
-
+	if (!(fs_info->sb->s_flags & MS_RDONLY)) {
+		ret = btrfs_find_dead_roots(fs_info->tree_root,
+					    root->root_key.objectid, root);
+		BUG_ON(ret);
+		btrfs_orphan_cleanup(root);
+	}
 	return root;
 }
 
@@ -1700,7 +1702,8 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 
 	btrfs_read_block_groups(extent_root);
 
-	fs_info->generation = btrfs_super_generation(disk_super) + 1;
+	fs_info->generation = generation + 1;
+	fs_info->last_trans_committed = generation;
 	fs_info->data_alloc_profile = (u64)-1;
 	fs_info->metadata_alloc_profile = (u64)-1;
 	fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
@@ -1715,6 +1718,9 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	if (!fs_info->transaction_kthread)
 		goto fail_cleaner;
 
+	if (sb->s_flags & MS_RDONLY)
+		return tree_root;
+
 	if (btrfs_super_log_root(disk_super) != 0) {
 		u32 blocksize;
 		u64 bytenr = btrfs_super_log_root(disk_super);
@@ -1735,7 +1741,6 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 		ret = btrfs_recover_log_trees(log_tree_root);
 		BUG_ON(ret);
 	}
-	fs_info->last_trans_committed = btrfs_super_generation(disk_super);
 
 	ret = btrfs_cleanup_reloc_trees(tree_root);
 	BUG_ON(ret);
@@ -1955,28 +1960,69 @@ static int del_fs_roots(struct btrfs_fs_info *fs_info)
 	return 0;
 }
 
-int close_ctree(struct btrfs_root *root)
+int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
 {
+	u64 root_objectid = 0;
+	struct btrfs_root *gang[8];
+	int i;
 	int ret;
-	struct btrfs_trans_handle *trans;
-	struct btrfs_fs_info *fs_info = root->fs_info;
 
-	fs_info->closing = 1;
-	smp_mb();
+	while (1) {
+		ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
+					     (void **)gang, root_objectid,
+					     ARRAY_SIZE(gang));
+		if (!ret)
+			break;
+		for (i = 0; i < ret; i++) {
+			root_objectid = gang[i]->root_key.objectid;
+			ret = btrfs_find_dead_roots(fs_info->tree_root,
+						    root_objectid, gang[i]);
+			BUG_ON(ret);
+			btrfs_orphan_cleanup(gang[i]);
+		}
+		root_objectid++;
+	}
+	return 0;
+}
 
-	kthread_stop(root->fs_info->transaction_kthread);
-	kthread_stop(root->fs_info->cleaner_kthread);
+int btrfs_commit_super(struct btrfs_root *root)
+{
+	struct btrfs_trans_handle *trans;
+	int ret;
 
+	mutex_lock(&root->fs_info->cleaner_mutex);
 	btrfs_clean_old_snapshots(root);
+	mutex_unlock(&root->fs_info->cleaner_mutex);
 	trans = btrfs_start_transaction(root, 1);
 	ret = btrfs_commit_transaction(trans, root);
-	/* run commit again to  drop the original snapshot */
+	BUG_ON(ret);
+	/* run commit again to drop the original snapshot */
 	trans = btrfs_start_transaction(root, 1);
 	btrfs_commit_transaction(trans, root);
 	ret = btrfs_write_and_wait_transaction(NULL, root);
 	BUG_ON(ret);
 
-	write_ctree_super(NULL, root);
+	ret = write_ctree_super(NULL, root);
+	return ret;
+}
+
+int close_ctree(struct btrfs_root *root)
+{
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	int ret;
+
+	fs_info->closing = 1;
+	smp_mb();
+
+	kthread_stop(root->fs_info->transaction_kthread);
+	kthread_stop(root->fs_info->cleaner_kthread);
+
+	if (!(fs_info->sb->s_flags & MS_RDONLY)) {
+		ret =  btrfs_commit_super(root);
+		if (ret) {
+			printk("btrfs: commit super returns %d\n", ret);
+		}
+	}
 
 	if (fs_info->delalloc_bytes) {
 		printk("btrfs: at unmount delalloc count %Lu\n",
@@ -2000,12 +2046,10 @@ int close_ctree(struct btrfs_root *root)
 		free_extent_buffer(root->fs_info->dev_root->node);
 
 	btrfs_free_block_groups(root->fs_info);
-	fs_info->closing = 2;
-	del_fs_roots(fs_info);
 
-	filemap_write_and_wait(fs_info->btree_inode->i_mapping);
+	del_fs_roots(fs_info);
 
-	truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
+	iput(fs_info->btree_inode);
 
 	btrfs_stop_workers(&fs_info->fixup_workers);
 	btrfs_stop_workers(&fs_info->delalloc_workers);
@@ -2014,7 +2058,6 @@ int close_ctree(struct btrfs_root *root)
 	btrfs_stop_workers(&fs_info->endio_write_workers);
 	btrfs_stop_workers(&fs_info->submit_workers);
 
-	iput(fs_info->btree_inode);
 #if 0
 	while(!list_empty(&fs_info->hashers)) {
 		struct btrfs_hasher *hasher;
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index b8d5948fa279..717e94811e4e 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -38,6 +38,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 int close_ctree(struct btrfs_root *root);
 int write_ctree_super(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root);
+int btrfs_commit_super(struct btrfs_root *root);
 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
 					    u64 bytenr, u32 blocksize);
 struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
@@ -49,6 +50,7 @@ struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
 					       struct btrfs_key *location);
 struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
 					      struct btrfs_key *location);
+int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info);
 int btrfs_insert_dev_radix(struct btrfs_root *root,
 			   struct block_device *bdev,
 			   u64 device_id,
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index e785f0a0632b..af2de30dbeac 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1794,7 +1794,7 @@ static int update_space_info(struct btrfs_fs_info *info, u64 flags,
 		*space_info = found;
 		return 0;
 	}
-	found = kmalloc(sizeof(*found), GFP_NOFS);
+	found = kzalloc(sizeof(*found), GFP_NOFS);
 	if (!found)
 		return -ENOMEM;
 
@@ -1807,6 +1807,7 @@ static int update_space_info(struct btrfs_fs_info *info, u64 flags,
 	found->bytes_used = bytes_used;
 	found->bytes_pinned = 0;
 	found->bytes_reserved = 0;
+	found->bytes_readonly = 0;
 	found->full = 0;
 	found->force_alloc = 0;
 	*space_info = found;
@@ -1829,6 +1830,19 @@ static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
 	}
 }
 
+static void set_block_group_readonly(struct btrfs_block_group_cache *cache)
+{
+	spin_lock(&cache->space_info->lock);
+	spin_lock(&cache->lock);
+	if (!cache->ro) {
+		cache->space_info->bytes_readonly += cache->key.offset -
+					btrfs_block_group_used(&cache->item);
+		cache->ro = 1;
+	}
+	spin_unlock(&cache->lock);
+	spin_unlock(&cache->space_info->lock);
+}
+
 static u64 reduce_alloc_profile(struct btrfs_root *root, u64 flags)
 {
 	u64 num_devices = root->fs_info->fs_devices->num_devices;
@@ -1865,7 +1879,9 @@ static int do_chunk_alloc(struct btrfs_trans_handle *trans,
 	u64 thresh;
 	u64 start;
 	u64 num_bytes;
-	int ret = 0, waited = 0;
+	int ret = 0;
+
+	mutex_lock(&extent_root->fs_info->chunk_mutex);
 
 	flags = reduce_alloc_profile(extent_root, flags);
 
@@ -1887,46 +1903,28 @@ static int do_chunk_alloc(struct btrfs_trans_handle *trans,
 		goto out;
 	}
 
-	thresh = div_factor(space_info->total_bytes, 6);
+	thresh = space_info->total_bytes - space_info->bytes_readonly;
+	thresh = div_factor(thresh, 6);
 	if (!force &&
 	   (space_info->bytes_used + space_info->bytes_pinned +
 	    space_info->bytes_reserved + alloc_bytes) < thresh) {
 		spin_unlock(&space_info->lock);
 		goto out;
 	}
-
 	spin_unlock(&space_info->lock);
 
-	ret = mutex_trylock(&extent_root->fs_info->chunk_mutex);
-	if (!ret && !force) {
-		goto out;
-	} else if (!ret) {
-		mutex_lock(&extent_root->fs_info->chunk_mutex);
-		waited = 1;
-	}
-
-	if (waited) {
-		spin_lock(&space_info->lock);
-		if (space_info->full) {
-			spin_unlock(&space_info->lock);
-			goto out_unlock;
-		}
-		spin_unlock(&space_info->lock);
-	}
-
 	ret = btrfs_alloc_chunk(trans, extent_root, &start, &num_bytes, flags);
 	if (ret) {
 printk("space info full %Lu\n", flags);
 		space_info->full = 1;
-		goto out_unlock;
+		goto out;
 	}
 
 	ret = btrfs_make_block_group(trans, extent_root, 0, flags,
 		     BTRFS_FIRST_CHUNK_TREE_OBJECTID, start, num_bytes);
 	BUG_ON(ret);
-out_unlock:
-	mutex_unlock(&extent_root->fs_info->chunk_mutex);
 out:
+	mutex_unlock(&extent_root->fs_info->chunk_mutex);
 	return ret;
 }
 
@@ -1956,12 +1954,18 @@ static int update_block_group(struct btrfs_trans_handle *trans,
 		if (alloc) {
 			old_val += num_bytes;
 			cache->space_info->bytes_used += num_bytes;
+			if (cache->ro) {
+				cache->space_info->bytes_readonly -= num_bytes;
+				WARN_ON(1);
+			}
 			btrfs_set_block_group_used(&cache->item, old_val);
 			spin_unlock(&cache->lock);
 			spin_unlock(&cache->space_info->lock);
 		} else {
 			old_val -= num_bytes;
 			cache->space_info->bytes_used -= num_bytes;
+			if (cache->ro)
+				cache->space_info->bytes_readonly += num_bytes;
 			btrfs_set_block_group_used(&cache->item, old_val);
 			spin_unlock(&cache->lock);
 			spin_unlock(&cache->space_info->lock);
@@ -5560,8 +5564,7 @@ int btrfs_relocate_block_group(struct btrfs_root *root, u64 group_start)
 	BUG_ON(IS_ERR(reloc_inode));
 
 	__alloc_chunk_for_shrink(root, block_group, 1);
-	block_group->ro = 1;
-	block_group->space_info->total_bytes -= block_group->key.offset;
+	set_block_group_readonly(block_group);
 
 	btrfs_start_delalloc_inodes(info->tree_root);
 	btrfs_wait_ordered_extents(info->tree_root, 0);
@@ -5868,6 +5871,7 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
 
 	block_group = btrfs_lookup_block_group(root->fs_info, group_start);
 	BUG_ON(!block_group);
+	BUG_ON(!block_group->ro);
 
 	memcpy(&key, &block_group->key, sizeof(key));
 
@@ -5881,6 +5885,11 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
 	list_del(&block_group->list);
 	up_write(&block_group->space_info->groups_sem);
 
+	spin_lock(&block_group->space_info->lock);
+	block_group->space_info->total_bytes -= block_group->key.offset;
+	block_group->space_info->bytes_readonly -= block_group->key.offset;
+	spin_unlock(&block_group->space_info->lock);
+
 	/*
 	memset(shrink_block_group, 0, sizeof(*shrink_block_group));
 	kfree(shrink_block_group);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 2ed2deacde90..3e3620e69bb9 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -1808,10 +1808,6 @@ void btrfs_orphan_cleanup(struct btrfs_root *root)
 	struct inode *inode;
 	int ret = 0, nr_unlink = 0, nr_truncate = 0;
 
-	/* don't do orphan cleanup if the fs is readonly. */
-	if (root->fs_info->sb->s_flags & MS_RDONLY)
-		return;
-
 	path = btrfs_alloc_path();
 	if (!path)
 		return;
@@ -3050,7 +3046,7 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
 	struct btrfs_root *root = bi->root;
 	struct btrfs_root *sub_root = root;
 	struct btrfs_key location;
-	int ret, new, do_orphan = 0;
+	int ret, new;
 
 	if (dentry->d_name.len > BTRFS_NAME_LEN)
 		return ERR_PTR(-ENAMETOOLONG);
@@ -3076,13 +3072,9 @@ static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
 		if (new && root != sub_root) {
 			igrab(inode);
 			sub_root->inode = inode;
-			do_orphan = 1;
 		}
 	}
 
-	if (unlikely(do_orphan))
-		btrfs_orphan_cleanup(sub_root);
-
 	return d_splice_alias(inode, dentry);
 }
 
@@ -3237,7 +3229,7 @@ int btrfs_write_inode(struct inode *inode, int wait)
 	struct btrfs_trans_handle *trans;
 	int ret = 0;
 
-	if (root->fs_info->closing > 1)
+	if (root->fs_info->btree_inode == inode)
 		return 0;
 
 	if (wait) {
@@ -4625,6 +4617,9 @@ int btrfs_start_delalloc_inodes(struct btrfs_root *root)
 	struct inode *inode;
 	unsigned long flags;
 
+	if (root->fs_info->sb->s_flags & MS_RDONLY)
+		return -EROFS;
+
 	spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
 	while(!list_empty(head)) {
 		binode = list_entry(head->next, struct btrfs_inode,
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 4d7cc7c504d0..52863cebd594 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -378,6 +378,9 @@ static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
 	int namelen;
 	int mod = 0;
 
+	if (root->fs_info->sb->s_flags & MS_RDONLY)
+		return -EROFS;
+
 	vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
 
 	if (!vol_args)
@@ -478,6 +481,9 @@ static noinline int btrfs_ioctl_snap_create(struct file *file,
 	int namelen;
 	int ret;
 
+	if (root->fs_info->sb->s_flags & MS_RDONLY)
+		return -EROFS;
+
 	vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
 
 	if (!vol_args)
@@ -534,6 +540,11 @@ static int btrfs_ioctl_defrag(struct file *file)
 {
 	struct inode *inode = fdentry(file)->d_inode;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
+	int ret;
+
+	ret = mnt_want_write(file->f_path.mnt);
+	if (ret)
+		return ret;
 
 	switch (inode->i_mode & S_IFMT) {
 	case S_IFDIR:
@@ -575,6 +586,9 @@ long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
 	struct btrfs_ioctl_vol_args *vol_args;
 	int ret;
 
+	if (root->fs_info->sb->s_flags & MS_RDONLY)
+		return -EROFS;
+
 	vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
 
 	if (!vol_args)
@@ -621,6 +635,10 @@ long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, u64 off,
 	 *   they don't overlap)?
 	 */
 
+	ret = mnt_want_write(file->f_path.mnt);
+	if (ret)
+		return ret;
+
 	src_file = fget(srcfd);
 	if (!src_file)
 		return -EBADF;
@@ -958,6 +976,10 @@ long btrfs_ioctl_trans_start(struct file *file)
 		goto out;
 	}
 
+	ret = mnt_want_write(file->f_path.mnt);
+	if (ret)
+		goto out;
+
 	mutex_lock(&root->fs_info->trans_mutex);
 	root->fs_info->open_ioctl_trans++;
 	mutex_unlock(&root->fs_info->trans_mutex);
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index ab9d5e89ed13..04a3bf816509 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -370,6 +370,9 @@ int btrfs_sync_fs(struct super_block *sb, int wait)
 	int ret;
 	root = btrfs_sb(sb);
 
+	if (sb->s_flags & MS_RDONLY)
+		return 0;
+
 	sb->s_dirt = 0;
 	if (!wait) {
 		filemap_flush(root->fs_info->btree_inode->i_mapping);
@@ -438,7 +441,7 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
 			up_write(&s->s_umount);
 			deactivate_super(s);
 			error = -EBUSY;
-			goto error_bdev;
+			goto error_close_devices;
 		}
 
 	} else {
@@ -487,7 +490,7 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
 
 error_s:
 	error = PTR_ERR(s);
-error_bdev:
+error_close_devices:
 	btrfs_close_devices(fs_devices);
 error_free_subvol_name:
 	kfree(subvol_name);
@@ -495,6 +498,35 @@ error:
 	return error;
 }
 
+static int btrfs_remount(struct super_block *sb, int *flags, char *data)
+{
+	struct btrfs_root *root = btrfs_sb(sb);
+	int ret;
+
+	if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
+		return 0;
+
+	if (*flags & MS_RDONLY) {
+		sb->s_flags |= MS_RDONLY;
+
+		ret =  btrfs_commit_super(root);
+		WARN_ON(ret);
+	} else {
+		if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
+			return -EINVAL;
+
+		ret = btrfs_cleanup_reloc_trees(root);
+		WARN_ON(ret);
+
+		ret = btrfs_cleanup_fs_roots(root->fs_info);
+		WARN_ON(ret);
+
+		sb->s_flags &= ~MS_RDONLY;
+	}
+
+	return 0;
+}
+
 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 {
 	struct btrfs_root *root = btrfs_sb(dentry->d_sb);
@@ -582,6 +614,7 @@ static struct super_operations btrfs_super_ops = {
 	.alloc_inode	= btrfs_alloc_inode,
 	.destroy_inode	= btrfs_destroy_inode,
 	.statfs		= btrfs_statfs,
+	.remount_fs	= btrfs_remount,
 	.write_super_lockfs = btrfs_write_super_lockfs,
 	.unlockfs	= btrfs_unlockfs,
 };
-- 
cgit v1.2.3


From a512bbf855ff0af474257475f2e6da7acd854f52 Mon Sep 17 00:00:00 2001
From: Yan Zheng <zheng.yan@oracle.com>
Date: Mon, 8 Dec 2008 16:46:26 -0500
Subject: Btrfs: superblock duplication

This patch implements superblock duplication. Superblocks
are stored at offset 16K, 64M and 256G on every devices.
Spaces used by superblocks are preserved by the allocator,
which uses a reverse mapping function to find the logical
addresses that correspond to superblocks. Thank you,

Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
---
 fs/btrfs/disk-io.c          | 208 ++++++++++++++++++++++++++++++--------------
 fs/btrfs/disk-io.h          |  17 +++-
 fs/btrfs/extent-tree.c      |  54 ++++++------
 fs/btrfs/free-space-cache.c |   1 -
 fs/btrfs/transaction.c      |   2 +-
 fs/btrfs/tree-log.c         |   3 +-
 fs/btrfs/volumes.c          | 107 ++++++++++++++++++-----
 fs/btrfs/volumes.h          |   6 +-
 8 files changed, 279 insertions(+), 119 deletions(-)

(limited to 'fs/btrfs/disk-io.h')

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 61dc3b2c834b..c72f4f3b912c 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1595,8 +1595,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 		     fs_info, BTRFS_ROOT_TREE_OBJECTID);
 
 
-	bh = __bread(fs_devices->latest_bdev,
-		     BTRFS_SUPER_INFO_OFFSET / 4096, 4096);
+	bh = btrfs_read_dev_super(fs_devices->latest_bdev);
 	if (!bh)
 		goto fail_iput;
 
@@ -1710,7 +1709,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	}
 
 	mutex_lock(&fs_info->chunk_mutex);
-	ret = btrfs_read_sys_array(tree_root);
+	ret = btrfs_read_sys_array(tree_root, btrfs_super_bytenr(disk_super));
 	mutex_unlock(&fs_info->chunk_mutex);
 	if (ret) {
 		printk("btrfs: failed to read the system array on %s\n",
@@ -1905,19 +1904,147 @@ static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
 	put_bh(bh);
 }
 
-static int write_all_supers(struct btrfs_root *root)
+struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
+{
+	struct buffer_head *bh;
+	struct buffer_head *latest = NULL;
+	struct btrfs_super_block *super;
+	int i;
+	u64 transid = 0;
+	u64 bytenr;
+
+	/* we would like to check all the supers, but that would make
+	 * a btrfs mount succeed after a mkfs from a different FS.
+	 * So, we need to add a special mount option to scan for
+	 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
+	 */
+	for (i = 0; i < 1; i++) {
+		bytenr = btrfs_sb_offset(i);
+		if (bytenr + 4096 >= i_size_read(bdev->bd_inode))
+			break;
+		bh = __bread(bdev, bytenr / 4096, 4096);
+		if (!bh)
+			continue;
+
+		super = (struct btrfs_super_block *)bh->b_data;
+		if (btrfs_super_bytenr(super) != bytenr ||
+		    strncmp((char *)(&super->magic), BTRFS_MAGIC,
+			    sizeof(super->magic))) {
+			brelse(bh);
+			continue;
+		}
+
+		if (!latest || btrfs_super_generation(super) > transid) {
+			brelse(latest);
+			latest = bh;
+			transid = btrfs_super_generation(super);
+		} else {
+			brelse(bh);
+		}
+	}
+	return latest;
+}
+
+static int write_dev_supers(struct btrfs_device *device,
+			    struct btrfs_super_block *sb,
+			    int do_barriers, int wait, int max_mirrors)
+{
+	struct buffer_head *bh;
+	int i;
+	int ret;
+	int errors = 0;
+	u32 crc;
+	u64 bytenr;
+	int last_barrier = 0;
+
+	if (max_mirrors == 0)
+		max_mirrors = BTRFS_SUPER_MIRROR_MAX;
+
+	/* make sure only the last submit_bh does a barrier */
+	if (do_barriers) {
+		for (i = 0; i < max_mirrors; i++) {
+			bytenr = btrfs_sb_offset(i);
+			if (bytenr + BTRFS_SUPER_INFO_SIZE >=
+			    device->total_bytes)
+				break;
+			last_barrier = i;
+		}
+	}
+
+	for (i = 0; i < max_mirrors; i++) {
+		bytenr = btrfs_sb_offset(i);
+		if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
+			break;
+
+		if (wait) {
+			bh = __find_get_block(device->bdev, bytenr / 4096,
+					      BTRFS_SUPER_INFO_SIZE);
+			BUG_ON(!bh);
+			brelse(bh);
+			wait_on_buffer(bh);
+			if (buffer_uptodate(bh)) {
+				brelse(bh);
+				continue;
+			}
+		} else {
+			btrfs_set_super_bytenr(sb, bytenr);
+
+			crc = ~(u32)0;
+			crc = btrfs_csum_data(NULL, (char *)sb +
+					      BTRFS_CSUM_SIZE, crc,
+					      BTRFS_SUPER_INFO_SIZE -
+					      BTRFS_CSUM_SIZE);
+			btrfs_csum_final(crc, sb->csum);
+
+			bh = __getblk(device->bdev, bytenr / 4096,
+				      BTRFS_SUPER_INFO_SIZE);
+			memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
+
+			set_buffer_uptodate(bh);
+			get_bh(bh);
+			lock_buffer(bh);
+			bh->b_end_io = btrfs_end_buffer_write_sync;
+		}
+
+		if (i == last_barrier && do_barriers && device->barriers) {
+			ret = submit_bh(WRITE_BARRIER, bh);
+			if (ret == -EOPNOTSUPP) {
+				printk("btrfs: disabling barriers on dev %s\n",
+				       device->name);
+				set_buffer_uptodate(bh);
+				device->barriers = 0;
+				get_bh(bh);
+				lock_buffer(bh);
+				ret = submit_bh(WRITE, bh);
+			}
+		} else {
+			ret = submit_bh(WRITE, bh);
+		}
+
+		if (!ret && wait) {
+			wait_on_buffer(bh);
+			if (!buffer_uptodate(bh))
+				errors++;
+		} else if (ret) {
+			errors++;
+		}
+		if (wait)
+			brelse(bh);
+	}
+	return errors < i ? 0 : -1;
+}
+
+int write_all_supers(struct btrfs_root *root, int max_mirrors)
 {
 	struct list_head *cur;
 	struct list_head *head = &root->fs_info->fs_devices->devices;
 	struct btrfs_device *dev;
 	struct btrfs_super_block *sb;
 	struct btrfs_dev_item *dev_item;
-	struct buffer_head *bh;
 	int ret;
 	int do_barriers;
 	int max_errors;
 	int total_errors = 0;
-	u32 crc;
 	u64 flags;
 
 	max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
@@ -1944,40 +2071,11 @@ static int write_all_supers(struct btrfs_root *root)
 		btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
 		memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
 		memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
+
 		flags = btrfs_super_flags(sb);
 		btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
 
-
-		crc = ~(u32)0;
-		crc = btrfs_csum_data(root, (char *)sb + BTRFS_CSUM_SIZE, crc,
-				      BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
-		btrfs_csum_final(crc, sb->csum);
-
-		bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET / 4096,
-			      BTRFS_SUPER_INFO_SIZE);
-
-		memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
-		dev->pending_io = bh;
-
-		get_bh(bh);
-		set_buffer_uptodate(bh);
-		lock_buffer(bh);
-		bh->b_end_io = btrfs_end_buffer_write_sync;
-
-		if (do_barriers && dev->barriers) {
-			ret = submit_bh(WRITE_BARRIER, bh);
-			if (ret == -EOPNOTSUPP) {
-				printk("btrfs: disabling barriers on dev %s\n",
-				       dev->name);
-				set_buffer_uptodate(bh);
-				dev->barriers = 0;
-				get_bh(bh);
-				lock_buffer(bh);
-				ret = submit_bh(WRITE, bh);
-			}
-		} else {
-			ret = submit_bh(WRITE, bh);
-		}
+		ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
 		if (ret)
 			total_errors++;
 	}
@@ -1985,8 +2083,8 @@ static int write_all_supers(struct btrfs_root *root)
 		printk("btrfs: %d errors while writing supers\n", total_errors);
 		BUG();
 	}
-	total_errors = 0;
 
+	total_errors = 0;
 	list_for_each(cur, head) {
 		dev = list_entry(cur, struct btrfs_device, dev_list);
 		if (!dev->bdev)
@@ -1994,29 +2092,9 @@ static int write_all_supers(struct btrfs_root *root)
 		if (!dev->in_fs_metadata || !dev->writeable)
 			continue;
 
-		BUG_ON(!dev->pending_io);
-		bh = dev->pending_io;
-		wait_on_buffer(bh);
-		if (!buffer_uptodate(dev->pending_io)) {
-			if (do_barriers && dev->barriers) {
-				printk("btrfs: disabling barriers on dev %s\n",
-				       dev->name);
-				set_buffer_uptodate(bh);
-				get_bh(bh);
-				lock_buffer(bh);
-				dev->barriers = 0;
-				ret = submit_bh(WRITE, bh);
-				BUG_ON(ret);
-				wait_on_buffer(bh);
-				if (!buffer_uptodate(bh))
-					total_errors++;
-			} else {
-				total_errors++;
-			}
-
-		}
-		dev->pending_io = NULL;
-		brelse(bh);
+		ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
+		if (ret)
+			total_errors++;
 	}
 	if (total_errors > max_errors) {
 		printk("btrfs: %d errors while writing supers\n", total_errors);
@@ -2025,12 +2103,12 @@ static int write_all_supers(struct btrfs_root *root)
 	return 0;
 }
 
-int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root)
+int write_ctree_super(struct btrfs_trans_handle *trans,
+		      struct btrfs_root *root, int max_mirrors)
 {
 	int ret;
 
-	ret = write_all_supers(root);
+	ret = write_all_supers(root, max_mirrors);
 	return ret;
 }
 
@@ -2116,7 +2194,7 @@ int btrfs_commit_super(struct btrfs_root *root)
 	ret = btrfs_write_and_wait_transaction(NULL, root);
 	BUG_ON(ret);
 
-	ret = write_ctree_super(NULL, root);
+	ret = write_ctree_super(NULL, root, 0);
 	return ret;
 }
 
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 717e94811e4e..c0ff404c31b7 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -19,8 +19,20 @@
 #ifndef __DISKIO__
 #define __DISKIO__
 
-#define BTRFS_SUPER_INFO_OFFSET (16 * 1024)
+#define BTRFS_SUPER_INFO_OFFSET (64 * 1024)
 #define BTRFS_SUPER_INFO_SIZE 4096
+
+#define BTRFS_SUPER_MIRROR_MAX	 3
+#define BTRFS_SUPER_MIRROR_SHIFT 12
+
+static inline u64 btrfs_sb_offset(int mirror)
+{
+	u64 start = 16 * 1024;
+	if (mirror)
+		return start << (BTRFS_SUPER_MIRROR_SHIFT * mirror);
+	return BTRFS_SUPER_INFO_OFFSET;
+}
+
 struct btrfs_device;
 struct btrfs_fs_devices;
 
@@ -37,7 +49,8 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 			      char *options);
 int close_ctree(struct btrfs_root *root);
 int write_ctree_super(struct btrfs_trans_handle *trans,
-		      struct btrfs_root *root);
+		      struct btrfs_root *root, int max_mirrors);
+struct buffer_head *btrfs_read_dev_super(struct block_device *bdev);
 int btrfs_commit_super(struct btrfs_root *root);
 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
 					    u64 bytenr, u32 blocksize);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index d15638529389..803647bc8400 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -189,6 +189,29 @@ static int add_new_free_space(struct btrfs_block_group_cache *block_group,
 	return 0;
 }
 
+static int remove_sb_from_cache(struct btrfs_root *root,
+				struct btrfs_block_group_cache *cache)
+{
+	u64 bytenr;
+	u64 *logical;
+	int stripe_len;
+	int i, nr, ret;
+
+	for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
+		bytenr = btrfs_sb_offset(i);
+		ret = btrfs_rmap_block(&root->fs_info->mapping_tree,
+				       cache->key.objectid, bytenr, 0,
+				       &logical, &nr, &stripe_len);
+		BUG_ON(ret);
+		while (nr--) {
+			btrfs_remove_free_space(cache, logical[nr],
+						stripe_len);
+		}
+		kfree(logical);
+	}
+	return 0;
+}
+
 static int cache_block_group(struct btrfs_root *root,
 			     struct btrfs_block_group_cache *block_group)
 {
@@ -197,9 +220,7 @@ static int cache_block_group(struct btrfs_root *root,
 	struct btrfs_key key;
 	struct extent_buffer *leaf;
 	int slot;
-	u64 last = 0;
-	u64 first_free;
-	int found = 0;
+	u64 last = block_group->key.objectid;
 
 	if (!block_group)
 		return 0;
@@ -220,23 +241,13 @@ static int cache_block_group(struct btrfs_root *root,
 	 * skip the locking here
 	 */
 	path->skip_locking = 1;
-	first_free = max_t(u64, block_group->key.objectid,
-			   BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE);
-	key.objectid = block_group->key.objectid;
+	key.objectid = last;
 	key.offset = 0;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 	if (ret < 0)
 		goto err;
-	ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY);
-	if (ret < 0)
-		goto err;
-	if (ret == 0) {
-		leaf = path->nodes[0];
-		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
-		if (key.objectid + key.offset > first_free)
-			first_free = key.objectid + key.offset;
-	}
+
 	while(1) {
 		leaf = path->nodes[0];
 		slot = path->slots[0];
@@ -258,11 +269,6 @@ static int cache_block_group(struct btrfs_root *root,
 			break;
 
 		if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) {
-			if (!found) {
-				last = first_free;
-				found = 1;
-			}
-
 			add_new_free_space(block_group, root->fs_info, last,
 					   key.objectid);
 
@@ -272,13 +278,11 @@ next:
 		path->slots[0]++;
 	}
 
-	if (!found)
-		last = first_free;
-
 	add_new_free_space(block_group, root->fs_info, last,
 			   block_group->key.objectid +
 			   block_group->key.offset);
 
+	remove_sb_from_cache(root, block_group);
 	block_group->cached = 1;
 	ret = 0;
 err:
@@ -1974,10 +1978,8 @@ static int update_block_group(struct btrfs_trans_handle *trans,
 		if (alloc) {
 			old_val += num_bytes;
 			cache->space_info->bytes_used += num_bytes;
-			if (cache->ro) {
+			if (cache->ro)
 				cache->space_info->bytes_readonly -= num_bytes;
-				WARN_ON(1);
-			}
 			btrfs_set_block_group_used(&cache->item, old_val);
 			spin_unlock(&cache->lock);
 			spin_unlock(&cache->space_info->lock);
diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c
index 09462adfbe33..2e69b9c30437 100644
--- a/fs/btrfs/free-space-cache.c
+++ b/fs/btrfs/free-space-cache.c
@@ -290,7 +290,6 @@ __btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
 			ret = -EINVAL;
 			goto out;
 		}
-
 		unlink_free_space(block_group, info);
 
 		if (info->bytes == bytes) {
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index c38f6a0e30b1..47cd5fcad2c8 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -1038,7 +1038,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	mutex_unlock(&root->fs_info->trans_mutex);
 	ret = btrfs_write_and_wait_transaction(trans, root);
 	BUG_ON(ret);
-	write_ctree_super(trans, root);
+	write_ctree_super(trans, root, 0);
 
 	/*
 	 * the super is written, we can safely allow the tree-loggers
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index 08469ec05850..d3f9c2c663c4 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -1996,7 +1996,7 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 	btrfs_set_super_log_root_level(&root->fs_info->super_for_commit,
 		       btrfs_header_level(log->fs_info->log_root_tree->node));
 
-	write_ctree_super(trans, log->fs_info->tree_root);
+	write_ctree_super(trans, log->fs_info->tree_root, 2);
 	log->fs_info->tree_log_transid++;
 	log->fs_info->tree_log_batch = 0;
 	atomic_set(&log->fs_info->tree_log_commit, 0);
@@ -2006,7 +2006,6 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 out:
 	mutex_unlock(&log->fs_info->tree_log_mutex);
 	return 0;
-
 }
 
 /* * free all the extents used by the tree log.  This should be called
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 2049d179ccd5..a79b3cc09e94 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -423,15 +423,11 @@ int __btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 		}
 		set_blocksize(bdev, 4096);
 
-		bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096);
+		bh = btrfs_read_dev_super(bdev);
 		if (!bh)
 			goto error_close;
 
 		disk_super = (struct btrfs_super_block *)bh->b_data;
-		if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
-		    sizeof(disk_super->magic)))
-			goto error_brelse;
-
 		devid = le64_to_cpu(disk_super->dev_item.devid);
 		if (devid != device->devid)
 			goto error_brelse;
@@ -529,17 +525,12 @@ int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder,
 	ret = set_blocksize(bdev, 4096);
 	if (ret)
 		goto error_close;
-	bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096);
+	bh = btrfs_read_dev_super(bdev);
 	if (!bh) {
 		ret = -EIO;
 		goto error_close;
 	}
 	disk_super = (struct btrfs_super_block *)bh->b_data;
-	if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
-	    sizeof(disk_super->magic))) {
-		ret = -EINVAL;
-		goto error_brelse;
-	}
 	devid = le64_to_cpu(disk_super->dev_item.devid);
 	transid = btrfs_super_generation(disk_super);
 	if (disk_super->label[0])
@@ -553,7 +544,6 @@ int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder,
 	printk("devid %Lu transid %Lu %s\n", devid, transid, path);
 	ret = device_list_add(path, disk_super, devid, fs_devices_ret);
 
-error_brelse:
 	brelse(bh);
 error_close:
 	close_bdev_exclusive(bdev, flags);
@@ -1016,17 +1006,12 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
 		}
 
 		set_blocksize(bdev, 4096);
-		bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096);
+		bh = btrfs_read_dev_super(bdev);
 		if (!bh) {
 			ret = -EIO;
 			goto error_close;
 		}
 		disk_super = (struct btrfs_super_block *)bh->b_data;
-		if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
-			    sizeof(disk_super->magic))) {
-			ret = -ENOENT;
-			goto error_brelse;
-		}
 		devid = le64_to_cpu(disk_super->dev_item.devid);
 		dev_uuid = disk_super->dev_item.uuid;
 		device = btrfs_find_device(root, devid, dev_uuid,
@@ -2563,6 +2548,88 @@ int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
 				 mirror_num, NULL);
 }
 
+int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree,
+		     u64 chunk_start, u64 physical, u64 devid,
+		     u64 **logical, int *naddrs, int *stripe_len)
+{
+	struct extent_map_tree *em_tree = &map_tree->map_tree;
+	struct extent_map *em;
+	struct map_lookup *map;
+	u64 *buf;
+	u64 bytenr;
+	u64 length;
+	u64 stripe_nr;
+	int i, j, nr = 0;
+
+	spin_lock(&em_tree->lock);
+	em = lookup_extent_mapping(em_tree, chunk_start, 1);
+	spin_unlock(&em_tree->lock);
+
+	BUG_ON(!em || em->start != chunk_start);
+	map = (struct map_lookup *)em->bdev;
+
+	length = em->len;
+	if (map->type & BTRFS_BLOCK_GROUP_RAID10)
+		do_div(length, map->num_stripes / map->sub_stripes);
+	else if (map->type & BTRFS_BLOCK_GROUP_RAID0)
+		do_div(length, map->num_stripes);
+
+	buf = kzalloc(sizeof(u64) * map->num_stripes, GFP_NOFS);
+	BUG_ON(!buf);
+
+	for (i = 0; i < map->num_stripes; i++) {
+		if (devid && map->stripes[i].dev->devid != devid)
+			continue;
+		if (map->stripes[i].physical > physical ||
+		    map->stripes[i].physical + length <= physical)
+			continue;
+
+		stripe_nr = physical - map->stripes[i].physical;
+		do_div(stripe_nr, map->stripe_len);
+
+		if (map->type & BTRFS_BLOCK_GROUP_RAID10) {
+			stripe_nr = stripe_nr * map->num_stripes + i;
+			do_div(stripe_nr, map->sub_stripes);
+		} else if (map->type & BTRFS_BLOCK_GROUP_RAID0) {
+			stripe_nr = stripe_nr * map->num_stripes + i;
+		}
+		bytenr = chunk_start + stripe_nr * map->stripe_len;
+		for (j = 0; j < nr; j++) {
+			if (buf[j] == bytenr)
+				break;
+		}
+		if (j == nr)
+			buf[nr++] = bytenr;
+	}
+
+	for (i = 0; i > nr; i++) {
+		struct btrfs_multi_bio *multi;
+		struct btrfs_bio_stripe *stripe;
+		int ret;
+
+		length = 1;
+		ret = btrfs_map_block(map_tree, WRITE, buf[i],
+				      &length, &multi, 0);
+		BUG_ON(ret);
+
+		stripe = multi->stripes;
+		for (j = 0; j < multi->num_stripes; j++) {
+			if (stripe->physical >= physical &&
+			    physical < stripe->physical + length)
+				break;
+		}
+		BUG_ON(j >= multi->num_stripes);
+		kfree(multi);
+	}
+
+	*logical = buf;
+	*naddrs = nr;
+	*stripe_len = map->stripe_len;
+
+	free_extent_map(em);
+	return 0;
+}
+
 int btrfs_unplug_page(struct btrfs_mapping_tree *map_tree,
 		      u64 logical, struct page *page)
 {
@@ -3003,7 +3070,7 @@ int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf)
 	return read_one_dev(root, buf, dev_item);
 }
 
-int btrfs_read_sys_array(struct btrfs_root *root)
+int btrfs_read_sys_array(struct btrfs_root *root, u64 sb_bytenr)
 {
 	struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
 	struct extent_buffer *sb;
@@ -3018,7 +3085,7 @@ int btrfs_read_sys_array(struct btrfs_root *root)
 	u32 cur;
 	struct btrfs_key key;
 
-	sb = btrfs_find_create_tree_block(root, BTRFS_SUPER_INFO_OFFSET,
+	sb = btrfs_find_create_tree_block(root, sb_bytenr,
 					  BTRFS_SUPER_INFO_SIZE);
 	if (!sb)
 		return -ENOMEM;
diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h
index fcbdcb3ae13e..bdebe83c3195 100644
--- a/fs/btrfs/volumes.h
+++ b/fs/btrfs/volumes.h
@@ -28,7 +28,6 @@ struct btrfs_device {
 	struct list_head dev_alloc_list;
 	struct btrfs_fs_devices *fs_devices;
 	struct btrfs_root *dev_root;
-	struct buffer_head *pending_io;
 	struct bio *pending_bios;
 	struct bio *pending_bio_tail;
 	int running_pending;
@@ -125,7 +124,10 @@ int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
 int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
 		    u64 logical, u64 *length,
 		    struct btrfs_multi_bio **multi_ret, int mirror_num);
-int btrfs_read_sys_array(struct btrfs_root *root);
+int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree,
+		     u64 chunk_start, u64 physical, u64 devid,
+		     u64 **logical, int *naddrs, int *stripe_len);
+int btrfs_read_sys_array(struct btrfs_root *root, u64 sb_bytenr);
 int btrfs_read_chunk_tree(struct btrfs_root *root);
 int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *extent_root, u64 type);
-- 
cgit v1.2.3