Merge commit 'v2.6.35' into kbuild/kbuild

Conflicts:
	arch/powerpc/Makefile

75 files changed, 6095 insertions, 3910 deletions

diff --git a/fs/ocfs2/Makefile b/fs/ocfs2/Makefile
index 600d2d2ade11..07d9fd854350 100644
--- a/fs/ocfs2/Makefile
+++ b/fs/ocfs2/Makefile
@@ -29,6 +29,7 @@ ocfs2-objs := \
 	mmap.o 			\
 	namei.o 		\
 	refcounttree.o		\
+	reservations.o		\
 	resize.o		\
 	slot_map.o 		\
 	suballoc.o 		\
@@ -46,6 +47,7 @@ ocfs2_stackglue-objs := stackglue.o
 ocfs2_stack_o2cb-objs := stack_o2cb.o
 ocfs2_stack_user-objs := stack_user.o
 
+obj-$(CONFIG_OCFS2_FS) += dlmfs/
 # cluster/ is always needed when OCFS2_FS for masklog support
 obj-$(CONFIG_OCFS2_FS) += cluster/
 obj-$(CONFIG_OCFS2_FS_O2CB) += dlm/
diff --git a/fs/ocfs2/acl.c b/fs/ocfs2/acl.c
index 0501974bedd0..da702294d7e7 100644
--- a/fs/ocfs2/acl.c
+++ b/fs/ocfs2/acl.c
@@ -21,6 +21,7 @@
 
 #include <linux/init.h>
 #include <linux/module.h>
+#include <linux/slab.h>
 #include <linux/string.h>
 
 #define MLOG_MASK_PREFIX ML_INODE
@@ -30,6 +31,8 @@
 #include "alloc.h"
 #include "dlmglue.h"
 #include "file.h"
+#include "inode.h"
+#include "journal.h"
 #include "ocfs2_fs.h"
 
 #include "xattr.h"
@@ -166,6 +169,60 @@ static struct posix_acl *ocfs2_get_acl(struct inode *inode, int type)
 }
 
 /*
+ * Helper function to set i_mode in memory and disk. Some call paths
+ * will not have di_bh or a journal handle to pass, in which case it
+ * will create it's own.
+ */
+static int ocfs2_acl_set_mode(struct inode *inode, struct buffer_head *di_bh,
+			      handle_t *handle, umode_t new_mode)
+{
+	int ret, commit_handle = 0;
+	struct ocfs2_dinode *di;
+
+	if (di_bh == NULL) {
+		ret = ocfs2_read_inode_block(inode, &di_bh);
+		if (ret) {
+			mlog_errno(ret);
+			goto out;
+		}
+	} else
+		get_bh(di_bh);
+
+	if (handle == NULL) {
+		handle = ocfs2_start_trans(OCFS2_SB(inode->i_sb),
+					   OCFS2_INODE_UPDATE_CREDITS);
+		if (IS_ERR(handle)) {
+			ret = PTR_ERR(handle);
+			mlog_errno(ret);
+			goto out_brelse;
+		}
+
+		commit_handle = 1;
+	}
+
+	di = (struct ocfs2_dinode *)di_bh->b_data;
+	ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (ret) {
+		mlog_errno(ret);
+		goto out_commit;
+	}
+
+	inode->i_mode = new_mode;
+	di->i_mode = cpu_to_le16(inode->i_mode);
+
+	ocfs2_journal_dirty(handle, di_bh);
+
+out_commit:
+	if (commit_handle)
+		ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle);
+out_brelse:
+	brelse(di_bh);
+out:
+	return ret;
+}
+
+/*
  * Set the access or default ACL of an inode.
  */
 static int ocfs2_set_acl(handle_t *handle,
@@ -193,9 +250,14 @@ static int ocfs2_set_acl(handle_t *handle,
 			if (ret < 0)
 				return ret;
 			else {
-				inode->i_mode = mode;
 				if (ret == 0)
 					acl = NULL;
+
+				ret = ocfs2_acl_set_mode(inode, di_bh,
+							 handle, mode);
+				if (ret)
+					return ret;
+
 			}
 		}
 		break;
@@ -283,6 +345,7 @@ int ocfs2_init_acl(handle_t *handle,
 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 	struct posix_acl *acl = NULL;
 	int ret = 0;
+	mode_t mode;
 
 	if (!S_ISLNK(inode->i_mode)) {
 		if (osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) {
@@ -291,12 +354,17 @@ int ocfs2_init_acl(handle_t *handle,
 			if (IS_ERR(acl))
 				return PTR_ERR(acl);
 		}
-		if (!acl)
-			inode->i_mode &= ~current_umask();
+		if (!acl) {
+			mode = inode->i_mode & ~current_umask();
+			ret = ocfs2_acl_set_mode(inode, di_bh, handle, mode);
+			if (ret) {
+				mlog_errno(ret);
+				goto cleanup;
+			}
+		}
 	}
 	if ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) && acl) {
 		struct posix_acl *clone;
-		mode_t mode;
 
 		if (S_ISDIR(inode->i_mode)) {
 			ret = ocfs2_set_acl(handle, inode, di_bh,
@@ -313,7 +381,7 @@ int ocfs2_init_acl(handle_t *handle,
 		mode = inode->i_mode;
 		ret = posix_acl_create_masq(clone, &mode);
 		if (ret >= 0) {
-			inode->i_mode = mode;
+			ret = ocfs2_acl_set_mode(inode, di_bh, handle, mode);
 			if (ret > 0) {
 				ret = ocfs2_set_acl(handle, inode,
 						    di_bh, ACL_TYPE_ACCESS,
@@ -421,7 +489,7 @@ cleanup:
 	return ret;
 }
 
-struct xattr_handler ocfs2_xattr_acl_access_handler = {
+const struct xattr_handler ocfs2_xattr_acl_access_handler = {
 	.prefix	= POSIX_ACL_XATTR_ACCESS,
 	.flags	= ACL_TYPE_ACCESS,
 	.list	= ocfs2_xattr_list_acl_access,
@@ -429,7 +497,7 @@ struct xattr_handler ocfs2_xattr_acl_access_handler = {
 	.set	= ocfs2_xattr_set_acl,
 };
 
-struct xattr_handler ocfs2_xattr_acl_default_handler = {
+const struct xattr_handler ocfs2_xattr_acl_default_handler = {
 	.prefix	= POSIX_ACL_XATTR_DEFAULT,
 	.flags	= ACL_TYPE_DEFAULT,
 	.list	= ocfs2_xattr_list_acl_default,
diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
index d17bdc718f74..215e12ce1d85 100644
--- a/fs/ocfs2/alloc.c
+++ b/fs/ocfs2/alloc.c
@@ -1006,7 +1006,7 @@ static int ocfs2_create_new_meta_bhs(handle_t *handle,
 	int count, status, i;
 	u16 suballoc_bit_start;
 	u32 num_got;
-	u64 first_blkno;
+	u64 suballoc_loc, first_blkno;
 	struct ocfs2_super *osb =
 		OCFS2_SB(ocfs2_metadata_cache_get_super(et->et_ci));
 	struct ocfs2_extent_block *eb;
@@ -1015,10 +1015,10 @@ static int ocfs2_create_new_meta_bhs(handle_t *handle,
 
 	count = 0;
 	while (count < wanted) {
-		status = ocfs2_claim_metadata(osb,
-					      handle,
+		status = ocfs2_claim_metadata(handle,
 					      meta_ac,
 					      wanted - count,
+					      &suballoc_loc,
 					      &suballoc_bit_start,
 					      &num_got,
 					      &first_blkno);
@@ -1050,7 +1050,9 @@ static int ocfs2_create_new_meta_bhs(handle_t *handle,
 			strcpy(eb->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE);
 			eb->h_blkno = cpu_to_le64(first_blkno);
 			eb->h_fs_generation = cpu_to_le32(osb->fs_generation);
-			eb->h_suballoc_slot = cpu_to_le16(osb->slot_num);
+			eb->h_suballoc_slot =
+				cpu_to_le16(meta_ac->ac_alloc_slot);
+			eb->h_suballoc_loc = cpu_to_le64(suballoc_loc);
 			eb->h_suballoc_bit = cpu_to_le16(suballoc_bit_start);
 			eb->h_list.l_count =
 				cpu_to_le16(ocfs2_extent_recs_per_eb(osb->sb));
@@ -1060,11 +1062,7 @@ static int ocfs2_create_new_meta_bhs(handle_t *handle,
 
 			/* We'll also be dirtied by the caller, so
 			 * this isn't absolutely necessary. */
-			status = ocfs2_journal_dirty(handle, bhs[i]);
-			if (status < 0) {
-				mlog_errno(status);
-				goto bail;
-			}
+			ocfs2_journal_dirty(handle, bhs[i]);
 		}
 
 		count += num_got;
@@ -1128,8 +1126,7 @@ static int ocfs2_adjust_rightmost_branch(handle_t *handle,
 		goto out;
 	}
 
-	status = ocfs2_extend_trans(handle, path_num_items(path) +
-				    handle->h_buffer_credits);
+	status = ocfs2_extend_trans(handle, path_num_items(path));
 	if (status < 0) {
 		mlog_errno(status);
 		goto out;
@@ -1269,12 +1266,7 @@ static int ocfs2_add_branch(handle_t *handle,
 		if (!eb_el->l_tree_depth)
 			new_last_eb_blk = le64_to_cpu(eb->h_blkno);
 
-		status = ocfs2_journal_dirty(handle, bh);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
-
+		ocfs2_journal_dirty(handle, bh);
 		next_blkno = le64_to_cpu(eb->h_blkno);
 	}
 
@@ -1320,17 +1312,10 @@ static int ocfs2_add_branch(handle_t *handle,
 	eb = (struct ocfs2_extent_block *) (*last_eb_bh)->b_data;
 	eb->h_next_leaf_blk = cpu_to_le64(new_last_eb_blk);
 
-	status = ocfs2_journal_dirty(handle, *last_eb_bh);
-	if (status < 0)
-		mlog_errno(status);
-	status = ocfs2_journal_dirty(handle, et->et_root_bh);
-	if (status < 0)
-		mlog_errno(status);
-	if (eb_bh) {
-		status = ocfs2_journal_dirty(handle, eb_bh);
-		if (status < 0)
-			mlog_errno(status);
-	}
+	ocfs2_journal_dirty(handle, *last_eb_bh);
+	ocfs2_journal_dirty(handle, et->et_root_bh);
+	if (eb_bh)
+		ocfs2_journal_dirty(handle, eb_bh);
 
 	/*
 	 * Some callers want to track the rightmost leaf so pass it
@@ -1398,11 +1383,7 @@ static int ocfs2_shift_tree_depth(handle_t *handle,
 	for (i = 0; i < le16_to_cpu(root_el->l_next_free_rec); i++)
 		eb_el->l_recs[i] = root_el->l_recs[i];
 
-	status = ocfs2_journal_dirty(handle, new_eb_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	ocfs2_journal_dirty(handle, new_eb_bh);
 
 	status = ocfs2_et_root_journal_access(handle, et,
 					      OCFS2_JOURNAL_ACCESS_WRITE);
@@ -1427,11 +1408,7 @@ static int ocfs2_shift_tree_depth(handle_t *handle,
 	if (root_el->l_tree_depth == cpu_to_le16(1))
 		ocfs2_et_set_last_eb_blk(et, le64_to_cpu(eb->h_blkno));
 
-	status = ocfs2_journal_dirty(handle, et->et_root_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	ocfs2_journal_dirty(handle, et->et_root_bh);
 
 	*ret_new_eb_bh = new_eb_bh;
 	new_eb_bh = NULL;
@@ -2063,7 +2040,7 @@ static void ocfs2_complete_edge_insert(handle_t *handle,
 				       struct ocfs2_path *right_path,
 				       int subtree_index)
 {
-	int ret, i, idx;
+	int i, idx;
 	struct ocfs2_extent_list *el, *left_el, *right_el;
 	struct ocfs2_extent_rec *left_rec, *right_rec;
 	struct buffer_head *root_bh = left_path->p_node[subtree_index].bh;
@@ -2101,13 +2078,8 @@ static void ocfs2_complete_edge_insert(handle_t *handle,
 		ocfs2_adjust_adjacent_records(left_rec, left_el, right_rec,
 					      right_el);
 
-		ret = ocfs2_journal_dirty(handle, left_path->p_node[i].bh);
-		if (ret)
-			mlog_errno(ret);
-
-		ret = ocfs2_journal_dirty(handle, right_path->p_node[i].bh);
-		if (ret)
-			mlog_errno(ret);
+		ocfs2_journal_dirty(handle, left_path->p_node[i].bh);
+		ocfs2_journal_dirty(handle, right_path->p_node[i].bh);
 
 		/*
 		 * Setup our list pointers now so that the current
@@ -2131,9 +2103,7 @@ static void ocfs2_complete_edge_insert(handle_t *handle,
 
 	root_bh = left_path->p_node[subtree_index].bh;
 
-	ret = ocfs2_journal_dirty(handle, root_bh);
-	if (ret)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, root_bh);
 }
 
 static int ocfs2_rotate_subtree_right(handle_t *handle,
@@ -2206,11 +2176,7 @@ static int ocfs2_rotate_subtree_right(handle_t *handle,
 
 	ocfs2_create_empty_extent(right_el);
 
-	ret = ocfs2_journal_dirty(handle, right_leaf_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out;
-	}
+	ocfs2_journal_dirty(handle, right_leaf_bh);
 
 	/* Do the copy now. */
 	i = le16_to_cpu(left_el->l_next_free_rec) - 1;
@@ -2229,11 +2195,7 @@ static int ocfs2_rotate_subtree_right(handle_t *handle,
 	memset(&left_el->l_recs[0], 0, sizeof(struct ocfs2_extent_rec));
 	le16_add_cpu(&left_el->l_next_free_rec, 1);
 
-	ret = ocfs2_journal_dirty(handle, left_leaf_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out;
-	}
+	ocfs2_journal_dirty(handle, left_leaf_bh);
 
 	ocfs2_complete_edge_insert(handle, left_path, right_path,
 				   subtree_index);
@@ -2248,8 +2210,8 @@ out:
  *
  * Will return zero if the path passed in is already the leftmost path.
  */
-static int ocfs2_find_cpos_for_left_leaf(struct super_block *sb,
-					 struct ocfs2_path *path, u32 *cpos)
+int ocfs2_find_cpos_for_left_leaf(struct super_block *sb,
+				  struct ocfs2_path *path, u32 *cpos)
 {
 	int i, j, ret = 0;
 	u64 blkno;
@@ -2326,20 +2288,14 @@ static int ocfs2_extend_rotate_transaction(handle_t *handle, int subtree_depth,
 					   int op_credits,
 					   struct ocfs2_path *path)
 {
-	int ret;
+	int ret = 0;
 	int credits = (path->p_tree_depth - subtree_depth) * 2 + 1 + op_credits;
 
-	if (handle->h_buffer_credits < credits) {
+	if (handle->h_buffer_credits < credits)
 		ret = ocfs2_extend_trans(handle,
 					 credits - handle->h_buffer_credits);
-		if (ret)
-			return ret;
 
-		if (unlikely(handle->h_buffer_credits < credits))
-			return ocfs2_extend_trans(handle, credits);
-	}
-
-	return 0;
+	return ret;
 }
 
 /*
@@ -2583,8 +2539,7 @@ static int ocfs2_update_edge_lengths(handle_t *handle,
 	 * records for all the bh in the path.
 	 * So we have to allocate extra credits and access them.
 	 */
-	ret = ocfs2_extend_trans(handle,
-				 handle->h_buffer_credits + subtree_index);
+	ret = ocfs2_extend_trans(handle, subtree_index);
 	if (ret) {
 		mlog_errno(ret);
 		goto out;
@@ -2822,12 +2777,8 @@ static int ocfs2_rotate_subtree_left(handle_t *handle,
 		ocfs2_remove_empty_extent(right_leaf_el);
 	}
 
-	ret = ocfs2_journal_dirty(handle, path_leaf_bh(left_path));
-	if (ret)
-		mlog_errno(ret);
-	ret = ocfs2_journal_dirty(handle, path_leaf_bh(right_path));
-	if (ret)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, path_leaf_bh(left_path));
+	ocfs2_journal_dirty(handle, path_leaf_bh(right_path));
 
 	if (del_right_subtree) {
 		ocfs2_unlink_subtree(handle, et, left_path, right_path,
@@ -2850,9 +2801,7 @@ static int ocfs2_rotate_subtree_left(handle_t *handle,
 		if (right_has_empty)
 			ocfs2_remove_empty_extent(left_leaf_el);
 
-		ret = ocfs2_journal_dirty(handle, et_root_bh);
-		if (ret)
-			mlog_errno(ret);
+		ocfs2_journal_dirty(handle, et_root_bh);
 
 		*deleted = 1;
 	} else
@@ -2961,10 +2910,7 @@ static int ocfs2_rotate_rightmost_leaf_left(handle_t *handle,
 	}
 
 	ocfs2_remove_empty_extent(el);
-
-	ret = ocfs2_journal_dirty(handle, bh);
-	if (ret)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, bh);
 
 out:
 	return ret;
@@ -3505,15 +3451,9 @@ static int ocfs2_merge_rec_right(struct ocfs2_path *left_path,
 
 	ocfs2_cleanup_merge(el, index);
 
-	ret = ocfs2_journal_dirty(handle, bh);
-	if (ret)
-		mlog_errno(ret);
-
+	ocfs2_journal_dirty(handle, bh);
 	if (right_path) {
-		ret = ocfs2_journal_dirty(handle, path_leaf_bh(right_path));
-		if (ret)
-			mlog_errno(ret);
-
+		ocfs2_journal_dirty(handle, path_leaf_bh(right_path));
 		ocfs2_complete_edge_insert(handle, left_path, right_path,
 					   subtree_index);
 	}
@@ -3682,14 +3622,9 @@ static int ocfs2_merge_rec_left(struct ocfs2_path *right_path,
 
 	ocfs2_cleanup_merge(el, index);
 
-	ret = ocfs2_journal_dirty(handle, bh);
-	if (ret)
-		mlog_errno(ret);
-
+	ocfs2_journal_dirty(handle, bh);
 	if (left_path) {
-		ret = ocfs2_journal_dirty(handle, path_leaf_bh(left_path));
-		if (ret)
-			mlog_errno(ret);
+		ocfs2_journal_dirty(handle, path_leaf_bh(left_path));
 
 		/*
 		 * In the situation that the right_rec is empty and the extent
@@ -4015,10 +3950,7 @@ static void ocfs2_adjust_rightmost_records(handle_t *handle,
 		le32_add_cpu(&rec->e_int_clusters,
 			     -le32_to_cpu(rec->e_cpos));
 
-		ret = ocfs2_journal_dirty(handle, bh);
-		if (ret)
-			mlog_errno(ret);
-
+		ocfs2_journal_dirty(handle, bh);
 	}
 }
 
@@ -4202,17 +4134,13 @@ static int ocfs2_insert_path(handle_t *handle,
 	struct buffer_head *leaf_bh = path_leaf_bh(right_path);
 
 	if (left_path) {
-		int credits = handle->h_buffer_credits;
-
 		/*
 		 * There's a chance that left_path got passed back to
 		 * us without being accounted for in the
 		 * journal. Extend our transaction here to be sure we
 		 * can change those blocks.
 		 */
-		credits += left_path->p_tree_depth;
-
-		ret = ocfs2_extend_trans(handle, credits);
+		ret = ocfs2_extend_trans(handle, left_path->p_tree_depth);
 		if (ret < 0) {
 			mlog_errno(ret);
 			goto out;
@@ -4250,17 +4178,13 @@ static int ocfs2_insert_path(handle_t *handle,
 		 * dirty this for us.
 		 */
 		if (left_path)
-			ret = ocfs2_journal_dirty(handle,
-						  path_leaf_bh(left_path));
-			if (ret)
-				mlog_errno(ret);
+			ocfs2_journal_dirty(handle,
+					    path_leaf_bh(left_path));
 	} else
 		ocfs2_insert_at_leaf(et, insert_rec, path_leaf_el(right_path),
 				     insert);
 
-	ret = ocfs2_journal_dirty(handle, leaf_bh);
-	if (ret)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, leaf_bh);
 
 	if (left_path) {
 		/*
@@ -4383,9 +4307,7 @@ out_update_clusters:
 		ocfs2_et_update_clusters(et,
 					 le16_to_cpu(insert_rec->e_leaf_clusters));
 
-	ret = ocfs2_journal_dirty(handle, et->et_root_bh);
-	if (ret)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, et->et_root_bh);
 
 out:
 	ocfs2_free_path(left_path);
@@ -4865,7 +4787,7 @@ int ocfs2_add_clusters_in_btree(handle_t *handle,
 		goto leave;
 	}
 
-	status = __ocfs2_claim_clusters(osb, handle, data_ac, 1,
+	status = __ocfs2_claim_clusters(handle, data_ac, 1,
 					clusters_to_add, &bit_off, &num_bits);
 	if (status < 0) {
 		if (status != -ENOSPC)
@@ -4894,11 +4816,7 @@ int ocfs2_add_clusters_in_btree(handle_t *handle,
 		goto leave;
 	}
 
-	status = ocfs2_journal_dirty(handle, et->et_root_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto leave;
-	}
+	ocfs2_journal_dirty(handle, et->et_root_bh);
 
 	clusters_to_add -= num_bits;
 	*logical_offset += num_bits;
@@ -5308,7 +5226,7 @@ static int ocfs2_split_tree(handle_t *handle, struct ocfs2_extent_tree *et,
 			    int index, u32 new_range,
 			    struct ocfs2_alloc_context *meta_ac)
 {
-	int ret, depth, credits = handle->h_buffer_credits;
+	int ret, depth, credits;
 	struct buffer_head *last_eb_bh = NULL;
 	struct ocfs2_extent_block *eb;
 	struct ocfs2_extent_list *rightmost_el, *el;
@@ -5339,8 +5257,8 @@ static int ocfs2_split_tree(handle_t *handle, struct ocfs2_extent_tree *et,
 	} else
 		rightmost_el = path_leaf_el(path);
 
-	credits += path->p_tree_depth +
-		   ocfs2_extend_meta_needed(et->et_root_el);
+	credits = path->p_tree_depth +
+		  ocfs2_extend_meta_needed(et->et_root_el);
 	ret = ocfs2_extend_trans(handle, credits);
 	if (ret) {
 		mlog_errno(ret);
@@ -5670,19 +5588,97 @@ out:
 	return ret;
 }
 
+/*
+ * ocfs2_reserve_blocks_for_rec_trunc() would look basically the
+ * same as ocfs2_lock_alloctors(), except for it accepts a blocks
+ * number to reserve some extra blocks, and it only handles meta
+ * data allocations.
+ *
+ * Currently, only ocfs2_remove_btree_range() uses it for truncating
+ * and punching holes.
+ */
+static int ocfs2_reserve_blocks_for_rec_trunc(struct inode *inode,
+					      struct ocfs2_extent_tree *et,
+					      u32 extents_to_split,
+					      struct ocfs2_alloc_context **ac,
+					      int extra_blocks)
+{
+	int ret = 0, num_free_extents;
+	unsigned int max_recs_needed = 2 * extents_to_split;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+	*ac = NULL;
+
+	num_free_extents = ocfs2_num_free_extents(osb, et);
+	if (num_free_extents < 0) {
+		ret = num_free_extents;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	if (!num_free_extents ||
+	    (ocfs2_sparse_alloc(osb) && num_free_extents < max_recs_needed))
+		extra_blocks += ocfs2_extend_meta_needed(et->et_root_el);
+
+	if (extra_blocks) {
+		ret = ocfs2_reserve_new_metadata_blocks(osb, extra_blocks, ac);
+		if (ret < 0) {
+			if (ret != -ENOSPC)
+				mlog_errno(ret);
+			goto out;
+		}
+	}
+
+out:
+	if (ret) {
+		if (*ac) {
+			ocfs2_free_alloc_context(*ac);
+			*ac = NULL;
+		}
+	}
+
+	return ret;
+}
+
 int ocfs2_remove_btree_range(struct inode *inode,
 			     struct ocfs2_extent_tree *et,
-			     u32 cpos, u32 phys_cpos, u32 len,
-			     struct ocfs2_cached_dealloc_ctxt *dealloc)
+			     u32 cpos, u32 phys_cpos, u32 len, int flags,
+			     struct ocfs2_cached_dealloc_ctxt *dealloc,
+			     u64 refcount_loc)
 {
-	int ret;
+	int ret, credits = 0, extra_blocks = 0;
 	u64 phys_blkno = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos);
 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 	struct inode *tl_inode = osb->osb_tl_inode;
 	handle_t *handle;
 	struct ocfs2_alloc_context *meta_ac = NULL;
+	struct ocfs2_refcount_tree *ref_tree = NULL;
+
+	if ((flags & OCFS2_EXT_REFCOUNTED) && len) {
+		BUG_ON(!(OCFS2_I(inode)->ip_dyn_features &
+			 OCFS2_HAS_REFCOUNT_FL));
+
+		ret = ocfs2_lock_refcount_tree(osb, refcount_loc, 1,
+					       &ref_tree, NULL);
+		if (ret) {
+			mlog_errno(ret);
+			goto out;
+		}
 
-	ret = ocfs2_lock_allocators(inode, et, 0, 1, NULL, &meta_ac);
+		ret = ocfs2_prepare_refcount_change_for_del(inode,
+							    refcount_loc,
+							    phys_blkno,
+							    len,
+							    &credits,
+							    &extra_blocks);
+		if (ret < 0) {
+			mlog_errno(ret);
+			goto out;
+		}
+	}
+
+	ret = ocfs2_reserve_blocks_for_rec_trunc(inode, et, 1, &meta_ac,
+						 extra_blocks);
 	if (ret) {
 		mlog_errno(ret);
 		return ret;
@@ -5698,7 +5694,8 @@ int ocfs2_remove_btree_range(struct inode *inode,
 		}
 	}
 
-	handle = ocfs2_start_trans(osb, ocfs2_remove_extent_credits(osb->sb));
+	handle = ocfs2_start_trans(osb,
+			ocfs2_remove_extent_credits(osb->sb) + credits);
 	if (IS_ERR(handle)) {
 		ret = PTR_ERR(handle);
 		mlog_errno(ret);
@@ -5712,7 +5709,7 @@ int ocfs2_remove_btree_range(struct inode *inode,
 		goto out;
 	}
 
-	vfs_dq_free_space_nodirty(inode,
+	dquot_free_space_nodirty(inode,
 				  ocfs2_clusters_to_bytes(inode->i_sb, len));
 
 	ret = ocfs2_remove_extent(handle, et, cpos, len, meta_ac, dealloc);
@@ -5723,15 +5720,22 @@ int ocfs2_remove_btree_range(struct inode *inode,
 
 	ocfs2_et_update_clusters(et, -len);
 
-	ret = ocfs2_journal_dirty(handle, et->et_root_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out_commit;
-	}
+	ocfs2_journal_dirty(handle, et->et_root_bh);
 
-	ret = ocfs2_truncate_log_append(osb, handle, phys_blkno, len);
-	if (ret)
-		mlog_errno(ret);
+	if (phys_blkno) {
+		if (flags & OCFS2_EXT_REFCOUNTED)
+			ret = ocfs2_decrease_refcount(inode, handle,
+					ocfs2_blocks_to_clusters(osb->sb,
+								 phys_blkno),
+					len, meta_ac,
+					dealloc, 1);
+		else
+			ret = ocfs2_truncate_log_append(osb, handle,
+							phys_blkno, len);
+		if (ret)
+			mlog_errno(ret);
+
+	}
 
 out_commit:
 	ocfs2_commit_trans(osb, handle);
@@ -5741,6 +5745,9 @@ out:
 	if (meta_ac)
 		ocfs2_free_alloc_context(meta_ac);
 
+	if (ref_tree)
+		ocfs2_unlock_refcount_tree(osb, ref_tree, 1);
+
 	return ret;
 }
 
@@ -5849,11 +5856,7 @@ int ocfs2_truncate_log_append(struct ocfs2_super *osb,
 	}
 	tl->tl_recs[index].t_clusters = cpu_to_le32(num_clusters);
 
-	status = ocfs2_journal_dirty(handle, tl_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	ocfs2_journal_dirty(handle, tl_bh);
 
 bail:
 	mlog_exit(status);
@@ -5892,11 +5895,7 @@ static int ocfs2_replay_truncate_records(struct ocfs2_super *osb,
 
 		tl->tl_used = cpu_to_le16(i);
 
-		status = ocfs2_journal_dirty(handle, tl_bh);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
+		ocfs2_journal_dirty(handle, tl_bh);
 
 		/* TODO: Perhaps we can calculate the bulk of the
 		 * credits up front rather than extending like
@@ -6037,7 +6036,7 @@ static void ocfs2_truncate_log_worker(struct work_struct *work)
 	if (status < 0)
 		mlog_errno(status);
 	else
-		ocfs2_init_inode_steal_slot(osb);
+		ocfs2_init_steal_slots(osb);
 
 	mlog_exit(status);
 }
@@ -6297,6 +6296,7 @@ int ocfs2_truncate_log_init(struct ocfs2_super *osb)
  */
 struct ocfs2_cached_block_free {
 	struct ocfs2_cached_block_free		*free_next;
+	u64					free_bg;
 	u64					free_blk;
 	unsigned int				free_bit;
 };
@@ -6343,8 +6343,11 @@ static int ocfs2_free_cached_blocks(struct ocfs2_super *osb,
 	}
 
 	while (head) {
-		bg_blkno = ocfs2_which_suballoc_group(head->free_blk,
-						      head->free_bit);
+		if (head->free_bg)
+			bg_blkno = head->free_bg;
+		else
+			bg_blkno = ocfs2_which_suballoc_group(head->free_blk,
+							      head->free_bit);
 		mlog(0, "Free bit: (bit %u, blkno %llu)\n",
 		     head->free_bit, (unsigned long long)head->free_blk);
 
@@ -6392,7 +6395,7 @@ int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
 	int ret = 0;
 	struct ocfs2_cached_block_free *item;
 
-	item = kmalloc(sizeof(*item), GFP_NOFS);
+	item = kzalloc(sizeof(*item), GFP_NOFS);
 	if (item == NULL) {
 		ret = -ENOMEM;
 		mlog_errno(ret);
@@ -6532,8 +6535,8 @@ ocfs2_find_per_slot_free_list(int type,
 }
 
 int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
-			      int type, int slot, u64 blkno,
-			      unsigned int bit)
+			      int type, int slot, u64 suballoc,
+			      u64 blkno, unsigned int bit)
 {
 	int ret;
 	struct ocfs2_per_slot_free_list *fl;
@@ -6546,7 +6549,7 @@ int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
 		goto out;
 	}
 
-	item = kmalloc(sizeof(*item), GFP_NOFS);
+	item = kzalloc(sizeof(*item), GFP_NOFS);
 	if (item == NULL) {
 		ret = -ENOMEM;
 		mlog_errno(ret);
@@ -6556,6 +6559,7 @@ int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
 	mlog(0, "Insert: (type %d, slot %u, bit %u, blk %llu)\n",
 	     type, slot, bit, (unsigned long long)blkno);
 
+	item->free_bg = suballoc;
 	item->free_blk = blkno;
 	item->free_bit = bit;
 	item->free_next = fl->f_first;
@@ -6572,433 +6576,11 @@ static int ocfs2_cache_extent_block_free(struct ocfs2_cached_dealloc_ctxt *ctxt,
 {
 	return ocfs2_cache_block_dealloc(ctxt, EXTENT_ALLOC_SYSTEM_INODE,
 					 le16_to_cpu(eb->h_suballoc_slot),
+					 le64_to_cpu(eb->h_suballoc_loc),
 					 le64_to_cpu(eb->h_blkno),
 					 le16_to_cpu(eb->h_suballoc_bit));
 }
 
-/* This function will figure out whether the currently last extent
- * block will be deleted, and if it will, what the new last extent
- * block will be so we can update his h_next_leaf_blk field, as well
- * as the dinodes i_last_eb_blk */
-static int ocfs2_find_new_last_ext_blk(struct inode *inode,
-				       unsigned int clusters_to_del,
-				       struct ocfs2_path *path,
-				       struct buffer_head **new_last_eb)
-{
-	int next_free, ret = 0;
-	u32 cpos;
-	struct ocfs2_extent_rec *rec;
-	struct ocfs2_extent_block *eb;
-	struct ocfs2_extent_list *el;
-	struct buffer_head *bh = NULL;
-
-	*new_last_eb = NULL;
-
-	/* we have no tree, so of course, no last_eb. */
-	if (!path->p_tree_depth)
-		goto out;
-
-	/* trunc to zero special case - this makes tree_depth = 0
-	 * regardless of what it is.  */
-	if (OCFS2_I(inode)->ip_clusters == clusters_to_del)
-		goto out;
-
-	el = path_leaf_el(path);
-	BUG_ON(!el->l_next_free_rec);
-
-	/*
-	 * Make sure that this extent list will actually be empty
-	 * after we clear away the data. We can shortcut out if
-	 * there's more than one non-empty extent in the
-	 * list. Otherwise, a check of the remaining extent is
-	 * necessary.
-	 */
-	next_free = le16_to_cpu(el->l_next_free_rec);
-	rec = NULL;
-	if (ocfs2_is_empty_extent(&el->l_recs[0])) {
-		if (next_free > 2)
-			goto out;
-
-		/* We may have a valid extent in index 1, check it. */
-		if (next_free == 2)
-			rec = &el->l_recs[1];
-
-		/*
-		 * Fall through - no more nonempty extents, so we want
-		 * to delete this leaf.
-		 */
-	} else {
-		if (next_free > 1)
-			goto out;
-
-		rec = &el->l_recs[0];
-	}
-
-	if (rec) {
-		/*
-		 * Check it we'll only be trimming off the end of this
-		 * cluster.
-		 */
-		if (le16_to_cpu(rec->e_leaf_clusters) > clusters_to_del)
-			goto out;
-	}
-
-	ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb, path, &cpos);
-	if (ret) {
-		mlog_errno(ret);
-		goto out;
-	}
-
-	ret = ocfs2_find_leaf(INODE_CACHE(inode), path_root_el(path), cpos, &bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out;
-	}
-
-	eb = (struct ocfs2_extent_block *) bh->b_data;
-	el = &eb->h_list;
-
-	/* ocfs2_find_leaf() gets the eb from ocfs2_read_extent_block().
-	 * Any corruption is a code bug. */
-	BUG_ON(!OCFS2_IS_VALID_EXTENT_BLOCK(eb));
-
-	*new_last_eb = bh;
-	get_bh(*new_last_eb);
-	mlog(0, "returning block %llu, (cpos: %u)\n",
-	     (unsigned long long)le64_to_cpu(eb->h_blkno), cpos);
-out:
-	brelse(bh);
-
-	return ret;
-}
-
-/*
- * Trim some clusters off the rightmost edge of a tree. Only called
- * during truncate.
- *
- * The caller needs to:
- *   - start journaling of each path component.
- *   - compute and fully set up any new last ext block
- */
-static int ocfs2_trim_tree(struct inode *inode, struct ocfs2_path *path,
-			   handle_t *handle, struct ocfs2_truncate_context *tc,
-			   u32 clusters_to_del, u64 *delete_start, u8 *flags)
-{
-	int ret, i, index = path->p_tree_depth;
-	u32 new_edge = 0;
-	u64 deleted_eb = 0;
-	struct buffer_head *bh;
-	struct ocfs2_extent_list *el;
-	struct ocfs2_extent_rec *rec;
-
-	*delete_start = 0;
-	*flags = 0;
-
-	while (index >= 0) {
-		bh = path->p_node[index].bh;
-		el = path->p_node[index].el;
-
-		mlog(0, "traveling tree (index = %d, block = %llu)\n",
-		     index,  (unsigned long long)bh->b_blocknr);
-
-		BUG_ON(le16_to_cpu(el->l_next_free_rec) == 0);
-
-		if (index !=
-		    (path->p_tree_depth - le16_to_cpu(el->l_tree_depth))) {
-			ocfs2_error(inode->i_sb,
-				    "Inode %lu has invalid ext. block %llu",
-				    inode->i_ino,
-				    (unsigned long long)bh->b_blocknr);
-			ret = -EROFS;
-			goto out;
-		}
-
-find_tail_record:
-		i = le16_to_cpu(el->l_next_free_rec) - 1;
-		rec = &el->l_recs[i];
-
-		mlog(0, "Extent list before: record %d: (%u, %u, %llu), "
-		     "next = %u\n", i, le32_to_cpu(rec->e_cpos),
-		     ocfs2_rec_clusters(el, rec),
-		     (unsigned long long)le64_to_cpu(rec->e_blkno),
-		     le16_to_cpu(el->l_next_free_rec));
-
-		BUG_ON(ocfs2_rec_clusters(el, rec) < clusters_to_del);
-
-		if (le16_to_cpu(el->l_tree_depth) == 0) {
-			/*
-			 * If the leaf block contains a single empty
-			 * extent and no records, we can just remove
-			 * the block.
-			 */
-			if (i == 0 && ocfs2_is_empty_extent(rec)) {
-				memset(rec, 0,
-				       sizeof(struct ocfs2_extent_rec));
-				el->l_next_free_rec = cpu_to_le16(0);
-
-				goto delete;
-			}
-
-			/*
-			 * Remove any empty extents by shifting things
-			 * left. That should make life much easier on
-			 * the code below. This condition is rare
-			 * enough that we shouldn't see a performance
-			 * hit.
-			 */
-			if (ocfs2_is_empty_extent(&el->l_recs[0])) {
-				le16_add_cpu(&el->l_next_free_rec, -1);
-
-				for(i = 0;
-				    i < le16_to_cpu(el->l_next_free_rec); i++)
-					el->l_recs[i] = el->l_recs[i + 1];
-
-				memset(&el->l_recs[i], 0,
-				       sizeof(struct ocfs2_extent_rec));
-
-				/*
-				 * We've modified our extent list. The
-				 * simplest way to handle this change
-				 * is to being the search from the
-				 * start again.
-				 */
-				goto find_tail_record;
-			}
-
-			le16_add_cpu(&rec->e_leaf_clusters, -clusters_to_del);
-
-			/*
-			 * We'll use "new_edge" on our way back up the
-			 * tree to know what our rightmost cpos is.
-			 */
-			new_edge = le16_to_cpu(rec->e_leaf_clusters);
-			new_edge += le32_to_cpu(rec->e_cpos);
-
-			/*
-			 * The caller will use this to delete data blocks.
-			 */
-			*delete_start = le64_to_cpu(rec->e_blkno)
-				+ ocfs2_clusters_to_blocks(inode->i_sb,
-					le16_to_cpu(rec->e_leaf_clusters));
-			*flags = rec->e_flags;
-
-			/*
-			 * If it's now empty, remove this record.
-			 */
-			if (le16_to_cpu(rec->e_leaf_clusters) == 0) {
-				memset(rec, 0,
-				       sizeof(struct ocfs2_extent_rec));
-				le16_add_cpu(&el->l_next_free_rec, -1);
-			}
-		} else {
-			if (le64_to_cpu(rec->e_blkno) == deleted_eb) {
-				memset(rec, 0,
-				       sizeof(struct ocfs2_extent_rec));
-				le16_add_cpu(&el->l_next_free_rec, -1);
-
-				goto delete;
-			}
-
-			/* Can this actually happen? */
-			if (le16_to_cpu(el->l_next_free_rec) == 0)
-				goto delete;
-
-			/*
-			 * We never actually deleted any clusters
-			 * because our leaf was empty. There's no
-			 * reason to adjust the rightmost edge then.
-			 */
-			if (new_edge == 0)
-				goto delete;
-
-			rec->e_int_clusters = cpu_to_le32(new_edge);
-			le32_add_cpu(&rec->e_int_clusters,
-				     -le32_to_cpu(rec->e_cpos));
-
-			 /*
-			  * A deleted child record should have been
-			  * caught above.
-			  */
-			 BUG_ON(le32_to_cpu(rec->e_int_clusters) == 0);
-		}
-
-delete:
-		ret = ocfs2_journal_dirty(handle, bh);
-		if (ret) {
-			mlog_errno(ret);
-			goto out;
-		}
-
-		mlog(0, "extent list container %llu, after: record %d: "
-		     "(%u, %u, %llu), next = %u.\n",
-		     (unsigned long long)bh->b_blocknr, i,
-		     le32_to_cpu(rec->e_cpos), ocfs2_rec_clusters(el, rec),
-		     (unsigned long long)le64_to_cpu(rec->e_blkno),
-		     le16_to_cpu(el->l_next_free_rec));
-
-		/*
-		 * We must be careful to only attempt delete of an
-		 * extent block (and not the root inode block).
-		 */
-		if (index > 0 && le16_to_cpu(el->l_next_free_rec) == 0) {
-			struct ocfs2_extent_block *eb =
-				(struct ocfs2_extent_block *)bh->b_data;
-
-			/*
-			 * Save this for use when processing the
-			 * parent block.
-			 */
-			deleted_eb = le64_to_cpu(eb->h_blkno);
-
-			mlog(0, "deleting this extent block.\n");
-
-			ocfs2_remove_from_cache(INODE_CACHE(inode), bh);
-
-			BUG_ON(ocfs2_rec_clusters(el, &el->l_recs[0]));
-			BUG_ON(le32_to_cpu(el->l_recs[0].e_cpos));
-			BUG_ON(le64_to_cpu(el->l_recs[0].e_blkno));
-
-			ret = ocfs2_cache_extent_block_free(&tc->tc_dealloc, eb);
-			/* An error here is not fatal. */
-			if (ret < 0)
-				mlog_errno(ret);
-		} else {
-			deleted_eb = 0;
-		}
-
-		index--;
-	}
-
-	ret = 0;
-out:
-	return ret;
-}
-
-static int ocfs2_do_truncate(struct ocfs2_super *osb,
-			     unsigned int clusters_to_del,
-			     struct inode *inode,
-			     struct buffer_head *fe_bh,
-			     handle_t *handle,
-			     struct ocfs2_truncate_context *tc,
-			     struct ocfs2_path *path,
-			     struct ocfs2_alloc_context *meta_ac)
-{
-	int status;
-	struct ocfs2_dinode *fe;
-	struct ocfs2_extent_block *last_eb = NULL;
-	struct ocfs2_extent_list *el;
-	struct buffer_head *last_eb_bh = NULL;
-	u64 delete_blk = 0;
-	u8 rec_flags;
-
-	fe = (struct ocfs2_dinode *) fe_bh->b_data;
-
-	status = ocfs2_find_new_last_ext_blk(inode, clusters_to_del,
-					     path, &last_eb_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
-
-	/*
-	 * Each component will be touched, so we might as well journal
-	 * here to avoid having to handle errors later.
-	 */
-	status = ocfs2_journal_access_path(INODE_CACHE(inode), handle, path);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
-
-	if (last_eb_bh) {
-		status = ocfs2_journal_access_eb(handle, INODE_CACHE(inode), last_eb_bh,
-						 OCFS2_JOURNAL_ACCESS_WRITE);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
-
-		last_eb = (struct ocfs2_extent_block *) last_eb_bh->b_data;
-	}
-
-	el = &(fe->id2.i_list);
-
-	/*
-	 * Lower levels depend on this never happening, but it's best
-	 * to check it up here before changing the tree.
-	 */
-	if (el->l_tree_depth && el->l_recs[0].e_int_clusters == 0) {
-		ocfs2_error(inode->i_sb,
-			    "Inode %lu has an empty extent record, depth %u\n",
-			    inode->i_ino, le16_to_cpu(el->l_tree_depth));
-		status = -EROFS;
-		goto bail;
-	}
-
-	vfs_dq_free_space_nodirty(inode,
-			ocfs2_clusters_to_bytes(osb->sb, clusters_to_del));
-	spin_lock(&OCFS2_I(inode)->ip_lock);
-	OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters) -
-				      clusters_to_del;
-	spin_unlock(&OCFS2_I(inode)->ip_lock);
-	le32_add_cpu(&fe->i_clusters, -clusters_to_del);
-	inode->i_blocks = ocfs2_inode_sector_count(inode);
-
-	status = ocfs2_trim_tree(inode, path, handle, tc,
-				 clusters_to_del, &delete_blk, &rec_flags);
-	if (status) {
-		mlog_errno(status);
-		goto bail;
-	}
-
-	if (le32_to_cpu(fe->i_clusters) == 0) {
-		/* trunc to zero is a special case. */
-		el->l_tree_depth = 0;
-		fe->i_last_eb_blk = 0;
-	} else if (last_eb)
-		fe->i_last_eb_blk = last_eb->h_blkno;
-
-	status = ocfs2_journal_dirty(handle, fe_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
-
-	if (last_eb) {
-		/* If there will be a new last extent block, then by
-		 * definition, there cannot be any leaves to the right of
-		 * him. */
-		last_eb->h_next_leaf_blk = 0;
-		status = ocfs2_journal_dirty(handle, last_eb_bh);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
-	}
-
-	if (delete_blk) {
-		if (rec_flags & OCFS2_EXT_REFCOUNTED)
-			status = ocfs2_decrease_refcount(inode, handle,
-					ocfs2_blocks_to_clusters(osb->sb,
-								 delete_blk),
-					clusters_to_del, meta_ac,
-					&tc->tc_dealloc, 1);
-		else
-			status = ocfs2_truncate_log_append(osb, handle,
-							   delete_blk,
-							   clusters_to_del);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
-	}
-	status = 0;
-bail:
-	brelse(last_eb_bh);
-	mlog_exit(status);
-	return status;
-}
-
 static int ocfs2_zero_func(handle_t *handle, struct buffer_head *bh)
 {
 	set_buffer_uptodate(bh);
@@ -7300,14 +6882,15 @@ int ocfs2_convert_inline_data_to_extents(struct inode *inode,
 		unsigned int page_end;
 		u64 phys;
 
-		if (vfs_dq_alloc_space_nodirty(inode,
-				       ocfs2_clusters_to_bytes(osb->sb, 1))) {
-			ret = -EDQUOT;
+		ret = dquot_alloc_space_nodirty(inode,
+				       ocfs2_clusters_to_bytes(osb->sb, 1));
+		if (ret)
 			goto out_commit;
-		}
 		did_quota = 1;
 
-		ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off,
+		data_ac->ac_resv = &OCFS2_I(inode)->ip_la_data_resv;
+
+		ret = ocfs2_claim_clusters(handle, data_ac, 1, &bit_off,
 					   &num);
 		if (ret) {
 			mlog_errno(ret);
@@ -7380,7 +6963,7 @@ int ocfs2_convert_inline_data_to_extents(struct inode *inode,
 
 out_commit:
 	if (ret < 0 && did_quota)
-		vfs_dq_free_space_nodirty(inode,
+		dquot_free_space_nodirty(inode,
 					  ocfs2_clusters_to_bytes(osb->sb, 1));
 
 	ocfs2_commit_trans(osb, handle);
@@ -7406,26 +6989,29 @@ out:
  */
 int ocfs2_commit_truncate(struct ocfs2_super *osb,
 			  struct inode *inode,
-			  struct buffer_head *fe_bh,
-			  struct ocfs2_truncate_context *tc)
+			  struct buffer_head *di_bh)
 {
-	int status, i, credits, tl_sem = 0;
-	u32 clusters_to_del, new_highest_cpos, range;
+	int status = 0, i, flags = 0;
+	u32 new_highest_cpos, range, trunc_cpos, trunc_len, phys_cpos, coff;
 	u64 blkno = 0;
 	struct ocfs2_extent_list *el;
-	handle_t *handle = NULL;
-	struct inode *tl_inode = osb->osb_tl_inode;
+	struct ocfs2_extent_rec *rec;
 	struct ocfs2_path *path = NULL;
-	struct ocfs2_dinode *di = (struct ocfs2_dinode *)fe_bh->b_data;
-	struct ocfs2_alloc_context *meta_ac = NULL;
-	struct ocfs2_refcount_tree *ref_tree = NULL;
+	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
+	struct ocfs2_extent_list *root_el = &(di->id2.i_list);
+	u64 refcount_loc = le64_to_cpu(di->i_refcount_loc);
+	struct ocfs2_extent_tree et;
+	struct ocfs2_cached_dealloc_ctxt dealloc;
 
 	mlog_entry_void();
 
+	ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), di_bh);
+	ocfs2_init_dealloc_ctxt(&dealloc);
+
 	new_highest_cpos = ocfs2_clusters_for_bytes(osb->sb,
 						     i_size_read(inode));
 
-	path = ocfs2_new_path(fe_bh, &di->id2.i_list,
+	path = ocfs2_new_path(di_bh, &di->id2.i_list,
 			      ocfs2_journal_access_di);
 	if (!path) {
 		status = -ENOMEM;
@@ -7444,8 +7030,6 @@ start:
 		goto bail;
 	}
 
-	credits = 0;
-
 	/*
 	 * Truncate always works against the rightmost tree branch.
 	 */
@@ -7480,101 +7064,62 @@ start:
 	}
 
 	i = le16_to_cpu(el->l_next_free_rec) - 1;
-	range = le32_to_cpu(el->l_recs[i].e_cpos) +
-		ocfs2_rec_clusters(el, &el->l_recs[i]);
-	if (i == 0 && ocfs2_is_empty_extent(&el->l_recs[i])) {
-		clusters_to_del = 0;
-	} else if (le32_to_cpu(el->l_recs[i].e_cpos) >= new_highest_cpos) {
-		clusters_to_del = ocfs2_rec_clusters(el, &el->l_recs[i]);
-		blkno = le64_to_cpu(el->l_recs[i].e_blkno);
+	rec = &el->l_recs[i];
+	flags = rec->e_flags;
+	range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec);
+
+	if (i == 0 && ocfs2_is_empty_extent(rec)) {
+		/*
+		 * Lower levels depend on this never happening, but it's best
+		 * to check it up here before changing the tree.
+		*/
+		if (root_el->l_tree_depth && rec->e_int_clusters == 0) {
+			ocfs2_error(inode->i_sb, "Inode %lu has an empty "
+				    "extent record, depth %u\n", inode->i_ino,
+				    le16_to_cpu(root_el->l_tree_depth));
+			status = -EROFS;
+			goto bail;
+		}
+		trunc_cpos = le32_to_cpu(rec->e_cpos);
+		trunc_len = 0;
+		blkno = 0;
+	} else if (le32_to_cpu(rec->e_cpos) >= new_highest_cpos) {
+		/*
+		 * Truncate entire record.
+		 */
+		trunc_cpos = le32_to_cpu(rec->e_cpos);
+		trunc_len = ocfs2_rec_clusters(el, rec);
+		blkno = le64_to_cpu(rec->e_blkno);
 	} else if (range > new_highest_cpos) {
-		clusters_to_del = (ocfs2_rec_clusters(el, &el->l_recs[i]) +
-				   le32_to_cpu(el->l_recs[i].e_cpos)) -
-				  new_highest_cpos;
-		blkno = le64_to_cpu(el->l_recs[i].e_blkno) +
-			ocfs2_clusters_to_blocks(inode->i_sb,
-				ocfs2_rec_clusters(el, &el->l_recs[i]) -
-				clusters_to_del);
+		/*
+		 * Partial truncate. it also should be
+		 * the last truncate we're doing.
+		 */
+		trunc_cpos = new_highest_cpos;
+		trunc_len = range - new_highest_cpos;
+		coff = new_highest_cpos - le32_to_cpu(rec->e_cpos);
+		blkno = le64_to_cpu(rec->e_blkno) +
+				ocfs2_clusters_to_blocks(inode->i_sb, coff);
 	} else {
+		/*
+		 * Truncate completed, leave happily.
+		 */
 		status = 0;
 		goto bail;
 	}
 
-	mlog(0, "clusters_to_del = %u in this pass, tail blk=%llu\n",
-	     clusters_to_del, (unsigned long long)path_leaf_bh(path)->b_blocknr);
-
-	if (el->l_recs[i].e_flags & OCFS2_EXT_REFCOUNTED && clusters_to_del) {
-		BUG_ON(!(OCFS2_I(inode)->ip_dyn_features &
-			 OCFS2_HAS_REFCOUNT_FL));
-
-		status = ocfs2_lock_refcount_tree(osb,
-						le64_to_cpu(di->i_refcount_loc),
-						1, &ref_tree, NULL);
-		if (status) {
-			mlog_errno(status);
-			goto bail;
-		}
-
-		status = ocfs2_prepare_refcount_change_for_del(inode, fe_bh,
-							       blkno,
-							       clusters_to_del,
-							       &credits,
-							       &meta_ac);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
-	}
-
-	mutex_lock(&tl_inode->i_mutex);
-	tl_sem = 1;
-	/* ocfs2_truncate_log_needs_flush guarantees us at least one
-	 * record is free for use. If there isn't any, we flush to get
-	 * an empty truncate log.  */
-	if (ocfs2_truncate_log_needs_flush(osb)) {
-		status = __ocfs2_flush_truncate_log(osb);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
-	}
-
-	credits += ocfs2_calc_tree_trunc_credits(osb->sb, clusters_to_del,
-						(struct ocfs2_dinode *)fe_bh->b_data,
-						el);
-	handle = ocfs2_start_trans(osb, credits);
-	if (IS_ERR(handle)) {
-		status = PTR_ERR(handle);
-		handle = NULL;
-		mlog_errno(status);
-		goto bail;
-	}
+	phys_cpos = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
 
-	status = ocfs2_do_truncate(osb, clusters_to_del, inode, fe_bh, handle,
-				   tc, path, meta_ac);
+	status = ocfs2_remove_btree_range(inode, &et, trunc_cpos,
+					  phys_cpos, trunc_len, flags, &dealloc,
+					  refcount_loc);
 	if (status < 0) {
 		mlog_errno(status);
 		goto bail;
 	}
 
-	mutex_unlock(&tl_inode->i_mutex);
-	tl_sem = 0;
-
-	ocfs2_commit_trans(osb, handle);
-	handle = NULL;
-
 	ocfs2_reinit_path(path, 1);
 
-	if (meta_ac) {
-		ocfs2_free_alloc_context(meta_ac);
-		meta_ac = NULL;
-	}
-
-	if (ref_tree) {
-		ocfs2_unlock_refcount_tree(osb, ref_tree, 1);
-		ref_tree = NULL;
-	}
-
 	/*
 	 * The check above will catch the case where we've truncated
 	 * away all allocation.
@@ -7585,25 +7130,10 @@ bail:
 
 	ocfs2_schedule_truncate_log_flush(osb, 1);
 
-	if (tl_sem)
-		mutex_unlock(&tl_inode->i_mutex);
-
-	if (handle)
-		ocfs2_commit_trans(osb, handle);
-
-	if (meta_ac)
-		ocfs2_free_alloc_context(meta_ac);
-
-	if (ref_tree)
-		ocfs2_unlock_refcount_tree(osb, ref_tree, 1);
-
-	ocfs2_run_deallocs(osb, &tc->tc_dealloc);
+	ocfs2_run_deallocs(osb, &dealloc);
 
 	ocfs2_free_path(path);
 
-	/* This will drop the ext_alloc cluster lock for us */
-	ocfs2_free_truncate_context(tc);
-
 	mlog_exit(status);
 	return status;
 }
diff --git a/fs/ocfs2/alloc.h b/fs/ocfs2/alloc.h
index 1db4359ccb90..55762b554b99 100644
--- a/fs/ocfs2/alloc.h
+++ b/fs/ocfs2/alloc.h
@@ -140,8 +140,9 @@ int ocfs2_remove_extent(handle_t *handle, struct ocfs2_extent_tree *et,
 			struct ocfs2_cached_dealloc_ctxt *dealloc);
 int ocfs2_remove_btree_range(struct inode *inode,
 			     struct ocfs2_extent_tree *et,
-			     u32 cpos, u32 phys_cpos, u32 len,
-			     struct ocfs2_cached_dealloc_ctxt *dealloc);
+			     u32 cpos, u32 phys_cpos, u32 len, int flags,
+			     struct ocfs2_cached_dealloc_ctxt *dealloc,
+			     u64 refcount_loc);
 
 int ocfs2_num_free_extents(struct ocfs2_super *osb,
 			   struct ocfs2_extent_tree *et);
@@ -209,7 +210,7 @@ static inline void ocfs2_init_dealloc_ctxt(struct ocfs2_cached_dealloc_ctxt *c)
 int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
 				u64 blkno, unsigned int bit);
 int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
-			      int type, int slot, u64 blkno,
+			      int type, int slot, u64 suballoc, u64 blkno,
 			      unsigned int bit);
 static inline int ocfs2_dealloc_has_cluster(struct ocfs2_cached_dealloc_ctxt *c)
 {
@@ -233,8 +234,7 @@ int ocfs2_prepare_truncate(struct ocfs2_super *osb,
 			   struct ocfs2_truncate_context **tc);
 int ocfs2_commit_truncate(struct ocfs2_super *osb,
 			  struct inode *inode,
-			  struct buffer_head *fe_bh,
-			  struct ocfs2_truncate_context *tc);
+			  struct buffer_head *di_bh);
 int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
 			  unsigned int start, unsigned int end, int trunc);
 
@@ -319,6 +319,8 @@ int ocfs2_journal_access_path(struct ocfs2_caching_info *ci,
 			      struct ocfs2_path *path);
 int ocfs2_find_cpos_for_right_leaf(struct super_block *sb,
 				   struct ocfs2_path *path, u32 *cpos);
+int ocfs2_find_cpos_for_left_leaf(struct super_block *sb,
+				  struct ocfs2_path *path, u32 *cpos);
 int ocfs2_find_subtree_root(struct ocfs2_extent_tree *et,
 			    struct ocfs2_path *left,
 			    struct ocfs2_path *right);
diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c
index 3dae4a13f6e4..356e976772bf 100644
--- a/fs/ocfs2/aops.c
+++ b/fs/ocfs2/aops.c
@@ -196,15 +196,14 @@ int ocfs2_get_block(struct inode *inode, sector_t iblock,
 			dump_stack();
 			goto bail;
 		}
-
-		past_eof = ocfs2_blocks_for_bytes(inode->i_sb, i_size_read(inode));
-		mlog(0, "Inode %lu, past_eof = %llu\n", inode->i_ino,
-		     (unsigned long long)past_eof);
-
-		if (create && (iblock >= past_eof))
-			set_buffer_new(bh_result);
 	}
 
+	past_eof = ocfs2_blocks_for_bytes(inode->i_sb, i_size_read(inode));
+	mlog(0, "Inode %lu, past_eof = %llu\n", inode->i_ino,
+	     (unsigned long long)past_eof);
+	if (create && (iblock >= past_eof))
+		set_buffer_new(bh_result);
+
 bail:
 	if (err < 0)
 		err = -EIO;
@@ -459,36 +458,6 @@ int walk_page_buffers(	handle_t *handle,
 	return ret;
 }
 
-handle_t *ocfs2_start_walk_page_trans(struct inode *inode,
-							 struct page *page,
-							 unsigned from,
-							 unsigned to)
-{
-	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
-	handle_t *handle;
-	int ret = 0;
-
-	handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
-	if (IS_ERR(handle)) {
-		ret = -ENOMEM;
-		mlog_errno(ret);
-		goto out;
-	}
-
-	if (ocfs2_should_order_data(inode)) {
-		ret = ocfs2_jbd2_file_inode(handle, inode);
-		if (ret < 0)
-			mlog_errno(ret);
-	}
-out:
-	if (ret) {
-		if (!IS_ERR(handle))
-			ocfs2_commit_trans(osb, handle);
-		handle = ERR_PTR(ret);
-	}
-	return handle;
-}
-
 static sector_t ocfs2_bmap(struct address_space *mapping, sector_t block)
 {
 	sector_t status;
@@ -577,8 +546,9 @@ static int ocfs2_direct_IO_get_blocks(struct inode *inode, sector_t iblock,
 		goto bail;
 	}
 
-	/* We should already CoW the refcounted extent. */
-	BUG_ON(ext_flags & OCFS2_EXT_REFCOUNTED);
+	/* We should already CoW the refcounted extent in case of create. */
+	BUG_ON(create && (ext_flags & OCFS2_EXT_REFCOUNTED));
+
 	/*
 	 * get_more_blocks() expects us to describe a hole by clearing
 	 * the mapped bit on bh_result().
@@ -599,7 +569,7 @@ bail:
 	return ret;
 }
 
-/* 
+/*
  * ocfs2_dio_end_io is called by the dio core when a dio is finished.  We're
  * particularly interested in the aio/dio case.  Like the core uses
  * i_alloc_sem, we use the rw_lock DLM lock to protect io on one node from
@@ -670,7 +640,7 @@ static ssize_t ocfs2_direct_IO(int rw,
 
 	ret = blockdev_direct_IO_no_locking(rw, iocb, inode,
 					    inode->i_sb->s_bdev, iov, offset,
-					    nr_segs, 
+					    nr_segs,
 					    ocfs2_direct_IO_get_blocks,
 					    ocfs2_dio_end_io);
 
@@ -1130,23 +1100,37 @@ out:
  */
 static int ocfs2_grab_pages_for_write(struct address_space *mapping,
 				      struct ocfs2_write_ctxt *wc,
-				      u32 cpos, loff_t user_pos, int new,
+				      u32 cpos, loff_t user_pos,
+				      unsigned user_len, int new,
 				      struct page *mmap_page)
 {
 	int ret = 0, i;
-	unsigned long start, target_index, index;
+	unsigned long start, target_index, end_index, index;
 	struct inode *inode = mapping->host;
+	loff_t last_byte;
 
 	target_index = user_pos >> PAGE_CACHE_SHIFT;
 
 	/*
 	 * Figure out how many pages we'll be manipulating here. For
 	 * non allocating write, we just change the one
-	 * page. Otherwise, we'll need a whole clusters worth.
+	 * page. Otherwise, we'll need a whole clusters worth.  If we're
+	 * writing past i_size, we only need enough pages to cover the
+	 * last page of the write.
 	 */
 	if (new) {
 		wc->w_num_pages = ocfs2_pages_per_cluster(inode->i_sb);
 		start = ocfs2_align_clusters_to_page_index(inode->i_sb, cpos);
+		/*
+		 * We need the index *past* the last page we could possibly
+		 * touch.  This is the page past the end of the write or
+		 * i_size, whichever is greater.
+		 */
+		last_byte = max(user_pos + user_len, i_size_read(inode));
+		BUG_ON(last_byte < 1);
+		end_index = ((last_byte - 1) >> PAGE_CACHE_SHIFT) + 1;
+		if ((start + wc->w_num_pages) > end_index)
+			wc->w_num_pages = end_index - start;
 	} else {
 		wc->w_num_pages = 1;
 		start = target_index;
@@ -1619,21 +1603,20 @@ out:
  * write path can treat it as an non-allocating write, which has no
  * special case code for sparse/nonsparse files.
  */
-static int ocfs2_expand_nonsparse_inode(struct inode *inode, loff_t pos,
-					unsigned len,
+static int ocfs2_expand_nonsparse_inode(struct inode *inode,
+					struct buffer_head *di_bh,
+					loff_t pos, unsigned len,
 					struct ocfs2_write_ctxt *wc)
 {
 	int ret;
-	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 	loff_t newsize = pos + len;
 
-	if (ocfs2_sparse_alloc(osb))
-		return 0;
+	BUG_ON(ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb)));
 
 	if (newsize <= i_size_read(inode))
 		return 0;
 
-	ret = ocfs2_extend_no_holes(inode, newsize, pos);
+	ret = ocfs2_extend_no_holes(inode, di_bh, newsize, pos);
 	if (ret)
 		mlog_errno(ret);
 
@@ -1643,6 +1626,18 @@ static int ocfs2_expand_nonsparse_inode(struct inode *inode, loff_t pos,
 	return ret;
 }
 
+static int ocfs2_zero_tail(struct inode *inode, struct buffer_head *di_bh,
+			   loff_t pos)
+{
+	int ret = 0;
+
+	BUG_ON(!ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb)));
+	if (pos > i_size_read(inode))
+		ret = ocfs2_zero_extend(inode, di_bh, pos);
+
+	return ret;
+}
+
 int ocfs2_write_begin_nolock(struct address_space *mapping,
 			     loff_t pos, unsigned len, unsigned flags,
 			     struct page **pagep, void **fsdata,
@@ -1678,7 +1673,11 @@ int ocfs2_write_begin_nolock(struct address_space *mapping,
 		}
 	}
 
-	ret = ocfs2_expand_nonsparse_inode(inode, pos, len, wc);
+	if (ocfs2_sparse_alloc(osb))
+		ret = ocfs2_zero_tail(inode, di_bh, pos);
+	else
+		ret = ocfs2_expand_nonsparse_inode(inode, di_bh, pos, len,
+						   wc);
 	if (ret) {
 		mlog_errno(ret);
 		goto out;
@@ -1734,6 +1733,9 @@ int ocfs2_write_begin_nolock(struct address_space *mapping,
 			goto out;
 		}
 
+		if (data_ac)
+			data_ac->ac_resv = &OCFS2_I(inode)->ip_la_data_resv;
+
 		credits = ocfs2_calc_extend_credits(inode->i_sb,
 						    &di->id2.i_list,
 						    clusters_to_alloc);
@@ -1763,10 +1765,11 @@ int ocfs2_write_begin_nolock(struct address_space *mapping,
 
 	wc->w_handle = handle;
 
-	if (clusters_to_alloc && vfs_dq_alloc_space_nodirty(inode,
-			ocfs2_clusters_to_bytes(osb->sb, clusters_to_alloc))) {
-		ret = -EDQUOT;
-		goto out_commit;
+	if (clusters_to_alloc) {
+		ret = dquot_alloc_space_nodirty(inode,
+			ocfs2_clusters_to_bytes(osb->sb, clusters_to_alloc));
+		if (ret)
+			goto out_commit;
 	}
 	/*
 	 * We don't want this to fail in ocfs2_write_end(), so do it
@@ -1784,7 +1787,7 @@ int ocfs2_write_begin_nolock(struct address_space *mapping,
 	 * that we can zero and flush if we error after adding the
 	 * extent.
 	 */
-	ret = ocfs2_grab_pages_for_write(mapping, wc, wc->w_cpos, pos,
+	ret = ocfs2_grab_pages_for_write(mapping, wc, wc->w_cpos, pos, len,
 					 cluster_of_pages, mmap_page);
 	if (ret) {
 		mlog_errno(ret);
@@ -1809,7 +1812,7 @@ success:
 	return 0;
 out_quota:
 	if (clusters_to_alloc)
-		vfs_dq_free_space(inode,
+		dquot_free_space(inode,
 			  ocfs2_clusters_to_bytes(osb->sb, clusters_to_alloc));
 out_commit:
 	ocfs2_commit_trans(osb, handle);
diff --git a/fs/ocfs2/blockcheck.c b/fs/ocfs2/blockcheck.c
index b7428c5d0d3b..ec6d12339593 100644
--- a/fs/ocfs2/blockcheck.c
+++ b/fs/ocfs2/blockcheck.c
@@ -403,7 +403,7 @@ void ocfs2_block_check_compute(void *data, size_t blocksize,
 	 * No ecc'd ocfs2 structure is larger than 4K, so ecc will be no
 	 * larger than 16 bits.
 	 */
-	BUG_ON(ecc > USHORT_MAX);
+	BUG_ON(ecc > USHRT_MAX);
 
 	bc->bc_crc32e = cpu_to_le32(crc);
 	bc->bc_ecc = cpu_to_le16((u16)ecc);
@@ -508,7 +508,7 @@ void ocfs2_block_check_compute_bhs(struct buffer_head **bhs, int nr,
 	 * No ecc'd ocfs2 structure is larger than 4K, so ecc will be no
 	 * larger than 16 bits.
 	 */
-	BUG_ON(ecc > USHORT_MAX);
+	BUG_ON(ecc > USHRT_MAX);
 
 	bc->bc_crc32e = cpu_to_le32(crc);
 	bc->bc_ecc = cpu_to_le16((u16)ecc);
diff --git a/fs/ocfs2/buffer_head_io.c b/fs/ocfs2/buffer_head_io.c
index d43d34a1dd31..f9d5d3ffc75a 100644
--- a/fs/ocfs2/buffer_head_io.c
+++ b/fs/ocfs2/buffer_head_io.c
@@ -25,7 +25,6 @@
 
 #include <linux/fs.h>
 #include <linux/types.h>
-#include <linux/slab.h>
 #include <linux/highmem.h>
 
 #include <cluster/masklog.h>
@@ -368,7 +367,7 @@ int ocfs2_read_blocks(struct ocfs2_caching_info *ci, u64 block, int nr,
 	}
 	ocfs2_metadata_cache_io_unlock(ci);
 
-	mlog(ML_BH_IO, "block=(%llu), nr=(%d), cached=%s, flags=0x%x\n", 
+	mlog(ML_BH_IO, "block=(%llu), nr=(%d), cached=%s, flags=0x%x\n",
 	     (unsigned long long)block, nr,
 	     ((flags & OCFS2_BH_IGNORE_CACHE) || ignore_cache) ? "no" : "yes",
 	     flags);
@@ -407,6 +406,7 @@ int ocfs2_write_super_or_backup(struct ocfs2_super *osb,
 				struct buffer_head *bh)
 {
 	int ret = 0;
+	struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
 
 	mlog_entry_void();
 
@@ -426,6 +426,7 @@ int ocfs2_write_super_or_backup(struct ocfs2_super *osb,
 
 	get_bh(bh); /* for end_buffer_write_sync() */
 	bh->b_end_io = end_buffer_write_sync;
+	ocfs2_compute_meta_ecc(osb->sb, bh->b_data, &di->i_check);
 	submit_bh(WRITE, bh);
 
 	wait_on_buffer(bh);
diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c
index eda5b8bcddd5..41d5f1f92d56 100644
--- a/fs/ocfs2/cluster/heartbeat.c
+++ b/fs/ocfs2/cluster/heartbeat.c
@@ -34,6 +34,7 @@
 #include <linux/crc32.h>
 #include <linux/time.h>
 #include <linux/debugfs.h>
+#include <linux/slab.h>
 
 #include "heartbeat.h"
 #include "tcp.h"
@@ -78,7 +79,7 @@ static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type);
 
 unsigned int o2hb_dead_threshold = O2HB_DEFAULT_DEAD_THRESHOLD;
 
-/* Only sets a new threshold if there are no active regions. 
+/* Only sets a new threshold if there are no active regions.
  *
  * No locking or otherwise interesting code is required for reading
  * o2hb_dead_threshold as it can't change once regions are active and
@@ -170,7 +171,7 @@ static void o2hb_write_timeout(struct work_struct *work)
 
 	mlog(ML_ERROR, "Heartbeat write timeout to device %s after %u "
 	     "milliseconds\n", reg->hr_dev_name,
-	     jiffies_to_msecs(jiffies - reg->hr_last_timeout_start)); 
+	     jiffies_to_msecs(jiffies - reg->hr_last_timeout_start));
 	o2quo_disk_timeout();
 }
 
@@ -624,7 +625,7 @@ static int o2hb_check_slot(struct o2hb_region *reg,
 	     "seq %llu last %llu changed %u equal %u\n",
 	     slot->ds_node_num, (long long)slot->ds_last_generation,
 	     le32_to_cpu(hb_block->hb_cksum),
-	     (unsigned long long)le64_to_cpu(hb_block->hb_seq), 
+	     (unsigned long long)le64_to_cpu(hb_block->hb_seq),
 	     (unsigned long long)slot->ds_last_time, slot->ds_changed_samples,
 	     slot->ds_equal_samples);
 
diff --git a/fs/ocfs2/cluster/masklog.c b/fs/ocfs2/cluster/masklog.c
index 1cd2934de615..c7fba396392d 100644
--- a/fs/ocfs2/cluster/masklog.c
+++ b/fs/ocfs2/cluster/masklog.c
@@ -112,9 +112,11 @@ static struct mlog_attribute mlog_attrs[MLOG_MAX_BITS] = {
 	define_mask(XATTR),
 	define_mask(QUOTA),
 	define_mask(REFCOUNT),
+	define_mask(BASTS),
 	define_mask(ERROR),
 	define_mask(NOTICE),
 	define_mask(KTHREAD),
+	define_mask(RESERVATIONS),
 };
 
 static struct attribute *mlog_attr_ptrs[MLOG_MAX_BITS] = {NULL, };
@@ -135,7 +137,7 @@ static ssize_t mlog_store(struct kobject *obj, struct attribute *attr,
 	return mlog_mask_store(mlog_attr->mask, buf, count);
 }
 
-static struct sysfs_ops mlog_attr_ops = {
+static const struct sysfs_ops mlog_attr_ops = {
 	.show  = mlog_show,
 	.store = mlog_store,
 };
diff --git a/fs/ocfs2/cluster/masklog.h b/fs/ocfs2/cluster/masklog.h
index 9b4d11726cf2..fd96e2a2fa56 100644
--- a/fs/ocfs2/cluster/masklog.h
+++ b/fs/ocfs2/cluster/masklog.h
@@ -114,10 +114,12 @@
 #define ML_XATTR	0x0000000020000000ULL /* ocfs2 extended attributes */
 #define ML_QUOTA	0x0000000040000000ULL /* ocfs2 quota operations */
 #define ML_REFCOUNT	0x0000000080000000ULL /* refcount tree operations */
+#define ML_BASTS	0x0000001000000000ULL /* dlmglue asts and basts */
 /* bits that are infrequently given and frequently matched in the high word */
 #define ML_ERROR	0x0000000100000000ULL /* sent to KERN_ERR */
 #define ML_NOTICE	0x0000000200000000ULL /* setn to KERN_NOTICE */
 #define ML_KTHREAD	0x0000000400000000ULL /* kernel thread activity */
+#define	ML_RESERVATIONS	0x0000000800000000ULL /* ocfs2 alloc reservations */
 
 #define MLOG_INITIAL_AND_MASK (ML_ERROR|ML_NOTICE)
 #define MLOG_INITIAL_NOT_MASK (ML_ENTRY|ML_EXIT)
@@ -194,9 +196,9 @@ extern struct mlog_bits mlog_and_bits, mlog_not_bits;
  * previous token if args expands to nothing.
  */
 #define __mlog_printk(level, fmt, args...)				\
-	printk(level "(%u,%lu):%s:%d " fmt, task_pid_nr(current),	\
-	       __mlog_cpu_guess, __PRETTY_FUNCTION__, __LINE__ ,	\
-	       ##args)
+	printk(level "(%s,%u,%lu):%s:%d " fmt, current->comm,		\
+	       task_pid_nr(current), __mlog_cpu_guess,			\
+	       __PRETTY_FUNCTION__, __LINE__ , ##args)
 
 #define mlog(mask, fmt, args...) do {					\
 	u64 __m = MLOG_MASK_PREFIX | (mask);				\
diff --git a/fs/ocfs2/cluster/nodemanager.c b/fs/ocfs2/cluster/nodemanager.c
index c81142e3ef84..ed0c9f367fed 100644
--- a/fs/ocfs2/cluster/nodemanager.c
+++ b/fs/ocfs2/cluster/nodemanager.c
@@ -19,6 +19,7 @@
  * Boston, MA 021110-1307, USA.
  */
 
+#include <linux/slab.h>
 #include <linux/kernel.h>
 #include <linux/module.h>
 #include <linux/configfs.h>
diff --git a/fs/ocfs2/cluster/quorum.c b/fs/ocfs2/cluster/quorum.c
index 639024033fce..cf3e16696216 100644
--- a/fs/ocfs2/cluster/quorum.c
+++ b/fs/ocfs2/cluster/quorum.c
@@ -44,7 +44,6 @@
  * and if they're the last, they fire off the decision.
  */
 #include <linux/kernel.h>
-#include <linux/slab.h>
 #include <linux/workqueue.h>
 #include <linux/reboot.h>
 
diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index 334f231a422c..aa75ca3f78da 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -72,9 +72,9 @@
 
 #include "tcp_internal.h"
 
-#define SC_NODEF_FMT "node %s (num %u) at %u.%u.%u.%u:%u"
+#define SC_NODEF_FMT "node %s (num %u) at %pI4:%u"
 #define SC_NODEF_ARGS(sc) sc->sc_node->nd_name, sc->sc_node->nd_num,	\
-			  NIPQUAD(sc->sc_node->nd_ipv4_address),	\
+			  &sc->sc_node->nd_ipv4_address,		\
 			  ntohs(sc->sc_node->nd_ipv4_port)
 
 /*
@@ -485,7 +485,7 @@ static void o2net_set_nn_state(struct o2net_node *nn,
 	}
 
 	if (was_valid && !valid) {
-		printk(KERN_INFO "o2net: no longer connected to "
+		printk(KERN_NOTICE "o2net: no longer connected to "
 		       SC_NODEF_FMT "\n", SC_NODEF_ARGS(old_sc));
 		o2net_complete_nodes_nsw(nn);
 	}
@@ -493,7 +493,7 @@ static void o2net_set_nn_state(struct o2net_node *nn,
 	if (!was_valid && valid) {
 		o2quo_conn_up(o2net_num_from_nn(nn));
 		cancel_delayed_work(&nn->nn_connect_expired);
-		printk(KERN_INFO "o2net: %s " SC_NODEF_FMT "\n",
+		printk(KERN_NOTICE "o2net: %s " SC_NODEF_FMT "\n",
 		       o2nm_this_node() > sc->sc_node->nd_num ?
 		       		"connected to" : "accepted connection from",
 		       SC_NODEF_ARGS(sc));
@@ -583,6 +583,9 @@ static void o2net_state_change(struct sock *sk)
 			o2net_sc_queue_work(sc, &sc->sc_connect_work);
 			break;
 		default:
+			printk(KERN_INFO "o2net: connection to " SC_NODEF_FMT
+			      " shutdown, state %d\n",
+			      SC_NODEF_ARGS(sc), sk->sk_state);
 			o2net_sc_queue_work(sc, &sc->sc_shutdown_work);
 			break;
 	}
@@ -930,7 +933,7 @@ static void o2net_sendpage(struct o2net_sock_container *sc,
 			cond_resched();
 			continue;
 		}
-		mlog(ML_ERROR, "sendpage of size %zu to " SC_NODEF_FMT 
+		mlog(ML_ERROR, "sendpage of size %zu to " SC_NODEF_FMT
 		     " failed with %zd\n", size, SC_NODEF_ARGS(sc), ret);
 		o2net_ensure_shutdown(nn, sc, 0);
 		break;
@@ -1476,14 +1479,14 @@ static void o2net_idle_timer(unsigned long data)
 
 	do_gettimeofday(&now);
 
-	printk(KERN_INFO "o2net: connection to " SC_NODEF_FMT " has been idle for %u.%u "
+	printk(KERN_NOTICE "o2net: connection to " SC_NODEF_FMT " has been idle for %u.%u "
 	     "seconds, shutting it down.\n", SC_NODEF_ARGS(sc),
 		     o2net_idle_timeout() / 1000,
 		     o2net_idle_timeout() % 1000);
 	mlog(ML_NOTICE, "here are some times that might help debug the "
 	     "situation: (tmr %ld.%ld now %ld.%ld dr %ld.%ld adv "
 	     "%ld.%ld:%ld.%ld func (%08x:%u) %ld.%ld:%ld.%ld)\n",
-	     sc->sc_tv_timer.tv_sec, (long) sc->sc_tv_timer.tv_usec, 
+	     sc->sc_tv_timer.tv_sec, (long) sc->sc_tv_timer.tv_usec,
 	     now.tv_sec, (long) now.tv_usec,
 	     sc->sc_tv_data_ready.tv_sec, (long) sc->sc_tv_data_ready.tv_usec,
 	     sc->sc_tv_advance_start.tv_sec,
diff --git a/fs/ocfs2/cluster/tcp_internal.h b/fs/ocfs2/cluster/tcp_internal.h
index 8d58cfe410b1..96fa7ebc530c 100644
--- a/fs/ocfs2/cluster/tcp_internal.h
+++ b/fs/ocfs2/cluster/tcp_internal.h
@@ -32,10 +32,10 @@
  * on their number */
 #define O2NET_QUORUM_DELAY_MS	((o2hb_dead_threshold + 2) * O2HB_REGION_TIMEOUT_MS)
 
-/* 
+/*
  * This version number represents quite a lot, unfortunately.  It not
  * only represents the raw network message protocol on the wire but also
- * locking semantics of the file system using the protocol.  It should 
+ * locking semantics of the file system using the protocol.  It should
  * be somewhere else, I'm sure, but right now it isn't.
  *
  * With version 11, we separate out the filesystem locking portion.  The
diff --git a/fs/ocfs2/dir.c b/fs/ocfs2/dir.c
index 28c3ec238796..f04ebcfffc4a 100644
--- a/fs/ocfs2/dir.c
+++ b/fs/ocfs2/dir.c
@@ -1194,7 +1194,7 @@ static int __ocfs2_delete_entry(handle_t *handle, struct inode *dir,
 			else
 				de->inode = 0;
 			dir->i_version++;
-			status = ocfs2_journal_dirty(handle, bh);
+			ocfs2_journal_dirty(handle, bh);
 			goto bail;
 		}
 		i += le16_to_cpu(de->rec_len);
@@ -1752,7 +1752,7 @@ int __ocfs2_add_entry(handle_t *handle,
 				ocfs2_recalc_free_list(dir, handle, lookup);
 
 			dir->i_version++;
-			status = ocfs2_journal_dirty(handle, insert_bh);
+			ocfs2_journal_dirty(handle, insert_bh);
 			retval = 0;
 			goto bail;
 		}
@@ -2297,12 +2297,7 @@ static int ocfs2_fill_new_dir_id(struct ocfs2_super *osb,
 	}
 
 	ocfs2_fill_initial_dirents(inode, parent, data->id_data, size);
-
 	ocfs2_journal_dirty(handle, di_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out;
-	}
 
 	i_size_write(inode, size);
 	inode->i_nlink = 2;
@@ -2366,11 +2361,7 @@ static int ocfs2_fill_new_dir_el(struct ocfs2_super *osb,
 		ocfs2_init_dir_trailer(inode, new_bh, size);
 	}
 
-	status = ocfs2_journal_dirty(handle, new_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	ocfs2_journal_dirty(handle, new_bh);
 
 	i_size_write(inode, inode->i_sb->s_blocksize);
 	inode->i_nlink = 2;
@@ -2404,15 +2395,15 @@ static int ocfs2_dx_dir_attach_index(struct ocfs2_super *osb,
 	int ret;
 	struct ocfs2_dinode *di = (struct ocfs2_dinode *) di_bh->b_data;
 	u16 dr_suballoc_bit;
-	u64 dr_blkno;
+	u64 suballoc_loc, dr_blkno;
 	unsigned int num_bits;
 	struct buffer_head *dx_root_bh = NULL;
 	struct ocfs2_dx_root_block *dx_root;
 	struct ocfs2_dir_block_trailer *trailer =
 		ocfs2_trailer_from_bh(dirdata_bh, dir->i_sb);
 
-	ret = ocfs2_claim_metadata(osb, handle, meta_ac, 1, &dr_suballoc_bit,
-				   &num_bits, &dr_blkno);
+	ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc,
+				   &dr_suballoc_bit, &num_bits, &dr_blkno);
 	if (ret) {
 		mlog_errno(ret);
 		goto out;
@@ -2439,7 +2430,8 @@ static int ocfs2_dx_dir_attach_index(struct ocfs2_super *osb,
 	dx_root = (struct ocfs2_dx_root_block *)dx_root_bh->b_data;
 	memset(dx_root, 0, osb->sb->s_blocksize);
 	strcpy(dx_root->dr_signature, OCFS2_DX_ROOT_SIGNATURE);
-	dx_root->dr_suballoc_slot = cpu_to_le16(osb->slot_num);
+	dx_root->dr_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot);
+	dx_root->dr_suballoc_loc = cpu_to_le64(suballoc_loc);
 	dx_root->dr_suballoc_bit = cpu_to_le16(dr_suballoc_bit);
 	dx_root->dr_fs_generation = cpu_to_le32(osb->fs_generation);
 	dx_root->dr_blkno = cpu_to_le64(dr_blkno);
@@ -2458,10 +2450,7 @@ static int ocfs2_dx_dir_attach_index(struct ocfs2_super *osb,
 		dx_root->dr_list.l_count =
 			cpu_to_le16(ocfs2_extent_recs_per_dx_root(osb->sb));
 	}
-
-	ret = ocfs2_journal_dirty(handle, dx_root_bh);
-	if (ret)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, dx_root_bh);
 
 	ret = ocfs2_journal_access_di(handle, INODE_CACHE(dir), di_bh,
 				      OCFS2_JOURNAL_ACCESS_CREATE);
@@ -2475,9 +2464,7 @@ static int ocfs2_dx_dir_attach_index(struct ocfs2_super *osb,
 	OCFS2_I(dir)->ip_dyn_features |= OCFS2_INDEXED_DIR_FL;
 	di->i_dyn_features = cpu_to_le16(OCFS2_I(dir)->ip_dyn_features);
 
-	ret = ocfs2_journal_dirty(handle, di_bh);
-	if (ret)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, di_bh);
 
 	*ret_dx_root_bh = dx_root_bh;
 	dx_root_bh = NULL;
@@ -2558,7 +2545,7 @@ static int __ocfs2_dx_dir_new_cluster(struct inode *dir,
 	 * chance of contiguousness as the directory grows in number
 	 * of entries.
 	 */
-	ret = __ocfs2_claim_clusters(osb, handle, data_ac, 1, 1, &phys, &num);
+	ret = __ocfs2_claim_clusters(handle, data_ac, 1, 1, &phys, &num);
 	if (ret) {
 		mlog_errno(ret);
 		goto out;
@@ -2964,12 +2951,10 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
 		goto out;
 	}
 
-	if (vfs_dq_alloc_space_nodirty(dir,
-				ocfs2_clusters_to_bytes(osb->sb,
-							alloc + dx_alloc))) {
-		ret = -EDQUOT;
+	ret = dquot_alloc_space_nodirty(dir,
+		ocfs2_clusters_to_bytes(osb->sb, alloc + dx_alloc));
+	if (ret)
 		goto out_commit;
-	}
 	did_quota = 1;
 
 	if (ocfs2_supports_indexed_dirs(osb) && !dx_inline) {
@@ -2993,7 +2978,9 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
 	 * if we only get one now, that's enough to continue. The rest
 	 * will be claimed after the conversion to extents.
 	 */
-	ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off, &len);
+	if (ocfs2_dir_resv_allowed(osb))
+		data_ac->ac_resv = &oi->ip_la_data_resv;
+	ret = ocfs2_claim_clusters(handle, data_ac, 1, &bit_off, &len);
 	if (ret) {
 		mlog_errno(ret);
 		goto out_commit;
@@ -3036,11 +3023,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
 		ocfs2_init_dir_trailer(dir, dirdata_bh, i);
 	}
 
-	ret = ocfs2_journal_dirty(handle, dirdata_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out_commit;
-	}
+	ocfs2_journal_dirty(handle, dirdata_bh);
 
 	if (ocfs2_supports_indexed_dirs(osb) && !dx_inline) {
 		/*
@@ -3106,11 +3089,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
 	 */
 	dir->i_blocks = ocfs2_inode_sector_count(dir);
 
-	ret = ocfs2_journal_dirty(handle, di_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out_commit;
-	}
+	ocfs2_journal_dirty(handle, di_bh);
 
 	if (ocfs2_supports_indexed_dirs(osb)) {
 		ret = ocfs2_dx_dir_attach_index(osb, handle, dir, di_bh,
@@ -3140,7 +3119,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
 	 * pass. Claim the 2nd cluster as a separate extent.
 	 */
 	if (alloc > len) {
-		ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off,
+		ret = ocfs2_claim_clusters(handle, data_ac, 1, &bit_off,
 					   &len);
 		if (ret) {
 			mlog_errno(ret);
@@ -3178,7 +3157,7 @@ static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh,
 
 out_commit:
 	if (ret < 0 && did_quota)
-		vfs_dq_free_space_nodirty(dir, bytes_allocated);
+		dquot_free_space_nodirty(dir, bytes_allocated);
 
 	ocfs2_commit_trans(osb, handle);
 
@@ -3221,11 +3200,10 @@ static int ocfs2_do_extend_dir(struct super_block *sb,
 	if (extend) {
 		u32 offset = OCFS2_I(dir)->ip_clusters;
 
-		if (vfs_dq_alloc_space_nodirty(dir,
-					ocfs2_clusters_to_bytes(sb, 1))) {
-			status = -EDQUOT;
+		status = dquot_alloc_space_nodirty(dir,
+					ocfs2_clusters_to_bytes(sb, 1));
+		if (status)
 			goto bail;
-		}
 		did_quota = 1;
 
 		status = ocfs2_add_inode_data(OCFS2_SB(sb), dir, &offset,
@@ -3254,7 +3232,7 @@ static int ocfs2_do_extend_dir(struct super_block *sb,
 	status = 0;
 bail:
 	if (did_quota && status < 0)
-		vfs_dq_free_space_nodirty(dir, ocfs2_clusters_to_bytes(sb, 1));
+		dquot_free_space_nodirty(dir, ocfs2_clusters_to_bytes(sb, 1));
 	mlog_exit(status);
 	return status;
 }
@@ -3372,6 +3350,9 @@ static int ocfs2_extend_dir(struct ocfs2_super *osb,
 			goto bail;
 		}
 
+		if (ocfs2_dir_resv_allowed(osb))
+			data_ac->ac_resv = &OCFS2_I(dir)->ip_la_data_resv;
+
 		credits = ocfs2_calc_extend_credits(sb, el, 1);
 	} else {
 		spin_unlock(&OCFS2_I(dir)->ip_lock);
@@ -3426,11 +3407,7 @@ do_extend:
 	} else {
 		de->rec_len = cpu_to_le16(sb->s_blocksize);
 	}
-	status = ocfs2_journal_dirty(handle, new_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	ocfs2_journal_dirty(handle, new_bh);
 
 	dir_i_size += dir->i_sb->s_blocksize;
 	i_size_write(dir, dir_i_size);
@@ -3889,11 +3866,10 @@ static int ocfs2_dx_dir_rebalance(struct ocfs2_super *osb, struct inode *dir,
 		goto out;
 	}
 
-	if (vfs_dq_alloc_space_nodirty(dir,
-				       ocfs2_clusters_to_bytes(dir->i_sb, 1))) {
-		ret = -EDQUOT;
+	ret = dquot_alloc_space_nodirty(dir,
+				       ocfs2_clusters_to_bytes(dir->i_sb, 1));
+	if (ret)
 		goto out_commit;
-	}
 	did_quota = 1;
 
 	ret = ocfs2_journal_access_dl(handle, INODE_CACHE(dir), dx_leaf_bh,
@@ -3910,11 +3886,7 @@ static int ocfs2_dx_dir_rebalance(struct ocfs2_super *osb, struct inode *dir,
 	     sizeof(struct ocfs2_dx_entry), dx_leaf_sort_cmp,
 	     dx_leaf_sort_swap);
 
-	ret = ocfs2_journal_dirty(handle, dx_leaf_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out_commit;
-	}
+	ocfs2_journal_dirty(handle, dx_leaf_bh);
 
 	ret = ocfs2_dx_dir_find_leaf_split(dx_leaf, leaf_cpos, insert_hash,
 					   &split_hash);
@@ -3983,7 +3955,7 @@ static int ocfs2_dx_dir_rebalance(struct ocfs2_super *osb, struct inode *dir,
 
 out_commit:
 	if (ret < 0 && did_quota)
-		vfs_dq_free_space_nodirty(dir,
+		dquot_free_space_nodirty(dir,
 				ocfs2_clusters_to_bytes(dir->i_sb, 1));
 
 	ocfs2_commit_trans(osb, handle);
@@ -4165,11 +4137,10 @@ static int ocfs2_expand_inline_dx_root(struct inode *dir,
 		goto out;
 	}
 
-	if (vfs_dq_alloc_space_nodirty(dir,
-				       ocfs2_clusters_to_bytes(osb->sb, 1))) {
-		ret = -EDQUOT;
+	ret = dquot_alloc_space_nodirty(dir,
+				       ocfs2_clusters_to_bytes(osb->sb, 1));
+	if (ret)
 		goto out_commit;
-	}
 	did_quota = 1;
 
 	/*
@@ -4229,7 +4200,7 @@ static int ocfs2_expand_inline_dx_root(struct inode *dir,
 
 out_commit:
 	if (ret < 0 && did_quota)
-		vfs_dq_free_space_nodirty(dir,
+		dquot_free_space_nodirty(dir,
 					  ocfs2_clusters_to_bytes(dir->i_sb, 1));
 
 	ocfs2_commit_trans(osb, handle);
@@ -4495,7 +4466,10 @@ static int ocfs2_dx_dir_remove_index(struct inode *dir,
 
 	blk = le64_to_cpu(dx_root->dr_blkno);
 	bit = le16_to_cpu(dx_root->dr_suballoc_bit);
-	bg_blkno = ocfs2_which_suballoc_group(blk, bit);
+	if (dx_root->dr_suballoc_loc)
+		bg_blkno = le64_to_cpu(dx_root->dr_suballoc_loc);
+	else
+		bg_blkno = ocfs2_which_suballoc_group(blk, bit);
 	ret = ocfs2_free_suballoc_bits(handle, dx_alloc_inode, dx_alloc_bh,
 				       bit, bg_blkno, 1);
 	if (ret)
@@ -4556,8 +4530,8 @@ int ocfs2_dx_dir_truncate(struct inode *dir, struct buffer_head *di_bh)
 
 		p_cpos = ocfs2_blocks_to_clusters(dir->i_sb, blkno);
 
-		ret = ocfs2_remove_btree_range(dir, &et, cpos, p_cpos, clen,
-					       &dealloc);
+		ret = ocfs2_remove_btree_range(dir, &et, cpos, p_cpos, clen, 0,
+					       &dealloc, 0);
 		if (ret) {
 			mlog_errno(ret);
 			goto out;
diff --git a/fs/ocfs2/dlm/Makefile b/fs/ocfs2/dlm/Makefile
index 190361375700..dcebf0d920fa 100644
--- a/fs/ocfs2/dlm/Makefile
+++ b/fs/ocfs2/dlm/Makefile
@@ -1,8 +1,7 @@
 EXTRA_CFLAGS += -Ifs/ocfs2
 
-obj-$(CONFIG_OCFS2_FS_O2CB) += ocfs2_dlm.o ocfs2_dlmfs.o
+obj-$(CONFIG_OCFS2_FS_O2CB) += ocfs2_dlm.o
 
 ocfs2_dlm-objs := dlmdomain.o dlmdebug.o dlmthread.o dlmrecovery.o \
 	dlmmaster.o dlmast.o dlmconvert.o dlmlock.o dlmunlock.o dlmver.o
 
-ocfs2_dlmfs-objs := userdlm.o dlmfs.o dlmfsver.o
diff --git a/fs/ocfs2/dlm/dlmapi.h b/fs/ocfs2/dlm/dlmapi.h
index b5786a787fab..3cfa114aa391 100644
--- a/fs/ocfs2/dlm/dlmapi.h
+++ b/fs/ocfs2/dlm/dlmapi.h
@@ -95,7 +95,7 @@ const char *dlm_errname(enum dlm_status err);
 		mlog(ML_ERROR, "dlm status = %s\n", dlm_errname((st)));	\
 } while (0)
 
-#define DLM_LKSB_UNUSED1           0x01  
+#define DLM_LKSB_UNUSED1           0x01
 #define DLM_LKSB_PUT_LVB           0x02
 #define DLM_LKSB_GET_LVB           0x04
 #define DLM_LKSB_UNUSED2           0x08
diff --git a/fs/ocfs2/dlm/dlmast.c b/fs/ocfs2/dlm/dlmast.c
index 01cf8cc3d286..f44999156839 100644
--- a/fs/ocfs2/dlm/dlmast.c
+++ b/fs/ocfs2/dlm/dlmast.c
@@ -28,7 +28,6 @@
 #include <linux/module.h>
 #include <linux/fs.h>
 #include <linux/types.h>
-#include <linux/slab.h>
 #include <linux/highmem.h>
 #include <linux/init.h>
 #include <linux/sysctl.h>
@@ -89,7 +88,7 @@ static int dlm_should_cancel_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
 	return 0;
 }
 
-static void __dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
+void __dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
 {
 	mlog_entry_void();
 
@@ -123,7 +122,7 @@ static void __dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
 		dlm_lock_put(lock);
 		/* free up the reserved bast that we are cancelling.
 		 * guaranteed that this will not be the last reserved
-		 * ast because *both* an ast and a bast were reserved 
+		 * ast because *both* an ast and a bast were reserved
 		 * to get to this point.  the res->spinlock will not be
 		 * taken here */
 		dlm_lockres_release_ast(dlm, res);
@@ -146,7 +145,7 @@ void dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
 }
 
 
-static void __dlm_queue_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
+void __dlm_queue_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock)
 {
 	mlog_entry_void();
 
@@ -185,9 +184,8 @@ static void dlm_update_lvb(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
 	BUG_ON(!lksb);
 
 	/* only updates if this node masters the lockres */
+	spin_lock(&res->spinlock);
 	if (res->owner == dlm->node_num) {
-
-		spin_lock(&res->spinlock);
 		/* check the lksb flags for the direction */
 		if (lksb->flags & DLM_LKSB_GET_LVB) {
 			mlog(0, "getting lvb from lockres for %s node\n",
@@ -202,8 +200,8 @@ static void dlm_update_lvb(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
  		 * here. In the future we might want to clear it at the time
  		 * the put is actually done.
 		 */
-		spin_unlock(&res->spinlock);
 	}
+	spin_unlock(&res->spinlock);
 
 	/* reset any lvb flags on the lksb */
 	lksb->flags &= ~(DLM_LKSB_PUT_LVB|DLM_LKSB_GET_LVB);
@@ -453,7 +451,9 @@ int dlm_send_proxy_ast_msg(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
 	ret = o2net_send_message_vec(DLM_PROXY_AST_MSG, dlm->key, vec, veclen,
 				     lock->ml.node, &status);
 	if (ret < 0)
-		mlog_errno(ret);
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", ret, DLM_PROXY_AST_MSG, dlm->key,
+		     lock->ml.node);
 	else {
 		if (status == DLM_RECOVERING) {
 			mlog(ML_ERROR, "sent AST to node %u, it thinks this "
diff --git a/fs/ocfs2/dlm/dlmcommon.h b/fs/ocfs2/dlm/dlmcommon.h
index 0102be35980c..4b6ae2c13b47 100644
--- a/fs/ocfs2/dlm/dlmcommon.h
+++ b/fs/ocfs2/dlm/dlmcommon.h
@@ -37,7 +37,7 @@
 #define DLM_THREAD_SHUFFLE_INTERVAL    5     // flush everything every 5 passes
 #define DLM_THREAD_MS                  200   // flush at least every 200 ms
 
-#define DLM_HASH_SIZE_DEFAULT	(1 << 14)
+#define DLM_HASH_SIZE_DEFAULT	(1 << 17)
 #if DLM_HASH_SIZE_DEFAULT < PAGE_SIZE
 # define DLM_HASH_PAGES		1
 #else
@@ -904,6 +904,8 @@ void __dlm_lockres_grab_inflight_ref(struct dlm_ctxt *dlm,
 
 void dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock);
 void dlm_queue_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock);
+void __dlm_queue_ast(struct dlm_ctxt *dlm, struct dlm_lock *lock);
+void __dlm_queue_bast(struct dlm_ctxt *dlm, struct dlm_lock *lock);
 void dlm_do_local_ast(struct dlm_ctxt *dlm,
 		      struct dlm_lock_resource *res,
 		      struct dlm_lock *lock);
diff --git a/fs/ocfs2/dlm/dlmconvert.c b/fs/ocfs2/dlm/dlmconvert.c
index ca96bce50e18..9f30491e5e88 100644
--- a/fs/ocfs2/dlm/dlmconvert.c
+++ b/fs/ocfs2/dlm/dlmconvert.c
@@ -28,7 +28,6 @@
 #include <linux/module.h>
 #include <linux/fs.h>
 #include <linux/types.h>
-#include <linux/slab.h>
 #include <linux/highmem.h>
 #include <linux/init.h>
 #include <linux/sysctl.h>
@@ -391,12 +390,14 @@ static enum dlm_status dlm_send_remote_convert_request(struct dlm_ctxt *dlm,
 		} else if (ret != DLM_NORMAL && ret != DLM_NOTQUEUED)
 			dlm_error(ret);
 	} else {
-		mlog_errno(tmpret);
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", tmpret, DLM_CONVERT_LOCK_MSG, dlm->key,
+		     res->owner);
 		if (dlm_is_host_down(tmpret)) {
 			/* instead of logging the same network error over
 			 * and over, sleep here and wait for the heartbeat
 			 * to notice the node is dead.  times out after 5s. */
-			dlm_wait_for_node_death(dlm, res->owner, 
+			dlm_wait_for_node_death(dlm, res->owner,
 						DLM_NODE_DEATH_WAIT_MAX);
 			ret = DLM_RECOVERING;
 			mlog(0, "node %u died so returning DLM_RECOVERING "
diff --git a/fs/ocfs2/dlm/dlmdebug.c b/fs/ocfs2/dlm/dlmdebug.c
index 42b0bad7a612..0cd24cf54396 100644
--- a/fs/ocfs2/dlm/dlmdebug.c
+++ b/fs/ocfs2/dlm/dlmdebug.c
@@ -102,7 +102,7 @@ void __dlm_print_one_lock_resource(struct dlm_lock_resource *res)
 	assert_spin_locked(&res->spinlock);
 
 	stringify_lockname(res->lockname.name, res->lockname.len,
-			   buf, sizeof(buf) - 1);
+			   buf, sizeof(buf));
 	printk("lockres: %s, owner=%u, state=%u\n",
 	       buf, res->owner, res->state);
 	printk("  last used: %lu, refcnt: %u, on purge list: %s\n",
diff --git a/fs/ocfs2/dlm/dlmdomain.c b/fs/ocfs2/dlm/dlmdomain.c
index 0334000676d3..153abb5abef0 100644
--- a/fs/ocfs2/dlm/dlmdomain.c
+++ b/fs/ocfs2/dlm/dlmdomain.c
@@ -511,7 +511,7 @@ static void __dlm_print_nodes(struct dlm_ctxt *dlm)
 
 	assert_spin_locked(&dlm->spinlock);
 
-	printk(KERN_INFO "ocfs2_dlm: Nodes in domain (\"%s\"): ", dlm->name);
+	printk(KERN_NOTICE "o2dlm: Nodes in domain %s: ", dlm->name);
 
 	while ((node = find_next_bit(dlm->domain_map, O2NM_MAX_NODES,
 				     node + 1)) < O2NM_MAX_NODES) {
@@ -534,7 +534,7 @@ static int dlm_exit_domain_handler(struct o2net_msg *msg, u32 len, void *data,
 
 	node = exit_msg->node_idx;
 
-	printk(KERN_INFO "ocfs2_dlm: Node %u leaves domain %s\n", node, dlm->name);
+	printk(KERN_NOTICE "o2dlm: Node %u leaves domain %s\n", node, dlm->name);
 
 	spin_lock(&dlm->spinlock);
 	clear_bit(node, dlm->domain_map);
@@ -565,7 +565,9 @@ static int dlm_send_one_domain_exit(struct dlm_ctxt *dlm,
 	status = o2net_send_message(DLM_EXIT_DOMAIN_MSG, dlm->key,
 				    &leave_msg, sizeof(leave_msg), node,
 				    NULL);
-
+	if (status < 0)
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", status, DLM_EXIT_DOMAIN_MSG, dlm->key, node);
 	mlog(0, "status return %d from o2net_send_message\n", status);
 
 	return status;
@@ -816,7 +818,7 @@ static int dlm_query_join_handler(struct o2net_msg *msg, u32 len, void *data,
 	}
 
 	/* Once the dlm ctxt is marked as leaving then we don't want
-	 * to be put in someone's domain map. 
+	 * to be put in someone's domain map.
 	 * Also, explicitly disallow joining at certain troublesome
 	 * times (ie. during recovery). */
 	if (dlm && dlm->dlm_state != DLM_CTXT_LEAVING) {
@@ -904,7 +906,7 @@ static int dlm_assert_joined_handler(struct o2net_msg *msg, u32 len, void *data,
 		set_bit(assert->node_idx, dlm->domain_map);
 		__dlm_set_joining_node(dlm, DLM_LOCK_RES_OWNER_UNKNOWN);
 
-		printk(KERN_INFO "ocfs2_dlm: Node %u joins domain %s\n",
+		printk(KERN_NOTICE "o2dlm: Node %u joins domain %s\n",
 		       assert->node_idx, dlm->name);
 		__dlm_print_nodes(dlm);
 
@@ -962,7 +964,9 @@ static int dlm_send_one_join_cancel(struct dlm_ctxt *dlm,
 				    &cancel_msg, sizeof(cancel_msg), node,
 				    NULL);
 	if (status < 0) {
-		mlog_errno(status);
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", status, DLM_CANCEL_JOIN_MSG, DLM_MOD_KEY,
+		     node);
 		goto bail;
 	}
 
@@ -1029,10 +1033,11 @@ static int dlm_request_join(struct dlm_ctxt *dlm,
 	byte_copymap(join_msg.node_map, dlm->live_nodes_map, O2NM_MAX_NODES);
 
 	status = o2net_send_message(DLM_QUERY_JOIN_MSG, DLM_MOD_KEY, &join_msg,
-				    sizeof(join_msg), node,
-				    &join_resp);
+				    sizeof(join_msg), node, &join_resp);
 	if (status < 0 && status != -ENOPROTOOPT) {
-		mlog_errno(status);
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", status, DLM_QUERY_JOIN_MSG, DLM_MOD_KEY,
+		     node);
 		goto bail;
 	}
 	dlm_query_join_wire_to_packet(join_resp, &packet);
@@ -1103,7 +1108,9 @@ static int dlm_send_one_join_assert(struct dlm_ctxt *dlm,
 				    &assert_msg, sizeof(assert_msg), node,
 				    NULL);
 	if (status < 0)
-		mlog_errno(status);
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", status, DLM_ASSERT_JOINED_MSG, DLM_MOD_KEY,
+		     node);
 
 	return status;
 }
@@ -1516,7 +1523,7 @@ static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain,
 		goto leave;
 	}
 
-	dlm->name = kmalloc(strlen(domain) + 1, GFP_KERNEL);
+	dlm->name = kstrdup(domain, GFP_KERNEL);
 	if (dlm->name == NULL) {
 		mlog_errno(-ENOMEM);
 		kfree(dlm);
@@ -1550,7 +1557,6 @@ static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain,
 	for (i = 0; i < DLM_HASH_BUCKETS; i++)
 		INIT_HLIST_HEAD(dlm_master_hash(dlm, i));
 
-	strcpy(dlm->name, domain);
 	dlm->key = key;
 	dlm->node_num = o2nm_this_node();
 
@@ -1665,7 +1671,7 @@ struct dlm_ctxt * dlm_register_domain(const char *domain,
 	struct dlm_ctxt *dlm = NULL;
 	struct dlm_ctxt *new_ctxt = NULL;
 
-	if (strlen(domain) > O2NM_MAX_NAME_LEN) {
+	if (strlen(domain) >= O2NM_MAX_NAME_LEN) {
 		ret = -ENAMETOOLONG;
 		mlog(ML_ERROR, "domain name length too long\n");
 		goto leave;
@@ -1703,6 +1709,7 @@ retry:
 		}
 
 		if (dlm_protocol_compare(&dlm->fs_locking_proto, fs_proto)) {
+			spin_unlock(&dlm_domain_lock);
 			mlog(ML_ERROR,
 			     "Requested locking protocol version is not "
 			     "compatible with already registered domain "
diff --git a/fs/ocfs2/dlm/dlmlock.c b/fs/ocfs2/dlm/dlmlock.c
index 437698e9465f..69cf369961c4 100644
--- a/fs/ocfs2/dlm/dlmlock.c
+++ b/fs/ocfs2/dlm/dlmlock.c
@@ -269,7 +269,7 @@ static enum dlm_status dlmlock_remote(struct dlm_ctxt *dlm,
 		}
 		dlm_revert_pending_lock(res, lock);
 		dlm_lock_put(lock);
-	} else if (dlm_is_recovery_lock(res->lockname.name, 
+	} else if (dlm_is_recovery_lock(res->lockname.name,
 					res->lockname.len)) {
 		/* special case for the $RECOVERY lock.
 		 * there will never be an AST delivered to put
@@ -329,7 +329,9 @@ static enum dlm_status dlm_send_remote_lock_request(struct dlm_ctxt *dlm,
 			BUG();
 		}
 	} else {
-		mlog_errno(tmpret);
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", tmpret, DLM_CREATE_LOCK_MSG, dlm->key,
+		     res->owner);
 		if (dlm_is_host_down(tmpret)) {
 			ret = DLM_RECOVERING;
 			mlog(0, "node %u died so returning DLM_RECOVERING "
@@ -429,7 +431,7 @@ struct dlm_lock * dlm_new_lock(int type, u8 node, u64 cookie,
 	struct dlm_lock *lock;
 	int kernel_allocated = 0;
 
-	lock = (struct dlm_lock *) kmem_cache_zalloc(dlm_lock_cache, GFP_NOFS);
+	lock = kmem_cache_zalloc(dlm_lock_cache, GFP_NOFS);
 	if (!lock)
 		return NULL;
 
diff --git a/fs/ocfs2/dlm/dlmmaster.c b/fs/ocfs2/dlm/dlmmaster.c
index 03ccf9a7b1f4..94b97fc6a88e 100644
--- a/fs/ocfs2/dlm/dlmmaster.c
+++ b/fs/ocfs2/dlm/dlmmaster.c
@@ -366,7 +366,7 @@ void dlm_hb_event_notify_attached(struct dlm_ctxt *dlm, int idx, int node_up)
 	struct dlm_master_list_entry *mle;
 
 	assert_spin_locked(&dlm->spinlock);
-	
+
 	list_for_each_entry(mle, &dlm->mle_hb_events, hb_events) {
 		if (node_up)
 			dlm_mle_node_up(dlm, mle, NULL, idx);
@@ -617,13 +617,11 @@ struct dlm_lock_resource *dlm_new_lockres(struct dlm_ctxt *dlm,
 {
 	struct dlm_lock_resource *res = NULL;
 
-	res = (struct dlm_lock_resource *)
-				kmem_cache_zalloc(dlm_lockres_cache, GFP_NOFS);
+	res = kmem_cache_zalloc(dlm_lockres_cache, GFP_NOFS);
 	if (!res)
 		goto error;
 
-	res->lockname.name = (char *)
-				kmem_cache_zalloc(dlm_lockname_cache, GFP_NOFS);
+	res->lockname.name = kmem_cache_zalloc(dlm_lockname_cache, GFP_NOFS);
 	if (!res->lockname.name)
 		goto error;
 
@@ -757,8 +755,7 @@ lookup:
 		spin_unlock(&dlm->spinlock);
 		mlog(0, "allocating a new resource\n");
 		/* nothing found and we need to allocate one. */
-		alloc_mle = (struct dlm_master_list_entry *)
-			kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
+		alloc_mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
 		if (!alloc_mle)
 			goto leave;
 		res = dlm_new_lockres(dlm, lockid, namelen);
@@ -833,7 +830,7 @@ lookup:
 		__dlm_insert_mle(dlm, mle);
 
 		/* still holding the dlm spinlock, check the recovery map
-		 * to see if there are any nodes that still need to be 
+		 * to see if there are any nodes that still need to be
 		 * considered.  these will not appear in the mle nodemap
 		 * but they might own this lockres.  wait on them. */
 		bit = find_next_bit(dlm->recovery_map, O2NM_MAX_NODES, 0);
@@ -883,7 +880,7 @@ redo_request:
 				msleep(500);
 			}
 			continue;
-		} 
+		}
 
 		dlm_kick_recovery_thread(dlm);
 		msleep(1000);
@@ -939,8 +936,8 @@ wait:
 		     res->lockname.name, blocked);
 		if (++tries > 20) {
 			mlog(ML_ERROR, "%s:%.*s: spinning on "
-			     "dlm_wait_for_lock_mastery, blocked=%d\n", 
-			     dlm->name, res->lockname.len, 
+			     "dlm_wait_for_lock_mastery, blocked=%d\n",
+			     dlm->name, res->lockname.len,
 			     res->lockname.name, blocked);
 			dlm_print_one_lock_resource(res);
 			dlm_print_one_mle(mle);
@@ -1029,7 +1026,7 @@ recheck:
 		ret = dlm_restart_lock_mastery(dlm, res, mle, *blocked);
 		b = (mle->type == DLM_MLE_BLOCK);
 		if ((*blocked && !b) || (!*blocked && b)) {
-			mlog(0, "%s:%.*s: status change: old=%d new=%d\n", 
+			mlog(0, "%s:%.*s: status change: old=%d new=%d\n",
 			     dlm->name, res->lockname.len, res->lockname.name,
 			     *blocked, b);
 			*blocked = b;
@@ -1542,8 +1539,7 @@ way_up_top:
 			spin_unlock(&dlm->master_lock);
 			spin_unlock(&dlm->spinlock);
 
-			mle = (struct dlm_master_list_entry *)
-				kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
+			mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
 			if (!mle) {
 				response = DLM_MASTER_RESP_ERROR;
 				mlog_errno(-ENOMEM);
@@ -1602,7 +1598,7 @@ send_response:
 		}
 		mlog(0, "%u is the owner of %.*s, cleaning everyone else\n",
 			     dlm->node_num, res->lockname.len, res->lockname.name);
-		ret = dlm_dispatch_assert_master(dlm, res, 0, request->node_idx, 
+		ret = dlm_dispatch_assert_master(dlm, res, 0, request->node_idx,
 						 DLM_ASSERT_MASTER_MLE_CLEANUP);
 		if (ret < 0) {
 			mlog(ML_ERROR, "failed to dispatch assert master work\n");
@@ -1666,7 +1662,9 @@ again:
 		tmpret = o2net_send_message(DLM_ASSERT_MASTER_MSG, dlm->key,
 					    &assert, sizeof(assert), to, &r);
 		if (tmpret < 0) {
-			mlog(0, "assert_master returned %d!\n", tmpret);
+			mlog(ML_ERROR, "Error %d when sending message %u (key "
+			     "0x%x) to node %u\n", tmpret,
+			     DLM_ASSERT_MASTER_MSG, dlm->key, to);
 			if (!dlm_is_host_down(tmpret)) {
 				mlog(ML_ERROR, "unhandled error=%d!\n", tmpret);
 				BUG();
@@ -1701,7 +1699,7 @@ again:
 
 		if (r & DLM_ASSERT_RESPONSE_REASSERT) {
 			mlog(0, "%.*s: node %u create mles on other "
-			     "nodes and requests a re-assert\n", 
+			     "nodes and requests a re-assert\n",
 			     namelen, lockname, to);
 			reassert = 1;
 		}
@@ -1812,7 +1810,7 @@ int dlm_assert_master_handler(struct o2net_msg *msg, u32 len, void *data,
 				spin_unlock(&dlm->master_lock);
 				spin_unlock(&dlm->spinlock);
 				goto done;
-			}	
+			}
 		}
 	}
 	spin_unlock(&dlm->master_lock);
@@ -1875,7 +1873,6 @@ int dlm_assert_master_handler(struct o2net_msg *msg, u32 len, void *data,
 ok:
 		spin_unlock(&res->spinlock);
 	}
-	spin_unlock(&dlm->spinlock);
 
 	// mlog(0, "woo!  got an assert_master from node %u!\n",
 	// 	     assert->node_idx);
@@ -1883,7 +1880,7 @@ ok:
 		int extra_ref = 0;
 		int nn = -1;
 		int rr, err = 0;
-		
+
 		spin_lock(&mle->spinlock);
 		if (mle->type == DLM_MLE_BLOCK || mle->type == DLM_MLE_MIGRATION)
 			extra_ref = 1;
@@ -1891,7 +1888,7 @@ ok:
 			/* MASTER mle: if any bits set in the response map
 			 * then the calling node needs to re-assert to clear
 			 * up nodes that this node contacted */
-			while ((nn = find_next_bit (mle->response_map, O2NM_MAX_NODES, 
+			while ((nn = find_next_bit (mle->response_map, O2NM_MAX_NODES,
 						    nn+1)) < O2NM_MAX_NODES) {
 				if (nn != dlm->node_num && nn != assert->node_idx)
 					master_request = 1;
@@ -1926,7 +1923,6 @@ ok:
 		/* master is known, detach if not already detached.
 		 * ensures that only one assert_master call will happen
 		 * on this mle. */
-		spin_lock(&dlm->spinlock);
 		spin_lock(&dlm->master_lock);
 
 		rr = atomic_read(&mle->mle_refs.refcount);
@@ -1959,7 +1955,6 @@ ok:
 			__dlm_put_mle(mle);
 		}
 		spin_unlock(&dlm->master_lock);
-		spin_unlock(&dlm->spinlock);
 	} else if (res) {
 		if (res->owner != assert->node_idx) {
 			mlog(0, "assert_master from %u, but current "
@@ -1967,6 +1962,7 @@ ok:
 			     res->owner, namelen, name);
 		}
 	}
+	spin_unlock(&dlm->spinlock);
 
 done:
 	ret = 0;
@@ -2002,7 +1998,7 @@ kill:
 	__dlm_print_one_lock_resource(res);
 	spin_unlock(&res->spinlock);
 	spin_unlock(&dlm->spinlock);
-	*ret_data = (void *)res; 
+	*ret_data = (void *)res;
 	dlm_put(dlm);
 	return -EINVAL;
 }
@@ -2040,10 +2036,10 @@ int dlm_dispatch_assert_master(struct dlm_ctxt *dlm,
 	item->u.am.request_from = request_from;
 	item->u.am.flags = flags;
 
-	if (ignore_higher) 
-		mlog(0, "IGNORE HIGHER: %.*s\n", res->lockname.len, 
+	if (ignore_higher)
+		mlog(0, "IGNORE HIGHER: %.*s\n", res->lockname.len,
 		     res->lockname.name);
-		
+
 	spin_lock(&dlm->work_lock);
 	list_add_tail(&item->list, &dlm->work_list);
 	spin_unlock(&dlm->work_lock);
@@ -2133,7 +2129,7 @@ put:
  * think that $RECOVERY is currently mastered by a dead node.  If so,
  * we wait a short time to allow that node to get notified by its own
  * heartbeat stack, then check again.  All $RECOVERY lock resources
- * mastered by dead nodes are purged when the hearbeat callback is 
+ * mastered by dead nodes are purged when the hearbeat callback is
  * fired, so we can know for sure that it is safe to continue once
  * the node returns a live node or no node.  */
 static int dlm_pre_master_reco_lockres(struct dlm_ctxt *dlm,
@@ -2174,7 +2170,7 @@ static int dlm_pre_master_reco_lockres(struct dlm_ctxt *dlm,
 				ret = -EAGAIN;
 			}
 			spin_unlock(&dlm->spinlock);
-			mlog(0, "%s: reco lock master is %u\n", dlm->name, 
+			mlog(0, "%s: reco lock master is %u\n", dlm->name,
 			     master);
 			break;
 		}
@@ -2207,7 +2203,9 @@ int dlm_drop_lockres_ref(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
 	ret = o2net_send_message(DLM_DEREF_LOCKRES_MSG, dlm->key,
 				 &deref, sizeof(deref), res->owner, &r);
 	if (ret < 0)
-		mlog_errno(ret);
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", ret, DLM_DEREF_LOCKRES_MSG, dlm->key,
+		     res->owner);
 	else if (r < 0) {
 		/* BAD.  other node says I did not have a ref. */
 		mlog(ML_ERROR,"while dropping ref on %s:%.*s "
@@ -2454,8 +2452,7 @@ static int dlm_migrate_lockres(struct dlm_ctxt *dlm,
 		goto leave;
 	}
 
-	mle = (struct dlm_master_list_entry *) kmem_cache_alloc(dlm_mle_cache,
-								GFP_NOFS);
+	mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
 	if (!mle) {
 		mlog_errno(ret);
 		goto leave;
@@ -2602,7 +2599,7 @@ fail:
 
 			mlog(0, "%s:%.*s: timed out during migration\n",
 			     dlm->name, res->lockname.len, res->lockname.name);
-			/* avoid hang during shutdown when migrating lockres 
+			/* avoid hang during shutdown when migrating lockres
 			 * to a node which also goes down */
 			if (dlm_is_node_dead(dlm, target)) {
 				mlog(0, "%s:%.*s: expected migration "
@@ -2738,7 +2735,7 @@ static int dlm_migration_can_proceed(struct dlm_ctxt *dlm,
 	can_proceed = !!(res->state & DLM_LOCK_RES_MIGRATING);
 	spin_unlock(&res->spinlock);
 
-	/* target has died, so make the caller break out of the 
+	/* target has died, so make the caller break out of the
 	 * wait_event, but caller must recheck the domain_map */
 	spin_lock(&dlm->spinlock);
 	if (!test_bit(mig_target, dlm->domain_map))
@@ -2811,14 +2808,8 @@ again:
 		mlog(0, "trying again...\n");
 		goto again;
 	}
-	/* now that we are sure the MIGRATING state is there, drop
-	 * the unneded state which blocked threads trying to DIRTY */
-	spin_lock(&res->spinlock);
-	BUG_ON(!(res->state & DLM_LOCK_RES_BLOCK_DIRTY));
-	BUG_ON(!(res->state & DLM_LOCK_RES_MIGRATING));
-	res->state &= ~DLM_LOCK_RES_BLOCK_DIRTY;
-	spin_unlock(&res->spinlock);
 
+	ret = 0;
 	/* did the target go down or die? */
 	spin_lock(&dlm->spinlock);
 	if (!test_bit(target, dlm->domain_map)) {
@@ -2829,9 +2820,21 @@ again:
 	spin_unlock(&dlm->spinlock);
 
 	/*
+	 * if target is down, we need to clear DLM_LOCK_RES_BLOCK_DIRTY for
+	 * another try; otherwise, we are sure the MIGRATING state is there,
+	 * drop the unneded state which blocked threads trying to DIRTY
+	 */
+	spin_lock(&res->spinlock);
+	BUG_ON(!(res->state & DLM_LOCK_RES_BLOCK_DIRTY));
+	res->state &= ~DLM_LOCK_RES_BLOCK_DIRTY;
+	if (!ret)
+		BUG_ON(!(res->state & DLM_LOCK_RES_MIGRATING));
+	spin_unlock(&res->spinlock);
+
+	/*
 	 * at this point:
 	 *
-	 *   o the DLM_LOCK_RES_MIGRATING flag is set
+	 *   o the DLM_LOCK_RES_MIGRATING flag is set if target not down
 	 *   o there are no pending asts on this lockres
 	 *   o all processes trying to reserve an ast on this
 	 *     lockres must wait for the MIGRATING flag to clear
@@ -2977,7 +2980,9 @@ static int dlm_do_migrate_request(struct dlm_ctxt *dlm,
 					 &migrate, sizeof(migrate), nodenum,
 					 &status);
 		if (ret < 0) {
-			mlog(0, "migrate_request returned %d!\n", ret);
+			mlog(ML_ERROR, "Error %d when sending message %u (key "
+			     "0x%x) to node %u\n", ret, DLM_MIGRATE_REQUEST_MSG,
+			     dlm->key, nodenum);
 			if (!dlm_is_host_down(ret)) {
 				mlog(ML_ERROR, "unhandled error=%d!\n", ret);
 				BUG();
@@ -3035,8 +3040,7 @@ int dlm_migrate_request_handler(struct o2net_msg *msg, u32 len, void *data,
 	hash = dlm_lockid_hash(name, namelen);
 
 	/* preallocate.. if this fails, abort */
-	mle = (struct dlm_master_list_entry *) kmem_cache_alloc(dlm_mle_cache,
-							 GFP_NOFS);
+	mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
 
 	if (!mle) {
 		ret = -ENOMEM;
diff --git a/fs/ocfs2/dlm/dlmrecovery.c b/fs/ocfs2/dlm/dlmrecovery.c
index 2f9e4e19a4f2..9dfaac73b36d 100644
--- a/fs/ocfs2/dlm/dlmrecovery.c
+++ b/fs/ocfs2/dlm/dlmrecovery.c
@@ -310,7 +310,7 @@ static int dlm_recovery_thread(void *data)
 	mlog(0, "dlm thread running for %s...\n", dlm->name);
 
 	while (!kthread_should_stop()) {
-		if (dlm_joined(dlm)) {
+		if (dlm_domain_fully_joined(dlm)) {
 			status = dlm_do_recovery(dlm);
 			if (status == -EAGAIN) {
 				/* do not sleep, recheck immediately. */
@@ -463,7 +463,7 @@ static int dlm_do_recovery(struct dlm_ctxt *dlm)
 	if (dlm->reco.dead_node == O2NM_INVALID_NODE_NUM) {
 		int bit;
 
-		bit = find_next_bit (dlm->recovery_map, O2NM_MAX_NODES+1, 0);
+		bit = find_next_bit (dlm->recovery_map, O2NM_MAX_NODES, 0);
 		if (bit >= O2NM_MAX_NODES || bit < 0)
 			dlm_set_reco_dead_node(dlm, O2NM_INVALID_NODE_NUM);
 		else
@@ -803,7 +803,9 @@ static int dlm_request_all_locks(struct dlm_ctxt *dlm, u8 request_from,
 
 	/* negative status is handled by caller */
 	if (ret < 0)
-		mlog_errno(ret);
+		mlog(ML_ERROR, "Error %d when sending message %u (key "
+		     "0x%x) to node %u\n", ret, DLM_LOCK_REQUEST_MSG,
+		     dlm->key, request_from);
 
 	// return from here, then
 	// sleep until all received or error
@@ -955,10 +957,10 @@ static int dlm_send_all_done_msg(struct dlm_ctxt *dlm, u8 dead_node, u8 send_to)
 	ret = o2net_send_message(DLM_RECO_DATA_DONE_MSG, dlm->key, &done_msg,
 				 sizeof(done_msg), send_to, &tmpret);
 	if (ret < 0) {
+		mlog(ML_ERROR, "Error %d when sending message %u (key "
+		     "0x%x) to node %u\n", ret, DLM_RECO_DATA_DONE_MSG,
+		     dlm->key, send_to);
 		if (!dlm_is_host_down(ret)) {
-			mlog_errno(ret);
-			mlog(ML_ERROR, "%s: unknown error sending data-done "
-			     "to %u\n", dlm->name, send_to);
 			BUG();
 		}
 	} else
@@ -1050,7 +1052,7 @@ static void dlm_move_reco_locks_to_list(struct dlm_ctxt *dlm,
 				if (lock->ml.node == dead_node) {
 					mlog(0, "AHA! there was "
 					     "a $RECOVERY lock for dead "
-					     "node %u (%s)!\n", 
+					     "node %u (%s)!\n",
 					     dead_node, dlm->name);
 					list_del_init(&lock->list);
 					dlm_lock_put(lock);
@@ -1126,7 +1128,9 @@ static int dlm_send_mig_lockres_msg(struct dlm_ctxt *dlm,
 	if (ret < 0) {
 		/* XXX: negative status is not handled.
 		 * this will end up killing this node. */
-		mlog_errno(ret);
+		mlog(ML_ERROR, "Error %d when sending message %u (key "
+		     "0x%x) to node %u\n", ret, DLM_MIG_LOCKRES_MSG,
+		     dlm->key, send_to);
 	} else {
 		/* might get an -ENOMEM back here */
 		ret = status;
@@ -1164,6 +1168,39 @@ static void dlm_init_migratable_lockres(struct dlm_migratable_lockres *mres,
 	mres->master = master;
 }
 
+static void dlm_prepare_lvb_for_migration(struct dlm_lock *lock,
+					  struct dlm_migratable_lockres *mres,
+					  int queue)
+{
+	if (!lock->lksb)
+	       return;
+
+	/* Ignore lvb in all locks in the blocked list */
+	if (queue == DLM_BLOCKED_LIST)
+		return;
+
+	/* Only consider lvbs in locks with granted EX or PR lock levels */
+	if (lock->ml.type != LKM_EXMODE && lock->ml.type != LKM_PRMODE)
+		return;
+
+	if (dlm_lvb_is_empty(mres->lvb)) {
+		memcpy(mres->lvb, lock->lksb->lvb, DLM_LVB_LEN);
+		return;
+	}
+
+	/* Ensure the lvb copied for migration matches in other valid locks */
+	if (!memcmp(mres->lvb, lock->lksb->lvb, DLM_LVB_LEN))
+		return;
+
+	mlog(ML_ERROR, "Mismatched lvb in lock cookie=%u:%llu, name=%.*s, "
+	     "node=%u\n",
+	     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
+	     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
+	     lock->lockres->lockname.len, lock->lockres->lockname.name,
+	     lock->ml.node);
+	dlm_print_one_lock_resource(lock->lockres);
+	BUG();
+}
 
 /* returns 1 if this lock fills the network structure,
  * 0 otherwise */
@@ -1181,20 +1218,7 @@ static int dlm_add_lock_to_array(struct dlm_lock *lock,
 	ml->list = queue;
 	if (lock->lksb) {
 		ml->flags = lock->lksb->flags;
-		/* send our current lvb */
-		if (ml->type == LKM_EXMODE ||
-		    ml->type == LKM_PRMODE) {
-			/* if it is already set, this had better be a PR
-			 * and it has to match */
-			if (!dlm_lvb_is_empty(mres->lvb) &&
-			    (ml->type == LKM_EXMODE ||
-			     memcmp(mres->lvb, lock->lksb->lvb, DLM_LVB_LEN))) {
-				mlog(ML_ERROR, "mismatched lvbs!\n");
-				dlm_print_one_lock_resource(lock->lockres);
-				BUG();
-			}
-			memcpy(mres->lvb, lock->lksb->lvb, DLM_LVB_LEN);
-		}
+		dlm_prepare_lvb_for_migration(lock, mres, queue);
 	}
 	ml->node = lock->ml.node;
 	mres->num_locks++;
@@ -1622,7 +1646,9 @@ int dlm_do_master_requery(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
 				 &req, sizeof(req), nodenum, &status);
 	/* XXX: negative status not handled properly here. */
 	if (ret < 0)
-		mlog_errno(ret);
+		mlog(ML_ERROR, "Error %d when sending message %u (key "
+		     "0x%x) to node %u\n", ret, DLM_MASTER_REQUERY_MSG,
+		     dlm->key, nodenum);
 	else {
 		BUG_ON(status < 0);
 		BUG_ON(status > DLM_LOCK_RES_OWNER_UNKNOWN);
@@ -1730,6 +1756,7 @@ static int dlm_process_recovery_data(struct dlm_ctxt *dlm,
 	struct dlm_lock *lock = NULL;
 	u8 from = O2NM_MAX_NODES;
 	unsigned int added = 0;
+	__be64 c;
 
 	mlog(0, "running %d locks for this lockres\n", mres->num_locks);
 	for (i=0; i<mres->num_locks; i++) {
@@ -1777,19 +1804,48 @@ static int dlm_process_recovery_data(struct dlm_ctxt *dlm,
 			/* lock is always created locally first, and
 			 * destroyed locally last.  it must be on the list */
 			if (!lock) {
-				__be64 c = ml->cookie;
-				mlog(ML_ERROR, "could not find local lock "
-					       "with cookie %u:%llu!\n",
+				c = ml->cookie;
+				mlog(ML_ERROR, "Could not find local lock "
+					       "with cookie %u:%llu, node %u, "
+					       "list %u, flags 0x%x, type %d, "
+					       "conv %d, highest blocked %d\n",
 				     dlm_get_lock_cookie_node(be64_to_cpu(c)),
-				     dlm_get_lock_cookie_seq(be64_to_cpu(c)));
+				     dlm_get_lock_cookie_seq(be64_to_cpu(c)),
+				     ml->node, ml->list, ml->flags, ml->type,
+				     ml->convert_type, ml->highest_blocked);
+				__dlm_print_one_lock_resource(res);
+				BUG();
+			}
+
+			if (lock->ml.node != ml->node) {
+				c = lock->ml.cookie;
+				mlog(ML_ERROR, "Mismatched node# in lock "
+				     "cookie %u:%llu, name %.*s, node %u\n",
+				     dlm_get_lock_cookie_node(be64_to_cpu(c)),
+				     dlm_get_lock_cookie_seq(be64_to_cpu(c)),
+				     res->lockname.len, res->lockname.name,
+				     lock->ml.node);
+				c = ml->cookie;
+				mlog(ML_ERROR, "Migrate lock cookie %u:%llu, "
+				     "node %u, list %u, flags 0x%x, type %d, "
+				     "conv %d, highest blocked %d\n",
+				     dlm_get_lock_cookie_node(be64_to_cpu(c)),
+				     dlm_get_lock_cookie_seq(be64_to_cpu(c)),
+				     ml->node, ml->list, ml->flags, ml->type,
+				     ml->convert_type, ml->highest_blocked);
 				__dlm_print_one_lock_resource(res);
 				BUG();
 			}
-			BUG_ON(lock->ml.node != ml->node);
 
 			if (tmpq != queue) {
-				mlog(0, "lock was on %u instead of %u for %.*s\n",
-				     j, ml->list, res->lockname.len, res->lockname.name);
+				c = ml->cookie;
+				mlog(0, "Lock cookie %u:%llu was on list %u "
+				     "instead of list %u for %.*s\n",
+				     dlm_get_lock_cookie_node(be64_to_cpu(c)),
+				     dlm_get_lock_cookie_seq(be64_to_cpu(c)),
+				     j, ml->list, res->lockname.len,
+				     res->lockname.name);
+				__dlm_print_one_lock_resource(res);
 				spin_unlock(&res->spinlock);
 				continue;
 			}
@@ -1839,7 +1895,7 @@ static int dlm_process_recovery_data(struct dlm_ctxt *dlm,
 				 * the lvb. */
 				memcpy(res->lvb, mres->lvb, DLM_LVB_LEN);
 			} else {
-				/* otherwise, the node is sending its 
+				/* otherwise, the node is sending its
 				 * most recent valid lvb info */
 				BUG_ON(ml->type != LKM_EXMODE &&
 				       ml->type != LKM_PRMODE);
@@ -1886,7 +1942,7 @@ skip_lvb:
 		spin_lock(&res->spinlock);
 		list_for_each_entry(lock, queue, list) {
 			if (lock->ml.cookie == ml->cookie) {
-				__be64 c = lock->ml.cookie;
+				c = lock->ml.cookie;
 				mlog(ML_ERROR, "%s:%.*s: %u:%llu: lock already "
 				     "exists on this lockres!\n", dlm->name,
 				     res->lockname.len, res->lockname.name,
@@ -2114,7 +2170,7 @@ static void dlm_revalidate_lvb(struct dlm_ctxt *dlm,
 	assert_spin_locked(&res->spinlock);
 
 	if (res->owner == dlm->node_num)
-		/* if this node owned the lockres, and if the dead node 
+		/* if this node owned the lockres, and if the dead node
 		 * had an EX when he died, blank out the lvb */
 		search_node = dead_node;
 	else {
@@ -2152,7 +2208,7 @@ static void dlm_free_dead_locks(struct dlm_ctxt *dlm,
 
 	/* this node is the lockres master:
 	 * 1) remove any stale locks for the dead node
-	 * 2) if the dead node had an EX when he died, blank out the lvb 
+	 * 2) if the dead node had an EX when he died, blank out the lvb
 	 */
 	assert_spin_locked(&dlm->spinlock);
 	assert_spin_locked(&res->spinlock);
@@ -2193,7 +2249,12 @@ static void dlm_free_dead_locks(struct dlm_ctxt *dlm,
 		mlog(0, "%s:%.*s: freed %u locks for dead node %u, "
 		     "dropping ref from lockres\n", dlm->name,
 		     res->lockname.len, res->lockname.name, freed, dead_node);
-		BUG_ON(!test_bit(dead_node, res->refmap));
+		if(!test_bit(dead_node, res->refmap)) {
+			mlog(ML_ERROR, "%s:%.*s: freed %u locks for dead node %u, "
+			     "but ref was not set\n", dlm->name,
+			     res->lockname.len, res->lockname.name, freed, dead_node);
+			__dlm_print_one_lock_resource(res);
+		}
 		dlm_lockres_clear_refmap_bit(dead_node, res);
 	} else if (test_bit(dead_node, res->refmap)) {
 		mlog(0, "%s:%.*s: dead node %u had a ref, but had "
@@ -2260,7 +2321,7 @@ static void dlm_do_local_recovery_cleanup(struct dlm_ctxt *dlm, u8 dead_node)
 				}
 				spin_unlock(&res->spinlock);
 				continue;
-			}			
+			}
 			spin_lock(&res->spinlock);
 			/* zero the lvb if necessary */
 			dlm_revalidate_lvb(dlm, res, dead_node);
@@ -2411,7 +2472,7 @@ static void dlm_reco_unlock_ast(void *astdata, enum dlm_status st)
  * this function on each node racing to become the recovery
  * master will not stop attempting this until either:
  * a) this node gets the EX (and becomes the recovery master),
- * or b) dlm->reco.new_master gets set to some nodenum 
+ * or b) dlm->reco.new_master gets set to some nodenum
  * != O2NM_INVALID_NODE_NUM (another node will do the reco).
  * so each time a recovery master is needed, the entire cluster
  * will sync at this point.  if the new master dies, that will
@@ -2424,7 +2485,7 @@ static int dlm_pick_recovery_master(struct dlm_ctxt *dlm)
 
 	mlog(0, "starting recovery of %s at %lu, dead=%u, this=%u\n",
 	     dlm->name, jiffies, dlm->reco.dead_node, dlm->node_num);
-again:	
+again:
 	memset(&lksb, 0, sizeof(lksb));
 
 	ret = dlmlock(dlm, LKM_EXMODE, &lksb, LKM_NOQUEUE|LKM_RECOVERY,
@@ -2437,8 +2498,8 @@ again:
 	if (ret == DLM_NORMAL) {
 		mlog(0, "dlm=%s dlmlock says I got it (this=%u)\n",
 		     dlm->name, dlm->node_num);
-		
-		/* got the EX lock.  check to see if another node 
+
+		/* got the EX lock.  check to see if another node
 		 * just became the reco master */
 		if (dlm_reco_master_ready(dlm)) {
 			mlog(0, "%s: got reco EX lock, but %u will "
@@ -2451,12 +2512,12 @@ again:
 			/* see if recovery was already finished elsewhere */
 			spin_lock(&dlm->spinlock);
 			if (dlm->reco.dead_node == O2NM_INVALID_NODE_NUM) {
-				status = -EINVAL;	
+				status = -EINVAL;
 				mlog(0, "%s: got reco EX lock, but "
 				     "node got recovered already\n", dlm->name);
 				if (dlm->reco.new_master != O2NM_INVALID_NODE_NUM) {
 					mlog(ML_ERROR, "%s: new master is %u "
-					     "but no dead node!\n", 
+					     "but no dead node!\n",
 					     dlm->name, dlm->reco.new_master);
 					BUG();
 				}
@@ -2468,7 +2529,7 @@ again:
 		 * set the master and send the messages to begin recovery */
 		if (!status) {
 			mlog(0, "%s: dead=%u, this=%u, sending "
-			     "begin_reco now\n", dlm->name, 
+			     "begin_reco now\n", dlm->name,
 			     dlm->reco.dead_node, dlm->node_num);
 			status = dlm_send_begin_reco_message(dlm,
 				      dlm->reco.dead_node);
@@ -2501,7 +2562,7 @@ again:
 		mlog(0, "dlm=%s dlmlock says another node got it (this=%u)\n",
 		     dlm->name, dlm->node_num);
 		/* another node is master. wait on
-		 * reco.new_master != O2NM_INVALID_NODE_NUM 
+		 * reco.new_master != O2NM_INVALID_NODE_NUM
 		 * for at most one second */
 		wait_event_timeout(dlm->dlm_reco_thread_wq,
 					 dlm_reco_master_ready(dlm),
@@ -2585,11 +2646,17 @@ retry:
 		if (dlm_is_host_down(ret)) {
 			/* node is down.  not involved in recovery
 			 * so just keep going */
-			mlog(0, "%s: node %u was down when sending "
+			mlog(ML_NOTICE, "%s: node %u was down when sending "
 			     "begin reco msg (%d)\n", dlm->name, nodenum, ret);
 			ret = 0;
 		}
-		if (ret == -EAGAIN) {
+
+		/*
+		 * Prior to commit aad1b15310b9bcd59fa81ab8f2b1513b59553ea8,
+		 * dlm_begin_reco_handler() returned EAGAIN and not -EAGAIN.
+		 * We are handling both for compatibility reasons.
+		 */
+		if (ret == -EAGAIN || ret == EAGAIN) {
 			mlog(0, "%s: trying to start recovery of node "
 			     "%u, but node %u is waiting for last recovery "
 			     "to complete, backoff for a bit\n", dlm->name,
@@ -2599,11 +2666,12 @@ retry:
 		}
 		if (ret < 0) {
 			struct dlm_lock_resource *res;
-			/* this is now a serious problem, possibly ENOMEM 
+
+			/* this is now a serious problem, possibly ENOMEM
 			 * in the network stack.  must retry */
 			mlog_errno(ret);
 			mlog(ML_ERROR, "begin reco of dlm %s to node %u "
-			    " returned %d\n", dlm->name, nodenum, ret);
+			     "returned %d\n", dlm->name, nodenum, ret);
 			res = dlm_lookup_lockres(dlm, DLM_RECOVERY_LOCK_NAME,
 						 DLM_RECOVERY_LOCK_NAME_LEN);
 			if (res) {
@@ -2612,7 +2680,7 @@ retry:
 			} else {
 				mlog(ML_ERROR, "recovery lock not found\n");
 			}
-			/* sleep for a bit in hopes that we can avoid 
+			/* sleep for a bit in hopes that we can avoid
 			 * another ENOMEM */
 			msleep(100);
 			goto retry;
@@ -2664,7 +2732,7 @@ int dlm_begin_reco_handler(struct o2net_msg *msg, u32 len, void *data,
 	}
 	if (dlm->reco.dead_node != O2NM_INVALID_NODE_NUM) {
 		mlog(ML_NOTICE, "%s: dead_node previously set to %u, "
-		     "node %u changing it to %u\n", dlm->name, 
+		     "node %u changing it to %u\n", dlm->name,
 		     dlm->reco.dead_node, br->node_idx, br->dead_node);
 	}
 	dlm_set_reco_master(dlm, br->node_idx);
@@ -2728,10 +2796,12 @@ stage2:
 		if (ret >= 0)
 			ret = status;
 		if (ret < 0) {
-			mlog_errno(ret);
+			mlog(ML_ERROR, "Error %d when sending message %u (key "
+			     "0x%x) to node %u\n", ret, DLM_FINALIZE_RECO_MSG,
+			     dlm->key, nodenum);
 			if (dlm_is_host_down(ret)) {
-				/* this has no effect on this recovery 
-				 * session, so set the status to zero to 
+				/* this has no effect on this recovery
+				 * session, so set the status to zero to
 				 * finish out the last recovery */
 				mlog(ML_ERROR, "node %u went down after this "
 				     "node finished recovery.\n", nodenum);
@@ -2768,7 +2838,7 @@ int dlm_finalize_reco_handler(struct o2net_msg *msg, u32 len, void *data,
 	mlog(0, "%s: node %u finalizing recovery stage%d of "
 	     "node %u (%u:%u)\n", dlm->name, fr->node_idx, stage,
 	     fr->dead_node, dlm->reco.dead_node, dlm->reco.new_master);
- 
+
 	spin_lock(&dlm->spinlock);
 
 	if (dlm->reco.new_master != fr->node_idx) {
diff --git a/fs/ocfs2/dlm/dlmthread.c b/fs/ocfs2/dlm/dlmthread.c
index 52ec020ea78b..d4f73ca68fe5 100644
--- a/fs/ocfs2/dlm/dlmthread.c
+++ b/fs/ocfs2/dlm/dlmthread.c
@@ -28,7 +28,6 @@
 #include <linux/module.h>
 #include <linux/fs.h>
 #include <linux/types.h>
-#include <linux/slab.h>
 #include <linux/highmem.h>
 #include <linux/init.h>
 #include <linux/sysctl.h>
@@ -310,6 +309,7 @@ static void dlm_shuffle_lists(struct dlm_ctxt *dlm,
 	 * spinlock, and because we know that it is not migrating/
 	 * recovering/in-progress, it is fine to reserve asts and
 	 * basts right before queueing them all throughout */
+	assert_spin_locked(&dlm->ast_lock);
 	assert_spin_locked(&res->spinlock);
 	BUG_ON((res->state & (DLM_LOCK_RES_MIGRATING|
 			      DLM_LOCK_RES_RECOVERING|
@@ -338,7 +338,7 @@ converting:
 			/* queue the BAST if not already */
 			if (lock->ml.highest_blocked == LKM_IVMODE) {
 				__dlm_lockres_reserve_ast(res);
-				dlm_queue_bast(dlm, lock);
+				__dlm_queue_bast(dlm, lock);
 			}
 			/* update the highest_blocked if needed */
 			if (lock->ml.highest_blocked < target->ml.convert_type)
@@ -356,7 +356,7 @@ converting:
 			can_grant = 0;
 			if (lock->ml.highest_blocked == LKM_IVMODE) {
 				__dlm_lockres_reserve_ast(res);
-				dlm_queue_bast(dlm, lock);
+				__dlm_queue_bast(dlm, lock);
 			}
 			if (lock->ml.highest_blocked < target->ml.convert_type)
 				lock->ml.highest_blocked =
@@ -384,7 +384,7 @@ converting:
 		spin_unlock(&target->spinlock);
 
 		__dlm_lockres_reserve_ast(res);
-		dlm_queue_ast(dlm, target);
+		__dlm_queue_ast(dlm, target);
 		/* go back and check for more */
 		goto converting;
 	}
@@ -403,7 +403,7 @@ blocked:
 			can_grant = 0;
 			if (lock->ml.highest_blocked == LKM_IVMODE) {
 				__dlm_lockres_reserve_ast(res);
-				dlm_queue_bast(dlm, lock);
+				__dlm_queue_bast(dlm, lock);
 			}
 			if (lock->ml.highest_blocked < target->ml.type)
 				lock->ml.highest_blocked = target->ml.type;
@@ -419,7 +419,7 @@ blocked:
 			can_grant = 0;
 			if (lock->ml.highest_blocked == LKM_IVMODE) {
 				__dlm_lockres_reserve_ast(res);
-				dlm_queue_bast(dlm, lock);
+				__dlm_queue_bast(dlm, lock);
 			}
 			if (lock->ml.highest_blocked < target->ml.type)
 				lock->ml.highest_blocked = target->ml.type;
@@ -445,7 +445,7 @@ blocked:
 		spin_unlock(&target->spinlock);
 
 		__dlm_lockres_reserve_ast(res);
-		dlm_queue_ast(dlm, target);
+		__dlm_queue_ast(dlm, target);
 		/* go back and check for more */
 		goto converting;
 	}
@@ -675,6 +675,7 @@ static int dlm_thread(void *data)
 		 	/* lockres can be re-dirtied/re-added to the
 			 * dirty_list in this gap, but that is ok */
 
+			spin_lock(&dlm->ast_lock);
 			spin_lock(&res->spinlock);
 			if (res->owner != dlm->node_num) {
 				__dlm_print_one_lock_resource(res);
@@ -695,6 +696,7 @@ static int dlm_thread(void *data)
 				/* move it to the tail and keep going */
 				res->state &= ~DLM_LOCK_RES_DIRTY;
 				spin_unlock(&res->spinlock);
+				spin_unlock(&dlm->ast_lock);
 				mlog(0, "delaying list shuffling for in-"
 				     "progress lockres %.*s, state=%d\n",
 				     res->lockname.len, res->lockname.name,
@@ -716,6 +718,7 @@ static int dlm_thread(void *data)
 			dlm_shuffle_lists(dlm, res);
 			res->state &= ~DLM_LOCK_RES_DIRTY;
 			spin_unlock(&res->spinlock);
+			spin_unlock(&dlm->ast_lock);
 
 			dlm_lockres_calc_usage(dlm, res);
 
diff --git a/fs/ocfs2/dlm/dlmunlock.c b/fs/ocfs2/dlm/dlmunlock.c
index 00f53b2aea76..817287c6a6db 100644
--- a/fs/ocfs2/dlm/dlmunlock.c
+++ b/fs/ocfs2/dlm/dlmunlock.c
@@ -28,7 +28,6 @@
 #include <linux/module.h>
 #include <linux/fs.h>
 #include <linux/types.h>
-#include <linux/slab.h>
 #include <linux/highmem.h>
 #include <linux/init.h>
 #include <linux/sysctl.h>
@@ -190,8 +189,8 @@ static enum dlm_status dlmunlock_common(struct dlm_ctxt *dlm,
 			actions &= ~(DLM_UNLOCK_REMOVE_LOCK|
 				     DLM_UNLOCK_REGRANT_LOCK|
 				     DLM_UNLOCK_CLEAR_CONVERT_TYPE);
-		} else if (status == DLM_RECOVERING || 
-			   status == DLM_MIGRATING || 
+		} else if (status == DLM_RECOVERING ||
+			   status == DLM_MIGRATING ||
 			   status == DLM_FORWARD) {
 			/* must clear the actions because this unlock
 			 * is about to be retried.  cannot free or do
@@ -355,7 +354,8 @@ static enum dlm_status dlm_send_remote_unlock_request(struct dlm_ctxt *dlm,
 			mlog(0, "master was in-progress.  retry\n");
 		ret = status;
 	} else {
-		mlog_errno(tmpret);
+		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
+		     "node %u\n", tmpret, DLM_UNLOCK_LOCK_MSG, dlm->key, owner);
 		if (dlm_is_host_down(tmpret)) {
 			/* NOTE: this seems strange, but it is what we want.
 			 * when the master goes down during a cancel or
@@ -661,14 +661,14 @@ retry:
 	if (call_ast) {
 		mlog(0, "calling unlockast(%p, %d)\n", data, status);
 		if (is_master) {
-			/* it is possible that there is one last bast 
+			/* it is possible that there is one last bast
 			 * pending.  make sure it is flushed, then
 			 * call the unlockast.
 			 * not an issue if this is a mastered remotely,
 			 * since this lock has been removed from the
 			 * lockres queues and cannot be found. */
 			dlm_kick_thread(dlm, NULL);
-			wait_event(dlm->ast_wq, 
+			wait_event(dlm->ast_wq,
 				   dlm_lock_basts_flushed(dlm, lock));
 		}
 		(*unlockast)(data, status);
diff --git a/fs/ocfs2/dlmfs/Makefile b/fs/ocfs2/dlmfs/Makefile
new file mode 100644
index 000000000000..df69b4856d0d
--- /dev/null
+++ b/fs/ocfs2/dlmfs/Makefile
@@ -0,0 +1,5 @@
+EXTRA_CFLAGS += -Ifs/ocfs2
+
+obj-$(CONFIG_OCFS2_FS) += ocfs2_dlmfs.o
+
+ocfs2_dlmfs-objs := userdlm.o dlmfs.o dlmfsver.o
diff --git a/fs/ocfs2/dlm/dlmfs.c b/fs/ocfs2/dlmfs/dlmfs.c
index 02bf17808bdc..b83d6107a1f5 100644
--- a/fs/ocfs2/dlm/dlmfs.c
+++ b/fs/ocfs2/dlmfs/dlmfs.c
@@ -43,24 +43,17 @@
 #include <linux/init.h>
 #include <linux/string.h>
 #include <linux/backing-dev.h>
+#include <linux/poll.h>
 
 #include <asm/uaccess.h>
 
-
-#include "cluster/nodemanager.h"
-#include "cluster/heartbeat.h"
-#include "cluster/tcp.h"
-
-#include "dlmapi.h"
-
+#include "stackglue.h"
 #include "userdlm.h"
-
 #include "dlmfsver.h"
 
 #define MLOG_MASK_PREFIX ML_DLMFS
 #include "cluster/masklog.h"
 
-#include "ocfs2_lockingver.h"
 
 static const struct super_operations dlmfs_ops;
 static const struct file_operations dlmfs_file_operations;
@@ -71,15 +64,46 @@ static struct kmem_cache *dlmfs_inode_cache;
 
 struct workqueue_struct *user_dlm_worker;
 
+
+
 /*
- * This is the userdlmfs locking protocol version.
+ * These are the ABI capabilities of dlmfs.
+ *
+ * Over time, dlmfs has added some features that were not part of the
+ * initial ABI.  Unfortunately, some of these features are not detectable
+ * via standard usage.  For example, Linux's default poll always returns
+ * POLLIN, so there is no way for a caller of poll(2) to know when dlmfs
+ * added poll support.  Instead, we provide this list of new capabilities.
+ *
+ * Capabilities is a read-only attribute.  We do it as a module parameter
+ * so we can discover it whether dlmfs is built in, loaded, or even not
+ * loaded.
  *
- * See fs/ocfs2/dlmglue.c for more details on locking versions.
+ * The ABI features are local to this machine's dlmfs mount.  This is
+ * distinct from the locking protocol, which is concerned with inter-node
+ * interaction.
+ *
+ * Capabilities:
+ * - bast	: POLLIN against the file descriptor of a held lock
+ *		  signifies a bast fired on the lock.
  */
-static const struct dlm_protocol_version user_locking_protocol = {
-	.pv_major = OCFS2_LOCKING_PROTOCOL_MAJOR,
-	.pv_minor = OCFS2_LOCKING_PROTOCOL_MINOR,
-};
+#define DLMFS_CAPABILITIES "bast stackglue"
+extern int param_set_dlmfs_capabilities(const char *val,
+					struct kernel_param *kp)
+{
+	printk(KERN_ERR "%s: readonly parameter\n", kp->name);
+	return -EINVAL;
+}
+static int param_get_dlmfs_capabilities(char *buffer,
+					struct kernel_param *kp)
+{
+	return strlcpy(buffer, DLMFS_CAPABILITIES,
+		       strlen(DLMFS_CAPABILITIES) + 1);
+}
+module_param_call(capabilities, param_set_dlmfs_capabilities,
+		  param_get_dlmfs_capabilities, NULL, 0444);
+MODULE_PARM_DESC(capabilities, DLMFS_CAPABILITIES);
+
 
 /*
  * decodes a set of open flags into a valid lock level and a set of flags.
@@ -88,20 +112,20 @@ static const struct dlm_protocol_version user_locking_protocol = {
  * O_RDONLY -> PRMODE level
  * O_WRONLY -> EXMODE level
  *
- * O_NONBLOCK -> LKM_NOQUEUE
+ * O_NONBLOCK -> NOQUEUE
  */
 static int dlmfs_decode_open_flags(int open_flags,
 				   int *level,
 				   int *flags)
 {
 	if (open_flags & (O_WRONLY|O_RDWR))
-		*level = LKM_EXMODE;
+		*level = DLM_LOCK_EX;
 	else
-		*level = LKM_PRMODE;
+		*level = DLM_LOCK_PR;
 
 	*flags = 0;
 	if (open_flags & O_NONBLOCK)
-		*flags |= LKM_NOQUEUE;
+		*flags |= DLM_LKF_NOQUEUE;
 
 	return 0;
 }
@@ -142,7 +166,7 @@ static int dlmfs_file_open(struct inode *inode,
 		 * to be able userspace to be able to distinguish a
 		 * valid lock request from one that simply couldn't be
 		 * granted. */
-		if (flags & LKM_NOQUEUE && status == -EAGAIN)
+		if (flags & DLM_LKF_NOQUEUE && status == -EAGAIN)
 			status = -ETXTBSY;
 		kfree(fp);
 		goto bail;
@@ -169,7 +193,7 @@ static int dlmfs_file_release(struct inode *inode,
 	status = 0;
 	if (fp) {
 		level = fp->fp_lock_level;
-		if (level != LKM_IVMODE)
+		if (level != DLM_LOCK_IV)
 			user_dlm_cluster_unlock(&ip->ip_lockres, level);
 
 		kfree(fp);
@@ -179,13 +203,46 @@ static int dlmfs_file_release(struct inode *inode,
 	return 0;
 }
 
+/*
+ * We do ->setattr() just to override size changes.  Our size is the size
+ * of the LVB and nothing else.
+ */
+static int dlmfs_file_setattr(struct dentry *dentry, struct iattr *attr)
+{
+	int error;
+	struct inode *inode = dentry->d_inode;
+
+	attr->ia_valid &= ~ATTR_SIZE;
+	error = inode_change_ok(inode, attr);
+	if (!error)
+		error = inode_setattr(inode, attr);
+
+	return error;
+}
+
+static unsigned int dlmfs_file_poll(struct file *file, poll_table *wait)
+{
+	int event = 0;
+	struct inode *inode = file->f_path.dentry->d_inode;
+	struct dlmfs_inode_private *ip = DLMFS_I(inode);
+
+	poll_wait(file, &ip->ip_lockres.l_event, wait);
+
+	spin_lock(&ip->ip_lockres.l_lock);
+	if (ip->ip_lockres.l_flags & USER_LOCK_BLOCKED)
+		event = POLLIN | POLLRDNORM;
+	spin_unlock(&ip->ip_lockres.l_lock);
+
+	return event;
+}
+
 static ssize_t dlmfs_file_read(struct file *filp,
 			       char __user *buf,
 			       size_t count,
 			       loff_t *ppos)
 {
 	int bytes_left;
-	ssize_t readlen;
+	ssize_t readlen, got;
 	char *lvb_buf;
 	struct inode *inode = filp->f_path.dentry->d_inode;
 
@@ -205,15 +262,19 @@ static ssize_t dlmfs_file_read(struct file *filp,
 	if ((count + *ppos) > i_size_read(inode))
 		readlen = i_size_read(inode) - *ppos;
 	else
-		readlen = count - *ppos;
+		readlen = count;
 
 	lvb_buf = kmalloc(readlen, GFP_NOFS);
 	if (!lvb_buf)
 		return -ENOMEM;
 
-	user_dlm_read_lvb(inode, lvb_buf, readlen);
-	bytes_left = __copy_to_user(buf, lvb_buf, readlen);
-	readlen -= bytes_left;
+	got = user_dlm_read_lvb(inode, lvb_buf, readlen);
+	if (got) {
+		BUG_ON(got != readlen);
+		bytes_left = __copy_to_user(buf, lvb_buf, readlen);
+		readlen -= bytes_left;
+	} else
+		readlen = 0;
 
 	kfree(lvb_buf);
 
@@ -272,7 +333,7 @@ static void dlmfs_init_once(void *foo)
 	struct dlmfs_inode_private *ip =
 		(struct dlmfs_inode_private *) foo;
 
-	ip->ip_dlm = NULL;
+	ip->ip_conn = NULL;
 	ip->ip_parent = NULL;
 
 	inode_init_once(&ip->ip_vfs_inode);
@@ -314,14 +375,14 @@ static void dlmfs_clear_inode(struct inode *inode)
 		goto clear_fields;
 	}
 
-	mlog(0, "we're a directory, ip->ip_dlm = 0x%p\n", ip->ip_dlm);
+	mlog(0, "we're a directory, ip->ip_conn = 0x%p\n", ip->ip_conn);
 	/* we must be a directory. If required, lets unregister the
 	 * dlm context now. */
-	if (ip->ip_dlm)
-		user_dlm_unregister_context(ip->ip_dlm);
+	if (ip->ip_conn)
+		user_dlm_unregister(ip->ip_conn);
 clear_fields:
 	ip->ip_parent = NULL;
-	ip->ip_dlm = NULL;
+	ip->ip_conn = NULL;
 }
 
 static struct backing_dev_info dlmfs_backing_dev_info = {
@@ -371,7 +432,7 @@ static struct inode *dlmfs_get_inode(struct inode *parent,
 	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
 
 	ip = DLMFS_I(inode);
-	ip->ip_dlm = DLMFS_I(parent)->ip_dlm;
+	ip->ip_conn = DLMFS_I(parent)->ip_conn;
 
 	switch (mode & S_IFMT) {
 	default:
@@ -425,13 +486,12 @@ static int dlmfs_mkdir(struct inode * dir,
 	struct inode *inode = NULL;
 	struct qstr *domain = &dentry->d_name;
 	struct dlmfs_inode_private *ip;
-	struct dlm_ctxt *dlm;
-	struct dlm_protocol_version proto = user_locking_protocol;
+	struct ocfs2_cluster_connection *conn;
 
 	mlog(0, "mkdir %.*s\n", domain->len, domain->name);
 
 	/* verify that we have a proper domain */
-	if (domain->len >= O2NM_MAX_NAME_LEN) {
+	if (domain->len >= GROUP_NAME_MAX) {
 		status = -EINVAL;
 		mlog(ML_ERROR, "invalid domain name for directory.\n");
 		goto bail;
@@ -446,14 +506,14 @@ static int dlmfs_mkdir(struct inode * dir,
 
 	ip = DLMFS_I(inode);
 
-	dlm = user_dlm_register_context(domain, &proto);
-	if (IS_ERR(dlm)) {
-		status = PTR_ERR(dlm);
+	conn = user_dlm_register(domain);
+	if (IS_ERR(conn)) {
+		status = PTR_ERR(conn);
 		mlog(ML_ERROR, "Error %d could not register domain \"%.*s\"\n",
 		     status, domain->len, domain->name);
 		goto bail;
 	}
-	ip->ip_dlm = dlm;
+	ip->ip_conn = conn;
 
 	inc_nlink(dir);
 	d_instantiate(dentry, inode);
@@ -549,6 +609,7 @@ static int dlmfs_fill_super(struct super_block * sb,
 static const struct file_operations dlmfs_file_operations = {
 	.open		= dlmfs_file_open,
 	.release	= dlmfs_file_release,
+	.poll		= dlmfs_file_poll,
 	.read		= dlmfs_file_read,
 	.write		= dlmfs_file_write,
 };
@@ -576,6 +637,7 @@ static const struct super_operations dlmfs_ops = {
 
 static const struct inode_operations dlmfs_file_inode_operations = {
 	.getattr	= simple_getattr,
+	.setattr	= dlmfs_file_setattr,
 };
 
 static int dlmfs_get_sb(struct file_system_type *fs_type,
@@ -620,6 +682,7 @@ static int __init init_dlmfs_fs(void)
 	}
 	cleanup_worker = 1;
 
+	user_dlm_set_locking_protocol();
 	status = register_filesystem(&dlmfs_fs_type);
 bail:
 	if (status) {
diff --git a/fs/ocfs2/dlm/dlmfsver.c b/fs/ocfs2/dlmfs/dlmfsver.c
index a733b3321f83..a733b3321f83 100644
--- a/fs/ocfs2/dlm/dlmfsver.c
+++ b/fs/ocfs2/dlmfs/dlmfsver.c
diff --git a/fs/ocfs2/dlm/dlmfsver.h b/fs/ocfs2/dlmfs/dlmfsver.h
index f35eadbed25c..f35eadbed25c 100644
--- a/fs/ocfs2/dlm/dlmfsver.h
+++ b/fs/ocfs2/dlmfs/dlmfsver.h
diff --git a/fs/ocfs2/dlm/userdlm.c b/fs/ocfs2/dlmfs/userdlm.c
index 4cb1d3dae250..0499e3fb7bdb 100644
--- a/fs/ocfs2/dlm/userdlm.c
+++ b/fs/ocfs2/dlmfs/userdlm.c
@@ -34,18 +34,19 @@
 #include <linux/types.h>
 #include <linux/crc32.h>
 
-
-#include "cluster/nodemanager.h"
-#include "cluster/heartbeat.h"
-#include "cluster/tcp.h"
-
-#include "dlmapi.h"
-
+#include "ocfs2_lockingver.h"
+#include "stackglue.h"
 #include "userdlm.h"
 
 #define MLOG_MASK_PREFIX ML_DLMFS
 #include "cluster/masklog.h"
 
+
+static inline struct user_lock_res *user_lksb_to_lock_res(struct ocfs2_dlm_lksb *lksb)
+{
+	return container_of(lksb, struct user_lock_res, l_lksb);
+}
+
 static inline int user_check_wait_flag(struct user_lock_res *lockres,
 				       int flag)
 {
@@ -73,15 +74,15 @@ static inline void user_wait_on_blocked_lock(struct user_lock_res *lockres)
 }
 
 /* I heart container_of... */
-static inline struct dlm_ctxt *
-dlm_ctxt_from_user_lockres(struct user_lock_res *lockres)
+static inline struct ocfs2_cluster_connection *
+cluster_connection_from_user_lockres(struct user_lock_res *lockres)
 {
 	struct dlmfs_inode_private *ip;
 
 	ip = container_of(lockres,
 			  struct dlmfs_inode_private,
 			  ip_lockres);
-	return ip->ip_dlm;
+	return ip->ip_conn;
 }
 
 static struct inode *
@@ -103,9 +104,9 @@ static inline void user_recover_from_dlm_error(struct user_lock_res *lockres)
 }
 
 #define user_log_dlm_error(_func, _stat, _lockres) do {			\
-	mlog(ML_ERROR, "Dlm error \"%s\" while calling %s on "		\
-		"resource %.*s: %s\n", dlm_errname(_stat), _func,	\
-		_lockres->l_namelen, _lockres->l_name, dlm_errmsg(_stat)); \
+	mlog(ML_ERROR, "Dlm error %d while calling %s on "		\
+		"resource %.*s\n", _stat, _func,			\
+		_lockres->l_namelen, _lockres->l_name); 		\
 } while (0)
 
 /* WARNING: This function lives in a world where the only three lock
@@ -113,34 +114,35 @@ static inline void user_recover_from_dlm_error(struct user_lock_res *lockres)
  * lock types are added. */
 static inline int user_highest_compat_lock_level(int level)
 {
-	int new_level = LKM_EXMODE;
+	int new_level = DLM_LOCK_EX;
 
-	if (level == LKM_EXMODE)
-		new_level = LKM_NLMODE;
-	else if (level == LKM_PRMODE)
-		new_level = LKM_PRMODE;
+	if (level == DLM_LOCK_EX)
+		new_level = DLM_LOCK_NL;
+	else if (level == DLM_LOCK_PR)
+		new_level = DLM_LOCK_PR;
 	return new_level;
 }
 
-static void user_ast(void *opaque)
+static void user_ast(struct ocfs2_dlm_lksb *lksb)
 {
-	struct user_lock_res *lockres = opaque;
-	struct dlm_lockstatus *lksb;
+	struct user_lock_res *lockres = user_lksb_to_lock_res(lksb);
+	int status;
 
-	mlog(0, "AST fired for lockres %.*s\n", lockres->l_namelen,
-	     lockres->l_name);
+	mlog(ML_BASTS, "AST fired for lockres %.*s, level %d => %d\n",
+	     lockres->l_namelen, lockres->l_name, lockres->l_level,
+	     lockres->l_requested);
 
 	spin_lock(&lockres->l_lock);
 
-	lksb = &(lockres->l_lksb);
-	if (lksb->status != DLM_NORMAL) {
+	status = ocfs2_dlm_lock_status(&lockres->l_lksb);
+	if (status) {
 		mlog(ML_ERROR, "lksb status value of %u on lockres %.*s\n",
-		     lksb->status, lockres->l_namelen, lockres->l_name);
+		     status, lockres->l_namelen, lockres->l_name);
 		spin_unlock(&lockres->l_lock);
 		return;
 	}
 
-	mlog_bug_on_msg(lockres->l_requested == LKM_IVMODE,
+	mlog_bug_on_msg(lockres->l_requested == DLM_LOCK_IV,
 			"Lockres %.*s, requested ivmode. flags 0x%x\n",
 			lockres->l_namelen, lockres->l_name, lockres->l_flags);
 
@@ -148,13 +150,13 @@ static void user_ast(void *opaque)
 	if (lockres->l_requested < lockres->l_level) {
 		if (lockres->l_requested <=
 		    user_highest_compat_lock_level(lockres->l_blocking)) {
-			lockres->l_blocking = LKM_NLMODE;
+			lockres->l_blocking = DLM_LOCK_NL;
 			lockres->l_flags &= ~USER_LOCK_BLOCKED;
 		}
 	}
 
 	lockres->l_level = lockres->l_requested;
-	lockres->l_requested = LKM_IVMODE;
+	lockres->l_requested = DLM_LOCK_IV;
 	lockres->l_flags |= USER_LOCK_ATTACHED;
 	lockres->l_flags &= ~USER_LOCK_BUSY;
 
@@ -193,11 +195,11 @@ static void __user_dlm_cond_queue_lockres(struct user_lock_res *lockres)
 		return;
 
 	switch (lockres->l_blocking) {
-	case LKM_EXMODE:
+	case DLM_LOCK_EX:
 		if (!lockres->l_ex_holders && !lockres->l_ro_holders)
 			queue = 1;
 		break;
-	case LKM_PRMODE:
+	case DLM_LOCK_PR:
 		if (!lockres->l_ex_holders)
 			queue = 1;
 		break;
@@ -209,12 +211,12 @@ static void __user_dlm_cond_queue_lockres(struct user_lock_res *lockres)
 		__user_dlm_queue_lockres(lockres);
 }
 
-static void user_bast(void *opaque, int level)
+static void user_bast(struct ocfs2_dlm_lksb *lksb, int level)
 {
-	struct user_lock_res *lockres = opaque;
+	struct user_lock_res *lockres = user_lksb_to_lock_res(lksb);
 
-	mlog(0, "Blocking AST fired for lockres %.*s. Blocking level %d\n",
-	     lockres->l_namelen, lockres->l_name, level);
+	mlog(ML_BASTS, "BAST fired for lockres %.*s, blocking %d, level %d\n",
+	     lockres->l_namelen, lockres->l_name, level, lockres->l_level);
 
 	spin_lock(&lockres->l_lock);
 	lockres->l_flags |= USER_LOCK_BLOCKED;
@@ -227,15 +229,15 @@ static void user_bast(void *opaque, int level)
 	wake_up(&lockres->l_event);
 }
 
-static void user_unlock_ast(void *opaque, enum dlm_status status)
+static void user_unlock_ast(struct ocfs2_dlm_lksb *lksb, int status)
 {
-	struct user_lock_res *lockres = opaque;
+	struct user_lock_res *lockres = user_lksb_to_lock_res(lksb);
 
-	mlog(0, "UNLOCK AST called on lock %.*s\n", lockres->l_namelen,
-	     lockres->l_name);
+	mlog(ML_BASTS, "UNLOCK AST fired for lockres %.*s, flags 0x%x\n",
+	     lockres->l_namelen, lockres->l_name, lockres->l_flags);
 
-	if (status != DLM_NORMAL && status != DLM_CANCELGRANT)
-		mlog(ML_ERROR, "Dlm returns status %d\n", status);
+	if (status)
+		mlog(ML_ERROR, "dlm returns status %d\n", status);
 
 	spin_lock(&lockres->l_lock);
 	/* The teardown flag gets set early during the unlock process,
@@ -243,7 +245,7 @@ static void user_unlock_ast(void *opaque, enum dlm_status status)
 	 * for a concurrent cancel. */
 	if (lockres->l_flags & USER_LOCK_IN_TEARDOWN
 	    && !(lockres->l_flags & USER_LOCK_IN_CANCEL)) {
-		lockres->l_level = LKM_IVMODE;
+		lockres->l_level = DLM_LOCK_IV;
 	} else if (status == DLM_CANCELGRANT) {
 		/* We tried to cancel a convert request, but it was
 		 * already granted. Don't clear the busy flag - the
@@ -254,7 +256,7 @@ static void user_unlock_ast(void *opaque, enum dlm_status status)
 	} else {
 		BUG_ON(!(lockres->l_flags & USER_LOCK_IN_CANCEL));
 		/* Cancel succeeded, we want to re-queue */
-		lockres->l_requested = LKM_IVMODE; /* cancel an
+		lockres->l_requested = DLM_LOCK_IV; /* cancel an
 						    * upconvert
 						    * request. */
 		lockres->l_flags &= ~USER_LOCK_IN_CANCEL;
@@ -271,6 +273,21 @@ out_noclear:
 	wake_up(&lockres->l_event);
 }
 
+/*
+ * This is the userdlmfs locking protocol version.
+ *
+ * See fs/ocfs2/dlmglue.c for more details on locking versions.
+ */
+static struct ocfs2_locking_protocol user_dlm_lproto = {
+	.lp_max_version = {
+		.pv_major = OCFS2_LOCKING_PROTOCOL_MAJOR,
+		.pv_minor = OCFS2_LOCKING_PROTOCOL_MINOR,
+	},
+	.lp_lock_ast		= user_ast,
+	.lp_blocking_ast	= user_bast,
+	.lp_unlock_ast		= user_unlock_ast,
+};
+
 static inline void user_dlm_drop_inode_ref(struct user_lock_res *lockres)
 {
 	struct inode *inode;
@@ -283,10 +300,10 @@ static void user_dlm_unblock_lock(struct work_struct *work)
 	int new_level, status;
 	struct user_lock_res *lockres =
 		container_of(work, struct user_lock_res, l_work);
-	struct dlm_ctxt *dlm = dlm_ctxt_from_user_lockres(lockres);
+	struct ocfs2_cluster_connection *conn =
+		cluster_connection_from_user_lockres(lockres);
 
-	mlog(0, "processing lockres %.*s\n", lockres->l_namelen,
-	     lockres->l_name);
+	mlog(0, "lockres %.*s\n", lockres->l_namelen, lockres->l_name);
 
 	spin_lock(&lockres->l_lock);
 
@@ -304,17 +321,23 @@ static void user_dlm_unblock_lock(struct work_struct *work)
 	 * flag, and finally we might get another bast which re-queues
 	 * us before our ast for the downconvert is called. */
 	if (!(lockres->l_flags & USER_LOCK_BLOCKED)) {
+		mlog(ML_BASTS, "lockres %.*s USER_LOCK_BLOCKED\n",
+		     lockres->l_namelen, lockres->l_name);
 		spin_unlock(&lockres->l_lock);
 		goto drop_ref;
 	}
 
 	if (lockres->l_flags & USER_LOCK_IN_TEARDOWN) {
+		mlog(ML_BASTS, "lockres %.*s USER_LOCK_IN_TEARDOWN\n",
+		     lockres->l_namelen, lockres->l_name);
 		spin_unlock(&lockres->l_lock);
 		goto drop_ref;
 	}
 
 	if (lockres->l_flags & USER_LOCK_BUSY) {
 		if (lockres->l_flags & USER_LOCK_IN_CANCEL) {
+			mlog(ML_BASTS, "lockres %.*s USER_LOCK_IN_CANCEL\n",
+			     lockres->l_namelen, lockres->l_name);
 			spin_unlock(&lockres->l_lock);
 			goto drop_ref;
 		}
@@ -322,32 +345,31 @@ static void user_dlm_unblock_lock(struct work_struct *work)
 		lockres->l_flags |= USER_LOCK_IN_CANCEL;
 		spin_unlock(&lockres->l_lock);
 
-		status = dlmunlock(dlm,
-				   &lockres->l_lksb,
-				   LKM_CANCEL,
-				   user_unlock_ast,
-				   lockres);
-		if (status != DLM_NORMAL)
-			user_log_dlm_error("dlmunlock", status, lockres);
+		status = ocfs2_dlm_unlock(conn, &lockres->l_lksb,
+					  DLM_LKF_CANCEL);
+		if (status)
+			user_log_dlm_error("ocfs2_dlm_unlock", status, lockres);
 		goto drop_ref;
 	}
 
 	/* If there are still incompat holders, we can exit safely
 	 * without worrying about re-queueing this lock as that will
 	 * happen on the last call to user_cluster_unlock. */
-	if ((lockres->l_blocking == LKM_EXMODE)
+	if ((lockres->l_blocking == DLM_LOCK_EX)
 	    && (lockres->l_ex_holders || lockres->l_ro_holders)) {
 		spin_unlock(&lockres->l_lock);
-		mlog(0, "can't downconvert for ex: ro = %u, ex = %u\n",
-			lockres->l_ro_holders, lockres->l_ex_holders);
+		mlog(ML_BASTS, "lockres %.*s, EX/PR Holders %u,%u\n",
+		     lockres->l_namelen, lockres->l_name,
+		     lockres->l_ex_holders, lockres->l_ro_holders);
 		goto drop_ref;
 	}
 
-	if ((lockres->l_blocking == LKM_PRMODE)
+	if ((lockres->l_blocking == DLM_LOCK_PR)
 	    && lockres->l_ex_holders) {
 		spin_unlock(&lockres->l_lock);
-		mlog(0, "can't downconvert for pr: ex = %u\n",
-			lockres->l_ex_holders);
+		mlog(ML_BASTS, "lockres %.*s, EX Holders %u\n",
+		     lockres->l_namelen, lockres->l_name,
+		     lockres->l_ex_holders);
 		goto drop_ref;
 	}
 
@@ -355,22 +377,17 @@ static void user_dlm_unblock_lock(struct work_struct *work)
 	new_level = user_highest_compat_lock_level(lockres->l_blocking);
 	lockres->l_requested = new_level;
 	lockres->l_flags |= USER_LOCK_BUSY;
-	mlog(0, "Downconvert lock from %d to %d\n",
-		lockres->l_level, new_level);
+	mlog(ML_BASTS, "lockres %.*s, downconvert %d => %d\n",
+	     lockres->l_namelen, lockres->l_name, lockres->l_level, new_level);
 	spin_unlock(&lockres->l_lock);
 
 	/* need lock downconvert request now... */
-	status = dlmlock(dlm,
-			 new_level,
-			 &lockres->l_lksb,
-			 LKM_CONVERT|LKM_VALBLK,
-			 lockres->l_name,
-			 lockres->l_namelen,
-			 user_ast,
-			 lockres,
-			 user_bast);
-	if (status != DLM_NORMAL) {
-		user_log_dlm_error("dlmlock", status, lockres);
+	status = ocfs2_dlm_lock(conn, new_level, &lockres->l_lksb,
+				DLM_LKF_CONVERT|DLM_LKF_VALBLK,
+				lockres->l_name,
+				lockres->l_namelen);
+	if (status) {
+		user_log_dlm_error("ocfs2_dlm_lock", status, lockres);
 		user_recover_from_dlm_error(lockres);
 	}
 
@@ -382,10 +399,10 @@ static inline void user_dlm_inc_holders(struct user_lock_res *lockres,
 					int level)
 {
 	switch(level) {
-	case LKM_EXMODE:
+	case DLM_LOCK_EX:
 		lockres->l_ex_holders++;
 		break;
-	case LKM_PRMODE:
+	case DLM_LOCK_PR:
 		lockres->l_ro_holders++;
 		break;
 	default:
@@ -410,20 +427,19 @@ int user_dlm_cluster_lock(struct user_lock_res *lockres,
 			  int lkm_flags)
 {
 	int status, local_flags;
-	struct dlm_ctxt *dlm = dlm_ctxt_from_user_lockres(lockres);
+	struct ocfs2_cluster_connection *conn =
+		cluster_connection_from_user_lockres(lockres);
 
-	if (level != LKM_EXMODE &&
-	    level != LKM_PRMODE) {
+	if (level != DLM_LOCK_EX &&
+	    level != DLM_LOCK_PR) {
 		mlog(ML_ERROR, "lockres %.*s: invalid request!\n",
 		     lockres->l_namelen, lockres->l_name);
 		status = -EINVAL;
 		goto bail;
 	}
 
-	mlog(0, "lockres %.*s: asking for %s lock, passed flags = 0x%x\n",
-	     lockres->l_namelen, lockres->l_name,
-	     (level == LKM_EXMODE) ? "LKM_EXMODE" : "LKM_PRMODE",
-	     lkm_flags);
+	mlog(ML_BASTS, "lockres %.*s, level %d, flags = 0x%x\n",
+	     lockres->l_namelen, lockres->l_name, level, lkm_flags);
 
 again:
 	if (signal_pending(current)) {
@@ -457,35 +473,26 @@ again:
 	}
 
 	if (level > lockres->l_level) {
-		local_flags = lkm_flags | LKM_VALBLK;
-		if (lockres->l_level != LKM_IVMODE)
-			local_flags |= LKM_CONVERT;
+		local_flags = lkm_flags | DLM_LKF_VALBLK;
+		if (lockres->l_level != DLM_LOCK_IV)
+			local_flags |= DLM_LKF_CONVERT;
 
 		lockres->l_requested = level;
 		lockres->l_flags |= USER_LOCK_BUSY;
 		spin_unlock(&lockres->l_lock);
 
-		BUG_ON(level == LKM_IVMODE);
-		BUG_ON(level == LKM_NLMODE);
+		BUG_ON(level == DLM_LOCK_IV);
+		BUG_ON(level == DLM_LOCK_NL);
 
 		/* call dlm_lock to upgrade lock now */
-		status = dlmlock(dlm,
-				 level,
-				 &lockres->l_lksb,
-				 local_flags,
-				 lockres->l_name,
-				 lockres->l_namelen,
-				 user_ast,
-				 lockres,
-				 user_bast);
-		if (status != DLM_NORMAL) {
-			if ((lkm_flags & LKM_NOQUEUE) &&
-			    (status == DLM_NOTQUEUED))
-				status = -EAGAIN;
-			else {
-				user_log_dlm_error("dlmlock", status, lockres);
-				status = -EINVAL;
-			}
+		status = ocfs2_dlm_lock(conn, level, &lockres->l_lksb,
+					local_flags, lockres->l_name,
+					lockres->l_namelen);
+		if (status) {
+			if ((lkm_flags & DLM_LKF_NOQUEUE) &&
+			    (status != -EAGAIN))
+				user_log_dlm_error("ocfs2_dlm_lock",
+						   status, lockres);
 			user_recover_from_dlm_error(lockres);
 			goto bail;
 		}
@@ -506,11 +513,11 @@ static inline void user_dlm_dec_holders(struct user_lock_res *lockres,
 					int level)
 {
 	switch(level) {
-	case LKM_EXMODE:
+	case DLM_LOCK_EX:
 		BUG_ON(!lockres->l_ex_holders);
 		lockres->l_ex_holders--;
 		break;
-	case LKM_PRMODE:
+	case DLM_LOCK_PR:
 		BUG_ON(!lockres->l_ro_holders);
 		lockres->l_ro_holders--;
 		break;
@@ -522,8 +529,8 @@ static inline void user_dlm_dec_holders(struct user_lock_res *lockres,
 void user_dlm_cluster_unlock(struct user_lock_res *lockres,
 			     int level)
 {
-	if (level != LKM_EXMODE &&
-	    level != LKM_PRMODE) {
+	if (level != DLM_LOCK_EX &&
+	    level != DLM_LOCK_PR) {
 		mlog(ML_ERROR, "lockres %.*s: invalid request!\n",
 		     lockres->l_namelen, lockres->l_name);
 		return;
@@ -540,33 +547,40 @@ void user_dlm_write_lvb(struct inode *inode,
 			unsigned int len)
 {
 	struct user_lock_res *lockres = &DLMFS_I(inode)->ip_lockres;
-	char *lvb = lockres->l_lksb.lvb;
+	char *lvb;
 
 	BUG_ON(len > DLM_LVB_LEN);
 
 	spin_lock(&lockres->l_lock);
 
-	BUG_ON(lockres->l_level < LKM_EXMODE);
+	BUG_ON(lockres->l_level < DLM_LOCK_EX);
+	lvb = ocfs2_dlm_lvb(&lockres->l_lksb);
 	memcpy(lvb, val, len);
 
 	spin_unlock(&lockres->l_lock);
 }
 
-void user_dlm_read_lvb(struct inode *inode,
-		       char *val,
-		       unsigned int len)
+ssize_t user_dlm_read_lvb(struct inode *inode,
+			  char *val,
+			  unsigned int len)
 {
 	struct user_lock_res *lockres = &DLMFS_I(inode)->ip_lockres;
-	char *lvb = lockres->l_lksb.lvb;
+	char *lvb;
+	ssize_t ret = len;
 
 	BUG_ON(len > DLM_LVB_LEN);
 
 	spin_lock(&lockres->l_lock);
 
-	BUG_ON(lockres->l_level < LKM_PRMODE);
-	memcpy(val, lvb, len);
+	BUG_ON(lockres->l_level < DLM_LOCK_PR);
+	if (ocfs2_dlm_lvb_valid(&lockres->l_lksb)) {
+		lvb = ocfs2_dlm_lvb(&lockres->l_lksb);
+		memcpy(val, lvb, len);
+	} else
+		ret = 0;
 
 	spin_unlock(&lockres->l_lock);
+	return ret;
 }
 
 void user_dlm_lock_res_init(struct user_lock_res *lockres,
@@ -576,9 +590,9 @@ void user_dlm_lock_res_init(struct user_lock_res *lockres,
 
 	spin_lock_init(&lockres->l_lock);
 	init_waitqueue_head(&lockres->l_event);
-	lockres->l_level = LKM_IVMODE;
-	lockres->l_requested = LKM_IVMODE;
-	lockres->l_blocking = LKM_IVMODE;
+	lockres->l_level = DLM_LOCK_IV;
+	lockres->l_requested = DLM_LOCK_IV;
+	lockres->l_blocking = DLM_LOCK_IV;
 
 	/* should have been checked before getting here. */
 	BUG_ON(dentry->d_name.len >= USER_DLM_LOCK_ID_MAX_LEN);
@@ -592,9 +606,10 @@ void user_dlm_lock_res_init(struct user_lock_res *lockres,
 int user_dlm_destroy_lock(struct user_lock_res *lockres)
 {
 	int status = -EBUSY;
-	struct dlm_ctxt *dlm = dlm_ctxt_from_user_lockres(lockres);
+	struct ocfs2_cluster_connection *conn =
+		cluster_connection_from_user_lockres(lockres);
 
-	mlog(0, "asked to destroy %.*s\n", lockres->l_namelen, lockres->l_name);
+	mlog(ML_BASTS, "lockres %.*s\n", lockres->l_namelen, lockres->l_name);
 
 	spin_lock(&lockres->l_lock);
 	if (lockres->l_flags & USER_LOCK_IN_TEARDOWN) {
@@ -627,14 +642,9 @@ int user_dlm_destroy_lock(struct user_lock_res *lockres)
 	lockres->l_flags |= USER_LOCK_BUSY;
 	spin_unlock(&lockres->l_lock);
 
-	status = dlmunlock(dlm,
-			   &lockres->l_lksb,
-			   LKM_VALBLK,
-			   user_unlock_ast,
-			   lockres);
-	if (status != DLM_NORMAL) {
-		user_log_dlm_error("dlmunlock", status, lockres);
-		status = -EINVAL;
+	status = ocfs2_dlm_unlock(conn, &lockres->l_lksb, DLM_LKF_VALBLK);
+	if (status) {
+		user_log_dlm_error("ocfs2_dlm_unlock", status, lockres);
 		goto bail;
 	}
 
@@ -645,32 +655,34 @@ bail:
 	return status;
 }
 
-struct dlm_ctxt *user_dlm_register_context(struct qstr *name,
-					   struct dlm_protocol_version *proto)
+static void user_dlm_recovery_handler_noop(int node_num,
+					   void *recovery_data)
 {
-	struct dlm_ctxt *dlm;
-	u32 dlm_key;
-	char *domain;
-
-	domain = kmalloc(name->len + 1, GFP_NOFS);
-	if (!domain) {
-		mlog_errno(-ENOMEM);
-		return ERR_PTR(-ENOMEM);
-	}
+	/* We ignore recovery events */
+	return;
+}
 
-	dlm_key = crc32_le(0, name->name, name->len);
+void user_dlm_set_locking_protocol(void)
+{
+	ocfs2_stack_glue_set_max_proto_version(&user_dlm_lproto.lp_max_version);
+}
 
-	snprintf(domain, name->len + 1, "%.*s", name->len, name->name);
+struct ocfs2_cluster_connection *user_dlm_register(struct qstr *name)
+{
+	int rc;
+	struct ocfs2_cluster_connection *conn;
 
-	dlm = dlm_register_domain(domain, dlm_key, proto);
-	if (IS_ERR(dlm))
-		mlog_errno(PTR_ERR(dlm));
+	rc = ocfs2_cluster_connect_agnostic(name->name, name->len,
+					    &user_dlm_lproto,
+					    user_dlm_recovery_handler_noop,
+					    NULL, &conn);
+	if (rc)
+		mlog_errno(rc);
 
-	kfree(domain);
-	return dlm;
+	return rc ? ERR_PTR(rc) : conn;
 }
 
-void user_dlm_unregister_context(struct dlm_ctxt *dlm)
+void user_dlm_unregister(struct ocfs2_cluster_connection *conn)
 {
-	dlm_unregister_domain(dlm);
+	ocfs2_cluster_disconnect(conn, 0);
 }
diff --git a/fs/ocfs2/dlm/userdlm.h b/fs/ocfs2/dlmfs/userdlm.h
index 0c3cc03c61fa..3b42d79531d7 100644
--- a/fs/ocfs2/dlm/userdlm.h
+++ b/fs/ocfs2/dlmfs/userdlm.h
@@ -57,7 +57,7 @@ struct user_lock_res {
 	int                      l_level;
 	unsigned int             l_ro_holders;
 	unsigned int             l_ex_holders;
-	struct dlm_lockstatus    l_lksb;
+	struct ocfs2_dlm_lksb    l_lksb;
 
 	int                      l_requested;
 	int                      l_blocking;
@@ -80,15 +80,15 @@ void user_dlm_cluster_unlock(struct user_lock_res *lockres,
 void user_dlm_write_lvb(struct inode *inode,
 			const char *val,
 			unsigned int len);
-void user_dlm_read_lvb(struct inode *inode,
-		       char *val,
-		       unsigned int len);
-struct dlm_ctxt *user_dlm_register_context(struct qstr *name,
-					   struct dlm_protocol_version *proto);
-void user_dlm_unregister_context(struct dlm_ctxt *dlm);
+ssize_t user_dlm_read_lvb(struct inode *inode,
+			  char *val,
+			  unsigned int len);
+struct ocfs2_cluster_connection *user_dlm_register(struct qstr *name);
+void user_dlm_unregister(struct ocfs2_cluster_connection *conn);
+void user_dlm_set_locking_protocol(void);
 
 struct dlmfs_inode_private {
-	struct dlm_ctxt             *ip_dlm;
+	struct ocfs2_cluster_connection	*ip_conn;
 
 	struct user_lock_res ip_lockres; /* unused for directories. */
 	struct inode         *ip_parent;
diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c
index c5e4a49e3a12..39eb16ac5f98 100644
--- a/fs/ocfs2/dlmglue.c
+++ b/fs/ocfs2/dlmglue.c
@@ -297,6 +297,11 @@ static inline int ocfs2_is_inode_lock(struct ocfs2_lock_res *lockres)
 		lockres->l_type == OCFS2_LOCK_TYPE_OPEN;
 }
 
+static inline struct ocfs2_lock_res *ocfs2_lksb_to_lock_res(struct ocfs2_dlm_lksb *lksb)
+{
+	return container_of(lksb, struct ocfs2_lock_res, l_lksb);
+}
+
 static inline struct inode *ocfs2_lock_res_inode(struct ocfs2_lock_res *lockres)
 {
 	BUG_ON(!ocfs2_is_inode_lock(lockres));
@@ -875,6 +880,14 @@ static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lo
 		lockres_or_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH);
 
 	lockres->l_level = lockres->l_requested;
+
+	/*
+	 * We set the OCFS2_LOCK_UPCONVERT_FINISHING flag before clearing
+	 * the OCFS2_LOCK_BUSY flag to prevent the dc thread from
+	 * downconverting the lock before the upconvert has fully completed.
+	 */
+	lockres_or_flags(lockres, OCFS2_LOCK_UPCONVERT_FINISHING);
+
 	lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
 
 	mlog_exit_void();
@@ -907,8 +920,6 @@ static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres,
 
 	assert_spin_locked(&lockres->l_lock);
 
-	lockres_or_flags(lockres, OCFS2_LOCK_BLOCKED);
-
 	if (level > lockres->l_blocking) {
 		/* only schedule a downconvert if we haven't already scheduled
 		 * one that goes low enough to satisfy the level we're
@@ -921,6 +932,13 @@ static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres,
 		lockres->l_blocking = level;
 	}
 
+	mlog(ML_BASTS, "lockres %s, block %d, level %d, l_block %d, dwn %d\n",
+	     lockres->l_name, level, lockres->l_level, lockres->l_blocking,
+	     needs_downconvert);
+
+	if (needs_downconvert)
+		lockres_or_flags(lockres, OCFS2_LOCK_BLOCKED);
+
 	mlog_exit(needs_downconvert);
 	return needs_downconvert;
 }
@@ -1031,18 +1049,17 @@ static unsigned int lockres_set_pending(struct ocfs2_lock_res *lockres)
 	return lockres->l_pending_gen;
 }
 
-
-static void ocfs2_blocking_ast(void *opaque, int level)
+static void ocfs2_blocking_ast(struct ocfs2_dlm_lksb *lksb, int level)
 {
-	struct ocfs2_lock_res *lockres = opaque;
+	struct ocfs2_lock_res *lockres = ocfs2_lksb_to_lock_res(lksb);
 	struct ocfs2_super *osb = ocfs2_get_lockres_osb(lockres);
 	int needs_downconvert;
 	unsigned long flags;
 
 	BUG_ON(level <= DLM_LOCK_NL);
 
-	mlog(0, "BAST fired for lockres %s, blocking %d, level %d type %s\n",
-	     lockres->l_name, level, lockres->l_level,
+	mlog(ML_BASTS, "BAST fired for lockres %s, blocking %d, level %d, "
+	     "type %s\n", lockres->l_name, level, lockres->l_level,
 	     ocfs2_lock_type_string(lockres->l_type));
 
 	/*
@@ -1063,9 +1080,9 @@ static void ocfs2_blocking_ast(void *opaque, int level)
 	ocfs2_wake_downconvert_thread(osb);
 }
 
-static void ocfs2_locking_ast(void *opaque)
+static void ocfs2_locking_ast(struct ocfs2_dlm_lksb *lksb)
 {
-	struct ocfs2_lock_res *lockres = opaque;
+	struct ocfs2_lock_res *lockres = ocfs2_lksb_to_lock_res(lksb);
 	struct ocfs2_super *osb = ocfs2_get_lockres_osb(lockres);
 	unsigned long flags;
 	int status;
@@ -1086,6 +1103,10 @@ static void ocfs2_locking_ast(void *opaque)
 		return;
 	}
 
+	mlog(ML_BASTS, "AST fired for lockres %s, action %d, unlock %d, "
+	     "level %d => %d\n", lockres->l_name, lockres->l_action,
+	     lockres->l_unlock_action, lockres->l_level, lockres->l_requested);
+
 	switch(lockres->l_action) {
 	case OCFS2_AST_ATTACH:
 		ocfs2_generic_handle_attach_action(lockres);
@@ -1098,8 +1119,8 @@ static void ocfs2_locking_ast(void *opaque)
 		ocfs2_generic_handle_downconvert_action(lockres);
 		break;
 	default:
-		mlog(ML_ERROR, "lockres %s: ast fired with invalid action: %u "
-		     "lockres flags = 0x%lx, unlock action: %u\n",
+		mlog(ML_ERROR, "lockres %s: AST fired with invalid action: %u, "
+		     "flags 0x%lx, unlock: %u\n",
 		     lockres->l_name, lockres->l_action, lockres->l_flags,
 		     lockres->l_unlock_action);
 		BUG();
@@ -1125,6 +1146,88 @@ out:
 	spin_unlock_irqrestore(&lockres->l_lock, flags);
 }
 
+static void ocfs2_unlock_ast(struct ocfs2_dlm_lksb *lksb, int error)
+{
+	struct ocfs2_lock_res *lockres = ocfs2_lksb_to_lock_res(lksb);
+	unsigned long flags;
+
+	mlog_entry_void();
+
+	mlog(ML_BASTS, "UNLOCK AST fired for lockres %s, action = %d\n",
+	     lockres->l_name, lockres->l_unlock_action);
+
+	spin_lock_irqsave(&lockres->l_lock, flags);
+	if (error) {
+		mlog(ML_ERROR, "Dlm passes error %d for lock %s, "
+		     "unlock_action %d\n", error, lockres->l_name,
+		     lockres->l_unlock_action);
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+		mlog_exit_void();
+		return;
+	}
+
+	switch(lockres->l_unlock_action) {
+	case OCFS2_UNLOCK_CANCEL_CONVERT:
+		mlog(0, "Cancel convert success for %s\n", lockres->l_name);
+		lockres->l_action = OCFS2_AST_INVALID;
+		/* Downconvert thread may have requeued this lock, we
+		 * need to wake it. */
+		if (lockres->l_flags & OCFS2_LOCK_BLOCKED)
+			ocfs2_wake_downconvert_thread(ocfs2_get_lockres_osb(lockres));
+		break;
+	case OCFS2_UNLOCK_DROP_LOCK:
+		lockres->l_level = DLM_LOCK_IV;
+		break;
+	default:
+		BUG();
+	}
+
+	lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
+	lockres->l_unlock_action = OCFS2_UNLOCK_INVALID;
+	wake_up(&lockres->l_event);
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+	mlog_exit_void();
+}
+
+/*
+ * This is the filesystem locking protocol.  It provides the lock handling
+ * hooks for the underlying DLM.  It has a maximum version number.
+ * The version number allows interoperability with systems running at
+ * the same major number and an equal or smaller minor number.
+ *
+ * Whenever the filesystem does new things with locks (adds or removes a
+ * lock, orders them differently, does different things underneath a lock),
+ * the version must be changed.  The protocol is negotiated when joining
+ * the dlm domain.  A node may join the domain if its major version is
+ * identical to all other nodes and its minor version is greater than
+ * or equal to all other nodes.  When its minor version is greater than
+ * the other nodes, it will run at the minor version specified by the
+ * other nodes.
+ *
+ * If a locking change is made that will not be compatible with older
+ * versions, the major number must be increased and the minor version set
+ * to zero.  If a change merely adds a behavior that can be disabled when
+ * speaking to older versions, the minor version must be increased.  If a
+ * change adds a fully backwards compatible change (eg, LVB changes that
+ * are just ignored by older versions), the version does not need to be
+ * updated.
+ */
+static struct ocfs2_locking_protocol lproto = {
+	.lp_max_version = {
+		.pv_major = OCFS2_LOCKING_PROTOCOL_MAJOR,
+		.pv_minor = OCFS2_LOCKING_PROTOCOL_MINOR,
+	},
+	.lp_lock_ast		= ocfs2_locking_ast,
+	.lp_blocking_ast	= ocfs2_blocking_ast,
+	.lp_unlock_ast		= ocfs2_unlock_ast,
+};
+
+void ocfs2_set_locking_protocol(void)
+{
+	ocfs2_stack_glue_set_max_proto_version(&lproto.lp_max_version);
+}
+
 static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres,
 						int convert)
 {
@@ -1133,6 +1236,7 @@ static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres,
 	mlog_entry_void();
 	spin_lock_irqsave(&lockres->l_lock, flags);
 	lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
+	lockres_clear_flags(lockres, OCFS2_LOCK_UPCONVERT_FINISHING);
 	if (convert)
 		lockres->l_action = OCFS2_AST_INVALID;
 	else
@@ -1179,8 +1283,7 @@ static int ocfs2_lock_create(struct ocfs2_super *osb,
 			     &lockres->l_lksb,
 			     dlm_flags,
 			     lockres->l_name,
-			     OCFS2_LOCK_ID_MAX_LEN - 1,
-			     lockres);
+			     OCFS2_LOCK_ID_MAX_LEN - 1);
 	lockres_clear_pending(lockres, gen, osb);
 	if (ret) {
 		ocfs2_log_dlm_error("ocfs2_dlm_lock", ret, lockres);
@@ -1323,13 +1426,13 @@ static int __ocfs2_cluster_lock(struct ocfs2_super *osb,
 again:
 	wait = 0;
 
+	spin_lock_irqsave(&lockres->l_lock, flags);
+
 	if (catch_signals && signal_pending(current)) {
 		ret = -ERESTARTSYS;
-		goto out;
+		goto unlock;
 	}
 
-	spin_lock_irqsave(&lockres->l_lock, flags);
-
 	mlog_bug_on_msg(lockres->l_flags & OCFS2_LOCK_FREEING,
 			"Cluster lock called on freeing lockres %s! flags "
 			"0x%lx\n", lockres->l_name, lockres->l_flags);
@@ -1346,6 +1449,25 @@ again:
 		goto unlock;
 	}
 
+	if (lockres->l_flags & OCFS2_LOCK_UPCONVERT_FINISHING) {
+		/*
+		 * We've upconverted. If the lock now has a level we can
+		 * work with, we take it. If, however, the lock is not at the
+		 * required level, we go thru the full cycle. One way this could
+		 * happen is if a process requesting an upconvert to PR is
+		 * closely followed by another requesting upconvert to an EX.
+		 * If the process requesting EX lands here, we want it to
+		 * continue attempting to upconvert and let the process
+		 * requesting PR take the lock.
+		 * If multiple processes request upconvert to PR, the first one
+		 * here will take the lock. The others will have to go thru the
+		 * OCFS2_LOCK_BLOCKED check to ensure that there is no pending
+		 * downconvert request.
+		 */
+		if (level <= lockres->l_level)
+			goto update_holders;
+	}
+
 	if (lockres->l_flags & OCFS2_LOCK_BLOCKED &&
 	    !ocfs2_may_continue_on_blocked_lock(lockres, level)) {
 		/* is the lock is currently blocked on behalf of
@@ -1383,7 +1505,7 @@ again:
 		BUG_ON(level == DLM_LOCK_IV);
 		BUG_ON(level == DLM_LOCK_NL);
 
-		mlog(0, "lock %s, convert from %d to level = %d\n",
+		mlog(ML_BASTS, "lockres %s, convert from %d to %d\n",
 		     lockres->l_name, lockres->l_level, level);
 
 		/* call dlm_lock to upgrade lock now */
@@ -1392,8 +1514,7 @@ again:
 				     &lockres->l_lksb,
 				     lkm_flags,
 				     lockres->l_name,
-				     OCFS2_LOCK_ID_MAX_LEN - 1,
-				     lockres);
+				     OCFS2_LOCK_ID_MAX_LEN - 1);
 		lockres_clear_pending(lockres, gen, osb);
 		if (ret) {
 			if (!(lkm_flags & DLM_LKF_NOQUEUE) ||
@@ -1416,11 +1537,14 @@ again:
 		goto again;
 	}
 
+update_holders:
 	/* Ok, if we get here then we're good to go. */
 	ocfs2_inc_holders(lockres, level);
 
 	ret = 0;
 unlock:
+	lockres_clear_flags(lockres, OCFS2_LOCK_UPCONVERT_FINISHING);
+
 	spin_unlock_irqrestore(&lockres->l_lock, flags);
 out:
 	/*
@@ -1757,7 +1881,7 @@ out:
  * ocfs2_file_lock() and ocfs2_file_unlock() map to a single pair of
  * flock() calls. The locking approach this requires is sufficiently
  * different from all other cluster lock types that we implement a
- * seperate path to the "low-level" dlm calls. In particular:
+ * separate path to the "low-level" dlm calls. In particular:
  *
  * - No optimization of lock levels is done - we take at exactly
  *   what's been requested.
@@ -1827,8 +1951,7 @@ int ocfs2_file_lock(struct file *file, int ex, int trylock)
 	spin_unlock_irqrestore(&lockres->l_lock, flags);
 
 	ret = ocfs2_dlm_lock(osb->cconn, level, &lockres->l_lksb, lkm_flags,
-			     lockres->l_name, OCFS2_LOCK_ID_MAX_LEN - 1,
-			     lockres);
+			     lockres->l_name, OCFS2_LOCK_ID_MAX_LEN - 1);
 	if (ret) {
 		if (!trylock || (ret != -EAGAIN)) {
 			ocfs2_log_dlm_error("ocfs2_dlm_lock", ret, lockres);
@@ -2957,7 +3080,7 @@ int ocfs2_dlm_init(struct ocfs2_super *osb)
 	status = ocfs2_cluster_connect(osb->osb_cluster_stack,
 				       osb->uuid_str,
 				       strlen(osb->uuid_str),
-				       ocfs2_do_node_down, osb,
+				       &lproto, ocfs2_do_node_down, osb,
 				       &conn);
 	if (status) {
 		mlog_errno(status);
@@ -3024,50 +3147,6 @@ void ocfs2_dlm_shutdown(struct ocfs2_super *osb,
 	mlog_exit_void();
 }
 
-static void ocfs2_unlock_ast(void *opaque, int error)
-{
-	struct ocfs2_lock_res *lockres = opaque;
-	unsigned long flags;
-
-	mlog_entry_void();
-
-	mlog(0, "UNLOCK AST called on lock %s, action = %d\n", lockres->l_name,
-	     lockres->l_unlock_action);
-
-	spin_lock_irqsave(&lockres->l_lock, flags);
-	if (error) {
-		mlog(ML_ERROR, "Dlm passes error %d for lock %s, "
-		     "unlock_action %d\n", error, lockres->l_name,
-		     lockres->l_unlock_action);
-		spin_unlock_irqrestore(&lockres->l_lock, flags);
-		mlog_exit_void();
-		return;
-	}
-
-	switch(lockres->l_unlock_action) {
-	case OCFS2_UNLOCK_CANCEL_CONVERT:
-		mlog(0, "Cancel convert success for %s\n", lockres->l_name);
-		lockres->l_action = OCFS2_AST_INVALID;
-		/* Downconvert thread may have requeued this lock, we
-		 * need to wake it. */
-		if (lockres->l_flags & OCFS2_LOCK_BLOCKED)
-			ocfs2_wake_downconvert_thread(ocfs2_get_lockres_osb(lockres));
-		break;
-	case OCFS2_UNLOCK_DROP_LOCK:
-		lockres->l_level = DLM_LOCK_IV;
-		break;
-	default:
-		BUG();
-	}
-
-	lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
-	lockres->l_unlock_action = OCFS2_UNLOCK_INVALID;
-	wake_up(&lockres->l_event);
-	spin_unlock_irqrestore(&lockres->l_lock, flags);
-
-	mlog_exit_void();
-}
-
 static int ocfs2_drop_lock(struct ocfs2_super *osb,
 			   struct ocfs2_lock_res *lockres)
 {
@@ -3135,8 +3214,7 @@ static int ocfs2_drop_lock(struct ocfs2_super *osb,
 
 	mlog(0, "lock %s\n", lockres->l_name);
 
-	ret = ocfs2_dlm_unlock(osb->cconn, &lockres->l_lksb, lkm_flags,
-			       lockres);
+	ret = ocfs2_dlm_unlock(osb->cconn, &lockres->l_lksb, lkm_flags);
 	if (ret) {
 		ocfs2_log_dlm_error("ocfs2_dlm_unlock", ret, lockres);
 		mlog(ML_ERROR, "lockres flags: %lu\n", lockres->l_flags);
@@ -3155,7 +3233,7 @@ out:
 /* Mark the lockres as being dropped. It will no longer be
  * queued if blocking, but we still may have to wait on it
  * being dequeued from the downconvert thread before we can consider
- * it safe to drop. 
+ * it safe to drop.
  *
  * You can *not* attempt to call cluster_lock on this lockres anymore. */
 void ocfs2_mark_lockres_freeing(struct ocfs2_lock_res *lockres)
@@ -3244,13 +3322,20 @@ static unsigned int ocfs2_prepare_downconvert(struct ocfs2_lock_res *lockres,
 	BUG_ON(lockres->l_blocking <= DLM_LOCK_NL);
 
 	if (lockres->l_level <= new_level) {
-		mlog(ML_ERROR, "lockres->l_level (%d) <= new_level (%d)\n",
-		     lockres->l_level, new_level);
+		mlog(ML_ERROR, "lockres %s, lvl %d <= %d, blcklst %d, mask %d, "
+		     "type %d, flags 0x%lx, hold %d %d, act %d %d, req %d, "
+		     "block %d, pgen %d\n", lockres->l_name, lockres->l_level,
+		     new_level, list_empty(&lockres->l_blocked_list),
+		     list_empty(&lockres->l_mask_waiters), lockres->l_type,
+		     lockres->l_flags, lockres->l_ro_holders,
+		     lockres->l_ex_holders, lockres->l_action,
+		     lockres->l_unlock_action, lockres->l_requested,
+		     lockres->l_blocking, lockres->l_pending_gen);
 		BUG();
 	}
 
-	mlog(0, "lock %s, new_level = %d, l_blocking = %d\n",
-	     lockres->l_name, new_level, lockres->l_blocking);
+	mlog(ML_BASTS, "lockres %s, level %d => %d, blocking %d\n",
+	     lockres->l_name, lockres->l_level, new_level, lockres->l_blocking);
 
 	lockres->l_action = OCFS2_AST_DOWNCONVERT;
 	lockres->l_requested = new_level;
@@ -3269,6 +3354,9 @@ static int ocfs2_downconvert_lock(struct ocfs2_super *osb,
 
 	mlog_entry_void();
 
+	mlog(ML_BASTS, "lockres %s, level %d => %d\n", lockres->l_name,
+	     lockres->l_level, new_level);
+
 	if (lvb)
 		dlm_flags |= DLM_LKF_VALBLK;
 
@@ -3277,8 +3365,7 @@ static int ocfs2_downconvert_lock(struct ocfs2_super *osb,
 			     &lockres->l_lksb,
 			     dlm_flags,
 			     lockres->l_name,
-			     OCFS2_LOCK_ID_MAX_LEN - 1,
-			     lockres);
+			     OCFS2_LOCK_ID_MAX_LEN - 1);
 	lockres_clear_pending(lockres, generation, osb);
 	if (ret) {
 		ocfs2_log_dlm_error("ocfs2_dlm_lock", ret, lockres);
@@ -3299,14 +3386,12 @@ static int ocfs2_prepare_cancel_convert(struct ocfs2_super *osb,
 	assert_spin_locked(&lockres->l_lock);
 
 	mlog_entry_void();
-	mlog(0, "lock %s\n", lockres->l_name);
 
 	if (lockres->l_unlock_action == OCFS2_UNLOCK_CANCEL_CONVERT) {
 		/* If we're already trying to cancel a lock conversion
 		 * then just drop the spinlock and allow the caller to
 		 * requeue this lock. */
-
-		mlog(0, "Lockres %s, skip convert\n", lockres->l_name);
+		mlog(ML_BASTS, "lockres %s, skip convert\n", lockres->l_name);
 		return 0;
 	}
 
@@ -3321,6 +3406,8 @@ static int ocfs2_prepare_cancel_convert(struct ocfs2_super *osb,
 			"lock %s, invalid flags: 0x%lx\n",
 			lockres->l_name, lockres->l_flags);
 
+	mlog(ML_BASTS, "lockres %s\n", lockres->l_name);
+
 	return 1;
 }
 
@@ -3330,16 +3417,15 @@ static int ocfs2_cancel_convert(struct ocfs2_super *osb,
 	int ret;
 
 	mlog_entry_void();
-	mlog(0, "lock %s\n", lockres->l_name);
 
 	ret = ocfs2_dlm_unlock(osb->cconn, &lockres->l_lksb,
-			       DLM_LKF_CANCEL, lockres);
+			       DLM_LKF_CANCEL);
 	if (ret) {
 		ocfs2_log_dlm_error("ocfs2_dlm_unlock", ret, lockres);
 		ocfs2_recover_from_dlm_error(lockres, 0);
 	}
 
-	mlog(0, "lock %s return from ocfs2_dlm_unlock\n", lockres->l_name);
+	mlog(ML_BASTS, "lockres %s\n", lockres->l_name);
 
 	mlog_exit(ret);
 	return ret;
@@ -3352,6 +3438,7 @@ static int ocfs2_unblock_lock(struct ocfs2_super *osb,
 	unsigned long flags;
 	int blocking;
 	int new_level;
+	int level;
 	int ret = 0;
 	int set_lvb = 0;
 	unsigned int gen;
@@ -3360,9 +3447,17 @@ static int ocfs2_unblock_lock(struct ocfs2_super *osb,
 
 	spin_lock_irqsave(&lockres->l_lock, flags);
 
-	BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED));
-
 recheck:
+	/*
+	 * Is it still blocking? If not, we have no more work to do.
+	 */
+	if (!(lockres->l_flags & OCFS2_LOCK_BLOCKED)) {
+		BUG_ON(lockres->l_blocking != DLM_LOCK_NL);
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+		ret = 0;
+		goto leave;
+	}
+
 	if (lockres->l_flags & OCFS2_LOCK_BUSY) {
 		/* XXX
 		 * This is a *big* race.  The OCFS2_LOCK_PENDING flag
@@ -3387,8 +3482,11 @@ recheck:
 		 * at the same time they set OCFS2_DLM_BUSY.  They must
 		 * clear OCFS2_DLM_PENDING after dlm_lock() returns.
 		 */
-		if (lockres->l_flags & OCFS2_LOCK_PENDING)
+		if (lockres->l_flags & OCFS2_LOCK_PENDING) {
+			mlog(ML_BASTS, "lockres %s, ReQ: Pending\n",
+			     lockres->l_name);
 			goto leave_requeue;
+		}
 
 		ctl->requeue = 1;
 		ret = ocfs2_prepare_cancel_convert(osb, lockres);
@@ -3401,31 +3499,70 @@ recheck:
 		goto leave;
 	}
 
+	/*
+	 * This prevents livelocks. OCFS2_LOCK_UPCONVERT_FINISHING flag is
+	 * set when the ast is received for an upconvert just before the
+	 * OCFS2_LOCK_BUSY flag is cleared. Now if the fs received a bast
+	 * on the heels of the ast, we want to delay the downconvert just
+	 * enough to allow the up requestor to do its task. Because this
+	 * lock is in the blocked queue, the lock will be downconverted
+	 * as soon as the requestor is done with the lock.
+	 */
+	if (lockres->l_flags & OCFS2_LOCK_UPCONVERT_FINISHING)
+		goto leave_requeue;
+
+	/*
+	 * How can we block and yet be at NL?  We were trying to upconvert
+	 * from NL and got canceled.  The code comes back here, and now
+	 * we notice and clear BLOCKING.
+	 */
+	if (lockres->l_level == DLM_LOCK_NL) {
+		BUG_ON(lockres->l_ex_holders || lockres->l_ro_holders);
+		mlog(ML_BASTS, "lockres %s, Aborting dc\n", lockres->l_name);
+		lockres->l_blocking = DLM_LOCK_NL;
+		lockres_clear_flags(lockres, OCFS2_LOCK_BLOCKED);
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+		goto leave;
+	}
+
 	/* if we're blocking an exclusive and we have *any* holders,
 	 * then requeue. */
 	if ((lockres->l_blocking == DLM_LOCK_EX)
-	    && (lockres->l_ex_holders || lockres->l_ro_holders))
+	    && (lockres->l_ex_holders || lockres->l_ro_holders)) {
+		mlog(ML_BASTS, "lockres %s, ReQ: EX/PR Holders %u,%u\n",
+		     lockres->l_name, lockres->l_ex_holders,
+		     lockres->l_ro_holders);
 		goto leave_requeue;
+	}
 
 	/* If it's a PR we're blocking, then only
 	 * requeue if we've got any EX holders */
 	if (lockres->l_blocking == DLM_LOCK_PR &&
-	    lockres->l_ex_holders)
+	    lockres->l_ex_holders) {
+		mlog(ML_BASTS, "lockres %s, ReQ: EX Holders %u\n",
+		     lockres->l_name, lockres->l_ex_holders);
 		goto leave_requeue;
+	}
 
 	/*
 	 * Can we get a lock in this state if the holder counts are
 	 * zero? The meta data unblock code used to check this.
 	 */
 	if ((lockres->l_ops->flags & LOCK_TYPE_REQUIRES_REFRESH)
-	    && (lockres->l_flags & OCFS2_LOCK_REFRESHING))
+	    && (lockres->l_flags & OCFS2_LOCK_REFRESHING)) {
+		mlog(ML_BASTS, "lockres %s, ReQ: Lock Refreshing\n",
+		     lockres->l_name);
 		goto leave_requeue;
+	}
 
 	new_level = ocfs2_highest_compat_lock_level(lockres->l_blocking);
 
 	if (lockres->l_ops->check_downconvert
-	    && !lockres->l_ops->check_downconvert(lockres, new_level))
+	    && !lockres->l_ops->check_downconvert(lockres, new_level)) {
+		mlog(ML_BASTS, "lockres %s, ReQ: Checkpointing\n",
+		     lockres->l_name);
 		goto leave_requeue;
+	}
 
 	/* If we get here, then we know that there are no more
 	 * incompatible holders (and anyone asking for an incompatible
@@ -3438,17 +3575,24 @@ recheck:
 	 * may sleep, so we save off a copy of what we're blocking as
 	 * it may change while we're not holding the spin lock. */
 	blocking = lockres->l_blocking;
+	level = lockres->l_level;
 	spin_unlock_irqrestore(&lockres->l_lock, flags);
 
 	ctl->unblock_action = lockres->l_ops->downconvert_worker(lockres, blocking);
 
-	if (ctl->unblock_action == UNBLOCK_STOP_POST)
+	if (ctl->unblock_action == UNBLOCK_STOP_POST) {
+		mlog(ML_BASTS, "lockres %s, UNBLOCK_STOP_POST\n",
+		     lockres->l_name);
 		goto leave;
+	}
 
 	spin_lock_irqsave(&lockres->l_lock, flags);
-	if (blocking != lockres->l_blocking) {
+	if ((blocking != lockres->l_blocking) || (level != lockres->l_level)) {
 		/* If this changed underneath us, then we can't drop
 		 * it just yet. */
+		mlog(ML_BASTS, "lockres %s, block=%d:%d, level=%d:%d, "
+		     "Recheck\n", lockres->l_name, blocking,
+		     lockres->l_blocking, level, lockres->l_level);
 		goto recheck;
 	}
 
@@ -3753,7 +3897,8 @@ static int ocfs2_refresh_qinfo(struct ocfs2_mem_dqinfo *oinfo)
 		oinfo->dqi_gi.dqi_free_entry =
 					be32_to_cpu(lvb->lvb_free_entry);
 	} else {
-		status = ocfs2_read_quota_block(oinfo->dqi_gqinode, 0, &bh);
+		status = ocfs2_read_quota_phys_block(oinfo->dqi_gqinode,
+						     oinfo->dqi_giblk, &bh);
 		if (status) {
 			mlog_errno(status);
 			goto bail;
@@ -3843,45 +3988,6 @@ void ocfs2_refcount_unlock(struct ocfs2_refcount_tree *ref_tree, int ex)
 		ocfs2_cluster_unlock(osb, lockres, level);
 }
 
-/*
- * This is the filesystem locking protocol.  It provides the lock handling
- * hooks for the underlying DLM.  It has a maximum version number.
- * The version number allows interoperability with systems running at
- * the same major number and an equal or smaller minor number.
- *
- * Whenever the filesystem does new things with locks (adds or removes a
- * lock, orders them differently, does different things underneath a lock),
- * the version must be changed.  The protocol is negotiated when joining
- * the dlm domain.  A node may join the domain if its major version is
- * identical to all other nodes and its minor version is greater than
- * or equal to all other nodes.  When its minor version is greater than
- * the other nodes, it will run at the minor version specified by the
- * other nodes.
- *
- * If a locking change is made that will not be compatible with older
- * versions, the major number must be increased and the minor version set
- * to zero.  If a change merely adds a behavior that can be disabled when
- * speaking to older versions, the minor version must be increased.  If a
- * change adds a fully backwards compatible change (eg, LVB changes that
- * are just ignored by older versions), the version does not need to be
- * updated.
- */
-static struct ocfs2_locking_protocol lproto = {
-	.lp_max_version = {
-		.pv_major = OCFS2_LOCKING_PROTOCOL_MAJOR,
-		.pv_minor = OCFS2_LOCKING_PROTOCOL_MINOR,
-	},
-	.lp_lock_ast		= ocfs2_locking_ast,
-	.lp_blocking_ast	= ocfs2_blocking_ast,
-	.lp_unlock_ast		= ocfs2_unlock_ast,
-};
-
-void ocfs2_set_locking_protocol(void)
-{
-	ocfs2_stack_glue_set_locking_protocol(&lproto);
-}
-
-
 static void ocfs2_process_blocked_lock(struct ocfs2_super *osb,
 				       struct ocfs2_lock_res *lockres)
 {
@@ -3898,7 +4004,7 @@ static void ocfs2_process_blocked_lock(struct ocfs2_super *osb,
 	BUG_ON(!lockres);
 	BUG_ON(!lockres->l_ops);
 
-	mlog(0, "lockres %s blocked.\n", lockres->l_name);
+	mlog(ML_BASTS, "lockres %s blocked\n", lockres->l_name);
 
 	/* Detect whether a lock has been marked as going away while
 	 * the downconvert thread was processing other things. A lock can
@@ -3921,7 +4027,7 @@ unqueue:
 	} else
 		ocfs2_schedule_blocked_lock(osb, lockres);
 
-	mlog(0, "lockres %s, requeue = %s.\n", lockres->l_name,
+	mlog(ML_BASTS, "lockres %s, requeue = %s.\n", lockres->l_name,
 	     ctl.requeue ? "yes" : "no");
 	spin_unlock_irqrestore(&lockres->l_lock, flags);
 
@@ -3943,7 +4049,7 @@ static void ocfs2_schedule_blocked_lock(struct ocfs2_super *osb,
 		/* Do not schedule a lock for downconvert when it's on
 		 * the way to destruction - any nodes wanting access
 		 * to the resource will get it soon. */
-		mlog(0, "Lockres %s won't be scheduled: flags 0x%lx\n",
+		mlog(ML_BASTS, "lockres %s won't be scheduled: flags 0x%lx\n",
 		     lockres->l_name, lockres->l_flags);
 		return;
 	}
diff --git a/fs/ocfs2/export.c b/fs/ocfs2/export.c
index 15713cbb865c..19ad145d2af3 100644
--- a/fs/ocfs2/export.c
+++ b/fs/ocfs2/export.c
@@ -239,7 +239,7 @@ static int ocfs2_encode_fh(struct dentry *dentry, u32 *fh_in, int *max_len,
 		mlog(0, "Encoding parent: blkno: %llu, generation: %u\n",
 		     (unsigned long long)blkno, generation);
 	}
-	
+
 	*max_len = len;
 
 bail:
diff --git a/fs/ocfs2/extent_map.c b/fs/ocfs2/extent_map.c
index d35a27f4523e..09e3fdfa6d33 100644
--- a/fs/ocfs2/extent_map.c
+++ b/fs/ocfs2/extent_map.c
@@ -24,6 +24,7 @@
 
 #include <linux/fs.h>
 #include <linux/init.h>
+#include <linux/slab.h>
 #include <linux/types.h>
 #include <linux/fiemap.h>
 
@@ -192,7 +193,7 @@ static int ocfs2_try_to_merge_extent_map(struct ocfs2_extent_map_item *emi,
 		emi->ei_clusters += ins->ei_clusters;
 		return 1;
 	} else if ((ins->ei_phys + ins->ei_clusters) == emi->ei_phys &&
-		   (ins->ei_cpos + ins->ei_clusters) == emi->ei_phys &&
+		   (ins->ei_cpos + ins->ei_clusters) == emi->ei_cpos &&
 		   ins->ei_flags == emi->ei_flags) {
 		emi->ei_phys = ins->ei_phys;
 		emi->ei_cpos = ins->ei_cpos;
@@ -453,7 +454,7 @@ static int ocfs2_get_clusters_nocache(struct inode *inode,
 	if (i == -1) {
 		/*
 		 * Holes can be larger than the maximum size of an
-		 * extent, so we return their lengths in a seperate
+		 * extent, so we return their lengths in a separate
 		 * field.
 		 */
 		if (hole_len) {
diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
index 06ccf6a86d35..2b10b36d1577 100644
--- a/fs/ocfs2/file.c
+++ b/fs/ocfs2/file.c
@@ -107,6 +107,9 @@ static int ocfs2_file_open(struct inode *inode, struct file *file)
 	mlog_entry("(0x%p, 0x%p, '%.*s')\n", inode, file,
 		   file->f_path.dentry->d_name.len, file->f_path.dentry->d_name.name);
 
+	if (file->f_mode & FMODE_WRITE)
+		dquot_initialize(inode);
+
 	spin_lock(&oi->ip_lock);
 
 	/* Check that the inode hasn't been wiped from disk by another
@@ -172,13 +175,12 @@ static int ocfs2_dir_release(struct inode *inode, struct file *file)
 	return 0;
 }
 
-static int ocfs2_sync_file(struct file *file,
-			   struct dentry *dentry,
-			   int datasync)
+static int ocfs2_sync_file(struct file *file, int datasync)
 {
 	int err = 0;
 	journal_t *journal;
-	struct inode *inode = dentry->d_inode;
+	struct dentry *dentry = file->f_path.dentry;
+	struct inode *inode = file->f_mapping->host;
 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 
 	mlog_entry("(0x%p, 0x%p, %d, '%.*s')\n", file, dentry, datasync,
@@ -275,10 +277,7 @@ int ocfs2_update_inode_atime(struct inode *inode,
 	inode->i_atime = CURRENT_TIME;
 	di->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
 	di->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
-
-	ret = ocfs2_journal_dirty(handle, bh);
-	if (ret < 0)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, bh);
 
 out_commit:
 	ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle);
@@ -427,9 +426,7 @@ static int ocfs2_orphan_for_truncate(struct ocfs2_super *osb,
 	di->i_ctime = di->i_mtime = cpu_to_le64(inode->i_ctime.tv_sec);
 	di->i_ctime_nsec = di->i_mtime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
 
-	status = ocfs2_journal_dirty(handle, fe_bh);
-	if (status < 0)
-		mlog_errno(status);
+	ocfs2_journal_dirty(handle, fe_bh);
 
 out_commit:
 	ocfs2_commit_trans(osb, handle);
@@ -446,7 +443,6 @@ static int ocfs2_truncate_file(struct inode *inode,
 	int status = 0;
 	struct ocfs2_dinode *fe = NULL;
 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
-	struct ocfs2_truncate_context *tc = NULL;
 
 	mlog_entry("(inode = %llu, new_i_size = %llu\n",
 		   (unsigned long long)OCFS2_I(inode)->ip_blkno,
@@ -485,6 +481,9 @@ static int ocfs2_truncate_file(struct inode *inode,
 
 	down_write(&OCFS2_I(inode)->ip_alloc_sem);
 
+	ocfs2_resv_discard(&osb->osb_la_resmap,
+			   &OCFS2_I(inode)->ip_la_data_resv);
+
 	/*
 	 * The inode lock forced other nodes to sync and drop their
 	 * pages, which (correctly) happens even if we have a truncate
@@ -514,13 +513,7 @@ static int ocfs2_truncate_file(struct inode *inode,
 		goto bail_unlock_sem;
 	}
 
-	status = ocfs2_prepare_truncate(osb, inode, di_bh, &tc);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail_unlock_sem;
-	}
-
-	status = ocfs2_commit_truncate(osb, inode, di_bh, tc);
+	status = ocfs2_commit_truncate(osb, inode, di_bh);
 	if (status < 0) {
 		mlog_errno(status);
 		goto bail_unlock_sem;
@@ -629,11 +622,10 @@ restart_all:
 	}
 
 restarted_transaction:
-	if (vfs_dq_alloc_space_nodirty(inode, ocfs2_clusters_to_bytes(osb->sb,
-	    clusters_to_add))) {
-		status = -EDQUOT;
+	status = dquot_alloc_space_nodirty(inode,
+			ocfs2_clusters_to_bytes(osb->sb, clusters_to_add));
+	if (status)
 		goto leave;
-	}
 	did_quota = 1;
 
 	/* reserve a write to the file entry early on - that we if we
@@ -664,17 +656,13 @@ restarted_transaction:
 		goto leave;
 	}
 
-	status = ocfs2_journal_dirty(handle, bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto leave;
-	}
+	ocfs2_journal_dirty(handle, bh);
 
 	spin_lock(&OCFS2_I(inode)->ip_lock);
 	clusters_to_add -= (OCFS2_I(inode)->ip_clusters - prev_clusters);
 	spin_unlock(&OCFS2_I(inode)->ip_lock);
 	/* Release unused quota reservation */
-	vfs_dq_free_space(inode,
+	dquot_free_space(inode,
 			ocfs2_clusters_to_bytes(osb->sb, clusters_to_add));
 	did_quota = 0;
 
@@ -682,6 +670,7 @@ restarted_transaction:
 		if (why == RESTART_META) {
 			mlog(0, "restarting function.\n");
 			restart_func = 1;
+			status = 0;
 		} else {
 			BUG_ON(why != RESTART_TRANS);
 
@@ -710,7 +699,7 @@ restarted_transaction:
 
 leave:
 	if (status < 0 && did_quota)
-		vfs_dq_free_space(inode,
+		dquot_free_space(inode,
 			ocfs2_clusters_to_bytes(osb->sb, clusters_to_add));
 	if (handle) {
 		ocfs2_commit_trans(osb, handle);
@@ -735,28 +724,55 @@ leave:
 	return status;
 }
 
+/*
+ * While a write will already be ordering the data, a truncate will not.
+ * Thus, we need to explicitly order the zeroed pages.
+ */
+static handle_t *ocfs2_zero_start_ordered_transaction(struct inode *inode)
+{
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	handle_t *handle = NULL;
+	int ret = 0;
+
+	if (!ocfs2_should_order_data(inode))
+		goto out;
+
+	handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
+	if (IS_ERR(handle)) {
+		ret = -ENOMEM;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	ret = ocfs2_jbd2_file_inode(handle, inode);
+	if (ret < 0)
+		mlog_errno(ret);
+
+out:
+	if (ret) {
+		if (!IS_ERR(handle))
+			ocfs2_commit_trans(osb, handle);
+		handle = ERR_PTR(ret);
+	}
+	return handle;
+}
+
 /* Some parts of this taken from generic_cont_expand, which turned out
  * to be too fragile to do exactly what we need without us having to
  * worry about recursive locking in ->write_begin() and ->write_end(). */
-static int ocfs2_write_zero_page(struct inode *inode,
-				 u64 size)
+static int ocfs2_write_zero_page(struct inode *inode, u64 abs_from,
+				 u64 abs_to)
 {
 	struct address_space *mapping = inode->i_mapping;
 	struct page *page;
-	unsigned long index;
-	unsigned int offset;
+	unsigned long index = abs_from >> PAGE_CACHE_SHIFT;
 	handle_t *handle = NULL;
-	int ret;
+	int ret = 0;
+	unsigned zero_from, zero_to, block_start, block_end;
 
-	offset = (size & (PAGE_CACHE_SIZE-1)); /* Within page */
-	/* ugh.  in prepare/commit_write, if from==to==start of block, we 
-	** skip the prepare.  make sure we never send an offset for the start
-	** of a block
-	*/
-	if ((offset & (inode->i_sb->s_blocksize - 1)) == 0) {
-		offset++;
-	}
-	index = size >> PAGE_CACHE_SHIFT;
+	BUG_ON(abs_from >= abs_to);
+	BUG_ON(abs_to > (((u64)index + 1) << PAGE_CACHE_SHIFT));
+	BUG_ON(abs_from & (inode->i_blkbits - 1));
 
 	page = grab_cache_page(mapping, index);
 	if (!page) {
@@ -765,31 +781,56 @@ static int ocfs2_write_zero_page(struct inode *inode,
 		goto out;
 	}
 
-	ret = ocfs2_prepare_write_nolock(inode, page, offset, offset);
-	if (ret < 0) {
-		mlog_errno(ret);
-		goto out_unlock;
-	}
+	/* Get the offsets within the page that we want to zero */
+	zero_from = abs_from & (PAGE_CACHE_SIZE - 1);
+	zero_to = abs_to & (PAGE_CACHE_SIZE - 1);
+	if (!zero_to)
+		zero_to = PAGE_CACHE_SIZE;
 
-	if (ocfs2_should_order_data(inode)) {
-		handle = ocfs2_start_walk_page_trans(inode, page, offset,
-						     offset);
-		if (IS_ERR(handle)) {
-			ret = PTR_ERR(handle);
-			handle = NULL;
+	mlog(0,
+	     "abs_from = %llu, abs_to = %llu, index = %lu, zero_from = %u, zero_to = %u\n",
+	     (unsigned long long)abs_from, (unsigned long long)abs_to,
+	     index, zero_from, zero_to);
+
+	/* We know that zero_from is block aligned */
+	for (block_start = zero_from; block_start < zero_to;
+	     block_start = block_end) {
+		block_end = block_start + (1 << inode->i_blkbits);
+
+		/*
+		 * block_start is block-aligned.  Bump it by one to
+		 * force ocfs2_{prepare,commit}_write() to zero the
+		 * whole block.
+		 */
+		ret = ocfs2_prepare_write_nolock(inode, page,
+						 block_start + 1,
+						 block_start + 1);
+		if (ret < 0) {
+			mlog_errno(ret);
 			goto out_unlock;
 		}
-	}
 
-	/* must not update i_size! */
-	ret = block_commit_write(page, offset, offset);
-	if (ret < 0)
-		mlog_errno(ret);
-	else
-		ret = 0;
+		if (!handle) {
+			handle = ocfs2_zero_start_ordered_transaction(inode);
+			if (IS_ERR(handle)) {
+				ret = PTR_ERR(handle);
+				handle = NULL;
+				break;
+			}
+		}
+
+		/* must not update i_size! */
+		ret = block_commit_write(page, block_start + 1,
+					 block_start + 1);
+		if (ret < 0)
+			mlog_errno(ret);
+		else
+			ret = 0;
+	}
 
 	if (handle)
 		ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle);
+
 out_unlock:
 	unlock_page(page);
 	page_cache_release(page);
@@ -797,22 +838,114 @@ out:
 	return ret;
 }
 
-static int ocfs2_zero_extend(struct inode *inode,
-			     u64 zero_to_size)
+/*
+ * Find the next range to zero.  We do this in terms of bytes because
+ * that's what ocfs2_zero_extend() wants, and it is dealing with the
+ * pagecache.  We may return multiple extents.
+ *
+ * zero_start and zero_end are ocfs2_zero_extend()s current idea of what
+ * needs to be zeroed.  range_start and range_end return the next zeroing
+ * range.  A subsequent call should pass the previous range_end as its
+ * zero_start.  If range_end is 0, there's nothing to do.
+ *
+ * Unwritten extents are skipped over.  Refcounted extents are CoWd.
+ */
+static int ocfs2_zero_extend_get_range(struct inode *inode,
+				       struct buffer_head *di_bh,
+				       u64 zero_start, u64 zero_end,
+				       u64 *range_start, u64 *range_end)
 {
-	int ret = 0;
-	u64 start_off;
-	struct super_block *sb = inode->i_sb;
+	int rc = 0, needs_cow = 0;
+	u32 p_cpos, zero_clusters = 0;
+	u32 zero_cpos =
+		zero_start >> OCFS2_SB(inode->i_sb)->s_clustersize_bits;
+	u32 last_cpos = ocfs2_clusters_for_bytes(inode->i_sb, zero_end);
+	unsigned int num_clusters = 0;
+	unsigned int ext_flags = 0;
 
-	start_off = ocfs2_align_bytes_to_blocks(sb, i_size_read(inode));
-	while (start_off < zero_to_size) {
-		ret = ocfs2_write_zero_page(inode, start_off);
-		if (ret < 0) {
-			mlog_errno(ret);
+	while (zero_cpos < last_cpos) {
+		rc = ocfs2_get_clusters(inode, zero_cpos, &p_cpos,
+					&num_clusters, &ext_flags);
+		if (rc) {
+			mlog_errno(rc);
 			goto out;
 		}
 
-		start_off += sb->s_blocksize;
+		if (p_cpos && !(ext_flags & OCFS2_EXT_UNWRITTEN)) {
+			zero_clusters = num_clusters;
+			if (ext_flags & OCFS2_EXT_REFCOUNTED)
+				needs_cow = 1;
+			break;
+		}
+
+		zero_cpos += num_clusters;
+	}
+	if (!zero_clusters) {
+		*range_end = 0;
+		goto out;
+	}
+
+	while ((zero_cpos + zero_clusters) < last_cpos) {
+		rc = ocfs2_get_clusters(inode, zero_cpos + zero_clusters,
+					&p_cpos, &num_clusters,
+					&ext_flags);
+		if (rc) {
+			mlog_errno(rc);
+			goto out;
+		}
+
+		if (!p_cpos || (ext_flags & OCFS2_EXT_UNWRITTEN))
+			break;
+		if (ext_flags & OCFS2_EXT_REFCOUNTED)
+			needs_cow = 1;
+		zero_clusters += num_clusters;
+	}
+	if ((zero_cpos + zero_clusters) > last_cpos)
+		zero_clusters = last_cpos - zero_cpos;
+
+	if (needs_cow) {
+		rc = ocfs2_refcount_cow(inode, di_bh, zero_cpos, zero_clusters,
+					UINT_MAX);
+		if (rc) {
+			mlog_errno(rc);
+			goto out;
+		}
+	}
+
+	*range_start = ocfs2_clusters_to_bytes(inode->i_sb, zero_cpos);
+	*range_end = ocfs2_clusters_to_bytes(inode->i_sb,
+					     zero_cpos + zero_clusters);
+
+out:
+	return rc;
+}
+
+/*
+ * Zero one range returned from ocfs2_zero_extend_get_range().  The caller
+ * has made sure that the entire range needs zeroing.
+ */
+static int ocfs2_zero_extend_range(struct inode *inode, u64 range_start,
+				   u64 range_end)
+{
+	int rc = 0;
+	u64 next_pos;
+	u64 zero_pos = range_start;
+
+	mlog(0, "range_start = %llu, range_end = %llu\n",
+	     (unsigned long long)range_start,
+	     (unsigned long long)range_end);
+	BUG_ON(range_start >= range_end);
+
+	while (zero_pos < range_end) {
+		next_pos = (zero_pos & PAGE_CACHE_MASK) + PAGE_CACHE_SIZE;
+		if (next_pos > range_end)
+			next_pos = range_end;
+		rc = ocfs2_write_zero_page(inode, zero_pos, next_pos);
+		if (rc < 0) {
+			mlog_errno(rc);
+			break;
+		}
+		zero_pos = next_pos;
 
 		/*
 		 * Very large extends have the potential to lock up
@@ -821,16 +954,63 @@ static int ocfs2_zero_extend(struct inode *inode,
 		cond_resched();
 	}
 
-out:
+	return rc;
+}
+
+int ocfs2_zero_extend(struct inode *inode, struct buffer_head *di_bh,
+		      loff_t zero_to_size)
+{
+	int ret = 0;
+	u64 zero_start, range_start = 0, range_end = 0;
+	struct super_block *sb = inode->i_sb;
+
+	zero_start = ocfs2_align_bytes_to_blocks(sb, i_size_read(inode));
+	mlog(0, "zero_start %llu for i_size %llu\n",
+	     (unsigned long long)zero_start,
+	     (unsigned long long)i_size_read(inode));
+	while (zero_start < zero_to_size) {
+		ret = ocfs2_zero_extend_get_range(inode, di_bh, zero_start,
+						  zero_to_size,
+						  &range_start,
+						  &range_end);
+		if (ret) {
+			mlog_errno(ret);
+			break;
+		}
+		if (!range_end)
+			break;
+		/* Trim the ends */
+		if (range_start < zero_start)
+			range_start = zero_start;
+		if (range_end > zero_to_size)
+			range_end = zero_to_size;
+
+		ret = ocfs2_zero_extend_range(inode, range_start,
+					      range_end);
+		if (ret) {
+			mlog_errno(ret);
+			break;
+		}
+		zero_start = range_end;
+	}
+
 	return ret;
 }
 
-int ocfs2_extend_no_holes(struct inode *inode, u64 new_i_size, u64 zero_to)
+int ocfs2_extend_no_holes(struct inode *inode, struct buffer_head *di_bh,
+			  u64 new_i_size, u64 zero_to)
 {
 	int ret;
 	u32 clusters_to_add;
 	struct ocfs2_inode_info *oi = OCFS2_I(inode);
 
+	/*
+	 * Only quota files call this without a bh, and they can't be
+	 * refcounted.
+	 */
+	BUG_ON(!di_bh && (oi->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL));
+	BUG_ON(!di_bh && !(oi->ip_flags & OCFS2_INODE_SYSTEM_FILE));
+
 	clusters_to_add = ocfs2_clusters_for_bytes(inode->i_sb, new_i_size);
 	if (clusters_to_add < oi->ip_clusters)
 		clusters_to_add = 0;
@@ -851,7 +1031,7 @@ int ocfs2_extend_no_holes(struct inode *inode, u64 new_i_size, u64 zero_to)
 	 * still need to zero the area between the old i_size and the
 	 * new i_size.
 	 */
-	ret = ocfs2_zero_extend(inode, zero_to);
+	ret = ocfs2_zero_extend(inode, di_bh, zero_to);
 	if (ret < 0)
 		mlog_errno(ret);
 
@@ -873,27 +1053,15 @@ static int ocfs2_extend_file(struct inode *inode,
 		goto out;
 
 	if (i_size_read(inode) == new_i_size)
-  		goto out;
+		goto out;
 	BUG_ON(new_i_size < i_size_read(inode));
 
 	/*
-	 * Fall through for converting inline data, even if the fs
-	 * supports sparse files.
-	 *
-	 * The check for inline data here is legal - nobody can add
-	 * the feature since we have i_mutex. We must check it again
-	 * after acquiring ip_alloc_sem though, as paths like mmap
-	 * might have raced us to converting the inode to extents.
-	 */
-	if (!(oi->ip_dyn_features & OCFS2_INLINE_DATA_FL)
-	    && ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb)))
-		goto out_update_size;
-
-	/*
 	 * The alloc sem blocks people in read/write from reading our
 	 * allocation until we're done changing it. We depend on
 	 * i_mutex to block other extend/truncate calls while we're
-	 * here.
+	 * here.  We even have to hold it for sparse files because there
+	 * might be some tail zeroing.
 	 */
 	down_write(&oi->ip_alloc_sem);
 
@@ -910,14 +1078,16 @@ static int ocfs2_extend_file(struct inode *inode,
 		ret = ocfs2_convert_inline_data_to_extents(inode, di_bh);
 		if (ret) {
 			up_write(&oi->ip_alloc_sem);
-
 			mlog_errno(ret);
 			goto out;
 		}
 	}
 
-	if (!ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb)))
-		ret = ocfs2_extend_no_holes(inode, new_i_size, new_i_size);
+	if (ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb)))
+		ret = ocfs2_zero_extend(inode, di_bh, new_i_size);
+	else
+		ret = ocfs2_extend_no_holes(inode, di_bh, new_i_size,
+					    new_i_size);
 
 	up_write(&oi->ip_alloc_sem);
 
@@ -943,9 +1113,8 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
 	struct ocfs2_super *osb = OCFS2_SB(sb);
 	struct buffer_head *bh = NULL;
 	handle_t *handle = NULL;
-	int qtype;
-	struct dquot *transfer_from[MAXQUOTAS] = { };
 	struct dquot *transfer_to[MAXQUOTAS] = { };
+	int qtype;
 
 	mlog_entry("(0x%p, '%.*s')\n", dentry,
 	           dentry->d_name.len, dentry->d_name.name);
@@ -976,6 +1145,8 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
 	if (status)
 		return status;
 
+	if (is_quota_modification(inode, attr))
+		dquot_initialize(inode);
 	size_change = S_ISREG(inode->i_mode) && attr->ia_valid & ATTR_SIZE;
 	if (size_change) {
 		status = ocfs2_rw_lock(inode, 1);
@@ -993,10 +1164,9 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
 	}
 
 	if (size_change && attr->ia_size != i_size_read(inode)) {
-		if (attr->ia_size > sb->s_maxbytes) {
-			status = -EFBIG;
+		status = inode_newsize_ok(inode, attr->ia_size);
+		if (status)
 			goto bail_unlock;
-		}
 
 		if (i_size_read(inode) > attr->ia_size) {
 			if (ocfs2_should_order_data(inode)) {
@@ -1021,16 +1191,14 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
 		/*
 		 * Gather pointers to quota structures so that allocation /
 		 * freeing of quota structures happens here and not inside
-		 * vfs_dq_transfer() where we have problems with lock ordering
+		 * dquot_transfer() where we have problems with lock ordering
 		 */
 		if (attr->ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid
 		    && OCFS2_HAS_RO_COMPAT_FEATURE(sb,
 		    OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
 			transfer_to[USRQUOTA] = dqget(sb, attr->ia_uid,
 						      USRQUOTA);
-			transfer_from[USRQUOTA] = dqget(sb, inode->i_uid,
-							USRQUOTA);
-			if (!transfer_to[USRQUOTA] || !transfer_from[USRQUOTA]) {
+			if (!transfer_to[USRQUOTA]) {
 				status = -ESRCH;
 				goto bail_unlock;
 			}
@@ -1040,9 +1208,7 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
 		    OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
 			transfer_to[GRPQUOTA] = dqget(sb, attr->ia_gid,
 						      GRPQUOTA);
-			transfer_from[GRPQUOTA] = dqget(sb, inode->i_gid,
-							GRPQUOTA);
-			if (!transfer_to[GRPQUOTA] || !transfer_from[GRPQUOTA]) {
+			if (!transfer_to[GRPQUOTA]) {
 				status = -ESRCH;
 				goto bail_unlock;
 			}
@@ -1054,7 +1220,7 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
 			mlog_errno(status);
 			goto bail_unlock;
 		}
-		status = vfs_dq_transfer(inode, attr) ? -EDQUOT : 0;
+		status = __dquot_transfer(inode, transfer_to);
 		if (status < 0)
 			goto bail_commit;
 	} else {
@@ -1067,7 +1233,7 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
 	}
 
 	/*
-	 * This will intentionally not wind up calling vmtruncate(),
+	 * This will intentionally not wind up calling simple_setsize(),
 	 * since all the work for a size change has been done above.
 	 * Otherwise, we could get into problems with truncate as
 	 * ip_alloc_sem is used there to protect against i_size
@@ -1094,10 +1260,8 @@ bail:
 	brelse(bh);
 
 	/* Release quota pointers in case we acquired them */
-	for (qtype = 0; qtype < MAXQUOTAS; qtype++) {
+	for (qtype = 0; qtype < MAXQUOTAS; qtype++)
 		dqput(transfer_to[qtype]);
-		dqput(transfer_from[qtype]);
-	}
 
 	if (!status && attr->ia_valid & ATTR_MODE) {
 		status = ocfs2_acl_chmod(inode);
@@ -1191,9 +1355,7 @@ static int __ocfs2_write_remove_suid(struct inode *inode,
 	di = (struct ocfs2_dinode *) bh->b_data;
 	di->i_mode = cpu_to_le16(inode->i_mode);
 
-	ret = ocfs2_journal_dirty(handle, bh);
-	if (ret < 0)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, bh);
 
 out_trans:
 	ocfs2_commit_trans(osb, handle);
@@ -1430,16 +1592,90 @@ out:
 	return ret;
 }
 
+static int ocfs2_find_rec(struct ocfs2_extent_list *el, u32 pos)
+{
+	int i;
+	struct ocfs2_extent_rec *rec = NULL;
+
+	for (i = le16_to_cpu(el->l_next_free_rec) - 1; i >= 0; i--) {
+
+		rec = &el->l_recs[i];
+
+		if (le32_to_cpu(rec->e_cpos) < pos)
+			break;
+	}
+
+	return i;
+}
+
+/*
+ * Helper to calculate the punching pos and length in one run, we handle the
+ * following three cases in order:
+ *
+ * - remove the entire record
+ * - remove a partial record
+ * - no record needs to be removed (hole-punching completed)
+*/
+static void ocfs2_calc_trunc_pos(struct inode *inode,
+				 struct ocfs2_extent_list *el,
+				 struct ocfs2_extent_rec *rec,
+				 u32 trunc_start, u32 *trunc_cpos,
+				 u32 *trunc_len, u32 *trunc_end,
+				 u64 *blkno, int *done)
+{
+	int ret = 0;
+	u32 coff, range;
+
+	range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec);
+
+	if (le32_to_cpu(rec->e_cpos) >= trunc_start) {
+		*trunc_cpos = le32_to_cpu(rec->e_cpos);
+		/*
+		 * Skip holes if any.
+		 */
+		if (range < *trunc_end)
+			*trunc_end = range;
+		*trunc_len = *trunc_end - le32_to_cpu(rec->e_cpos);
+		*blkno = le64_to_cpu(rec->e_blkno);
+		*trunc_end = le32_to_cpu(rec->e_cpos);
+	} else if (range > trunc_start) {
+		*trunc_cpos = trunc_start;
+		*trunc_len = *trunc_end - trunc_start;
+		coff = trunc_start - le32_to_cpu(rec->e_cpos);
+		*blkno = le64_to_cpu(rec->e_blkno) +
+				ocfs2_clusters_to_blocks(inode->i_sb, coff);
+		*trunc_end = trunc_start;
+	} else {
+		/*
+		 * It may have two following possibilities:
+		 *
+		 * - last record has been removed
+		 * - trunc_start was within a hole
+		 *
+		 * both two cases mean the completion of hole punching.
+		 */
+		ret = 1;
+	}
+
+	*done = ret;
+}
+
 static int ocfs2_remove_inode_range(struct inode *inode,
 				    struct buffer_head *di_bh, u64 byte_start,
 				    u64 byte_len)
 {
-	int ret = 0;
-	u32 trunc_start, trunc_len, cpos, phys_cpos, alloc_size;
+	int ret = 0, flags = 0, done = 0, i;
+	u32 trunc_start, trunc_len, trunc_end, trunc_cpos, phys_cpos;
+	u32 cluster_in_el;
 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 	struct ocfs2_cached_dealloc_ctxt dealloc;
 	struct address_space *mapping = inode->i_mapping;
 	struct ocfs2_extent_tree et;
+	struct ocfs2_path *path = NULL;
+	struct ocfs2_extent_list *el = NULL;
+	struct ocfs2_extent_rec *rec = NULL;
+	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
+	u64 blkno, refcount_loc = le64_to_cpu(di->i_refcount_loc);
 
 	ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), di_bh);
 	ocfs2_init_dealloc_ctxt(&dealloc);
@@ -1465,17 +1701,35 @@ static int ocfs2_remove_inode_range(struct inode *inode,
 		goto out;
 	}
 
+	/*
+	 * For reflinks, we may need to CoW 2 clusters which might be
+	 * partially zero'd later, if hole's start and end offset were
+	 * within one cluster(means is not exactly aligned to clustersize).
+	 */
+
+	if (OCFS2_I(inode)->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL) {
+
+		ret = ocfs2_cow_file_pos(inode, di_bh, byte_start);
+		if (ret) {
+			mlog_errno(ret);
+			goto out;
+		}
+
+		ret = ocfs2_cow_file_pos(inode, di_bh, byte_start + byte_len);
+		if (ret) {
+			mlog_errno(ret);
+			goto out;
+		}
+	}
+
 	trunc_start = ocfs2_clusters_for_bytes(osb->sb, byte_start);
-	trunc_len = (byte_start + byte_len) >> osb->s_clustersize_bits;
-	if (trunc_len >= trunc_start)
-		trunc_len -= trunc_start;
-	else
-		trunc_len = 0;
+	trunc_end = (byte_start + byte_len) >> osb->s_clustersize_bits;
+	cluster_in_el = trunc_end;
 
-	mlog(0, "Inode: %llu, start: %llu, len: %llu, cstart: %u, clen: %u\n",
+	mlog(0, "Inode: %llu, start: %llu, len: %llu, cstart: %u, cend: %u\n",
 	     (unsigned long long)OCFS2_I(inode)->ip_blkno,
 	     (unsigned long long)byte_start,
-	     (unsigned long long)byte_len, trunc_start, trunc_len);
+	     (unsigned long long)byte_len, trunc_start, trunc_end);
 
 	ret = ocfs2_zero_partial_clusters(inode, byte_start, byte_len);
 	if (ret) {
@@ -1483,31 +1737,79 @@ static int ocfs2_remove_inode_range(struct inode *inode,
 		goto out;
 	}
 
-	cpos = trunc_start;
-	while (trunc_len) {
-		ret = ocfs2_get_clusters(inode, cpos, &phys_cpos,
-					 &alloc_size, NULL);
+	path = ocfs2_new_path_from_et(&et);
+	if (!path) {
+		ret = -ENOMEM;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	while (trunc_end > trunc_start) {
+
+		ret = ocfs2_find_path(INODE_CACHE(inode), path,
+				      cluster_in_el);
 		if (ret) {
 			mlog_errno(ret);
 			goto out;
 		}
 
-		if (alloc_size > trunc_len)
-			alloc_size = trunc_len;
+		el = path_leaf_el(path);
 
-		/* Only do work for non-holes */
-		if (phys_cpos != 0) {
-			ret = ocfs2_remove_btree_range(inode, &et, cpos,
-						       phys_cpos, alloc_size,
-						       &dealloc);
+		i = ocfs2_find_rec(el, trunc_end);
+		/*
+		 * Need to go to previous extent block.
+		 */
+		if (i < 0) {
+			if (path->p_tree_depth == 0)
+				break;
+
+			ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb,
+							    path,
+							    &cluster_in_el);
 			if (ret) {
 				mlog_errno(ret);
 				goto out;
 			}
+
+			/*
+			 * We've reached the leftmost extent block,
+			 * it's safe to leave.
+			 */
+			if (cluster_in_el == 0)
+				break;
+
+			/*
+			 * The 'pos' searched for previous extent block is
+			 * always one cluster less than actual trunc_end.
+			 */
+			trunc_end = cluster_in_el + 1;
+
+			ocfs2_reinit_path(path, 1);
+
+			continue;
+
+		} else
+			rec = &el->l_recs[i];
+
+		ocfs2_calc_trunc_pos(inode, el, rec, trunc_start, &trunc_cpos,
+				     &trunc_len, &trunc_end, &blkno, &done);
+		if (done)
+			break;
+
+		flags = rec->e_flags;
+		phys_cpos = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
+
+		ret = ocfs2_remove_btree_range(inode, &et, trunc_cpos,
+					       phys_cpos, trunc_len, flags,
+					       &dealloc, refcount_loc);
+		if (ret < 0) {
+			mlog_errno(ret);
+			goto out;
 		}
 
-		cpos += alloc_size;
-		trunc_len -= alloc_size;
+		cluster_in_el = trunc_end;
+
+		ocfs2_reinit_path(path, 1);
 	}
 
 	ocfs2_truncate_cluster_pages(inode, byte_start, byte_len);
@@ -1779,7 +2081,7 @@ static int ocfs2_prepare_inode_for_write(struct dentry *dentry,
 	struct inode *inode = dentry->d_inode;
 	loff_t saved_pos, end;
 
-	/* 
+	/*
 	 * We start with a read level meta lock and only jump to an ex
 	 * if we need to make modifications here.
 	 */
@@ -1836,6 +2138,8 @@ static int ocfs2_prepare_inode_for_write(struct dentry *dentry,
 							       &meta_level);
 			if (has_refcount)
 				*has_refcount = 1;
+			if (direct_io)
+				*direct_io = 0;
 		}
 
 		if (ret < 0) {
@@ -1859,10 +2163,6 @@ static int ocfs2_prepare_inode_for_write(struct dentry *dentry,
 			break;
 		}
 
-		if (has_refcount && *has_refcount == 1) {
-			*direct_io = 0;
-			break;
-		}
 		/*
 		 * Allowing concurrent direct writes means
 		 * i_size changes wouldn't be synchronized, so
@@ -1980,18 +2280,18 @@ relock:
 	/* communicate with ocfs2_dio_end_io */
 	ocfs2_iocb_set_rw_locked(iocb, rw_level);
 
-	if (direct_io) {
-		ret = generic_segment_checks(iov, &nr_segs, &ocount,
-					     VERIFY_READ);
-		if (ret)
-			goto out_dio;
+	ret = generic_segment_checks(iov, &nr_segs, &ocount,
+				     VERIFY_READ);
+	if (ret)
+		goto out_dio;
 
-		count = ocount;
-		ret = generic_write_checks(file, ppos, &count,
-					   S_ISBLK(inode->i_mode));
-		if (ret)
-			goto out_dio;
+	count = ocount;
+	ret = generic_write_checks(file, ppos, &count,
+				   S_ISBLK(inode->i_mode));
+	if (ret)
+		goto out_dio;
 
+	if (direct_io) {
 		written = generic_file_direct_write(iocb, iov, &nr_segs, *ppos,
 						    ppos, count, ocount);
 		if (written < 0) {
@@ -1999,30 +2299,37 @@ relock:
 			 * direct write may have instantiated a few
 			 * blocks outside i_size. Trim these off again.
 			 * Don't need i_size_read because we hold i_mutex.
+			 *
+			 * XXX(hch): this looks buggy because ocfs2 did not
+			 * actually implement ->truncate.  Take a look at
+			 * the new truncate sequence and update this accordingly
 			 */
 			if (*ppos + count > inode->i_size)
-				vmtruncate(inode, inode->i_size);
+				simple_setsize(inode, inode->i_size);
 			ret = written;
 			goto out_dio;
 		}
 	} else {
-		written = __generic_file_aio_write(iocb, iov, nr_segs, ppos);
+		current->backing_dev_info = file->f_mapping->backing_dev_info;
+		written = generic_file_buffered_write(iocb, iov, nr_segs, *ppos,
+						      ppos, count, 0);
+		current->backing_dev_info = NULL;
 	}
 
 out_dio:
 	/* buffered aio wouldn't have proper lock coverage today */
 	BUG_ON(ret == -EIOCBQUEUED && !(file->f_flags & O_DIRECT));
 
-	if ((file->f_flags & O_DSYNC && !direct_io) || IS_SYNC(inode) ||
-	    (file->f_flags & O_DIRECT && has_refcount)) {
+	if (((file->f_flags & O_DSYNC) && !direct_io) || IS_SYNC(inode) ||
+	    ((file->f_flags & O_DIRECT) && has_refcount)) {
 		ret = filemap_fdatawrite_range(file->f_mapping, pos,
 					       pos + count - 1);
 		if (ret < 0)
 			written = ret;
 
-		if (!ret && (old_size != i_size_read(inode) ||
-		    old_clusters != OCFS2_I(inode)->ip_clusters ||
-		    has_refcount)) {
+		if (!ret && ((old_size != i_size_read(inode)) ||
+			     (old_clusters != OCFS2_I(inode)->ip_clusters) ||
+			     has_refcount)) {
 			ret = jbd2_journal_force_commit(osb->journal->j_journal);
 			if (ret < 0)
 				written = ret;
@@ -2033,7 +2340,7 @@ out_dio:
 						      pos + count - 1);
 	}
 
-	/* 
+	/*
 	 * deep in g_f_a_w_n()->ocfs2_direct_IO we pass in a ocfs2_dio_end_io
 	 * function pointer which is called when o_direct io completes so that
 	 * it can unlock our rw lock.  (it's the clustered equivalent of
@@ -2043,7 +2350,7 @@ out_dio:
 	 * async dio is going to do it in the future or an end_io after an
 	 * error has already done it.
 	 */
-	if (ret == -EIOCBQUEUED || !ocfs2_iocb_is_rw_locked(iocb)) {
+	if ((ret == -EIOCBQUEUED) || (!ocfs2_iocb_is_rw_locked(iocb))) {
 		rw_level = -1;
 		have_alloc_sem = 0;
 	}
@@ -2198,7 +2505,7 @@ static ssize_t ocfs2_file_aio_read(struct kiocb *iocb,
 		goto bail;
 	}
 
-	/* 
+	/*
 	 * buffered reads protect themselves in ->readpage().  O_DIRECT reads
 	 * need locks to protect pending reads from racing with truncate.
 	 */
@@ -2220,10 +2527,10 @@ static ssize_t ocfs2_file_aio_read(struct kiocb *iocb,
 	 * We're fine letting folks race truncates and extending
 	 * writes with read across the cluster, just like they can
 	 * locally. Hence no rw_lock during read.
-	 * 
+	 *
 	 * Take and drop the meta data lock to update inode fields
 	 * like i_size. This allows the checks down below
-	 * generic_file_aio_read() a chance of actually working. 
+	 * generic_file_aio_read() a chance of actually working.
 	 */
 	ret = ocfs2_inode_lock_atime(inode, filp->f_vfsmnt, &lock_level);
 	if (ret < 0) {
@@ -2248,7 +2555,7 @@ static ssize_t ocfs2_file_aio_read(struct kiocb *iocb,
 bail:
 	if (have_alloc_sem)
 		up_read(&inode->i_alloc_sem);
-	if (rw_level != -1) 
+	if (rw_level != -1)
 		ocfs2_rw_unlock(inode, rw_level);
 	mlog_exit(ret);
 
diff --git a/fs/ocfs2/file.h b/fs/ocfs2/file.h
index d66cf4f7c70e..97bf761c9e7c 100644
--- a/fs/ocfs2/file.h
+++ b/fs/ocfs2/file.h
@@ -54,8 +54,10 @@ int ocfs2_add_inode_data(struct ocfs2_super *osb,
 int ocfs2_simple_size_update(struct inode *inode,
 			     struct buffer_head *di_bh,
 			     u64 new_i_size);
-int ocfs2_extend_no_holes(struct inode *inode, u64 new_i_size,
-			  u64 zero_to);
+int ocfs2_extend_no_holes(struct inode *inode, struct buffer_head *di_bh,
+			  u64 new_i_size, u64 zero_to);
+int ocfs2_zero_extend(struct inode *inode, struct buffer_head *di_bh,
+		      loff_t zero_to);
 int ocfs2_setattr(struct dentry *dentry, struct iattr *attr);
 int ocfs2_getattr(struct vfsmount *mnt, struct dentry *dentry,
 		  struct kstat *stat);
diff --git a/fs/ocfs2/heartbeat.c b/fs/ocfs2/heartbeat.c
index c6e7213db868..1aa863dd901f 100644
--- a/fs/ocfs2/heartbeat.c
+++ b/fs/ocfs2/heartbeat.c
@@ -26,7 +26,6 @@
 
 #include <linux/fs.h>
 #include <linux/types.h>
-#include <linux/slab.h>
 #include <linux/highmem.h>
 
 #define MLOG_MASK_PREFIX ML_SUPER
diff --git a/fs/ocfs2/inode.c b/fs/ocfs2/inode.c
index 0297fb8982b8..abb0a95cc717 100644
--- a/fs/ocfs2/inode.c
+++ b/fs/ocfs2/inode.c
@@ -25,7 +25,6 @@
 
 #include <linux/fs.h>
 #include <linux/types.h>
-#include <linux/slab.h>
 #include <linux/highmem.h>
 #include <linux/pagemap.h>
 #include <linux/quotaops.h>
@@ -377,6 +376,10 @@ void ocfs2_populate_inode(struct inode *inode, struct ocfs2_dinode *fe,
 
 	OCFS2_I(inode)->ip_last_used_slot = 0;
 	OCFS2_I(inode)->ip_last_used_group = 0;
+
+	if (S_ISDIR(inode->i_mode))
+		ocfs2_resv_set_type(&OCFS2_I(inode)->ip_la_data_resv,
+				    OCFS2_RESV_FLAG_DIR);
 	mlog_exit_void();
 }
 
@@ -475,7 +478,7 @@ static int ocfs2_read_locked_inode(struct inode *inode,
 	if (args->fi_flags & OCFS2_FI_FLAG_ORPHAN_RECOVERY) {
 		status = ocfs2_try_open_lock(inode, 0);
 		if (status) {
-			make_bad_inode(inode);	
+			make_bad_inode(inode);
 			return status;
 		}
 	}
@@ -540,7 +543,6 @@ static int ocfs2_truncate_for_delete(struct ocfs2_super *osb,
 				     struct buffer_head *fe_bh)
 {
 	int status = 0;
-	struct ocfs2_truncate_context *tc = NULL;
 	struct ocfs2_dinode *fe;
 	handle_t *handle = NULL;
 
@@ -559,6 +561,7 @@ static int ocfs2_truncate_for_delete(struct ocfs2_super *osb,
 		handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
 		if (IS_ERR(handle)) {
 			status = PTR_ERR(handle);
+			handle = NULL;
 			mlog_errno(status);
 			goto out;
 		}
@@ -582,13 +585,7 @@ static int ocfs2_truncate_for_delete(struct ocfs2_super *osb,
 		ocfs2_commit_trans(osb, handle);
 		handle = NULL;
 
-		status = ocfs2_prepare_truncate(osb, inode, fe_bh, &tc);
-		if (status < 0) {
-			mlog_errno(status);
-			goto out;
-		}
-
-		status = ocfs2_commit_truncate(osb, inode, fe_bh, tc);
+		status = ocfs2_commit_truncate(osb, inode, fe_bh);
 		if (status < 0) {
 			mlog_errno(status);
 			goto out;
@@ -640,11 +637,13 @@ static int ocfs2_remove_inode(struct inode *inode,
 		goto bail_unlock;
 	}
 
-	status = ocfs2_orphan_del(osb, handle, orphan_dir_inode, inode,
-				  orphan_dir_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail_commit;
+	if (!(OCFS2_I(inode)->ip_flags & OCFS2_INODE_SKIP_ORPHAN_DIR)) {
+		status = ocfs2_orphan_del(osb, handle, orphan_dir_inode, inode,
+					  orphan_dir_bh);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail_commit;
+		}
 	}
 
 	/* set the inodes dtime */
@@ -657,15 +656,10 @@ static int ocfs2_remove_inode(struct inode *inode,
 
 	di->i_dtime = cpu_to_le64(CURRENT_TIME.tv_sec);
 	di->i_flags &= cpu_to_le32(~(OCFS2_VALID_FL | OCFS2_ORPHANED_FL));
-
-	status = ocfs2_journal_dirty(handle, di_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail_commit;
-	}
+	ocfs2_journal_dirty(handle, di_bh);
 
 	ocfs2_remove_from_cache(INODE_CACHE(inode), di_bh);
-	vfs_dq_free_inode(inode);
+	dquot_free_inode(inode);
 
 	status = ocfs2_free_dinode(handle, inode_alloc_inode,
 				   inode_alloc_bh, di);
@@ -684,7 +678,7 @@ bail:
 	return status;
 }
 
-/* 
+/*
  * Serialize with orphan dir recovery. If the process doing
  * recovery on this orphan dir does an iget() with the dir
  * i_mutex held, we'll deadlock here. Instead we detect this
@@ -723,38 +717,39 @@ static void ocfs2_signal_wipe_completion(struct ocfs2_super *osb,
 static int ocfs2_wipe_inode(struct inode *inode,
 			    struct buffer_head *di_bh)
 {
-	int status, orphaned_slot;
+	int status, orphaned_slot = -1;
 	struct inode *orphan_dir_inode = NULL;
 	struct buffer_head *orphan_dir_bh = NULL;
 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
-	struct ocfs2_dinode *di;
+	struct ocfs2_dinode *di = (struct ocfs2_dinode *) di_bh->b_data;
 
-	di = (struct ocfs2_dinode *) di_bh->b_data;
-	orphaned_slot = le16_to_cpu(di->i_orphaned_slot);
+	if (!(OCFS2_I(inode)->ip_flags & OCFS2_INODE_SKIP_ORPHAN_DIR)) {
+		orphaned_slot = le16_to_cpu(di->i_orphaned_slot);
 
-	status = ocfs2_check_orphan_recovery_state(osb, orphaned_slot);
-	if (status)
-		return status;
+		status = ocfs2_check_orphan_recovery_state(osb, orphaned_slot);
+		if (status)
+			return status;
 
-	orphan_dir_inode = ocfs2_get_system_file_inode(osb,
-						       ORPHAN_DIR_SYSTEM_INODE,
-						       orphaned_slot);
-	if (!orphan_dir_inode) {
-		status = -EEXIST;
-		mlog_errno(status);
-		goto bail;
-	}
+		orphan_dir_inode = ocfs2_get_system_file_inode(osb,
+							       ORPHAN_DIR_SYSTEM_INODE,
+							       orphaned_slot);
+		if (!orphan_dir_inode) {
+			status = -EEXIST;
+			mlog_errno(status);
+			goto bail;
+		}
 
-	/* Lock the orphan dir. The lock will be held for the entire
-	 * delete_inode operation. We do this now to avoid races with
-	 * recovery completion on other nodes. */
-	mutex_lock(&orphan_dir_inode->i_mutex);
-	status = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1);
-	if (status < 0) {
-		mutex_unlock(&orphan_dir_inode->i_mutex);
+		/* Lock the orphan dir. The lock will be held for the entire
+		 * delete_inode operation. We do this now to avoid races with
+		 * recovery completion on other nodes. */
+		mutex_lock(&orphan_dir_inode->i_mutex);
+		status = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1);
+		if (status < 0) {
+			mutex_unlock(&orphan_dir_inode->i_mutex);
 
-		mlog_errno(status);
-		goto bail;
+			mlog_errno(status);
+			goto bail;
+		}
 	}
 
 	/* we do this while holding the orphan dir lock because we
@@ -795,6 +790,9 @@ static int ocfs2_wipe_inode(struct inode *inode,
 		mlog_errno(status);
 
 bail_unlock_dir:
+	if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SKIP_ORPHAN_DIR)
+		return status;
+
 	ocfs2_inode_unlock(orphan_dir_inode, 1);
 	mutex_unlock(&orphan_dir_inode->i_mutex);
 	brelse(orphan_dir_bh);
@@ -890,7 +888,23 @@ static int ocfs2_query_inode_wipe(struct inode *inode,
 
 	/* Do some basic inode verification... */
 	di = (struct ocfs2_dinode *) di_bh->b_data;
-	if (!(di->i_flags & cpu_to_le32(OCFS2_ORPHANED_FL))) {
+	if (!(di->i_flags & cpu_to_le32(OCFS2_ORPHANED_FL)) &&
+	    !(oi->ip_flags & OCFS2_INODE_SKIP_ORPHAN_DIR)) {
+		/*
+		 * Inodes in the orphan dir must have ORPHANED_FL.  The only
+		 * inodes that come back out of the orphan dir are reflink
+		 * targets. A reflink target may be moved out of the orphan
+		 * dir between the time we scan the directory and the time we
+		 * process it. This would lead to HAS_REFCOUNT_FL being set but
+		 * ORPHANED_FL not.
+		 */
+		if (di->i_dyn_features & cpu_to_le16(OCFS2_HAS_REFCOUNT_FL)) {
+			mlog(0, "Reflinked inode %llu is no longer orphaned.  "
+			     "it shouldn't be deleted\n",
+			     (unsigned long long)oi->ip_blkno);
+			goto bail;
+		}
+
 		/* for lack of a better error? */
 		status = -EEXIST;
 		mlog(ML_ERROR,
@@ -958,7 +972,7 @@ static void ocfs2_cleanup_delete_inode(struct inode *inode,
 void ocfs2_delete_inode(struct inode *inode)
 {
 	int wipe, status;
-	sigset_t blocked, oldset;
+	sigset_t oldset;
 	struct buffer_head *di_bh = NULL;
 
 	mlog_entry("(inode->i_ino = %lu)\n", inode->i_ino);
@@ -971,6 +985,8 @@ void ocfs2_delete_inode(struct inode *inode)
 		goto bail;
 	}
 
+	dquot_initialize(inode);
+
 	if (!ocfs2_inode_is_valid_to_delete(inode)) {
 		/* It's probably not necessary to truncate_inode_pages
 		 * here but we do it for safety anyway (it will most
@@ -983,13 +999,7 @@ void ocfs2_delete_inode(struct inode *inode)
 	 * messaging paths may return us -ERESTARTSYS. Which would
 	 * cause us to exit early, resulting in inodes being orphaned
 	 * forever. */
-	sigfillset(&blocked);
-	status = sigprocmask(SIG_BLOCK, &blocked, &oldset);
-	if (status < 0) {
-		mlog_errno(status);
-		ocfs2_cleanup_delete_inode(inode, 1);
-		goto bail;
-	}
+	ocfs2_block_signals(&oldset);
 
 	/*
 	 * Synchronize us against ocfs2_get_dentry. We take this in
@@ -1063,9 +1073,7 @@ bail_unlock_nfs_sync:
 	ocfs2_nfs_sync_unlock(OCFS2_SB(inode->i_sb), 0);
 
 bail_unblock:
-	status = sigprocmask(SIG_SETMASK, &oldset, NULL);
-	if (status < 0)
-		mlog_errno(status);
+	ocfs2_unblock_signals(&oldset);
 bail:
 	clear_inode(inode);
 	mlog_exit_void();
@@ -1087,6 +1095,8 @@ void ocfs2_clear_inode(struct inode *inode)
 	mlog_bug_on_msg(OCFS2_SB(inode->i_sb) == NULL,
 			"Inode=%lu\n", inode->i_ino);
 
+	dquot_drop(inode);
+
 	/* To preven remote deletes we hold open lock before, now it
 	 * is time to unlock PR and EX open locks. */
 	ocfs2_open_unlock(inode);
@@ -1097,6 +1107,10 @@ void ocfs2_clear_inode(struct inode *inode)
 	ocfs2_mark_lockres_freeing(&oi->ip_inode_lockres);
 	ocfs2_mark_lockres_freeing(&oi->ip_open_lockres);
 
+	ocfs2_resv_discard(&OCFS2_SB(inode->i_sb)->osb_la_resmap,
+			   &oi->ip_la_data_resv);
+	ocfs2_resv_init_once(&oi->ip_la_data_resv);
+
 	/* We very well may get a clear_inode before all an inodes
 	 * metadata has hit disk. Of course, we can't drop any cluster
 	 * locks until the journal has finished with it. The only
@@ -1272,13 +1286,8 @@ int ocfs2_mark_inode_dirty(handle_t *handle,
 	fe->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec);
 	fe->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
 
-	status = ocfs2_journal_dirty(handle, bh);
-	if (status < 0)
-		mlog_errno(status);
-
-	status = 0;
+	ocfs2_journal_dirty(handle, bh);
 leave:
-
 	mlog_exit(status);
 	return status;
 }
diff --git a/fs/ocfs2/inode.h b/fs/ocfs2/inode.h
index ba4fe07b293c..9f5f5fcadc45 100644
--- a/fs/ocfs2/inode.h
+++ b/fs/ocfs2/inode.h
@@ -70,6 +70,8 @@ struct ocfs2_inode_info
 	/* Only valid if the inode is the dir. */
 	u32				ip_last_used_slot;
 	u64				ip_last_used_group;
+
+	struct ocfs2_alloc_reservation	ip_la_data_resv;
 };
 
 /*
@@ -100,6 +102,8 @@ struct ocfs2_inode_info
 #define OCFS2_INODE_MAYBE_ORPHANED	0x00000020
 /* Does someone have the file open O_DIRECT */
 #define OCFS2_INODE_OPEN_DIRECT		0x00000040
+/* Tell the inode wipe code it's not in orphan dir */
+#define OCFS2_INODE_SKIP_ORPHAN_DIR     0x00000080
 
 static inline struct ocfs2_inode_info *OCFS2_I(struct inode *inode)
 {
diff --git a/fs/ocfs2/ioctl.c b/fs/ocfs2/ioctl.c
index 31fbb0619510..7d9d9c132cef 100644
--- a/fs/ocfs2/ioctl.c
+++ b/fs/ocfs2/ioctl.c
@@ -7,6 +7,7 @@
 
 #include <linux/fs.h>
 #include <linux/mount.h>
+#include <linux/compat.h>
 
 #define MLOG_MASK_PREFIX ML_INODE
 #include <cluster/masklog.h>
@@ -181,6 +182,10 @@ long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 #ifdef CONFIG_COMPAT
 long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
 {
+	bool preserve;
+	struct reflink_arguments args;
+	struct inode *inode = file->f_path.dentry->d_inode;
+
 	switch (cmd) {
 	case OCFS2_IOC32_GETFLAGS:
 		cmd = OCFS2_IOC_GETFLAGS;
@@ -195,8 +200,15 @@ long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
 	case OCFS2_IOC_GROUP_EXTEND:
 	case OCFS2_IOC_GROUP_ADD:
 	case OCFS2_IOC_GROUP_ADD64:
-	case OCFS2_IOC_REFLINK:
 		break;
+	case OCFS2_IOC_REFLINK:
+		if (copy_from_user(&args, (struct reflink_arguments *)arg,
+				   sizeof(args)))
+			return -EFAULT;
+		preserve = (args.preserve != 0);
+
+		return ocfs2_reflink_ioctl(inode, compat_ptr(args.old_path),
+					   compat_ptr(args.new_path), preserve);
 	default:
 		return -ENOIOCTLCMD;
 	}
diff --git a/fs/ocfs2/ioctl.h b/fs/ocfs2/ioctl.h
index cf9a5ee30fef..0cd5323bd3f0 100644
--- a/fs/ocfs2/ioctl.h
+++ b/fs/ocfs2/ioctl.h
@@ -7,10 +7,10 @@
  *
  */
 
-#ifndef OCFS2_IOCTL_H
-#define OCFS2_IOCTL_H
+#ifndef OCFS2_IOCTL_PROTO_H
+#define OCFS2_IOCTL_PROTO_H
 
 long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg);
 long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg);
 
-#endif /* OCFS2_IOCTL_H */
+#endif /* OCFS2_IOCTL_PROTO_H */
diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c
index bf34c491ae96..625de9d7088c 100644
--- a/fs/ocfs2/journal.c
+++ b/fs/ocfs2/journal.c
@@ -402,9 +402,7 @@ int ocfs2_commit_trans(struct ocfs2_super *osb,
 }
 
 /*
- * 'nblocks' is what you want to add to the current
- * transaction. extend_trans will either extend the current handle by
- * nblocks, or commit it and start a new one with nblocks credits.
+ * 'nblocks' is what you want to add to the current transaction.
  *
  * This might call jbd2_journal_restart() which will commit dirty buffers
  * and then restart the transaction. Before calling
@@ -422,11 +420,15 @@ int ocfs2_commit_trans(struct ocfs2_super *osb,
  */
 int ocfs2_extend_trans(handle_t *handle, int nblocks)
 {
-	int status;
+	int status, old_nblocks;
 
 	BUG_ON(!handle);
-	BUG_ON(!nblocks);
+	BUG_ON(nblocks < 0);
+
+	if (!nblocks)
+		return 0;
 
+	old_nblocks = handle->h_buffer_credits;
 	mlog_entry_void();
 
 	mlog(0, "Trying to extend transaction by %d blocks\n", nblocks);
@@ -445,7 +447,8 @@ int ocfs2_extend_trans(handle_t *handle, int nblocks)
 		mlog(0,
 		     "jbd2_journal_extend failed, trying "
 		     "jbd2_journal_restart\n");
-		status = jbd2_journal_restart(handle, nblocks);
+		status = jbd2_journal_restart(handle,
+					      old_nblocks + nblocks);
 		if (status < 0) {
 			mlog_errno(status);
 			goto bail;
@@ -469,7 +472,7 @@ static inline struct ocfs2_triggers *to_ocfs2_trigger(struct jbd2_buffer_trigger
 	return container_of(triggers, struct ocfs2_triggers, ot_triggers);
 }
 
-static void ocfs2_commit_trigger(struct jbd2_buffer_trigger_type *triggers,
+static void ocfs2_frozen_trigger(struct jbd2_buffer_trigger_type *triggers,
 				 struct buffer_head *bh,
 				 void *data, size_t size)
 {
@@ -488,7 +491,7 @@ static void ocfs2_commit_trigger(struct jbd2_buffer_trigger_type *triggers,
  * Quota blocks have their own trigger because the struct ocfs2_block_check
  * offset depends on the blocksize.
  */
-static void ocfs2_dq_commit_trigger(struct jbd2_buffer_trigger_type *triggers,
+static void ocfs2_dq_frozen_trigger(struct jbd2_buffer_trigger_type *triggers,
 				 struct buffer_head *bh,
 				 void *data, size_t size)
 {
@@ -508,7 +511,7 @@ static void ocfs2_dq_commit_trigger(struct jbd2_buffer_trigger_type *triggers,
  * Directory blocks also have their own trigger because the
  * struct ocfs2_block_check offset depends on the blocksize.
  */
-static void ocfs2_db_commit_trigger(struct jbd2_buffer_trigger_type *triggers,
+static void ocfs2_db_frozen_trigger(struct jbd2_buffer_trigger_type *triggers,
 				 struct buffer_head *bh,
 				 void *data, size_t size)
 {
@@ -541,7 +544,7 @@ static void ocfs2_abort_trigger(struct jbd2_buffer_trigger_type *triggers,
 
 static struct ocfs2_triggers di_triggers = {
 	.ot_triggers = {
-		.t_commit = ocfs2_commit_trigger,
+		.t_frozen = ocfs2_frozen_trigger,
 		.t_abort = ocfs2_abort_trigger,
 	},
 	.ot_offset	= offsetof(struct ocfs2_dinode, i_check),
@@ -549,7 +552,7 @@ static struct ocfs2_triggers di_triggers = {
 
 static struct ocfs2_triggers eb_triggers = {
 	.ot_triggers = {
-		.t_commit = ocfs2_commit_trigger,
+		.t_frozen = ocfs2_frozen_trigger,
 		.t_abort = ocfs2_abort_trigger,
 	},
 	.ot_offset	= offsetof(struct ocfs2_extent_block, h_check),
@@ -557,7 +560,7 @@ static struct ocfs2_triggers eb_triggers = {
 
 static struct ocfs2_triggers rb_triggers = {
 	.ot_triggers = {
-		.t_commit = ocfs2_commit_trigger,
+		.t_frozen = ocfs2_frozen_trigger,
 		.t_abort = ocfs2_abort_trigger,
 	},
 	.ot_offset	= offsetof(struct ocfs2_refcount_block, rf_check),
@@ -565,7 +568,7 @@ static struct ocfs2_triggers rb_triggers = {
 
 static struct ocfs2_triggers gd_triggers = {
 	.ot_triggers = {
-		.t_commit = ocfs2_commit_trigger,
+		.t_frozen = ocfs2_frozen_trigger,
 		.t_abort = ocfs2_abort_trigger,
 	},
 	.ot_offset	= offsetof(struct ocfs2_group_desc, bg_check),
@@ -573,14 +576,14 @@ static struct ocfs2_triggers gd_triggers = {
 
 static struct ocfs2_triggers db_triggers = {
 	.ot_triggers = {
-		.t_commit = ocfs2_db_commit_trigger,
+		.t_frozen = ocfs2_db_frozen_trigger,
 		.t_abort = ocfs2_abort_trigger,
 	},
 };
 
 static struct ocfs2_triggers xb_triggers = {
 	.ot_triggers = {
-		.t_commit = ocfs2_commit_trigger,
+		.t_frozen = ocfs2_frozen_trigger,
 		.t_abort = ocfs2_abort_trigger,
 	},
 	.ot_offset	= offsetof(struct ocfs2_xattr_block, xb_check),
@@ -588,14 +591,14 @@ static struct ocfs2_triggers xb_triggers = {
 
 static struct ocfs2_triggers dq_triggers = {
 	.ot_triggers = {
-		.t_commit = ocfs2_dq_commit_trigger,
+		.t_frozen = ocfs2_dq_frozen_trigger,
 		.t_abort = ocfs2_abort_trigger,
 	},
 };
 
 static struct ocfs2_triggers dr_triggers = {
 	.ot_triggers = {
-		.t_commit = ocfs2_commit_trigger,
+		.t_frozen = ocfs2_frozen_trigger,
 		.t_abort = ocfs2_abort_trigger,
 	},
 	.ot_offset	= offsetof(struct ocfs2_dx_root_block, dr_check),
@@ -603,7 +606,7 @@ static struct ocfs2_triggers dr_triggers = {
 
 static struct ocfs2_triggers dl_triggers = {
 	.ot_triggers = {
-		.t_commit = ocfs2_commit_trigger,
+		.t_frozen = ocfs2_frozen_trigger,
 		.t_abort = ocfs2_abort_trigger,
 	},
 	.ot_offset	= offsetof(struct ocfs2_dx_leaf, dl_check),
@@ -734,8 +737,7 @@ int ocfs2_journal_access(handle_t *handle, struct ocfs2_caching_info *ci,
 	return __ocfs2_journal_access(handle, ci, bh, NULL, type);
 }
 
-int ocfs2_journal_dirty(handle_t *handle,
-			struct buffer_head *bh)
+void ocfs2_journal_dirty(handle_t *handle, struct buffer_head *bh)
 {
 	int status;
 
@@ -743,13 +745,9 @@ int ocfs2_journal_dirty(handle_t *handle,
 		   (unsigned long long)bh->b_blocknr);
 
 	status = jbd2_journal_dirty_metadata(handle, bh);
-	if (status < 0)
-		mlog(ML_ERROR, "Could not dirty metadata buffer. "
-		     "(bh->b_blocknr=%llu)\n",
-		     (unsigned long long)bh->b_blocknr);
+	BUG_ON(status);
 
-	mlog_exit(status);
-	return status;
+	mlog_exit_void();
 }
 
 #define OCFS2_DEFAULT_COMMIT_INTERVAL	(HZ * JBD2_DEFAULT_MAX_COMMIT_AGE)
@@ -1938,7 +1936,7 @@ void ocfs2_orphan_scan_work(struct work_struct *work)
 	mutex_lock(&os->os_lock);
 	ocfs2_queue_orphan_scan(osb);
 	if (atomic_read(&os->os_state) == ORPHAN_SCAN_ACTIVE)
-		schedule_delayed_work(&os->os_orphan_scan_work,
+		queue_delayed_work(ocfs2_wq, &os->os_orphan_scan_work,
 				      ocfs2_orphan_scan_timeout());
 	mutex_unlock(&os->os_lock);
 }
@@ -1978,8 +1976,8 @@ void ocfs2_orphan_scan_start(struct ocfs2_super *osb)
 		atomic_set(&os->os_state, ORPHAN_SCAN_INACTIVE);
 	else {
 		atomic_set(&os->os_state, ORPHAN_SCAN_ACTIVE);
-		schedule_delayed_work(&os->os_orphan_scan_work,
-				      ocfs2_orphan_scan_timeout());
+		queue_delayed_work(ocfs2_wq, &os->os_orphan_scan_work,
+				   ocfs2_orphan_scan_timeout());
 	}
 }
 
@@ -2034,7 +2032,7 @@ static int ocfs2_queue_orphans(struct ocfs2_super *osb,
 		status = -ENOENT;
 		mlog_errno(status);
 		return status;
-	}	
+	}
 
 	mutex_lock(&orphan_dir_inode->i_mutex);
 	status = ocfs2_inode_lock(orphan_dir_inode, NULL, 0);
diff --git a/fs/ocfs2/journal.h b/fs/ocfs2/journal.h
index 3f74e09b0d80..b5baaa8e710f 100644
--- a/fs/ocfs2/journal.h
+++ b/fs/ocfs2/journal.h
@@ -325,8 +325,7 @@ int ocfs2_journal_access(handle_t *handle, struct ocfs2_caching_info *ci,
  *	<modify the bh>
  * 	ocfs2_journal_dirty(handle, bh);
  */
-int                  ocfs2_journal_dirty(handle_t *handle,
-					 struct buffer_head *bh);
+void ocfs2_journal_dirty(handle_t *handle, struct buffer_head *bh);
 
 /*
  *  Credit Macros:
@@ -562,6 +561,18 @@ static inline int ocfs2_calc_group_alloc_credits(struct super_block *sb,
 	return blocks;
 }
 
+/*
+ * Allocating a discontiguous block group requires the credits from
+ * ocfs2_calc_group_alloc_credits() as well as enough credits to fill
+ * the group descriptor's extent list.  The caller already has started
+ * the transaction with ocfs2_calc_group_alloc_credits().  They extend
+ * it with these credits.
+ */
+static inline int ocfs2_calc_bg_discontig_credits(struct super_block *sb)
+{
+	return ocfs2_extent_recs_per_gd(sb);
+}
+
 static inline int ocfs2_calc_tree_trunc_credits(struct super_block *sb,
 						unsigned int clusters_to_del,
 						struct ocfs2_dinode *fe,
diff --git a/fs/ocfs2/localalloc.c b/fs/ocfs2/localalloc.c
index ac10f83edb95..ec6adbf8f551 100644
--- a/fs/ocfs2/localalloc.c
+++ b/fs/ocfs2/localalloc.c
@@ -52,7 +52,8 @@ static u32 ocfs2_local_alloc_count_bits(struct ocfs2_dinode *alloc);
 
 static int ocfs2_local_alloc_find_clear_bits(struct ocfs2_super *osb,
 					     struct ocfs2_dinode *alloc,
-					     u32 numbits);
+					     u32 *numbits,
+					     struct ocfs2_alloc_reservation *resv);
 
 static void ocfs2_clear_local_alloc(struct ocfs2_dinode *alloc);
 
@@ -74,6 +75,151 @@ static int ocfs2_local_alloc_new_window(struct ocfs2_super *osb,
 static int ocfs2_local_alloc_slide_window(struct ocfs2_super *osb,
 					  struct inode *local_alloc_inode);
 
+/*
+ * ocfs2_la_default_mb() - determine a default size, in megabytes of
+ * the local alloc.
+ *
+ * Generally, we'd like to pick as large a local alloc as
+ * possible. Performance on large workloads tends to scale
+ * proportionally to la size. In addition to that, the reservations
+ * code functions more efficiently as it can reserve more windows for
+ * write.
+ *
+ * Some things work against us when trying to choose a large local alloc:
+ *
+ * - We need to ensure our sizing is picked to leave enough space in
+ *   group descriptors for other allocations (such as block groups,
+ *   etc). Picking default sizes which are a multiple of 4 could help
+ *   - block groups are allocated in 2mb and 4mb chunks.
+ *
+ * - Likewise, we don't want to starve other nodes of bits on small
+ *   file systems. This can easily be taken care of by limiting our
+ *   default to a reasonable size (256M) on larger cluster sizes.
+ *
+ * - Some file systems can't support very large sizes - 4k and 8k in
+ *   particular are limited to less than 128 and 256 megabytes respectively.
+ *
+ * The following reference table shows group descriptor and local
+ * alloc maximums at various cluster sizes (4k blocksize)
+ *
+ * csize: 4K	group: 126M	la: 121M
+ * csize: 8K	group: 252M	la: 243M
+ * csize: 16K	group: 504M	la: 486M
+ * csize: 32K	group: 1008M	la: 972M
+ * csize: 64K	group: 2016M	la: 1944M
+ * csize: 128K	group: 4032M	la: 3888M
+ * csize: 256K	group: 8064M	la: 7776M
+ * csize: 512K	group: 16128M	la: 15552M
+ * csize: 1024K	group: 32256M	la: 31104M
+ */
+#define	OCFS2_LA_MAX_DEFAULT_MB	256
+#define	OCFS2_LA_OLD_DEFAULT	8
+unsigned int ocfs2_la_default_mb(struct ocfs2_super *osb)
+{
+	unsigned int la_mb;
+	unsigned int gd_mb;
+	unsigned int la_max_mb;
+	unsigned int megs_per_slot;
+	struct super_block *sb = osb->sb;
+
+	gd_mb = ocfs2_clusters_to_megabytes(osb->sb,
+		8 * ocfs2_group_bitmap_size(sb, 0, osb->s_feature_incompat));
+
+	/*
+	 * This takes care of files systems with very small group
+	 * descriptors - 512 byte blocksize at cluster sizes lower
+	 * than 16K and also 1k blocksize with 4k cluster size.
+	 */
+	if ((sb->s_blocksize == 512 && osb->s_clustersize <= 8192)
+	    || (sb->s_blocksize == 1024 && osb->s_clustersize == 4096))
+		return OCFS2_LA_OLD_DEFAULT;
+
+	/*
+	 * Leave enough room for some block groups and make the final
+	 * value we work from a multiple of 4.
+	 */
+	gd_mb -= 16;
+	gd_mb &= 0xFFFFFFFB;
+
+	la_mb = gd_mb;
+
+	/*
+	 * Keep window sizes down to a reasonable default
+	 */
+	if (la_mb > OCFS2_LA_MAX_DEFAULT_MB) {
+		/*
+		 * Some clustersize / blocksize combinations will have
+		 * given us a larger than OCFS2_LA_MAX_DEFAULT_MB
+		 * default size, but get poor distribution when
+		 * limited to exactly 256 megabytes.
+		 *
+		 * As an example, 16K clustersize at 4K blocksize
+		 * gives us a cluster group size of 504M. Paring the
+		 * local alloc size down to 256 however, would give us
+		 * only one window and around 200MB left in the
+		 * cluster group. Instead, find the first size below
+		 * 256 which would give us an even distribution.
+		 *
+		 * Larger cluster group sizes actually work out pretty
+		 * well when pared to 256, so we don't have to do this
+		 * for any group that fits more than two
+		 * OCFS2_LA_MAX_DEFAULT_MB windows.
+		 */
+		if (gd_mb > (2 * OCFS2_LA_MAX_DEFAULT_MB))
+			la_mb = 256;
+		else {
+			unsigned int gd_mult = gd_mb;
+
+			while (gd_mult > 256)
+				gd_mult = gd_mult >> 1;
+
+			la_mb = gd_mult;
+		}
+	}
+
+	megs_per_slot = osb->osb_clusters_at_boot / osb->max_slots;
+	megs_per_slot = ocfs2_clusters_to_megabytes(osb->sb, megs_per_slot);
+	/* Too many nodes, too few disk clusters. */
+	if (megs_per_slot < la_mb)
+		la_mb = megs_per_slot;
+
+	/* We can't store more bits than we can in a block. */
+	la_max_mb = ocfs2_clusters_to_megabytes(osb->sb,
+						ocfs2_local_alloc_size(sb) * 8);
+	if (la_mb > la_max_mb)
+		la_mb = la_max_mb;
+
+	return la_mb;
+}
+
+void ocfs2_la_set_sizes(struct ocfs2_super *osb, int requested_mb)
+{
+	struct super_block *sb = osb->sb;
+	unsigned int la_default_mb = ocfs2_la_default_mb(osb);
+	unsigned int la_max_mb;
+
+	la_max_mb = ocfs2_clusters_to_megabytes(sb,
+						ocfs2_local_alloc_size(sb) * 8);
+
+	mlog(0, "requested: %dM, max: %uM, default: %uM\n",
+	     requested_mb, la_max_mb, la_default_mb);
+
+	if (requested_mb == -1) {
+		/* No user request - use defaults */
+		osb->local_alloc_default_bits =
+			ocfs2_megabytes_to_clusters(sb, la_default_mb);
+	} else if (requested_mb > la_max_mb) {
+		/* Request is too big, we give the maximum available */
+		osb->local_alloc_default_bits =
+			ocfs2_megabytes_to_clusters(sb, la_max_mb);
+	} else {
+		osb->local_alloc_default_bits =
+			ocfs2_megabytes_to_clusters(sb, requested_mb);
+	}
+
+	osb->local_alloc_bits = osb->local_alloc_default_bits;
+}
+
 static inline int ocfs2_la_state_enabled(struct ocfs2_super *osb)
 {
 	return (osb->local_alloc_state == OCFS2_LA_THROTTLED ||
@@ -156,7 +302,7 @@ int ocfs2_load_local_alloc(struct ocfs2_super *osb)
 		     osb->local_alloc_bits, (osb->bitmap_cpg - 1));
 		osb->local_alloc_bits =
 			ocfs2_megabytes_to_clusters(osb->sb,
-						    OCFS2_DEFAULT_LOCAL_ALLOC_SIZE);
+						    ocfs2_la_default_mb(osb));
 	}
 
 	/* read the alloc off disk */
@@ -262,6 +408,8 @@ void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb)
 
 	osb->local_alloc_state = OCFS2_LA_DISABLED;
 
+	ocfs2_resmap_uninit(&osb->osb_la_resmap);
+
 	main_bm_inode = ocfs2_get_system_file_inode(osb,
 						    GLOBAL_BITMAP_SYSTEM_INODE,
 						    OCFS2_INVALID_SLOT);
@@ -305,12 +453,7 @@ void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb)
 	}
 
 	ocfs2_clear_local_alloc(alloc);
-
-	status = ocfs2_journal_dirty(handle, bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out_commit;
-	}
+	ocfs2_journal_dirty(handle, bh);
 
 	brelse(bh);
 	osb->local_alloc_bh = NULL;
@@ -476,51 +619,11 @@ out_mutex:
 
 out:
 	if (!status)
-		ocfs2_init_inode_steal_slot(osb);
+		ocfs2_init_steal_slots(osb);
 	mlog_exit(status);
 	return status;
 }
 
-/* Check to see if the local alloc window is within ac->ac_max_block */
-static int ocfs2_local_alloc_in_range(struct inode *inode,
-				      struct ocfs2_alloc_context *ac,
-				      u32 bits_wanted)
-{
-	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
-	struct ocfs2_dinode *alloc;
-	struct ocfs2_local_alloc *la;
-	int start;
-	u64 block_off;
-
-	if (!ac->ac_max_block)
-		return 1;
-
-	alloc = (struct ocfs2_dinode *) osb->local_alloc_bh->b_data;
-	la = OCFS2_LOCAL_ALLOC(alloc);
-
-	start = ocfs2_local_alloc_find_clear_bits(osb, alloc, bits_wanted);
-	if (start == -1) {
-		mlog_errno(-ENOSPC);
-		return 0;
-	}
-
-	/*
-	 * Converting (bm_off + start + bits_wanted) to blocks gives us
-	 * the blkno just past our actual allocation.  This is perfect
-	 * to compare with ac_max_block.
-	 */
-	block_off = ocfs2_clusters_to_blocks(inode->i_sb,
-					     le32_to_cpu(la->la_bm_off) +
-					     start + bits_wanted);
-	mlog(0, "Checking %llu against %llu\n",
-	     (unsigned long long)block_off,
-	     (unsigned long long)ac->ac_max_block);
-	if (block_off > ac->ac_max_block)
-		return 0;
-
-	return 1;
-}
-
 /*
  * make sure we've got at least bits_wanted contiguous bits in the
  * local alloc. You lose them when you drop i_mutex.
@@ -613,17 +716,6 @@ int ocfs2_reserve_local_alloc_bits(struct ocfs2_super *osb,
 		mlog(0, "Calling in_range for max block %llu\n",
 		     (unsigned long long)ac->ac_max_block);
 
-	if (!ocfs2_local_alloc_in_range(local_alloc_inode, ac,
-					bits_wanted)) {
-		/*
-		 * The window is outside ac->ac_max_block.
-		 * This errno tells the caller to keep localalloc enabled
-		 * but to get the allocation from the main bitmap.
-		 */
-		status = -EFBIG;
-		goto bail;
-	}
-
 	ac->ac_inode = local_alloc_inode;
 	/* We should never use localalloc from another slot */
 	ac->ac_alloc_slot = osb->slot_num;
@@ -664,7 +756,8 @@ int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb,
 	alloc = (struct ocfs2_dinode *) osb->local_alloc_bh->b_data;
 	la = OCFS2_LOCAL_ALLOC(alloc);
 
-	start = ocfs2_local_alloc_find_clear_bits(osb, alloc, bits_wanted);
+	start = ocfs2_local_alloc_find_clear_bits(osb, alloc, &bits_wanted,
+						  ac->ac_resv);
 	if (start == -1) {
 		/* TODO: Shouldn't we just BUG here? */
 		status = -ENOSPC;
@@ -674,8 +767,6 @@ int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb,
 
 	bitmap = la->la_bitmap;
 	*bit_off = le32_to_cpu(la->la_bm_off) + start;
-	/* local alloc is always contiguous by nature -- we never
-	 * delete bits from it! */
 	*num_bits = bits_wanted;
 
 	status = ocfs2_journal_access_di(handle,
@@ -687,18 +778,15 @@ int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb,
 		goto bail;
 	}
 
+	ocfs2_resmap_claimed_bits(&osb->osb_la_resmap, ac->ac_resv, start,
+				  bits_wanted);
+
 	while(bits_wanted--)
 		ocfs2_set_bit(start++, bitmap);
 
 	le32_add_cpu(&alloc->id1.bitmap1.i_used, *num_bits);
+	ocfs2_journal_dirty(handle, osb->local_alloc_bh);
 
-	status = ocfs2_journal_dirty(handle, osb->local_alloc_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
-
-	status = 0;
 bail:
 	mlog_exit(status);
 	return status;
@@ -722,13 +810,17 @@ static u32 ocfs2_local_alloc_count_bits(struct ocfs2_dinode *alloc)
 }
 
 static int ocfs2_local_alloc_find_clear_bits(struct ocfs2_super *osb,
-					     struct ocfs2_dinode *alloc,
-					     u32 numbits)
+				     struct ocfs2_dinode *alloc,
+				     u32 *numbits,
+				     struct ocfs2_alloc_reservation *resv)
 {
 	int numfound, bitoff, left, startoff, lastzero;
+	int local_resv = 0;
+	struct ocfs2_alloc_reservation r;
 	void *bitmap = NULL;
+	struct ocfs2_reservation_map *resmap = &osb->osb_la_resmap;
 
-	mlog_entry("(numbits wanted = %u)\n", numbits);
+	mlog_entry("(numbits wanted = %u)\n", *numbits);
 
 	if (!alloc->id1.bitmap1.i_total) {
 		mlog(0, "No bits in my window!\n");
@@ -736,6 +828,30 @@ static int ocfs2_local_alloc_find_clear_bits(struct ocfs2_super *osb,
 		goto bail;
 	}
 
+	if (!resv) {
+		local_resv = 1;
+		ocfs2_resv_init_once(&r);
+		ocfs2_resv_set_type(&r, OCFS2_RESV_FLAG_TMP);
+		resv = &r;
+	}
+
+	numfound = *numbits;
+	if (ocfs2_resmap_resv_bits(resmap, resv, &bitoff, &numfound) == 0) {
+		if (numfound < *numbits)
+			*numbits = numfound;
+		goto bail;
+	}
+
+	/*
+	 * Code error. While reservations are enabled, local
+	 * allocation should _always_ go through them.
+	 */
+	BUG_ON(osb->osb_resv_level != 0);
+
+	/*
+	 * Reservations are disabled. Handle this the old way.
+	 */
+
 	bitmap = OCFS2_LOCAL_ALLOC(alloc)->la_bitmap;
 
 	numfound = bitoff = startoff = 0;
@@ -761,7 +877,7 @@ static int ocfs2_local_alloc_find_clear_bits(struct ocfs2_super *osb,
 			startoff = bitoff+1;
 		}
 		/* we got everything we needed */
-		if (numfound == numbits) {
+		if (numfound == *numbits) {
 			/* mlog(0, "Found it all!\n"); */
 			break;
 		}
@@ -770,12 +886,15 @@ static int ocfs2_local_alloc_find_clear_bits(struct ocfs2_super *osb,
 	mlog(0, "Exiting loop, bitoff = %d, numfound = %d\n", bitoff,
 	     numfound);
 
-	if (numfound == numbits)
+	if (numfound == *numbits)
 		bitoff = startoff - numfound;
 	else
 		bitoff = -1;
 
 bail:
+	if (local_resv)
+		ocfs2_resv_discard(resmap, resv);
+
 	mlog_exit(bitoff);
 	return bitoff;
 }
@@ -872,8 +991,10 @@ static int ocfs2_sync_local_to_main(struct ocfs2_super *osb,
 			     (unsigned long long)la_start_blk,
 			     (unsigned long long)blkno);
 
-			status = ocfs2_free_clusters(handle, main_bm_inode,
-						     main_bm_bh, blkno, count);
+			status = ocfs2_release_clusters(handle,
+							main_bm_inode,
+							main_bm_bh, blkno,
+							count);
 			if (status < 0) {
 				mlog_errno(status);
 				goto bail;
@@ -984,8 +1105,7 @@ static int ocfs2_local_alloc_reserve_for_window(struct ocfs2_super *osb,
 	}
 
 retry_enospc:
-	(*ac)->ac_bits_wanted = osb->local_alloc_bits;
-
+	(*ac)->ac_bits_wanted = osb->local_alloc_default_bits;
 	status = ocfs2_reserve_cluster_bitmap_bits(osb, *ac);
 	if (status == -ENOSPC) {
 		if (ocfs2_recalc_la_window(osb, OCFS2_LA_EVENT_ENOSPC) ==
@@ -1048,7 +1168,7 @@ static int ocfs2_local_alloc_new_window(struct ocfs2_super *osb,
 	/* we used the generic suballoc reserve function, but we set
 	 * everything up nicely, so there's no reason why we can't use
 	 * the more specific cluster api to claim bits. */
-	status = ocfs2_claim_clusters(osb, handle, ac, osb->local_alloc_bits,
+	status = ocfs2_claim_clusters(handle, ac, osb->local_alloc_bits,
 				      &cluster_off, &cluster_count);
 	if (status == -ENOSPC) {
 retry_enospc:
@@ -1061,7 +1181,8 @@ retry_enospc:
 		    OCFS2_LA_DISABLED)
 			goto bail;
 
-		status = ocfs2_claim_clusters(osb, handle, ac,
+		ac->ac_bits_wanted = osb->local_alloc_default_bits;
+		status = ocfs2_claim_clusters(handle, ac,
 					      osb->local_alloc_bits,
 					      &cluster_off,
 					      &cluster_count);
@@ -1096,6 +1217,9 @@ retry_enospc:
 	memset(OCFS2_LOCAL_ALLOC(alloc)->la_bitmap, 0,
 	       le16_to_cpu(la->la_size));
 
+	ocfs2_resmap_restart(&osb->osb_la_resmap, cluster_count,
+			     OCFS2_LOCAL_ALLOC(alloc)->la_bitmap);
+
 	mlog(0, "New window allocated:\n");
 	mlog(0, "window la_bm_off = %u\n",
 	     OCFS2_LOCAL_ALLOC(alloc)->la_bm_off);
@@ -1167,12 +1291,7 @@ static int ocfs2_local_alloc_slide_window(struct ocfs2_super *osb,
 	}
 
 	ocfs2_clear_local_alloc(alloc);
-
-	status = ocfs2_journal_dirty(handle, osb->local_alloc_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	ocfs2_journal_dirty(handle, osb->local_alloc_bh);
 
 	status = ocfs2_sync_local_to_main(osb, handle, alloc_copy,
 					  main_bm_inode, main_bm_bh);
@@ -1190,7 +1309,6 @@ static int ocfs2_local_alloc_slide_window(struct ocfs2_super *osb,
 
 	atomic_inc(&osb->alloc_stats.moves);
 
-	status = 0;
 bail:
 	if (handle)
 		ocfs2_commit_trans(osb, handle);
diff --git a/fs/ocfs2/localalloc.h b/fs/ocfs2/localalloc.h
index ac5ea9f86653..1be9b5864460 100644
--- a/fs/ocfs2/localalloc.h
+++ b/fs/ocfs2/localalloc.h
@@ -30,6 +30,9 @@ int ocfs2_load_local_alloc(struct ocfs2_super *osb);
 
 void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb);
 
+void ocfs2_la_set_sizes(struct ocfs2_super *osb, int requested_mb);
+unsigned int ocfs2_la_default_mb(struct ocfs2_super *osb);
+
 int ocfs2_begin_local_alloc_recovery(struct ocfs2_super *osb,
 				     int node_num,
 				     struct ocfs2_dinode **alloc_copy);
diff --git a/fs/ocfs2/locks.c b/fs/ocfs2/locks.c
index 544ac6245175..b5cb3ede9408 100644
--- a/fs/ocfs2/locks.c
+++ b/fs/ocfs2/locks.c
@@ -133,7 +133,7 @@ int ocfs2_lock(struct file *file, int cmd, struct file_lock *fl)
 
 	if (!(fl->fl_flags & FL_POSIX))
 		return -ENOLCK;
-	if (__mandatory_lock(inode))
+	if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
 		return -ENOLCK;
 
 	return ocfs2_plock(osb->cconn, OCFS2_I(inode)->ip_blkno, file, cmd, fl);
diff --git a/fs/ocfs2/mmap.c b/fs/ocfs2/mmap.c
index 39737613424a..af2b8fe1f139 100644
--- a/fs/ocfs2/mmap.c
+++ b/fs/ocfs2/mmap.c
@@ -25,7 +25,6 @@
 
 #include <linux/fs.h>
 #include <linux/types.h>
-#include <linux/slab.h>
 #include <linux/highmem.h>
 #include <linux/pagemap.h>
 #include <linux/uio.h>
@@ -42,44 +41,20 @@
 #include "file.h"
 #include "inode.h"
 #include "mmap.h"
+#include "super.h"
 
-static inline int ocfs2_vm_op_block_sigs(sigset_t *blocked, sigset_t *oldset)
-{
-	/* The best way to deal with signals in the vm path is
-	 * to block them upfront, rather than allowing the
-	 * locking paths to return -ERESTARTSYS. */
-	sigfillset(blocked);
-
-	/* We should technically never get a bad return value
-	 * from sigprocmask */
-	return sigprocmask(SIG_BLOCK, blocked, oldset);
-}
-
-static inline int ocfs2_vm_op_unblock_sigs(sigset_t *oldset)
-{
-	return sigprocmask(SIG_SETMASK, oldset, NULL);
-}
 
 static int ocfs2_fault(struct vm_area_struct *area, struct vm_fault *vmf)
 {
-	sigset_t blocked, oldset;
-	int error, ret;
+	sigset_t oldset;
+	int ret;
 
 	mlog_entry("(area=%p, page offset=%lu)\n", area, vmf->pgoff);
 
-	error = ocfs2_vm_op_block_sigs(&blocked, &oldset);
-	if (error < 0) {
-		mlog_errno(error);
-		ret = VM_FAULT_SIGBUS;
-		goto out;
-	}
-
+	ocfs2_block_signals(&oldset);
 	ret = filemap_fault(area, vmf);
+	ocfs2_unblock_signals(&oldset);
 
-	error = ocfs2_vm_op_unblock_sigs(&oldset);
-	if (error < 0)
-		mlog_errno(error);
-out:
 	mlog_exit_ptr(vmf->page);
 	return ret;
 }
@@ -159,14 +134,10 @@ static int ocfs2_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
 	struct page *page = vmf->page;
 	struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
 	struct buffer_head *di_bh = NULL;
-	sigset_t blocked, oldset;
-	int ret, ret2;
+	sigset_t oldset;
+	int ret;
 
-	ret = ocfs2_vm_op_block_sigs(&blocked, &oldset);
-	if (ret < 0) {
-		mlog_errno(ret);
-		return ret;
-	}
+	ocfs2_block_signals(&oldset);
 
 	/*
 	 * The cluster locks taken will block a truncate from another
@@ -194,9 +165,7 @@ static int ocfs2_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
 	ocfs2_inode_unlock(inode, 1);
 
 out:
-	ret2 = ocfs2_vm_op_unblock_sigs(&oldset);
-	if (ret2 < 0)
-		mlog_errno(ret2);
+	ocfs2_unblock_signals(&oldset);
 	if (ret)
 		ret = VM_FAULT_SIGBUS;
 	return ret;
diff --git a/fs/ocfs2/namei.c b/fs/ocfs2/namei.c
index 50fb26a6a5f5..f171b51a74f7 100644
--- a/fs/ocfs2/namei.c
+++ b/fs/ocfs2/namei.c
@@ -84,7 +84,7 @@ static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb,
 static int ocfs2_orphan_add(struct ocfs2_super *osb,
 			    handle_t *handle,
 			    struct inode *inode,
-			    struct ocfs2_dinode *fe,
+			    struct buffer_head *fe_bh,
 			    char *name,
 			    struct ocfs2_dir_lookup_result *lookup,
 			    struct inode *orphan_dir_inode);
@@ -204,15 +204,8 @@ static struct inode *ocfs2_get_init_inode(struct inode *dir, int mode)
 		inode->i_nlink = 2;
 	else
 		inode->i_nlink = 1;
-	inode->i_uid = current_fsuid();
-	if (dir->i_mode & S_ISGID) {
-		inode->i_gid = dir->i_gid;
-		if (S_ISDIR(mode))
-			mode |= S_ISGID;
-	} else
-		inode->i_gid = current_fsgid();
-	inode->i_mode = mode;
-	vfs_dq_init(inode);
+	inode_init_owner(inode, dir, mode);
+	dquot_initialize(inode);
 	return inode;
 }
 
@@ -239,11 +232,15 @@ static int ocfs2_mknod(struct inode *dir,
 	};
 	int did_quota_inode = 0;
 	struct ocfs2_dir_lookup_result lookup = { NULL, };
+	sigset_t oldset;
+	int did_block_signals = 0;
 
 	mlog_entry("(0x%p, 0x%p, %d, %lu, '%.*s')\n", dir, dentry, mode,
 		   (unsigned long)dev, dentry->d_name.len,
 		   dentry->d_name.name);
 
+	dquot_initialize(dir);
+
 	/* get our super block */
 	osb = OCFS2_SB(dir->i_sb);
 
@@ -348,13 +345,13 @@ static int ocfs2_mknod(struct inode *dir,
 		goto leave;
 	}
 
-	/* We don't use standard VFS wrapper because we don't want vfs_dq_init
-	 * to be called. */
-	if (sb_any_quota_active(osb->sb) &&
-	    osb->sb->dq_op->alloc_inode(inode, 1) == NO_QUOTA) {
-		status = -EDQUOT;
+	/* Starting to change things, restart is no longer possible. */
+	ocfs2_block_signals(&oldset);
+	did_block_signals = 1;
+
+	status = dquot_alloc_inode(inode);
+	if (status)
 		goto leave;
-	}
 	did_quota_inode = 1;
 
 	mlog_entry("(0x%p, 0x%p, %d, %lu, '%.*s')\n", dir, dentry,
@@ -386,11 +383,7 @@ static int ocfs2_mknod(struct inode *dir,
 			goto leave;
 		}
 		ocfs2_add_links_count(dirfe, 1);
-		status = ocfs2_journal_dirty(handle, parent_fe_bh);
-		if (status < 0) {
-			mlog_errno(status);
-			goto leave;
-		}
+		ocfs2_journal_dirty(handle, parent_fe_bh);
 		inc_nlink(dir);
 	}
 
@@ -410,32 +403,39 @@ static int ocfs2_mknod(struct inode *dir,
 		}
 	}
 
-	status = ocfs2_add_entry(handle, dentry, inode,
-				 OCFS2_I(inode)->ip_blkno, parent_fe_bh,
-				 &lookup);
-	if (status < 0) {
+	/*
+	 * Do this before adding the entry to the directory. We add
+	 * also set d_op after success so that ->d_iput() will cleanup
+	 * the dentry lock even if ocfs2_add_entry() fails below.
+	 */
+	status = ocfs2_dentry_attach_lock(dentry, inode,
+					  OCFS2_I(dir)->ip_blkno);
+	if (status) {
 		mlog_errno(status);
 		goto leave;
 	}
+	dentry->d_op = &ocfs2_dentry_ops;
 
-	status = ocfs2_dentry_attach_lock(dentry, inode,
-					  OCFS2_I(dir)->ip_blkno);
-	if (status) {
+	status = ocfs2_add_entry(handle, dentry, inode,
+				 OCFS2_I(inode)->ip_blkno, parent_fe_bh,
+				 &lookup);
+	if (status < 0) {
 		mlog_errno(status);
 		goto leave;
 	}
 
 	insert_inode_hash(inode);
-	dentry->d_op = &ocfs2_dentry_ops;
 	d_instantiate(dentry, inode);
 	status = 0;
 leave:
 	if (status < 0 && did_quota_inode)
-		vfs_dq_free_inode(inode);
+		dquot_free_inode(inode);
 	if (handle)
 		ocfs2_commit_trans(osb, handle);
 
 	ocfs2_inode_unlock(dir, 1);
+	if (did_block_signals)
+		ocfs2_unblock_signals(&oldset);
 
 	if (status == -ENOSPC)
 		mlog(0, "Disk is full\n");
@@ -447,11 +447,6 @@ leave:
 
 	ocfs2_free_dir_lookup_result(&lookup);
 
-	if ((status < 0) && inode) {
-		clear_nlink(inode);
-		iput(inode);
-	}
-
 	if (inode_ac)
 		ocfs2_free_alloc_context(inode_ac);
 
@@ -461,6 +456,17 @@ leave:
 	if (meta_ac)
 		ocfs2_free_alloc_context(meta_ac);
 
+	/*
+	 * We should call iput after the i_mutex of the bitmap been
+	 * unlocked in ocfs2_free_alloc_context, or the
+	 * ocfs2_delete_inode will mutex_lock again.
+	 */
+	if ((status < 0) && inode) {
+		OCFS2_I(inode)->ip_flags |= OCFS2_INODE_SKIP_ORPHAN_DIR;
+		clear_nlink(inode);
+		iput(inode);
+	}
+
 	mlog_exit(status);
 
 	return status;
@@ -478,14 +484,15 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb,
 	int status = 0;
 	struct ocfs2_dinode *fe = NULL;
 	struct ocfs2_extent_list *fel;
-	u64 fe_blkno = 0;
+	u64 suballoc_loc, fe_blkno = 0;
 	u16 suballoc_bit;
 	u16 feat;
 
 	*new_fe_bh = NULL;
 
-	status = ocfs2_claim_new_inode(osb, handle, dir, parent_fe_bh,
-				       inode_ac, &suballoc_bit, &fe_blkno);
+	status = ocfs2_claim_new_inode(handle, dir, parent_fe_bh,
+				       inode_ac, &suballoc_loc,
+				       &suballoc_bit, &fe_blkno);
 	if (status < 0) {
 		mlog_errno(status);
 		goto leave;
@@ -522,6 +529,7 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb,
 	fe->i_generation = cpu_to_le32(inode->i_generation);
 	fe->i_fs_generation = cpu_to_le32(osb->fs_generation);
 	fe->i_blkno = cpu_to_le64(fe_blkno);
+	fe->i_suballoc_loc = cpu_to_le64(suballoc_loc);
 	fe->i_suballoc_bit = cpu_to_le16(suballoc_bit);
 	fe->i_suballoc_slot = cpu_to_le16(inode_ac->ac_alloc_slot);
 	fe->i_uid = cpu_to_le32(inode->i_uid);
@@ -558,11 +566,7 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb,
 		fel->l_count = cpu_to_le16(ocfs2_extent_recs_per_inode(osb->sb));
 	}
 
-	status = ocfs2_journal_dirty(handle, *new_fe_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto leave;
-	}
+	ocfs2_journal_dirty(handle, *new_fe_bh);
 
 	ocfs2_populate_inode(inode, fe, 1);
 	ocfs2_ci_set_new(osb, INODE_CACHE(inode));
@@ -628,6 +632,7 @@ static int ocfs2_link(struct dentry *old_dentry,
 	struct ocfs2_dinode *fe = NULL;
 	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
 	struct ocfs2_dir_lookup_result lookup = { NULL, };
+	sigset_t oldset;
 
 	mlog_entry("(inode=%lu, old='%.*s' new='%.*s')\n", inode->i_ino,
 		   old_dentry->d_name.len, old_dentry->d_name.name,
@@ -636,6 +641,8 @@ static int ocfs2_link(struct dentry *old_dentry,
 	if (S_ISDIR(inode->i_mode))
 		return -EPERM;
 
+	dquot_initialize(dir);
+
 	err = ocfs2_inode_lock_nested(dir, &parent_fe_bh, 1, OI_LS_PARENT);
 	if (err < 0) {
 		if (err != -ENOENT)
@@ -682,6 +689,9 @@ static int ocfs2_link(struct dentry *old_dentry,
 		goto out_unlock_inode;
 	}
 
+	/* Starting to change things, restart is no longer possible. */
+	ocfs2_block_signals(&oldset);
+
 	err = ocfs2_journal_access_di(handle, INODE_CACHE(inode), fe_bh,
 				      OCFS2_JOURNAL_ACCESS_WRITE);
 	if (err < 0) {
@@ -694,14 +704,7 @@ static int ocfs2_link(struct dentry *old_dentry,
 	ocfs2_set_links_count(fe, inode->i_nlink);
 	fe->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
 	fe->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
-
-	err = ocfs2_journal_dirty(handle, fe_bh);
-	if (err < 0) {
-		ocfs2_add_links_count(fe, -1);
-		drop_nlink(inode);
-		mlog_errno(err);
-		goto out_commit;
-	}
+	ocfs2_journal_dirty(handle, fe_bh);
 
 	err = ocfs2_add_entry(handle, dentry, inode,
 			      OCFS2_I(inode)->ip_blkno,
@@ -725,6 +728,7 @@ static int ocfs2_link(struct dentry *old_dentry,
 
 out_commit:
 	ocfs2_commit_trans(osb, handle);
+	ocfs2_unblock_signals(&oldset);
 out_unlock_inode:
 	ocfs2_inode_unlock(inode, 1);
 
@@ -791,6 +795,8 @@ static int ocfs2_unlink(struct inode *dir,
 	mlog_entry("(0x%p, 0x%p, '%.*s')\n", dir, dentry,
 		   dentry->d_name.len, dentry->d_name.name);
 
+	dquot_initialize(dir);
+
 	BUG_ON(dentry->d_parent->d_inode != dir);
 
 	mlog(0, "ino = %llu\n", (unsigned long long)OCFS2_I(inode)->ip_blkno);
@@ -877,7 +883,7 @@ static int ocfs2_unlink(struct inode *dir,
 	fe = (struct ocfs2_dinode *) fe_bh->b_data;
 
 	if (inode_is_unlinkable(inode)) {
-		status = ocfs2_orphan_add(osb, handle, inode, fe, orphan_name,
+		status = ocfs2_orphan_add(osb, handle, inode, fe_bh, orphan_name,
 					  &orphan_insert, orphan_dir);
 		if (status < 0) {
 			mlog_errno(status);
@@ -896,12 +902,7 @@ static int ocfs2_unlink(struct inode *dir,
 		drop_nlink(inode);
 	drop_nlink(inode);
 	ocfs2_set_links_count(fe, inode->i_nlink);
-
-	status = ocfs2_journal_dirty(handle, fe_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto leave;
-	}
+	ocfs2_journal_dirty(handle, fe_bh);
 
 	dir->i_ctime = dir->i_mtime = CURRENT_TIME;
 	if (S_ISDIR(inode->i_mode))
@@ -1051,6 +1052,9 @@ static int ocfs2_rename(struct inode *old_dir,
 		   old_dentry->d_name.len, old_dentry->d_name.name,
 		   new_dentry->d_name.len, new_dentry->d_name.name);
 
+	dquot_initialize(old_dir);
+	dquot_initialize(new_dir);
+
 	osb = OCFS2_SB(old_dir->i_sb);
 
 	if (new_inode) {
@@ -1295,7 +1299,7 @@ static int ocfs2_rename(struct inode *old_dir,
 		if (S_ISDIR(new_inode->i_mode) ||
 		    (ocfs2_read_links_count(newfe) == 1)) {
 			status = ocfs2_orphan_add(osb, handle, new_inode,
-						  newfe, orphan_name,
+						  newfe_bh, orphan_name,
 						  &orphan_insert, orphan_dir);
 			if (status < 0) {
 				mlog_errno(status);
@@ -1316,12 +1320,7 @@ static int ocfs2_rename(struct inode *old_dir,
 			ocfs2_set_links_count(newfe, 0);
 		else
 			ocfs2_add_links_count(newfe, -1);
-
-		status = ocfs2_journal_dirty(handle, newfe_bh);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
+		ocfs2_journal_dirty(handle, newfe_bh);
 	} else {
 		/* if the name was not found in new_dir, add it now */
 		status = ocfs2_add_entry(handle, new_dentry, old_inode,
@@ -1340,10 +1339,7 @@ static int ocfs2_rename(struct inode *old_dir,
 
 		old_di->i_ctime = cpu_to_le64(old_inode->i_ctime.tv_sec);
 		old_di->i_ctime_nsec = cpu_to_le32(old_inode->i_ctime.tv_nsec);
-
-		status = ocfs2_journal_dirty(handle, old_inode_bh);
-		if (status < 0)
-			mlog_errno(status);
+		ocfs2_journal_dirty(handle, old_inode_bh);
 	} else
 		mlog_errno(status);
 
@@ -1415,7 +1411,7 @@ static int ocfs2_rename(struct inode *old_dir,
 							 OCFS2_JOURNAL_ACCESS_WRITE);
 			fe = (struct ocfs2_dinode *) old_dir_bh->b_data;
 			ocfs2_set_links_count(fe, old_dir->i_nlink);
-			status = ocfs2_journal_dirty(handle, old_dir_bh);
+			ocfs2_journal_dirty(handle, old_dir_bh);
 		}
 	}
 	ocfs2_dentry_move(old_dentry, new_dentry, old_dir, new_dir);
@@ -1547,11 +1543,7 @@ static int ocfs2_create_symlink_data(struct ocfs2_super *osb,
 		       (bytes_left > sb->s_blocksize) ? sb->s_blocksize :
 		       bytes_left);
 
-		status = ocfs2_journal_dirty(handle, bhs[virtual]);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
+		ocfs2_journal_dirty(handle, bhs[virtual]);
 
 		virtual++;
 		p_blkno++;
@@ -1595,10 +1587,14 @@ static int ocfs2_symlink(struct inode *dir,
 	};
 	int did_quota = 0, did_quota_inode = 0;
 	struct ocfs2_dir_lookup_result lookup = { NULL, };
+	sigset_t oldset;
+	int did_block_signals = 0;
 
 	mlog_entry("(0x%p, 0x%p, symname='%s' actual='%.*s')\n", dir,
 		   dentry, symname, dentry->d_name.len, dentry->d_name.name);
 
+	dquot_initialize(dir);
+
 	sb = dir->i_sb;
 	osb = OCFS2_SB(sb);
 
@@ -1688,13 +1684,13 @@ static int ocfs2_symlink(struct inode *dir,
 		goto bail;
 	}
 
-	/* We don't use standard VFS wrapper because we don't want vfs_dq_init
-	 * to be called. */
-	if (sb_any_quota_active(osb->sb) &&
-	    osb->sb->dq_op->alloc_inode(inode, 1) == NO_QUOTA) {
-		status = -EDQUOT;
+	/* Starting to change things, restart is no longer possible. */
+	ocfs2_block_signals(&oldset);
+	did_block_signals = 1;
+
+	status = dquot_alloc_inode(inode);
+	if (status)
 		goto bail;
-	}
 	did_quota_inode = 1;
 
 	mlog_entry("(0x%p, 0x%p, %d, '%.*s')\n", dir, dentry,
@@ -1716,11 +1712,10 @@ static int ocfs2_symlink(struct inode *dir,
 		u32 offset = 0;
 
 		inode->i_op = &ocfs2_symlink_inode_operations;
-		if (vfs_dq_alloc_space_nodirty(inode,
-		    ocfs2_clusters_to_bytes(osb->sb, 1))) {
-			status = -EDQUOT;
+		status = dquot_alloc_space_nodirty(inode,
+		    ocfs2_clusters_to_bytes(osb->sb, 1));
+		if (status)
 			goto bail;
-		}
 		did_quota = 1;
 		status = ocfs2_add_inode_data(osb, inode, &offset, 1, 0,
 					      new_fe_bh,
@@ -1769,33 +1764,40 @@ static int ocfs2_symlink(struct inode *dir,
 		}
 	}
 
-	status = ocfs2_add_entry(handle, dentry, inode,
-				 le64_to_cpu(fe->i_blkno), parent_fe_bh,
-				 &lookup);
-	if (status < 0) {
+	/*
+	 * Do this before adding the entry to the directory. We add
+	 * also set d_op after success so that ->d_iput() will cleanup
+	 * the dentry lock even if ocfs2_add_entry() fails below.
+	 */
+	status = ocfs2_dentry_attach_lock(dentry, inode, OCFS2_I(dir)->ip_blkno);
+	if (status) {
 		mlog_errno(status);
 		goto bail;
 	}
+	dentry->d_op = &ocfs2_dentry_ops;
 
-	status = ocfs2_dentry_attach_lock(dentry, inode, OCFS2_I(dir)->ip_blkno);
-	if (status) {
+	status = ocfs2_add_entry(handle, dentry, inode,
+				 le64_to_cpu(fe->i_blkno), parent_fe_bh,
+				 &lookup);
+	if (status < 0) {
 		mlog_errno(status);
 		goto bail;
 	}
 
 	insert_inode_hash(inode);
-	dentry->d_op = &ocfs2_dentry_ops;
 	d_instantiate(dentry, inode);
 bail:
 	if (status < 0 && did_quota)
-		vfs_dq_free_space_nodirty(inode,
+		dquot_free_space_nodirty(inode,
 					ocfs2_clusters_to_bytes(osb->sb, 1));
 	if (status < 0 && did_quota_inode)
-		vfs_dq_free_inode(inode);
+		dquot_free_inode(inode);
 	if (handle)
 		ocfs2_commit_trans(osb, handle);
 
 	ocfs2_inode_unlock(dir, 1);
+	if (did_block_signals)
+		ocfs2_unblock_signals(&oldset);
 
 	brelse(new_fe_bh);
 	brelse(parent_fe_bh);
@@ -1809,6 +1811,7 @@ bail:
 	if (xattr_ac)
 		ocfs2_free_alloc_context(xattr_ac);
 	if ((status < 0) && inode) {
+		OCFS2_I(inode)->ip_flags |= OCFS2_INODE_SKIP_ORPHAN_DIR;
 		clear_nlink(inode);
 		iput(inode);
 	}
@@ -1909,7 +1912,7 @@ leave:
 static int ocfs2_orphan_add(struct ocfs2_super *osb,
 			    handle_t *handle,
 			    struct inode *inode,
-			    struct ocfs2_dinode *fe,
+			    struct buffer_head *fe_bh,
 			    char *name,
 			    struct ocfs2_dir_lookup_result *lookup,
 			    struct inode *orphan_dir_inode)
@@ -1917,6 +1920,7 @@ static int ocfs2_orphan_add(struct ocfs2_super *osb,
 	struct buffer_head *orphan_dir_bh = NULL;
 	int status = 0;
 	struct ocfs2_dinode *orphan_fe;
+	struct ocfs2_dinode *fe = (struct ocfs2_dinode *) fe_bh->b_data;
 
 	mlog_entry("(inode->i_ino = %lu)\n", inode->i_ino);
 
@@ -1941,29 +1945,42 @@ static int ocfs2_orphan_add(struct ocfs2_super *osb,
 	if (S_ISDIR(inode->i_mode))
 		ocfs2_add_links_count(orphan_fe, 1);
 	orphan_dir_inode->i_nlink = ocfs2_read_links_count(orphan_fe);
+	ocfs2_journal_dirty(handle, orphan_dir_bh);
 
-	status = ocfs2_journal_dirty(handle, orphan_dir_bh);
+	status = __ocfs2_add_entry(handle, orphan_dir_inode, name,
+				   OCFS2_ORPHAN_NAMELEN, inode,
+				   OCFS2_I(inode)->ip_blkno,
+				   orphan_dir_bh, lookup);
 	if (status < 0) {
 		mlog_errno(status);
 		goto leave;
 	}
 
-	status = __ocfs2_add_entry(handle, orphan_dir_inode, name,
-				   OCFS2_ORPHAN_NAMELEN, inode,
-				   OCFS2_I(inode)->ip_blkno,
-				   orphan_dir_bh, lookup);
+	/*
+	 * We're going to journal the change of i_flags and i_orphaned_slot.
+	 * It's safe anyway, though some callers may duplicate the journaling.
+	 * Journaling within the func just make the logic look more
+	 * straightforward.
+	 */
+	status = ocfs2_journal_access_di(handle,
+					 INODE_CACHE(inode),
+					 fe_bh,
+					 OCFS2_JOURNAL_ACCESS_WRITE);
 	if (status < 0) {
 		mlog_errno(status);
 		goto leave;
 	}
 
 	le32_add_cpu(&fe->i_flags, OCFS2_ORPHANED_FL);
+	OCFS2_I(inode)->ip_flags &= ~OCFS2_INODE_SKIP_ORPHAN_DIR;
 
 	/* Record which orphan dir our inode now resides
 	 * in. delete_inode will use this to determine which orphan
 	 * dir to lock. */
 	fe->i_orphaned_slot = cpu_to_le16(osb->slot_num);
 
+	ocfs2_journal_dirty(handle, fe_bh);
+
 	mlog(0, "Inode %llu orphaned in slot %d\n",
 	     (unsigned long long)OCFS2_I(inode)->ip_blkno, osb->slot_num);
 
@@ -2027,12 +2044,7 @@ int ocfs2_orphan_del(struct ocfs2_super *osb,
 	if (S_ISDIR(inode->i_mode))
 		ocfs2_add_links_count(orphan_fe, -1);
 	orphan_dir_inode->i_nlink = ocfs2_read_links_count(orphan_fe);
-
-	status = ocfs2_journal_dirty(handle, orphan_dir_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto leave;
-	}
+	ocfs2_journal_dirty(handle, orphan_dir_bh);
 
 leave:
 	ocfs2_free_dir_lookup_result(&lookup);
@@ -2099,13 +2111,9 @@ int ocfs2_create_inode_in_orphan(struct inode *dir,
 		goto leave;
 	}
 
-	/* We don't use standard VFS wrapper because we don't want vfs_dq_init
-	 * to be called. */
-	if (sb_any_quota_active(osb->sb) &&
-	    osb->sb->dq_op->alloc_inode(inode, 1) == NO_QUOTA) {
-		status = -EDQUOT;
+	status = dquot_alloc_inode(inode);
+	if (status)
 		goto leave;
-	}
 	did_quota_inode = 1;
 
 	inode->i_nlink = 0;
@@ -2125,7 +2133,7 @@ int ocfs2_create_inode_in_orphan(struct inode *dir,
 	}
 
 	di = (struct ocfs2_dinode *)new_di_bh->b_data;
-	status = ocfs2_orphan_add(osb, handle, inode, di, orphan_name,
+	status = ocfs2_orphan_add(osb, handle, inode, new_di_bh, orphan_name,
 				  &orphan_insert, orphan_dir);
 	if (status < 0) {
 		mlog_errno(status);
@@ -2140,7 +2148,7 @@ int ocfs2_create_inode_in_orphan(struct inode *dir,
 	insert_inode_hash(inode);
 leave:
 	if (status < 0 && did_quota_inode)
-		vfs_dq_free_inode(inode);
+		dquot_free_inode(inode);
 	if (handle)
 		ocfs2_commit_trans(osb, handle);
 
diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h
index 9362eea7424b..c67003b6b5a2 100644
--- a/fs/ocfs2/ocfs2.h
+++ b/fs/ocfs2/ocfs2.h
@@ -42,10 +42,12 @@
 
 #include "ocfs2_fs.h"
 #include "ocfs2_lockid.h"
+#include "ocfs2_ioctl.h"
 
 /* For struct ocfs2_blockcheck_stats */
 #include "blockcheck.h"
 
+#include "reservations.h"
 
 /* Caching of metadata buffers */
 
@@ -136,6 +138,10 @@ enum ocfs2_unlock_action {
 #define OCFS2_LOCK_PENDING       (0x00000400) /* This lockres is pending a
 						 call to dlm_lock.  Only
 						 exists with BUSY set. */
+#define OCFS2_LOCK_UPCONVERT_FINISHING (0x00000800) /* blocks the dc thread
+						     * from downconverting
+						     * before the upconvert
+						     * has completed */
 
 struct ocfs2_lock_res_ops;
 
@@ -155,7 +161,7 @@ struct ocfs2_lock_res {
 	int                      l_level;
 	unsigned int             l_ro_holders;
 	unsigned int             l_ex_holders;
-	union ocfs2_dlm_lksb     l_lksb;
+	struct ocfs2_dlm_lksb    l_lksb;
 
 	/* used from AST/BAST funcs. */
 	enum ocfs2_ast_action    l_action;
@@ -301,7 +307,9 @@ struct ocfs2_super
 	u32 s_next_generation;
 	unsigned long osb_flags;
 	s16 s_inode_steal_slot;
+	s16 s_meta_steal_slot;
 	atomic_t s_num_inodes_stolen;
+	atomic_t s_num_meta_stolen;
 
 	unsigned long s_mount_opt;
 	unsigned int s_atime_quantum;
@@ -334,6 +342,9 @@ struct ocfs2_super
 	 */
 	unsigned int local_alloc_bits;
 	unsigned int local_alloc_default_bits;
+	/* osb_clusters_at_boot can become stale! Do not trust it to
+	 * be up to date. */
+	unsigned int osb_clusters_at_boot;
 
 	enum ocfs2_local_alloc_state local_alloc_state; /* protected
 							 * by osb_lock */
@@ -342,6 +353,11 @@ struct ocfs2_super
 
 	u64 la_last_gd;
 
+	struct ocfs2_reservation_map	osb_la_resmap;
+
+	unsigned int	osb_resv_level;
+	unsigned int	osb_dir_resv_level;
+
 	/* Next three fields are for local node slot recovery during
 	 * mount. */
 	int dirty;
@@ -475,6 +491,13 @@ static inline int ocfs2_supports_indexed_dirs(struct ocfs2_super *osb)
 	return 0;
 }
 
+static inline int ocfs2_supports_discontig_bg(struct ocfs2_super *osb)
+{
+	if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG)
+		return 1;
+	return 0;
+}
+
 static inline unsigned int ocfs2_link_max(struct ocfs2_super *osb)
 {
 	if (ocfs2_supports_indexed_dirs(osb))
@@ -756,35 +779,24 @@ static inline unsigned int ocfs2_megabytes_to_clusters(struct super_block *sb,
 	return megs << (20 - OCFS2_SB(sb)->s_clustersize_bits);
 }
 
-static inline void ocfs2_init_inode_steal_slot(struct ocfs2_super *osb)
+static inline unsigned int ocfs2_clusters_to_megabytes(struct super_block *sb,
+						       unsigned int clusters)
 {
-	spin_lock(&osb->osb_lock);
-	osb->s_inode_steal_slot = OCFS2_INVALID_SLOT;
-	spin_unlock(&osb->osb_lock);
-	atomic_set(&osb->s_num_inodes_stolen, 0);
+	return clusters >> (20 - OCFS2_SB(sb)->s_clustersize_bits);
 }
 
-static inline void ocfs2_set_inode_steal_slot(struct ocfs2_super *osb,
-					      s16 slot)
+static inline void _ocfs2_set_bit(unsigned int bit, unsigned long *bitmap)
 {
-	spin_lock(&osb->osb_lock);
-	osb->s_inode_steal_slot = slot;
-	spin_unlock(&osb->osb_lock);
+	ext2_set_bit(bit, bitmap);
 }
+#define ocfs2_set_bit(bit, addr) _ocfs2_set_bit((bit), (unsigned long *)(addr))
 
-static inline s16 ocfs2_get_inode_steal_slot(struct ocfs2_super *osb)
+static inline void _ocfs2_clear_bit(unsigned int bit, unsigned long *bitmap)
 {
-	s16 slot;
-
-	spin_lock(&osb->osb_lock);
-	slot = osb->s_inode_steal_slot;
-	spin_unlock(&osb->osb_lock);
-
-	return slot;
+	ext2_clear_bit(bit, bitmap);
 }
+#define ocfs2_clear_bit(bit, addr) _ocfs2_clear_bit((bit), (unsigned long *)(addr))
 
-#define ocfs2_set_bit ext2_set_bit
-#define ocfs2_clear_bit ext2_clear_bit
 #define ocfs2_test_bit ext2_test_bit
 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit
 #define ocfs2_find_next_bit ext2_find_next_bit
diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
index 1a1a679e51b5..33f1c9a8258d 100644
--- a/fs/ocfs2/ocfs2_fs.h
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -100,7 +100,8 @@
 					 | OCFS2_FEATURE_INCOMPAT_XATTR \
 					 | OCFS2_FEATURE_INCOMPAT_META_ECC \
 					 | OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS \
-					 | OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE)
+					 | OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE \
+					 | OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG)
 #define OCFS2_FEATURE_RO_COMPAT_SUPP	(OCFS2_FEATURE_RO_COMPAT_UNWRITTEN \
 					 | OCFS2_FEATURE_RO_COMPAT_USRQUOTA \
 					 | OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
@@ -165,6 +166,9 @@
 /* Refcount tree support */
 #define OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE	0x1000
 
+/* Discontigous block groups */
+#define OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG	0x2000
+
 /*
  * backup superblock flag is used to indicate that this volume
  * has backup superblocks.
@@ -254,63 +258,6 @@
 						 * refcount tree */
 
 /*
- * ioctl commands
- */
-#define OCFS2_IOC_GETFLAGS	_IOR('f', 1, long)
-#define OCFS2_IOC_SETFLAGS	_IOW('f', 2, long)
-#define OCFS2_IOC32_GETFLAGS	_IOR('f', 1, int)
-#define OCFS2_IOC32_SETFLAGS	_IOW('f', 2, int)
-
-/*
- * Space reservation / allocation / free ioctls and argument structure
- * are designed to be compatible with XFS.
- *
- * ALLOCSP* and FREESP* are not and will never be supported, but are
- * included here for completeness.
- */
-struct ocfs2_space_resv {
-	__s16		l_type;
-	__s16		l_whence;
-	__s64		l_start;
-	__s64		l_len;		/* len == 0 means until end of file */
-	__s32		l_sysid;
-	__u32		l_pid;
-	__s32		l_pad[4];	/* reserve area			    */
-};
-
-#define OCFS2_IOC_ALLOCSP		_IOW ('X', 10, struct ocfs2_space_resv)
-#define OCFS2_IOC_FREESP		_IOW ('X', 11, struct ocfs2_space_resv)
-#define OCFS2_IOC_RESVSP		_IOW ('X', 40, struct ocfs2_space_resv)
-#define OCFS2_IOC_UNRESVSP	_IOW ('X', 41, struct ocfs2_space_resv)
-#define OCFS2_IOC_ALLOCSP64	_IOW ('X', 36, struct ocfs2_space_resv)
-#define OCFS2_IOC_FREESP64	_IOW ('X', 37, struct ocfs2_space_resv)
-#define OCFS2_IOC_RESVSP64	_IOW ('X', 42, struct ocfs2_space_resv)
-#define OCFS2_IOC_UNRESVSP64	_IOW ('X', 43, struct ocfs2_space_resv)
-
-/* Used to pass group descriptor data when online resize is done */
-struct ocfs2_new_group_input {
-	__u64 group;		/* Group descriptor's blkno. */
-	__u32 clusters;		/* Total number of clusters in this group */
-	__u32 frees;		/* Total free clusters in this group */
-	__u16 chain;		/* Chain for this group */
-	__u16 reserved1;
-	__u32 reserved2;
-};
-
-#define OCFS2_IOC_GROUP_EXTEND	_IOW('o', 1, int)
-#define OCFS2_IOC_GROUP_ADD	_IOW('o', 2,struct ocfs2_new_group_input)
-#define OCFS2_IOC_GROUP_ADD64	_IOW('o', 3,struct ocfs2_new_group_input)
-
-/* Used to pass 2 file names to reflink. */
-struct reflink_arguments {
-	__u64 old_path;
-	__u64 new_path;
-	__u64 preserve;
-};
-#define OCFS2_IOC_REFLINK	_IOW('o', 4, struct reflink_arguments)
-
-
-/*
  * Journal Flags (ocfs2_dinode.id1.journal1.i_flags)
  */
 #define OCFS2_JOURNAL_DIRTY_FL	(0x00000001)	/* Journal needs recovery */
@@ -340,14 +287,6 @@ struct reflink_arguments {
 #define OCFS2_MIN_JOURNAL_SIZE		(4 * 1024 * 1024)
 
 /*
- * Default local alloc size (in megabytes)
- *
- * The value chosen should be such that most allocations, including new
- * block groups, use local alloc.
- */
-#define OCFS2_DEFAULT_LOCAL_ALLOC_SIZE	8
-
-/*
  * Inline extended attribute size (in bytes)
  * The value chosen should be aligned to 16 byte boundaries.
  */
@@ -569,7 +508,10 @@ struct ocfs2_extent_block
 					   block group */
 	__le32 h_fs_generation;		/* Must match super block */
 	__le64 h_blkno;			/* Offset on disk, in blocks */
-/*20*/	__le64 h_reserved3;
+/*20*/	__le64 h_suballoc_loc;		/* Suballocator block group this
+					   eb belongs to.  Only valid
+					   if allocated from a
+					   discontiguous block group */
 	__le64 h_next_leaf_blk;		/* Offset on disk, in blocks,
 					   of next leaf header pointing
 					   to data */
@@ -736,7 +678,11 @@ struct ocfs2_dinode {
 /*80*/	struct ocfs2_block_check i_check;	/* Error checking */
 /*88*/	__le64 i_dx_root;		/* Pointer to dir index root block */
 /*90*/	__le64 i_refcount_loc;
-	__le64 i_reserved2[4];
+	__le64 i_suballoc_loc;		/* Suballocator block group this
+					   inode belongs to.  Only valid
+					   if allocated from a
+					   discontiguous block group */
+/*A0*/	__le64 i_reserved2[3];
 /*B8*/	union {
 		__le64 i_pad1;		/* Generic way to refer to this
 					   64bit union */
@@ -871,7 +817,12 @@ struct ocfs2_dx_root_block {
 	__le32		dr_reserved2;
 	__le64		dr_free_blk;		/* Pointer to head of free
 						 * unindexed block list. */
-	__le64		dr_reserved3[15];
+	__le64		dr_suballoc_loc;	/* Suballocator block group
+						   this root belongs to.
+						   Only valid if allocated
+						   from a discontiguous
+						   block group */
+	__le64		dr_reserved3[14];
 	union {
 		struct ocfs2_extent_list dr_list; /* Keep this aligned to 128
 						   * bits for maximum space
@@ -897,6 +848,13 @@ struct ocfs2_dx_leaf {
 };
 
 /*
+ * Largest bitmap for a block (suballocator) group in bytes.  This limit
+ * does not affect cluster groups (global allocator).  Cluster group
+ * bitmaps run to the end of the block.
+ */
+#define OCFS2_MAX_BG_BITMAP_SIZE	256
+
+/*
  * On disk allocator group structure for OCFS2
  */
 struct ocfs2_group_desc
@@ -917,7 +875,29 @@ struct ocfs2_group_desc
 	__le64   bg_blkno;               /* Offset on disk, in blocks */
 /*30*/	struct ocfs2_block_check bg_check;	/* Error checking */
 	__le64   bg_reserved2;
-/*40*/	__u8    bg_bitmap[0];
+/*40*/	union {
+		__u8    bg_bitmap[0];
+		struct {
+			/*
+			 * Block groups may be discontiguous when
+			 * OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG is set.
+			 * The extents of a discontigous block group are
+			 * stored in bg_list.  It is a flat list.
+			 * l_tree_depth must always be zero.  A
+			 * discontiguous group is signified by a non-zero
+			 * bg_list->l_next_free_rec.  Only block groups
+			 * can be discontiguous; Cluster groups cannot.
+			 * We've never made a block group with more than
+			 * 2048 blocks (256 bytes of bg_bitmap).  This
+			 * codifies that limit so that we can fit bg_list.
+			 * bg_size of a discontiguous block group will
+			 * be 256 to match bg_bitmap_filler.
+			 */
+			__u8 bg_bitmap_filler[OCFS2_MAX_BG_BITMAP_SIZE];
+/*140*/			struct ocfs2_extent_list bg_list;
+		};
+	};
+/* Actual on-disk size is one block */
 };
 
 struct ocfs2_refcount_rec {
@@ -962,7 +942,11 @@ struct ocfs2_refcount_block {
 /*40*/	__le32 rf_generation;		/* generation number. all be the same
 					 * for the same refcount tree. */
 	__le32 rf_reserved0;
-	__le64 rf_reserved1[7];
+	__le64 rf_suballoc_loc;		/* Suballocator block group this
+					   refcount block belongs to. Only
+					   valid if allocated from a
+					   discontiguous block group */
+/*50*/	__le64 rf_reserved1[6];
 /*80*/	union {
 		struct ocfs2_refcount_list rf_records;  /* List of refcount
 							  records */
@@ -1074,7 +1058,10 @@ struct ocfs2_xattr_block {
 					real xattr or a xattr tree. */
 	__le16	xb_reserved0;
 	__le32  xb_reserved1;
-	__le64	xb_reserved2;
+	__le64	xb_suballoc_loc;	/* Suballocator block group this
+					   xattr block belongs to. Only
+					   valid if allocated from a
+					   discontiguous block group */
 /*30*/	union {
 		struct ocfs2_xattr_header xb_header; /* xattr header if this
 							block contains xattr */
@@ -1311,6 +1298,16 @@ static inline u16 ocfs2_extent_recs_per_eb(struct super_block *sb)
 	return size / sizeof(struct ocfs2_extent_rec);
 }
 
+static inline u16 ocfs2_extent_recs_per_gd(struct super_block *sb)
+{
+	int size;
+
+	size = sb->s_blocksize -
+		offsetof(struct ocfs2_group_desc, bg_list.l_recs);
+
+	return size / sizeof(struct ocfs2_extent_rec);
+}
+
 static inline int ocfs2_dx_entries_per_leaf(struct super_block *sb)
 {
 	int size;
@@ -1341,13 +1338,23 @@ static inline u16 ocfs2_local_alloc_size(struct super_block *sb)
 	return size;
 }
 
-static inline int ocfs2_group_bitmap_size(struct super_block *sb)
+static inline int ocfs2_group_bitmap_size(struct super_block *sb,
+					  int suballocator,
+					  u32 feature_incompat)
 {
-	int size;
-
-	size = sb->s_blocksize -
+	int size = sb->s_blocksize -
 		offsetof(struct ocfs2_group_desc, bg_bitmap);
 
+	/*
+	 * The cluster allocator uses the entire block.  Suballocators have
+	 * never used more than OCFS2_MAX_BG_BITMAP_SIZE.  Unfortunately, older
+	 * code expects bg_size set to the maximum.  Thus we must keep
+	 * bg_size as-is unless discontig_bg is enabled.
+	 */
+	if (suballocator &&
+	    (feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG))
+		size = OCFS2_MAX_BG_BITMAP_SIZE;
+
 	return size;
 }
 
@@ -1417,9 +1424,16 @@ static inline int ocfs2_fast_symlink_chars(int blocksize)
 	return blocksize - offsetof(struct ocfs2_dinode, id2.i_symlink);
 }
 
-static inline int ocfs2_max_inline_data(int blocksize)
+static inline int ocfs2_max_inline_data_with_xattr(int blocksize,
+						   struct ocfs2_dinode *di)
 {
-	return blocksize - offsetof(struct ocfs2_dinode, id2.i_data.id_data);
+	if (di && (di->i_dyn_features & OCFS2_INLINE_XATTR_FL))
+		return blocksize -
+			offsetof(struct ocfs2_dinode, id2.i_data.id_data) -
+			di->i_xattr_inline_size;
+	else
+		return blocksize -
+			offsetof(struct ocfs2_dinode, id2.i_data.id_data);
 }
 
 static inline int ocfs2_extent_recs_per_inode(int blocksize)
@@ -1452,23 +1466,43 @@ static inline int ocfs2_extent_recs_per_eb(int blocksize)
 	return size / sizeof(struct ocfs2_extent_rec);
 }
 
-static inline int ocfs2_local_alloc_size(int blocksize)
+static inline int ocfs2_extent_recs_per_gd(int blocksize)
 {
 	int size;
 
 	size = blocksize -
-		offsetof(struct ocfs2_dinode, id2.i_lab.la_bitmap);
+		offsetof(struct ocfs2_group_desc, bg_list.l_recs);
 
-	return size;
+	return size / sizeof(struct ocfs2_extent_rec);
 }
 
-static inline int ocfs2_group_bitmap_size(int blocksize)
+static inline int ocfs2_local_alloc_size(int blocksize)
 {
 	int size;
 
 	size = blocksize -
+		offsetof(struct ocfs2_dinode, id2.i_lab.la_bitmap);
+
+	return size;
+}
+
+static inline int ocfs2_group_bitmap_size(int blocksize,
+					  int suballocator,
+					  uint32_t feature_incompat)
+{
+	int size = sb->s_blocksize -
 		offsetof(struct ocfs2_group_desc, bg_bitmap);
 
+	/*
+	 * The cluster allocator uses the entire block.  Suballocators have
+	 * never used more than OCFS2_MAX_BG_BITMAP_SIZE.  Unfortunately, older
+	 * code expects bg_size set to the maximum.  Thus we must keep
+	 * bg_size as-is unless discontig_bg is enabled.
+	 */
+	if (suballocator &&
+	    (feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG))
+		size = OCFS2_MAX_BG_BITMAP_SIZE;
+
 	return size;
 }
 
@@ -1541,5 +1575,19 @@ static inline void ocfs2_set_de_type(struct ocfs2_dir_entry *de,
 	de->file_type = ocfs2_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
 }
 
+static inline int ocfs2_gd_is_discontig(struct ocfs2_group_desc *gd)
+{
+	if ((offsetof(struct ocfs2_group_desc, bg_bitmap) +
+	     le16_to_cpu(gd->bg_size)) !=
+	    offsetof(struct ocfs2_group_desc, bg_list))
+		return 0;
+	/*
+	 * Only valid to check l_next_free_rec if
+	 * bg_bitmap + bg_size == bg_list.
+	 */
+	if (!gd->bg_list.l_next_free_rec)
+		return 0;
+	return 1;
+}
 #endif  /* _OCFS2_FS_H */
 
diff --git a/fs/ocfs2/ocfs2_ioctl.h b/fs/ocfs2/ocfs2_ioctl.h
new file mode 100644
index 000000000000..2d3420af1a83
--- /dev/null
+++ b/fs/ocfs2/ocfs2_ioctl.h
@@ -0,0 +1,79 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ocfs2_ioctl.h
+ *
+ * Defines OCFS2 ioctls.
+ *
+ * Copyright (C) 2010 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, version 2, 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.
+ */
+
+#ifndef OCFS2_IOCTL_H
+#define OCFS2_IOCTL_H
+
+/*
+ * ioctl commands
+ */
+#define OCFS2_IOC_GETFLAGS	_IOR('f', 1, long)
+#define OCFS2_IOC_SETFLAGS	_IOW('f', 2, long)
+#define OCFS2_IOC32_GETFLAGS	_IOR('f', 1, int)
+#define OCFS2_IOC32_SETFLAGS	_IOW('f', 2, int)
+
+/*
+ * Space reservation / allocation / free ioctls and argument structure
+ * are designed to be compatible with XFS.
+ *
+ * ALLOCSP* and FREESP* are not and will never be supported, but are
+ * included here for completeness.
+ */
+struct ocfs2_space_resv {
+	__s16		l_type;
+	__s16		l_whence;
+	__s64		l_start;
+	__s64		l_len;		/* len == 0 means until end of file */
+	__s32		l_sysid;
+	__u32		l_pid;
+	__s32		l_pad[4];	/* reserve area			    */
+};
+
+#define OCFS2_IOC_ALLOCSP		_IOW ('X', 10, struct ocfs2_space_resv)
+#define OCFS2_IOC_FREESP		_IOW ('X', 11, struct ocfs2_space_resv)
+#define OCFS2_IOC_RESVSP		_IOW ('X', 40, struct ocfs2_space_resv)
+#define OCFS2_IOC_UNRESVSP	_IOW ('X', 41, struct ocfs2_space_resv)
+#define OCFS2_IOC_ALLOCSP64	_IOW ('X', 36, struct ocfs2_space_resv)
+#define OCFS2_IOC_FREESP64	_IOW ('X', 37, struct ocfs2_space_resv)
+#define OCFS2_IOC_RESVSP64	_IOW ('X', 42, struct ocfs2_space_resv)
+#define OCFS2_IOC_UNRESVSP64	_IOW ('X', 43, struct ocfs2_space_resv)
+
+/* Used to pass group descriptor data when online resize is done */
+struct ocfs2_new_group_input {
+	__u64 group;		/* Group descriptor's blkno. */
+	__u32 clusters;		/* Total number of clusters in this group */
+	__u32 frees;		/* Total free clusters in this group */
+	__u16 chain;		/* Chain for this group */
+	__u16 reserved1;
+	__u32 reserved2;
+};
+
+#define OCFS2_IOC_GROUP_EXTEND	_IOW('o', 1, int)
+#define OCFS2_IOC_GROUP_ADD	_IOW('o', 2,struct ocfs2_new_group_input)
+#define OCFS2_IOC_GROUP_ADD64	_IOW('o', 3,struct ocfs2_new_group_input)
+
+/* Used to pass 2 file names to reflink. */
+struct reflink_arguments {
+	__u64 old_path;
+	__u64 new_path;
+	__u64 preserve;
+};
+#define OCFS2_IOC_REFLINK	_IOW('o', 4, struct reflink_arguments)
+
+#endif /* OCFS2_IOCTL_H */
diff --git a/fs/ocfs2/ocfs2_lockingver.h b/fs/ocfs2/ocfs2_lockingver.h
index 82d5eeac0fff..2e45c8d2ea7e 100644
--- a/fs/ocfs2/ocfs2_lockingver.h
+++ b/fs/ocfs2/ocfs2_lockingver.h
@@ -23,6 +23,8 @@
 /*
  * The protocol version for ocfs2 cluster locking.  See dlmglue.c for
  * more details.
+ *
+ * 1.0 - Initial locking version from ocfs2 1.4.
  */
 #define OCFS2_LOCKING_PROTOCOL_MAJOR 1
 #define OCFS2_LOCKING_PROTOCOL_MINOR 0
diff --git a/fs/ocfs2/quota.h b/fs/ocfs2/quota.h
index 123bc520a2c0..196fcb52d95d 100644
--- a/fs/ocfs2/quota.h
+++ b/fs/ocfs2/quota.h
@@ -23,6 +23,7 @@
 struct ocfs2_dquot {
 	struct dquot dq_dquot;	/* Generic VFS dquot */
 	loff_t dq_local_off;	/* Offset in the local quota file */
+	u64 dq_local_phys_blk;	/* Physical block carrying quota structure */
 	struct ocfs2_quota_chunk *dq_chunk;	/* Chunk dquot is in */
 	unsigned int dq_use_count;	/* Number of nodes having reference to this entry in global quota file */
 	s64 dq_origspace;	/* Last globally synced space usage */
@@ -51,8 +52,9 @@ struct ocfs2_mem_dqinfo {
 	struct ocfs2_lock_res dqi_gqlock;	/* Lock protecting quota information structure */
 	struct buffer_head *dqi_gqi_bh;	/* Buffer head with global quota file inode - set only if inode lock is obtained */
 	int dqi_gqi_count;		/* Number of holders of dqi_gqi_bh */
+	u64 dqi_giblk;			/* Number of block with global information header */
 	struct buffer_head *dqi_lqi_bh;	/* Buffer head with local quota file inode */
-	struct buffer_head *dqi_ibh;	/* Buffer with information header */
+	struct buffer_head *dqi_libh;	/* Buffer with local information header */
 	struct qtree_mem_dqinfo dqi_gi;	/* Info about global file */
 	struct delayed_work dqi_sync_work;	/* Work for syncing dquots */
 	struct ocfs2_quota_recovery *dqi_rec;	/* Pointer to recovery
@@ -102,8 +104,12 @@ static inline int ocfs2_global_release_dquot(struct dquot *dquot)
 
 int ocfs2_lock_global_qf(struct ocfs2_mem_dqinfo *oinfo, int ex);
 void ocfs2_unlock_global_qf(struct ocfs2_mem_dqinfo *oinfo, int ex);
-int ocfs2_read_quota_block(struct inode *inode, u64 v_block,
-			   struct buffer_head **bh);
+int ocfs2_validate_quota_block(struct super_block *sb, struct buffer_head *bh);
+int ocfs2_read_quota_phys_block(struct inode *inode, u64 p_block,
+				struct buffer_head **bh);
+int ocfs2_create_local_dquot(struct dquot *dquot);
+int ocfs2_local_release_dquot(handle_t *handle, struct dquot *dquot);
+int ocfs2_local_write_dquot(struct dquot *dquot);
 
 extern const struct dquot_operations ocfs2_quota_operations;
 extern struct quota_format_type ocfs2_quota_format;
diff --git a/fs/ocfs2/quota_global.c b/fs/ocfs2/quota_global.c
index b437dc0c4cad..4607923eb24c 100644
--- a/fs/ocfs2/quota_global.c
+++ b/fs/ocfs2/quota_global.c
@@ -3,6 +3,7 @@
  */
 #include <linux/spinlock.h>
 #include <linux/fs.h>
+#include <linux/slab.h>
 #include <linux/quota.h>
 #include <linux/quotaops.h>
 #include <linux/dqblk_qtree.h>
@@ -24,8 +25,44 @@
 #include "dlmglue.h"
 #include "uptodate.h"
 #include "super.h"
+#include "buffer_head_io.h"
 #include "quota.h"
 
+/*
+ * Locking of quotas with OCFS2 is rather complex. Here are rules that
+ * should be obeyed by all the functions:
+ * - any write of quota structure (either to local or global file) is protected
+ *   by dqio_mutex or dquot->dq_lock.
+ * - any modification of global quota file holds inode cluster lock, i_mutex,
+ *   and ip_alloc_sem of the global quota file (achieved by
+ *   ocfs2_lock_global_qf). It also has to hold qinfo_lock.
+ * - an allocation of new blocks for local quota file is protected by
+ *   its ip_alloc_sem
+ *
+ * A rough sketch of locking dependencies (lf = local file, gf = global file):
+ * Normal filesystem operation:
+ *   start_trans -> dqio_mutex -> write to lf
+ * Syncing of local and global file:
+ *   ocfs2_lock_global_qf -> start_trans -> dqio_mutex -> qinfo_lock ->
+ *     write to gf
+ *						       -> write to lf
+ * Acquire dquot for the first time:
+ *   dq_lock -> ocfs2_lock_global_qf -> qinfo_lock -> read from gf
+ *				     -> alloc space for gf
+ *				     -> start_trans -> qinfo_lock -> write to gf
+ *	     -> ip_alloc_sem of lf -> alloc space for lf
+ *	     -> write to lf
+ * Release last reference to dquot:
+ *   dq_lock -> ocfs2_lock_global_qf -> start_trans -> qinfo_lock -> write to gf
+ *	     -> write to lf
+ * Note that all the above operations also hold the inode cluster lock of lf.
+ * Recovery:
+ *   inode cluster lock of recovered lf
+ *     -> read bitmaps -> ip_alloc_sem of lf
+ *     -> ocfs2_lock_global_qf -> start_trans -> dqio_mutex -> qinfo_lock ->
+ *        write to gf
+ */
+
 static struct workqueue_struct *ocfs2_quota_wq = NULL;
 
 static void qsync_work_fn(struct work_struct *work);
@@ -90,8 +127,7 @@ struct qtree_fmt_operations ocfs2_global_ops = {
 	.is_id = ocfs2_global_is_id,
 };
 
-static int ocfs2_validate_quota_block(struct super_block *sb,
-				      struct buffer_head *bh)
+int ocfs2_validate_quota_block(struct super_block *sb, struct buffer_head *bh)
 {
 	struct ocfs2_disk_dqtrailer *dqt =
 		ocfs2_block_dqtrailer(sb->s_blocksize, bh->b_data);
@@ -109,54 +145,19 @@ static int ocfs2_validate_quota_block(struct super_block *sb,
 	return ocfs2_validate_meta_ecc(sb, bh->b_data, &dqt->dq_check);
 }
 
-int ocfs2_read_quota_block(struct inode *inode, u64 v_block,
-			   struct buffer_head **bh)
+int ocfs2_read_quota_phys_block(struct inode *inode, u64 p_block,
+				struct buffer_head **bhp)
 {
-	int rc = 0;
-	struct buffer_head *tmp = *bh;
-
-	if (i_size_read(inode) >> inode->i_sb->s_blocksize_bits <= v_block) {
-		ocfs2_error(inode->i_sb,
-			    "Quota file %llu is probably corrupted! Requested "
-			    "to read block %Lu but file has size only %Lu\n",
-			    (unsigned long long)OCFS2_I(inode)->ip_blkno,
-			    (unsigned long long)v_block,
-			    (unsigned long long)i_size_read(inode));
-		return -EIO;
-	}
-	rc = ocfs2_read_virt_blocks(inode, v_block, 1, &tmp, 0,
-				    ocfs2_validate_quota_block);
+	int rc;
+
+	*bhp = NULL;
+	rc = ocfs2_read_blocks(INODE_CACHE(inode), p_block, 1, bhp, 0,
+			       ocfs2_validate_quota_block);
 	if (rc)
 		mlog_errno(rc);
-
-	/* If ocfs2_read_virt_blocks() got us a new bh, pass it up. */
-	if (!rc && !*bh)
-		*bh = tmp;
-
 	return rc;
 }
 
-static int ocfs2_get_quota_block(struct inode *inode, int block,
-				 struct buffer_head **bh)
-{
-	u64 pblock, pcount;
-	int err;
-
-	down_read(&OCFS2_I(inode)->ip_alloc_sem);
-	err = ocfs2_extent_map_get_blocks(inode, block, &pblock, &pcount, NULL);
-	up_read(&OCFS2_I(inode)->ip_alloc_sem);
-	if (err) {
-		mlog_errno(err);
-		return err;
-	}
-	*bh = sb_getblk(inode->i_sb, pblock);
-	if (!*bh) {
-		err = -EIO;
-		mlog_errno(err);
-	}
-	return err;
-}
-
 /* Read data from global quotafile - avoid pagecache and such because we cannot
  * afford acquiring the locks... We use quota cluster lock to serialize
  * operations. Caller is responsible for acquiring it. */
@@ -171,6 +172,7 @@ ssize_t ocfs2_quota_read(struct super_block *sb, int type, char *data,
 	int err = 0;
 	struct buffer_head *bh;
 	size_t toread, tocopy;
+	u64 pblock = 0, pcount = 0;
 
 	if (off > i_size)
 		return 0;
@@ -179,8 +181,19 @@ ssize_t ocfs2_quota_read(struct super_block *sb, int type, char *data,
 	toread = len;
 	while (toread > 0) {
 		tocopy = min_t(size_t, (sb->s_blocksize - offset), toread);
+		if (!pcount) {
+			err = ocfs2_extent_map_get_blocks(gqinode, blk, &pblock,
+							  &pcount, NULL);
+			if (err) {
+				mlog_errno(err);
+				return err;
+			}
+		} else {
+			pcount--;
+			pblock++;
+		}
 		bh = NULL;
-		err = ocfs2_read_quota_block(gqinode, blk, &bh);
+		err = ocfs2_read_quota_phys_block(gqinode, pblock, &bh);
 		if (err) {
 			mlog_errno(err);
 			return err;
@@ -208,6 +221,7 @@ ssize_t ocfs2_quota_write(struct super_block *sb, int type,
 	int err = 0, new = 0, ja_type;
 	struct buffer_head *bh = NULL;
 	handle_t *handle = journal_current_handle();
+	u64 pblock, pcount;
 
 	if (!handle) {
 		mlog(ML_ERROR, "Quota write (off=%llu, len=%llu) cancelled "
@@ -220,12 +234,11 @@ ssize_t ocfs2_quota_write(struct super_block *sb, int type,
 		len = sb->s_blocksize - OCFS2_QBLK_RESERVED_SPACE - offset;
 	}
 
-	mutex_lock_nested(&gqinode->i_mutex, I_MUTEX_QUOTA);
 	if (gqinode->i_size < off + len) {
 		loff_t rounded_end =
 				ocfs2_align_bytes_to_blocks(sb, off + len);
 
-		/* Space is already allocated in ocfs2_global_read_dquot() */
+		/* Space is already allocated in ocfs2_acquire_dquot() */
 		err = ocfs2_simple_size_update(gqinode,
 					       oinfo->dqi_gqi_bh,
 					       rounded_end);
@@ -233,13 +246,20 @@ ssize_t ocfs2_quota_write(struct super_block *sb, int type,
 			goto out;
 		new = 1;
 	}
+	err = ocfs2_extent_map_get_blocks(gqinode, blk, &pblock, &pcount, NULL);
+	if (err) {
+		mlog_errno(err);
+		goto out;
+	}
 	/* Not rewriting whole block? */
 	if ((offset || len < sb->s_blocksize - OCFS2_QBLK_RESERVED_SPACE) &&
 	    !new) {
-		err = ocfs2_read_quota_block(gqinode, blk, &bh);
+		err = ocfs2_read_quota_phys_block(gqinode, pblock, &bh);
 		ja_type = OCFS2_JOURNAL_ACCESS_WRITE;
 	} else {
-		err = ocfs2_get_quota_block(gqinode, blk, &bh);
+		bh = sb_getblk(sb, pblock);
+		if (!bh)
+			err = -ENOMEM;
 		ja_type = OCFS2_JOURNAL_ACCESS_CREATE;
 	}
 	if (err) {
@@ -260,19 +280,15 @@ ssize_t ocfs2_quota_write(struct super_block *sb, int type,
 		brelse(bh);
 		goto out;
 	}
-	err = ocfs2_journal_dirty(handle, bh);
+	ocfs2_journal_dirty(handle, bh);
 	brelse(bh);
-	if (err < 0)
-		goto out;
 out:
 	if (err) {
-		mutex_unlock(&gqinode->i_mutex);
 		mlog_errno(err);
 		return err;
 	}
 	gqinode->i_version++;
 	ocfs2_mark_inode_dirty(handle, gqinode, oinfo->dqi_gqi_bh);
-	mutex_unlock(&gqinode->i_mutex);
 	return len;
 }
 
@@ -290,11 +306,23 @@ int ocfs2_lock_global_qf(struct ocfs2_mem_dqinfo *oinfo, int ex)
 	else
 		WARN_ON(bh != oinfo->dqi_gqi_bh);
 	spin_unlock(&dq_data_lock);
+	if (ex) {
+		mutex_lock(&oinfo->dqi_gqinode->i_mutex);
+		down_write(&OCFS2_I(oinfo->dqi_gqinode)->ip_alloc_sem);
+	} else {
+		down_read(&OCFS2_I(oinfo->dqi_gqinode)->ip_alloc_sem);
+	}
 	return 0;
 }
 
 void ocfs2_unlock_global_qf(struct ocfs2_mem_dqinfo *oinfo, int ex)
 {
+	if (ex) {
+		up_write(&OCFS2_I(oinfo->dqi_gqinode)->ip_alloc_sem);
+		mutex_unlock(&oinfo->dqi_gqinode->i_mutex);
+	} else {
+		up_read(&OCFS2_I(oinfo->dqi_gqinode)->ip_alloc_sem);
+	}
 	ocfs2_inode_unlock(oinfo->dqi_gqinode, ex);
 	brelse(oinfo->dqi_gqi_bh);
 	spin_lock(&dq_data_lock);
@@ -312,6 +340,7 @@ int ocfs2_global_read_info(struct super_block *sb, int type)
 	struct ocfs2_global_disk_dqinfo dinfo;
 	struct mem_dqinfo *info = sb_dqinfo(sb, type);
 	struct ocfs2_mem_dqinfo *oinfo = info->dqi_priv;
+	u64 pcount;
 	int status;
 
 	mlog_entry_void();
@@ -338,9 +367,19 @@ int ocfs2_global_read_info(struct super_block *sb, int type)
 		mlog_errno(status);
 		goto out_err;
 	}
+
+	status = ocfs2_extent_map_get_blocks(gqinode, 0, &oinfo->dqi_giblk,
+					     &pcount, NULL);
+	if (status < 0)
+		goto out_unlock;
+
+	status = ocfs2_qinfo_lock(oinfo, 0);
+	if (status < 0)
+		goto out_unlock;
 	status = sb->s_op->quota_read(sb, type, (char *)&dinfo,
 				      sizeof(struct ocfs2_global_disk_dqinfo),
 				      OCFS2_GLOBAL_INFO_OFF);
+	ocfs2_qinfo_unlock(oinfo, 0);
 	ocfs2_unlock_global_qf(oinfo, 0);
 	if (status != sizeof(struct ocfs2_global_disk_dqinfo)) {
 		mlog(ML_ERROR, "Cannot read global quota info (%d).\n",
@@ -367,6 +406,10 @@ int ocfs2_global_read_info(struct super_block *sb, int type)
 out_err:
 	mlog_exit(status);
 	return status;
+out_unlock:
+	ocfs2_unlock_global_qf(oinfo, 0);
+	mlog_errno(status);
+	goto out_err;
 }
 
 /* Write information to global quota file. Expects exlusive lock on quota
@@ -425,78 +468,10 @@ static int ocfs2_global_qinit_alloc(struct super_block *sb, int type)
 
 static int ocfs2_calc_global_qinit_credits(struct super_block *sb, int type)
 {
-	/* We modify all the allocated blocks, tree root, and info block */
+	/* We modify all the allocated blocks, tree root, info block and
+	 * the inode */
 	return (ocfs2_global_qinit_alloc(sb, type) + 2) *
-			OCFS2_QUOTA_BLOCK_UPDATE_CREDITS;
-}
-
-/* Read in information from global quota file and acquire a reference to it.
- * dquot_acquire() has already started the transaction and locked quota file */
-int ocfs2_global_read_dquot(struct dquot *dquot)
-{
-	int err, err2, ex = 0;
-	struct super_block *sb = dquot->dq_sb;
-	int type = dquot->dq_type;
-	struct ocfs2_mem_dqinfo *info = sb_dqinfo(sb, type)->dqi_priv;
-	struct ocfs2_super *osb = OCFS2_SB(sb);
-	struct inode *gqinode = info->dqi_gqinode;
-	int need_alloc = ocfs2_global_qinit_alloc(sb, type);
-	handle_t *handle = NULL;
-
-	err = ocfs2_qinfo_lock(info, 0);
-	if (err < 0)
-		goto out;
-	err = qtree_read_dquot(&info->dqi_gi, dquot);
-	if (err < 0)
-		goto out_qlock;
-	OCFS2_DQUOT(dquot)->dq_use_count++;
-	OCFS2_DQUOT(dquot)->dq_origspace = dquot->dq_dqb.dqb_curspace;
-	OCFS2_DQUOT(dquot)->dq_originodes = dquot->dq_dqb.dqb_curinodes;
-	ocfs2_qinfo_unlock(info, 0);
-
-	if (!dquot->dq_off) {	/* No real quota entry? */
-		ex = 1;
-		/*
-		 * Add blocks to quota file before we start a transaction since
-		 * locking allocators ranks above a transaction start
-		 */
-		WARN_ON(journal_current_handle());
-		down_write(&OCFS2_I(gqinode)->ip_alloc_sem);
-		err = ocfs2_extend_no_holes(gqinode,
-			gqinode->i_size + (need_alloc << sb->s_blocksize_bits),
-			gqinode->i_size);
-		up_write(&OCFS2_I(gqinode)->ip_alloc_sem);
-		if (err < 0)
-			goto out;
-	}
-
-	handle = ocfs2_start_trans(osb,
-				   ocfs2_calc_global_qinit_credits(sb, type));
-	if (IS_ERR(handle)) {
-		err = PTR_ERR(handle);
-		goto out;
-	}
-	err = ocfs2_qinfo_lock(info, ex);
-	if (err < 0)
-		goto out_trans;
-	err = qtree_write_dquot(&info->dqi_gi, dquot);
-	if (ex && info_dirty(sb_dqinfo(dquot->dq_sb, dquot->dq_type))) {
-		err2 = __ocfs2_global_write_info(dquot->dq_sb, dquot->dq_type);
-		if (!err)
-			err = err2;
-	}
-out_qlock:
-	if (ex)
-		ocfs2_qinfo_unlock(info, 1);
-	else
-		ocfs2_qinfo_unlock(info, 0);
-out_trans:
-	if (handle)
-		ocfs2_commit_trans(osb, handle);
-out:
-	if (err < 0)
-		mlog_errno(err);
-	return err;
+			OCFS2_QUOTA_BLOCK_UPDATE_CREDITS + 1;
 }
 
 /* Sync local information about quota modifications with global quota file.
@@ -637,14 +612,13 @@ static int ocfs2_sync_dquot_helper(struct dquot *dquot, unsigned long type)
 	}
 	mutex_lock(&sb_dqopt(sb)->dqio_mutex);
 	status = ocfs2_sync_dquot(dquot);
-	mutex_unlock(&sb_dqopt(sb)->dqio_mutex);
 	if (status < 0)
 		mlog_errno(status);
 	/* We have to write local structure as well... */
-	dquot_mark_dquot_dirty(dquot);
-	status = dquot_commit(dquot);
+	status = ocfs2_local_write_dquot(dquot);
 	if (status < 0)
 		mlog_errno(status);
+	mutex_unlock(&sb_dqopt(sb)->dqio_mutex);
 	ocfs2_commit_trans(osb, handle);
 out_ilock:
 	ocfs2_unlock_global_qf(oinfo, 1);
@@ -683,7 +657,9 @@ static int ocfs2_write_dquot(struct dquot *dquot)
 		mlog_errno(status);
 		goto out;
 	}
-	status = dquot_commit(dquot);
+	mutex_lock(&sb_dqopt(dquot->dq_sb)->dqio_mutex);
+	status = ocfs2_local_write_dquot(dquot);
+	mutex_unlock(&sb_dqopt(dquot->dq_sb)->dqio_mutex);
 	ocfs2_commit_trans(osb, handle);
 out:
 	mlog_exit(status);
@@ -714,6 +690,10 @@ static int ocfs2_release_dquot(struct dquot *dquot)
 
 	mlog_entry("id=%u, type=%d", dquot->dq_id, dquot->dq_type);
 
+	mutex_lock(&dquot->dq_lock);
+	/* Check whether we are not racing with some other dqget() */
+	if (atomic_read(&dquot->dq_count) > 1)
+		goto out;
 	status = ocfs2_lock_global_qf(oinfo, 1);
 	if (status < 0)
 		goto out;
@@ -724,30 +704,113 @@ static int ocfs2_release_dquot(struct dquot *dquot)
 		mlog_errno(status);
 		goto out_ilock;
 	}
-	status = dquot_release(dquot);
+
+	status = ocfs2_global_release_dquot(dquot);
+	if (status < 0) {
+		mlog_errno(status);
+		goto out_trans;
+	}
+	status = ocfs2_local_release_dquot(handle, dquot);
+	/*
+	 * If we fail here, we cannot do much as global structure is
+	 * already released. So just complain...
+	 */
+	if (status < 0)
+		mlog_errno(status);
+	clear_bit(DQ_ACTIVE_B, &dquot->dq_flags);
+out_trans:
 	ocfs2_commit_trans(osb, handle);
 out_ilock:
 	ocfs2_unlock_global_qf(oinfo, 1);
 out:
+	mutex_unlock(&dquot->dq_lock);
 	mlog_exit(status);
 	return status;
 }
 
+/*
+ * Read global dquot structure from disk or create it if it does
+ * not exist. Also update use count of the global structure and
+ * create structure in node-local quota file.
+ */
 static int ocfs2_acquire_dquot(struct dquot *dquot)
 {
-	struct ocfs2_mem_dqinfo *oinfo =
-			sb_dqinfo(dquot->dq_sb, dquot->dq_type)->dqi_priv;
-	int status = 0;
+	int status = 0, err;
+	int ex = 0;
+	struct super_block *sb = dquot->dq_sb;
+	struct ocfs2_super *osb = OCFS2_SB(sb);
+	int type = dquot->dq_type;
+	struct ocfs2_mem_dqinfo *info = sb_dqinfo(sb, type)->dqi_priv;
+	struct inode *gqinode = info->dqi_gqinode;
+	int need_alloc = ocfs2_global_qinit_alloc(sb, type);
+	handle_t *handle;
 
-	mlog_entry("id=%u, type=%d", dquot->dq_id, dquot->dq_type);
-	/* We need an exclusive lock, because we're going to update use count
-	 * and instantiate possibly new dquot structure */
-	status = ocfs2_lock_global_qf(oinfo, 1);
+	mlog_entry("id=%u, type=%d", dquot->dq_id, type);
+	mutex_lock(&dquot->dq_lock);
+	/*
+	 * We need an exclusive lock, because we're going to update use count
+	 * and instantiate possibly new dquot structure
+	 */
+	status = ocfs2_lock_global_qf(info, 1);
 	if (status < 0)
 		goto out;
-	status = dquot_acquire(dquot);
-	ocfs2_unlock_global_qf(oinfo, 1);
+	if (!test_bit(DQ_READ_B, &dquot->dq_flags)) {
+		status = ocfs2_qinfo_lock(info, 0);
+		if (status < 0)
+			goto out_dq;
+		status = qtree_read_dquot(&info->dqi_gi, dquot);
+		ocfs2_qinfo_unlock(info, 0);
+		if (status < 0)
+			goto out_dq;
+	}
+	set_bit(DQ_READ_B, &dquot->dq_flags);
+
+	OCFS2_DQUOT(dquot)->dq_use_count++;
+	OCFS2_DQUOT(dquot)->dq_origspace = dquot->dq_dqb.dqb_curspace;
+	OCFS2_DQUOT(dquot)->dq_originodes = dquot->dq_dqb.dqb_curinodes;
+	if (!dquot->dq_off) {	/* No real quota entry? */
+		ex = 1;
+		/*
+		 * Add blocks to quota file before we start a transaction since
+		 * locking allocators ranks above a transaction start
+		 */
+		WARN_ON(journal_current_handle());
+		status = ocfs2_extend_no_holes(gqinode, NULL,
+			gqinode->i_size + (need_alloc << sb->s_blocksize_bits),
+			gqinode->i_size);
+		if (status < 0)
+			goto out_dq;
+	}
+
+	handle = ocfs2_start_trans(osb,
+				   ocfs2_calc_global_qinit_credits(sb, type));
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		goto out_dq;
+	}
+	status = ocfs2_qinfo_lock(info, ex);
+	if (status < 0)
+		goto out_trans;
+	status = qtree_write_dquot(&info->dqi_gi, dquot);
+	if (ex && info_dirty(sb_dqinfo(sb, type))) {
+		err = __ocfs2_global_write_info(sb, type);
+		if (!status)
+			status = err;
+	}
+	ocfs2_qinfo_unlock(info, ex);
+out_trans:
+	ocfs2_commit_trans(osb, handle);
+out_dq:
+	ocfs2_unlock_global_qf(info, 1);
+	if (status < 0)
+		goto out;
+
+	status = ocfs2_create_local_dquot(dquot);
+	if (status < 0)
+		goto out;
+	set_bit(DQ_ACTIVE_B, &dquot->dq_flags);
 out:
+	mutex_unlock(&dquot->dq_lock);
 	mlog_exit(status);
 	return status;
 }
@@ -769,7 +832,6 @@ static int ocfs2_mark_dquot_dirty(struct dquot *dquot)
 	struct ocfs2_super *osb = OCFS2_SB(sb);
 
 	mlog_entry("id=%u, type=%d", dquot->dq_id, type);
-	dquot_mark_dquot_dirty(dquot);
 
 	/* In case user set some limits, sync dquot immediately to global
 	 * quota file so that information propagates quicker */
@@ -792,14 +854,16 @@ static int ocfs2_mark_dquot_dirty(struct dquot *dquot)
 		mlog_errno(status);
 		goto out_ilock;
 	}
+	mutex_lock(&sb_dqopt(sb)->dqio_mutex);
 	status = ocfs2_sync_dquot(dquot);
 	if (status < 0) {
 		mlog_errno(status);
-		goto out_trans;
+		goto out_dlock;
 	}
 	/* Now write updated local dquot structure */
-	status = dquot_commit(dquot);
-out_trans:
+	status = ocfs2_local_write_dquot(dquot);
+out_dlock:
+	mutex_unlock(&sb_dqopt(sb)->dqio_mutex);
 	ocfs2_commit_trans(osb, handle);
 out_ilock:
 	ocfs2_unlock_global_qf(oinfo, 1);
@@ -851,14 +915,7 @@ static void ocfs2_destroy_dquot(struct dquot *dquot)
 }
 
 const struct dquot_operations ocfs2_quota_operations = {
-	.initialize	= dquot_initialize,
-	.drop		= dquot_drop,
-	.alloc_space	= dquot_alloc_space,
-	.alloc_inode	= dquot_alloc_inode,
-	.free_space	= dquot_free_space,
-	.free_inode	= dquot_free_inode,
-	.transfer	= dquot_transfer,
-	.write_dquot	= ocfs2_write_dquot,
+	/* We never make dquot dirty so .write_dquot is never called */
 	.acquire_dquot	= ocfs2_acquire_dquot,
 	.release_dquot	= ocfs2_release_dquot,
 	.mark_dirty	= ocfs2_mark_dquot_dirty,
diff --git a/fs/ocfs2/quota_local.c b/fs/ocfs2/quota_local.c
index 21f9e71223ca..dc78764ccc4c 100644
--- a/fs/ocfs2/quota_local.c
+++ b/fs/ocfs2/quota_local.c
@@ -3,6 +3,7 @@
  */
 
 #include <linux/fs.h>
+#include <linux/slab.h>
 #include <linux/quota.h>
 #include <linux/quotaops.h>
 #include <linux/module.h>
@@ -21,6 +22,7 @@
 #include "dlmglue.h"
 #include "quota.h"
 #include "uptodate.h"
+#include "super.h"
 
 /* Number of local quota structures per block */
 static inline unsigned int ol_quota_entries_per_block(struct super_block *sb)
@@ -118,12 +120,8 @@ static int ocfs2_modify_bh(struct inode *inode, struct buffer_head *bh,
 	lock_buffer(bh);
 	modify(bh, private);
 	unlock_buffer(bh);
-	status = ocfs2_journal_dirty(handle, bh);
-	if (status < 0) {
-		mlog_errno(status);
-		ocfs2_commit_trans(OCFS2_SB(sb), handle);
-		return status;
-	}
+	ocfs2_journal_dirty(handle, bh);
+
 	status = ocfs2_commit_trans(OCFS2_SB(sb), handle);
 	if (status < 0) {
 		mlog_errno(status);
@@ -132,6 +130,39 @@ static int ocfs2_modify_bh(struct inode *inode, struct buffer_head *bh,
 	return 0;
 }
 
+/*
+ * Read quota block from a given logical offset.
+ *
+ * This function acquires ip_alloc_sem and thus it must not be called with a
+ * transaction started.
+ */
+static int ocfs2_read_quota_block(struct inode *inode, u64 v_block,
+				  struct buffer_head **bh)
+{
+	int rc = 0;
+	struct buffer_head *tmp = *bh;
+
+	if (i_size_read(inode) >> inode->i_sb->s_blocksize_bits <= v_block) {
+		ocfs2_error(inode->i_sb,
+			    "Quota file %llu is probably corrupted! Requested "
+			    "to read block %Lu but file has size only %Lu\n",
+			    (unsigned long long)OCFS2_I(inode)->ip_blkno,
+			    (unsigned long long)v_block,
+			    (unsigned long long)i_size_read(inode));
+		return -EIO;
+	}
+	rc = ocfs2_read_virt_blocks(inode, v_block, 1, &tmp, 0,
+				    ocfs2_validate_quota_block);
+	if (rc)
+		mlog_errno(rc);
+
+	/* If ocfs2_read_virt_blocks() got us a new bh, pass it up. */
+	if (!rc && !*bh)
+		*bh = tmp;
+
+	return rc;
+}
+
 /* Check whether we understand format of quota files */
 static int ocfs2_local_check_quota_file(struct super_block *sb, int type)
 {
@@ -457,7 +488,7 @@ static int ocfs2_recover_local_quota_file(struct inode *lqinode,
 			break;
 		}
 		dchunk = (struct ocfs2_local_disk_chunk *)hbh->b_data;
-		for_each_bit(bit, rchunk->rc_bitmap, ol_chunk_entries(sb)) {
+		for_each_set_bit(bit, rchunk->rc_bitmap, ol_chunk_entries(sb)) {
 			qbh = NULL;
 			status = ocfs2_read_quota_block(lqinode,
 						ol_dqblk_block(sb, chunk, bit),
@@ -522,9 +553,7 @@ static int ocfs2_recover_local_quota_file(struct inode *lqinode,
 			ocfs2_clear_bit(bit, dchunk->dqc_bitmap);
 			le32_add_cpu(&dchunk->dqc_free, 1);
 			unlock_buffer(qbh);
-			status = ocfs2_journal_dirty(handle, qbh);
-			if (status < 0)
-				mlog_errno(status);
+			ocfs2_journal_dirty(handle, qbh);
 out_commit:
 			mutex_unlock(&sb_dqopt(sb)->dqio_mutex);
 			ocfs2_commit_trans(OCFS2_SB(sb), handle);
@@ -630,9 +659,7 @@ int ocfs2_finish_quota_recovery(struct ocfs2_super *osb,
 		lock_buffer(bh);
 		ldinfo->dqi_flags = cpu_to_le32(flags | OLQF_CLEAN);
 		unlock_buffer(bh);
-		status = ocfs2_journal_dirty(handle, bh);
-		if (status < 0)
-			mlog_errno(status);
+		ocfs2_journal_dirty(handle, bh);
 out_trans:
 		ocfs2_commit_trans(osb, handle);
 out_bh:
@@ -678,7 +705,7 @@ static int ocfs2_local_read_info(struct super_block *sb, int type)
 	INIT_LIST_HEAD(&oinfo->dqi_chunk);
 	oinfo->dqi_rec = NULL;
 	oinfo->dqi_lqi_bh = NULL;
-	oinfo->dqi_ibh = NULL;
+	oinfo->dqi_libh = NULL;
 
 	status = ocfs2_global_read_info(sb, type);
 	if (status < 0)
@@ -704,7 +731,7 @@ static int ocfs2_local_read_info(struct super_block *sb, int type)
 	info->dqi_flags = le32_to_cpu(ldinfo->dqi_flags);
 	oinfo->dqi_chunks = le32_to_cpu(ldinfo->dqi_chunks);
 	oinfo->dqi_blocks = le32_to_cpu(ldinfo->dqi_blocks);
-	oinfo->dqi_ibh = bh;
+	oinfo->dqi_libh = bh;
 
 	/* We crashed when using local quota file? */
 	if (!(info->dqi_flags & OLQF_CLEAN)) {
@@ -766,7 +793,7 @@ static int ocfs2_local_write_info(struct super_block *sb, int type)
 {
 	struct mem_dqinfo *info = sb_dqinfo(sb, type);
 	struct buffer_head *bh = ((struct ocfs2_mem_dqinfo *)info->dqi_priv)
-						->dqi_ibh;
+						->dqi_libh;
 	int status;
 
 	status = ocfs2_modify_bh(sb_dqopt(sb)->files[type], bh, olq_update_info,
@@ -789,10 +816,6 @@ static int ocfs2_local_free_info(struct super_block *sb, int type)
 	int mark_clean = 1, len;
 	int status;
 
-	/* At this point we know there are no more dquots and thus
-	 * even if there's some sync in the pdflush queue, it won't
-	 * find any dquots and return without doing anything */
-	cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 	iput(oinfo->dqi_gqinode);
 	ocfs2_simple_drop_lockres(OCFS2_SB(sb), &oinfo->dqi_gqlock);
 	ocfs2_lock_res_free(&oinfo->dqi_gqlock);
@@ -827,7 +850,7 @@ static int ocfs2_local_free_info(struct super_block *sb, int type)
 	/* Mark local file as clean */
 	info->dqi_flags |= OLQF_CLEAN;
 	status = ocfs2_modify_bh(sb_dqopt(sb)->files[type],
-				 oinfo->dqi_ibh,
+				 oinfo->dqi_libh,
 				 olq_update_info,
 				 info);
 	if (status < 0) {
@@ -837,7 +860,7 @@ static int ocfs2_local_free_info(struct super_block *sb, int type)
 
 out:
 	ocfs2_inode_unlock(sb_dqopt(sb)->files[type], 1);
-	brelse(oinfo->dqi_ibh);
+	brelse(oinfo->dqi_libh);
 	brelse(oinfo->dqi_lqi_bh);
 	kfree(oinfo);
 	return 0;
@@ -865,22 +888,21 @@ static void olq_set_dquot(struct buffer_head *bh, void *private)
 }
 
 /* Write dquot to local quota file */
-static int ocfs2_local_write_dquot(struct dquot *dquot)
+int ocfs2_local_write_dquot(struct dquot *dquot)
 {
 	struct super_block *sb = dquot->dq_sb;
 	struct ocfs2_dquot *od = OCFS2_DQUOT(dquot);
-	struct buffer_head *bh = NULL;
+	struct buffer_head *bh;
+	struct inode *lqinode = sb_dqopt(sb)->files[dquot->dq_type];
 	int status;
 
-	status = ocfs2_read_quota_block(sb_dqopt(sb)->files[dquot->dq_type],
-				    ol_dqblk_file_block(sb, od->dq_local_off),
-				    &bh);
+	status = ocfs2_read_quota_phys_block(lqinode, od->dq_local_phys_blk,
+					     &bh);
 	if (status) {
 		mlog_errno(status);
 		goto out;
 	}
-	status = ocfs2_modify_bh(sb_dqopt(sb)->files[dquot->dq_type], bh,
-				 olq_set_dquot, od);
+	status = ocfs2_modify_bh(lqinode, bh, olq_set_dquot, od);
 	if (status < 0) {
 		mlog_errno(status);
 		goto out;
@@ -949,7 +971,7 @@ static struct ocfs2_quota_chunk *ocfs2_local_quota_add_chunk(
 	u64 p_blkno;
 
 	/* We are protected by dqio_sem so no locking needed */
-	status = ocfs2_extend_no_holes(lqinode,
+	status = ocfs2_extend_no_holes(lqinode, NULL,
 				       lqinode->i_size + 2 * sb->s_blocksize,
 				       lqinode->i_size);
 	if (status < 0) {
@@ -980,10 +1002,8 @@ static struct ocfs2_quota_chunk *ocfs2_local_quota_add_chunk(
 	}
 
 	/* Initialize chunk header */
-	down_read(&OCFS2_I(lqinode)->ip_alloc_sem);
 	status = ocfs2_extent_map_get_blocks(lqinode, oinfo->dqi_blocks,
 					     &p_blkno, NULL, NULL);
-	up_read(&OCFS2_I(lqinode)->ip_alloc_sem);
 	if (status < 0) {
 		mlog_errno(status);
 		goto out_trans;
@@ -1008,17 +1028,11 @@ static struct ocfs2_quota_chunk *ocfs2_local_quota_add_chunk(
 	       sb->s_blocksize - sizeof(struct ocfs2_local_disk_chunk) -
 	       OCFS2_QBLK_RESERVED_SPACE);
 	unlock_buffer(bh);
-	status = ocfs2_journal_dirty(handle, bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out_trans;
-	}
+	ocfs2_journal_dirty(handle, bh);
 
 	/* Initialize new block with structures */
-	down_read(&OCFS2_I(lqinode)->ip_alloc_sem);
 	status = ocfs2_extent_map_get_blocks(lqinode, oinfo->dqi_blocks + 1,
 					     &p_blkno, NULL, NULL);
-	up_read(&OCFS2_I(lqinode)->ip_alloc_sem);
 	if (status < 0) {
 		mlog_errno(status);
 		goto out_trans;
@@ -1039,11 +1053,7 @@ static struct ocfs2_quota_chunk *ocfs2_local_quota_add_chunk(
 	lock_buffer(dbh);
 	memset(dbh->b_data, 0, sb->s_blocksize - OCFS2_QBLK_RESERVED_SPACE);
 	unlock_buffer(dbh);
-	status = ocfs2_journal_dirty(handle, dbh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out_trans;
-	}
+	ocfs2_journal_dirty(handle, dbh);
 
 	/* Update local quotafile info */
 	oinfo->dqi_blocks += 2;
@@ -1104,7 +1114,7 @@ static struct ocfs2_quota_chunk *ocfs2_extend_local_quota_file(
 		return ocfs2_local_quota_add_chunk(sb, type, offset);
 
 	/* We are protected by dqio_sem so no locking needed */
-	status = ocfs2_extend_no_holes(lqinode,
+	status = ocfs2_extend_no_holes(lqinode, NULL,
 				       lqinode->i_size + sb->s_blocksize,
 				       lqinode->i_size);
 	if (status < 0) {
@@ -1119,10 +1129,8 @@ static struct ocfs2_quota_chunk *ocfs2_extend_local_quota_file(
 	}
 
 	/* Get buffer from the just added block */
-	down_read(&OCFS2_I(lqinode)->ip_alloc_sem);
 	status = ocfs2_extent_map_get_blocks(lqinode, oinfo->dqi_blocks,
 					     &p_blkno, NULL, NULL);
-	up_read(&OCFS2_I(lqinode)->ip_alloc_sem);
 	if (status < 0) {
 		mlog_errno(status);
 		goto out;
@@ -1154,11 +1162,8 @@ static struct ocfs2_quota_chunk *ocfs2_extend_local_quota_file(
 	lock_buffer(bh);
 	memset(bh->b_data, 0, sb->s_blocksize);
 	unlock_buffer(bh);
-	status = ocfs2_journal_dirty(handle, bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out_trans;
-	}
+	ocfs2_journal_dirty(handle, bh);
+
 	/* Update chunk header */
 	status = ocfs2_journal_access_dq(handle, INODE_CACHE(lqinode),
 					 chunk->qc_headerbh,
@@ -1172,11 +1177,8 @@ static struct ocfs2_quota_chunk *ocfs2_extend_local_quota_file(
 	lock_buffer(chunk->qc_headerbh);
 	le32_add_cpu(&dchunk->dqc_free, ol_quota_entries_per_block(sb));
 	unlock_buffer(chunk->qc_headerbh);
-	status = ocfs2_journal_dirty(handle, chunk->qc_headerbh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out_trans;
-	}
+	ocfs2_journal_dirty(handle, chunk->qc_headerbh);
+
 	/* Update file header */
 	oinfo->dqi_blocks++;
 	status = ocfs2_local_write_info(sb, type);
@@ -1209,7 +1211,7 @@ static void olq_alloc_dquot(struct buffer_head *bh, void *private)
 }
 
 /* Create dquot in the local file for given id */
-static int ocfs2_create_local_dquot(struct dquot *dquot)
+int ocfs2_create_local_dquot(struct dquot *dquot)
 {
 	struct super_block *sb = dquot->dq_sb;
 	int type = dquot->dq_type;
@@ -1218,17 +1220,27 @@ static int ocfs2_create_local_dquot(struct dquot *dquot)
 	struct ocfs2_dquot *od = OCFS2_DQUOT(dquot);
 	int offset;
 	int status;
+	u64 pcount;
 
+	down_write(&OCFS2_I(lqinode)->ip_alloc_sem);
 	chunk = ocfs2_find_free_entry(sb, type, &offset);
 	if (!chunk) {
 		chunk = ocfs2_extend_local_quota_file(sb, type, &offset);
-		if (IS_ERR(chunk))
-			return PTR_ERR(chunk);
+		if (IS_ERR(chunk)) {
+			status = PTR_ERR(chunk);
+			goto out;
+		}
 	} else if (IS_ERR(chunk)) {
-		return PTR_ERR(chunk);
+		status = PTR_ERR(chunk);
+		goto out;
 	}
 	od->dq_local_off = ol_dqblk_off(sb, chunk->qc_num, offset);
 	od->dq_chunk = chunk;
+	status = ocfs2_extent_map_get_blocks(lqinode,
+				     ol_dqblk_block(sb, chunk->qc_num, offset),
+				     &od->dq_local_phys_blk,
+				     &pcount,
+				     NULL);
 
 	/* Initialize dquot structure on disk */
 	status = ocfs2_local_write_dquot(dquot);
@@ -1245,39 +1257,15 @@ static int ocfs2_create_local_dquot(struct dquot *dquot)
 		goto out;
 	}
 out:
+	up_write(&OCFS2_I(lqinode)->ip_alloc_sem);
 	return status;
 }
 
-/* Create entry in local file for dquot, load data from the global file */
-static int ocfs2_local_read_dquot(struct dquot *dquot)
-{
-	int status;
-
-	mlog_entry("id=%u, type=%d\n", dquot->dq_id, dquot->dq_type);
-
-	status = ocfs2_global_read_dquot(dquot);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out_err;
-	}
-
-	/* Now create entry in the local quota file */
-	status = ocfs2_create_local_dquot(dquot);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out_err;
-	}
-	mlog_exit(0);
-	return 0;
-out_err:
-	mlog_exit(status);
-	return status;
-}
-
-/* Release dquot structure from local quota file. ocfs2_release_dquot() has
- * already started a transaction and obtained exclusive lock for global
- * quota file. */
-static int ocfs2_local_release_dquot(struct dquot *dquot)
+/*
+ * Release dquot structure from local quota file. ocfs2_release_dquot() has
+ * already started a transaction and written all changes to global quota file
+ */
+int ocfs2_local_release_dquot(handle_t *handle, struct dquot *dquot)
 {
 	int status;
 	int type = dquot->dq_type;
@@ -1285,15 +1273,6 @@ static int ocfs2_local_release_dquot(struct dquot *dquot)
 	struct super_block *sb = dquot->dq_sb;
 	struct ocfs2_local_disk_chunk *dchunk;
 	int offset;
-	handle_t *handle = journal_current_handle();
-
-	BUG_ON(!handle);
-	/* First write all local changes to global file */
-	status = ocfs2_global_release_dquot(dquot);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out;
-	}
 
 	status = ocfs2_journal_access_dq(handle,
 			INODE_CACHE(sb_dqopt(sb)->files[type]),
@@ -1311,12 +1290,8 @@ static int ocfs2_local_release_dquot(struct dquot *dquot)
 	ocfs2_clear_bit(offset, dchunk->dqc_bitmap);
 	le32_add_cpu(&dchunk->dqc_free, 1);
 	unlock_buffer(od->dq_chunk->qc_headerbh);
-	status = ocfs2_journal_dirty(handle, od->dq_chunk->qc_headerbh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out;
-	}
-	status = 0;
+	ocfs2_journal_dirty(handle, od->dq_chunk->qc_headerbh);
+
 out:
 	/* Clear the read bit so that next time someone uses this
 	 * dquot he reads fresh info from disk and allocates local
@@ -1330,9 +1305,6 @@ static const struct quota_format_ops ocfs2_format_ops = {
 	.read_file_info		= ocfs2_local_read_info,
 	.write_file_info	= ocfs2_global_write_info,
 	.free_file_info		= ocfs2_local_free_info,
-	.read_dqblk		= ocfs2_local_read_dquot,
-	.commit_dqblk		= ocfs2_local_write_dquot,
-	.release_dqblk		= ocfs2_local_release_dquot,
 };
 
 struct quota_format_type ocfs2_quota_format = {
diff --git a/fs/ocfs2/refcounttree.c b/fs/ocfs2/refcounttree.c
index 74db2be75dd6..3ac5aa733e9c 100644
--- a/fs/ocfs2/refcounttree.c
+++ b/fs/ocfs2/refcounttree.c
@@ -37,7 +37,6 @@
 
 #include <linux/bio.h>
 #include <linux/blkdev.h>
-#include <linux/gfp.h>
 #include <linux/slab.h>
 #include <linux/writeback.h>
 #include <linux/pagevec.h>
@@ -571,7 +570,7 @@ static int ocfs2_create_refcount_tree(struct inode *inode,
 	struct ocfs2_refcount_tree *new_tree = NULL, *tree = NULL;
 	u16 suballoc_bit_start;
 	u32 num_got;
-	u64 first_blkno;
+	u64 suballoc_loc, first_blkno;
 
 	BUG_ON(oi->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL);
 
@@ -597,7 +596,7 @@ static int ocfs2_create_refcount_tree(struct inode *inode,
 		goto out_commit;
 	}
 
-	ret = ocfs2_claim_metadata(osb, handle, meta_ac, 1,
+	ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc,
 				   &suballoc_bit_start, &num_got,
 				   &first_blkno);
 	if (ret) {
@@ -626,7 +625,8 @@ static int ocfs2_create_refcount_tree(struct inode *inode,
 	rb = (struct ocfs2_refcount_block *)new_bh->b_data;
 	memset(rb, 0, inode->i_sb->s_blocksize);
 	strcpy((void *)rb, OCFS2_REFCOUNT_BLOCK_SIGNATURE);
-	rb->rf_suballoc_slot = cpu_to_le16(osb->slot_num);
+	rb->rf_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot);
+	rb->rf_suballoc_loc = cpu_to_le64(suballoc_loc);
 	rb->rf_suballoc_bit = cpu_to_le16(suballoc_bit_start);
 	rb->rf_fs_generation = cpu_to_le32(osb->fs_generation);
 	rb->rf_blkno = cpu_to_le64(first_blkno);
@@ -791,7 +791,10 @@ int ocfs2_remove_refcount_tree(struct inode *inode, struct buffer_head *di_bh)
 	if (le32_to_cpu(rb->rf_count) == 1) {
 		blk = le64_to_cpu(rb->rf_blkno);
 		bit = le16_to_cpu(rb->rf_suballoc_bit);
-		bg_blkno = ocfs2_which_suballoc_group(blk, bit);
+		if (rb->rf_suballoc_loc)
+			bg_blkno = le64_to_cpu(rb->rf_suballoc_loc);
+		else
+			bg_blkno = ocfs2_which_suballoc_group(blk, bit);
 
 		alloc_inode = ocfs2_get_system_file_inode(osb,
 					EXTENT_ALLOC_SYSTEM_INODE,
@@ -1269,9 +1272,7 @@ static int ocfs2_change_refcount_rec(handle_t *handle,
 	} else if (merge)
 		ocfs2_refcount_rec_merge(rb, index);
 
-	ret = ocfs2_journal_dirty(handle, ref_leaf_bh);
-	if (ret)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, ref_leaf_bh);
 out:
 	return ret;
 }
@@ -1285,7 +1286,7 @@ static int ocfs2_expand_inline_ref_root(handle_t *handle,
 	int ret;
 	u16 suballoc_bit_start;
 	u32 num_got;
-	u64 blkno;
+	u64 suballoc_loc, blkno;
 	struct super_block *sb = ocfs2_metadata_cache_get_super(ci);
 	struct buffer_head *new_bh = NULL;
 	struct ocfs2_refcount_block *new_rb;
@@ -1299,7 +1300,7 @@ static int ocfs2_expand_inline_ref_root(handle_t *handle,
 		goto out;
 	}
 
-	ret = ocfs2_claim_metadata(OCFS2_SB(sb), handle, meta_ac, 1,
+	ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc,
 				   &suballoc_bit_start, &num_got,
 				   &blkno);
 	if (ret) {
@@ -1330,7 +1331,8 @@ static int ocfs2_expand_inline_ref_root(handle_t *handle,
 	memcpy(new_bh->b_data, ref_root_bh->b_data, sb->s_blocksize);
 
 	new_rb = (struct ocfs2_refcount_block *)new_bh->b_data;
-	new_rb->rf_suballoc_slot = cpu_to_le16(OCFS2_SB(sb)->slot_num);
+	new_rb->rf_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot);
+	new_rb->rf_suballoc_loc = cpu_to_le64(suballoc_loc);
 	new_rb->rf_suballoc_bit = cpu_to_le16(suballoc_bit_start);
 	new_rb->rf_blkno = cpu_to_le64(blkno);
 	new_rb->rf_cpos = cpu_to_le32(0);
@@ -1525,7 +1527,7 @@ static int ocfs2_new_leaf_refcount_block(handle_t *handle,
 	int ret;
 	u16 suballoc_bit_start;
 	u32 num_got, new_cpos;
-	u64 blkno;
+	u64 suballoc_loc, blkno;
 	struct super_block *sb = ocfs2_metadata_cache_get_super(ci);
 	struct ocfs2_refcount_block *root_rb =
 			(struct ocfs2_refcount_block *)ref_root_bh->b_data;
@@ -1549,7 +1551,7 @@ static int ocfs2_new_leaf_refcount_block(handle_t *handle,
 		goto out;
 	}
 
-	ret = ocfs2_claim_metadata(OCFS2_SB(sb), handle, meta_ac, 1,
+	ret = ocfs2_claim_metadata(handle, meta_ac, 1, &suballoc_loc,
 				   &suballoc_bit_start, &num_got,
 				   &blkno);
 	if (ret) {
@@ -1576,7 +1578,8 @@ static int ocfs2_new_leaf_refcount_block(handle_t *handle,
 	new_rb = (struct ocfs2_refcount_block *)new_bh->b_data;
 	memset(new_rb, 0, sb->s_blocksize);
 	strcpy((void *)new_rb, OCFS2_REFCOUNT_BLOCK_SIGNATURE);
-	new_rb->rf_suballoc_slot = cpu_to_le16(OCFS2_SB(sb)->slot_num);
+	new_rb->rf_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot);
+	new_rb->rf_suballoc_loc = cpu_to_le64(suballoc_loc);
 	new_rb->rf_suballoc_bit = cpu_to_le16(suballoc_bit_start);
 	new_rb->rf_fs_generation = cpu_to_le32(OCFS2_SB(sb)->fs_generation);
 	new_rb->rf_blkno = cpu_to_le64(blkno);
@@ -1695,7 +1698,7 @@ static int ocfs2_adjust_refcount_rec(handle_t *handle,
 	 * 2 more credits, one for the leaf refcount block, one for
 	 * the extent block contains the extent rec.
 	 */
-	ret = ocfs2_extend_trans(handle, handle->h_buffer_credits + 2);
+	ret = ocfs2_extend_trans(handle, 2);
 	if (ret < 0) {
 		mlog_errno(ret);
 		goto out;
@@ -1803,11 +1806,7 @@ static int ocfs2_insert_refcount_rec(handle_t *handle,
 	if (merge)
 		ocfs2_refcount_rec_merge(rb, index);
 
-	ret = ocfs2_journal_dirty(handle, ref_leaf_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out;
-	}
+	ocfs2_journal_dirty(handle, ref_leaf_bh);
 
 	if (index == 0) {
 		ret = ocfs2_adjust_refcount_rec(handle, ci,
@@ -1978,9 +1977,7 @@ static int ocfs2_split_refcount_rec(handle_t *handle,
 			ocfs2_refcount_rec_merge(rb, index);
 	}
 
-	ret = ocfs2_journal_dirty(handle, ref_leaf_bh);
-	if (ret)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, ref_leaf_bh);
 
 out:
 	brelse(new_bh);
@@ -2113,6 +2110,7 @@ static int ocfs2_remove_refcount_extent(handle_t *handle,
 	 */
 	ret = ocfs2_cache_block_dealloc(dealloc, EXTENT_ALLOC_SYSTEM_INODE,
 					le16_to_cpu(rb->rf_suballoc_slot),
+					le64_to_cpu(rb->rf_suballoc_loc),
 					le64_to_cpu(rb->rf_blkno),
 					le16_to_cpu(rb->rf_suballoc_bit));
 	if (ret) {
@@ -2517,20 +2515,19 @@ out:
  *
  * Normally the refcount blocks store these refcount should be
  * contiguous also, so that we can get the number easily.
- * As for meta_ac, we will at most add split 2 refcount record and
- * 2 more refcount block, so just check it in a rough way.
+ * We will at most add split 2 refcount records and 2 more
+ * refcount blocks, so just check it in a rough way.
  *
  * Caller must hold refcount tree lock.
  */
 int ocfs2_prepare_refcount_change_for_del(struct inode *inode,
-					  struct buffer_head *di_bh,
+					  u64 refcount_loc,
 					  u64 phys_blkno,
 					  u32 clusters,
 					  int *credits,
-					  struct ocfs2_alloc_context **meta_ac)
+					  int *ref_blocks)
 {
-	int ret, ref_blocks = 0;
-	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
+	int ret;
 	struct ocfs2_inode_info *oi = OCFS2_I(inode);
 	struct buffer_head *ref_root_bh = NULL;
 	struct ocfs2_refcount_tree *tree;
@@ -2547,14 +2544,13 @@ int ocfs2_prepare_refcount_change_for_del(struct inode *inode,
 	BUG_ON(!(oi->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL));
 
 	ret = ocfs2_get_refcount_tree(OCFS2_SB(inode->i_sb),
-				      le64_to_cpu(di->i_refcount_loc), &tree);
+				      refcount_loc, &tree);
 	if (ret) {
 		mlog_errno(ret);
 		goto out;
 	}
 
-	ret = ocfs2_read_refcount_block(&tree->rf_ci,
-					le64_to_cpu(di->i_refcount_loc),
+	ret = ocfs2_read_refcount_block(&tree->rf_ci, refcount_loc,
 					&ref_root_bh);
 	if (ret) {
 		mlog_errno(ret);
@@ -2565,21 +2561,14 @@ int ocfs2_prepare_refcount_change_for_del(struct inode *inode,
 					       &tree->rf_ci,
 					       ref_root_bh,
 					       start_cpos, clusters,
-					       &ref_blocks, credits);
+					       ref_blocks, credits);
 	if (ret) {
 		mlog_errno(ret);
 		goto out;
 	}
 
-	mlog(0, "reserve new metadata %d, credits = %d\n",
-	     ref_blocks, *credits);
-
-	if (ref_blocks) {
-		ret = ocfs2_reserve_new_metadata_blocks(OCFS2_SB(inode->i_sb),
-							ref_blocks, meta_ac);
-		if (ret)
-			mlog_errno(ret);
-	}
+	mlog(0, "reserve new metadata %d blocks, credits = %d\n",
+	     *ref_blocks, *credits);
 
 out:
 	brelse(ref_root_bh);
@@ -2942,10 +2931,16 @@ static int ocfs2_duplicate_clusters_by_page(handle_t *handle,
 
 	offset = ((loff_t)cpos) << OCFS2_SB(sb)->s_clustersize_bits;
 	end = offset + (new_len << OCFS2_SB(sb)->s_clustersize_bits);
+	/*
+	 * We only duplicate pages until we reach the page contains i_size - 1.
+	 * So trim 'end' to i_size.
+	 */
+	if (end > i_size_read(context->inode))
+		end = i_size_read(context->inode);
 
 	while (offset < end) {
 		page_index = offset >> PAGE_CACHE_SHIFT;
-		map_end = (page_index + 1) << PAGE_CACHE_SHIFT;
+		map_end = ((loff_t)page_index + 1) << PAGE_CACHE_SHIFT;
 		if (map_end > end)
 			map_end = end;
 
@@ -2957,8 +2952,12 @@ static int ocfs2_duplicate_clusters_by_page(handle_t *handle,
 
 		page = grab_cache_page(mapping, page_index);
 
-		/* This page can't be dirtied before we CoW it out. */
-		BUG_ON(PageDirty(page));
+		/*
+		 * In case PAGE_CACHE_SIZE <= CLUSTER_SIZE, This page
+		 * can't be dirtied before we CoW it out.
+		 */
+		if (PAGE_CACHE_SIZE <= OCFS2_SB(sb)->s_clustersize)
+			BUG_ON(PageDirty(page));
 
 		if (!PageUptodate(page)) {
 			ret = block_read_full_page(page, ocfs2_get_block);
@@ -3037,11 +3036,7 @@ static int ocfs2_duplicate_clusters_by_jbd(handle_t *handle,
 		}
 
 		memcpy(new_bh->b_data, old_bh->b_data, sb->s_blocksize);
-		ret = ocfs2_journal_dirty(handle, new_bh);
-		if (ret) {
-			mlog_errno(ret);
-			break;
-		}
+		ocfs2_journal_dirty(handle, new_bh);
 
 		brelse(new_bh);
 		brelse(old_bh);
@@ -3170,7 +3165,7 @@ static int ocfs2_cow_sync_writeback(struct super_block *sb,
 
 	while (offset < end) {
 		page_index = offset >> PAGE_CACHE_SHIFT;
-		map_end = (page_index + 1) << PAGE_CACHE_SHIFT;
+		map_end = ((loff_t)page_index + 1) << PAGE_CACHE_SHIFT;
 		if (map_end > end)
 			map_end = end;
 
@@ -3279,7 +3274,7 @@ static int ocfs2_make_clusters_writable(struct super_block *sb,
 		} else {
 			delete = 1;
 
-			ret = __ocfs2_claim_clusters(osb, handle,
+			ret = __ocfs2_claim_clusters(handle,
 						     context->data_ac,
 						     1, set_len,
 						     &new_bit, &new_len);
@@ -4071,6 +4066,7 @@ static int ocfs2_complete_reflink(struct inode *s_inode,
 	OCFS2_I(t_inode)->ip_dyn_features = OCFS2_I(s_inode)->ip_dyn_features;
 	spin_unlock(&OCFS2_I(t_inode)->ip_lock);
 	i_size_write(t_inode, size);
+	t_inode->i_blocks = s_inode->i_blocks;
 
 	di->i_xattr_inline_size = s_di->i_xattr_inline_size;
 	di->i_clusters = s_di->i_clusters;
@@ -4079,6 +4075,9 @@ static int ocfs2_complete_reflink(struct inode *s_inode,
 	di->i_attr = s_di->i_attr;
 
 	if (preserve) {
+		t_inode->i_uid = s_inode->i_uid;
+		t_inode->i_gid = s_inode->i_gid;
+		t_inode->i_mode = s_inode->i_mode;
 		di->i_uid = s_di->i_uid;
 		di->i_gid = s_di->i_gid;
 		di->i_mode = s_di->i_mode;
@@ -4173,6 +4172,12 @@ static int __ocfs2_reflink(struct dentry *old_dentry,
 	struct inode *inode = old_dentry->d_inode;
 	struct buffer_head *new_bh = NULL;
 
+	if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SYSTEM_FILE) {
+		ret = -EINVAL;
+		mlog_errno(ret);
+		goto out;
+	}
+
 	ret = filemap_fdatawrite(inode->i_mapping);
 	if (ret) {
 		mlog_errno(ret);
@@ -4386,7 +4391,7 @@ static int ocfs2_vfs_reflink(struct dentry *old_dentry, struct inode *dir,
 	}
 
 	mutex_lock(&inode->i_mutex);
-	vfs_dq_init(dir);
+	dquot_initialize(dir);
 	error = ocfs2_reflink(old_dentry, dir, new_dentry, preserve);
 	mutex_unlock(&inode->i_mutex);
 	if (!error)
diff --git a/fs/ocfs2/refcounttree.h b/fs/ocfs2/refcounttree.h
index c1d19b1d3ecc..9983ba1570e2 100644
--- a/fs/ocfs2/refcounttree.h
+++ b/fs/ocfs2/refcounttree.h
@@ -47,11 +47,11 @@ int ocfs2_decrease_refcount(struct inode *inode,
 			    struct ocfs2_cached_dealloc_ctxt *dealloc,
 			    int delete);
 int ocfs2_prepare_refcount_change_for_del(struct inode *inode,
-					  struct buffer_head *di_bh,
+					  u64 refcount_loc,
 					  u64 phys_blkno,
 					  u32 clusters,
 					  int *credits,
-					  struct ocfs2_alloc_context **meta_ac);
+					  int *ref_blocks);
 int ocfs2_refcount_cow(struct inode *inode, struct buffer_head *di_bh,
 		       u32 cpos, u32 write_len, u32 max_cpos);
 
diff --git a/fs/ocfs2/reservations.c b/fs/ocfs2/reservations.c
new file mode 100644
index 000000000000..d8b6e4259b80
--- /dev/null
+++ b/fs/ocfs2/reservations.c
@@ -0,0 +1,846 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * reservations.c
+ *
+ * Allocation reservations implementation
+ *
+ * Some code borrowed from fs/ext3/balloc.c and is:
+ *
+ * Copyright (C) 1992, 1993, 1994, 1995
+ * Remy Card (card@masi.ibp.fr)
+ * Laboratoire MASI - Institut Blaise Pascal
+ * Universite Pierre et Marie Curie (Paris VI)
+ *
+ * The rest is copyright (C) 2010 Novell.  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 version 2 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.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/highmem.h>
+#include <linux/bitops.h>
+#include <linux/list.h>
+
+#define MLOG_MASK_PREFIX ML_RESERVATIONS
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#ifdef CONFIG_OCFS2_DEBUG_FS
+#define OCFS2_CHECK_RESERVATIONS
+#endif
+
+DEFINE_SPINLOCK(resv_lock);
+
+#define	OCFS2_MIN_RESV_WINDOW_BITS	8
+#define	OCFS2_MAX_RESV_WINDOW_BITS	1024
+
+int ocfs2_dir_resv_allowed(struct ocfs2_super *osb)
+{
+	return (osb->osb_resv_level && osb->osb_dir_resv_level);
+}
+
+static unsigned int ocfs2_resv_window_bits(struct ocfs2_reservation_map *resmap,
+					   struct ocfs2_alloc_reservation *resv)
+{
+	struct ocfs2_super *osb = resmap->m_osb;
+	unsigned int bits;
+
+	if (!(resv->r_flags & OCFS2_RESV_FLAG_DIR)) {
+		/* 8, 16, 32, 64, 128, 256, 512, 1024 */
+		bits = 4 << osb->osb_resv_level;
+	} else {
+		bits = 4 << osb->osb_dir_resv_level;
+	}
+	return bits;
+}
+
+static inline unsigned int ocfs2_resv_end(struct ocfs2_alloc_reservation *resv)
+{
+	if (resv->r_len)
+		return resv->r_start + resv->r_len - 1;
+	return resv->r_start;
+}
+
+static inline int ocfs2_resv_empty(struct ocfs2_alloc_reservation *resv)
+{
+	return !!(resv->r_len == 0);
+}
+
+static inline int ocfs2_resmap_disabled(struct ocfs2_reservation_map *resmap)
+{
+	if (resmap->m_osb->osb_resv_level == 0)
+		return 1;
+	return 0;
+}
+
+static void ocfs2_dump_resv(struct ocfs2_reservation_map *resmap)
+{
+	struct ocfs2_super *osb = resmap->m_osb;
+	struct rb_node *node;
+	struct ocfs2_alloc_reservation *resv;
+	int i = 0;
+
+	mlog(ML_NOTICE, "Dumping resmap for device %s. Bitmap length: %u\n",
+	     osb->dev_str, resmap->m_bitmap_len);
+
+	node = rb_first(&resmap->m_reservations);
+	while (node) {
+		resv = rb_entry(node, struct ocfs2_alloc_reservation, r_node);
+
+		mlog(ML_NOTICE, "start: %u\tend: %u\tlen: %u\tlast_start: %u"
+		     "\tlast_len: %u\n", resv->r_start,
+		     ocfs2_resv_end(resv), resv->r_len, resv->r_last_start,
+		     resv->r_last_len);
+
+		node = rb_next(node);
+		i++;
+	}
+
+	mlog(ML_NOTICE, "%d reservations found. LRU follows\n", i);
+
+	i = 0;
+	list_for_each_entry(resv, &resmap->m_lru, r_lru) {
+		mlog(ML_NOTICE, "LRU(%d) start: %u\tend: %u\tlen: %u\t"
+		     "last_start: %u\tlast_len: %u\n", i, resv->r_start,
+		     ocfs2_resv_end(resv), resv->r_len, resv->r_last_start,
+		     resv->r_last_len);
+
+		i++;
+	}
+}
+
+#ifdef OCFS2_CHECK_RESERVATIONS
+static int ocfs2_validate_resmap_bits(struct ocfs2_reservation_map *resmap,
+				      int i,
+				      struct ocfs2_alloc_reservation *resv)
+{
+	char *disk_bitmap = resmap->m_disk_bitmap;
+	unsigned int start = resv->r_start;
+	unsigned int end = ocfs2_resv_end(resv);
+
+	while (start <= end) {
+		if (ocfs2_test_bit(start, disk_bitmap)) {
+			mlog(ML_ERROR,
+			     "reservation %d covers an allocated area "
+			     "starting at bit %u!\n", i, start);
+			return 1;
+		}
+
+		start++;
+	}
+	return 0;
+}
+
+static void ocfs2_check_resmap(struct ocfs2_reservation_map *resmap)
+{
+	unsigned int off = 0;
+	int i = 0;
+	struct rb_node *node;
+	struct ocfs2_alloc_reservation *resv;
+
+	node = rb_first(&resmap->m_reservations);
+	while (node) {
+		resv = rb_entry(node, struct ocfs2_alloc_reservation, r_node);
+
+		if (i > 0 && resv->r_start <= off) {
+			mlog(ML_ERROR, "reservation %d has bad start off!\n",
+			     i);
+			goto bad;
+		}
+
+		if (resv->r_len == 0) {
+			mlog(ML_ERROR, "reservation %d has no length!\n",
+			     i);
+			goto bad;
+		}
+
+		if (resv->r_start > ocfs2_resv_end(resv)) {
+			mlog(ML_ERROR, "reservation %d has invalid range!\n",
+			     i);
+			goto bad;
+		}
+
+		if (ocfs2_resv_end(resv) >= resmap->m_bitmap_len) {
+			mlog(ML_ERROR, "reservation %d extends past bitmap!\n",
+			     i);
+			goto bad;
+		}
+
+		if (ocfs2_validate_resmap_bits(resmap, i, resv))
+			goto bad;
+
+		off = ocfs2_resv_end(resv);
+		node = rb_next(node);
+
+		i++;
+	}
+	return;
+
+bad:
+	ocfs2_dump_resv(resmap);
+	BUG();
+}
+#else
+static inline void ocfs2_check_resmap(struct ocfs2_reservation_map *resmap)
+{
+
+}
+#endif
+
+void ocfs2_resv_init_once(struct ocfs2_alloc_reservation *resv)
+{
+	memset(resv, 0, sizeof(*resv));
+	INIT_LIST_HEAD(&resv->r_lru);
+}
+
+void ocfs2_resv_set_type(struct ocfs2_alloc_reservation *resv,
+			 unsigned int flags)
+{
+	BUG_ON(flags & ~OCFS2_RESV_TYPES);
+
+	resv->r_flags |= flags;
+}
+
+int ocfs2_resmap_init(struct ocfs2_super *osb,
+		      struct ocfs2_reservation_map *resmap)
+{
+	memset(resmap, 0, sizeof(*resmap));
+
+	resmap->m_osb = osb;
+	resmap->m_reservations = RB_ROOT;
+	/* m_bitmap_len is initialized to zero by the above memset. */
+	INIT_LIST_HEAD(&resmap->m_lru);
+
+	return 0;
+}
+
+static void ocfs2_resv_mark_lru(struct ocfs2_reservation_map *resmap,
+				struct ocfs2_alloc_reservation *resv)
+{
+	assert_spin_locked(&resv_lock);
+
+	if (!list_empty(&resv->r_lru))
+		list_del_init(&resv->r_lru);
+
+	list_add_tail(&resv->r_lru, &resmap->m_lru);
+}
+
+static void __ocfs2_resv_trunc(struct ocfs2_alloc_reservation *resv)
+{
+	resv->r_len = 0;
+	resv->r_start = 0;
+}
+
+static void ocfs2_resv_remove(struct ocfs2_reservation_map *resmap,
+			      struct ocfs2_alloc_reservation *resv)
+{
+	if (resv->r_flags & OCFS2_RESV_FLAG_INUSE) {
+		list_del_init(&resv->r_lru);
+		rb_erase(&resv->r_node, &resmap->m_reservations);
+		resv->r_flags &= ~OCFS2_RESV_FLAG_INUSE;
+	}
+}
+
+static void __ocfs2_resv_discard(struct ocfs2_reservation_map *resmap,
+				 struct ocfs2_alloc_reservation *resv)
+{
+	assert_spin_locked(&resv_lock);
+
+	__ocfs2_resv_trunc(resv);
+	/*
+	 * last_len and last_start no longer make sense if
+	 * we're changing the range of our allocations.
+	 */
+	resv->r_last_len = resv->r_last_start = 0;
+
+	ocfs2_resv_remove(resmap, resv);
+}
+
+/* does nothing if 'resv' is null */
+void ocfs2_resv_discard(struct ocfs2_reservation_map *resmap,
+			struct ocfs2_alloc_reservation *resv)
+{
+	if (resv) {
+		spin_lock(&resv_lock);
+		__ocfs2_resv_discard(resmap, resv);
+		spin_unlock(&resv_lock);
+	}
+}
+
+static void ocfs2_resmap_clear_all_resv(struct ocfs2_reservation_map *resmap)
+{
+	struct rb_node *node;
+	struct ocfs2_alloc_reservation *resv;
+
+	assert_spin_locked(&resv_lock);
+
+	while ((node = rb_last(&resmap->m_reservations)) != NULL) {
+		resv = rb_entry(node, struct ocfs2_alloc_reservation, r_node);
+
+		__ocfs2_resv_discard(resmap, resv);
+	}
+}
+
+void ocfs2_resmap_restart(struct ocfs2_reservation_map *resmap,
+			  unsigned int clen, char *disk_bitmap)
+{
+	if (ocfs2_resmap_disabled(resmap))
+		return;
+
+	spin_lock(&resv_lock);
+
+	ocfs2_resmap_clear_all_resv(resmap);
+	resmap->m_bitmap_len = clen;
+	resmap->m_disk_bitmap = disk_bitmap;
+
+	spin_unlock(&resv_lock);
+}
+
+void ocfs2_resmap_uninit(struct ocfs2_reservation_map *resmap)
+{
+	/* Does nothing for now. Keep this around for API symmetry */
+}
+
+static void ocfs2_resv_insert(struct ocfs2_reservation_map *resmap,
+			      struct ocfs2_alloc_reservation *new)
+{
+	struct rb_root *root = &resmap->m_reservations;
+	struct rb_node *parent = NULL;
+	struct rb_node **p = &root->rb_node;
+	struct ocfs2_alloc_reservation *tmp;
+
+	assert_spin_locked(&resv_lock);
+
+	mlog(0, "Insert reservation start: %u len: %u\n", new->r_start,
+	     new->r_len);
+
+	while (*p) {
+		parent = *p;
+
+		tmp = rb_entry(parent, struct ocfs2_alloc_reservation, r_node);
+
+		if (new->r_start < tmp->r_start) {
+			p = &(*p)->rb_left;
+
+			/*
+			 * This is a good place to check for
+			 * overlapping reservations.
+			 */
+			BUG_ON(ocfs2_resv_end(new) >= tmp->r_start);
+		} else if (new->r_start > ocfs2_resv_end(tmp)) {
+			p = &(*p)->rb_right;
+		} else {
+			/* This should never happen! */
+			mlog(ML_ERROR, "Duplicate reservation window!\n");
+			BUG();
+		}
+	}
+
+	rb_link_node(&new->r_node, parent, p);
+	rb_insert_color(&new->r_node, root);
+	new->r_flags |= OCFS2_RESV_FLAG_INUSE;
+
+	ocfs2_resv_mark_lru(resmap, new);
+
+	ocfs2_check_resmap(resmap);
+}
+
+/**
+ * ocfs2_find_resv_lhs() - find the window which contains goal
+ * @resmap: reservation map to search
+ * @goal: which bit to search for
+ *
+ * If a window containing that goal is not found, we return the window
+ * which comes before goal. Returns NULL on empty rbtree or no window
+ * before goal.
+ */
+static struct ocfs2_alloc_reservation *
+ocfs2_find_resv_lhs(struct ocfs2_reservation_map *resmap, unsigned int goal)
+{
+	struct ocfs2_alloc_reservation *resv = NULL;
+	struct ocfs2_alloc_reservation *prev_resv = NULL;
+	struct rb_node *node = resmap->m_reservations.rb_node;
+
+	assert_spin_locked(&resv_lock);
+
+	if (!node)
+		return NULL;
+
+	node = rb_first(&resmap->m_reservations);
+	while (node) {
+		resv = rb_entry(node, struct ocfs2_alloc_reservation, r_node);
+
+		if (resv->r_start <= goal && ocfs2_resv_end(resv) >= goal)
+			break;
+
+		/* Check if we overshot the reservation just before goal? */
+		if (resv->r_start > goal) {
+			resv = prev_resv;
+			break;
+		}
+
+		prev_resv = resv;
+		node = rb_next(node);
+	}
+
+	return resv;
+}
+
+/*
+ * We are given a range within the bitmap, which corresponds to a gap
+ * inside the reservations tree (search_start, search_len). The range
+ * can be anything from the whole bitmap, to a gap between
+ * reservations.
+ *
+ * The start value of *rstart is insignificant.
+ *
+ * This function searches the bitmap range starting at search_start
+ * with length search_len for a set of contiguous free bits. We try
+ * to find up to 'wanted' bits, but can sometimes return less.
+ *
+ * Returns the length of allocation, 0 if no free bits are found.
+ *
+ * *cstart and *clen will also be populated with the result.
+ */
+static int ocfs2_resmap_find_free_bits(struct ocfs2_reservation_map *resmap,
+				       unsigned int wanted,
+				       unsigned int search_start,
+				       unsigned int search_len,
+				       unsigned int *rstart,
+				       unsigned int *rlen)
+{
+	void *bitmap = resmap->m_disk_bitmap;
+	unsigned int best_start, best_len = 0;
+	int offset, start, found;
+
+	mlog(0, "Find %u bits within range (%u, len %u) resmap len: %u\n",
+	     wanted, search_start, search_len, resmap->m_bitmap_len);
+
+	found = best_start = best_len = 0;
+
+	start = search_start;
+	while ((offset = ocfs2_find_next_zero_bit(bitmap, resmap->m_bitmap_len,
+						 start)) != -1) {
+		/* Search reached end of the region */
+		if (offset >= (search_start + search_len))
+			break;
+
+		if (offset == start) {
+			/* we found a zero */
+			found++;
+			/* move start to the next bit to test */
+			start++;
+		} else {
+			/* got a zero after some ones */
+			found = 1;
+			start = offset + 1;
+		}
+		if (found > best_len) {
+			best_len = found;
+			best_start = start - found;
+		}
+
+		if (found >= wanted)
+			break;
+	}
+
+	if (best_len == 0)
+		return 0;
+
+	if (best_len >= wanted)
+		best_len = wanted;
+
+	*rlen = best_len;
+	*rstart = best_start;
+
+	mlog(0, "Found start: %u len: %u\n", best_start, best_len);
+
+	return *rlen;
+}
+
+static void __ocfs2_resv_find_window(struct ocfs2_reservation_map *resmap,
+				     struct ocfs2_alloc_reservation *resv,
+				     unsigned int goal, unsigned int wanted)
+{
+	struct rb_root *root = &resmap->m_reservations;
+	unsigned int gap_start, gap_end, gap_len;
+	struct ocfs2_alloc_reservation *prev_resv, *next_resv;
+	struct rb_node *prev, *next;
+	unsigned int cstart, clen;
+	unsigned int best_start = 0, best_len = 0;
+
+	/*
+	 * Nasty cases to consider:
+	 *
+	 * - rbtree is empty
+	 * - our window should be first in all reservations
+	 * - our window should be last in all reservations
+	 * - need to make sure we don't go past end of bitmap
+	 */
+
+	mlog(0, "resv start: %u resv end: %u goal: %u wanted: %u\n",
+	     resv->r_start, ocfs2_resv_end(resv), goal, wanted);
+
+	assert_spin_locked(&resv_lock);
+
+	if (RB_EMPTY_ROOT(root)) {
+		/*
+		 * Easiest case - empty tree. We can just take
+		 * whatever window of free bits we want.
+		 */
+
+		mlog(0, "Empty root\n");
+
+		clen = ocfs2_resmap_find_free_bits(resmap, wanted, goal,
+						   resmap->m_bitmap_len - goal,
+						   &cstart, &clen);
+
+		/*
+		 * This should never happen - the local alloc window
+		 * will always have free bits when we're called.
+		 */
+		BUG_ON(goal == 0 && clen == 0);
+
+		if (clen == 0)
+			return;
+
+		resv->r_start = cstart;
+		resv->r_len = clen;
+
+		ocfs2_resv_insert(resmap, resv);
+		return;
+	}
+
+	prev_resv = ocfs2_find_resv_lhs(resmap, goal);
+
+	if (prev_resv == NULL) {
+		mlog(0, "Goal on LHS of leftmost window\n");
+
+		/*
+		 * A NULL here means that the search code couldn't
+		 * find a window that starts before goal.
+		 *
+		 * However, we can take the first window after goal,
+		 * which is also by definition, the leftmost window in
+		 * the entire tree. If we can find free bits in the
+		 * gap between goal and the LHS window, then the
+		 * reservation can safely be placed there.
+		 *
+		 * Otherwise we fall back to a linear search, checking
+		 * the gaps in between windows for a place to
+		 * allocate.
+		 */
+
+		next = rb_first(root);
+		next_resv = rb_entry(next, struct ocfs2_alloc_reservation,
+				     r_node);
+
+		/*
+		 * The search should never return such a window. (see
+		 * comment above
+		 */
+		if (next_resv->r_start <= goal) {
+			mlog(ML_ERROR, "goal: %u next_resv: start %u len %u\n",
+			     goal, next_resv->r_start, next_resv->r_len);
+			ocfs2_dump_resv(resmap);
+			BUG();
+		}
+
+		clen = ocfs2_resmap_find_free_bits(resmap, wanted, goal,
+						   next_resv->r_start - goal,
+						   &cstart, &clen);
+		if (clen) {
+			best_len = clen;
+			best_start = cstart;
+			if (best_len == wanted)
+				goto out_insert;
+		}
+
+		prev_resv = next_resv;
+		next_resv = NULL;
+	}
+
+	prev = &prev_resv->r_node;
+
+	/* Now we do a linear search for a window, starting at 'prev_rsv' */
+	while (1) {
+		next = rb_next(prev);
+		if (next) {
+			mlog(0, "One more resv found in linear search\n");
+			next_resv = rb_entry(next,
+					     struct ocfs2_alloc_reservation,
+					     r_node);
+
+			gap_start = ocfs2_resv_end(prev_resv) + 1;
+			gap_end = next_resv->r_start - 1;
+			gap_len = gap_end - gap_start + 1;
+		} else {
+			mlog(0, "No next node\n");
+			/*
+			 * We're at the rightmost edge of the
+			 * tree. See if a reservation between this
+			 * window and the end of the bitmap will work.
+			 */
+			gap_start = ocfs2_resv_end(prev_resv) + 1;
+			gap_len = resmap->m_bitmap_len - gap_start;
+			gap_end = resmap->m_bitmap_len - 1;
+		}
+
+		/*
+		 * No need to check this gap if we have already found
+		 * a larger region of free bits.
+		 */
+		if (gap_len <= best_len)
+			goto next_resv;
+
+		clen = ocfs2_resmap_find_free_bits(resmap, wanted, gap_start,
+						   gap_len, &cstart, &clen);
+		if (clen == wanted) {
+			best_len = clen;
+			best_start = cstart;
+			goto out_insert;
+		} else if (clen > best_len) {
+			best_len = clen;
+			best_start = cstart;
+		}
+
+next_resv:
+		if (!next)
+			break;
+
+		prev = next;
+		prev_resv = rb_entry(prev, struct ocfs2_alloc_reservation,
+				     r_node);
+	}
+
+out_insert:
+	if (best_len) {
+		resv->r_start = best_start;
+		resv->r_len = best_len;
+		ocfs2_resv_insert(resmap, resv);
+	}
+}
+
+static void ocfs2_cannibalize_resv(struct ocfs2_reservation_map *resmap,
+				   struct ocfs2_alloc_reservation *resv,
+				   unsigned int wanted)
+{
+	struct ocfs2_alloc_reservation *lru_resv;
+	int tmpwindow = !!(resv->r_flags & OCFS2_RESV_FLAG_TMP);
+	unsigned int min_bits;
+
+	if (!tmpwindow)
+		min_bits = ocfs2_resv_window_bits(resmap, resv) >> 1;
+	else
+		min_bits = wanted; /* We at know the temp window will use all
+				    * of these bits */
+
+	/*
+	 * Take the first reservation off the LRU as our 'target'. We
+	 * don't try to be smart about it. There might be a case for
+	 * searching based on size but I don't have enough data to be
+	 * sure. --Mark (3/16/2010)
+	 */
+	lru_resv = list_first_entry(&resmap->m_lru,
+				    struct ocfs2_alloc_reservation, r_lru);
+
+	mlog(0, "lru resv: start: %u len: %u end: %u\n", lru_resv->r_start,
+	     lru_resv->r_len, ocfs2_resv_end(lru_resv));
+
+	/*
+	 * Cannibalize (some or all) of the target reservation and
+	 * feed it to the current window.
+	 */
+	if (lru_resv->r_len <= min_bits) {
+		/*
+		 * Discard completely if size is less than or equal to a
+		 * reasonable threshold - 50% of window bits for non temporary
+		 * windows.
+		 */
+		resv->r_start = lru_resv->r_start;
+		resv->r_len = lru_resv->r_len;
+
+		__ocfs2_resv_discard(resmap, lru_resv);
+	} else {
+		unsigned int shrink;
+		if (tmpwindow)
+			shrink = min_bits;
+		else
+			shrink = lru_resv->r_len / 2;
+
+		lru_resv->r_len -= shrink;
+
+		resv->r_start = ocfs2_resv_end(lru_resv) + 1;
+		resv->r_len = shrink;
+	}
+
+	mlog(0, "Reservation now looks like: r_start: %u r_end: %u "
+	     "r_len: %u r_last_start: %u r_last_len: %u\n",
+	     resv->r_start, ocfs2_resv_end(resv), resv->r_len,
+	     resv->r_last_start, resv->r_last_len);
+
+	ocfs2_resv_insert(resmap, resv);
+}
+
+static void ocfs2_resv_find_window(struct ocfs2_reservation_map *resmap,
+				   struct ocfs2_alloc_reservation *resv,
+				   unsigned int wanted)
+{
+	unsigned int goal = 0;
+
+	BUG_ON(!ocfs2_resv_empty(resv));
+
+	/*
+	 * Begin by trying to get a window as close to the previous
+	 * one as possible. Using the most recent allocation as a
+	 * start goal makes sense.
+	 */
+	if (resv->r_last_len) {
+		goal = resv->r_last_start + resv->r_last_len;
+		if (goal >= resmap->m_bitmap_len)
+			goal = 0;
+	}
+
+	__ocfs2_resv_find_window(resmap, resv, goal, wanted);
+
+	/* Search from last alloc didn't work, try once more from beginning. */
+	if (ocfs2_resv_empty(resv) && goal != 0)
+		__ocfs2_resv_find_window(resmap, resv, 0, wanted);
+
+	if (ocfs2_resv_empty(resv)) {
+		/*
+		 * Still empty? Pull oldest one off the LRU, remove it from
+		 * tree, put this one in it's place.
+		 */
+		ocfs2_cannibalize_resv(resmap, resv, wanted);
+	}
+
+	BUG_ON(ocfs2_resv_empty(resv));
+}
+
+int ocfs2_resmap_resv_bits(struct ocfs2_reservation_map *resmap,
+			   struct ocfs2_alloc_reservation *resv,
+			   int *cstart, int *clen)
+{
+	unsigned int wanted = *clen;
+
+	if (resv == NULL || ocfs2_resmap_disabled(resmap))
+		return -ENOSPC;
+
+	spin_lock(&resv_lock);
+
+	/*
+	 * We don't want to over-allocate for temporary
+	 * windows. Otherwise, we run the risk of fragmenting the
+	 * allocation space.
+	 */
+	wanted = ocfs2_resv_window_bits(resmap, resv);
+	if ((resv->r_flags & OCFS2_RESV_FLAG_TMP) || wanted < *clen)
+		wanted = *clen;
+
+	if (ocfs2_resv_empty(resv)) {
+		mlog(0, "empty reservation, find new window\n");
+
+		/*
+		 * Try to get a window here. If it works, we must fall
+		 * through and test the bitmap . This avoids some
+		 * ping-ponging of windows due to non-reserved space
+		 * being allocation before we initialize a window for
+		 * that inode.
+		 */
+		ocfs2_resv_find_window(resmap, resv, wanted);
+	}
+
+	BUG_ON(ocfs2_resv_empty(resv));
+
+	*cstart = resv->r_start;
+	*clen = resv->r_len;
+
+	spin_unlock(&resv_lock);
+	return 0;
+}
+
+static void
+	ocfs2_adjust_resv_from_alloc(struct ocfs2_reservation_map *resmap,
+				     struct ocfs2_alloc_reservation *resv,
+				     unsigned int start, unsigned int end)
+{
+	unsigned int rhs = 0;
+	unsigned int old_end = ocfs2_resv_end(resv);
+
+	BUG_ON(start != resv->r_start || old_end < end);
+
+	/*
+	 * Completely used? We can remove it then.
+	 */
+	if (old_end == end) {
+		__ocfs2_resv_discard(resmap, resv);
+		return;
+	}
+
+	rhs = old_end - end;
+
+	/*
+	 * This should have been trapped above.
+	 */
+	BUG_ON(rhs == 0);
+
+	resv->r_start = end + 1;
+	resv->r_len = old_end - resv->r_start + 1;
+}
+
+void ocfs2_resmap_claimed_bits(struct ocfs2_reservation_map *resmap,
+			       struct ocfs2_alloc_reservation *resv,
+			       u32 cstart, u32 clen)
+{
+	unsigned int cend = cstart + clen - 1;
+
+	if (resmap == NULL || ocfs2_resmap_disabled(resmap))
+		return;
+
+	if (resv == NULL)
+		return;
+
+	BUG_ON(cstart != resv->r_start);
+
+	spin_lock(&resv_lock);
+
+	mlog(0, "claim bits: cstart: %u cend: %u clen: %u r_start: %u "
+	     "r_end: %u r_len: %u, r_last_start: %u r_last_len: %u\n",
+	     cstart, cend, clen, resv->r_start, ocfs2_resv_end(resv),
+	     resv->r_len, resv->r_last_start, resv->r_last_len);
+
+	BUG_ON(cstart < resv->r_start);
+	BUG_ON(cstart > ocfs2_resv_end(resv));
+	BUG_ON(cend > ocfs2_resv_end(resv));
+
+	ocfs2_adjust_resv_from_alloc(resmap, resv, cstart, cend);
+	resv->r_last_start = cstart;
+	resv->r_last_len = clen;
+
+	/*
+	 * May have been discarded above from
+	 * ocfs2_adjust_resv_from_alloc().
+	 */
+	if (!ocfs2_resv_empty(resv))
+		ocfs2_resv_mark_lru(resmap, resv);
+
+	mlog(0, "Reservation now looks like: r_start: %u r_end: %u "
+	     "r_len: %u r_last_start: %u r_last_len: %u\n",
+	     resv->r_start, ocfs2_resv_end(resv), resv->r_len,
+	     resv->r_last_start, resv->r_last_len);
+
+	ocfs2_check_resmap(resmap);
+
+	spin_unlock(&resv_lock);
+}
diff --git a/fs/ocfs2/reservations.h b/fs/ocfs2/reservations.h
new file mode 100644
index 000000000000..1e49cc29d06c
--- /dev/null
+++ b/fs/ocfs2/reservations.h
@@ -0,0 +1,159 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * reservations.h
+ *
+ * Allocation reservations function prototypes and structures.
+ *
+ * Copyright (C) 2010 Novell.  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 version 2 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.
+ */
+
+#ifndef	OCFS2_RESERVATIONS_H
+#define	OCFS2_RESERVATIONS_H
+
+#include <linux/rbtree.h>
+
+#define OCFS2_DEFAULT_RESV_LEVEL	2
+#define OCFS2_MAX_RESV_LEVEL	9
+#define OCFS2_MIN_RESV_LEVEL	0
+
+struct ocfs2_alloc_reservation {
+	struct rb_node	r_node;
+
+	unsigned int	r_start;	/* Begining of current window */
+	unsigned int	r_len;		/* Length of the window */
+
+	unsigned int	r_last_len;	/* Length of most recent alloc */
+	unsigned int	r_last_start;	/* Start of most recent alloc */
+	struct list_head	r_lru;	/* LRU list head */
+
+	unsigned int	r_flags;
+};
+
+#define	OCFS2_RESV_FLAG_INUSE	0x01	/* Set when r_node is part of a btree */
+#define	OCFS2_RESV_FLAG_TMP	0x02	/* Temporary reservation, will be
+					 * destroyed immedately after use */
+#define	OCFS2_RESV_FLAG_DIR	0x04	/* Reservation is for an unindexed
+					 * directory btree */
+
+struct ocfs2_reservation_map {
+	struct rb_root		m_reservations;
+	char			*m_disk_bitmap;
+
+	struct ocfs2_super	*m_osb;
+
+	/* The following are not initialized to meaningful values until a disk
+	 * bitmap is provided. */
+	u32			m_bitmap_len;	/* Number of valid
+						 * bits available */
+
+	struct list_head	m_lru;		/* LRU of reservations
+						 * structures. */
+
+};
+
+void ocfs2_resv_init_once(struct ocfs2_alloc_reservation *resv);
+
+#define OCFS2_RESV_TYPES	(OCFS2_RESV_FLAG_TMP|OCFS2_RESV_FLAG_DIR)
+void ocfs2_resv_set_type(struct ocfs2_alloc_reservation *resv,
+			 unsigned int flags);
+
+int ocfs2_dir_resv_allowed(struct ocfs2_super *osb);
+
+/**
+ * ocfs2_resv_discard() - truncate a reservation
+ * @resmap:
+ * @resv: the reservation to truncate.
+ *
+ * After this function is called, the reservation will be empty, and
+ * unlinked from the rbtree.
+ */
+void ocfs2_resv_discard(struct ocfs2_reservation_map *resmap,
+			struct ocfs2_alloc_reservation *resv);
+
+
+/**
+ * ocfs2_resmap_init() - Initialize fields of a reservations bitmap
+ * @resmap: struct ocfs2_reservation_map to initialize
+ * @obj: unused for now
+ * @ops: unused for now
+ * @max_bitmap_bytes: Maximum size of the bitmap (typically blocksize)
+ *
+ * Only possible return value other than '0' is -ENOMEM for failure to
+ * allocation mirror bitmap.
+ */
+int ocfs2_resmap_init(struct ocfs2_super *osb,
+		      struct ocfs2_reservation_map *resmap);
+
+/**
+ * ocfs2_resmap_restart() - "restart" a reservation bitmap
+ * @resmap: reservations bitmap
+ * @clen: Number of valid bits in the bitmap
+ * @disk_bitmap: the disk bitmap this resmap should refer to.
+ *
+ * Re-initialize the parameters of a reservation bitmap. This is
+ * useful for local alloc window slides.
+ *
+ * This function will call ocfs2_trunc_resv against all existing
+ * reservations. A future version will recalculate existing
+ * reservations based on the new bitmap.
+ */
+void ocfs2_resmap_restart(struct ocfs2_reservation_map *resmap,
+			  unsigned int clen, char *disk_bitmap);
+
+/**
+ * ocfs2_resmap_uninit() - uninitialize a reservation bitmap structure
+ * @resmap: the struct ocfs2_reservation_map to uninitialize
+ */
+void ocfs2_resmap_uninit(struct ocfs2_reservation_map *resmap);
+
+/**
+ * ocfs2_resmap_resv_bits() - Return still-valid reservation bits
+ * @resmap: reservations bitmap
+ * @resv: reservation to base search from
+ * @cstart: start of proposed allocation
+ * @clen: length (in clusters) of proposed allocation
+ *
+ * Using the reservation data from resv, this function will compare
+ * resmap and resmap->m_disk_bitmap to determine what part (if any) of
+ * the reservation window is still clear to use. If resv is empty,
+ * this function will try to allocate a window for it.
+ *
+ * On success, zero is returned and the valid allocation area is set in cstart
+ * and clen.
+ *
+ * Returns -ENOSPC if reservations are disabled.
+ */
+int ocfs2_resmap_resv_bits(struct ocfs2_reservation_map *resmap,
+			   struct ocfs2_alloc_reservation *resv,
+			   int *cstart, int *clen);
+
+/**
+ * ocfs2_resmap_claimed_bits() - Tell the reservation code that bits were used.
+ * @resmap: reservations bitmap
+ * @resv: optional reservation to recalulate based on new bitmap
+ * @cstart: start of allocation in clusters
+ * @clen: end of allocation in clusters.
+ *
+ * Tell the reservation code that bits were used to fulfill allocation in
+ * resmap. The bits don't have to have been part of any existing
+ * reservation. But we must always call this function when bits are claimed.
+ * Internally, the reservations code will use this information to mark the
+ * reservations bitmap. If resv is passed, it's next allocation window will be
+ * calculated. It also expects that 'cstart' is the same as we passed back
+ * from ocfs2_resmap_resv_bits().
+ */
+void ocfs2_resmap_claimed_bits(struct ocfs2_reservation_map *resmap,
+			       struct ocfs2_alloc_reservation *resv,
+			       u32 cstart, u32 clen);
+
+#endif	/* OCFS2_RESERVATIONS_H */
diff --git a/fs/ocfs2/resize.c b/fs/ocfs2/resize.c
index 3c3d673a4d20..dacd553d8617 100644
--- a/fs/ocfs2/resize.c
+++ b/fs/ocfs2/resize.c
@@ -134,11 +134,7 @@ static int ocfs2_update_last_group_and_inode(handle_t *handle,
 		le16_add_cpu(&group->bg_free_bits_count, -1 * backups);
 	}
 
-	ret = ocfs2_journal_dirty(handle, group_bh);
-	if (ret < 0) {
-		mlog_errno(ret);
-		goto out_rollback;
-	}
+	ocfs2_journal_dirty(handle, group_bh);
 
 	/* update the inode accordingly. */
 	ret = ocfs2_journal_access_di(handle, INODE_CACHE(bm_inode), bm_bh,
@@ -319,7 +315,8 @@ int ocfs2_group_extend(struct inode * inode, int new_clusters)
 	BUG_ON(!OCFS2_IS_VALID_DINODE(fe));
 
 	if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
-				 ocfs2_group_bitmap_size(osb->sb) * 8) {
+		ocfs2_group_bitmap_size(osb->sb, 0,
+					osb->s_feature_incompat) * 8) {
 		mlog(ML_ERROR, "The disk is too old and small. "
 		     "Force to do offline resize.");
 		ret = -EINVAL;
@@ -500,7 +497,8 @@ int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input)
 	fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
 
 	if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
-				 ocfs2_group_bitmap_size(osb->sb) * 8) {
+		ocfs2_group_bitmap_size(osb->sb, 0,
+					osb->s_feature_incompat) * 8) {
 		mlog(ML_ERROR, "The disk is too old and small."
 		     " Force to do offline resize.");
 		ret = -EINVAL;
@@ -545,12 +543,7 @@ int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input)
 
 	group = (struct ocfs2_group_desc *)group_bh->b_data;
 	group->bg_next_group = cr->c_blkno;
-
-	ret = ocfs2_journal_dirty(handle, group_bh);
-	if (ret < 0) {
-		mlog_errno(ret);
-		goto out_commit;
-	}
+	ocfs2_journal_dirty(handle, group_bh);
 
 	ret = ocfs2_journal_access_di(handle, INODE_CACHE(main_bm_inode),
 				      main_bm_bh, OCFS2_JOURNAL_ACCESS_WRITE);
diff --git a/fs/ocfs2/stack_o2cb.c b/fs/ocfs2/stack_o2cb.c
index e49c41050264..0d3049f696c5 100644
--- a/fs/ocfs2/stack_o2cb.c
+++ b/fs/ocfs2/stack_o2cb.c
@@ -19,6 +19,7 @@
 
 #include <linux/kernel.h>
 #include <linux/crc32.h>
+#include <linux/slab.h>
 #include <linux/module.h>
 
 /* Needed for AOP_TRUNCATED_PAGE in mlog_errno() */
@@ -161,24 +162,23 @@ static int dlm_status_to_errno(enum dlm_status status)
 
 static void o2dlm_lock_ast_wrapper(void *astarg)
 {
-	BUG_ON(o2cb_stack.sp_proto == NULL);
+	struct ocfs2_dlm_lksb *lksb = astarg;
 
-	o2cb_stack.sp_proto->lp_lock_ast(astarg);
+	lksb->lksb_conn->cc_proto->lp_lock_ast(lksb);
 }
 
 static void o2dlm_blocking_ast_wrapper(void *astarg, int level)
 {
-	BUG_ON(o2cb_stack.sp_proto == NULL);
+	struct ocfs2_dlm_lksb *lksb = astarg;
 
-	o2cb_stack.sp_proto->lp_blocking_ast(astarg, level);
+	lksb->lksb_conn->cc_proto->lp_blocking_ast(lksb, level);
 }
 
 static void o2dlm_unlock_ast_wrapper(void *astarg, enum dlm_status status)
 {
+	struct ocfs2_dlm_lksb *lksb = astarg;
 	int error = dlm_status_to_errno(status);
 
-	BUG_ON(o2cb_stack.sp_proto == NULL);
-
 	/*
 	 * In o2dlm, you can get both the lock_ast() for the lock being
 	 * granted and the unlock_ast() for the CANCEL failing.  A
@@ -193,16 +193,15 @@ static void o2dlm_unlock_ast_wrapper(void *astarg, enum dlm_status status)
 	if (status == DLM_CANCELGRANT)
 		return;
 
-	o2cb_stack.sp_proto->lp_unlock_ast(astarg, error);
+	lksb->lksb_conn->cc_proto->lp_unlock_ast(lksb, error);
 }
 
 static int o2cb_dlm_lock(struct ocfs2_cluster_connection *conn,
 			 int mode,
-			 union ocfs2_dlm_lksb *lksb,
+			 struct ocfs2_dlm_lksb *lksb,
 			 u32 flags,
 			 void *name,
-			 unsigned int namelen,
-			 void *astarg)
+			 unsigned int namelen)
 {
 	enum dlm_status status;
 	int o2dlm_mode = mode_to_o2dlm(mode);
@@ -211,28 +210,27 @@ static int o2cb_dlm_lock(struct ocfs2_cluster_connection *conn,
 
 	status = dlmlock(conn->cc_lockspace, o2dlm_mode, &lksb->lksb_o2dlm,
 			 o2dlm_flags, name, namelen,
-			 o2dlm_lock_ast_wrapper, astarg,
+			 o2dlm_lock_ast_wrapper, lksb,
 			 o2dlm_blocking_ast_wrapper);
 	ret = dlm_status_to_errno(status);
 	return ret;
 }
 
 static int o2cb_dlm_unlock(struct ocfs2_cluster_connection *conn,
-			   union ocfs2_dlm_lksb *lksb,
-			   u32 flags,
-			   void *astarg)
+			   struct ocfs2_dlm_lksb *lksb,
+			   u32 flags)
 {
 	enum dlm_status status;
 	int o2dlm_flags = flags_to_o2dlm(flags);
 	int ret;
 
 	status = dlmunlock(conn->cc_lockspace, &lksb->lksb_o2dlm,
-			   o2dlm_flags, o2dlm_unlock_ast_wrapper, astarg);
+			   o2dlm_flags, o2dlm_unlock_ast_wrapper, lksb);
 	ret = dlm_status_to_errno(status);
 	return ret;
 }
 
-static int o2cb_dlm_lock_status(union ocfs2_dlm_lksb *lksb)
+static int o2cb_dlm_lock_status(struct ocfs2_dlm_lksb *lksb)
 {
 	return dlm_status_to_errno(lksb->lksb_o2dlm.status);
 }
@@ -242,17 +240,17 @@ static int o2cb_dlm_lock_status(union ocfs2_dlm_lksb *lksb)
  * contents, it will zero out the LVB.  Thus the caller can always trust
  * the contents.
  */
-static int o2cb_dlm_lvb_valid(union ocfs2_dlm_lksb *lksb)
+static int o2cb_dlm_lvb_valid(struct ocfs2_dlm_lksb *lksb)
 {
 	return 1;
 }
 
-static void *o2cb_dlm_lvb(union ocfs2_dlm_lksb *lksb)
+static void *o2cb_dlm_lvb(struct ocfs2_dlm_lksb *lksb)
 {
 	return (void *)(lksb->lksb_o2dlm.lvb);
 }
 
-static void o2cb_dump_lksb(union ocfs2_dlm_lksb *lksb)
+static void o2cb_dump_lksb(struct ocfs2_dlm_lksb *lksb)
 {
 	dlm_print_one_lock(lksb->lksb_o2dlm.lockid);
 }
@@ -277,10 +275,10 @@ static int o2cb_cluster_connect(struct ocfs2_cluster_connection *conn)
 	u32 dlm_key;
 	struct dlm_ctxt *dlm;
 	struct o2dlm_private *priv;
-	struct dlm_protocol_version dlm_version;
+	struct dlm_protocol_version fs_version;
 
 	BUG_ON(conn == NULL);
-	BUG_ON(o2cb_stack.sp_proto == NULL);
+	BUG_ON(conn->cc_proto == NULL);
 
 	/* for now we only have one cluster/node, make sure we see it
 	 * in the heartbeat universe */
@@ -304,18 +302,18 @@ static int o2cb_cluster_connect(struct ocfs2_cluster_connection *conn)
 	/* used by the dlm code to make message headers unique, each
 	 * node in this domain must agree on this. */
 	dlm_key = crc32_le(0, conn->cc_name, conn->cc_namelen);
-	dlm_version.pv_major = conn->cc_version.pv_major;
-	dlm_version.pv_minor = conn->cc_version.pv_minor;
+	fs_version.pv_major = conn->cc_version.pv_major;
+	fs_version.pv_minor = conn->cc_version.pv_minor;
 
-	dlm = dlm_register_domain(conn->cc_name, dlm_key, &dlm_version);
+	dlm = dlm_register_domain(conn->cc_name, dlm_key, &fs_version);
 	if (IS_ERR(dlm)) {
 		rc = PTR_ERR(dlm);
 		mlog_errno(rc);
 		goto out_free;
 	}
 
-	conn->cc_version.pv_major = dlm_version.pv_major;
-	conn->cc_version.pv_minor = dlm_version.pv_minor;
+	conn->cc_version.pv_major = fs_version.pv_major;
+	conn->cc_version.pv_minor = fs_version.pv_minor;
 	conn->cc_lockspace = dlm;
 
 	dlm_register_eviction_cb(dlm, &priv->op_eviction_cb);
diff --git a/fs/ocfs2/stack_user.c b/fs/ocfs2/stack_user.c
index da78a2a334fd..2dc57bca0688 100644
--- a/fs/ocfs2/stack_user.c
+++ b/fs/ocfs2/stack_user.c
@@ -21,11 +21,11 @@
 #include <linux/fs.h>
 #include <linux/miscdevice.h>
 #include <linux/mutex.h>
+#include <linux/slab.h>
 #include <linux/smp_lock.h>
 #include <linux/reboot.h>
 #include <asm/uaccess.h>
 
-#include "ocfs2.h"  /* For struct ocfs2_lock_res */
 #include "stackglue.h"
 
 #include <linux/dlm_plock.h>
@@ -63,8 +63,8 @@
  * negotiated by the client.  The client negotiates based on the maximum
  * version advertised in /sys/fs/ocfs2/max_locking_protocol.  The major
  * number from the "SETV" message must match
- * ocfs2_user_plugin.sp_proto->lp_max_version.pv_major, and the minor number
- * must be less than or equal to ...->lp_max_version.pv_minor.
+ * ocfs2_user_plugin.sp_max_proto.pv_major, and the minor number
+ * must be less than or equal to ...sp_max_version.pv_minor.
  *
  * Once this information has been set, mounts will be allowed.  From this
  * point on, the "DOWN" message can be sent for node down notification.
@@ -401,7 +401,7 @@ static int ocfs2_control_do_setversion_msg(struct file *file,
 	char *ptr = NULL;
 	struct ocfs2_control_private *p = file->private_data;
 	struct ocfs2_protocol_version *max =
-		&ocfs2_user_plugin.sp_proto->lp_max_version;
+		&ocfs2_user_plugin.sp_max_proto;
 
 	if (ocfs2_control_get_handshake_state(file) !=
 	    OCFS2_CONTROL_HANDSHAKE_PROTOCOL)
@@ -664,18 +664,10 @@ static void ocfs2_control_exit(void)
 		       -rc);
 }
 
-static struct dlm_lksb *fsdlm_astarg_to_lksb(void *astarg)
-{
-	struct ocfs2_lock_res *res = astarg;
-	return &res->l_lksb.lksb_fsdlm;
-}
-
 static void fsdlm_lock_ast_wrapper(void *astarg)
 {
-	struct dlm_lksb *lksb = fsdlm_astarg_to_lksb(astarg);
-	int status = lksb->sb_status;
-
-	BUG_ON(ocfs2_user_plugin.sp_proto == NULL);
+	struct ocfs2_dlm_lksb *lksb = astarg;
+	int status = lksb->lksb_fsdlm.sb_status;
 
 	/*
 	 * For now we're punting on the issue of other non-standard errors
@@ -688,25 +680,24 @@ static void fsdlm_lock_ast_wrapper(void *astarg)
 	 */
 
 	if (status == -DLM_EUNLOCK || status == -DLM_ECANCEL)
-		ocfs2_user_plugin.sp_proto->lp_unlock_ast(astarg, 0);
+		lksb->lksb_conn->cc_proto->lp_unlock_ast(lksb, 0);
 	else
-		ocfs2_user_plugin.sp_proto->lp_lock_ast(astarg);
+		lksb->lksb_conn->cc_proto->lp_lock_ast(lksb);
 }
 
 static void fsdlm_blocking_ast_wrapper(void *astarg, int level)
 {
-	BUG_ON(ocfs2_user_plugin.sp_proto == NULL);
+	struct ocfs2_dlm_lksb *lksb = astarg;
 
-	ocfs2_user_plugin.sp_proto->lp_blocking_ast(astarg, level);
+	lksb->lksb_conn->cc_proto->lp_blocking_ast(lksb, level);
 }
 
 static int user_dlm_lock(struct ocfs2_cluster_connection *conn,
 			 int mode,
-			 union ocfs2_dlm_lksb *lksb,
+			 struct ocfs2_dlm_lksb *lksb,
 			 u32 flags,
 			 void *name,
-			 unsigned int namelen,
-			 void *astarg)
+			 unsigned int namelen)
 {
 	int ret;
 
@@ -716,36 +707,35 @@ static int user_dlm_lock(struct ocfs2_cluster_connection *conn,
 
 	ret = dlm_lock(conn->cc_lockspace, mode, &lksb->lksb_fsdlm,
 		       flags|DLM_LKF_NODLCKWT, name, namelen, 0,
-		       fsdlm_lock_ast_wrapper, astarg,
+		       fsdlm_lock_ast_wrapper, lksb,
 		       fsdlm_blocking_ast_wrapper);
 	return ret;
 }
 
 static int user_dlm_unlock(struct ocfs2_cluster_connection *conn,
-			   union ocfs2_dlm_lksb *lksb,
-			   u32 flags,
-			   void *astarg)
+			   struct ocfs2_dlm_lksb *lksb,
+			   u32 flags)
 {
 	int ret;
 
 	ret = dlm_unlock(conn->cc_lockspace, lksb->lksb_fsdlm.sb_lkid,
-			 flags, &lksb->lksb_fsdlm, astarg);
+			 flags, &lksb->lksb_fsdlm, lksb);
 	return ret;
 }
 
-static int user_dlm_lock_status(union ocfs2_dlm_lksb *lksb)
+static int user_dlm_lock_status(struct ocfs2_dlm_lksb *lksb)
 {
 	return lksb->lksb_fsdlm.sb_status;
 }
 
-static int user_dlm_lvb_valid(union ocfs2_dlm_lksb *lksb)
+static int user_dlm_lvb_valid(struct ocfs2_dlm_lksb *lksb)
 {
 	int invalid = lksb->lksb_fsdlm.sb_flags & DLM_SBF_VALNOTVALID;
 
 	return !invalid;
 }
 
-static void *user_dlm_lvb(union ocfs2_dlm_lksb *lksb)
+static void *user_dlm_lvb(struct ocfs2_dlm_lksb *lksb)
 {
 	if (!lksb->lksb_fsdlm.sb_lvbptr)
 		lksb->lksb_fsdlm.sb_lvbptr = (char *)lksb +
@@ -753,7 +743,7 @@ static void *user_dlm_lvb(union ocfs2_dlm_lksb *lksb)
 	return (void *)(lksb->lksb_fsdlm.sb_lvbptr);
 }
 
-static void user_dlm_dump_lksb(union ocfs2_dlm_lksb *lksb)
+static void user_dlm_dump_lksb(struct ocfs2_dlm_lksb *lksb)
 {
 }
 
diff --git a/fs/ocfs2/stackglue.c b/fs/ocfs2/stackglue.c
index f3df0baa9a48..39abf89697ed 100644
--- a/fs/ocfs2/stackglue.c
+++ b/fs/ocfs2/stackglue.c
@@ -36,7 +36,7 @@
 #define OCFS2_STACK_PLUGIN_USER		"user"
 #define OCFS2_MAX_HB_CTL_PATH		256
 
-static struct ocfs2_locking_protocol *lproto;
+static struct ocfs2_protocol_version locking_max_version;
 static DEFINE_SPINLOCK(ocfs2_stack_lock);
 static LIST_HEAD(ocfs2_stack_list);
 static char cluster_stack_name[OCFS2_STACK_LABEL_LEN + 1];
@@ -176,7 +176,7 @@ int ocfs2_stack_glue_register(struct ocfs2_stack_plugin *plugin)
 	spin_lock(&ocfs2_stack_lock);
 	if (!ocfs2_stack_lookup(plugin->sp_name)) {
 		plugin->sp_count = 0;
-		plugin->sp_proto = lproto;
+		plugin->sp_max_proto = locking_max_version;
 		list_add(&plugin->sp_list, &ocfs2_stack_list);
 		printk(KERN_INFO "ocfs2: Registered cluster interface %s\n",
 		       plugin->sp_name);
@@ -213,77 +213,76 @@ void ocfs2_stack_glue_unregister(struct ocfs2_stack_plugin *plugin)
 }
 EXPORT_SYMBOL_GPL(ocfs2_stack_glue_unregister);
 
-void ocfs2_stack_glue_set_locking_protocol(struct ocfs2_locking_protocol *proto)
+void ocfs2_stack_glue_set_max_proto_version(struct ocfs2_protocol_version *max_proto)
 {
 	struct ocfs2_stack_plugin *p;
 
-	BUG_ON(proto == NULL);
-
 	spin_lock(&ocfs2_stack_lock);
-	BUG_ON(active_stack != NULL);
+	if (memcmp(max_proto, &locking_max_version,
+		   sizeof(struct ocfs2_protocol_version))) {
+		BUG_ON(locking_max_version.pv_major != 0);
 
-	lproto = proto;
-	list_for_each_entry(p, &ocfs2_stack_list, sp_list) {
-		p->sp_proto = lproto;
+		locking_max_version = *max_proto;
+		list_for_each_entry(p, &ocfs2_stack_list, sp_list) {
+			p->sp_max_proto = locking_max_version;
+		}
 	}
-
 	spin_unlock(&ocfs2_stack_lock);
 }
-EXPORT_SYMBOL_GPL(ocfs2_stack_glue_set_locking_protocol);
+EXPORT_SYMBOL_GPL(ocfs2_stack_glue_set_max_proto_version);
 
 
 /*
- * The ocfs2_dlm_lock() and ocfs2_dlm_unlock() functions take
- * "struct ocfs2_lock_res *astarg" instead of "void *astarg" because the
- * underlying stack plugins need to pilfer the lksb off of the lock_res.
- * If some other structure needs to be passed as an astarg, the plugins
- * will need to be given a different avenue to the lksb.
+ * The ocfs2_dlm_lock() and ocfs2_dlm_unlock() functions take no argument
+ * for the ast and bast functions.  They will pass the lksb to the ast
+ * and bast.  The caller can wrap the lksb with their own structure to
+ * get more information.
  */
 int ocfs2_dlm_lock(struct ocfs2_cluster_connection *conn,
 		   int mode,
-		   union ocfs2_dlm_lksb *lksb,
+		   struct ocfs2_dlm_lksb *lksb,
 		   u32 flags,
 		   void *name,
-		   unsigned int namelen,
-		   struct ocfs2_lock_res *astarg)
+		   unsigned int namelen)
 {
-	BUG_ON(lproto == NULL);
-
+	if (!lksb->lksb_conn)
+		lksb->lksb_conn = conn;
+	else
+		BUG_ON(lksb->lksb_conn != conn);
 	return active_stack->sp_ops->dlm_lock(conn, mode, lksb, flags,
-					      name, namelen, astarg);
+					      name, namelen);
 }
 EXPORT_SYMBOL_GPL(ocfs2_dlm_lock);
 
 int ocfs2_dlm_unlock(struct ocfs2_cluster_connection *conn,
-		     union ocfs2_dlm_lksb *lksb,
-		     u32 flags,
-		     struct ocfs2_lock_res *astarg)
+		     struct ocfs2_dlm_lksb *lksb,
+		     u32 flags)
 {
-	BUG_ON(lproto == NULL);
+	BUG_ON(lksb->lksb_conn == NULL);
 
-	return active_stack->sp_ops->dlm_unlock(conn, lksb, flags, astarg);
+	return active_stack->sp_ops->dlm_unlock(conn, lksb, flags);
 }
 EXPORT_SYMBOL_GPL(ocfs2_dlm_unlock);
 
-int ocfs2_dlm_lock_status(union ocfs2_dlm_lksb *lksb)
+int ocfs2_dlm_lock_status(struct ocfs2_dlm_lksb *lksb)
 {
 	return active_stack->sp_ops->lock_status(lksb);
 }
 EXPORT_SYMBOL_GPL(ocfs2_dlm_lock_status);
 
-int ocfs2_dlm_lvb_valid(union ocfs2_dlm_lksb *lksb)
+int ocfs2_dlm_lvb_valid(struct ocfs2_dlm_lksb *lksb)
 {
 	return active_stack->sp_ops->lvb_valid(lksb);
 }
 EXPORT_SYMBOL_GPL(ocfs2_dlm_lvb_valid);
 
-void *ocfs2_dlm_lvb(union ocfs2_dlm_lksb *lksb)
+void *ocfs2_dlm_lvb(struct ocfs2_dlm_lksb *lksb)
 {
 	return active_stack->sp_ops->lock_lvb(lksb);
 }
 EXPORT_SYMBOL_GPL(ocfs2_dlm_lvb);
 
-void ocfs2_dlm_dump_lksb(union ocfs2_dlm_lksb *lksb)
+void ocfs2_dlm_dump_lksb(struct ocfs2_dlm_lksb *lksb)
 {
 	active_stack->sp_ops->dump_lksb(lksb);
 }
@@ -312,6 +311,7 @@ EXPORT_SYMBOL_GPL(ocfs2_plock);
 int ocfs2_cluster_connect(const char *stack_name,
 			  const char *group,
 			  int grouplen,
+			  struct ocfs2_locking_protocol *lproto,
 			  void (*recovery_handler)(int node_num,
 						   void *recovery_data),
 			  void *recovery_data,
@@ -329,6 +329,12 @@ int ocfs2_cluster_connect(const char *stack_name,
 		goto out;
 	}
 
+	if (memcmp(&lproto->lp_max_version, &locking_max_version,
+		   sizeof(struct ocfs2_protocol_version))) {
+		rc = -EINVAL;
+		goto out;
+	}
+
 	new_conn = kzalloc(sizeof(struct ocfs2_cluster_connection),
 			   GFP_KERNEL);
 	if (!new_conn) {
@@ -341,6 +347,7 @@ int ocfs2_cluster_connect(const char *stack_name,
 	new_conn->cc_recovery_handler = recovery_handler;
 	new_conn->cc_recovery_data = recovery_data;
 
+	new_conn->cc_proto = lproto;
 	/* Start the new connection at our maximum compatibility level */
 	new_conn->cc_version = lproto->lp_max_version;
 
@@ -366,6 +373,24 @@ out:
 }
 EXPORT_SYMBOL_GPL(ocfs2_cluster_connect);
 
+/* The caller will ensure all nodes have the same cluster stack */
+int ocfs2_cluster_connect_agnostic(const char *group,
+				   int grouplen,
+				   struct ocfs2_locking_protocol *lproto,
+				   void (*recovery_handler)(int node_num,
+							    void *recovery_data),
+				   void *recovery_data,
+				   struct ocfs2_cluster_connection **conn)
+{
+	char *stack_name = NULL;
+
+	if (cluster_stack_name[0])
+		stack_name = cluster_stack_name;
+	return ocfs2_cluster_connect(stack_name, group, grouplen, lproto,
+				     recovery_handler, recovery_data, conn);
+}
+EXPORT_SYMBOL_GPL(ocfs2_cluster_connect_agnostic);
+
 /* If hangup_pending is 0, the stack driver will be dropped */
 int ocfs2_cluster_disconnect(struct ocfs2_cluster_connection *conn,
 			     int hangup_pending)
@@ -453,10 +478,10 @@ static ssize_t ocfs2_max_locking_protocol_show(struct kobject *kobj,
 	ssize_t ret = 0;
 
 	spin_lock(&ocfs2_stack_lock);
-	if (lproto)
+	if (locking_max_version.pv_major)
 		ret = snprintf(buf, PAGE_SIZE, "%u.%u\n",
-			       lproto->lp_max_version.pv_major,
-			       lproto->lp_max_version.pv_minor);
+			       locking_max_version.pv_major,
+			       locking_max_version.pv_minor);
 	spin_unlock(&ocfs2_stack_lock);
 
 	return ret;
@@ -685,7 +710,10 @@ static int __init ocfs2_stack_glue_init(void)
 
 static void __exit ocfs2_stack_glue_exit(void)
 {
-	lproto = NULL;
+	memset(&locking_max_version, 0,
+	       sizeof(struct ocfs2_protocol_version));
+	locking_max_version.pv_major = 0;
+	locking_max_version.pv_minor = 0;
 	ocfs2_sysfs_exit();
 	if (ocfs2_table_header)
 		unregister_sysctl_table(ocfs2_table_header);
diff --git a/fs/ocfs2/stackglue.h b/fs/ocfs2/stackglue.h
index 03a44d60eac9..8ce7398ae1d2 100644
--- a/fs/ocfs2/stackglue.h
+++ b/fs/ocfs2/stackglue.h
@@ -56,17 +56,6 @@ struct ocfs2_protocol_version {
 };
 
 /*
- * The ocfs2_locking_protocol defines the handlers called on ocfs2's behalf.
- */
-struct ocfs2_locking_protocol {
-	struct ocfs2_protocol_version lp_max_version;
-	void (*lp_lock_ast)(void *astarg);
-	void (*lp_blocking_ast)(void *astarg, int level);
-	void (*lp_unlock_ast)(void *astarg, int error);
-};
-
-
-/*
  * The dlm_lockstatus struct includes lvb space, but the dlm_lksb struct only
  * has a pointer to separately allocated lvb space.  This struct exists only to
  * include in the lksb union to make space for a combined dlm_lksb and lvb.
@@ -81,12 +70,27 @@ struct fsdlm_lksb_plus_lvb {
  * size of the union is known.  Lock status structures are embedded in
  * ocfs2 inodes.
  */
-union ocfs2_dlm_lksb {
-	struct dlm_lockstatus lksb_o2dlm;
-	struct dlm_lksb lksb_fsdlm;
-	struct fsdlm_lksb_plus_lvb padding;
+struct ocfs2_cluster_connection;
+struct ocfs2_dlm_lksb {
+	 union {
+		 struct dlm_lockstatus lksb_o2dlm;
+		 struct dlm_lksb lksb_fsdlm;
+		 struct fsdlm_lksb_plus_lvb padding;
+	 };
+	 struct ocfs2_cluster_connection *lksb_conn;
+};
+
+/*
+ * The ocfs2_locking_protocol defines the handlers called on ocfs2's behalf.
+ */
+struct ocfs2_locking_protocol {
+	struct ocfs2_protocol_version lp_max_version;
+	void (*lp_lock_ast)(struct ocfs2_dlm_lksb *lksb);
+	void (*lp_blocking_ast)(struct ocfs2_dlm_lksb *lksb, int level);
+	void (*lp_unlock_ast)(struct ocfs2_dlm_lksb *lksb, int error);
 };
 
+
 /*
  * A cluster connection.  Mostly opaque to ocfs2, the connection holds
  * state for the underlying stack.  ocfs2 does use cc_version to determine
@@ -96,6 +100,7 @@ struct ocfs2_cluster_connection {
 	char cc_name[GROUP_NAME_MAX];
 	int cc_namelen;
 	struct ocfs2_protocol_version cc_version;
+	struct ocfs2_locking_protocol *cc_proto;
 	void (*cc_recovery_handler)(int node_num, void *recovery_data);
 	void *cc_recovery_data;
 	void *cc_lockspace;
@@ -155,27 +160,29 @@ struct ocfs2_stack_operations {
 	 *
 	 * ast and bast functions are not part of the call because the
 	 * stack will likely want to wrap ast and bast calls before passing
-	 * them to stack->sp_proto.
+	 * them to stack->sp_proto.  There is no astarg.  The lksb will
+	 * be passed back to the ast and bast functions.  The caller can
+	 * use this to find their object.
 	 */
 	int (*dlm_lock)(struct ocfs2_cluster_connection *conn,
 			int mode,
-			union ocfs2_dlm_lksb *lksb,
+			struct ocfs2_dlm_lksb *lksb,
 			u32 flags,
 			void *name,
-			unsigned int namelen,
-			void *astarg);
+			unsigned int namelen);
 
 	/*
 	 * Call the underlying dlm unlock function.  The ->dlm_unlock()
 	 * function should convert the flags as appropriate.
 	 *
 	 * The unlock ast is not passed, as the stack will want to wrap
-	 * it before calling stack->sp_proto->lp_unlock_ast().
+	 * it before calling stack->sp_proto->lp_unlock_ast().  There is
+	 * no astarg.  The lksb will be passed back to the unlock ast
+	 * function.  The caller can use this to find their object.
 	 */
 	int (*dlm_unlock)(struct ocfs2_cluster_connection *conn,
-			  union ocfs2_dlm_lksb *lksb,
-			  u32 flags,
-			  void *astarg);
+			  struct ocfs2_dlm_lksb *lksb,
+			  u32 flags);
 
 	/*
 	 * Return the status of the current lock status block.  The fs
@@ -183,17 +190,17 @@ struct ocfs2_stack_operations {
 	 * callback pulls out the stack-specific lksb, converts the status
 	 * to a proper errno, and returns it.
 	 */
-	int (*lock_status)(union ocfs2_dlm_lksb *lksb);
+	int (*lock_status)(struct ocfs2_dlm_lksb *lksb);
 
 	/*
 	 * Return non-zero if the LVB is valid.
 	 */
-	int (*lvb_valid)(union ocfs2_dlm_lksb *lksb);
+	int (*lvb_valid)(struct ocfs2_dlm_lksb *lksb);
 
 	/*
 	 * Pull the lvb pointer off of the stack-specific lksb.
 	 */
-	void *(*lock_lvb)(union ocfs2_dlm_lksb *lksb);
+	void *(*lock_lvb)(struct ocfs2_dlm_lksb *lksb);
 
 	/*
 	 * Cluster-aware posix locks
@@ -210,7 +217,7 @@ struct ocfs2_stack_operations {
 	 * This is an optoinal debugging hook.  If provided, the
 	 * stack can dump debugging information about this lock.
 	 */
-	void (*dump_lksb)(union ocfs2_dlm_lksb *lksb);
+	void (*dump_lksb)(struct ocfs2_dlm_lksb *lksb);
 };
 
 /*
@@ -226,7 +233,7 @@ struct ocfs2_stack_plugin {
 	/* These are managed by the stackglue code. */
 	struct list_head sp_list;
 	unsigned int sp_count;
-	struct ocfs2_locking_protocol *sp_proto;
+	struct ocfs2_protocol_version sp_max_proto;
 };
 
 
@@ -234,10 +241,22 @@ struct ocfs2_stack_plugin {
 int ocfs2_cluster_connect(const char *stack_name,
 			  const char *group,
 			  int grouplen,
+			  struct ocfs2_locking_protocol *lproto,
 			  void (*recovery_handler)(int node_num,
 						   void *recovery_data),
 			  void *recovery_data,
 			  struct ocfs2_cluster_connection **conn);
+/*
+ * Used by callers that don't store their stack name.  They must ensure
+ * all nodes have the same stack.
+ */
+int ocfs2_cluster_connect_agnostic(const char *group,
+				   int grouplen,
+				   struct ocfs2_locking_protocol *lproto,
+				   void (*recovery_handler)(int node_num,
+							    void *recovery_data),
+				   void *recovery_data,
+				   struct ocfs2_cluster_connection **conn);
 int ocfs2_cluster_disconnect(struct ocfs2_cluster_connection *conn,
 			     int hangup_pending);
 void ocfs2_cluster_hangup(const char *group, int grouplen);
@@ -246,26 +265,24 @@ int ocfs2_cluster_this_node(unsigned int *node);
 struct ocfs2_lock_res;
 int ocfs2_dlm_lock(struct ocfs2_cluster_connection *conn,
 		   int mode,
-		   union ocfs2_dlm_lksb *lksb,
+		   struct ocfs2_dlm_lksb *lksb,
 		   u32 flags,
 		   void *name,
-		   unsigned int namelen,
-		   struct ocfs2_lock_res *astarg);
+		   unsigned int namelen);
 int ocfs2_dlm_unlock(struct ocfs2_cluster_connection *conn,
-		     union ocfs2_dlm_lksb *lksb,
-		     u32 flags,
-		     struct ocfs2_lock_res *astarg);
+		     struct ocfs2_dlm_lksb *lksb,
+		     u32 flags);
 
-int ocfs2_dlm_lock_status(union ocfs2_dlm_lksb *lksb);
-int ocfs2_dlm_lvb_valid(union ocfs2_dlm_lksb *lksb);
-void *ocfs2_dlm_lvb(union ocfs2_dlm_lksb *lksb);
-void ocfs2_dlm_dump_lksb(union ocfs2_dlm_lksb *lksb);
+int ocfs2_dlm_lock_status(struct ocfs2_dlm_lksb *lksb);
+int ocfs2_dlm_lvb_valid(struct ocfs2_dlm_lksb *lksb);
+void *ocfs2_dlm_lvb(struct ocfs2_dlm_lksb *lksb);
+void ocfs2_dlm_dump_lksb(struct ocfs2_dlm_lksb *lksb);
 
 int ocfs2_stack_supports_plocks(void);
 int ocfs2_plock(struct ocfs2_cluster_connection *conn, u64 ino,
 		struct file *file, int cmd, struct file_lock *fl);
 
-void ocfs2_stack_glue_set_locking_protocol(struct ocfs2_locking_protocol *proto);
+void ocfs2_stack_glue_set_max_proto_version(struct ocfs2_protocol_version *max_proto);
 
 
 /* Used by stack plugins */
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
index c30b644d9572..a8e6a95a353f 100644
--- a/fs/ocfs2/suballoc.c
+++ b/fs/ocfs2/suballoc.c
@@ -51,7 +51,16 @@
 #define ALLOC_NEW_GROUP			0x1
 #define ALLOC_GROUPS_FROM_GLOBAL	0x2
 
-#define OCFS2_MAX_INODES_TO_STEAL	1024
+#define OCFS2_MAX_TO_STEAL		1024
+
+struct ocfs2_suballoc_result {
+	u64		sr_bg_blkno;	/* The bg we allocated from.  Set
+					   to 0 when a block group is
+					   contiguous. */
+	u64		sr_blkno;	/* The first allocated block */
+	unsigned int	sr_bit_offset;	/* The bit in the bg */
+	unsigned int	sr_bits;	/* How many bits we claimed */
+};
 
 static inline void ocfs2_debug_bg(struct ocfs2_group_desc *bg);
 static inline void ocfs2_debug_suballoc_inode(struct ocfs2_dinode *fe);
@@ -60,6 +69,7 @@ static int ocfs2_block_group_fill(handle_t *handle,
 				  struct inode *alloc_inode,
 				  struct buffer_head *bg_bh,
 				  u64 group_blkno,
+				  unsigned int group_clusters,
 				  u16 my_chain,
 				  struct ocfs2_chain_list *cl);
 static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
@@ -73,20 +83,17 @@ static int ocfs2_cluster_group_search(struct inode *inode,
 				      struct buffer_head *group_bh,
 				      u32 bits_wanted, u32 min_bits,
 				      u64 max_block,
-				      u16 *bit_off, u16 *bits_found);
+				      struct ocfs2_suballoc_result *res);
 static int ocfs2_block_group_search(struct inode *inode,
 				    struct buffer_head *group_bh,
 				    u32 bits_wanted, u32 min_bits,
 				    u64 max_block,
-				    u16 *bit_off, u16 *bits_found);
-static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
-				     struct ocfs2_alloc_context *ac,
+				    struct ocfs2_suballoc_result *res);
+static int ocfs2_claim_suballoc_bits(struct ocfs2_alloc_context *ac,
 				     handle_t *handle,
 				     u32 bits_wanted,
 				     u32 min_bits,
-				     u16 *bit_off,
-				     unsigned int *num_bits,
-				     u64 *bg_blkno);
+				     struct ocfs2_suballoc_result *res);
 static int ocfs2_test_bg_bit_allocatable(struct buffer_head *bg_bh,
 					 int nr);
 static inline int ocfs2_block_group_set_bits(handle_t *handle,
@@ -95,13 +102,6 @@ static inline int ocfs2_block_group_set_bits(handle_t *handle,
 					     struct buffer_head *group_bh,
 					     unsigned int bit_off,
 					     unsigned int num_bits);
-static inline int ocfs2_block_group_clear_bits(handle_t *handle,
-					       struct inode *alloc_inode,
-					       struct ocfs2_group_desc *bg,
-					       struct buffer_head *group_bh,
-					       unsigned int bit_off,
-					       unsigned int num_bits);
-
 static int ocfs2_relink_block_group(handle_t *handle,
 				    struct inode *alloc_inode,
 				    struct buffer_head *fe_bh,
@@ -137,6 +137,7 @@ void ocfs2_free_ac_resource(struct ocfs2_alloc_context *ac)
 	}
 	brelse(ac->ac_bh);
 	ac->ac_bh = NULL;
+	ac->ac_resv = NULL;
 }
 
 void ocfs2_free_alloc_context(struct ocfs2_alloc_context *ac)
@@ -152,7 +153,7 @@ static u32 ocfs2_bits_per_group(struct ocfs2_chain_list *cl)
 
 #define do_error(fmt, ...)						\
 	do{								\
-		if (clean_error)					\
+		if (resize)					\
 			mlog(ML_ERROR, fmt "\n", ##__VA_ARGS__);	\
 		else							\
 			ocfs2_error(sb, fmt, ##__VA_ARGS__);		\
@@ -160,7 +161,7 @@ static u32 ocfs2_bits_per_group(struct ocfs2_chain_list *cl)
 
 static int ocfs2_validate_gd_self(struct super_block *sb,
 				  struct buffer_head *bh,
-				  int clean_error)
+				  int resize)
 {
 	struct ocfs2_group_desc *gd = (struct ocfs2_group_desc *)bh->b_data;
 
@@ -211,7 +212,7 @@ static int ocfs2_validate_gd_self(struct super_block *sb,
 static int ocfs2_validate_gd_parent(struct super_block *sb,
 				    struct ocfs2_dinode *di,
 				    struct buffer_head *bh,
-				    int clean_error)
+				    int resize)
 {
 	unsigned int max_bits;
 	struct ocfs2_group_desc *gd = (struct ocfs2_group_desc *)bh->b_data;
@@ -233,8 +234,11 @@ static int ocfs2_validate_gd_parent(struct super_block *sb,
 		return -EINVAL;
 	}
 
-	if (le16_to_cpu(gd->bg_chain) >=
-	    le16_to_cpu(di->id2.i_chain.cl_next_free_rec)) {
+	/* In resize, we may meet the case bg_chain == cl_next_free_rec. */
+	if ((le16_to_cpu(gd->bg_chain) >
+	     le16_to_cpu(di->id2.i_chain.cl_next_free_rec)) ||
+	    ((le16_to_cpu(gd->bg_chain) ==
+	     le16_to_cpu(di->id2.i_chain.cl_next_free_rec)) && !resize)) {
 		do_error("Group descriptor #%llu has bad chain %u",
 			 (unsigned long long)bh->b_blocknr,
 			 le16_to_cpu(gd->bg_chain));
@@ -329,14 +333,38 @@ out:
 	return rc;
 }
 
+static void ocfs2_bg_discontig_add_extent(struct ocfs2_super *osb,
+					  struct ocfs2_group_desc *bg,
+					  struct ocfs2_chain_list *cl,
+					  u64 p_blkno, u32 clusters)
+{
+	struct ocfs2_extent_list *el = &bg->bg_list;
+	struct ocfs2_extent_rec *rec;
+
+	BUG_ON(!ocfs2_supports_discontig_bg(osb));
+	if (!el->l_next_free_rec)
+		el->l_count = cpu_to_le16(ocfs2_extent_recs_per_gd(osb->sb));
+	rec = &el->l_recs[le16_to_cpu(el->l_next_free_rec)];
+	rec->e_blkno = cpu_to_le64(p_blkno);
+	rec->e_cpos = cpu_to_le32(le16_to_cpu(bg->bg_bits) /
+				  le16_to_cpu(cl->cl_bpc));
+	rec->e_leaf_clusters = cpu_to_le32(clusters);
+	le16_add_cpu(&bg->bg_bits, clusters * le16_to_cpu(cl->cl_bpc));
+	le16_add_cpu(&bg->bg_free_bits_count,
+		     clusters * le16_to_cpu(cl->cl_bpc));
+	le16_add_cpu(&el->l_next_free_rec, 1);
+}
+
 static int ocfs2_block_group_fill(handle_t *handle,
 				  struct inode *alloc_inode,
 				  struct buffer_head *bg_bh,
 				  u64 group_blkno,
+				  unsigned int group_clusters,
 				  u16 my_chain,
 				  struct ocfs2_chain_list *cl)
 {
 	int status = 0;
+	struct ocfs2_super *osb = OCFS2_SB(alloc_inode->i_sb);
 	struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) bg_bh->b_data;
 	struct super_block * sb = alloc_inode->i_sb;
 
@@ -363,19 +391,23 @@ static int ocfs2_block_group_fill(handle_t *handle,
 	memset(bg, 0, sb->s_blocksize);
 	strcpy(bg->bg_signature, OCFS2_GROUP_DESC_SIGNATURE);
 	bg->bg_generation = cpu_to_le32(OCFS2_SB(sb)->fs_generation);
-	bg->bg_size = cpu_to_le16(ocfs2_group_bitmap_size(sb));
-	bg->bg_bits = cpu_to_le16(ocfs2_bits_per_group(cl));
+	bg->bg_size = cpu_to_le16(ocfs2_group_bitmap_size(sb, 1,
+						osb->s_feature_incompat));
 	bg->bg_chain = cpu_to_le16(my_chain);
 	bg->bg_next_group = cl->cl_recs[my_chain].c_blkno;
 	bg->bg_parent_dinode = cpu_to_le64(OCFS2_I(alloc_inode)->ip_blkno);
 	bg->bg_blkno = cpu_to_le64(group_blkno);
+	if (group_clusters == le16_to_cpu(cl->cl_cpg))
+		bg->bg_bits = cpu_to_le16(ocfs2_bits_per_group(cl));
+	else
+		ocfs2_bg_discontig_add_extent(osb, bg, cl, group_blkno,
+					      group_clusters);
+
 	/* set the 1st bit in the bitmap to account for the descriptor block */
 	ocfs2_set_bit(0, (unsigned long *)bg->bg_bitmap);
 	bg->bg_free_bits_count = cpu_to_le16(le16_to_cpu(bg->bg_bits) - 1);
 
-	status = ocfs2_journal_dirty(handle, bg_bh);
-	if (status < 0)
-		mlog_errno(status);
+	ocfs2_journal_dirty(handle, bg_bh);
 
 	/* There is no need to zero out or otherwise initialize the
 	 * other blocks in a group - All valid FS metadata in a block
@@ -401,6 +433,238 @@ static inline u16 ocfs2_find_smallest_chain(struct ocfs2_chain_list *cl)
 	return best;
 }
 
+static struct buffer_head *
+ocfs2_block_group_alloc_contig(struct ocfs2_super *osb, handle_t *handle,
+			       struct inode *alloc_inode,
+			       struct ocfs2_alloc_context *ac,
+			       struct ocfs2_chain_list *cl)
+{
+	int status;
+	u32 bit_off, num_bits;
+	u64 bg_blkno;
+	struct buffer_head *bg_bh;
+	unsigned int alloc_rec = ocfs2_find_smallest_chain(cl);
+
+	status = ocfs2_claim_clusters(handle, ac,
+				      le16_to_cpu(cl->cl_cpg), &bit_off,
+				      &num_bits);
+	if (status < 0) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto bail;
+	}
+
+	/* setup the group */
+	bg_blkno = ocfs2_clusters_to_blocks(osb->sb, bit_off);
+	mlog(0, "new descriptor, record %u, at block %llu\n",
+	     alloc_rec, (unsigned long long)bg_blkno);
+
+	bg_bh = sb_getblk(osb->sb, bg_blkno);
+	if (!bg_bh) {
+		status = -EIO;
+		mlog_errno(status);
+		goto bail;
+	}
+	ocfs2_set_new_buffer_uptodate(INODE_CACHE(alloc_inode), bg_bh);
+
+	status = ocfs2_block_group_fill(handle, alloc_inode, bg_bh,
+					bg_blkno, num_bits, alloc_rec, cl);
+	if (status < 0) {
+		brelse(bg_bh);
+		mlog_errno(status);
+	}
+
+bail:
+	return status ? ERR_PTR(status) : bg_bh;
+}
+
+static int ocfs2_block_group_claim_bits(struct ocfs2_super *osb,
+					handle_t *handle,
+					struct ocfs2_alloc_context *ac,
+					unsigned int min_bits,
+					u32 *bit_off, u32 *num_bits)
+{
+	int status = 0;
+
+	while (min_bits) {
+		status = ocfs2_claim_clusters(handle, ac, min_bits,
+					      bit_off, num_bits);
+		if (status != -ENOSPC)
+			break;
+
+		min_bits >>= 1;
+	}
+
+	return status;
+}
+
+static int ocfs2_block_group_grow_discontig(handle_t *handle,
+					    struct inode *alloc_inode,
+					    struct buffer_head *bg_bh,
+					    struct ocfs2_alloc_context *ac,
+					    struct ocfs2_chain_list *cl,
+					    unsigned int min_bits)
+{
+	int status;
+	struct ocfs2_super *osb = OCFS2_SB(alloc_inode->i_sb);
+	struct ocfs2_group_desc *bg =
+		(struct ocfs2_group_desc *)bg_bh->b_data;
+	unsigned int needed = le16_to_cpu(cl->cl_cpg) -
+			 le16_to_cpu(bg->bg_bits) / le16_to_cpu(cl->cl_bpc);
+	u32 p_cpos, clusters;
+	u64 p_blkno;
+	struct ocfs2_extent_list *el = &bg->bg_list;
+
+	status = ocfs2_journal_access_gd(handle,
+					 INODE_CACHE(alloc_inode),
+					 bg_bh,
+					 OCFS2_JOURNAL_ACCESS_CREATE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	while ((needed > 0) && (le16_to_cpu(el->l_next_free_rec) <
+				le16_to_cpu(el->l_count))) {
+		if (min_bits > needed)
+			min_bits = needed;
+		status = ocfs2_block_group_claim_bits(osb, handle, ac,
+						      min_bits, &p_cpos,
+						      &clusters);
+		if (status < 0) {
+			if (status != -ENOSPC)
+				mlog_errno(status);
+			goto bail;
+		}
+		p_blkno = ocfs2_clusters_to_blocks(osb->sb, p_cpos);
+		ocfs2_bg_discontig_add_extent(osb, bg, cl, p_blkno,
+					      clusters);
+
+		min_bits = clusters;
+		needed = le16_to_cpu(cl->cl_cpg) -
+			 le16_to_cpu(bg->bg_bits) / le16_to_cpu(cl->cl_bpc);
+	}
+
+	if (needed > 0) {
+		/*
+		 * We have used up all the extent rec but can't fill up
+		 * the cpg. So bail out.
+		 */
+		status = -ENOSPC;
+		goto bail;
+	}
+
+	ocfs2_journal_dirty(handle, bg_bh);
+
+bail:
+	return status;
+}
+
+static void ocfs2_bg_alloc_cleanup(handle_t *handle,
+				   struct ocfs2_alloc_context *cluster_ac,
+				   struct inode *alloc_inode,
+				   struct buffer_head *bg_bh)
+{
+	int i, ret;
+	struct ocfs2_group_desc *bg;
+	struct ocfs2_extent_list *el;
+	struct ocfs2_extent_rec *rec;
+
+	if (!bg_bh)
+		return;
+
+	bg = (struct ocfs2_group_desc *)bg_bh->b_data;
+	el = &bg->bg_list;
+	for (i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) {
+		rec = &el->l_recs[i];
+		ret = ocfs2_free_clusters(handle, cluster_ac->ac_inode,
+					  cluster_ac->ac_bh,
+					  le64_to_cpu(rec->e_blkno),
+					  le32_to_cpu(rec->e_leaf_clusters));
+		if (ret)
+			mlog_errno(ret);
+		/* Try all the clusters to free */
+	}
+
+	ocfs2_remove_from_cache(INODE_CACHE(alloc_inode), bg_bh);
+	brelse(bg_bh);
+}
+
+static struct buffer_head *
+ocfs2_block_group_alloc_discontig(handle_t *handle,
+				  struct inode *alloc_inode,
+				  struct ocfs2_alloc_context *ac,
+				  struct ocfs2_chain_list *cl)
+{
+	int status;
+	u32 bit_off, num_bits;
+	u64 bg_blkno;
+	unsigned int min_bits = le16_to_cpu(cl->cl_cpg) >> 1;
+	struct buffer_head *bg_bh = NULL;
+	unsigned int alloc_rec = ocfs2_find_smallest_chain(cl);
+	struct ocfs2_super *osb = OCFS2_SB(alloc_inode->i_sb);
+
+	if (!ocfs2_supports_discontig_bg(osb)) {
+		status = -ENOSPC;
+		goto bail;
+	}
+
+	status = ocfs2_extend_trans(handle,
+				    ocfs2_calc_bg_discontig_credits(osb->sb));
+	if (status) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/*
+	 * We're going to be grabbing from multiple cluster groups.
+	 * We don't have enough credits to relink them all, and the
+	 * cluster groups will be staying in cache for the duration of
+	 * this operation.
+	 */
+	ac->ac_allow_chain_relink = 0;
+
+	/* Claim the first region */
+	status = ocfs2_block_group_claim_bits(osb, handle, ac, min_bits,
+					      &bit_off, &num_bits);
+	if (status < 0) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto bail;
+	}
+	min_bits = num_bits;
+
+	/* setup the group */
+	bg_blkno = ocfs2_clusters_to_blocks(osb->sb, bit_off);
+	mlog(0, "new descriptor, record %u, at block %llu\n",
+	     alloc_rec, (unsigned long long)bg_blkno);
+
+	bg_bh = sb_getblk(osb->sb, bg_blkno);
+	if (!bg_bh) {
+		status = -EIO;
+		mlog_errno(status);
+		goto bail;
+	}
+	ocfs2_set_new_buffer_uptodate(INODE_CACHE(alloc_inode), bg_bh);
+
+	status = ocfs2_block_group_fill(handle, alloc_inode, bg_bh,
+					bg_blkno, num_bits, alloc_rec, cl);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_block_group_grow_discontig(handle, alloc_inode,
+						  bg_bh, ac, cl, min_bits);
+	if (status)
+		mlog_errno(status);
+
+bail:
+	if (status)
+		ocfs2_bg_alloc_cleanup(handle, ac, alloc_inode, bg_bh);
+	return status ? ERR_PTR(status) : bg_bh;
+}
+
 /*
  * We expect the block group allocator to already be locked.
  */
@@ -416,9 +680,7 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
 	struct ocfs2_chain_list *cl;
 	struct ocfs2_alloc_context *ac = NULL;
 	handle_t *handle = NULL;
-	u32 bit_off, num_bits;
 	u16 alloc_rec;
-	u64 bg_blkno;
 	struct buffer_head *bg_bh = NULL;
 	struct ocfs2_group_desc *bg;
 
@@ -451,44 +713,20 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
 		     (unsigned long long)*last_alloc_group);
 		ac->ac_last_group = *last_alloc_group;
 	}
-	status = ocfs2_claim_clusters(osb,
-				      handle,
-				      ac,
-				      le16_to_cpu(cl->cl_cpg),
-				      &bit_off,
-				      &num_bits);
-	if (status < 0) {
+
+	bg_bh = ocfs2_block_group_alloc_contig(osb, handle, alloc_inode,
+					       ac, cl);
+	if (IS_ERR(bg_bh) && (PTR_ERR(bg_bh) == -ENOSPC))
+		bg_bh = ocfs2_block_group_alloc_discontig(handle,
+							  alloc_inode,
+							  ac, cl);
+	if (IS_ERR(bg_bh)) {
+		status = PTR_ERR(bg_bh);
+		bg_bh = NULL;
 		if (status != -ENOSPC)
 			mlog_errno(status);
 		goto bail;
 	}
-
-	alloc_rec = ocfs2_find_smallest_chain(cl);
-
-	/* setup the group */
-	bg_blkno = ocfs2_clusters_to_blocks(osb->sb, bit_off);
-	mlog(0, "new descriptor, record %u, at block %llu\n",
-	     alloc_rec, (unsigned long long)bg_blkno);
-
-	bg_bh = sb_getblk(osb->sb, bg_blkno);
-	if (!bg_bh) {
-		status = -EIO;
-		mlog_errno(status);
-		goto bail;
-	}
-	ocfs2_set_new_buffer_uptodate(INODE_CACHE(alloc_inode), bg_bh);
-
-	status = ocfs2_block_group_fill(handle,
-					alloc_inode,
-					bg_bh,
-					bg_blkno,
-					alloc_rec,
-					cl);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
-
 	bg = (struct ocfs2_group_desc *) bg_bh->b_data;
 
 	status = ocfs2_journal_access_di(handle, INODE_CACHE(alloc_inode),
@@ -498,10 +736,12 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
 		goto bail;
 	}
 
+	alloc_rec = le16_to_cpu(bg->bg_chain);
 	le32_add_cpu(&cl->cl_recs[alloc_rec].c_free,
 		     le16_to_cpu(bg->bg_free_bits_count));
-	le32_add_cpu(&cl->cl_recs[alloc_rec].c_total, le16_to_cpu(bg->bg_bits));
-	cl->cl_recs[alloc_rec].c_blkno  = cpu_to_le64(bg_blkno);
+	le32_add_cpu(&cl->cl_recs[alloc_rec].c_total,
+		     le16_to_cpu(bg->bg_bits));
+	cl->cl_recs[alloc_rec].c_blkno = bg->bg_blkno;
 	if (le16_to_cpu(cl->cl_next_free_rec) < le16_to_cpu(cl->cl_count))
 		le16_add_cpu(&cl->cl_next_free_rec, 1);
 
@@ -510,11 +750,7 @@ static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
 	le32_add_cpu(&fe->id1.bitmap1.i_total, le16_to_cpu(bg->bg_bits));
 	le32_add_cpu(&fe->i_clusters, le16_to_cpu(cl->cl_cpg));
 
-	status = ocfs2_journal_dirty(handle, bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	ocfs2_journal_dirty(handle, bh);
 
 	spin_lock(&OCFS2_I(alloc_inode)->ip_lock);
 	OCFS2_I(alloc_inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
@@ -637,12 +873,113 @@ bail:
 	return status;
 }
 
+static void ocfs2_init_inode_steal_slot(struct ocfs2_super *osb)
+{
+	spin_lock(&osb->osb_lock);
+	osb->s_inode_steal_slot = OCFS2_INVALID_SLOT;
+	spin_unlock(&osb->osb_lock);
+	atomic_set(&osb->s_num_inodes_stolen, 0);
+}
+
+static void ocfs2_init_meta_steal_slot(struct ocfs2_super *osb)
+{
+	spin_lock(&osb->osb_lock);
+	osb->s_meta_steal_slot = OCFS2_INVALID_SLOT;
+	spin_unlock(&osb->osb_lock);
+	atomic_set(&osb->s_num_meta_stolen, 0);
+}
+
+void ocfs2_init_steal_slots(struct ocfs2_super *osb)
+{
+	ocfs2_init_inode_steal_slot(osb);
+	ocfs2_init_meta_steal_slot(osb);
+}
+
+static void __ocfs2_set_steal_slot(struct ocfs2_super *osb, int slot, int type)
+{
+	spin_lock(&osb->osb_lock);
+	if (type == INODE_ALLOC_SYSTEM_INODE)
+		osb->s_inode_steal_slot = slot;
+	else if (type == EXTENT_ALLOC_SYSTEM_INODE)
+		osb->s_meta_steal_slot = slot;
+	spin_unlock(&osb->osb_lock);
+}
+
+static int __ocfs2_get_steal_slot(struct ocfs2_super *osb, int type)
+{
+	int slot = OCFS2_INVALID_SLOT;
+
+	spin_lock(&osb->osb_lock);
+	if (type == INODE_ALLOC_SYSTEM_INODE)
+		slot = osb->s_inode_steal_slot;
+	else if (type == EXTENT_ALLOC_SYSTEM_INODE)
+		slot = osb->s_meta_steal_slot;
+	spin_unlock(&osb->osb_lock);
+
+	return slot;
+}
+
+static int ocfs2_get_inode_steal_slot(struct ocfs2_super *osb)
+{
+	return __ocfs2_get_steal_slot(osb, INODE_ALLOC_SYSTEM_INODE);
+}
+
+static int ocfs2_get_meta_steal_slot(struct ocfs2_super *osb)
+{
+	return __ocfs2_get_steal_slot(osb, EXTENT_ALLOC_SYSTEM_INODE);
+}
+
+static int ocfs2_steal_resource(struct ocfs2_super *osb,
+				struct ocfs2_alloc_context *ac,
+				int type)
+{
+	int i, status = -ENOSPC;
+	int slot = __ocfs2_get_steal_slot(osb, type);
+
+	/* Start to steal resource from the first slot after ours. */
+	if (slot == OCFS2_INVALID_SLOT)
+		slot = osb->slot_num + 1;
+
+	for (i = 0; i < osb->max_slots; i++, slot++) {
+		if (slot == osb->max_slots)
+			slot = 0;
+
+		if (slot == osb->slot_num)
+			continue;
+
+		status = ocfs2_reserve_suballoc_bits(osb, ac,
+						     type,
+						     (u32)slot, NULL,
+						     NOT_ALLOC_NEW_GROUP);
+		if (status >= 0) {
+			__ocfs2_set_steal_slot(osb, slot, type);
+			break;
+		}
+
+		ocfs2_free_ac_resource(ac);
+	}
+
+	return status;
+}
+
+static int ocfs2_steal_inode(struct ocfs2_super *osb,
+			     struct ocfs2_alloc_context *ac)
+{
+	return ocfs2_steal_resource(osb, ac, INODE_ALLOC_SYSTEM_INODE);
+}
+
+static int ocfs2_steal_meta(struct ocfs2_super *osb,
+			    struct ocfs2_alloc_context *ac)
+{
+	return ocfs2_steal_resource(osb, ac, EXTENT_ALLOC_SYSTEM_INODE);
+}
+
 int ocfs2_reserve_new_metadata_blocks(struct ocfs2_super *osb,
 				      int blocks,
 				      struct ocfs2_alloc_context **ac)
 {
 	int status;
-	u32 slot;
+	int slot = ocfs2_get_meta_steal_slot(osb);
 
 	*ac = kzalloc(sizeof(struct ocfs2_alloc_context), GFP_KERNEL);
 	if (!(*ac)) {
@@ -653,12 +990,34 @@ int ocfs2_reserve_new_metadata_blocks(struct ocfs2_super *osb,
 
 	(*ac)->ac_bits_wanted = blocks;
 	(*ac)->ac_which = OCFS2_AC_USE_META;
-	slot = osb->slot_num;
 	(*ac)->ac_group_search = ocfs2_block_group_search;
 
+	if (slot != OCFS2_INVALID_SLOT &&
+		atomic_read(&osb->s_num_meta_stolen) < OCFS2_MAX_TO_STEAL)
+		goto extent_steal;
+
+	atomic_set(&osb->s_num_meta_stolen, 0);
 	status = ocfs2_reserve_suballoc_bits(osb, (*ac),
 					     EXTENT_ALLOC_SYSTEM_INODE,
-					     slot, NULL, ALLOC_NEW_GROUP);
+					     (u32)osb->slot_num, NULL,
+					     ALLOC_GROUPS_FROM_GLOBAL|ALLOC_NEW_GROUP);
+
+
+	if (status >= 0) {
+		status = 0;
+		if (slot != OCFS2_INVALID_SLOT)
+			ocfs2_init_meta_steal_slot(osb);
+		goto bail;
+	} else if (status < 0 && status != -ENOSPC) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	ocfs2_free_ac_resource(*ac);
+
+extent_steal:
+	status = ocfs2_steal_meta(osb, *ac);
+	atomic_inc(&osb->s_num_meta_stolen);
 	if (status < 0) {
 		if (status != -ENOSPC)
 			mlog_errno(status);
@@ -685,43 +1044,11 @@ int ocfs2_reserve_new_metadata(struct ocfs2_super *osb,
 					ac);
 }
 
-static int ocfs2_steal_inode_from_other_nodes(struct ocfs2_super *osb,
-					      struct ocfs2_alloc_context *ac)
-{
-	int i, status = -ENOSPC;
-	s16 slot = ocfs2_get_inode_steal_slot(osb);
-
-	/* Start to steal inodes from the first slot after ours. */
-	if (slot == OCFS2_INVALID_SLOT)
-		slot = osb->slot_num + 1;
-
-	for (i = 0; i < osb->max_slots; i++, slot++) {
-		if (slot == osb->max_slots)
-			slot = 0;
-
-		if (slot == osb->slot_num)
-			continue;
-
-		status = ocfs2_reserve_suballoc_bits(osb, ac,
-						     INODE_ALLOC_SYSTEM_INODE,
-						     slot, NULL,
-						     NOT_ALLOC_NEW_GROUP);
-		if (status >= 0) {
-			ocfs2_set_inode_steal_slot(osb, slot);
-			break;
-		}
-
-		ocfs2_free_ac_resource(ac);
-	}
-
-	return status;
-}
-
 int ocfs2_reserve_new_inode(struct ocfs2_super *osb,
 			    struct ocfs2_alloc_context **ac)
 {
 	int status;
-	s16 slot = ocfs2_get_inode_steal_slot(osb);
+	int slot = ocfs2_get_inode_steal_slot(osb);
 	u64 alloc_group;
 
 	*ac = kzalloc(sizeof(struct ocfs2_alloc_context), GFP_KERNEL);
@@ -754,14 +1081,14 @@ int ocfs2_reserve_new_inode(struct ocfs2_super *osb,
 	 * need to check our slots to see whether there is some space for us.
 	 */
 	if (slot != OCFS2_INVALID_SLOT &&
-	    atomic_read(&osb->s_num_inodes_stolen) < OCFS2_MAX_INODES_TO_STEAL)
+	    atomic_read(&osb->s_num_inodes_stolen) < OCFS2_MAX_TO_STEAL)
 		goto inode_steal;
 
 	atomic_set(&osb->s_num_inodes_stolen, 0);
 	alloc_group = osb->osb_inode_alloc_group;
 	status = ocfs2_reserve_suballoc_bits(osb, *ac,
 					     INODE_ALLOC_SYSTEM_INODE,
-					     osb->slot_num,
+					     (u32)osb->slot_num,
 					     &alloc_group,
 					     ALLOC_NEW_GROUP |
 					     ALLOC_GROUPS_FROM_GLOBAL);
@@ -789,7 +1116,7 @@ int ocfs2_reserve_new_inode(struct ocfs2_super *osb,
 	ocfs2_free_ac_resource(*ac);
 
 inode_steal:
-	status = ocfs2_steal_inode_from_other_nodes(osb, *ac);
+	status = ocfs2_steal_inode(osb, *ac);
 	atomic_inc(&osb->s_num_inodes_stolen);
 	if (status < 0) {
 		if (status != -ENOSPC)
@@ -859,11 +1186,7 @@ static int ocfs2_reserve_clusters_with_limit(struct ocfs2_super *osb,
 		status = ocfs2_reserve_local_alloc_bits(osb,
 							bits_wanted,
 							*ac);
-		if (status == -EFBIG) {
-			/* The local alloc window is outside ac_max_block.
-			 * use the main bitmap. */
-			status = -ENOSPC;
-		} else if ((status < 0) && (status != -ENOSPC)) {
+		if ((status < 0) && (status != -ENOSPC)) {
 			mlog_errno(status);
 			goto bail;
 		}
@@ -946,8 +1269,7 @@ static int ocfs2_block_group_find_clear_bits(struct ocfs2_super *osb,
 					     struct buffer_head *bg_bh,
 					     unsigned int bits_wanted,
 					     unsigned int total_bits,
-					     u16 *bit_off,
-					     u16 *bits_found)
+					     struct ocfs2_suballoc_result *res)
 {
 	void *bitmap;
 	u16 best_offset, best_size;
@@ -991,14 +1313,9 @@ static int ocfs2_block_group_find_clear_bits(struct ocfs2_super *osb,
 		}
 	}
 
-	/* XXX: I think the first clause is equivalent to the second
-	 * 	- jlbec */
-	if (found == bits_wanted) {
-		*bit_off = start - found;
-		*bits_found = found;
-	} else if (best_size) {
-		*bit_off = best_offset;
-		*bits_found = best_size;
+	if (best_size) {
+		res->sr_bit_offset = best_offset;
+		res->sr_bits = best_size;
 	} else {
 		status = -ENOSPC;
 		/* No error log here -- see the comment above
@@ -1042,16 +1359,10 @@ static inline int ocfs2_block_group_set_bits(handle_t *handle,
 	}
 
 	le16_add_cpu(&bg->bg_free_bits_count, -num_bits);
-
 	while(num_bits--)
 		ocfs2_set_bit(bit_off++, bitmap);
 
-	status = ocfs2_journal_dirty(handle,
-				     group_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	ocfs2_journal_dirty(handle, group_bh);
 
 bail:
 	mlog_exit(status);
@@ -1115,12 +1426,7 @@ static int ocfs2_relink_block_group(handle_t *handle,
 	}
 
 	prev_bg->bg_next_group = bg->bg_next_group;
-
-	status = ocfs2_journal_dirty(handle, prev_bg_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out_rollback;
-	}
+	ocfs2_journal_dirty(handle, prev_bg_bh);
 
 	status = ocfs2_journal_access_gd(handle, INODE_CACHE(alloc_inode),
 					 bg_bh, OCFS2_JOURNAL_ACCESS_WRITE);
@@ -1130,12 +1436,7 @@ static int ocfs2_relink_block_group(handle_t *handle,
 	}
 
 	bg->bg_next_group = fe->id2.i_chain.cl_recs[chain].c_blkno;
-
-	status = ocfs2_journal_dirty(handle, bg_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out_rollback;
-	}
+	ocfs2_journal_dirty(handle, bg_bh);
 
 	status = ocfs2_journal_access_di(handle, INODE_CACHE(alloc_inode),
 					 fe_bh, OCFS2_JOURNAL_ACCESS_WRITE);
@@ -1145,14 +1446,8 @@ static int ocfs2_relink_block_group(handle_t *handle,
 	}
 
 	fe->id2.i_chain.cl_recs[chain].c_blkno = bg->bg_blkno;
+	ocfs2_journal_dirty(handle, fe_bh);
 
-	status = ocfs2_journal_dirty(handle, fe_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto out_rollback;
-	}
-
-	status = 0;
 out_rollback:
 	if (status < 0) {
 		fe->id2.i_chain.cl_recs[chain].c_blkno = cpu_to_le64(fe_ptr);
@@ -1176,14 +1471,13 @@ static int ocfs2_cluster_group_search(struct inode *inode,
 				      struct buffer_head *group_bh,
 				      u32 bits_wanted, u32 min_bits,
 				      u64 max_block,
-				      u16 *bit_off, u16 *bits_found)
+				      struct ocfs2_suballoc_result *res)
 {
 	int search = -ENOSPC;
 	int ret;
 	u64 blkoff;
 	struct ocfs2_group_desc *gd = (struct ocfs2_group_desc *) group_bh->b_data;
 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
-	u16 tmp_off, tmp_found;
 	unsigned int max_bits, gd_cluster_off;
 
 	BUG_ON(!ocfs2_is_cluster_bitmap(inode));
@@ -1210,15 +1504,15 @@ static int ocfs2_cluster_group_search(struct inode *inode,
 
 		ret = ocfs2_block_group_find_clear_bits(OCFS2_SB(inode->i_sb),
 							group_bh, bits_wanted,
-							max_bits,
-							&tmp_off, &tmp_found);
+							max_bits, res);
 		if (ret)
 			return ret;
 
 		if (max_block) {
 			blkoff = ocfs2_clusters_to_blocks(inode->i_sb,
 							  gd_cluster_off +
-							  tmp_off + tmp_found);
+							  res->sr_bit_offset +
+							  res->sr_bits);
 			mlog(0, "Checking %llu against %llu\n",
 			     (unsigned long long)blkoff,
 			     (unsigned long long)max_block);
@@ -1230,16 +1524,14 @@ static int ocfs2_cluster_group_search(struct inode *inode,
 		 * return success, but we still want to return
 		 * -ENOSPC unless it found the minimum number
 		 * of bits. */
-		if (min_bits <= tmp_found) {
-			*bit_off = tmp_off;
-			*bits_found = tmp_found;
+		if (min_bits <= res->sr_bits)
 			search = 0; /* success */
-		} else if (tmp_found) {
+		else if (res->sr_bits) {
 			/*
 			 * Don't show bits which we'll be returning
 			 * for allocation to the local alloc bitmap.
 			 */
-			ocfs2_local_alloc_seen_free_bits(osb, tmp_found);
+			ocfs2_local_alloc_seen_free_bits(osb, res->sr_bits);
 		}
 	}
 
@@ -1250,7 +1542,7 @@ static int ocfs2_block_group_search(struct inode *inode,
 				    struct buffer_head *group_bh,
 				    u32 bits_wanted, u32 min_bits,
 				    u64 max_block,
-				    u16 *bit_off, u16 *bits_found)
+				    struct ocfs2_suballoc_result *res)
 {
 	int ret = -ENOSPC;
 	u64 blkoff;
@@ -1263,10 +1555,10 @@ static int ocfs2_block_group_search(struct inode *inode,
 		ret = ocfs2_block_group_find_clear_bits(OCFS2_SB(inode->i_sb),
 							group_bh, bits_wanted,
 							le16_to_cpu(bg->bg_bits),
-							bit_off, bits_found);
+							res);
 		if (!ret && max_block) {
-			blkoff = le64_to_cpu(bg->bg_blkno) + *bit_off +
-				*bits_found;
+			blkoff = le64_to_cpu(bg->bg_blkno) +
+				res->sr_bit_offset + res->sr_bits;
 			mlog(0, "Checking %llu against %llu\n",
 			     (unsigned long long)blkoff,
 			     (unsigned long long)max_block);
@@ -1299,33 +1591,76 @@ static int ocfs2_alloc_dinode_update_counts(struct inode *inode,
 	tmp_used = le32_to_cpu(di->id1.bitmap1.i_used);
 	di->id1.bitmap1.i_used = cpu_to_le32(num_bits + tmp_used);
 	le32_add_cpu(&cl->cl_recs[chain].c_free, -num_bits);
-
-	ret = ocfs2_journal_dirty(handle, di_bh);
-	if (ret < 0)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, di_bh);
 
 out:
 	return ret;
 }
 
+static int ocfs2_bg_discontig_fix_by_rec(struct ocfs2_suballoc_result *res,
+					 struct ocfs2_extent_rec *rec,
+					 struct ocfs2_chain_list *cl)
+{
+	unsigned int bpc = le16_to_cpu(cl->cl_bpc);
+	unsigned int bitoff = le32_to_cpu(rec->e_cpos) * bpc;
+	unsigned int bitcount = le32_to_cpu(rec->e_leaf_clusters) * bpc;
+
+	if (res->sr_bit_offset < bitoff)
+		return 0;
+	if (res->sr_bit_offset >= (bitoff + bitcount))
+		return 0;
+	res->sr_blkno = le64_to_cpu(rec->e_blkno) +
+		(res->sr_bit_offset - bitoff);
+	if ((res->sr_bit_offset + res->sr_bits) > (bitoff + bitcount))
+		res->sr_bits = (bitoff + bitcount) - res->sr_bit_offset;
+	return 1;
+}
+
+static void ocfs2_bg_discontig_fix_result(struct ocfs2_alloc_context *ac,
+					  struct ocfs2_group_desc *bg,
+					  struct ocfs2_suballoc_result *res)
+{
+	int i;
+	u64 bg_blkno = res->sr_bg_blkno;  /* Save off */
+	struct ocfs2_extent_rec *rec;
+	struct ocfs2_dinode *di = (struct ocfs2_dinode *)ac->ac_bh->b_data;
+	struct ocfs2_chain_list *cl = &di->id2.i_chain;
+
+	if (ocfs2_is_cluster_bitmap(ac->ac_inode)) {
+		res->sr_blkno = 0;
+		return;
+	}
+
+	res->sr_blkno = res->sr_bg_blkno + res->sr_bit_offset;
+	res->sr_bg_blkno = 0;  /* Clear it for contig block groups */
+	if (!ocfs2_supports_discontig_bg(OCFS2_SB(ac->ac_inode->i_sb)) ||
+	    !bg->bg_list.l_next_free_rec)
+		return;
+
+	for (i = 0; i < le16_to_cpu(bg->bg_list.l_next_free_rec); i++) {
+		rec = &bg->bg_list.l_recs[i];
+		if (ocfs2_bg_discontig_fix_by_rec(res, rec, cl)) {
+			res->sr_bg_blkno = bg_blkno;  /* Restore */
+			break;
+		}
+	}
+}
+
 static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac,
 				  handle_t *handle,
 				  u32 bits_wanted,
 				  u32 min_bits,
-				  u16 *bit_off,
-				  unsigned int *num_bits,
-				  u64 gd_blkno,
+				  struct ocfs2_suballoc_result *res,
 				  u16 *bits_left)
 {
 	int ret;
-	u16 found;
 	struct buffer_head *group_bh = NULL;
 	struct ocfs2_group_desc *gd;
 	struct ocfs2_dinode *di = (struct ocfs2_dinode *)ac->ac_bh->b_data;
 	struct inode *alloc_inode = ac->ac_inode;
 
-	ret = ocfs2_read_group_descriptor(alloc_inode, di, gd_blkno,
-					  &group_bh);
+	ret = ocfs2_read_group_descriptor(alloc_inode, di,
+					  res->sr_bg_blkno, &group_bh);
 	if (ret < 0) {
 		mlog_errno(ret);
 		return ret;
@@ -1333,17 +1668,18 @@ static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac,
 
 	gd = (struct ocfs2_group_desc *) group_bh->b_data;
 	ret = ac->ac_group_search(alloc_inode, group_bh, bits_wanted, min_bits,
-				  ac->ac_max_block, bit_off, &found);
+				  ac->ac_max_block, res);
 	if (ret < 0) {
 		if (ret != -ENOSPC)
 			mlog_errno(ret);
 		goto out;
 	}
 
-	*num_bits = found;
+	if (!ret)
+		ocfs2_bg_discontig_fix_result(ac, gd, res);
 
 	ret = ocfs2_alloc_dinode_update_counts(alloc_inode, handle, ac->ac_bh,
-					       *num_bits,
+					       res->sr_bits,
 					       le16_to_cpu(gd->bg_chain));
 	if (ret < 0) {
 		mlog_errno(ret);
@@ -1351,7 +1687,7 @@ static int ocfs2_search_one_group(struct ocfs2_alloc_context *ac,
 	}
 
 	ret = ocfs2_block_group_set_bits(handle, alloc_inode, gd, group_bh,
-					 *bit_off, *num_bits);
+					 res->sr_bit_offset, res->sr_bits);
 	if (ret < 0)
 		mlog_errno(ret);
 
@@ -1367,13 +1703,11 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
 			      handle_t *handle,
 			      u32 bits_wanted,
 			      u32 min_bits,
-			      u16 *bit_off,
-			      unsigned int *num_bits,
-			      u64 *bg_blkno,
+			      struct ocfs2_suballoc_result *res,
 			      u16 *bits_left)
 {
 	int status;
-	u16 chain, tmp_bits;
+	u16 chain;
 	u32 tmp_used;
 	u64 next_group;
 	struct inode *alloc_inode = ac->ac_inode;
@@ -1402,8 +1736,8 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
 	 * the 1st group with any empty bits. */
 	while ((status = ac->ac_group_search(alloc_inode, group_bh,
 					     bits_wanted, min_bits,
-					     ac->ac_max_block, bit_off,
-					     &tmp_bits)) == -ENOSPC) {
+					     ac->ac_max_block,
+					     res)) == -ENOSPC) {
 		if (!bg->bg_next_group)
 			break;
 
@@ -1428,11 +1762,14 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
 	}
 
 	mlog(0, "alloc succeeds: we give %u bits from block group %llu\n",
-	     tmp_bits, (unsigned long long)le64_to_cpu(bg->bg_blkno));
+	     res->sr_bits, (unsigned long long)le64_to_cpu(bg->bg_blkno));
 
-	*num_bits = tmp_bits;
+	res->sr_bg_blkno = le64_to_cpu(bg->bg_blkno);
+
+	BUG_ON(res->sr_bits == 0);
+	if (!status)
+		ocfs2_bg_discontig_fix_result(ac, bg, res);
 
-	BUG_ON(*num_bits == 0);
 
 	/*
 	 * Keep track of previous block descriptor read. When
@@ -1449,7 +1786,7 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
 	 */
 	if (ac->ac_allow_chain_relink &&
 	    (prev_group_bh) &&
-	    (ocfs2_block_group_reasonably_empty(bg, *num_bits))) {
+	    (ocfs2_block_group_reasonably_empty(bg, res->sr_bits))) {
 		status = ocfs2_relink_block_group(handle, alloc_inode,
 						  ac->ac_bh, group_bh,
 						  prev_group_bh, chain);
@@ -1471,31 +1808,24 @@ static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
 	}
 
 	tmp_used = le32_to_cpu(fe->id1.bitmap1.i_used);
-	fe->id1.bitmap1.i_used = cpu_to_le32(*num_bits + tmp_used);
-	le32_add_cpu(&cl->cl_recs[chain].c_free, -(*num_bits));
-
-	status = ocfs2_journal_dirty(handle,
-				     ac->ac_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	fe->id1.bitmap1.i_used = cpu_to_le32(res->sr_bits + tmp_used);
+	le32_add_cpu(&cl->cl_recs[chain].c_free, -res->sr_bits);
+	ocfs2_journal_dirty(handle, ac->ac_bh);
 
 	status = ocfs2_block_group_set_bits(handle,
 					    alloc_inode,
 					    bg,
 					    group_bh,
-					    *bit_off,
-					    *num_bits);
+					    res->sr_bit_offset,
+					    res->sr_bits);
 	if (status < 0) {
 		mlog_errno(status);
 		goto bail;
 	}
 
-	mlog(0, "Allocated %u bits from suballocator %llu\n", *num_bits,
+	mlog(0, "Allocated %u bits from suballocator %llu\n", res->sr_bits,
 	     (unsigned long long)le64_to_cpu(fe->i_blkno));
 
-	*bg_blkno = le64_to_cpu(bg->bg_blkno);
 	*bits_left = le16_to_cpu(bg->bg_free_bits_count);
 bail:
 	brelse(group_bh);
@@ -1506,19 +1836,15 @@ bail:
 }
 
 /* will give out up to bits_wanted contiguous bits. */
-static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
-				     struct ocfs2_alloc_context *ac,
+static int ocfs2_claim_suballoc_bits(struct ocfs2_alloc_context *ac,
 				     handle_t *handle,
 				     u32 bits_wanted,
 				     u32 min_bits,
-				     u16 *bit_off,
-				     unsigned int *num_bits,
-				     u64 *bg_blkno)
+				     struct ocfs2_suballoc_result *res)
 {
 	int status;
 	u16 victim, i;
 	u16 bits_left = 0;
-	u64 hint_blkno = ac->ac_last_group;
 	struct ocfs2_chain_list *cl;
 	struct ocfs2_dinode *fe;
 
@@ -1536,7 +1862,8 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
 
 	if (le32_to_cpu(fe->id1.bitmap1.i_used) >=
 	    le32_to_cpu(fe->id1.bitmap1.i_total)) {
-		ocfs2_error(osb->sb, "Chain allocator dinode %llu has %u used "
+		ocfs2_error(ac->ac_inode->i_sb,
+			    "Chain allocator dinode %llu has %u used "
 			    "bits but only %u total.",
 			    (unsigned long long)le64_to_cpu(fe->i_blkno),
 			    le32_to_cpu(fe->id1.bitmap1.i_used),
@@ -1545,22 +1872,16 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
 		goto bail;
 	}
 
-	if (hint_blkno) {
+	res->sr_bg_blkno = ac->ac_last_group;
+	if (res->sr_bg_blkno) {
 		/* Attempt to short-circuit the usual search mechanism
 		 * by jumping straight to the most recently used
 		 * allocation group. This helps us mantain some
 		 * contiguousness across allocations. */
 		status = ocfs2_search_one_group(ac, handle, bits_wanted,
-						min_bits, bit_off, num_bits,
-						hint_blkno, &bits_left);
-		if (!status) {
-			/* Be careful to update *bg_blkno here as the
-			 * caller is expecting it to be filled in, and
-			 * ocfs2_search_one_group() won't do that for
-			 * us. */
-			*bg_blkno = hint_blkno;
+						min_bits, res, &bits_left);
+		if (!status)
 			goto set_hint;
-		}
 		if (status < 0 && status != -ENOSPC) {
 			mlog_errno(status);
 			goto bail;
@@ -1573,8 +1894,8 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
 	ac->ac_chain = victim;
 	ac->ac_allow_chain_relink = 1;
 
-	status = ocfs2_search_chain(ac, handle, bits_wanted, min_bits, bit_off,
-				    num_bits, bg_blkno, &bits_left);
+	status = ocfs2_search_chain(ac, handle, bits_wanted, min_bits,
+				    res, &bits_left);
 	if (!status)
 		goto set_hint;
 	if (status < 0 && status != -ENOSPC) {
@@ -1598,8 +1919,7 @@ static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
 
 		ac->ac_chain = i;
 		status = ocfs2_search_chain(ac, handle, bits_wanted, min_bits,
-					    bit_off, num_bits, bg_blkno,
-					    &bits_left);
+					    res, &bits_left);
 		if (!status)
 			break;
 		if (status < 0 && status != -ENOSPC) {
@@ -1616,7 +1936,7 @@ set_hint:
 		if (bits_left < min_bits)
 			ac->ac_last_group = 0;
 		else
-			ac->ac_last_group = *bg_blkno;
+			ac->ac_last_group = res->sr_bg_blkno;
 	}
 
 bail:
@@ -1624,37 +1944,37 @@ bail:
 	return status;
 }
 
-int ocfs2_claim_metadata(struct ocfs2_super *osb,
-			 handle_t *handle,
+int ocfs2_claim_metadata(handle_t *handle,
 			 struct ocfs2_alloc_context *ac,
 			 u32 bits_wanted,
+			 u64 *suballoc_loc,
 			 u16 *suballoc_bit_start,
 			 unsigned int *num_bits,
 			 u64 *blkno_start)
 {
 	int status;
-	u64 bg_blkno;
+	struct ocfs2_suballoc_result res = { .sr_blkno = 0, };
 
 	BUG_ON(!ac);
 	BUG_ON(ac->ac_bits_wanted < (ac->ac_bits_given + bits_wanted));
 	BUG_ON(ac->ac_which != OCFS2_AC_USE_META);
 
-	status = ocfs2_claim_suballoc_bits(osb,
-					   ac,
+	status = ocfs2_claim_suballoc_bits(ac,
 					   handle,
 					   bits_wanted,
 					   1,
-					   suballoc_bit_start,
-					   num_bits,
-					   &bg_blkno);
+					   &res);
 	if (status < 0) {
 		mlog_errno(status);
 		goto bail;
 	}
-	atomic_inc(&osb->alloc_stats.bg_allocs);
+	atomic_inc(&OCFS2_SB(ac->ac_inode->i_sb)->alloc_stats.bg_allocs);
 
-	*blkno_start = bg_blkno + (u64) *suballoc_bit_start;
-	ac->ac_bits_given += (*num_bits);
+	*suballoc_loc = res.sr_bg_blkno;
+	*suballoc_bit_start = res.sr_bit_offset;
+	*blkno_start = res.sr_blkno;
+	ac->ac_bits_given += res.sr_bits;
+	*num_bits = res.sr_bits;
 	status = 0;
 bail:
 	mlog_exit(status);
@@ -1662,10 +1982,10 @@ bail:
 }
 
 static void ocfs2_init_inode_ac_group(struct inode *dir,
-				      struct buffer_head *parent_fe_bh,
+				      struct buffer_head *parent_di_bh,
 				      struct ocfs2_alloc_context *ac)
 {
-	struct ocfs2_dinode *fe = (struct ocfs2_dinode *)parent_fe_bh->b_data;
+	struct ocfs2_dinode *di = (struct ocfs2_dinode *)parent_di_bh->b_data;
 	/*
 	 * Try to allocate inodes from some specific group.
 	 *
@@ -1679,10 +1999,14 @@ static void ocfs2_init_inode_ac_group(struct inode *dir,
 	if (OCFS2_I(dir)->ip_last_used_group &&
 	    OCFS2_I(dir)->ip_last_used_slot == ac->ac_alloc_slot)
 		ac->ac_last_group = OCFS2_I(dir)->ip_last_used_group;
-	else if (le16_to_cpu(fe->i_suballoc_slot) == ac->ac_alloc_slot)
-		ac->ac_last_group = ocfs2_which_suballoc_group(
-					le64_to_cpu(fe->i_blkno),
-					le16_to_cpu(fe->i_suballoc_bit));
+	else if (le16_to_cpu(di->i_suballoc_slot) == ac->ac_alloc_slot) {
+		if (di->i_suballoc_loc)
+			ac->ac_last_group = le64_to_cpu(di->i_suballoc_loc);
+		else
+			ac->ac_last_group = ocfs2_which_suballoc_group(
+					le64_to_cpu(di->i_blkno),
+					le16_to_cpu(di->i_suballoc_bit));
+	}
 }
 
 static inline void ocfs2_save_inode_ac_group(struct inode *dir,
@@ -1692,17 +2016,16 @@ static inline void ocfs2_save_inode_ac_group(struct inode *dir,
 	OCFS2_I(dir)->ip_last_used_slot = ac->ac_alloc_slot;
 }
 
-int ocfs2_claim_new_inode(struct ocfs2_super *osb,
-			  handle_t *handle,
+int ocfs2_claim_new_inode(handle_t *handle,
 			  struct inode *dir,
 			  struct buffer_head *parent_fe_bh,
 			  struct ocfs2_alloc_context *ac,
+			  u64 *suballoc_loc,
 			  u16 *suballoc_bit,
 			  u64 *fe_blkno)
 {
 	int status;
-	unsigned int num_bits;
-	u64 bg_blkno;
+	struct ocfs2_suballoc_result res;
 
 	mlog_entry_void();
 
@@ -1713,23 +2036,22 @@ int ocfs2_claim_new_inode(struct ocfs2_super *osb,
 
 	ocfs2_init_inode_ac_group(dir, parent_fe_bh, ac);
 
-	status = ocfs2_claim_suballoc_bits(osb,
-					   ac,
+	status = ocfs2_claim_suballoc_bits(ac,
 					   handle,
 					   1,
 					   1,
-					   suballoc_bit,
-					   &num_bits,
-					   &bg_blkno);
+					   &res);
 	if (status < 0) {
 		mlog_errno(status);
 		goto bail;
 	}
-	atomic_inc(&osb->alloc_stats.bg_allocs);
+	atomic_inc(&OCFS2_SB(ac->ac_inode->i_sb)->alloc_stats.bg_allocs);
 
-	BUG_ON(num_bits != 1);
+	BUG_ON(res.sr_bits != 1);
 
-	*fe_blkno = bg_blkno + (u64) (*suballoc_bit);
+	*suballoc_loc = res.sr_bg_blkno;
+	*suballoc_bit = res.sr_bit_offset;
+	*fe_blkno = res.sr_blkno;
 	ac->ac_bits_given++;
 	ocfs2_save_inode_ac_group(dir, ac);
 	status = 0;
@@ -1799,8 +2121,7 @@ static inline void ocfs2_block_to_cluster_group(struct inode *inode,
  * contig. allocation, set to '1' to indicate we can deal with extents
  * of any size.
  */
-int __ocfs2_claim_clusters(struct ocfs2_super *osb,
-			   handle_t *handle,
+int __ocfs2_claim_clusters(handle_t *handle,
 			   struct ocfs2_alloc_context *ac,
 			   u32 min_clusters,
 			   u32 max_clusters,
@@ -1809,8 +2130,8 @@ int __ocfs2_claim_clusters(struct ocfs2_super *osb,
 {
 	int status;
 	unsigned int bits_wanted = max_clusters;
-	u64 bg_blkno = 0;
-	u16 bg_bit_off;
+	struct ocfs2_suballoc_result res = { .sr_blkno = 0, };
+	struct ocfs2_super *osb = OCFS2_SB(ac->ac_inode->i_sb);
 
 	mlog_entry_void();
 
@@ -1820,6 +2141,8 @@ int __ocfs2_claim_clusters(struct ocfs2_super *osb,
 	       && ac->ac_which != OCFS2_AC_USE_MAIN);
 
 	if (ac->ac_which == OCFS2_AC_USE_LOCAL) {
+		WARN_ON(min_clusters > 1);
+
 		status = ocfs2_claim_local_alloc_bits(osb,
 						      handle,
 						      ac,
@@ -1842,20 +2165,19 @@ int __ocfs2_claim_clusters(struct ocfs2_super *osb,
 		if (bits_wanted > (osb->bitmap_cpg - 1))
 			bits_wanted = osb->bitmap_cpg - 1;
 
-		status = ocfs2_claim_suballoc_bits(osb,
-						   ac,
+		status = ocfs2_claim_suballoc_bits(ac,
 						   handle,
 						   bits_wanted,
 						   min_clusters,
-						   &bg_bit_off,
-						   num_clusters,
-						   &bg_blkno);
+						   &res);
 		if (!status) {
+			BUG_ON(res.sr_blkno); /* cluster alloc can't set */
 			*cluster_start =
 				ocfs2_desc_bitmap_to_cluster_off(ac->ac_inode,
-								 bg_blkno,
-								 bg_bit_off);
+								 res.sr_bg_blkno,
+								 res.sr_bit_offset);
 			atomic_inc(&osb->alloc_stats.bitmap_data);
+			*num_clusters = res.sr_bits;
 		}
 	}
 	if (status < 0) {
@@ -1871,8 +2193,7 @@ bail:
 	return status;
 }
 
-int ocfs2_claim_clusters(struct ocfs2_super *osb,
-			 handle_t *handle,
+int ocfs2_claim_clusters(handle_t *handle,
 			 struct ocfs2_alloc_context *ac,
 			 u32 min_clusters,
 			 u32 *cluster_start,
@@ -1880,22 +2201,22 @@ int ocfs2_claim_clusters(struct ocfs2_super *osb,
 {
 	unsigned int bits_wanted = ac->ac_bits_wanted - ac->ac_bits_given;
 
-	return __ocfs2_claim_clusters(osb, handle, ac, min_clusters,
+	return __ocfs2_claim_clusters(handle, ac, min_clusters,
 				      bits_wanted, cluster_start, num_clusters);
 }
 
-static inline int ocfs2_block_group_clear_bits(handle_t *handle,
-					       struct inode *alloc_inode,
-					       struct ocfs2_group_desc *bg,
-					       struct buffer_head *group_bh,
-					       unsigned int bit_off,
-					       unsigned int num_bits)
+static int ocfs2_block_group_clear_bits(handle_t *handle,
+					struct inode *alloc_inode,
+					struct ocfs2_group_desc *bg,
+					struct buffer_head *group_bh,
+					unsigned int bit_off,
+					unsigned int num_bits,
+					void (*undo_fn)(unsigned int bit,
+							unsigned long *bmap))
 {
 	int status;
 	unsigned int tmp;
-	int journal_type = OCFS2_JOURNAL_ACCESS_WRITE;
 	struct ocfs2_group_desc *undo_bg = NULL;
-	int cluster_bitmap = 0;
 
 	mlog_entry_void();
 
@@ -1905,20 +2226,18 @@ static inline int ocfs2_block_group_clear_bits(handle_t *handle,
 
 	mlog(0, "off = %u, num = %u\n", bit_off, num_bits);
 
-	if (ocfs2_is_cluster_bitmap(alloc_inode))
-		journal_type = OCFS2_JOURNAL_ACCESS_UNDO;
-
+	BUG_ON(undo_fn && !ocfs2_is_cluster_bitmap(alloc_inode));
 	status = ocfs2_journal_access_gd(handle, INODE_CACHE(alloc_inode),
-					 group_bh, journal_type);
+					 group_bh,
+					 undo_fn ?
+					 OCFS2_JOURNAL_ACCESS_UNDO :
+					 OCFS2_JOURNAL_ACCESS_WRITE);
 	if (status < 0) {
 		mlog_errno(status);
 		goto bail;
 	}
 
-	if (ocfs2_is_cluster_bitmap(alloc_inode))
-		cluster_bitmap = 1;
-
-	if (cluster_bitmap) {
+	if (undo_fn) {
 		jbd_lock_bh_state(group_bh);
 		undo_bg = (struct ocfs2_group_desc *)
 					bh2jh(group_bh)->b_committed_data;
@@ -1929,18 +2248,16 @@ static inline int ocfs2_block_group_clear_bits(handle_t *handle,
 	while(tmp--) {
 		ocfs2_clear_bit((bit_off + tmp),
 				(unsigned long *) bg->bg_bitmap);
-		if (cluster_bitmap)
-			ocfs2_set_bit(bit_off + tmp,
-				      (unsigned long *) undo_bg->bg_bitmap);
+		if (undo_fn)
+			undo_fn(bit_off + tmp,
+				(unsigned long *) undo_bg->bg_bitmap);
 	}
 	le16_add_cpu(&bg->bg_free_bits_count, num_bits);
 
-	if (cluster_bitmap)
+	if (undo_fn)
 		jbd_unlock_bh_state(group_bh);
 
-	status = ocfs2_journal_dirty(handle, group_bh);
-	if (status < 0)
-		mlog_errno(status);
+	ocfs2_journal_dirty(handle, group_bh);
 bail:
 	return status;
 }
@@ -1948,12 +2265,14 @@ bail:
 /*
  * expects the suballoc inode to already be locked.
  */
-int ocfs2_free_suballoc_bits(handle_t *handle,
-			     struct inode *alloc_inode,
-			     struct buffer_head *alloc_bh,
-			     unsigned int start_bit,
-			     u64 bg_blkno,
-			     unsigned int count)
+static int _ocfs2_free_suballoc_bits(handle_t *handle,
+				     struct inode *alloc_inode,
+				     struct buffer_head *alloc_bh,
+				     unsigned int start_bit,
+				     u64 bg_blkno,
+				     unsigned int count,
+				     void (*undo_fn)(unsigned int bit,
+						     unsigned long *bitmap))
 {
 	int status = 0;
 	u32 tmp_used;
@@ -1988,7 +2307,7 @@ int ocfs2_free_suballoc_bits(handle_t *handle,
 
 	status = ocfs2_block_group_clear_bits(handle, alloc_inode,
 					      group, group_bh,
-					      start_bit, count);
+					      start_bit, count, undo_fn);
 	if (status < 0) {
 		mlog_errno(status);
 		goto bail;
@@ -2005,12 +2324,7 @@ int ocfs2_free_suballoc_bits(handle_t *handle,
 		     count);
 	tmp_used = le32_to_cpu(fe->id1.bitmap1.i_used);
 	fe->id1.bitmap1.i_used = cpu_to_le32(tmp_used - count);
-
-	status = ocfs2_journal_dirty(handle, alloc_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto bail;
-	}
+	ocfs2_journal_dirty(handle, alloc_bh);
 
 bail:
 	brelse(group_bh);
@@ -2019,6 +2333,17 @@ bail:
 	return status;
 }
 
+int ocfs2_free_suballoc_bits(handle_t *handle,
+			     struct inode *alloc_inode,
+			     struct buffer_head *alloc_bh,
+			     unsigned int start_bit,
+			     u64 bg_blkno,
+			     unsigned int count)
+{
+	return _ocfs2_free_suballoc_bits(handle, alloc_inode, alloc_bh,
+					 start_bit, bg_blkno, count, NULL);
+}
+
 int ocfs2_free_dinode(handle_t *handle,
 		      struct inode *inode_alloc_inode,
 		      struct buffer_head *inode_alloc_bh,
@@ -2028,15 +2353,19 @@ int ocfs2_free_dinode(handle_t *handle,
 	u16 bit = le16_to_cpu(di->i_suballoc_bit);
 	u64 bg_blkno = ocfs2_which_suballoc_group(blk, bit);
 
+	if (di->i_suballoc_loc)
+		bg_blkno = le64_to_cpu(di->i_suballoc_loc);
 	return ocfs2_free_suballoc_bits(handle, inode_alloc_inode,
 					inode_alloc_bh, bit, bg_blkno, 1);
 }
 
-int ocfs2_free_clusters(handle_t *handle,
-		       struct inode *bitmap_inode,
-		       struct buffer_head *bitmap_bh,
-		       u64 start_blk,
-		       unsigned int num_clusters)
+static int _ocfs2_free_clusters(handle_t *handle,
+				struct inode *bitmap_inode,
+				struct buffer_head *bitmap_bh,
+				u64 start_blk,
+				unsigned int num_clusters,
+				void (*undo_fn)(unsigned int bit,
+						unsigned long *bitmap))
 {
 	int status;
 	u16 bg_start_bit;
@@ -2063,9 +2392,9 @@ int ocfs2_free_clusters(handle_t *handle,
 	mlog(0, "bg_blkno = %llu, bg_start_bit = %u\n",
 	     (unsigned long long)bg_blkno, bg_start_bit);
 
-	status = ocfs2_free_suballoc_bits(handle, bitmap_inode, bitmap_bh,
-					  bg_start_bit, bg_blkno,
-					  num_clusters);
+	status = _ocfs2_free_suballoc_bits(handle, bitmap_inode, bitmap_bh,
+					   bg_start_bit, bg_blkno,
+					   num_clusters, undo_fn);
 	if (status < 0) {
 		mlog_errno(status);
 		goto out;
@@ -2079,6 +2408,32 @@ out:
 	return status;
 }
 
+int ocfs2_free_clusters(handle_t *handle,
+			struct inode *bitmap_inode,
+			struct buffer_head *bitmap_bh,
+			u64 start_blk,
+			unsigned int num_clusters)
+{
+	return _ocfs2_free_clusters(handle, bitmap_inode, bitmap_bh,
+				    start_blk, num_clusters,
+				    _ocfs2_set_bit);
+}
+
+/*
+ * Give never-used clusters back to the global bitmap.  We don't need
+ * to protect these bits in the undo buffer.
+ */
+int ocfs2_release_clusters(handle_t *handle,
+			   struct inode *bitmap_inode,
+			   struct buffer_head *bitmap_bh,
+			   u64 start_blk,
+			   unsigned int num_clusters)
+{
+	return _ocfs2_free_clusters(handle, bitmap_inode, bitmap_bh,
+				    start_blk, num_clusters,
+				    _ocfs2_clear_bit);
+}
+
 static inline void ocfs2_debug_bg(struct ocfs2_group_desc *bg)
 {
 	printk("Block Group:\n");
@@ -2269,7 +2624,7 @@ static int ocfs2_test_suballoc_bit(struct ocfs2_super *osb,
 				   struct buffer_head *alloc_bh, u64 blkno,
 				   u16 bit, int *res)
 {
-	struct ocfs2_dinode *alloc_fe;
+	struct ocfs2_dinode *alloc_di;
 	struct ocfs2_group_desc *group;
 	struct buffer_head *group_bh = NULL;
 	u64 bg_blkno;
@@ -2278,17 +2633,20 @@ static int ocfs2_test_suballoc_bit(struct ocfs2_super *osb,
 	mlog_entry("blkno: %llu bit: %u\n", (unsigned long long)blkno,
 		   (unsigned int)bit);
 
-	alloc_fe = (struct ocfs2_dinode *)alloc_bh->b_data;
-	if ((bit + 1) > ocfs2_bits_per_group(&alloc_fe->id2.i_chain)) {
+	alloc_di = (struct ocfs2_dinode *)alloc_bh->b_data;
+	if ((bit + 1) > ocfs2_bits_per_group(&alloc_di->id2.i_chain)) {
 		mlog(ML_ERROR, "suballoc bit %u out of range of %u\n",
 		     (unsigned int)bit,
-		     ocfs2_bits_per_group(&alloc_fe->id2.i_chain));
+		     ocfs2_bits_per_group(&alloc_di->id2.i_chain));
 		status = -EINVAL;
 		goto bail;
 	}
 
-	bg_blkno = ocfs2_which_suballoc_group(blkno, bit);
-	status = ocfs2_read_group_descriptor(suballoc, alloc_fe, bg_blkno,
+	if (alloc_di->i_suballoc_loc)
+		bg_blkno = le64_to_cpu(alloc_di->i_suballoc_loc);
+	else
+		bg_blkno = ocfs2_which_suballoc_group(blkno, bit);
+	status = ocfs2_read_group_descriptor(suballoc, alloc_di, bg_blkno,
 					     &group_bh);
 	if (status < 0) {
 		mlog(ML_ERROR, "read group %llu failed %d\n",
diff --git a/fs/ocfs2/suballoc.h b/fs/ocfs2/suballoc.h
index 8c9a78a43164..a017dd3ee7d9 100644
--- a/fs/ocfs2/suballoc.h
+++ b/fs/ocfs2/suballoc.h
@@ -26,13 +26,14 @@
 #ifndef _CHAINALLOC_H_
 #define _CHAINALLOC_H_
 
+struct ocfs2_suballoc_result;
 typedef int (group_search_t)(struct inode *,
 			     struct buffer_head *,
 			     u32,			/* bits_wanted */
 			     u32,			/* min_bits */
 			     u64,			/* max_block */
-			     u16 *,			/* *bit_off */
-			     u16 *);			/* *bits_found */
+			     struct ocfs2_suballoc_result *);
+							/* found bits */
 
 struct ocfs2_alloc_context {
 	struct inode *ac_inode;    /* which bitmap are we allocating from? */
@@ -54,8 +55,11 @@ struct ocfs2_alloc_context {
 	u64    ac_last_group;
 	u64    ac_max_block;  /* Highest block number to allocate. 0 is
 				 is the same as ~0 - unlimited */
+
+	struct ocfs2_alloc_reservation	*ac_resv;
 };
 
+void ocfs2_init_steal_slots(struct ocfs2_super *osb);
 void ocfs2_free_alloc_context(struct ocfs2_alloc_context *ac);
 static inline int ocfs2_alloc_context_bits_left(struct ocfs2_alloc_context *ac)
 {
@@ -79,22 +83,21 @@ int ocfs2_reserve_clusters(struct ocfs2_super *osb,
 			   u32 bits_wanted,
 			   struct ocfs2_alloc_context **ac);
 
-int ocfs2_claim_metadata(struct ocfs2_super *osb,
-			 handle_t *handle,
+int ocfs2_claim_metadata(handle_t *handle,
 			 struct ocfs2_alloc_context *ac,
 			 u32 bits_wanted,
+			 u64 *suballoc_loc,
 			 u16 *suballoc_bit_start,
 			 u32 *num_bits,
 			 u64 *blkno_start);
-int ocfs2_claim_new_inode(struct ocfs2_super *osb,
-			  handle_t *handle,
+int ocfs2_claim_new_inode(handle_t *handle,
 			  struct inode *dir,
 			  struct buffer_head *parent_fe_bh,
 			  struct ocfs2_alloc_context *ac,
+			  u64 *suballoc_loc,
 			  u16 *suballoc_bit,
 			  u64 *fe_blkno);
-int ocfs2_claim_clusters(struct ocfs2_super *osb,
-			 handle_t *handle,
+int ocfs2_claim_clusters(handle_t *handle,
 			 struct ocfs2_alloc_context *ac,
 			 u32 min_clusters,
 			 u32 *cluster_start,
@@ -103,8 +106,7 @@ int ocfs2_claim_clusters(struct ocfs2_super *osb,
  * Use this variant of ocfs2_claim_clusters to specify a maxiumum
  * number of clusters smaller than the allocation reserved.
  */
-int __ocfs2_claim_clusters(struct ocfs2_super *osb,
-			   handle_t *handle,
+int __ocfs2_claim_clusters(handle_t *handle,
 			   struct ocfs2_alloc_context *ac,
 			   u32 min_clusters,
 			   u32 max_clusters,
@@ -126,6 +128,11 @@ int ocfs2_free_clusters(handle_t *handle,
 			struct buffer_head *bitmap_bh,
 			u64 start_blk,
 			unsigned int num_clusters);
+int ocfs2_release_clusters(handle_t *handle,
+			   struct inode *bitmap_inode,
+			   struct buffer_head *bitmap_bh,
+			   u64 start_blk,
+			   unsigned int num_clusters);
 
 static inline u64 ocfs2_which_suballoc_group(u64 block, unsigned int bit)
 {
diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c
index 26069917a9f5..0eaa929a4dbf 100644
--- a/fs/ocfs2/super.c
+++ b/fs/ocfs2/super.c
@@ -69,6 +69,7 @@
 #include "xattr.h"
 #include "quota.h"
 #include "refcounttree.h"
+#include "suballoc.h"
 
 #include "buffer_head_io.h"
 
@@ -93,7 +94,9 @@ struct mount_options
 	unsigned long	mount_opt;
 	unsigned int	atime_quantum;
 	signed short	slot;
-	unsigned int	localalloc_opt;
+	int		localalloc_opt;
+	unsigned int	resv_level;
+	int		dir_resv_level;
 	char		cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
 };
 
@@ -175,6 +178,8 @@ enum {
 	Opt_noacl,
 	Opt_usrquota,
 	Opt_grpquota,
+	Opt_resv_level,
+	Opt_dir_resv_level,
 	Opt_err,
 };
 
@@ -201,6 +206,8 @@ static const match_table_t tokens = {
 	{Opt_noacl, "noacl"},
 	{Opt_usrquota, "usrquota"},
 	{Opt_grpquota, "grpquota"},
+	{Opt_resv_level, "resv_level=%u"},
+	{Opt_dir_resv_level, "dir_resv_level=%u"},
 	{Opt_err, NULL}
 };
 
@@ -301,9 +308,12 @@ static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
 
 	spin_lock(&osb->osb_lock);
 	out += snprintf(buf + out, len - out,
-			"%10s => Slot: %d  NumStolen: %d\n", "Steal",
+			"%10s => InodeSlot: %d  StolenInodes: %d, "
+			"MetaSlot: %d  StolenMeta: %d\n", "Steal",
 			osb->s_inode_steal_slot,
-			atomic_read(&osb->s_num_inodes_stolen));
+			atomic_read(&osb->s_num_inodes_stolen),
+			osb->s_meta_steal_slot,
+			atomic_read(&osb->s_num_meta_stolen));
 	spin_unlock(&osb->osb_lock);
 
 	out += snprintf(buf + out, len - out, "OrphanScan => ");
@@ -869,13 +879,15 @@ static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 		if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 			continue;
 		if (unsuspend)
-			status = vfs_quota_enable(
-					sb_dqopt(sb)->files[type],
-					type, QFMT_OCFS2,
-					DQUOT_SUSPENDED);
-		else
-			status = vfs_quota_disable(sb, type,
-						   DQUOT_SUSPENDED);
+			status = dquot_resume(sb, type);
+		else {
+			struct ocfs2_mem_dqinfo *oinfo;
+
+			/* Cancel periodic syncing before suspending */
+			oinfo = sb_dqinfo(sb, type)->dqi_priv;
+			cancel_delayed_work_sync(&oinfo->dqi_sync_work);
+			status = dquot_suspend(sb, type);
+		}
 		if (status < 0)
 			break;
 	}
@@ -906,8 +918,8 @@ static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 			status = -ENOENT;
 			goto out_quota_off;
 		}
-		status = vfs_quota_enable(inode[type], type, QFMT_OCFS2,
-						DQUOT_USAGE_ENABLED);
+		status = dquot_enable(inode[type], type, QFMT_OCFS2,
+				      DQUOT_USAGE_ENABLED);
 		if (status < 0)
 			goto out_quota_off;
 	}
@@ -928,18 +940,22 @@ static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 	int type;
 	struct inode *inode;
 	struct super_block *sb = osb->sb;
+	struct ocfs2_mem_dqinfo *oinfo;
 
 	/* We mostly ignore errors in this function because there's not much
 	 * we can do when we see them */
 	for (type = 0; type < MAXQUOTAS; type++) {
 		if (!sb_has_quota_loaded(sb, type))
 			continue;
+		/* Cancel periodic syncing before we grab dqonoff_mutex */
+		oinfo = sb_dqinfo(sb, type)->dqi_priv;
+		cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 		inode = igrab(sb->s_dquot.files[type]);
 		/* Turn off quotas. This will remove all dquot structures from
 		 * memory and so they will be automatically synced to global
 		 * quota files */
-		vfs_quota_disable(sb, type, DQUOT_USAGE_ENABLED |
-					    DQUOT_LIMITS_ENABLED);
+		dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
+					DQUOT_LIMITS_ENABLED);
 		if (!inode)
 			continue;
 		iput(inode);
@@ -948,7 +964,7 @@ static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 
 /* Handle quota on quotactl */
 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
-			  char *path, int remount)
+			  char *path)
 {
 	unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 					     OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
@@ -956,30 +972,24 @@ static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
 	if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 		return -EINVAL;
 
-	if (remount)
-		return 0;	/* Just ignore it has been handled in
-				 * ocfs2_remount() */
-	return vfs_quota_enable(sb_dqopt(sb)->files[type], type,
-				    format_id, DQUOT_LIMITS_ENABLED);
+	return dquot_enable(sb_dqopt(sb)->files[type], type,
+			    format_id, DQUOT_LIMITS_ENABLED);
 }
 
 /* Handle quota off quotactl */
-static int ocfs2_quota_off(struct super_block *sb, int type, int remount)
+static int ocfs2_quota_off(struct super_block *sb, int type)
 {
-	if (remount)
-		return 0;	/* Ignore now and handle later in
-				 * ocfs2_remount() */
-	return vfs_quota_disable(sb, type, DQUOT_LIMITS_ENABLED);
+	return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
 }
 
 static const struct quotactl_ops ocfs2_quotactl_ops = {
 	.quota_on	= ocfs2_quota_on,
 	.quota_off	= ocfs2_quota_off,
-	.quota_sync	= vfs_quota_sync,
-	.get_info	= vfs_get_dqinfo,
-	.set_info	= vfs_set_dqinfo,
-	.get_dqblk	= vfs_get_dqblk,
-	.set_dqblk	= vfs_set_dqblk,
+	.quota_sync	= dquot_quota_sync,
+	.get_info	= dquot_get_dqinfo,
+	.set_info	= dquot_set_dqinfo,
+	.get_dqblk	= dquot_get_dqblk,
+	.set_dqblk	= dquot_set_dqblk,
 };
 
 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
@@ -1024,8 +1034,14 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
 	osb->s_atime_quantum = parsed_options.atime_quantum;
 	osb->preferred_slot = parsed_options.slot;
 	osb->osb_commit_interval = parsed_options.commit_interval;
-	osb->local_alloc_default_bits = ocfs2_megabytes_to_clusters(sb, parsed_options.localalloc_opt);
-	osb->local_alloc_bits = osb->local_alloc_default_bits;
+
+	ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
+	osb->osb_resv_level = parsed_options.resv_level;
+	osb->osb_dir_resv_level = parsed_options.resv_level;
+	if (parsed_options.dir_resv_level == -1)
+		osb->osb_dir_resv_level = parsed_options.resv_level;
+	else
+		osb->osb_dir_resv_level = parsed_options.dir_resv_level;
 
 	status = ocfs2_verify_userspace_stack(osb, &parsed_options);
 	if (status)
@@ -1062,7 +1078,7 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
 				     "file system, but write access is "
 				     "unavailable.\n");
 			else
-				mlog_errno(status);			
+				mlog_errno(status);
 			goto read_super_error;
 		}
 
@@ -1281,11 +1297,13 @@ static int ocfs2_parse_options(struct super_block *sb,
 		   options ? options : "(none)");
 
 	mopt->commit_interval = 0;
-	mopt->mount_opt = 0;
+	mopt->mount_opt = OCFS2_MOUNT_NOINTR;
 	mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
 	mopt->slot = OCFS2_INVALID_SLOT;
-	mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE;
+	mopt->localalloc_opt = -1;
 	mopt->cluster_stack[0] = '\0';
+	mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
+	mopt->dir_resv_level = -1;
 
 	if (!options) {
 		status = 1;
@@ -1376,7 +1394,7 @@ static int ocfs2_parse_options(struct super_block *sb,
 				status = 0;
 				goto bail;
 			}
-			if (option >= 0 && (option <= ocfs2_local_alloc_size(sb) * 8))
+			if (option >= 0)
 				mopt->localalloc_opt = option;
 			break;
 		case Opt_localflocks:
@@ -1429,6 +1447,28 @@ static int ocfs2_parse_options(struct super_block *sb,
 			mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
 			mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
 			break;
+		case Opt_resv_level:
+			if (is_remount)
+				break;
+			if (match_int(&args[0], &option)) {
+				status = 0;
+				goto bail;
+			}
+			if (option >= OCFS2_MIN_RESV_LEVEL &&
+			    option < OCFS2_MAX_RESV_LEVEL)
+				mopt->resv_level = option;
+			break;
+		case Opt_dir_resv_level:
+			if (is_remount)
+				break;
+			if (match_int(&args[0], &option)) {
+				status = 0;
+				goto bail;
+			}
+			if (option >= OCFS2_MIN_RESV_LEVEL &&
+			    option < OCFS2_MAX_RESV_LEVEL)
+				mopt->dir_resv_level = option;
+			break;
 		default:
 			mlog(ML_ERROR,
 			     "Unrecognized mount option \"%s\" "
@@ -1483,7 +1523,7 @@ static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
 			   (unsigned) (osb->osb_commit_interval / HZ));
 
 	local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
-	if (local_alloc_megs != OCFS2_DEFAULT_LOCAL_ALLOC_SIZE)
+	if (local_alloc_megs != ocfs2_la_default_mb(osb))
 		seq_printf(s, ",localalloc=%d", local_alloc_megs);
 
 	if (opts & OCFS2_MOUNT_LOCALFLOCKS)
@@ -1510,6 +1550,12 @@ static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
 	else
 		seq_printf(s, ",noacl");
 
+	if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
+		seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
+
+	if (osb->osb_dir_resv_level != osb->osb_resv_level)
+		seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
+
 	return 0;
 }
 
@@ -1684,6 +1730,8 @@ static void ocfs2_inode_init_once(void *data)
 	oi->ip_blkno = 0ULL;
 	oi->ip_clusters = 0;
 
+	ocfs2_resv_init_once(&oi->ip_la_data_resv);
+
 	ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
 	ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
 	ocfs2_lock_res_init_once(&oi->ip_open_lockres);
@@ -1997,7 +2045,7 @@ static int ocfs2_initialize_super(struct super_block *sb,
 	osb->blocked_lock_count = 0;
 	spin_lock_init(&osb->osb_lock);
 	spin_lock_init(&osb->osb_xattr_lock);
-	ocfs2_init_inode_steal_slot(osb);
+	ocfs2_init_steal_slots(osb);
 
 	atomic_set(&osb->alloc_stats.moves, 0);
 	atomic_set(&osb->alloc_stats.local_data, 0);
@@ -2038,6 +2086,12 @@ static int ocfs2_initialize_super(struct super_block *sb,
 
 	init_waitqueue_head(&osb->osb_mount_event);
 
+	status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
+	if (status) {
+		mlog_errno(status);
+		goto bail;
+	}
+
 	osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
 	if (!osb->vol_label) {
 		mlog(ML_ERROR, "unable to alloc vol label\n");
@@ -2220,9 +2274,11 @@ static int ocfs2_initialize_super(struct super_block *sb,
 	}
 
 	osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
+	osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
 	iput(inode);
 
-	osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8;
+	osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
+				 osb->s_feature_incompat) * 8;
 
 	status = ocfs2_init_slot_info(osb);
 	if (status < 0) {
@@ -2505,5 +2561,25 @@ void __ocfs2_abort(struct super_block* sb,
 	ocfs2_handle_error(sb);
 }
 
+/*
+ * Void signal blockers, because in-kernel sigprocmask() only fails
+ * when SIG_* is wrong.
+ */
+void ocfs2_block_signals(sigset_t *oldset)
+{
+	int rc;
+	sigset_t blocked;
+
+	sigfillset(&blocked);
+	rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
+	BUG_ON(rc);
+}
+
+void ocfs2_unblock_signals(sigset_t *oldset)
+{
+	int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
+	BUG_ON(rc);
+}
+
 module_init(ocfs2_init);
 module_exit(ocfs2_exit);
diff --git a/fs/ocfs2/super.h b/fs/ocfs2/super.h
index 783f5270f2a1..40c7de084c10 100644
--- a/fs/ocfs2/super.h
+++ b/fs/ocfs2/super.h
@@ -45,4 +45,11 @@ void __ocfs2_abort(struct super_block *sb,
 
 #define ocfs2_abort(sb, fmt, args...) __ocfs2_abort(sb, __PRETTY_FUNCTION__, fmt, ##args)
 
+/*
+ * Void signal blockers, because in-kernel sigprocmask() only fails
+ * when SIG_* is wrong.
+ */
+void ocfs2_block_signals(sigset_t *oldset);
+void ocfs2_unblock_signals(sigset_t *oldset);
+
 #endif /* OCFS2_SUPER_H */
diff --git a/fs/ocfs2/symlink.c b/fs/ocfs2/symlink.c
index 49b133ccbf11..32499d213fc4 100644
--- a/fs/ocfs2/symlink.c
+++ b/fs/ocfs2/symlink.c
@@ -137,20 +137,20 @@ static void *ocfs2_fast_follow_link(struct dentry *dentry,
 	}
 
 	memcpy(link, target, len);
-	nd_set_link(nd, link);
 
 bail:
+	nd_set_link(nd, status ? ERR_PTR(status) : link);
 	brelse(bh);
 
 	mlog_exit(status);
-	return status ? ERR_PTR(status) : link;
+	return NULL;
 }
 
 static void ocfs2_fast_put_link(struct dentry *dentry, struct nameidata *nd, void *cookie)
 {
-	char *link = cookie;
-
-	kfree(link);
+	char *link = nd_get_link(nd);
+	if (!IS_ERR(link))
+		kfree(link);
 }
 
 const struct inode_operations ocfs2_symlink_inode_operations = {
diff --git a/fs/ocfs2/sysfile.c b/fs/ocfs2/sysfile.c
index 40e53702948c..bfe7190cdbf1 100644
--- a/fs/ocfs2/sysfile.c
+++ b/fs/ocfs2/sysfile.c
@@ -25,7 +25,6 @@
 
 #include <linux/fs.h>
 #include <linux/types.h>
-#include <linux/slab.h>
 #include <linux/highmem.h>
 
 #define MLOG_MASK_PREFIX ML_INODE
diff --git a/fs/ocfs2/uptodate.c b/fs/ocfs2/uptodate.c
index c61369342a27..a0a120e82b97 100644
--- a/fs/ocfs2/uptodate.c
+++ b/fs/ocfs2/uptodate.c
@@ -267,8 +267,8 @@ static int ocfs2_buffer_cached(struct ocfs2_caching_info *ci,
 }
 
 /* Warning: even if it returns true, this does *not* guarantee that
- * the block is stored in our inode metadata cache. 
- * 
+ * the block is stored in our inode metadata cache.
+ *
  * This can be called under lock_buffer()
  */
 int ocfs2_buffer_uptodate(struct ocfs2_caching_info *ci,
diff --git a/fs/ocfs2/xattr.c b/fs/ocfs2/xattr.c
index 8fc6fb071c6d..d03469f61801 100644
--- a/fs/ocfs2/xattr.c
+++ b/fs/ocfs2/xattr.c
@@ -79,6 +79,7 @@ struct ocfs2_xattr_set_ctxt {
 	struct ocfs2_alloc_context *meta_ac;
 	struct ocfs2_alloc_context *data_ac;
 	struct ocfs2_cached_dealloc_ctxt dealloc;
+	int set_abort;
 };
 
 #define OCFS2_XATTR_ROOT_SIZE	(sizeof(struct ocfs2_xattr_def_value_root))
@@ -96,7 +97,7 @@ static struct ocfs2_xattr_def_value_root def_xv = {
 	.xv.xr_list.l_count = cpu_to_le16(1),
 };
 
-struct xattr_handler *ocfs2_xattr_handlers[] = {
+const struct xattr_handler *ocfs2_xattr_handlers[] = {
 	&ocfs2_xattr_user_handler,
 	&ocfs2_xattr_acl_access_handler,
 	&ocfs2_xattr_acl_default_handler,
@@ -105,7 +106,7 @@ struct xattr_handler *ocfs2_xattr_handlers[] = {
 	NULL
 };
 
-static struct xattr_handler *ocfs2_xattr_handler_map[OCFS2_XATTR_MAX] = {
+static const struct xattr_handler *ocfs2_xattr_handler_map[OCFS2_XATTR_MAX] = {
 	[OCFS2_XATTR_INDEX_USER]	= &ocfs2_xattr_user_handler,
 	[OCFS2_XATTR_INDEX_POSIX_ACL_ACCESS]
 					= &ocfs2_xattr_acl_access_handler,
@@ -116,10 +117,11 @@ static struct xattr_handler *ocfs2_xattr_handler_map[OCFS2_XATTR_MAX] = {
 };
 
 struct ocfs2_xattr_info {
-	int name_index;
-	const char *name;
-	const void *value;
-	size_t value_len;
+	int		xi_name_index;
+	const char	*xi_name;
+	int		xi_name_len;
+	const void	*xi_value;
+	size_t		xi_value_len;
 };
 
 struct ocfs2_xattr_search {
@@ -137,6 +139,115 @@ struct ocfs2_xattr_search {
 	int not_found;
 };
 
+/* Operations on struct ocfs2_xa_entry */
+struct ocfs2_xa_loc;
+struct ocfs2_xa_loc_operations {
+	/*
+	 * Journal functions
+	 */
+	int (*xlo_journal_access)(handle_t *handle, struct ocfs2_xa_loc *loc,
+				  int type);
+	void (*xlo_journal_dirty)(handle_t *handle, struct ocfs2_xa_loc *loc);
+
+	/*
+	 * Return a pointer to the appropriate buffer in loc->xl_storage
+	 * at the given offset from loc->xl_header.
+	 */
+	void *(*xlo_offset_pointer)(struct ocfs2_xa_loc *loc, int offset);
+
+	/* Can we reuse the existing entry for the new value? */
+	int (*xlo_can_reuse)(struct ocfs2_xa_loc *loc,
+			     struct ocfs2_xattr_info *xi);
+
+	/* How much space is needed for the new value? */
+	int (*xlo_check_space)(struct ocfs2_xa_loc *loc,
+			       struct ocfs2_xattr_info *xi);
+
+	/*
+	 * Return the offset of the first name+value pair.  This is
+	 * the start of our downward-filling free space.
+	 */
+	int (*xlo_get_free_start)(struct ocfs2_xa_loc *loc);
+
+	/*
+	 * Remove the name+value at this location.  Do whatever is
+	 * appropriate with the remaining name+value pairs.
+	 */
+	void (*xlo_wipe_namevalue)(struct ocfs2_xa_loc *loc);
+
+	/* Fill xl_entry with a new entry */
+	void (*xlo_add_entry)(struct ocfs2_xa_loc *loc, u32 name_hash);
+
+	/* Add name+value storage to an entry */
+	void (*xlo_add_namevalue)(struct ocfs2_xa_loc *loc, int size);
+
+	/*
+	 * Initialize the value buf's access and bh fields for this entry.
+	 * ocfs2_xa_fill_value_buf() will handle the xv pointer.
+	 */
+	void (*xlo_fill_value_buf)(struct ocfs2_xa_loc *loc,
+				   struct ocfs2_xattr_value_buf *vb);
+};
+
+/*
+ * Describes an xattr entry location.  This is a memory structure
+ * tracking the on-disk structure.
+ */
+struct ocfs2_xa_loc {
+	/* This xattr belongs to this inode */
+	struct inode *xl_inode;
+
+	/* The ocfs2_xattr_header inside the on-disk storage. Not NULL. */
+	struct ocfs2_xattr_header *xl_header;
+
+	/* Bytes from xl_header to the end of the storage */
+	int xl_size;
+
+	/*
+	 * The ocfs2_xattr_entry this location describes.  If this is
+	 * NULL, this location describes the on-disk structure where it
+	 * would have been.
+	 */
+	struct ocfs2_xattr_entry *xl_entry;
+
+	/*
+	 * Internal housekeeping
+	 */
+
+	/* Buffer(s) containing this entry */
+	void *xl_storage;
+
+	/* Operations on the storage backing this location */
+	const struct ocfs2_xa_loc_operations *xl_ops;
+};
+
+/*
+ * Convenience functions to calculate how much space is needed for a
+ * given name+value pair
+ */
+static int namevalue_size(int name_len, uint64_t value_len)
+{
+	if (value_len > OCFS2_XATTR_INLINE_SIZE)
+		return OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE;
+	else
+		return OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_SIZE(value_len);
+}
+
+static int namevalue_size_xi(struct ocfs2_xattr_info *xi)
+{
+	return namevalue_size(xi->xi_name_len, xi->xi_value_len);
+}
+
+static int namevalue_size_xe(struct ocfs2_xattr_entry *xe)
+{
+	u64 value_len = le64_to_cpu(xe->xe_value_size);
+
+	BUG_ON((value_len > OCFS2_XATTR_INLINE_SIZE) &&
+	       ocfs2_xattr_is_local(xe));
+	return namevalue_size(xe->xe_name_len, value_len);
+}
+
+
 static int ocfs2_xattr_bucket_get_name_value(struct super_block *sb,
 					     struct ocfs2_xattr_header *xh,
 					     int index,
@@ -212,14 +323,6 @@ static inline u16 ocfs2_blocks_per_xattr_bucket(struct super_block *sb)
 	return OCFS2_XATTR_BUCKET_SIZE / (1 << sb->s_blocksize_bits);
 }
 
-static inline u16 ocfs2_xattr_max_xe_in_bucket(struct super_block *sb)
-{
-	u16 len = sb->s_blocksize -
-		 offsetof(struct ocfs2_xattr_header, xh_entries);
-
-	return len / sizeof(struct ocfs2_xattr_entry);
-}
-
 #define bucket_blkno(_b) ((_b)->bu_bhs[0]->b_blocknr)
 #define bucket_block(_b, _n) ((_b)->bu_bhs[(_n)]->b_data)
 #define bucket_xh(_b) ((struct ocfs2_xattr_header *)bucket_block((_b), 0))
@@ -437,7 +540,7 @@ static int ocfs2_read_xattr_block(struct inode *inode, u64 xb_blkno,
 
 static inline const char *ocfs2_xattr_prefix(int name_index)
 {
-	struct xattr_handler *handler = NULL;
+	const struct xattr_handler *handler = NULL;
 
 	if (name_index > 0 && name_index < OCFS2_XATTR_MAX)
 		handler = ocfs2_xattr_handler_map[name_index];
@@ -463,35 +566,22 @@ static u32 ocfs2_xattr_name_hash(struct inode *inode,
 	return hash;
 }
 
-/*
- * ocfs2_xattr_hash_entry()
- *
- * Compute the hash of an extended attribute.
- */
-static void ocfs2_xattr_hash_entry(struct inode *inode,
-				   struct ocfs2_xattr_header *header,
-				   struct ocfs2_xattr_entry *entry)
+static int ocfs2_xattr_entry_real_size(int name_len, size_t value_len)
 {
-	u32 hash = 0;
-	char *name = (char *)header + le16_to_cpu(entry->xe_name_offset);
-
-	hash = ocfs2_xattr_name_hash(inode, name, entry->xe_name_len);
-	entry->xe_name_hash = cpu_to_le32(hash);
-
-	return;
+	return namevalue_size(name_len, value_len) +
+		sizeof(struct ocfs2_xattr_entry);
 }
 
-static int ocfs2_xattr_entry_real_size(int name_len, size_t value_len)
+static int ocfs2_xi_entry_usage(struct ocfs2_xattr_info *xi)
 {
-	int size = 0;
-
-	if (value_len <= OCFS2_XATTR_INLINE_SIZE)
-		size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_SIZE(value_len);
-	else
-		size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE;
-	size += sizeof(struct ocfs2_xattr_entry);
+	return namevalue_size_xi(xi) +
+		sizeof(struct ocfs2_xattr_entry);
+}
 
-	return size;
+static int ocfs2_xe_entry_usage(struct ocfs2_xattr_entry *xe)
+{
+	return namevalue_size_xe(xe) +
+		sizeof(struct ocfs2_xattr_entry);
 }
 
 int ocfs2_calc_security_init(struct inode *dir,
@@ -619,7 +709,7 @@ static int ocfs2_xattr_extend_allocation(struct inode *inode,
 					 struct ocfs2_xattr_value_buf *vb,
 					 struct ocfs2_xattr_set_ctxt *ctxt)
 {
-	int status = 0;
+	int status = 0, credits;
 	handle_t *handle = ctxt->handle;
 	enum ocfs2_alloc_restarted why;
 	u32 prev_clusters, logical_start = le32_to_cpu(vb->vb_xv->xr_clusters);
@@ -629,42 +719,54 @@ static int ocfs2_xattr_extend_allocation(struct inode *inode,
 
 	ocfs2_init_xattr_value_extent_tree(&et, INODE_CACHE(inode), vb);
 
-	status = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh,
-			      OCFS2_JOURNAL_ACCESS_WRITE);
-	if (status < 0) {
-		mlog_errno(status);
-		goto leave;
-	}
-
-	prev_clusters = le32_to_cpu(vb->vb_xv->xr_clusters);
-	status = ocfs2_add_clusters_in_btree(handle,
-					     &et,
-					     &logical_start,
-					     clusters_to_add,
-					     0,
-					     ctxt->data_ac,
-					     ctxt->meta_ac,
-					     &why);
-	if (status < 0) {
-		mlog_errno(status);
-		goto leave;
-	}
+	while (clusters_to_add) {
+		status = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh,
+				       OCFS2_JOURNAL_ACCESS_WRITE);
+		if (status < 0) {
+			mlog_errno(status);
+			break;
+		}
 
-	status = ocfs2_journal_dirty(handle, vb->vb_bh);
-	if (status < 0) {
-		mlog_errno(status);
-		goto leave;
-	}
+		prev_clusters = le32_to_cpu(vb->vb_xv->xr_clusters);
+		status = ocfs2_add_clusters_in_btree(handle,
+						     &et,
+						     &logical_start,
+						     clusters_to_add,
+						     0,
+						     ctxt->data_ac,
+						     ctxt->meta_ac,
+						     &why);
+		if ((status < 0) && (status != -EAGAIN)) {
+			if (status != -ENOSPC)
+				mlog_errno(status);
+			break;
+		}
 
-	clusters_to_add -= le32_to_cpu(vb->vb_xv->xr_clusters) - prev_clusters;
+		ocfs2_journal_dirty(handle, vb->vb_bh);
 
-	/*
-	 * We should have already allocated enough space before the transaction,
-	 * so no need to restart.
-	 */
-	BUG_ON(why != RESTART_NONE || clusters_to_add);
+		clusters_to_add -= le32_to_cpu(vb->vb_xv->xr_clusters) -
+					 prev_clusters;
 
-leave:
+		if (why != RESTART_NONE && clusters_to_add) {
+			/*
+			 * We can only fail in case the alloc file doesn't give
+			 * up enough clusters.
+			 */
+			BUG_ON(why == RESTART_META);
+
+			mlog(0, "restarting xattr value extension for %u"
+			     " clusters,.\n", clusters_to_add);
+			credits = ocfs2_calc_extend_credits(inode->i_sb,
+							    &vb->vb_xv->xr_list,
+							    clusters_to_add);
+			status = ocfs2_extend_trans(handle, credits);
+			if (status < 0) {
+				status = -ENOMEM;
+				mlog_errno(status);
+				break;
+			}
+		}
+	}
 
 	return status;
 }
@@ -697,12 +799,7 @@ static int __ocfs2_remove_xattr_range(struct inode *inode,
 	}
 
 	le32_add_cpu(&vb->vb_xv->xr_clusters, -len);
-
-	ret = ocfs2_journal_dirty(handle, vb->vb_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out;
-	}
+	ocfs2_journal_dirty(handle, vb->vb_bh);
 
 	if (ext_flags & OCFS2_EXT_REFCOUNTED)
 		ret = ocfs2_decrease_refcount(inode, handle,
@@ -1285,11 +1382,7 @@ static int __ocfs2_xattr_set_value_outside(struct inode *inode,
 				memset(bh->b_data + cp_len, 0,
 				       blocksize - cp_len);
 
-			ret = ocfs2_journal_dirty(handle, bh);
-			if (ret < 0) {
-				mlog_errno(ret);
-				goto out;
-			}
+			ocfs2_journal_dirty(handle, bh);
 			brelse(bh);
 			bh = NULL;
 
@@ -1308,452 +1401,901 @@ out:
 	return ret;
 }
 
-static int ocfs2_xattr_cleanup(struct inode *inode,
-			       handle_t *handle,
-			       struct ocfs2_xattr_info *xi,
-			       struct ocfs2_xattr_search *xs,
-			       struct ocfs2_xattr_value_buf *vb,
-			       size_t offs)
+static int ocfs2_xa_check_space_helper(int needed_space, int free_start,
+				       int num_entries)
 {
-	int ret = 0;
-	size_t name_len = strlen(xi->name);
-	void *val = xs->base + offs;
-	size_t size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE;
+	int free_space;
 
-	ret = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh,
-			    OCFS2_JOURNAL_ACCESS_WRITE);
-	if (ret) {
-		mlog_errno(ret);
-		goto out;
-	}
-	/* Decrease xattr count */
-	le16_add_cpu(&xs->header->xh_count, -1);
-	/* Remove the xattr entry and tree root which has already be set*/
-	memset((void *)xs->here, 0, sizeof(struct ocfs2_xattr_entry));
-	memset(val, 0, size);
+	if (!needed_space)
+		return 0;
 
-	ret = ocfs2_journal_dirty(handle, vb->vb_bh);
-	if (ret < 0)
-		mlog_errno(ret);
-out:
-	return ret;
+	free_space = free_start -
+		sizeof(struct ocfs2_xattr_header) -
+		(num_entries * sizeof(struct ocfs2_xattr_entry)) -
+		OCFS2_XATTR_HEADER_GAP;
+	if (free_space < 0)
+		return -EIO;
+	if (free_space < needed_space)
+		return -ENOSPC;
+
+	return 0;
 }
 
-static int ocfs2_xattr_update_entry(struct inode *inode,
-				    handle_t *handle,
-				    struct ocfs2_xattr_info *xi,
-				    struct ocfs2_xattr_search *xs,
-				    struct ocfs2_xattr_value_buf *vb,
-				    size_t offs)
+static int ocfs2_xa_journal_access(handle_t *handle, struct ocfs2_xa_loc *loc,
+				   int type)
 {
-	int ret;
+	return loc->xl_ops->xlo_journal_access(handle, loc, type);
+}
 
-	ret = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh,
-			    OCFS2_JOURNAL_ACCESS_WRITE);
-	if (ret) {
-		mlog_errno(ret);
-		goto out;
-	}
+static void ocfs2_xa_journal_dirty(handle_t *handle, struct ocfs2_xa_loc *loc)
+{
+	loc->xl_ops->xlo_journal_dirty(handle, loc);
+}
 
-	xs->here->xe_name_offset = cpu_to_le16(offs);
-	xs->here->xe_value_size = cpu_to_le64(xi->value_len);
-	if (xi->value_len <= OCFS2_XATTR_INLINE_SIZE)
-		ocfs2_xattr_set_local(xs->here, 1);
-	else
-		ocfs2_xattr_set_local(xs->here, 0);
-	ocfs2_xattr_hash_entry(inode, xs->header, xs->here);
+/* Give a pointer into the storage for the given offset */
+static void *ocfs2_xa_offset_pointer(struct ocfs2_xa_loc *loc, int offset)
+{
+	BUG_ON(offset >= loc->xl_size);
+	return loc->xl_ops->xlo_offset_pointer(loc, offset);
+}
 
-	ret = ocfs2_journal_dirty(handle, vb->vb_bh);
-	if (ret < 0)
-		mlog_errno(ret);
-out:
-	return ret;
+/*
+ * Wipe the name+value pair and allow the storage to reclaim it.  This
+ * must be followed by either removal of the entry or a call to
+ * ocfs2_xa_add_namevalue().
+ */
+static void ocfs2_xa_wipe_namevalue(struct ocfs2_xa_loc *loc)
+{
+	loc->xl_ops->xlo_wipe_namevalue(loc);
 }
 
 /*
- * ocfs2_xattr_set_value_outside()
- *
- * Set large size value in B tree.
+ * Find lowest offset to a name+value pair.  This is the start of our
+ * downward-growing free space.
  */
-static int ocfs2_xattr_set_value_outside(struct inode *inode,
-					 struct ocfs2_xattr_info *xi,
-					 struct ocfs2_xattr_search *xs,
-					 struct ocfs2_xattr_set_ctxt *ctxt,
-					 struct ocfs2_xattr_value_buf *vb,
-					 size_t offs)
+static int ocfs2_xa_get_free_start(struct ocfs2_xa_loc *loc)
 {
-	size_t name_len = strlen(xi->name);
-	void *val = xs->base + offs;
-	struct ocfs2_xattr_value_root *xv = NULL;
-	size_t size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE;
-	int ret = 0;
+	return loc->xl_ops->xlo_get_free_start(loc);
+}
 
-	memset(val, 0, size);
-	memcpy(val, xi->name, name_len);
-	xv = (struct ocfs2_xattr_value_root *)
-		(val + OCFS2_XATTR_SIZE(name_len));
-	xv->xr_clusters = 0;
-	xv->xr_last_eb_blk = 0;
-	xv->xr_list.l_tree_depth = 0;
-	xv->xr_list.l_count = cpu_to_le16(1);
-	xv->xr_list.l_next_free_rec = 0;
-	vb->vb_xv = xv;
-
-	ret = ocfs2_xattr_value_truncate(inode, vb, xi->value_len, ctxt);
-	if (ret < 0) {
-		mlog_errno(ret);
-		return ret;
+/* Can we reuse loc->xl_entry for xi? */
+static int ocfs2_xa_can_reuse_entry(struct ocfs2_xa_loc *loc,
+				    struct ocfs2_xattr_info *xi)
+{
+	return loc->xl_ops->xlo_can_reuse(loc, xi);
+}
+
+/* How much free space is needed to set the new value */
+static int ocfs2_xa_check_space(struct ocfs2_xa_loc *loc,
+				struct ocfs2_xattr_info *xi)
+{
+	return loc->xl_ops->xlo_check_space(loc, xi);
+}
+
+static void ocfs2_xa_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash)
+{
+	loc->xl_ops->xlo_add_entry(loc, name_hash);
+	loc->xl_entry->xe_name_hash = cpu_to_le32(name_hash);
+	/*
+	 * We can't leave the new entry's xe_name_offset at zero or
+	 * add_namevalue() will go nuts.  We set it to the size of our
+	 * storage so that it can never be less than any other entry.
+	 */
+	loc->xl_entry->xe_name_offset = cpu_to_le16(loc->xl_size);
+}
+
+static void ocfs2_xa_add_namevalue(struct ocfs2_xa_loc *loc,
+				   struct ocfs2_xattr_info *xi)
+{
+	int size = namevalue_size_xi(xi);
+	int nameval_offset;
+	char *nameval_buf;
+
+	loc->xl_ops->xlo_add_namevalue(loc, size);
+	loc->xl_entry->xe_value_size = cpu_to_le64(xi->xi_value_len);
+	loc->xl_entry->xe_name_len = xi->xi_name_len;
+	ocfs2_xattr_set_type(loc->xl_entry, xi->xi_name_index);
+	ocfs2_xattr_set_local(loc->xl_entry,
+			      xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE);
+
+	nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset);
+	nameval_buf = ocfs2_xa_offset_pointer(loc, nameval_offset);
+	memset(nameval_buf, 0, size);
+	memcpy(nameval_buf, xi->xi_name, xi->xi_name_len);
+}
+
+static void ocfs2_xa_fill_value_buf(struct ocfs2_xa_loc *loc,
+				    struct ocfs2_xattr_value_buf *vb)
+{
+	int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset);
+	int name_size = OCFS2_XATTR_SIZE(loc->xl_entry->xe_name_len);
+
+	/* Value bufs are for value trees */
+	BUG_ON(ocfs2_xattr_is_local(loc->xl_entry));
+	BUG_ON(namevalue_size_xe(loc->xl_entry) !=
+	       (name_size + OCFS2_XATTR_ROOT_SIZE));
+
+	loc->xl_ops->xlo_fill_value_buf(loc, vb);
+	vb->vb_xv =
+		(struct ocfs2_xattr_value_root *)ocfs2_xa_offset_pointer(loc,
+							nameval_offset +
+							name_size);
+}
+
+static int ocfs2_xa_block_journal_access(handle_t *handle,
+					 struct ocfs2_xa_loc *loc, int type)
+{
+	struct buffer_head *bh = loc->xl_storage;
+	ocfs2_journal_access_func access;
+
+	if (loc->xl_size == (bh->b_size -
+			     offsetof(struct ocfs2_xattr_block,
+				      xb_attrs.xb_header)))
+		access = ocfs2_journal_access_xb;
+	else
+		access = ocfs2_journal_access_di;
+	return access(handle, INODE_CACHE(loc->xl_inode), bh, type);
+}
+
+static void ocfs2_xa_block_journal_dirty(handle_t *handle,
+					 struct ocfs2_xa_loc *loc)
+{
+	struct buffer_head *bh = loc->xl_storage;
+
+	ocfs2_journal_dirty(handle, bh);
+}
+
+static void *ocfs2_xa_block_offset_pointer(struct ocfs2_xa_loc *loc,
+					   int offset)
+{
+	return (char *)loc->xl_header + offset;
+}
+
+static int ocfs2_xa_block_can_reuse(struct ocfs2_xa_loc *loc,
+				    struct ocfs2_xattr_info *xi)
+{
+	/*
+	 * Block storage is strict.  If the sizes aren't exact, we will
+	 * remove the old one and reinsert the new.
+	 */
+	return namevalue_size_xe(loc->xl_entry) ==
+		namevalue_size_xi(xi);
+}
+
+static int ocfs2_xa_block_get_free_start(struct ocfs2_xa_loc *loc)
+{
+	struct ocfs2_xattr_header *xh = loc->xl_header;
+	int i, count = le16_to_cpu(xh->xh_count);
+	int offset, free_start = loc->xl_size;
+
+	for (i = 0; i < count; i++) {
+		offset = le16_to_cpu(xh->xh_entries[i].xe_name_offset);
+		if (offset < free_start)
+			free_start = offset;
 	}
-	ret = ocfs2_xattr_update_entry(inode, ctxt->handle, xi, xs, vb, offs);
-	if (ret < 0) {
-		mlog_errno(ret);
-		return ret;
+
+	return free_start;
+}
+
+static int ocfs2_xa_block_check_space(struct ocfs2_xa_loc *loc,
+				      struct ocfs2_xattr_info *xi)
+{
+	int count = le16_to_cpu(loc->xl_header->xh_count);
+	int free_start = ocfs2_xa_get_free_start(loc);
+	int needed_space = ocfs2_xi_entry_usage(xi);
+
+	/*
+	 * Block storage will reclaim the original entry before inserting
+	 * the new value, so we only need the difference.  If the new
+	 * entry is smaller than the old one, we don't need anything.
+	 */
+	if (loc->xl_entry) {
+		/* Don't need space if we're reusing! */
+		if (ocfs2_xa_can_reuse_entry(loc, xi))
+			needed_space = 0;
+		else
+			needed_space -= ocfs2_xe_entry_usage(loc->xl_entry);
 	}
-	ret = __ocfs2_xattr_set_value_outside(inode, ctxt->handle, vb,
-					      xi->value, xi->value_len);
-	if (ret < 0)
-		mlog_errno(ret);
+	if (needed_space < 0)
+		needed_space = 0;
+	return ocfs2_xa_check_space_helper(needed_space, free_start, count);
+}
 
-	return ret;
+/*
+ * Block storage for xattrs keeps the name+value pairs compacted.  When
+ * we remove one, we have to shift any that preceded it towards the end.
+ */
+static void ocfs2_xa_block_wipe_namevalue(struct ocfs2_xa_loc *loc)
+{
+	int i, offset;
+	int namevalue_offset, first_namevalue_offset, namevalue_size;
+	struct ocfs2_xattr_entry *entry = loc->xl_entry;
+	struct ocfs2_xattr_header *xh = loc->xl_header;
+	int count = le16_to_cpu(xh->xh_count);
+
+	namevalue_offset = le16_to_cpu(entry->xe_name_offset);
+	namevalue_size = namevalue_size_xe(entry);
+	first_namevalue_offset = ocfs2_xa_get_free_start(loc);
+
+	/* Shift the name+value pairs */
+	memmove((char *)xh + first_namevalue_offset + namevalue_size,
+		(char *)xh + first_namevalue_offset,
+		namevalue_offset - first_namevalue_offset);
+	memset((char *)xh + first_namevalue_offset, 0, namevalue_size);
+
+	/* Now tell xh->xh_entries about it */
+	for (i = 0; i < count; i++) {
+		offset = le16_to_cpu(xh->xh_entries[i].xe_name_offset);
+		if (offset <= namevalue_offset)
+			le16_add_cpu(&xh->xh_entries[i].xe_name_offset,
+				     namevalue_size);
+	}
+
+	/*
+	 * Note that we don't update xh_free_start or xh_name_value_len
+	 * because they're not used in block-stored xattrs.
+	 */
+}
+
+static void ocfs2_xa_block_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash)
+{
+	int count = le16_to_cpu(loc->xl_header->xh_count);
+	loc->xl_entry = &(loc->xl_header->xh_entries[count]);
+	le16_add_cpu(&loc->xl_header->xh_count, 1);
+	memset(loc->xl_entry, 0, sizeof(struct ocfs2_xattr_entry));
+}
+
+static void ocfs2_xa_block_add_namevalue(struct ocfs2_xa_loc *loc, int size)
+{
+	int free_start = ocfs2_xa_get_free_start(loc);
+
+	loc->xl_entry->xe_name_offset = cpu_to_le16(free_start - size);
+}
+
+static void ocfs2_xa_block_fill_value_buf(struct ocfs2_xa_loc *loc,
+					  struct ocfs2_xattr_value_buf *vb)
+{
+	struct buffer_head *bh = loc->xl_storage;
+
+	if (loc->xl_size == (bh->b_size -
+			     offsetof(struct ocfs2_xattr_block,
+				      xb_attrs.xb_header)))
+		vb->vb_access = ocfs2_journal_access_xb;
+	else
+		vb->vb_access = ocfs2_journal_access_di;
+	vb->vb_bh = bh;
 }
 
 /*
- * ocfs2_xattr_set_entry_local()
- *
- * Set, replace or remove extended attribute in local.
+ * Operations for xattrs stored in blocks.  This includes inline inode
+ * storage and unindexed ocfs2_xattr_blocks.
  */
-static void ocfs2_xattr_set_entry_local(struct inode *inode,
-					struct ocfs2_xattr_info *xi,
-					struct ocfs2_xattr_search *xs,
-					struct ocfs2_xattr_entry *last,
-					size_t min_offs)
+static const struct ocfs2_xa_loc_operations ocfs2_xa_block_loc_ops = {
+	.xlo_journal_access	= ocfs2_xa_block_journal_access,
+	.xlo_journal_dirty	= ocfs2_xa_block_journal_dirty,
+	.xlo_offset_pointer	= ocfs2_xa_block_offset_pointer,
+	.xlo_check_space	= ocfs2_xa_block_check_space,
+	.xlo_can_reuse		= ocfs2_xa_block_can_reuse,
+	.xlo_get_free_start	= ocfs2_xa_block_get_free_start,
+	.xlo_wipe_namevalue	= ocfs2_xa_block_wipe_namevalue,
+	.xlo_add_entry		= ocfs2_xa_block_add_entry,
+	.xlo_add_namevalue	= ocfs2_xa_block_add_namevalue,
+	.xlo_fill_value_buf	= ocfs2_xa_block_fill_value_buf,
+};
+
+static int ocfs2_xa_bucket_journal_access(handle_t *handle,
+					  struct ocfs2_xa_loc *loc, int type)
 {
-	size_t name_len = strlen(xi->name);
-	int i;
+	struct ocfs2_xattr_bucket *bucket = loc->xl_storage;
 
-	if (xi->value && xs->not_found) {
-		/* Insert the new xattr entry. */
-		le16_add_cpu(&xs->header->xh_count, 1);
-		ocfs2_xattr_set_type(last, xi->name_index);
-		ocfs2_xattr_set_local(last, 1);
-		last->xe_name_len = name_len;
-	} else {
-		void *first_val;
-		void *val;
-		size_t offs, size;
-
-		first_val = xs->base + min_offs;
-		offs = le16_to_cpu(xs->here->xe_name_offset);
-		val = xs->base + offs;
-
-		if (le64_to_cpu(xs->here->xe_value_size) >
-		    OCFS2_XATTR_INLINE_SIZE)
-			size = OCFS2_XATTR_SIZE(name_len) +
-				OCFS2_XATTR_ROOT_SIZE;
+	return ocfs2_xattr_bucket_journal_access(handle, bucket, type);
+}
+
+static void ocfs2_xa_bucket_journal_dirty(handle_t *handle,
+					  struct ocfs2_xa_loc *loc)
+{
+	struct ocfs2_xattr_bucket *bucket = loc->xl_storage;
+
+	ocfs2_xattr_bucket_journal_dirty(handle, bucket);
+}
+
+static void *ocfs2_xa_bucket_offset_pointer(struct ocfs2_xa_loc *loc,
+					    int offset)
+{
+	struct ocfs2_xattr_bucket *bucket = loc->xl_storage;
+	int block, block_offset;
+
+	/* The header is at the front of the bucket */
+	block = offset >> loc->xl_inode->i_sb->s_blocksize_bits;
+	block_offset = offset % loc->xl_inode->i_sb->s_blocksize;
+
+	return bucket_block(bucket, block) + block_offset;
+}
+
+static int ocfs2_xa_bucket_can_reuse(struct ocfs2_xa_loc *loc,
+				     struct ocfs2_xattr_info *xi)
+{
+	return namevalue_size_xe(loc->xl_entry) >=
+		namevalue_size_xi(xi);
+}
+
+static int ocfs2_xa_bucket_get_free_start(struct ocfs2_xa_loc *loc)
+{
+	struct ocfs2_xattr_bucket *bucket = loc->xl_storage;
+	return le16_to_cpu(bucket_xh(bucket)->xh_free_start);
+}
+
+static int ocfs2_bucket_align_free_start(struct super_block *sb,
+					 int free_start, int size)
+{
+	/*
+	 * We need to make sure that the name+value pair fits within
+	 * one block.
+	 */
+	if (((free_start - size) >> sb->s_blocksize_bits) !=
+	    ((free_start - 1) >> sb->s_blocksize_bits))
+		free_start -= free_start % sb->s_blocksize;
+
+	return free_start;
+}
+
+static int ocfs2_xa_bucket_check_space(struct ocfs2_xa_loc *loc,
+				       struct ocfs2_xattr_info *xi)
+{
+	int rc;
+	int count = le16_to_cpu(loc->xl_header->xh_count);
+	int free_start = ocfs2_xa_get_free_start(loc);
+	int needed_space = ocfs2_xi_entry_usage(xi);
+	int size = namevalue_size_xi(xi);
+	struct super_block *sb = loc->xl_inode->i_sb;
+
+	/*
+	 * Bucket storage does not reclaim name+value pairs it cannot
+	 * reuse.  They live as holes until the bucket fills, and then
+	 * the bucket is defragmented.  However, the bucket can reclaim
+	 * the ocfs2_xattr_entry.
+	 */
+	if (loc->xl_entry) {
+		/* Don't need space if we're reusing! */
+		if (ocfs2_xa_can_reuse_entry(loc, xi))
+			needed_space = 0;
 		else
-			size = OCFS2_XATTR_SIZE(name_len) +
-			OCFS2_XATTR_SIZE(le64_to_cpu(xs->here->xe_value_size));
-
-		if (xi->value && size == OCFS2_XATTR_SIZE(name_len) +
-				OCFS2_XATTR_SIZE(xi->value_len)) {
-			/* The old and the new value have the
-			   same size. Just replace the value. */
-			ocfs2_xattr_set_local(xs->here, 1);
-			xs->here->xe_value_size = cpu_to_le64(xi->value_len);
-			/* Clear value bytes. */
-			memset(val + OCFS2_XATTR_SIZE(name_len),
-			       0,
-			       OCFS2_XATTR_SIZE(xi->value_len));
-			memcpy(val + OCFS2_XATTR_SIZE(name_len),
-			       xi->value,
-			       xi->value_len);
-			return;
-		}
-		/* Remove the old name+value. */
-		memmove(first_val + size, first_val, val - first_val);
-		memset(first_val, 0, size);
-		xs->here->xe_name_hash = 0;
-		xs->here->xe_name_offset = 0;
-		ocfs2_xattr_set_local(xs->here, 1);
-		xs->here->xe_value_size = 0;
-
-		min_offs += size;
-
-		/* Adjust all value offsets. */
-		last = xs->header->xh_entries;
-		for (i = 0 ; i < le16_to_cpu(xs->header->xh_count); i++) {
-			size_t o = le16_to_cpu(last->xe_name_offset);
-
-			if (o < offs)
-				last->xe_name_offset = cpu_to_le16(o + size);
-			last += 1;
-		}
+			needed_space -= sizeof(struct ocfs2_xattr_entry);
+	}
+	BUG_ON(needed_space < 0);
 
-		if (!xi->value) {
-			/* Remove the old entry. */
-			last -= 1;
-			memmove(xs->here, xs->here + 1,
-				(void *)last - (void *)xs->here);
-			memset(last, 0, sizeof(struct ocfs2_xattr_entry));
-			le16_add_cpu(&xs->header->xh_count, -1);
-		}
+	if (free_start < size) {
+		if (needed_space)
+			return -ENOSPC;
+	} else {
+		/*
+		 * First we check if it would fit in the first place.
+		 * Below, we align the free start to a block.  This may
+		 * slide us below the minimum gap.  By checking unaligned
+		 * first, we avoid that error.
+		 */
+		rc = ocfs2_xa_check_space_helper(needed_space, free_start,
+						 count);
+		if (rc)
+			return rc;
+		free_start = ocfs2_bucket_align_free_start(sb, free_start,
+							   size);
 	}
-	if (xi->value) {
-		/* Insert the new name+value. */
-		size_t size = OCFS2_XATTR_SIZE(name_len) +
-				OCFS2_XATTR_SIZE(xi->value_len);
-		void *val = xs->base + min_offs - size;
+	return ocfs2_xa_check_space_helper(needed_space, free_start, count);
+}
 
-		xs->here->xe_name_offset = cpu_to_le16(min_offs - size);
-		memset(val, 0, size);
-		memcpy(val, xi->name, name_len);
-		memcpy(val + OCFS2_XATTR_SIZE(name_len),
-		       xi->value,
-		       xi->value_len);
-		xs->here->xe_value_size = cpu_to_le64(xi->value_len);
-		ocfs2_xattr_set_local(xs->here, 1);
-		ocfs2_xattr_hash_entry(inode, xs->header, xs->here);
+static void ocfs2_xa_bucket_wipe_namevalue(struct ocfs2_xa_loc *loc)
+{
+	le16_add_cpu(&loc->xl_header->xh_name_value_len,
+		     -namevalue_size_xe(loc->xl_entry));
+}
+
+static void ocfs2_xa_bucket_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash)
+{
+	struct ocfs2_xattr_header *xh = loc->xl_header;
+	int count = le16_to_cpu(xh->xh_count);
+	int low = 0, high = count - 1, tmp;
+	struct ocfs2_xattr_entry *tmp_xe;
+
+	/*
+	 * We keep buckets sorted by name_hash, so we need to find
+	 * our insert place.
+	 */
+	while (low <= high && count) {
+		tmp = (low + high) / 2;
+		tmp_xe = &xh->xh_entries[tmp];
+
+		if (name_hash > le32_to_cpu(tmp_xe->xe_name_hash))
+			low = tmp + 1;
+		else if (name_hash < le32_to_cpu(tmp_xe->xe_name_hash))
+			high = tmp - 1;
+		else {
+			low = tmp;
+			break;
+		}
 	}
 
-	return;
+	if (low != count)
+		memmove(&xh->xh_entries[low + 1],
+			&xh->xh_entries[low],
+			((count - low) * sizeof(struct ocfs2_xattr_entry)));
+
+	le16_add_cpu(&xh->xh_count, 1);
+	loc->xl_entry = &xh->xh_entries[low];
+	memset(loc->xl_entry, 0, sizeof(struct ocfs2_xattr_entry));
+}
+
+static void ocfs2_xa_bucket_add_namevalue(struct ocfs2_xa_loc *loc, int size)
+{
+	int free_start = ocfs2_xa_get_free_start(loc);
+	struct ocfs2_xattr_header *xh = loc->xl_header;
+	struct super_block *sb = loc->xl_inode->i_sb;
+	int nameval_offset;
+
+	free_start = ocfs2_bucket_align_free_start(sb, free_start, size);
+	nameval_offset = free_start - size;
+	loc->xl_entry->xe_name_offset = cpu_to_le16(nameval_offset);
+	xh->xh_free_start = cpu_to_le16(nameval_offset);
+	le16_add_cpu(&xh->xh_name_value_len, size);
+
+}
+
+static void ocfs2_xa_bucket_fill_value_buf(struct ocfs2_xa_loc *loc,
+					   struct ocfs2_xattr_value_buf *vb)
+{
+	struct ocfs2_xattr_bucket *bucket = loc->xl_storage;
+	struct super_block *sb = loc->xl_inode->i_sb;
+	int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset);
+	int size = namevalue_size_xe(loc->xl_entry);
+	int block_offset = nameval_offset >> sb->s_blocksize_bits;
+
+	/* Values are not allowed to straddle block boundaries */
+	BUG_ON(block_offset !=
+	       ((nameval_offset + size - 1) >> sb->s_blocksize_bits));
+	/* We expect the bucket to be filled in */
+	BUG_ON(!bucket->bu_bhs[block_offset]);
+
+	vb->vb_access = ocfs2_journal_access;
+	vb->vb_bh = bucket->bu_bhs[block_offset];
+}
+
+/* Operations for xattrs stored in buckets. */
+static const struct ocfs2_xa_loc_operations ocfs2_xa_bucket_loc_ops = {
+	.xlo_journal_access	= ocfs2_xa_bucket_journal_access,
+	.xlo_journal_dirty	= ocfs2_xa_bucket_journal_dirty,
+	.xlo_offset_pointer	= ocfs2_xa_bucket_offset_pointer,
+	.xlo_check_space	= ocfs2_xa_bucket_check_space,
+	.xlo_can_reuse		= ocfs2_xa_bucket_can_reuse,
+	.xlo_get_free_start	= ocfs2_xa_bucket_get_free_start,
+	.xlo_wipe_namevalue	= ocfs2_xa_bucket_wipe_namevalue,
+	.xlo_add_entry		= ocfs2_xa_bucket_add_entry,
+	.xlo_add_namevalue	= ocfs2_xa_bucket_add_namevalue,
+	.xlo_fill_value_buf	= ocfs2_xa_bucket_fill_value_buf,
+};
+
+static unsigned int ocfs2_xa_value_clusters(struct ocfs2_xa_loc *loc)
+{
+	struct ocfs2_xattr_value_buf vb;
+
+	if (ocfs2_xattr_is_local(loc->xl_entry))
+		return 0;
+
+	ocfs2_xa_fill_value_buf(loc, &vb);
+	return le32_to_cpu(vb.vb_xv->xr_clusters);
+}
+
+static int ocfs2_xa_value_truncate(struct ocfs2_xa_loc *loc, u64 bytes,
+				   struct ocfs2_xattr_set_ctxt *ctxt)
+{
+	int trunc_rc, access_rc;
+	struct ocfs2_xattr_value_buf vb;
+
+	ocfs2_xa_fill_value_buf(loc, &vb);
+	trunc_rc = ocfs2_xattr_value_truncate(loc->xl_inode, &vb, bytes,
+					      ctxt);
+
+	/*
+	 * The caller of ocfs2_xa_value_truncate() has already called
+	 * ocfs2_xa_journal_access on the loc.  However, The truncate code
+	 * calls ocfs2_extend_trans().  This may commit the previous
+	 * transaction and open a new one.  If this is a bucket, truncate
+	 * could leave only vb->vb_bh set up for journaling.  Meanwhile,
+	 * the caller is expecting to dirty the entire bucket.  So we must
+	 * reset the journal work.  We do this even if truncate has failed,
+	 * as it could have failed after committing the extend.
+	 */
+	access_rc = ocfs2_xa_journal_access(ctxt->handle, loc,
+					    OCFS2_JOURNAL_ACCESS_WRITE);
+
+	/* Errors in truncate take precedence */
+	return trunc_rc ? trunc_rc : access_rc;
+}
+
+static void ocfs2_xa_remove_entry(struct ocfs2_xa_loc *loc)
+{
+	int index, count;
+	struct ocfs2_xattr_header *xh = loc->xl_header;
+	struct ocfs2_xattr_entry *entry = loc->xl_entry;
+
+	ocfs2_xa_wipe_namevalue(loc);
+	loc->xl_entry = NULL;
+
+	le16_add_cpu(&xh->xh_count, -1);
+	count = le16_to_cpu(xh->xh_count);
+
+	/*
+	 * Only zero out the entry if there are more remaining.  This is
+	 * important for an empty bucket, as it keeps track of the
+	 * bucket's hash value.  It doesn't hurt empty block storage.
+	 */
+	if (count) {
+		index = ((char *)entry - (char *)&xh->xh_entries) /
+			sizeof(struct ocfs2_xattr_entry);
+		memmove(&xh->xh_entries[index], &xh->xh_entries[index + 1],
+			(count - index) * sizeof(struct ocfs2_xattr_entry));
+		memset(&xh->xh_entries[count], 0,
+		       sizeof(struct ocfs2_xattr_entry));
+	}
 }
 
 /*
- * ocfs2_xattr_set_entry()
+ * If we have a problem adjusting the size of an external value during
+ * ocfs2_xa_prepare_entry() or ocfs2_xa_remove(), we may have an xattr
+ * in an intermediate state.  For example, the value may be partially
+ * truncated.
+ *
+ * If the value tree hasn't changed, the extend/truncate went nowhere.
+ * We have nothing to do.  The caller can treat it as a straight error.
  *
- * Set extended attribute entry into inode or block.
+ * If the value tree got partially truncated, we now have a corrupted
+ * extended attribute.  We're going to wipe its entry and leak the
+ * clusters.  Better to leak some storage than leave a corrupt entry.
  *
- * If extended attribute value size > OCFS2_XATTR_INLINE_SIZE,
- * We first insert tree root(ocfs2_xattr_value_root) with set_entry_local(),
- * then set value in B tree with set_value_outside().
+ * If the value tree grew, it obviously didn't grow enough for the
+ * new entry.  We're not going to try and reclaim those clusters either.
+ * If there was already an external value there (orig_clusters != 0),
+ * the new clusters are attached safely and we can just leave the old
+ * value in place.  If there was no external value there, we remove
+ * the entry.
+ *
+ * This way, the xattr block we store in the journal will be consistent.
+ * If the size change broke because of the journal, no changes will hit
+ * disk anyway.
  */
-static int ocfs2_xattr_set_entry(struct inode *inode,
-				 struct ocfs2_xattr_info *xi,
-				 struct ocfs2_xattr_search *xs,
-				 struct ocfs2_xattr_set_ctxt *ctxt,
-				 int flag)
-{
-	struct ocfs2_xattr_entry *last;
-	struct ocfs2_inode_info *oi = OCFS2_I(inode);
-	struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data;
-	size_t min_offs = xs->end - xs->base, name_len = strlen(xi->name);
-	size_t size_l = 0;
-	handle_t *handle = ctxt->handle;
-	int free, i, ret;
-	struct ocfs2_xattr_info xi_l = {
-		.name_index = xi->name_index,
-		.name = xi->name,
-		.value = xi->value,
-		.value_len = xi->value_len,
-	};
-	struct ocfs2_xattr_value_buf vb = {
-		.vb_bh = xs->xattr_bh,
-		.vb_access = ocfs2_journal_access_di,
-	};
+static void ocfs2_xa_cleanup_value_truncate(struct ocfs2_xa_loc *loc,
+					    const char *what,
+					    unsigned int orig_clusters)
+{
+	unsigned int new_clusters = ocfs2_xa_value_clusters(loc);
+	char *nameval_buf = ocfs2_xa_offset_pointer(loc,
+				le16_to_cpu(loc->xl_entry->xe_name_offset));
+
+	if (new_clusters < orig_clusters) {
+		mlog(ML_ERROR,
+		     "Partial truncate while %s xattr %.*s.  Leaking "
+		     "%u clusters and removing the entry\n",
+		     what, loc->xl_entry->xe_name_len, nameval_buf,
+		     orig_clusters - new_clusters);
+		ocfs2_xa_remove_entry(loc);
+	} else if (!orig_clusters) {
+		mlog(ML_ERROR,
+		     "Unable to allocate an external value for xattr "
+		     "%.*s safely.  Leaking %u clusters and removing the "
+		     "entry\n",
+		     loc->xl_entry->xe_name_len, nameval_buf,
+		     new_clusters - orig_clusters);
+		ocfs2_xa_remove_entry(loc);
+	} else if (new_clusters > orig_clusters)
+		mlog(ML_ERROR,
+		     "Unable to grow xattr %.*s safely.  %u new clusters "
+		     "have been added, but the value will not be "
+		     "modified\n",
+		     loc->xl_entry->xe_name_len, nameval_buf,
+		     new_clusters - orig_clusters);
+}
+
+static int ocfs2_xa_remove(struct ocfs2_xa_loc *loc,
+			   struct ocfs2_xattr_set_ctxt *ctxt)
+{
+	int rc = 0;
+	unsigned int orig_clusters;
+
+	if (!ocfs2_xattr_is_local(loc->xl_entry)) {
+		orig_clusters = ocfs2_xa_value_clusters(loc);
+		rc = ocfs2_xa_value_truncate(loc, 0, ctxt);
+		if (rc) {
+			mlog_errno(rc);
+			/*
+			 * Since this is remove, we can return 0 if
+			 * ocfs2_xa_cleanup_value_truncate() is going to
+			 * wipe the entry anyway.  So we check the
+			 * cluster count as well.
+			 */
+			if (orig_clusters != ocfs2_xa_value_clusters(loc))
+				rc = 0;
+			ocfs2_xa_cleanup_value_truncate(loc, "removing",
+							orig_clusters);
+			if (rc)
+				goto out;
+		}
+	}
 
-	if (!(flag & OCFS2_INLINE_XATTR_FL)) {
-		BUG_ON(xs->xattr_bh == xs->inode_bh);
-		vb.vb_access = ocfs2_journal_access_xb;
-	} else
-		BUG_ON(xs->xattr_bh != xs->inode_bh);
+	ocfs2_xa_remove_entry(loc);
 
-	/* Compute min_offs, last and free space. */
-	last = xs->header->xh_entries;
+out:
+	return rc;
+}
 
-	for (i = 0 ; i < le16_to_cpu(xs->header->xh_count); i++) {
-		size_t offs = le16_to_cpu(last->xe_name_offset);
-		if (offs < min_offs)
-			min_offs = offs;
-		last += 1;
-	}
+static void ocfs2_xa_install_value_root(struct ocfs2_xa_loc *loc)
+{
+	int name_size = OCFS2_XATTR_SIZE(loc->xl_entry->xe_name_len);
+	char *nameval_buf;
 
-	free = min_offs - ((void *)last - xs->base) - OCFS2_XATTR_HEADER_GAP;
-	if (free < 0)
-		return -EIO;
+	nameval_buf = ocfs2_xa_offset_pointer(loc,
+				le16_to_cpu(loc->xl_entry->xe_name_offset));
+	memcpy(nameval_buf + name_size, &def_xv, OCFS2_XATTR_ROOT_SIZE);
+}
 
-	if (!xs->not_found) {
-		size_t size = 0;
-		if (ocfs2_xattr_is_local(xs->here))
-			size = OCFS2_XATTR_SIZE(name_len) +
-			OCFS2_XATTR_SIZE(le64_to_cpu(xs->here->xe_value_size));
-		else
-			size = OCFS2_XATTR_SIZE(name_len) +
-				OCFS2_XATTR_ROOT_SIZE;
-		free += (size + sizeof(struct ocfs2_xattr_entry));
-	}
-	/* Check free space in inode or block */
-	if (xi->value && xi->value_len > OCFS2_XATTR_INLINE_SIZE) {
-		if (free < sizeof(struct ocfs2_xattr_entry) +
-			   OCFS2_XATTR_SIZE(name_len) +
-			   OCFS2_XATTR_ROOT_SIZE) {
-			ret = -ENOSPC;
-			goto out;
+/*
+ * Take an existing entry and make it ready for the new value.  This
+ * won't allocate space, but it may free space.  It should be ready for
+ * ocfs2_xa_prepare_entry() to finish the work.
+ */
+static int ocfs2_xa_reuse_entry(struct ocfs2_xa_loc *loc,
+				struct ocfs2_xattr_info *xi,
+				struct ocfs2_xattr_set_ctxt *ctxt)
+{
+	int rc = 0;
+	int name_size = OCFS2_XATTR_SIZE(xi->xi_name_len);
+	unsigned int orig_clusters;
+	char *nameval_buf;
+	int xe_local = ocfs2_xattr_is_local(loc->xl_entry);
+	int xi_local = xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE;
+
+	BUG_ON(OCFS2_XATTR_SIZE(loc->xl_entry->xe_name_len) !=
+	       name_size);
+
+	nameval_buf = ocfs2_xa_offset_pointer(loc,
+				le16_to_cpu(loc->xl_entry->xe_name_offset));
+	if (xe_local) {
+		memset(nameval_buf + name_size, 0,
+		       namevalue_size_xe(loc->xl_entry) - name_size);
+		if (!xi_local)
+			ocfs2_xa_install_value_root(loc);
+	} else {
+		orig_clusters = ocfs2_xa_value_clusters(loc);
+		if (xi_local) {
+			rc = ocfs2_xa_value_truncate(loc, 0, ctxt);
+			if (rc < 0)
+				mlog_errno(rc);
+			else
+				memset(nameval_buf + name_size, 0,
+				       namevalue_size_xe(loc->xl_entry) -
+				       name_size);
+		} else if (le64_to_cpu(loc->xl_entry->xe_value_size) >
+			   xi->xi_value_len) {
+			rc = ocfs2_xa_value_truncate(loc, xi->xi_value_len,
+						     ctxt);
+			if (rc < 0)
+				mlog_errno(rc);
 		}
-		size_l = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE;
-		xi_l.value = (void *)&def_xv;
-		xi_l.value_len = OCFS2_XATTR_ROOT_SIZE;
-	} else if (xi->value) {
-		if (free < sizeof(struct ocfs2_xattr_entry) +
-			   OCFS2_XATTR_SIZE(name_len) +
-			   OCFS2_XATTR_SIZE(xi->value_len)) {
-			ret = -ENOSPC;
+
+		if (rc) {
+			ocfs2_xa_cleanup_value_truncate(loc, "reusing",
+							orig_clusters);
 			goto out;
 		}
 	}
 
-	if (!xs->not_found) {
-		/* For existing extended attribute */
-		size_t size = OCFS2_XATTR_SIZE(name_len) +
-			OCFS2_XATTR_SIZE(le64_to_cpu(xs->here->xe_value_size));
-		size_t offs = le16_to_cpu(xs->here->xe_name_offset);
-		void *val = xs->base + offs;
+	loc->xl_entry->xe_value_size = cpu_to_le64(xi->xi_value_len);
+	ocfs2_xattr_set_local(loc->xl_entry, xi_local);
 
-		if (ocfs2_xattr_is_local(xs->here) && size == size_l) {
-			/* Replace existing local xattr with tree root */
-			ret = ocfs2_xattr_set_value_outside(inode, xi, xs,
-							    ctxt, &vb, offs);
-			if (ret < 0)
-				mlog_errno(ret);
-			goto out;
-		} else if (!ocfs2_xattr_is_local(xs->here)) {
-			/* For existing xattr which has value outside */
-			vb.vb_xv = (struct ocfs2_xattr_value_root *)
-				(val + OCFS2_XATTR_SIZE(name_len));
+out:
+	return rc;
+}
 
-			if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) {
-				/*
-				 * If new value need set outside also,
-				 * first truncate old value to new value,
-				 * then set new value with set_value_outside().
-				 */
-				ret = ocfs2_xattr_value_truncate(inode,
-								 &vb,
-								 xi->value_len,
-								 ctxt);
-				if (ret < 0) {
-					mlog_errno(ret);
-					goto out;
-				}
+/*
+ * Prepares loc->xl_entry to receive the new xattr.  This includes
+ * properly setting up the name+value pair region.  If loc->xl_entry
+ * already exists, it will take care of modifying it appropriately.
+ *
+ * Note that this modifies the data.  You did journal_access already,
+ * right?
+ */
+static int ocfs2_xa_prepare_entry(struct ocfs2_xa_loc *loc,
+				  struct ocfs2_xattr_info *xi,
+				  u32 name_hash,
+				  struct ocfs2_xattr_set_ctxt *ctxt)
+{
+	int rc = 0;
+	unsigned int orig_clusters;
+	__le64 orig_value_size = 0;
 
-				ret = ocfs2_xattr_update_entry(inode,
-							       handle,
-							       xi,
-							       xs,
-							       &vb,
-							       offs);
-				if (ret < 0) {
-					mlog_errno(ret);
-					goto out;
-				}
+	rc = ocfs2_xa_check_space(loc, xi);
+	if (rc)
+		goto out;
 
-				ret = __ocfs2_xattr_set_value_outside(inode,
-								handle,
-								&vb,
-								xi->value,
-								xi->value_len);
-				if (ret < 0)
-					mlog_errno(ret);
+	if (loc->xl_entry) {
+		if (ocfs2_xa_can_reuse_entry(loc, xi)) {
+			orig_value_size = loc->xl_entry->xe_value_size;
+			rc = ocfs2_xa_reuse_entry(loc, xi, ctxt);
+			if (rc)
+				goto out;
+			goto alloc_value;
+		}
+
+		if (!ocfs2_xattr_is_local(loc->xl_entry)) {
+			orig_clusters = ocfs2_xa_value_clusters(loc);
+			rc = ocfs2_xa_value_truncate(loc, 0, ctxt);
+			if (rc) {
+				mlog_errno(rc);
+				ocfs2_xa_cleanup_value_truncate(loc,
+								"overwriting",
+								orig_clusters);
 				goto out;
-			} else {
-				/*
-				 * If new value need set in local,
-				 * just trucate old value to zero.
-				 */
-				 ret = ocfs2_xattr_value_truncate(inode,
-								  &vb,
-								  0,
-								  ctxt);
-				if (ret < 0)
-					mlog_errno(ret);
 			}
 		}
+		ocfs2_xa_wipe_namevalue(loc);
+	} else
+		ocfs2_xa_add_entry(loc, name_hash);
+
+	/*
+	 * If we get here, we have a blank entry.  Fill it.  We grow our
+	 * name+value pair back from the end.
+	 */
+	ocfs2_xa_add_namevalue(loc, xi);
+	if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE)
+		ocfs2_xa_install_value_root(loc);
+
+alloc_value:
+	if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) {
+		orig_clusters = ocfs2_xa_value_clusters(loc);
+		rc = ocfs2_xa_value_truncate(loc, xi->xi_value_len, ctxt);
+		if (rc < 0) {
+			ctxt->set_abort = 1;
+			ocfs2_xa_cleanup_value_truncate(loc, "growing",
+							orig_clusters);
+			/*
+			 * If we were growing an existing value,
+			 * ocfs2_xa_cleanup_value_truncate() won't remove
+			 * the entry. We need to restore the original value
+			 * size.
+			 */
+			if (loc->xl_entry) {
+				BUG_ON(!orig_value_size);
+				loc->xl_entry->xe_value_size = orig_value_size;
+			}
+			mlog_errno(rc);
+		}
 	}
 
-	ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), xs->inode_bh,
+out:
+	return rc;
+}
+
+/*
+ * Store the value portion of the name+value pair.  This will skip
+ * values that are stored externally.  Their tree roots were set up
+ * by ocfs2_xa_prepare_entry().
+ */
+static int ocfs2_xa_store_value(struct ocfs2_xa_loc *loc,
+				struct ocfs2_xattr_info *xi,
+				struct ocfs2_xattr_set_ctxt *ctxt)
+{
+	int rc = 0;
+	int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset);
+	int name_size = OCFS2_XATTR_SIZE(xi->xi_name_len);
+	char *nameval_buf;
+	struct ocfs2_xattr_value_buf vb;
+
+	nameval_buf = ocfs2_xa_offset_pointer(loc, nameval_offset);
+	if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) {
+		ocfs2_xa_fill_value_buf(loc, &vb);
+		rc = __ocfs2_xattr_set_value_outside(loc->xl_inode,
+						     ctxt->handle, &vb,
+						     xi->xi_value,
+						     xi->xi_value_len);
+	} else
+		memcpy(nameval_buf + name_size, xi->xi_value, xi->xi_value_len);
+
+	return rc;
+}
+
+static int ocfs2_xa_set(struct ocfs2_xa_loc *loc,
+			struct ocfs2_xattr_info *xi,
+			struct ocfs2_xattr_set_ctxt *ctxt)
+{
+	int ret;
+	u32 name_hash = ocfs2_xattr_name_hash(loc->xl_inode, xi->xi_name,
+					      xi->xi_name_len);
+
+	ret = ocfs2_xa_journal_access(ctxt->handle, loc,
 				      OCFS2_JOURNAL_ACCESS_WRITE);
 	if (ret) {
 		mlog_errno(ret);
 		goto out;
 	}
 
-	if (!(flag & OCFS2_INLINE_XATTR_FL)) {
-		ret = vb.vb_access(handle, INODE_CACHE(inode), vb.vb_bh,
-				   OCFS2_JOURNAL_ACCESS_WRITE);
-		if (ret) {
-			mlog_errno(ret);
-			goto out;
-		}
-	}
-
 	/*
-	 * Set value in local, include set tree root in local.
-	 * This is the first step for value size >INLINE_SIZE.
+	 * From here on out, everything is going to modify the buffer a
+	 * little.  Errors are going to leave the xattr header in a
+	 * sane state.  Thus, even with errors we dirty the sucker.
 	 */
-	ocfs2_xattr_set_entry_local(inode, &xi_l, xs, last, min_offs);
 
-	if (!(flag & OCFS2_INLINE_XATTR_FL)) {
-		ret = ocfs2_journal_dirty(handle, xs->xattr_bh);
-		if (ret < 0) {
-			mlog_errno(ret);
-			goto out;
-		}
+	/* Don't worry, we are never called with !xi_value and !xl_entry */
+	if (!xi->xi_value) {
+		ret = ocfs2_xa_remove(loc, ctxt);
+		goto out_dirty;
 	}
 
-	if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) &&
-	    (flag & OCFS2_INLINE_XATTR_FL)) {
-		struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
-		unsigned int xattrsize = osb->s_xattr_inline_size;
-
-		/*
-		 * Adjust extent record count or inline data size
-		 * to reserve space for extended attribute.
-		 */
-		if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
-			struct ocfs2_inline_data *idata = &di->id2.i_data;
-			le16_add_cpu(&idata->id_count, -xattrsize);
-		} else if (!(ocfs2_inode_is_fast_symlink(inode))) {
-			struct ocfs2_extent_list *el = &di->id2.i_list;
-			le16_add_cpu(&el->l_count, -(xattrsize /
-					sizeof(struct ocfs2_extent_rec)));
-		}
-		di->i_xattr_inline_size = cpu_to_le16(xattrsize);
+	ret = ocfs2_xa_prepare_entry(loc, xi, name_hash, ctxt);
+	if (ret) {
+		if (ret != -ENOSPC)
+			mlog_errno(ret);
+		goto out_dirty;
 	}
-	/* Update xattr flag */
-	spin_lock(&oi->ip_lock);
-	oi->ip_dyn_features |= flag;
-	di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features);
-	spin_unlock(&oi->ip_lock);
 
-	ret = ocfs2_journal_dirty(handle, xs->inode_bh);
-	if (ret < 0)
+	ret = ocfs2_xa_store_value(loc, xi, ctxt);
+	if (ret)
 		mlog_errno(ret);
 
-	if (!ret && xi->value_len > OCFS2_XATTR_INLINE_SIZE) {
-		/*
-		 * Set value outside in B tree.
-		 * This is the second step for value size > INLINE_SIZE.
-		 */
-		size_t offs = le16_to_cpu(xs->here->xe_name_offset);
-		ret = ocfs2_xattr_set_value_outside(inode, xi, xs, ctxt,
-						    &vb, offs);
-		if (ret < 0) {
-			int ret2;
+out_dirty:
+	ocfs2_xa_journal_dirty(ctxt->handle, loc);
 
-			mlog_errno(ret);
-			/*
-			 * If set value outside failed, we have to clean
-			 * the junk tree root we have already set in local.
-			 */
-			ret2 = ocfs2_xattr_cleanup(inode, ctxt->handle,
-						   xi, xs, &vb, offs);
-			if (ret2 < 0)
-				mlog_errno(ret2);
-		}
-	}
 out:
 	return ret;
 }
 
+static void ocfs2_init_dinode_xa_loc(struct ocfs2_xa_loc *loc,
+				     struct inode *inode,
+				     struct buffer_head *bh,
+				     struct ocfs2_xattr_entry *entry)
+{
+	struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
+
+	BUG_ON(!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_XATTR_FL));
+
+	loc->xl_inode = inode;
+	loc->xl_ops = &ocfs2_xa_block_loc_ops;
+	loc->xl_storage = bh;
+	loc->xl_entry = entry;
+	loc->xl_size = le16_to_cpu(di->i_xattr_inline_size);
+	loc->xl_header =
+		(struct ocfs2_xattr_header *)(bh->b_data + bh->b_size -
+					      loc->xl_size);
+}
+
+static void ocfs2_init_xattr_block_xa_loc(struct ocfs2_xa_loc *loc,
+					  struct inode *inode,
+					  struct buffer_head *bh,
+					  struct ocfs2_xattr_entry *entry)
+{
+	struct ocfs2_xattr_block *xb =
+		(struct ocfs2_xattr_block *)bh->b_data;
+
+	BUG_ON(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED);
+
+	loc->xl_inode = inode;
+	loc->xl_ops = &ocfs2_xa_block_loc_ops;
+	loc->xl_storage = bh;
+	loc->xl_header = &(xb->xb_attrs.xb_header);
+	loc->xl_entry = entry;
+	loc->xl_size = bh->b_size - offsetof(struct ocfs2_xattr_block,
+					     xb_attrs.xb_header);
+}
+
+static void ocfs2_init_xattr_bucket_xa_loc(struct ocfs2_xa_loc *loc,
+					   struct ocfs2_xattr_bucket *bucket,
+					   struct ocfs2_xattr_entry *entry)
+{
+	loc->xl_inode = bucket->bu_inode;
+	loc->xl_ops = &ocfs2_xa_bucket_loc_ops;
+	loc->xl_storage = bucket;
+	loc->xl_header = bucket_xh(bucket);
+	loc->xl_entry = entry;
+	loc->xl_size = OCFS2_XATTR_BUCKET_SIZE;
+}
+
 /*
  * In xattr remove, if it is stored outside and refcounted, we may have
  * the chance to split the refcount tree. So need the allocators.
@@ -1945,7 +2487,10 @@ static int ocfs2_xattr_free_block(struct inode *inode,
 	xb = (struct ocfs2_xattr_block *)blk_bh->b_data;
 	blk = le64_to_cpu(xb->xb_blkno);
 	bit = le16_to_cpu(xb->xb_suballoc_bit);
-	bg_blkno = ocfs2_which_suballoc_group(blk, bit);
+	if (xb->xb_suballoc_loc)
+		bg_blkno = le64_to_cpu(xb->xb_suballoc_loc);
+	else
+		bg_blkno = ocfs2_which_suballoc_group(blk, bit);
 
 	xb_alloc_inode = ocfs2_get_system_file_inode(osb,
 				EXTENT_ALLOC_SYSTEM_INODE,
@@ -2060,9 +2605,7 @@ int ocfs2_xattr_remove(struct inode *inode, struct buffer_head *di_bh)
 	di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features);
 	spin_unlock(&oi->ip_lock);
 
-	ret = ocfs2_journal_dirty(handle, di_bh);
-	if (ret < 0)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, di_bh);
 out_commit:
 	ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle);
 out:
@@ -2149,6 +2692,53 @@ static int ocfs2_xattr_ibody_find(struct inode *inode,
 	return 0;
 }
 
+static int ocfs2_xattr_ibody_init(struct inode *inode,
+				  struct buffer_head *di_bh,
+				  struct ocfs2_xattr_set_ctxt *ctxt)
+{
+	int ret;
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	unsigned int xattrsize = osb->s_xattr_inline_size;
+
+	if (!ocfs2_xattr_has_space_inline(inode, di)) {
+		ret = -ENOSPC;
+		goto out;
+	}
+
+	ret = ocfs2_journal_access_di(ctxt->handle, INODE_CACHE(inode), di_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (ret) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	/*
+	 * Adjust extent record count or inline data size
+	 * to reserve space for extended attribute.
+	 */
+	if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
+		struct ocfs2_inline_data *idata = &di->id2.i_data;
+		le16_add_cpu(&idata->id_count, -xattrsize);
+	} else if (!(ocfs2_inode_is_fast_symlink(inode))) {
+		struct ocfs2_extent_list *el = &di->id2.i_list;
+		le16_add_cpu(&el->l_count, -(xattrsize /
+					     sizeof(struct ocfs2_extent_rec)));
+	}
+	di->i_xattr_inline_size = cpu_to_le16(xattrsize);
+
+	spin_lock(&oi->ip_lock);
+	oi->ip_dyn_features |= OCFS2_INLINE_XATTR_FL|OCFS2_HAS_XATTR_FL;
+	di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features);
+	spin_unlock(&oi->ip_lock);
+
+	ocfs2_journal_dirty(ctxt->handle, di_bh);
+
+out:
+	return ret;
+}
+
 /*
  * ocfs2_xattr_ibody_set()
  *
@@ -2160,9 +2750,10 @@ static int ocfs2_xattr_ibody_set(struct inode *inode,
 				 struct ocfs2_xattr_search *xs,
 				 struct ocfs2_xattr_set_ctxt *ctxt)
 {
+	int ret;
 	struct ocfs2_inode_info *oi = OCFS2_I(inode);
 	struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data;
-	int ret;
+	struct ocfs2_xa_loc loc;
 
 	if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE)
 		return -ENOSPC;
@@ -2175,8 +2766,25 @@ static int ocfs2_xattr_ibody_set(struct inode *inode,
 		}
 	}
 
-	ret = ocfs2_xattr_set_entry(inode, xi, xs, ctxt,
-				(OCFS2_INLINE_XATTR_FL | OCFS2_HAS_XATTR_FL));
+	if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) {
+		ret = ocfs2_xattr_ibody_init(inode, xs->inode_bh, ctxt);
+		if (ret) {
+			if (ret != -ENOSPC)
+				mlog_errno(ret);
+			goto out;
+		}
+	}
+
+	ocfs2_init_dinode_xa_loc(&loc, inode, xs->inode_bh,
+				 xs->not_found ? NULL : xs->here);
+	ret = ocfs2_xa_set(&loc, xi, ctxt);
+	if (ret) {
+		if (ret != -ENOSPC)
+			mlog_errno(ret);
+		goto out;
+	}
+	xs->here = loc.xl_entry;
+
 out:
 	up_write(&oi->ip_alloc_sem);
 
@@ -2236,32 +2844,30 @@ cleanup:
 	return ret;
 }
 
-static int ocfs2_create_xattr_block(handle_t *handle,
-				    struct inode *inode,
+static int ocfs2_create_xattr_block(struct inode *inode,
 				    struct buffer_head *inode_bh,
-				    struct ocfs2_alloc_context *meta_ac,
-				    struct buffer_head **ret_bh,
-				    int indexed)
+				    struct ocfs2_xattr_set_ctxt *ctxt,
+				    int indexed,
+				    struct buffer_head **ret_bh)
 {
 	int ret;
 	u16 suballoc_bit_start;
 	u32 num_got;
-	u64 first_blkno;
+	u64 suballoc_loc, first_blkno;
 	struct ocfs2_dinode *di =  (struct ocfs2_dinode *)inode_bh->b_data;
-	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 	struct buffer_head *new_bh = NULL;
 	struct ocfs2_xattr_block *xblk;
 
-	ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), inode_bh,
-				      OCFS2_JOURNAL_ACCESS_CREATE);
+	ret = ocfs2_journal_access_di(ctxt->handle, INODE_CACHE(inode),
+				      inode_bh, OCFS2_JOURNAL_ACCESS_CREATE);
 	if (ret < 0) {
 		mlog_errno(ret);
 		goto end;
 	}
 
-	ret = ocfs2_claim_metadata(osb, handle, meta_ac, 1,
-				   &suballoc_bit_start, &num_got,
-				   &first_blkno);
+	ret = ocfs2_claim_metadata(ctxt->handle, ctxt->meta_ac, 1,
+				   &suballoc_loc, &suballoc_bit_start,
+				   &num_got, &first_blkno);
 	if (ret < 0) {
 		mlog_errno(ret);
 		goto end;
@@ -2270,7 +2876,7 @@ static int ocfs2_create_xattr_block(handle_t *handle,
 	new_bh = sb_getblk(inode->i_sb, first_blkno);
 	ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode), new_bh);
 
-	ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode),
+	ret = ocfs2_journal_access_xb(ctxt->handle, INODE_CACHE(inode),
 				      new_bh,
 				      OCFS2_JOURNAL_ACCESS_CREATE);
 	if (ret < 0) {
@@ -2282,11 +2888,12 @@ static int ocfs2_create_xattr_block(handle_t *handle,
 	xblk = (struct ocfs2_xattr_block *)new_bh->b_data;
 	memset(xblk, 0, inode->i_sb->s_blocksize);
 	strcpy((void *)xblk, OCFS2_XATTR_BLOCK_SIGNATURE);
-	xblk->xb_suballoc_slot = cpu_to_le16(osb->slot_num);
+	xblk->xb_suballoc_slot = cpu_to_le16(ctxt->meta_ac->ac_alloc_slot);
+	xblk->xb_suballoc_loc = cpu_to_le64(suballoc_loc);
 	xblk->xb_suballoc_bit = cpu_to_le16(suballoc_bit_start);
-	xblk->xb_fs_generation = cpu_to_le32(osb->fs_generation);
+	xblk->xb_fs_generation =
+		cpu_to_le32(OCFS2_SB(inode->i_sb)->fs_generation);
 	xblk->xb_blkno = cpu_to_le64(first_blkno);
-
 	if (indexed) {
 		struct ocfs2_xattr_tree_root *xr = &xblk->xb_attrs.xb_root;
 		xr->xt_clusters = cpu_to_le32(1);
@@ -2297,14 +2904,17 @@ static int ocfs2_create_xattr_block(handle_t *handle,
 		xr->xt_list.l_next_free_rec = cpu_to_le16(1);
 		xblk->xb_flags = cpu_to_le16(OCFS2_XATTR_INDEXED);
 	}
+	ocfs2_journal_dirty(ctxt->handle, new_bh);
 
-	ret = ocfs2_journal_dirty(handle, new_bh);
-	if (ret < 0) {
-		mlog_errno(ret);
-		goto end;
-	}
+	/* Add it to the inode */
 	di->i_xattr_loc = cpu_to_le64(first_blkno);
-	ocfs2_journal_dirty(handle, inode_bh);
+
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+	OCFS2_I(inode)->ip_dyn_features |= OCFS2_HAS_XATTR_FL;
+	di->i_dyn_features = cpu_to_le16(OCFS2_I(inode)->ip_dyn_features);
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+
+	ocfs2_journal_dirty(ctxt->handle, inode_bh);
 
 	*ret_bh = new_bh;
 	new_bh = NULL;
@@ -2326,13 +2936,13 @@ static int ocfs2_xattr_block_set(struct inode *inode,
 				 struct ocfs2_xattr_set_ctxt *ctxt)
 {
 	struct buffer_head *new_bh = NULL;
-	handle_t *handle = ctxt->handle;
 	struct ocfs2_xattr_block *xblk = NULL;
 	int ret;
+	struct ocfs2_xa_loc loc;
 
 	if (!xs->xattr_bh) {
-		ret = ocfs2_create_xattr_block(handle, inode, xs->inode_bh,
-					       ctxt->meta_ac, &new_bh, 0);
+		ret = ocfs2_create_xattr_block(inode, xs->inode_bh, ctxt,
+					       0, &new_bh);
 		if (ret) {
 			mlog_errno(ret);
 			goto end;
@@ -2348,21 +2958,25 @@ static int ocfs2_xattr_block_set(struct inode *inode,
 		xblk = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data;
 
 	if (!(le16_to_cpu(xblk->xb_flags) & OCFS2_XATTR_INDEXED)) {
-		/* Set extended attribute into external block */
-		ret = ocfs2_xattr_set_entry(inode, xi, xs, ctxt,
-					    OCFS2_HAS_XATTR_FL);
-		if (!ret || ret != -ENOSPC)
-			goto end;
+		ocfs2_init_xattr_block_xa_loc(&loc, inode, xs->xattr_bh,
+					      xs->not_found ? NULL : xs->here);
 
-		ret = ocfs2_xattr_create_index_block(inode, xs, ctxt);
-		if (ret)
+		ret = ocfs2_xa_set(&loc, xi, ctxt);
+		if (!ret)
+			xs->here = loc.xl_entry;
+		else if ((ret != -ENOSPC) || ctxt->set_abort)
 			goto end;
+		else {
+			ret = ocfs2_xattr_create_index_block(inode, xs, ctxt);
+			if (ret)
+				goto end;
+		}
 	}
 
-	ret = ocfs2_xattr_set_entry_index_block(inode, xi, xs, ctxt);
+	if (le16_to_cpu(xblk->xb_flags) & OCFS2_XATTR_INDEXED)
+		ret = ocfs2_xattr_set_entry_index_block(inode, xi, xs, ctxt);
 
 end:
-
 	return ret;
 }
 
@@ -2371,7 +2985,6 @@ static int ocfs2_xattr_can_be_in_inode(struct inode *inode,
 				       struct ocfs2_xattr_info *xi,
 				       struct ocfs2_xattr_search *xs)
 {
-	u64 value_size;
 	struct ocfs2_xattr_entry *last;
 	int free, i;
 	size_t min_offs = xs->end - xs->base;
@@ -2394,13 +3007,7 @@ static int ocfs2_xattr_can_be_in_inode(struct inode *inode,
 
 	BUG_ON(!xs->not_found);
 
-	if (xi->value_len > OCFS2_XATTR_INLINE_SIZE)
-		value_size = OCFS2_XATTR_ROOT_SIZE;
-	else
-		value_size = OCFS2_XATTR_SIZE(xi->value_len);
-
-	if (free >= sizeof(struct ocfs2_xattr_entry) +
-		   OCFS2_XATTR_SIZE(strlen(xi->name)) + value_size)
+	if (free >= (sizeof(struct ocfs2_xattr_entry) + namevalue_size_xi(xi)))
 		return 1;
 
 	return 0;
@@ -2424,7 +3031,7 @@ static int ocfs2_calc_xattr_set_need(struct inode *inode,
 	char *base = NULL;
 	int name_offset, name_len = 0;
 	u32 new_clusters = ocfs2_clusters_for_bytes(inode->i_sb,
-						    xi->value_len);
+						    xi->xi_value_len);
 	u64 value_size;
 
 	/*
@@ -2432,14 +3039,14 @@ static int ocfs2_calc_xattr_set_need(struct inode *inode,
 	 * No matter whether we replace an old one or add a new one,
 	 * we need this for writing.
 	 */
-	if (xi->value_len > OCFS2_XATTR_INLINE_SIZE)
+	if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE)
 		credits += new_clusters *
 			   ocfs2_clusters_to_blocks(inode->i_sb, 1);
 
 	if (xis->not_found && xbs->not_found) {
 		credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb);
 
-		if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) {
+		if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) {
 			clusters_add += new_clusters;
 			credits += ocfs2_calc_extend_credits(inode->i_sb,
 							&def_xv.xv.xr_list,
@@ -2484,7 +3091,7 @@ static int ocfs2_calc_xattr_set_need(struct inode *inode,
 	 * The credits for removing the value tree will be extended
 	 * by ocfs2_remove_extent itself.
 	 */
-	if (!xi->value) {
+	if (!xi->xi_value) {
 		if (!ocfs2_xattr_is_local(xe))
 			credits += ocfs2_remove_extent_credits(inode->i_sb);
 
@@ -2514,7 +3121,7 @@ static int ocfs2_calc_xattr_set_need(struct inode *inode,
 		}
 	}
 
-	if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) {
+	if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) {
 		/* the new values will be stored outside. */
 		u32 old_clusters = 0;
 
@@ -2547,9 +3154,10 @@ static int ocfs2_calc_xattr_set_need(struct inode *inode,
 		 * value, we don't need any allocation, otherwise we have
 		 * to guess metadata allocation.
 		 */
-		if ((ocfs2_xattr_is_local(xe) && value_size >= xi->value_len) ||
+		if ((ocfs2_xattr_is_local(xe) &&
+		     (value_size >= xi->xi_value_len)) ||
 		    (!ocfs2_xattr_is_local(xe) &&
-		     OCFS2_XATTR_ROOT_SIZE >= xi->value_len))
+		     OCFS2_XATTR_ROOT_SIZE >= xi->xi_value_len))
 			goto out;
 	}
 
@@ -2639,7 +3247,7 @@ static int ocfs2_init_xattr_set_ctxt(struct inode *inode,
 
 	meta_add += extra_meta;
 	mlog(0, "Set xattr %s, reserve meta blocks = %d, clusters = %d, "
-	     "credits = %d\n", xi->name, meta_add, clusters_add, *credits);
+	     "credits = %d\n", xi->xi_name, meta_add, clusters_add, *credits);
 
 	if (meta_add) {
 		ret = ocfs2_reserve_new_metadata_blocks(osb, meta_add,
@@ -2679,7 +3287,7 @@ static int __ocfs2_xattr_set_handle(struct inode *inode,
 {
 	int ret = 0, credits, old_found;
 
-	if (!xi->value) {
+	if (!xi->xi_value) {
 		/* Remove existing extended attribute */
 		if (!xis->not_found)
 			ret = ocfs2_xattr_ibody_set(inode, xi, xis, ctxt);
@@ -2693,8 +3301,8 @@ static int __ocfs2_xattr_set_handle(struct inode *inode,
 			 * If succeed and that extended attribute existing in
 			 * external block, then we will remove it.
 			 */
-			xi->value = NULL;
-			xi->value_len = 0;
+			xi->xi_value = NULL;
+			xi->xi_value_len = 0;
 
 			old_found = xis->not_found;
 			xis->not_found = -ENODATA;
@@ -2712,18 +3320,17 @@ static int __ocfs2_xattr_set_handle(struct inode *inode,
 				goto out;
 			}
 
-			ret = ocfs2_extend_trans(ctxt->handle, credits +
-					ctxt->handle->h_buffer_credits);
+			ret = ocfs2_extend_trans(ctxt->handle, credits);
 			if (ret) {
 				mlog_errno(ret);
 				goto out;
 			}
 			ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt);
-		} else if (ret == -ENOSPC) {
+		} else if ((ret == -ENOSPC) && !ctxt->set_abort) {
 			if (di->i_xattr_loc && !xbs->xattr_bh) {
 				ret = ocfs2_xattr_block_find(inode,
-							     xi->name_index,
-							     xi->name, xbs);
+							     xi->xi_name_index,
+							     xi->xi_name, xbs);
 				if (ret)
 					goto out;
 
@@ -2743,8 +3350,7 @@ static int __ocfs2_xattr_set_handle(struct inode *inode,
 					goto out;
 				}
 
-				ret = ocfs2_extend_trans(ctxt->handle, credits +
-					ctxt->handle->h_buffer_credits);
+				ret = ocfs2_extend_trans(ctxt->handle, credits);
 				if (ret) {
 					mlog_errno(ret);
 					goto out;
@@ -2762,8 +3368,8 @@ static int __ocfs2_xattr_set_handle(struct inode *inode,
 				 * If succeed and that extended attribute
 				 * existing in inode, we will remove it.
 				 */
-				xi->value = NULL;
-				xi->value_len = 0;
+				xi->xi_value = NULL;
+				xi->xi_value_len = 0;
 				xbs->not_found = -ENODATA;
 				ret = ocfs2_calc_xattr_set_need(inode,
 								di,
@@ -2778,8 +3384,7 @@ static int __ocfs2_xattr_set_handle(struct inode *inode,
 					goto out;
 				}
 
-				ret = ocfs2_extend_trans(ctxt->handle, credits +
-						ctxt->handle->h_buffer_credits);
+				ret = ocfs2_extend_trans(ctxt->handle, credits);
 				if (ret) {
 					mlog_errno(ret);
 					goto out;
@@ -2829,10 +3434,11 @@ int ocfs2_xattr_set_handle(handle_t *handle,
 	int ret;
 
 	struct ocfs2_xattr_info xi = {
-		.name_index = name_index,
-		.name = name,
-		.value = value,
-		.value_len = value_len,
+		.xi_name_index = name_index,
+		.xi_name = name,
+		.xi_name_len = strlen(name),
+		.xi_value = value,
+		.xi_value_len = value_len,
 	};
 
 	struct ocfs2_xattr_search xis = {
@@ -2912,10 +3518,11 @@ int ocfs2_xattr_set(struct inode *inode,
 	struct ocfs2_refcount_tree *ref_tree = NULL;
 
 	struct ocfs2_xattr_info xi = {
-		.name_index = name_index,
-		.name = name,
-		.value = value,
-		.value_len = value_len,
+		.xi_name_index = name_index,
+		.xi_name = name,
+		.xi_name_len = strlen(name),
+		.xi_value = value,
+		.xi_value_len = value_len,
 	};
 
 	struct ocfs2_xattr_search xis = {
@@ -3647,7 +4254,6 @@ static int ocfs2_xattr_create_index_block(struct inode *inode,
 	u32 bit_off, len;
 	u64 blkno;
 	handle_t *handle = ctxt->handle;
-	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 	struct ocfs2_inode_info *oi = OCFS2_I(inode);
 	struct buffer_head *xb_bh = xs->xattr_bh;
 	struct ocfs2_xattr_block *xb =
@@ -3675,7 +4281,7 @@ static int ocfs2_xattr_create_index_block(struct inode *inode,
 		goto out;
 	}
 
-	ret = __ocfs2_claim_clusters(osb, handle, ctxt->data_ac,
+	ret = __ocfs2_claim_clusters(handle, ctxt->data_ac,
 				     1, 1, &bit_off, &len);
 	if (ret) {
 		mlog_errno(ret);
@@ -3759,7 +4365,7 @@ static int ocfs2_defrag_xattr_bucket(struct inode *inode,
 				     struct ocfs2_xattr_bucket *bucket)
 {
 	int ret, i;
-	size_t end, offset, len, value_len;
+	size_t end, offset, len;
 	struct ocfs2_xattr_header *xh;
 	char *entries, *buf, *bucket_buf = NULL;
 	u64 blkno = bucket_blkno(bucket);
@@ -3813,12 +4419,7 @@ static int ocfs2_defrag_xattr_bucket(struct inode *inode,
 	end = OCFS2_XATTR_BUCKET_SIZE;
 	for (i = 0; i < le16_to_cpu(xh->xh_count); i++, xe++) {
 		offset = le16_to_cpu(xe->xe_name_offset);
-		if (ocfs2_xattr_is_local(xe))
-			value_len = OCFS2_XATTR_SIZE(
-					le64_to_cpu(xe->xe_value_size));
-		else
-			value_len = OCFS2_XATTR_ROOT_SIZE;
-		len = OCFS2_XATTR_SIZE(xe->xe_name_len) + value_len;
+		len = namevalue_size_xe(xe);
 
 		/*
 		 * We must make sure that the name/value pair
@@ -4007,7 +4608,7 @@ static int ocfs2_divide_xattr_bucket(struct inode *inode,
 				    int new_bucket_head)
 {
 	int ret, i;
-	int count, start, len, name_value_len = 0, xe_len, name_offset = 0;
+	int count, start, len, name_value_len = 0, name_offset = 0;
 	struct ocfs2_xattr_bucket *s_bucket = NULL, *t_bucket = NULL;
 	struct ocfs2_xattr_header *xh;
 	struct ocfs2_xattr_entry *xe;
@@ -4098,13 +4699,7 @@ static int ocfs2_divide_xattr_bucket(struct inode *inode,
 	name_value_len = 0;
 	for (i = 0; i < start; i++) {
 		xe = &xh->xh_entries[i];
-		xe_len = OCFS2_XATTR_SIZE(xe->xe_name_len);
-		if (ocfs2_xattr_is_local(xe))
-			xe_len +=
-			   OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size));
-		else
-			xe_len += OCFS2_XATTR_ROOT_SIZE;
-		name_value_len += xe_len;
+		name_value_len += namevalue_size_xe(xe);
 		if (le16_to_cpu(xe->xe_name_offset) < name_offset)
 			name_offset = le16_to_cpu(xe->xe_name_offset);
 	}
@@ -4134,12 +4729,6 @@ static int ocfs2_divide_xattr_bucket(struct inode *inode,
 	xh->xh_free_start = cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE);
 	for (i = 0; i < le16_to_cpu(xh->xh_count); i++) {
 		xe = &xh->xh_entries[i];
-		xe_len = OCFS2_XATTR_SIZE(xe->xe_name_len);
-		if (ocfs2_xattr_is_local(xe))
-			xe_len +=
-			   OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size));
-		else
-			xe_len += OCFS2_XATTR_ROOT_SIZE;
 		if (le16_to_cpu(xe->xe_name_offset) <
 		    le16_to_cpu(xh->xh_free_start))
 			xh->xh_free_start = xe->xe_name_offset;
@@ -4302,8 +4891,7 @@ static int ocfs2_mv_xattr_buckets(struct inode *inode, handle_t *handle,
 	 * We need to update the first bucket of the old extent and all
 	 * the buckets going to the new extent.
 	 */
-	credits = ((num_buckets + 1) * blks_per_bucket) +
-		handle->h_buffer_credits;
+	credits = ((num_buckets + 1) * blks_per_bucket);
 	ret = ocfs2_extend_trans(handle, credits);
 	if (ret) {
 		mlog_errno(ret);
@@ -4373,7 +4961,7 @@ static int ocfs2_divide_xattr_cluster(struct inode *inode,
 				      u32 *first_hash)
 {
 	u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb);
-	int ret, credits = 2 * blk_per_bucket + handle->h_buffer_credits;
+	int ret, credits = 2 * blk_per_bucket;
 
 	BUG_ON(OCFS2_XATTR_BUCKET_SIZE < OCFS2_SB(inode->i_sb)->s_clustersize);
 
@@ -4514,7 +5102,7 @@ static int ocfs2_add_new_xattr_cluster(struct inode *inode,
 		goto leave;
 	}
 
-	ret = __ocfs2_claim_clusters(osb, handle, ctxt->data_ac, 1,
+	ret = __ocfs2_claim_clusters(handle, ctxt->data_ac, 1,
 				     clusters_to_add, &bit_off, &num_bits);
 	if (ret < 0) {
 		if (ret != -ENOSPC)
@@ -4568,9 +5156,7 @@ static int ocfs2_add_new_xattr_cluster(struct inode *inode,
 		goto leave;
 	}
 
-	ret = ocfs2_journal_dirty(handle, root_bh);
-	if (ret < 0)
-		mlog_errno(ret);
+	ocfs2_journal_dirty(handle, root_bh);
 
 leave:
 	return ret;
@@ -4615,8 +5201,7 @@ static int ocfs2_extend_xattr_bucket(struct inode *inode,
 	 * existing bucket.  Then we add the last existing bucket, the
 	 * new bucket, and the first bucket (3 * blk_per_bucket).
 	 */
-	credits = (end_blk - target_blk) + (3 * blk_per_bucket) +
-		  handle->h_buffer_credits;
+	credits = (end_blk - target_blk) + (3 * blk_per_bucket);
 	ret = ocfs2_extend_trans(handle, credits);
 	if (ret) {
 		mlog_errno(ret);
@@ -4751,195 +5336,6 @@ static inline char *ocfs2_xattr_bucket_get_val(struct inode *inode,
 }
 
 /*
- * Handle the normal xattr set, including replace, delete and new.
- *
- * Note: "local" indicates the real data's locality. So we can't
- * just its bucket locality by its length.
- */
-static void ocfs2_xattr_set_entry_normal(struct inode *inode,
-					 struct ocfs2_xattr_info *xi,
-					 struct ocfs2_xattr_search *xs,
-					 u32 name_hash,
-					 int local)
-{
-	struct ocfs2_xattr_entry *last, *xe;
-	int name_len = strlen(xi->name);
-	struct ocfs2_xattr_header *xh = xs->header;
-	u16 count = le16_to_cpu(xh->xh_count), start;
-	size_t blocksize = inode->i_sb->s_blocksize;
-	char *val;
-	size_t offs, size, new_size;
-
-	last = &xh->xh_entries[count];
-	if (!xs->not_found) {
-		xe = xs->here;
-		offs = le16_to_cpu(xe->xe_name_offset);
-		if (ocfs2_xattr_is_local(xe))
-			size = OCFS2_XATTR_SIZE(name_len) +
-			OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size));
-		else
-			size = OCFS2_XATTR_SIZE(name_len) +
-			OCFS2_XATTR_SIZE(OCFS2_XATTR_ROOT_SIZE);
-
-		/*
-		 * If the new value will be stored outside, xi->value has been
-		 * initalized as an empty ocfs2_xattr_value_root, and the same
-		 * goes with xi->value_len, so we can set new_size safely here.
-		 * See ocfs2_xattr_set_in_bucket.
-		 */
-		new_size = OCFS2_XATTR_SIZE(name_len) +
-			   OCFS2_XATTR_SIZE(xi->value_len);
-
-		le16_add_cpu(&xh->xh_name_value_len, -size);
-		if (xi->value) {
-			if (new_size > size)
-				goto set_new_name_value;
-
-			/* Now replace the old value with new one. */
-			if (local)
-				xe->xe_value_size = cpu_to_le64(xi->value_len);
-			else
-				xe->xe_value_size = 0;
-
-			val = ocfs2_xattr_bucket_get_val(inode,
-							 xs->bucket, offs);
-			memset(val + OCFS2_XATTR_SIZE(name_len), 0,
-			       size - OCFS2_XATTR_SIZE(name_len));
-			if (OCFS2_XATTR_SIZE(xi->value_len) > 0)
-				memcpy(val + OCFS2_XATTR_SIZE(name_len),
-				       xi->value, xi->value_len);
-
-			le16_add_cpu(&xh->xh_name_value_len, new_size);
-			ocfs2_xattr_set_local(xe, local);
-			return;
-		} else {
-			/*
-			 * Remove the old entry if there is more than one.
-			 * We don't remove the last entry so that we can
-			 * use it to indicate the hash value of the empty
-			 * bucket.
-			 */
-			last -= 1;
-			le16_add_cpu(&xh->xh_count, -1);
-			if (xh->xh_count) {
-				memmove(xe, xe + 1,
-					(void *)last - (void *)xe);
-				memset(last, 0,
-				       sizeof(struct ocfs2_xattr_entry));
-			} else
-				xh->xh_free_start =
-					cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE);
-
-			return;
-		}
-	} else {
-		/* find a new entry for insert. */
-		int low = 0, high = count - 1, tmp;
-		struct ocfs2_xattr_entry *tmp_xe;
-
-		while (low <= high && count) {
-			tmp = (low + high) / 2;
-			tmp_xe = &xh->xh_entries[tmp];
-
-			if (name_hash > le32_to_cpu(tmp_xe->xe_name_hash))
-				low = tmp + 1;
-			else if (name_hash <
-				 le32_to_cpu(tmp_xe->xe_name_hash))
-				high = tmp - 1;
-			else {
-				low = tmp;
-				break;
-			}
-		}
-
-		xe = &xh->xh_entries[low];
-		if (low != count)
-			memmove(xe + 1, xe, (void *)last - (void *)xe);
-
-		le16_add_cpu(&xh->xh_count, 1);
-		memset(xe, 0, sizeof(struct ocfs2_xattr_entry));
-		xe->xe_name_hash = cpu_to_le32(name_hash);
-		xe->xe_name_len = name_len;
-		ocfs2_xattr_set_type(xe, xi->name_index);
-	}
-
-set_new_name_value:
-	/* Insert the new name+value. */
-	size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_SIZE(xi->value_len);
-
-	/*
-	 * We must make sure that the name/value pair
-	 * exists in the same block.
-	 */
-	offs = le16_to_cpu(xh->xh_free_start);
-	start = offs - size;
-
-	if (start >> inode->i_sb->s_blocksize_bits !=
-	    (offs - 1) >> inode->i_sb->s_blocksize_bits) {
-		offs = offs - offs % blocksize;
-		xh->xh_free_start = cpu_to_le16(offs);
-	}
-
-	val = ocfs2_xattr_bucket_get_val(inode, xs->bucket, offs - size);
-	xe->xe_name_offset = cpu_to_le16(offs - size);
-
-	memset(val, 0, size);
-	memcpy(val, xi->name, name_len);
-	memcpy(val + OCFS2_XATTR_SIZE(name_len), xi->value, xi->value_len);
-
-	xe->xe_value_size = cpu_to_le64(xi->value_len);
-	ocfs2_xattr_set_local(xe, local);
-	xs->here = xe;
-	le16_add_cpu(&xh->xh_free_start, -size);
-	le16_add_cpu(&xh->xh_name_value_len, size);
-
-	return;
-}
-
-/*
- * Set the xattr entry in the specified bucket.
- * The bucket is indicated by xs->bucket and it should have the enough
- * space for the xattr insertion.
- */
-static int ocfs2_xattr_set_entry_in_bucket(struct inode *inode,
-					   handle_t *handle,
-					   struct ocfs2_xattr_info *xi,
-					   struct ocfs2_xattr_search *xs,
-					   u32 name_hash,
-					   int local)
-{
-	int ret;
-	u64 blkno;
-
-	mlog(0, "Set xattr entry len = %lu index = %d in bucket %llu\n",
-	     (unsigned long)xi->value_len, xi->name_index,
-	     (unsigned long long)bucket_blkno(xs->bucket));
-
-	if (!xs->bucket->bu_bhs[1]) {
-		blkno = bucket_blkno(xs->bucket);
-		ocfs2_xattr_bucket_relse(xs->bucket);
-		ret = ocfs2_read_xattr_bucket(xs->bucket, blkno);
-		if (ret) {
-			mlog_errno(ret);
-			goto out;
-		}
-	}
-
-	ret = ocfs2_xattr_bucket_journal_access(handle, xs->bucket,
-						OCFS2_JOURNAL_ACCESS_WRITE);
-	if (ret < 0) {
-		mlog_errno(ret);
-		goto out;
-	}
-
-	ocfs2_xattr_set_entry_normal(inode, xi, xs, name_hash, local);
-	ocfs2_xattr_bucket_journal_dirty(handle, xs->bucket);
-
-out:
-	return ret;
-}
-
-/*
  * Truncate the specified xe_off entry in xattr bucket.
  * bucket is indicated by header_bh and len is the new length.
  * Both the ocfs2_xattr_value_root and the entry will be updated here.
@@ -5009,66 +5405,6 @@ out:
 	return ret;
 }
 
-static int ocfs2_xattr_bucket_value_truncate_xs(struct inode *inode,
-					struct ocfs2_xattr_search *xs,
-					int len,
-					struct ocfs2_xattr_set_ctxt *ctxt)
-{
-	int ret, offset;
-	struct ocfs2_xattr_entry *xe = xs->here;
-	struct ocfs2_xattr_header *xh = (struct ocfs2_xattr_header *)xs->base;
-
-	BUG_ON(!xs->bucket->bu_bhs[0] || !xe || ocfs2_xattr_is_local(xe));
-
-	offset = xe - xh->xh_entries;
-	ret = ocfs2_xattr_bucket_value_truncate(inode, xs->bucket,
-						offset, len, ctxt);
-	if (ret)
-		mlog_errno(ret);
-
-	return ret;
-}
-
-static int ocfs2_xattr_bucket_set_value_outside(struct inode *inode,
-						handle_t *handle,
-						struct ocfs2_xattr_search *xs,
-						char *val,
-						int value_len)
-{
-	int ret, offset, block_off;
-	struct ocfs2_xattr_value_root *xv;
-	struct ocfs2_xattr_entry *xe = xs->here;
-	struct ocfs2_xattr_header *xh = bucket_xh(xs->bucket);
-	void *base;
-	struct ocfs2_xattr_value_buf vb = {
-		.vb_access = ocfs2_journal_access,
-	};
-
-	BUG_ON(!xs->base || !xe || ocfs2_xattr_is_local(xe));
-
-	ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, xh,
-						xe - xh->xh_entries,
-						&block_off,
-						&offset);
-	if (ret) {
-		mlog_errno(ret);
-		goto out;
-	}
-
-	base = bucket_block(xs->bucket, block_off);
-	xv = (struct ocfs2_xattr_value_root *)(base + offset +
-		 OCFS2_XATTR_SIZE(xe->xe_name_len));
-
-	vb.vb_xv = xv;
-	vb.vb_bh = xs->bucket->bu_bhs[block_off];
-	ret = __ocfs2_xattr_set_value_outside(inode, handle,
-					      &vb, val, value_len);
-	if (ret)
-		mlog_errno(ret);
-out:
-	return ret;
-}
-
 static int ocfs2_rm_xattr_cluster(struct inode *inode,
 				  struct buffer_head *root_bh,
 				  u64 blkno,
@@ -5141,12 +5477,7 @@ static int ocfs2_rm_xattr_cluster(struct inode *inode,
 	}
 
 	le32_add_cpu(&xb->xb_attrs.xb_root.xt_clusters, -len);
-
-	ret = ocfs2_journal_dirty(handle, root_bh);
-	if (ret) {
-		mlog_errno(ret);
-		goto out_commit;
-	}
+	ocfs2_journal_dirty(handle, root_bh);
 
 	ret = ocfs2_truncate_log_append(osb, handle, blkno, len);
 	if (ret)
@@ -5167,128 +5498,6 @@ out:
 	return ret;
 }
 
-static void ocfs2_xattr_bucket_remove_xs(struct inode *inode,
-					 handle_t *handle,
-					 struct ocfs2_xattr_search *xs)
-{
-	struct ocfs2_xattr_header *xh = bucket_xh(xs->bucket);
-	struct ocfs2_xattr_entry *last = &xh->xh_entries[
-						le16_to_cpu(xh->xh_count) - 1];
-	int ret = 0;
-
-	ret = ocfs2_xattr_bucket_journal_access(handle, xs->bucket,
-						OCFS2_JOURNAL_ACCESS_WRITE);
-	if (ret) {
-		mlog_errno(ret);
-		return;
-	}
-
-	/* Remove the old entry. */
-	memmove(xs->here, xs->here + 1,
-		(void *)last - (void *)xs->here);
-	memset(last, 0, sizeof(struct ocfs2_xattr_entry));
-	le16_add_cpu(&xh->xh_count, -1);
-
-	ocfs2_xattr_bucket_journal_dirty(handle, xs->bucket);
-}
-
-/*
- * Set the xattr name/value in the bucket specified in xs.
- *
- * As the new value in xi may be stored in the bucket or in an outside cluster,
- * we divide the whole process into 3 steps:
- * 1. insert name/value in the bucket(ocfs2_xattr_set_entry_in_bucket)
- * 2. truncate of the outside cluster(ocfs2_xattr_bucket_value_truncate_xs)
- * 3. Set the value to the outside cluster(ocfs2_xattr_bucket_set_value_outside)
- * 4. If the clusters for the new outside value can't be allocated, we need
- *    to free the xattr we allocated in set.
- */
-static int ocfs2_xattr_set_in_bucket(struct inode *inode,
-				     struct ocfs2_xattr_info *xi,
-				     struct ocfs2_xattr_search *xs,
-				     struct ocfs2_xattr_set_ctxt *ctxt)
-{
-	int ret, local = 1;
-	size_t value_len;
-	char *val = (char *)xi->value;
-	struct ocfs2_xattr_entry *xe = xs->here;
-	u32 name_hash = ocfs2_xattr_name_hash(inode, xi->name,
-					      strlen(xi->name));
-
-	if (!xs->not_found && !ocfs2_xattr_is_local(xe)) {
-		/*
-		 * We need to truncate the xattr storage first.
-		 *
-		 * If both the old and new value are stored to
-		 * outside block, we only need to truncate
-		 * the storage and then set the value outside.
-		 *
-		 * If the new value should be stored within block,
-		 * we should free all the outside block first and
-		 * the modification to the xattr block will be done
-		 * by following steps.
-		 */
-		if (xi->value_len > OCFS2_XATTR_INLINE_SIZE)
-			value_len = xi->value_len;
-		else
-			value_len = 0;
-
-		ret = ocfs2_xattr_bucket_value_truncate_xs(inode, xs,
-							   value_len,
-							   ctxt);
-		if (ret)
-			goto out;
-
-		if (value_len)
-			goto set_value_outside;
-	}
-
-	value_len = xi->value_len;
-	/* So we have to handle the inside block change now. */
-	if (value_len > OCFS2_XATTR_INLINE_SIZE) {
-		/*
-		 * If the new value will be stored outside of block,
-		 * initalize a new empty value root and insert it first.
-		 */
-		local = 0;
-		xi->value = &def_xv;
-		xi->value_len = OCFS2_XATTR_ROOT_SIZE;
-	}
-
-	ret = ocfs2_xattr_set_entry_in_bucket(inode, ctxt->handle, xi, xs,
-					      name_hash, local);
-	if (ret) {
-		mlog_errno(ret);
-		goto out;
-	}
-
-	if (value_len <= OCFS2_XATTR_INLINE_SIZE)
-		goto out;
-
-	/* allocate the space now for the outside block storage. */
-	ret = ocfs2_xattr_bucket_value_truncate_xs(inode, xs,
-						   value_len, ctxt);
-	if (ret) {
-		mlog_errno(ret);
-
-		if (xs->not_found) {
-			/*
-			 * We can't allocate enough clusters for outside
-			 * storage and we have allocated xattr already,
-			 * so need to remove it.
-			 */
-			ocfs2_xattr_bucket_remove_xs(inode, ctxt->handle, xs);
-		}
-		goto out;
-	}
-
-set_value_outside:
-	ret = ocfs2_xattr_bucket_set_value_outside(inode, ctxt->handle,
-						   xs, val, value_len);
-out:
-	return ret;
-}
-
 /*
  * check whether the xattr bucket is filled up with the same hash value.
  * If we want to insert the xattr with the same hash, return -ENOSPC.
@@ -5317,156 +5526,116 @@ static int ocfs2_check_xattr_bucket_collision(struct inode *inode,
 	return 0;
 }
 
-static int ocfs2_xattr_set_entry_index_block(struct inode *inode,
-					     struct ocfs2_xattr_info *xi,
-					     struct ocfs2_xattr_search *xs,
-					     struct ocfs2_xattr_set_ctxt *ctxt)
+/*
+ * Try to set the entry in the current bucket.  If we fail, the caller
+ * will handle getting us another bucket.
+ */
+static int ocfs2_xattr_set_entry_bucket(struct inode *inode,
+					struct ocfs2_xattr_info *xi,
+					struct ocfs2_xattr_search *xs,
+					struct ocfs2_xattr_set_ctxt *ctxt)
 {
-	struct ocfs2_xattr_header *xh;
-	struct ocfs2_xattr_entry *xe;
-	u16 count, header_size, xh_free_start;
-	int free, max_free, need, old;
-	size_t value_size = 0, name_len = strlen(xi->name);
-	size_t blocksize = inode->i_sb->s_blocksize;
-	int ret, allocation = 0;
-
-	mlog_entry("Set xattr %s in xattr index block\n", xi->name);
-
-try_again:
-	xh = xs->header;
-	count = le16_to_cpu(xh->xh_count);
-	xh_free_start = le16_to_cpu(xh->xh_free_start);
-	header_size = sizeof(struct ocfs2_xattr_header) +
-			count * sizeof(struct ocfs2_xattr_entry);
-	max_free = OCFS2_XATTR_BUCKET_SIZE - header_size -
-		le16_to_cpu(xh->xh_name_value_len) - OCFS2_XATTR_HEADER_GAP;
-
-	mlog_bug_on_msg(header_size > blocksize, "bucket %llu has header size "
-			"of %u which exceed block size\n",
-			(unsigned long long)bucket_blkno(xs->bucket),
-			header_size);
+	int ret;
+	struct ocfs2_xa_loc loc;
 
-	if (xi->value && xi->value_len > OCFS2_XATTR_INLINE_SIZE)
-		value_size = OCFS2_XATTR_ROOT_SIZE;
-	else if (xi->value)
-		value_size = OCFS2_XATTR_SIZE(xi->value_len);
+	mlog_entry("Set xattr %s in xattr bucket\n", xi->xi_name);
 
-	if (xs->not_found)
-		need = sizeof(struct ocfs2_xattr_entry) +
-			OCFS2_XATTR_SIZE(name_len) + value_size;
-	else {
-		need = value_size + OCFS2_XATTR_SIZE(name_len);
+	ocfs2_init_xattr_bucket_xa_loc(&loc, xs->bucket,
+				       xs->not_found ? NULL : xs->here);
+	ret = ocfs2_xa_set(&loc, xi, ctxt);
+	if (!ret) {
+		xs->here = loc.xl_entry;
+		goto out;
+	}
+	if (ret != -ENOSPC) {
+		mlog_errno(ret);
+		goto out;
+	}
 
-		/*
-		 * We only replace the old value if the new length is smaller
-		 * than the old one. Otherwise we will allocate new space in the
-		 * bucket to store it.
-		 */
-		xe = xs->here;
-		if (ocfs2_xattr_is_local(xe))
-			old = OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size));
-		else
-			old = OCFS2_XATTR_SIZE(OCFS2_XATTR_ROOT_SIZE);
+	/* Ok, we need space.  Let's try defragmenting the bucket. */
+	ret = ocfs2_defrag_xattr_bucket(inode, ctxt->handle,
+					xs->bucket);
+	if (ret) {
+		mlog_errno(ret);
+		goto out;
+	}
 
-		if (old >= value_size)
-			need = 0;
+	ret = ocfs2_xa_set(&loc, xi, ctxt);
+	if (!ret) {
+		xs->here = loc.xl_entry;
+		goto out;
 	}
+	if (ret != -ENOSPC)
+		mlog_errno(ret);
 
-	free = xh_free_start - header_size - OCFS2_XATTR_HEADER_GAP;
-	/*
-	 * We need to make sure the new name/value pair
-	 * can exist in the same block.
-	 */
-	if (xh_free_start % blocksize < need)
-		free -= xh_free_start % blocksize;
-
-	mlog(0, "xs->not_found = %d, in xattr bucket %llu: free = %d, "
-	     "need = %d, max_free = %d, xh_free_start = %u, xh_name_value_len ="
-	     " %u\n", xs->not_found,
-	     (unsigned long long)bucket_blkno(xs->bucket),
-	     free, need, max_free, le16_to_cpu(xh->xh_free_start),
-	     le16_to_cpu(xh->xh_name_value_len));
-
-	if (free < need ||
-	    (xs->not_found &&
-	     count == ocfs2_xattr_max_xe_in_bucket(inode->i_sb))) {
-		if (need <= max_free &&
-		    count < ocfs2_xattr_max_xe_in_bucket(inode->i_sb)) {
-			/*
-			 * We can create the space by defragment. Since only the
-			 * name/value will be moved, the xe shouldn't be changed
-			 * in xs.
-			 */
-			ret = ocfs2_defrag_xattr_bucket(inode, ctxt->handle,
-							xs->bucket);
-			if (ret) {
-				mlog_errno(ret);
-				goto out;
-			}
 
-			xh_free_start = le16_to_cpu(xh->xh_free_start);
-			free = xh_free_start - header_size
-				- OCFS2_XATTR_HEADER_GAP;
-			if (xh_free_start % blocksize < need)
-				free -= xh_free_start % blocksize;
+out:
+	mlog_exit(ret);
+	return ret;
+}
 
-			if (free >= need)
-				goto xattr_set;
+static int ocfs2_xattr_set_entry_index_block(struct inode *inode,
+					     struct ocfs2_xattr_info *xi,
+					     struct ocfs2_xattr_search *xs,
+					     struct ocfs2_xattr_set_ctxt *ctxt)
+{
+	int ret;
 
-			mlog(0, "Can't get enough space for xattr insert by "
-			     "defragment. Need %u bytes, but we have %d, so "
-			     "allocate new bucket for it.\n", need, free);
-		}
+	mlog_entry("Set xattr %s in xattr index block\n", xi->xi_name);
 
-		/*
-		 * We have to add new buckets or clusters and one
-		 * allocation should leave us enough space for insert.
-		 */
-		BUG_ON(allocation);
+	ret = ocfs2_xattr_set_entry_bucket(inode, xi, xs, ctxt);
+	if (!ret)
+		goto out;
+	if (ret != -ENOSPC) {
+		mlog_errno(ret);
+		goto out;
+	}
 
-		/*
-		 * We do not allow for overlapping ranges between buckets. And
-		 * the maximum number of collisions we will allow for then is
-		 * one bucket's worth, so check it here whether we need to
-		 * add a new bucket for the insert.
-		 */
-		ret = ocfs2_check_xattr_bucket_collision(inode,
-							 xs->bucket,
-							 xi->name);
-		if (ret) {
-			mlog_errno(ret);
-			goto out;
-		}
+	/* Ack, need more space.  Let's try to get another bucket! */
 
-		ret = ocfs2_add_new_xattr_bucket(inode,
-						 xs->xattr_bh,
+	/*
+	 * We do not allow for overlapping ranges between buckets. And
+	 * the maximum number of collisions we will allow for then is
+	 * one bucket's worth, so check it here whether we need to
+	 * add a new bucket for the insert.
+	 */
+	ret = ocfs2_check_xattr_bucket_collision(inode,
 						 xs->bucket,
-						 ctxt);
-		if (ret) {
-			mlog_errno(ret);
-			goto out;
-		}
+						 xi->xi_name);
+	if (ret) {
+		mlog_errno(ret);
+		goto out;
+	}
 
-		/*
-		 * ocfs2_add_new_xattr_bucket() will have updated
-		 * xs->bucket if it moved, but it will not have updated
-		 * any of the other search fields.  Thus, we drop it and
-		 * re-search.  Everything should be cached, so it'll be
-		 * quick.
-		 */
-		ocfs2_xattr_bucket_relse(xs->bucket);
-		ret = ocfs2_xattr_index_block_find(inode, xs->xattr_bh,
-						   xi->name_index,
-						   xi->name, xs);
-		if (ret && ret != -ENODATA)
-			goto out;
-		xs->not_found = ret;
-		allocation = 1;
-		goto try_again;
+	ret = ocfs2_add_new_xattr_bucket(inode,
+					 xs->xattr_bh,
+					 xs->bucket,
+					 ctxt);
+	if (ret) {
+		mlog_errno(ret);
+		goto out;
 	}
 
-xattr_set:
-	ret = ocfs2_xattr_set_in_bucket(inode, xi, xs, ctxt);
+	/*
+	 * ocfs2_add_new_xattr_bucket() will have updated
+	 * xs->bucket if it moved, but it will not have updated
+	 * any of the other search fields.  Thus, we drop it and
+	 * re-search.  Everything should be cached, so it'll be
+	 * quick.
+	 */
+	ocfs2_xattr_bucket_relse(xs->bucket);
+	ret = ocfs2_xattr_index_block_find(inode, xs->xattr_bh,
+					   xi->xi_name_index,
+					   xi->xi_name, xs);
+	if (ret && ret != -ENODATA)
+		goto out;
+	xs->not_found = ret;
+
+	/* Ok, we have a new bucket, let's try again */
+	ret = ocfs2_xattr_set_entry_bucket(inode, xi, xs, ctxt);
+	if (ret && (ret != -ENOSPC))
+		mlog_errno(ret);
+
 out:
 	mlog_exit(ret);
 	return ret;
@@ -5678,7 +5847,7 @@ static int ocfs2_prepare_refcount_xattr(struct inode *inode,
 	 * refcount tree, and make the original extent become 3. So we will need
 	 * 2 * cluster more extent recs at most.
 	 */
-	if (!xi->value || xi->value_len <= OCFS2_XATTR_INLINE_SIZE) {
+	if (!xi->xi_value || xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE) {
 
 		ret = ocfs2_refcounted_xattr_delete_need(inode,
 							 &(*ref_tree)->rf_ci,
@@ -6354,33 +6523,33 @@ static int ocfs2_create_empty_xattr_block(struct inode *inode,
 					  int indexed)
 {
 	int ret;
-	handle_t *handle;
-	struct ocfs2_alloc_context *meta_ac;
 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	struct ocfs2_xattr_set_ctxt ctxt;
 
-	ret = ocfs2_reserve_new_metadata_blocks(osb, 1, &meta_ac);
+	memset(&ctxt, 0, sizeof(ctxt));
+	ret = ocfs2_reserve_new_metadata_blocks(osb, 1, &ctxt.meta_ac);
 	if (ret < 0) {
 		mlog_errno(ret);
 		return ret;
 	}
 
-	handle = ocfs2_start_trans(osb, OCFS2_XATTR_BLOCK_CREATE_CREDITS);
-	if (IS_ERR(handle)) {
-		ret = PTR_ERR(handle);
+	ctxt.handle = ocfs2_start_trans(osb, OCFS2_XATTR_BLOCK_CREATE_CREDITS);
+	if (IS_ERR(ctxt.handle)) {
+		ret = PTR_ERR(ctxt.handle);
 		mlog_errno(ret);
 		goto out;
 	}
 
 	mlog(0, "create new xattr block for inode %llu, index = %d\n",
 	     (unsigned long long)fe_bh->b_blocknr, indexed);
-	ret = ocfs2_create_xattr_block(handle, inode, fe_bh,
-				       meta_ac, ret_bh, indexed);
+	ret = ocfs2_create_xattr_block(inode, fe_bh, &ctxt, indexed,
+				       ret_bh);
 	if (ret)
 		mlog_errno(ret);
 
-	ocfs2_commit_trans(osb, handle);
+	ocfs2_commit_trans(osb, ctxt.handle);
 out:
-	ocfs2_free_alloc_context(meta_ac);
+	ocfs2_free_alloc_context(ctxt.meta_ac);
 	return ret;
 }
 
@@ -6635,16 +6804,15 @@ out:
 	return ret;
 }
 
-static int ocfs2_reflink_xattr_buckets(handle_t *handle,
+static int ocfs2_reflink_xattr_bucket(handle_t *handle,
 				u64 blkno, u64 new_blkno, u32 clusters,
+				u32 *cpos, int num_buckets,
 				struct ocfs2_alloc_context *meta_ac,
 				struct ocfs2_alloc_context *data_ac,
 				struct ocfs2_reflink_xattr_tree_args *args)
 {
 	int i, j, ret = 0;
 	struct super_block *sb = args->reflink->old_inode->i_sb;
-	u32 bpc = ocfs2_xattr_buckets_per_cluster(OCFS2_SB(sb));
-	u32 num_buckets = clusters * bpc;
 	int bpb = args->old_bucket->bu_blocks;
 	struct ocfs2_xattr_value_buf vb = {
 		.vb_access = ocfs2_journal_access,
@@ -6663,14 +6831,6 @@ static int ocfs2_reflink_xattr_buckets(handle_t *handle,
 			break;
 		}
 
-		/*
-		 * The real bucket num in this series of blocks is stored
-		 * in the 1st bucket.
-		 */
-		if (i == 0)
-			num_buckets = le16_to_cpu(
-				bucket_xh(args->old_bucket)->xh_num_buckets);
-
 		ret = ocfs2_xattr_bucket_journal_access(handle,
 						args->new_bucket,
 						OCFS2_JOURNAL_ACCESS_CREATE);
@@ -6684,6 +6844,18 @@ static int ocfs2_reflink_xattr_buckets(handle_t *handle,
 			       bucket_block(args->old_bucket, j),
 			       sb->s_blocksize);
 
+		/*
+		 * Record the start cpos so that we can use it to initialize
+		 * our xattr tree we also set the xh_num_bucket for the new
+		 * bucket.
+		 */
+		if (i == 0) {
+			*cpos = le32_to_cpu(bucket_xh(args->new_bucket)->
+					    xh_entries[0].xe_name_hash);
+			bucket_xh(args->new_bucket)->xh_num_buckets =
+				cpu_to_le16(num_buckets);
+		}
+
 		ocfs2_xattr_bucket_journal_dirty(handle, args->new_bucket);
 
 		ret = ocfs2_reflink_xattr_header(handle, args->reflink,
@@ -6713,6 +6885,7 @@ static int ocfs2_reflink_xattr_buckets(handle_t *handle,
 		}
 
 		ocfs2_xattr_bucket_journal_dirty(handle, args->new_bucket);
+
 		ocfs2_xattr_bucket_relse(args->old_bucket);
 		ocfs2_xattr_bucket_relse(args->new_bucket);
 	}
@@ -6721,6 +6894,75 @@ static int ocfs2_reflink_xattr_buckets(handle_t *handle,
 	ocfs2_xattr_bucket_relse(args->new_bucket);
 	return ret;
 }
+
+static int ocfs2_reflink_xattr_buckets(handle_t *handle,
+				struct inode *inode,
+				struct ocfs2_reflink_xattr_tree_args *args,
+				struct ocfs2_extent_tree *et,
+				struct ocfs2_alloc_context *meta_ac,
+				struct ocfs2_alloc_context *data_ac,
+				u64 blkno, u32 cpos, u32 len)
+{
+	int ret, first_inserted = 0;
+	u32 p_cluster, num_clusters, reflink_cpos = 0;
+	u64 new_blkno;
+	unsigned int num_buckets, reflink_buckets;
+	unsigned int bpc =
+		ocfs2_xattr_buckets_per_cluster(OCFS2_SB(inode->i_sb));
+
+	ret = ocfs2_read_xattr_bucket(args->old_bucket, blkno);
+	if (ret) {
+		mlog_errno(ret);
+		goto out;
+	}
+	num_buckets = le16_to_cpu(bucket_xh(args->old_bucket)->xh_num_buckets);
+	ocfs2_xattr_bucket_relse(args->old_bucket);
+
+	while (len && num_buckets) {
+		ret = ocfs2_claim_clusters(handle, data_ac,
+					   1, &p_cluster, &num_clusters);
+		if (ret) {
+			mlog_errno(ret);
+			goto out;
+		}
+
+		new_blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster);
+		reflink_buckets = min(num_buckets, bpc * num_clusters);
+
+		ret = ocfs2_reflink_xattr_bucket(handle, blkno,
+						 new_blkno, num_clusters,
+						 &reflink_cpos, reflink_buckets,
+						 meta_ac, data_ac, args);
+		if (ret) {
+			mlog_errno(ret);
+			goto out;
+		}
+
+		/*
+		 * For the 1st allocated cluster, we make it use the same cpos
+		 * so that the xattr tree looks the same as the original one
+		 * in the most case.
+		 */
+		if (!first_inserted) {
+			reflink_cpos = cpos;
+			first_inserted = 1;
+		}
+		ret = ocfs2_insert_extent(handle, et, reflink_cpos, new_blkno,
+					  num_clusters, 0, meta_ac);
+		if (ret)
+			mlog_errno(ret);
+
+		mlog(0, "insert new xattr extent rec start %llu len %u to %u\n",
+		     (unsigned long long)new_blkno, num_clusters, reflink_cpos);
+
+		len -= num_clusters;
+		blkno += ocfs2_clusters_to_blocks(inode->i_sb, num_clusters);
+		num_buckets -= reflink_buckets;
+	}
+out:
+	return ret;
+}
+
 /*
  * Create the same xattr extent record in the new inode's xattr tree.
  */
@@ -6732,8 +6974,6 @@ static int ocfs2_reflink_xattr_rec(struct inode *inode,
 				   void *para)
 {
 	int ret, credits = 0;
-	u32 p_cluster, num_clusters;
-	u64 new_blkno;
 	handle_t *handle;
 	struct ocfs2_reflink_xattr_tree_args *args =
 			(struct ocfs2_reflink_xattr_tree_args *)para;
@@ -6742,6 +6982,9 @@ static int ocfs2_reflink_xattr_rec(struct inode *inode,
 	struct ocfs2_alloc_context *data_ac = NULL;
 	struct ocfs2_extent_tree et;
 
+	mlog(0, "reflink xattr buckets %llu len %u\n",
+	     (unsigned long long)blkno, len);
+
 	ocfs2_init_xattr_tree_extent_tree(&et,
 					  INODE_CACHE(args->reflink->new_inode),
 					  args->new_blk_bh);
@@ -6761,32 +7004,12 @@ static int ocfs2_reflink_xattr_rec(struct inode *inode,
 		goto out;
 	}
 
-	ret = ocfs2_claim_clusters(osb, handle, data_ac,
-				   len, &p_cluster, &num_clusters);
-	if (ret) {
-		mlog_errno(ret);
-		goto out_commit;
-	}
-
-	new_blkno = ocfs2_clusters_to_blocks(osb->sb, p_cluster);
-
-	mlog(0, "reflink xattr buckets %llu to %llu, len %u\n",
-	     (unsigned long long)blkno, (unsigned long long)new_blkno, len);
-	ret = ocfs2_reflink_xattr_buckets(handle, blkno, new_blkno, len,
-					  meta_ac, data_ac, args);
-	if (ret) {
-		mlog_errno(ret);
-		goto out_commit;
-	}
-
-	mlog(0, "insert new xattr extent rec start %llu len %u to %u\n",
-	     (unsigned long long)new_blkno, len, cpos);
-	ret = ocfs2_insert_extent(handle, &et, cpos, new_blkno,
-				  len, 0, meta_ac);
+	ret = ocfs2_reflink_xattr_buckets(handle, inode, args, &et,
+					  meta_ac, data_ac,
+					  blkno, cpos, len);
 	if (ret)
 		mlog_errno(ret);
 
-out_commit:
 	ocfs2_commit_trans(osb, handle);
 
 out:
@@ -7060,7 +7283,7 @@ int ocfs2_init_security_set(handle_t *handle,
 				     xattr_ac, data_ac);
 }
 
-struct xattr_handler ocfs2_xattr_security_handler = {
+const struct xattr_handler ocfs2_xattr_security_handler = {
 	.prefix	= XATTR_SECURITY_PREFIX,
 	.list	= ocfs2_xattr_security_list,
 	.get	= ocfs2_xattr_security_get,
@@ -7104,7 +7327,7 @@ static int ocfs2_xattr_trusted_set(struct dentry *dentry, const char *name,
 			       name, value, size, flags);
 }
 
-struct xattr_handler ocfs2_xattr_trusted_handler = {
+const struct xattr_handler ocfs2_xattr_trusted_handler = {
 	.prefix	= XATTR_TRUSTED_PREFIX,
 	.list	= ocfs2_xattr_trusted_list,
 	.get	= ocfs2_xattr_trusted_get,
@@ -7160,7 +7383,7 @@ static int ocfs2_xattr_user_set(struct dentry *dentry, const char *name,
 			       name, value, size, flags);
 }
 
-struct xattr_handler ocfs2_xattr_user_handler = {
+const struct xattr_handler ocfs2_xattr_user_handler = {
 	.prefix	= XATTR_USER_PREFIX,
 	.list	= ocfs2_xattr_user_list,
 	.get	= ocfs2_xattr_user_get,
diff --git a/fs/ocfs2/xattr.h b/fs/ocfs2/xattr.h
index abd72a47f520..aa64bb37a65b 100644
--- a/fs/ocfs2/xattr.h
+++ b/fs/ocfs2/xattr.h
@@ -37,12 +37,12 @@ struct ocfs2_security_xattr_info {
 	size_t value_len;
 };
 
-extern struct xattr_handler ocfs2_xattr_user_handler;
-extern struct xattr_handler ocfs2_xattr_trusted_handler;
-extern struct xattr_handler ocfs2_xattr_security_handler;
-extern struct xattr_handler ocfs2_xattr_acl_access_handler;
-extern struct xattr_handler ocfs2_xattr_acl_default_handler;
-extern struct xattr_handler *ocfs2_xattr_handlers[];
+extern const struct xattr_handler ocfs2_xattr_user_handler;
+extern const struct xattr_handler ocfs2_xattr_trusted_handler;
+extern const struct xattr_handler ocfs2_xattr_security_handler;
+extern const struct xattr_handler ocfs2_xattr_acl_access_handler;
+extern const struct xattr_handler ocfs2_xattr_acl_default_handler;
+extern const struct xattr_handler *ocfs2_xattr_handlers[];
 
 ssize_t ocfs2_listxattr(struct dentry *, char *, size_t);
 int ocfs2_xattr_get_nolock(struct inode *, struct buffer_head *, int,