Rework btrfs_drop_inode to avoid scheduling

Signed-off-by: Chris Mason <chris.mason@oracle.com>

6 files changed, 44 insertions, 10 deletions

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index a2c2d6d82c5e..1e19f2d86334 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -349,6 +349,7 @@ struct btrfs_fs_info {
 
 	u64 total_pinned;
 	spinlock_t delalloc_lock;
+	spinlock_t new_trans_lock;
 	u64 delalloc_bytes;
 };
 /*
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 87daed632f37..cd29922d4073 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -639,6 +639,7 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	INIT_LIST_HEAD(&fs_info->hashers);
 	spin_lock_init(&fs_info->hash_lock);
 	spin_lock_init(&fs_info->delalloc_lock);
+	spin_lock_init(&fs_info->new_trans_lock);
 
 	memset(&fs_info->super_kobj, 0, sizeof(fs_info->super_kobj));
 	init_completion(&fs_info->kobj_unregister);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index e53d20331646..008e3445748e 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -993,20 +993,13 @@ fail:
 
 void btrfs_drop_inode(struct inode *inode)
 {
-	struct btrfs_trans_handle *trans;
-	struct btrfs_root *root = BTRFS_I(inode)->root;
-
-	if (!BTRFS_I(inode)->ordered_trans) {
+	if (!BTRFS_I(inode)->ordered_trans || inode->i_nlink) {
 		generic_drop_inode(inode);
 		return;
 	}
-	/* nasty, but it prevents a deadlock with data=ordered by preventing
-	 * a commit until after this inode is done
-	 */
-	trans = btrfs_start_transaction(root, 1);
+	/* FIXME, make sure this delete actually ends up in the transaction */
+	btrfs_del_ordered_inode(inode);
 	generic_drop_inode(inode);
-	/* note, the inode is now untouchable */
-	btrfs_end_transaction(trans, root);
 }
 
 void btrfs_delete_inode(struct inode *inode)
diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c
index 411aba84d305..b56011baa17c 100644
--- a/fs/btrfs/ordered-data.c
+++ b/fs/btrfs/ordered-data.c
@@ -219,3 +219,39 @@ int btrfs_find_del_first_ordered_inode(struct btrfs_ordered_inode_tree *tree,
 	kfree(entry);
 	return 1;
 }
+
+static int __btrfs_del_ordered_inode(struct btrfs_ordered_inode_tree *tree,
+				     u64 root_objectid, u64 objectid)
+{
+	struct tree_entry *entry;
+	struct rb_node *node;
+	struct rb_node *prev;
+
+	write_lock(&tree->lock);
+	node = __tree_search(&tree->tree, root_objectid, objectid, &prev);
+	if (!node) {
+		write_unlock(&tree->lock);
+		return 0;
+	}
+	rb_erase(node, &tree->tree);
+	write_unlock(&tree->lock);
+	entry = rb_entry(node, struct tree_entry, rb_node);
+	kfree(entry);
+	return 1;
+}
+
+int btrfs_del_ordered_inode(struct inode *inode)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	u64 root_objectid = root->root_key.objectid;
+
+	spin_lock(&root->fs_info->new_trans_lock);
+	if (root->fs_info->running_transaction) {
+		struct btrfs_ordered_inode_tree *tree;
+		tree = &root->fs_info->running_transaction->ordered_inode_tree;
+		__btrfs_del_ordered_inode(tree, root_objectid, inode->i_ino);
+	}
+	spin_unlock(&root->fs_info->new_trans_lock);
+	return 0;
+}
+
diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h
index aaf9eb142719..26b26212865b 100644
--- a/fs/btrfs/ordered-data.h
+++ b/fs/btrfs/ordered-data.h
@@ -36,4 +36,5 @@ int btrfs_find_del_first_ordered_inode(struct btrfs_ordered_inode_tree *tree,
 				       u64 *root_objectid, u64 *objectid);
 int btrfs_find_first_ordered_inode(struct btrfs_ordered_inode_tree *tree,
 				       u64 *root_objectid, u64 *objectid);
+int btrfs_del_ordered_inode(struct inode *inode);
 #endif
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 614903f5c884..a3205808ab2b 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -699,7 +699,9 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	BUG_ON(ret);
 
 	cur_trans = root->fs_info->running_transaction;
+	spin_lock(&root->fs_info->new_trans_lock);
 	root->fs_info->running_transaction = NULL;
+	spin_unlock(&root->fs_info->new_trans_lock);
 	btrfs_set_super_generation(&root->fs_info->super_copy,
 				   cur_trans->transid);
 	btrfs_set_super_root(&root->fs_info->super_copy,