Browse Source

Btrfs: avoid deadlocks during reservations in btrfs_truncate_block

The new enospc code makes it possible to deadlock if we don't use
FLUSH_LIMIT during reservations inside a transaction.  This enforces
the correct flush type to avoid both deadlocks and assertions

Signed-off-by: Chris Mason <clm@fb.com>
Signed-off-by: Josef Bacik <jbacik@fb.com>
Josef Bacik 9 years ago
parent
commit
bac357dcec
1 changed files with 5 additions and 0 deletions
  1. 5 0
      fs/btrfs/extent-tree.c

+ 5 - 0
fs/btrfs/extent-tree.c

@@ -5939,10 +5939,15 @@ int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes)
 	 * the middle of a transaction commit.  We also don't need the delalloc
 	 * mutex since we won't race with anybody.  We need this mostly to make
 	 * lockdep shut its filthy mouth.
+	 *
+	 * If we have a transaction open (can happen if we call truncate_block
+	 * from truncate), then we need FLUSH_LIMIT so we don't deadlock.
 	 */
 	if (btrfs_is_free_space_inode(inode)) {
 		flush = BTRFS_RESERVE_NO_FLUSH;
 		delalloc_lock = false;
+	} else if (current->journal_info) {
+		flush = BTRFS_RESERVE_FLUSH_LIMIT;
 	}
 
 	if (flush != BTRFS_RESERVE_NO_FLUSH &&