Btrfs: remove unnecessary btrfs_mark_buffer_dirty in split_leaf
authorLiu Bo <bo.li.liu@oracle.com>
Wed, 7 Sep 2016 21:48:28 +0000 (14:48 -0700)
committerDavid Sterba <dsterba@suse.com>
Mon, 26 Sep 2016 17:50:44 +0000 (19:50 +0200)
When we're not able to get enough space through splitting leaf,
we'd create a new sibling leaf instead, and it's possible that we return
 a zero-nritem sibling leaf and mark it dirty before it's in a consistent
state.  With CONFIG_BTRFS_FS_CHECK_INTEGRITY=y, the integrity check of
check_leaf will report panic due to this zero-nritem non-root leaf.

This removes the unnecessary btrfs_mark_buffer_dirty.

Reported-by: Filipe Manana <fdmanana@gmail.com>
Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
fs/btrfs/ctree.c

index 8f67a14..f6ba165 100644 (file)
@@ -4311,7 +4311,11 @@ again:
                        if (path->slots[1] == 0)
                                fixup_low_keys(fs_info, path, &disk_key, 1);
                }
-               btrfs_mark_buffer_dirty(right);
+               /*
+                * We create a new leaf 'right' for the required ins_len and
+                * we'll do btrfs_mark_buffer_dirty() on this leaf after copying
+                * the content of ins_len to 'right'.
+                */
                return ret;
        }