]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/commitdiff
btrfs: remove WARN_ON in log_dir_items
authorJosef Bacik <josef@toxicpanda.com>
Wed, 6 Mar 2019 22:13:04 +0000 (17:13 -0500)
committerKleber Sacilotto de Souza <kleber.souza@canonical.com>
Wed, 14 Aug 2019 09:18:49 +0000 (11:18 +0200)
BugLink: https://bugs.launchpad.net/bugs/1838116
commit 2cc8334270e281815c3850c3adea363c51f21e0d upstream.

When Filipe added the recursive directory logging stuff in
2f2ff0ee5e430 ("Btrfs: fix metadata inconsistencies after directory
fsync") he specifically didn't take the directory i_mutex for the
children directories that we need to log because of lockdep.  This is
generally fine, but can lead to this WARN_ON() tripping if we happen to
run delayed deletion's in between our first search and our second search
of dir_item/dir_indexes for this directory.  We expect this to happen,
so the WARN_ON() isn't necessary.  Drop the WARN_ON() and add a comment
so we know why this case can happen.

CC: stable@vger.kernel.org # 4.4+
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Kamal Mostafa <kamal@canonical.com>
Signed-off-by: Khalid Elmously <khalid.elmously@canonical.com>
fs/btrfs/tree-log.c

index 96e5dd95ff6a58efb5cf99a5ce1dc9606101b231..5e86dba989536aec2d9806741cafc5c899eb09b9 100644 (file)
@@ -3424,9 +3424,16 @@ static noinline int log_dir_items(struct btrfs_trans_handle *trans,
        }
        btrfs_release_path(path);
 
-       /* find the first key from this transaction again */
+       /*
+        * Find the first key from this transaction again.  See the note for
+        * log_new_dir_dentries, if we're logging a directory recursively we
+        * won't be holding its i_mutex, which means we can modify the directory
+        * while we're logging it.  If we remove an entry between our first
+        * search and this search we'll not find the key again and can just
+        * bail.
+        */
        ret = btrfs_search_slot(NULL, root, &min_key, path, 0, 0);
-       if (WARN_ON(ret != 0))
+       if (ret != 0)
                goto done;
 
        /*