btrfs: deal with errors when adding inode reference during log replay
authorFilipe Manana <fdmanana@suse.com>
Fri, 1 Oct 2021 12:52:32 +0000 (13:52 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 20 Oct 2021 09:44:59 +0000 (11:44 +0200)
commit 52db77791fe24538c8aa2a183248399715f6b380 upstream.

At __inode_add_ref(), we treating any error returned from
btrfs_lookup_dir_item() or from btrfs_lookup_dir_index_item() as meaning
that there is no existing directory entry in the fs/subvolume tree.
This is not correct since we can get errors such as, for example, -EIO
when reading extent buffers while searching the fs/subvolume's btree.

So fix that and return the error to the caller when it is not -ENOENT.

CC: stable@vger.kernel.org # 4.14+
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/btrfs/tree-log.c

index ca8fbc7..66d275b 100644 (file)
@@ -1137,7 +1137,10 @@ next:
        /* look for a conflicting sequence number */
        di = btrfs_lookup_dir_index_item(trans, root, path, btrfs_ino(dir),
                                         ref_index, name, namelen, 0);
-       if (di && !IS_ERR(di)) {
+       if (IS_ERR(di)) {
+               if (PTR_ERR(di) != -ENOENT)
+                       return PTR_ERR(di);
+       } else if (di) {
                ret = drop_one_dir_item(trans, root, path, dir, di);
                if (ret)
                        return ret;
@@ -1147,7 +1150,9 @@ next:
        /* look for a conflicting name */
        di = btrfs_lookup_dir_item(trans, root, path, btrfs_ino(dir),
                                   name, namelen, 0);
-       if (di && !IS_ERR(di)) {
+       if (IS_ERR(di)) {
+               return PTR_ERR(di);
+       } else if (di) {
                ret = drop_one_dir_item(trans, root, path, dir, di);
                if (ret)
                        return ret;