]> git.ipfire.org Git - thirdparty/xfsprogs-dev.git/commitdiff
xfs: No need for inode number error injection in __xfs_dir3_data_check
authorDave Chinner <dchinner@redhat.com>
Wed, 30 Jun 2021 22:28:28 +0000 (18:28 -0400)
committerEric Sandeen <sandeen@sandeen.net>
Wed, 30 Jun 2021 22:28:28 +0000 (18:28 -0400)
Source kernel commit: 39d3c0b5968b5421922e2fc939b6d6158df8ac1c

We call xfs_dir_ino_validate() for every dir entry in a directory
when doing validity checking of the directory. It calls
xfs_verify_dir_ino() then emits a corruption report if bad or does
error injection if good. It is extremely costly:

43.27%  [kernel]  [k] xfs_dir3_leaf_check_int
10.28%  [kernel]  [k] __xfs_dir3_data_check
6.61%  [kernel]  [k] xfs_verify_dir_ino
4.16%  [kernel]  [k] xfs_errortag_test
4.00%  [kernel]  [k] memcpy
3.48%  [kernel]  [k] xfs_dir_ino_validate

7% of the cpu usage in this directory traversal workload is
xfs_dir_ino_validate() doing absolutely nothing.

We don't need error injection to simulate a bad inode numbers in the
directory structure because we can do that by fuzzing the structure
on disk.

And we don't need a corruption report, because the
__xfs_dir3_data_check() will emit one if the inode number is bad.

So just call xfs_verify_dir_ino() directly here, and get rid of all
this unnecessary overhead:

40.30%  [kernel]  [k] xfs_dir3_leaf_check_int
10.98%  [kernel]  [k] __xfs_dir3_data_check
8.10%  [kernel]  [k] xfs_verify_dir_ino
4.42%  [kernel]  [k] memcpy
2.22%  [kernel]  [k] xfs_dir2_data_get_ftype
1.52%  [kernel]  [k] do_raw_spin_lock

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
libxfs/xfs_dir2_data.c

index ddd5e8856b0241efb162c230def22ba1372d2647..d1dbf73ae0469d7c1483f0e21bd6e77fa439aa76 100644 (file)
@@ -215,7 +215,7 @@ __xfs_dir3_data_check(
                 */
                if (dep->namelen == 0)
                        return __this_address;
-               if (xfs_dir_ino_validate(mp, be64_to_cpu(dep->inumber)))
+               if (!xfs_verify_dir_ino(mp, be64_to_cpu(dep->inumber)))
                        return __this_address;
                if (offset + xfs_dir2_data_entsize(mp, dep->namelen) > end)
                        return __this_address;