This test is motivated by an fsync issue discovered in btrfs.
The steps to trigger the issue were:
1) remove an hard link from an inode with a large number of hard links;
2) add a new hard link;
3) add another hard link with the same name as the one removed in step 1;
4) fsync the inode.
These steps made the btrfs fsync log replay fail (with the -EOVERFLOW
error), making the filesystem unmountable, requiring the use of
btrfs-zero-log (it wipes the fsync log) in order to make the filesystem
mountable again (but losing some data/metadata).
The btrfs issue was fixed by the following linux kernel patches:
Btrfs: fix fsync when extend references are added to an inode
Btrfs: fix fsync log replay for inodes with a mix of regular refs and extrefs
This issue was present in btrfs since the extrefs (extend references)
feature was added (2012).
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
This test is motivated by an fsync issue discovered in btrfs.
The issue in btrfs was that adding a new hard link to an inode that
already had a large number of hardlinks and fsync the inode, would
make the fsync log replay code update the inode with a wrong link count
(smaller than the correct value). This resulted later in dangling
directory index entries, after removing most of the hard links
(correct_value - wrong_value), that were visible to user space but it
was impossible to delete them or do any other operation on them (since
they pointed to an inode that didn't exist anymore, resulting in -ESTALE
errors).
The btrfs issue was fixed by the following linux kernel patch:
Btrfs: fix fsync when extend references are added to an inode
This issue was present in btrfs since the extrefs (extend references)
feature was added (2012).
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
This test is motivated by an fsync issue discovered in btrfs.
The issue was that after fsyncing an inode that got its link count
decremented, and the new link count is greater than zero, after the
fsync log replay the inode's parent directory metadata became
inconsistent - it had a wrong i_size and dangling index entries which
prevented the directory from ever being removed (rmdir always failed
with -ENOTEMPTY, even if the directory had no more child inodes).
The btrfs issue was fixed by the following linux kernel patch:
Btrfs: fix directory inconsistency after fsync log replay
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Commit 01d42b7efe broke the check
for the success status of running fstrim. The [ ] bracets should
have been killed. This made several tests being skipped even when
the test/scratch devices support trim/discard.
For reference:
$ [ fstrim /mnt/ ] || echo foobar
bash: [: fstrim: unary operator expected
foobar
$ fstrim /mnt/ || echo foobar
$ echo $?
0
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
After installing the test suite, src directory only contains binary
programs in the final building directory. Here executing "cp src/itrash.c
$SCRATCH_MNT/${I}" will output "cp: cannot stat 'src/itrash.c': No
such file or directory" error message. Fix it.
Signed-off-by: Xing Gu <gux.fnst@cn.fujitsu.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Previously this test was taking 6-7 minutes, and writing half a
gigabyte of data in the dump/restore test directory. Change this to
be about 60 megs, and to take ~20 seconds.
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Currently if _check_test_fs and/or _check_scratch_fs find corruption,
the test itself is still reported as pass, like
[root@hp-dl388eg8-01 xfstests]# ./check xfs/071 xfs/072
FSTYP -- xfs (non-debug)
PLATFORM -- Linux/x86_64 hp-dl388eg8-01 3.18.0-rc7+
MKFS_OPTIONS -- -f -bsize=4096 /dev/sda6
MOUNT_OPTIONS -- -o context=system_u:object_r:nfs_t:s0 /dev/sda6 /mnt/testarea/scratch
xfs/071 2s
_check_xfs_filesystem: filesystem on /dev/sda6 is inconsistent (r) (see /root/xfstests/results//xfs/071.full)
xfs/072 1s
Ran: xfs/071 xfs/072
Passed all 2 tests
[root@hp-dl388eg8-01 xfstests]# echo $?
0
Usually it's not a problem, but it does confuse scripts that depend on
return value of check. Update check to treat _check_{test,scratch}_fs
failures as test failures too, new test output is like
[root@hp-dl388eg8-01 xfstests]# ./check xfs/071 xfs/072
FSTYP -- xfs (non-debug)
PLATFORM -- Linux/x86_64 hp-dl388eg8-01 3.18.0-rc7+
MKFS_OPTIONS -- -f -bsize=4096 /dev/sda6
MOUNT_OPTIONS -- -o context=system_u:object_r:nfs_t:s0 /dev/sda6 /mnt/testarea/scratch
xfs/071 2s ... 2s
_check_xfs_filesystem: filesystem on /dev/sda6 is inconsistent (r) (see /root/xfstests/results//xfs/071.full)
xfs/072 1s ... 1s
Ran: xfs/071 xfs/072
Failures: xfs/071
Failed 1 of 2 tests
[root@hp-dl388eg8-01 xfstests]# echo $?
1
Signed-off-by: Eryu Guan <eguan@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
So the callers could know if these functions find corruptions by the
return value.
Signed-off-by: Eryu Guan <eguan@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
If _check_xxx_filesystem called by tests when corruption found
they exit, but we don't want the calls from the test harness itself to
exit when corruption is found.
_check_xfs_filesystem already behaves correctly, make
_check_generic_filesystem and _check_btrfs_filesystem follow the same
behavior.
Signed-off-by: Eryu Guan <eguan@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Some filesystems do not support O_DIRECT. Check whether TEST_DIR supports
it by running xfs_io with -d flag.
Signed-off-by: Junho Ryu <jayr@google.com>
Signed-off-by: Dushan Tcholich <dusanc@gmail.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
To check for FITRIM tests used _require_fstrim() and
_test_batched_discard() but as _test_batched_discard() already
includes _test_fstrim() unify FSTRIM check throughout xfstests with
_require_batched_discard().
Signed-off-by: Dushan Tcholich <dusanc@gmail.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
These tests use the falloc command in xfs_io, and there are some file
systems (ext3) or file system configurations (ext4 in ext3
compatibility mode) which do not support fallocate. So add the
explicit requirement to avoid false test failures.
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Regression test for a btrfs clone ioctl issue where races between
a clone operation and concurrent target file reads would result in
leaving stale data in the page cache. After the clone operation
finished, reading from the clone target file would return the old
and no longer valid data. This affected only buffered reads (i.e.
didn't affect direct IO reads).
This issue was fixed by the following linux kernel patch:
Btrfs: ensure readers see new data after a clone operation
(commit c125b8bff1d9f6c8c91ce4eb8bd5616058c7d510)
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Regression test for a btrfs issue where if right after the snapshot
creation ioctl started, a file write followed by a file truncate
happened, with both operations increasing the file's size, the created
snapshot would capture an inconsistent state of the file system tree.
That state reflected the file truncation but it didn't reflect the
write operation, and left a gap between two file extent items (and
that gap corresponded to the total or a partial area of the write
operation's range).
This issue was fixed by the following linux kernel patch:
Btrfs: fix snapshot inconsistency after a file write followed by truncate
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
It's possible based on a race conditions (and possibly the version of
coreutils which supplies /usr/bin/yes) that commands of the form:
yes | $MKFS_PROG ...
will end up causing the following failure:
shared/298 16s ... [23:49:03] [23:49:19] - output mismatch (see /results/results-4k/shared/298.out.bad)
--- tests/shared/298.out 2014-10-31 10:13:04.000000000 -0400
+++ /results/results-4k/shared/298.out.bad 2014-11-29 23:49:19.118138099 -0500
@@ -1,4 +1,6 @@
QA output created by 298
+yes: standard output: Broken pipe
+yes: write error
Generating garbage on loop...done.
Running fstrim...done.
Detecting interesting holes in image...done.
...
(Run 'diff -u tests/shared/298.out /results/results-4k/shared/298.out.bad' to see the entire diff)
Using mke2fs's -F option instad of "yes | $MKFS_PROG" fixes this issue.
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Fix two problems in generic/299
1. Remove $seqres.full before test, otherwise the file is growing all
the time.
2. Make sure fio really exits, otherwise fio would block umount. $pid is
the pid of function run_check not fio, sometimes fio is still there when
$pid is dead and blocking umount.
Signed-off-by: Eryu Guan <eguan@redhat.com>
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Some new btrfs groups have been added in the btrfs stress patchset, add
other tests to proper groups too.
Reviewed-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
_require_ext4_bigalloc and _require_ext4_mkfs_bigalloc create a
200k fs. DAX (as XIP before it) relies on a fs block size ==
system page size; ie 4k for us. With a 200k fs size limit,
mkfs refuses to create a filesystem with a 4k block size, so
the check fails, but for a bogus reason.
Increasing the fs size to 512m allows mkfs to create a filesystem
with 4k block size, so the test passes.
Signed-off-by: Vishal Verma <vishal.l.verma@linux.intel.com>
Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
_require_test will mount test dev if not mounted, fix the comment to
reflect that.
Signed-off-by: Eryu Guan <eguan@redhat.com>
Reviewed-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Filipe Manana
Jaegeuk Kim
Xing Gu
Dushan Tcholich
Theodore Ts'o
Eryu Guan
Liu Bo
Junho Ryu
David Disseldorp
Vishal Verma