A simply reproducer from Frank Sorenson:
ftruncate(fd, 65012224)
io_prep_pwrite(iocbs[0], fd, buf[0], 1048576, 63963648);
io_prep_pwrite(iocbs[1], fd, buf[1], 1048576, 65012224);
io_submit(io_ctx, 1, &iocbs[0]);
io_submit(io_ctx, 1, &iocbs[1]);
io_getevents(io_ctx, 2, 2, events, NULL)
help to find an ext4 corruption:
**************** **************** ****************
* page 1 * * page 2 * * page 3 *
**************** **************** ****************
existing 0000000000000000 0000000000000000 0000000000000000
write 1 AAAAAAAAAAAAAA AA
write 2 BBBBBBBBBBBBBB BB
result 00AAAAAAAAAAAAAA 00BBBBBBBBBBBBBB BB00000000000000
desired 00AAAAAAAAAAAAAA AABBBBBBBBBBBBBB BB00000000000000
This issue remind us we might miss unaligned AIO test for long time.
We thought fsx cover this part, but looks like it's not. So this case
trys to cover unaligned direct AIO write test on file with different
initial truncate i_size.
The following patches fix the issue on xfs and ext4.
xfs: serialize unaligned dio writes against all other dio writes
ext4: Fix data corruption caused by unaligned direct AIO
Signed-off-by: Zorro Lang <zlang@redhat.com>
Reviewed-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Older kernels (prior commit 494633fac7896 "vfs: vfs_dedupe_file_range()
doesn't return EOPNOTSUPP") will return EINVAL when operation is not
supported. Make fsx treat this error as a sign of unsupported
deduplication as well to make it usable with these older kernels.
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Commit "8309b39a fstests: fix broken _require_scratch usage" did
below change on xfs/065:
-_scratch_unmount 2>/dev/null
+_scratch_mkfs_xfs >> $seqres.full
It cause xfs/065 always fails now, as:
QA output created by 065
mkfs.xfs: /dev/sdb2 contains a mounted filesystem
...
So use _require_scratch, to make sure the SCRATCH_DEV is unmounted
before mkfs.
Signed-off-by: Zorro Lang <zlang@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
When this test finishes there is no 077.full file with output from
commands. Sometimes this information is useful for post mortem so
stop deleting the file upon test completion.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
This test makes sure that we can't use stale unrecovered fs metadata to
drive a DISCARD festival on a disk and thereby destroy user data by
accident.
The following patches fixed the bug on ext4, xfs and btrfs
ext4: prohibit fstrim in norecovery mode
xfs: prohibit fstrim in norecovery mode
Btrfs: do not allow trimming when a fs is mounted with the nologreplay option
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
When a device is removed from a btrfs filesystem its superblock copies
must be deleted. This test ensures this is indeed the case.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
XFS has historically had a stale data exposure window if a crash
occurs after a delalloc->physical extent conversion but before
writeback completes to the associated extent. While this should be a
rare occurrence in production environments due to typical writeback
ordering and such, it is not guaranteed in all cases until data
extents are initialized as unwritten (or otherwise zeroed) before
they are written.
Add a test that performs selective writeback ordering to reproduce
stale data exposure after a crash. Note that this test currently
fails on XFS.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Every call to _scratch_dev_pool_get must be paired with call to
_scratch_dev_pool_put otherwise the SCRATCH_POOL variable will have
less devices than it actually must.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Make sure that copy, metadump, and mdrestore work on a filesystem with
all known metadata types.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
$USE_EXTERNAL needs to be set when using external log devices. In such a
setup, tests which have "_require_prjquota
$SCRATCH_DEV" (e.g. generic/383) incorrectly end up being marked as
"not run" since the test "[ "$USE_EXTERNAL" = yes -a ! -z "$_dev" ]"
evaluates to true.
This commit fixes the bug by marking the test as "not run" only when
$USE_EXTERNAL is set and one of $TEST_RTDEV or $SCRATCH_RTDEV is set.
Signed-off-by: Chandan Rajendra <chandan@linux.ibm.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
_require_scratch doesn't actually format the scratch device with
anything, which means that tests are required to format them before
using them. Fix tests that don't do this correctly.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Enhance the populated fs metadump image cache to support multiple
configurations per filesystem so that we reduce the image creation
overhead even further.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Don't require scrub for ext4's populated fs creation test because there
is no general online scrub program for ext*.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Certain filesystems (ext4 w/ 1k block size) can run out of space while
running this test because they have very limited xattr storage
capabilities. If we run out of space while setting an attr, don't
bother continuing the test.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Fix the unwritten extent detector in this test to ignore post-eof
allocations because those are harmless.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
This test doesn't call fsync or sync to force writeback of the first 60k
of the file, which means that we could end up with a file full of
zeroes or an empty file. Since this is a regression test that looks for
stale disk contents slipping through, change the test to look for the
stale bytes.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
coreutils provides the shuf(1) utility that randomizes the order of a
list and seeds its random number generator with /dev/urandom. It's a
bit speedier than awk, so use it if available.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
awk doesn't have a particularly good random number generator -- it seeds
from the Unix epoch time in seconds, which means that the run order
across a bunch of VMs started at exactly the same time are unsettlingly
predictable. Therefore, at least try to seed it with bash's $RANDOM,
which is slightly less predictable.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
When test generic/131 on nfs, the 'server.out' maybe create
later than expect. Because the server is running on background,
we should ensure the 'server.out' is exist before 'cat' it.
So, let's create the server.out manually.
Signed-off-by: ZhangXiaoxu <zhangxiaoxu5@huawei.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Update generic/398 to pass after kernel commit f5e55e777cc9 ("fscrypt:
return -EXDEV for incompatible rename or link into encrypted dir"),
which intentionally changed some error codes from EPERM to EXDEV in
order to allow standard tools like 'mv' to move files into an encrypted
directory.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
The way we decided if an unwritten extent is considered a hole or
data is by checking if the page and/or blocks are marked uptodate,
that is contain valid data in the page cache.
xfs/420 and xfs/421 try to exercise SEEK_HOLE / SEEK_DATA in the
presence of cowextsize preallocations over holes in the data fork.
The current XFS code never actually uses those for buffer writes,
but a pending patch changes that. For SEEK_HOLE / SEEK_DATA to work
properly in that case we also need to look at the COW fork in their
implementations and thus have to rely on the unwritten extent page
cache probing. But the tests for it ensure we do have valid data in
the pagecache by calling md5sum on the test files, and thus reading
their contents (including the zero-filled holes) in, and thus making
them all valid data.
Fix that by dropping the page cache content after the md5sum calls.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
After fsync, filesystem should guarantee inode metadata including
permission info being persisted, so even after sudden power-cut,
during mount, we should recover i_mode fields correctly, in order
to not loss those meta info.
So adding this testcase to check whether generic filesystem can
guarantee that.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Test that if we truncate a file to reduce its size, rename it and then
fsync it, after a power failure the file has a correct size and name.
This test is motivated by a bug found in btrfs, which is fixed by a
patch for the linux kernel titled:
"Btrfs: fix incorrect file size after shrinking truncate and fsync"
This test currently passes on ext4, xfs, f2fs and patched btrfs.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Split out most of the user.* tests from 097 and move them to a new
test that only tests user.* xattrs.
This makes it possible to use this test on filesystems that can only
provide user.* xattrs such as CIFS.
Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Zorro Lang
Jan Kara
Nikolay Borisov
Darrick J. Wong
Brian Foster
Chandan Rajendra
zhangxiaoxu 00467499
Eric Biggers
Christoph Hellwig
Chao Yu
Filipe Manana
Ronnie Sahlberg