_get_file_block_size is intended to return the size (in bytes) of the
fundamental allocation unit for a file. This is required for remapping
operations like fallocate and reflink, which can only operate on
allocation units. Since the XFS realtime volume can be configure for
allocation units larger than 1 fs block, we need to factor that in here.
Note that ext* with bigalloc does not allocations to be aligned to the
cluster size, so no update is needed there.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Chandan Babu R <chandanrlinux@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
After fsx supports IO_URING read/write, add a test to do IO_URING
direct-IO test by fsx
Signed-off-by: Zorro Lang <zlang@redhat.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
After fsx supports IO_URING read/write, add a test to do IO_URING
soak test of fsx.
Signed-off-by: Zorro Lang <zlang@redhat.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
IO_URING is a new feature for GNU/Linux system, if someone case of
xfstests tests this feature, better to check if current system
supports it, or need _notrun.
Signed-off-by: Zorro Lang <zlang@redhat.com>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Test that if we keep overwriting an entire file, either with buffered
writes or direct IO writes, the number of used blocks reported by stat(2)
is never zero while the writes and writeback are in progress.
This is motivated by a bug in btrfs and currently fails on btrfs only. It
is fixed a patchset for btrfs that has the following patches:
btrfs: fix missing delalloc new bit for new delalloc ranges
btrfs: refactor btrfs_drop_extents() to make it easier to extend
btrfs: fix race when defragging that leads to unnecessary IO
btrfs: update the number of bytes used by an inode atomically
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
In commit 65cd8e8a8e ("fscrypt-crypt-util: fix IV incrementing for
--iv-ino-lblk-32") I mistakenly decreased the size of fscrypt_iv to 24
bytes, which is the most that is explicitly needed by any of the IV
generation methods. However, Adiantum encryption takes a 32-byte IV, so
the buffer still needs to be 32 bytes, with any extra bytes zeroed.
So restore the size to 32 bytes.
This fixes a buffer overread that caused generic/550 and generic/584 to
sometimes fail, depending on the build of the fscrypt-crypt-util binary.
(Most of the time it still worked by chance.)
Fixes: 65cd8e8a8e ("fscrypt-crypt-util: fix IV incrementing for --iv-ino-lblk-32")
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
attr_remove is deprecated, so replace it with lremovexattr.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
We shouldn't feed sizeof() to strncpy as the string length. Just use
snprintf, which at least doesn't have the zero termination problems.
In file included from /usr/include/string.h:495,
from ../src/global.h:73,
from fsx.c:16:
In function 'strncpy',
inlined from 'main' at fsx.c:2944:5:
/usr/include/x86_64-linux-gnu/bits/string_fortified.h:106:10: warning:
'__builtin_strncpy' specified bound 4096 equals destination size
[-Wstringop-truncation]
106 | return __builtin___strncpy_chk (__dest, __src, __len, __bos (__dest));
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In function 'strncpy',
inlined from 'main' at fsx.c:2914:4:
/usr/include/x86_64-linux-gnu/bits/string_fortified.h:106:10: warning:
'__builtin_strncpy' specified bound 1024 equals destination size
[-Wstringop-truncation]
106 | return __builtin___strncpy_chk (__dest, __src, __len, __bos (__dest));
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Newer versions of xfsprogs now complain if the rmap and refcount btree
levels are insane, so hide that error from the golden output.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Certain tests have certain requirements where the realtime parameters
are concerned. Fix them all.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
This is a regression test that tried to make sure that repair correctly
clears the XFS inode reflink flag when it detects files that do not
share any blocks. However, it does this checking by looking at the
(online) lsattr output. This worked fine during development when we
exposed the reflink state via the stat ioctls, but that has long since
been removed. Now the only way to check is via xfs_db, so switch it to
use that.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
The _get_file_block_size helper was added so that tests could find out
the size of a fundamental unit of allocation for a given file, which is
necessary for certain fallocate and clonerange tests.
On certain filesystem configurations (ocfs2 with clusters, xfs with a
large rt extent size), this is /not/ the same as the filesystem block
size, and these tests will fail. Fix them to use the correct helper.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Fix this test so that it works when the rt extent size is larger than
single block.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
This is a regression test for a specific bug that requires a specific
configuration of the data device. Realtime volumes and external logs
don't affect the efficacy of the test, but the test can fail mkfs if the
realtime device is very large.
Therefore, unset USE_EXTERNAL so that we always run this regression
test, even if the tester enabled realtime.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Test that after doing a memory mapped write to an empty file, a call to
stat(2) reports a non-zero number of used blocks.
This is motivated by a bug in btrfs where the number of blocks used does
not change. It currenly fails on btrfs and it is fixed by a patch that
has the following subject:
"btrfs: fix missing delalloc new bit for new delalloc ranges"
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
When append_reader gets bad data, print the actual and expected values.
This was helpful to me when hunting a failure in generic/069 on ceph
recently.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Filipe noticed that btrfs/220 started failing with some mount option
changes I made recently, but upon closer inspection this test actually
fails in a lot of different ways normally, specifically if you specify
MOUNT_OPTIONS, or if you make an fs with the free space tree.
Address all these issues by reworking how we test that the mount options
are what we expect. First get what the default mount options are for a
plain mount of SCRATCH_DEV. This is used as the baseline, so no matter
how the mount options change in the future it will always work properly.
Secondly instead of specifying a rigid order of the mount options we're
testing, which breaks if we adjust the order in /proc/self/mounts,
simply specify the options we're actually interested in checking. Then
in the test function combine the common options with the new options
we're testing, and then combine that with our actual options and use
some sort magic to see if there's any difference. If there's no
difference then we know we have everything set as expected, if not we
fail.
This patch addresses the initial issue that Filipe noticed, but also
fixes the failures when you specified MOUNT_OPTIONS, or if you made the
fs with the free space tree.
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Test that encryption nonces are unique and random, where randomness is
approximated as "incompressible by the xz program".
This gets indirectly tested by generic/399, but there are some gaps.
It's good to test for this directly too.
This test runs and passes on ext4 and f2fs. It doesn't currently run on
ubifs because _get_encryption_nonce() isn't implemented for ubifs yet.
(At some point I'll probably switch _get_encryption_nonce() to use
FS_IOC_GET_ENCRYPTION_NONCE, which was added in Linux 5.7. But for now
I'd like to keep the tests using it runnable on older kernels too.)
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
generic/398 contains workarounds to allow for renames of encrypted files
to fail with different error codes. However, these error codes were
fixed up by kernel commits f5e55e777cc9 ("fscrypt: return -EXDEV for
incompatible rename or link into encrypted dir") and 0c1ad5242d4f
("ubifs: switch to fscrypt_prepare_rename()").
It's been long enough, so update the test to expect the correct behavior
only, so we don't accidentally reintroduce the wrong behavior.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
generic/395 contains workarounds to allow for some of the fscrypt ioctls
to fail with different error codes. However, the error codes were all
fixed up and documented years ago:
- FS_IOC_GET_ENCRYPTION_POLICY on ext4 failed with ENOENT instead of
ENODATA on unencrypted files. Fixed by commit db717d8e26c2
("fscrypto: move ioctl processing more fully into common code").
- FS_IOC_SET_ENCRYPTION_POLICY failed with EINVAL instead of EEXIST
on encrypted files. Fixed by commit 8488cd96ff88 ("fscrypt: use
EEXIST when file already uses different policy").
- FS_IOC_SET_ENCRYPTION_POLICY failed with EINVAL instead of ENOTDIR
on nondirectories. Fixed by commit dffd0cfa06d4 ("fscrypt: use
ENOTDIR when setting encryption policy on nondirectory").
It's been long enough, so update the test to expect the correct behavior
only, so we don't accidentally reintroduce the wrong behavior.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
generic/397 contains workarounds to allow for kernel bugs where trying
to open or create files in an encrypted directory without the encryption
key failed with ENOENT, EACCES, or EPERM instead of the expected ENOKEY.
However, all these bugs have been fixed. ext4 and f2fs were fixed years
ago by commit 54475f531bb8 ("fscrypt: use ENOKEY when file cannot be
created w/o key"). ubifs was fixed by commit b01531db6cec ("fscrypt:
fix race where ->lookup() marks plaintext dentry as ciphertext").
It's been long enough, so update the test to expect the correct behavior
only, so we don't accidentally reintroduce the wrong behavior.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Some test cases are missing the 'rm -f $seqres.full' line but are appending
to that file, so everytime they run that file gets bigger and bigger (some
of them are using about a dozen megabytes on one of my test boxes).
So just add the 'rm -f $seqres.full' line to them, together with the comment
that the 'new' script generates for new test cases.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Darrick J. Wong
Zorro Lang
Filipe Manana
Eric Biggers
Xiao Yang
Jeff Layton
Josef Bacik