Although ro-compat features cannot be mounted rw, a bug allows an
ro->rw remount transition, and this will corrupt the filesystem if
it contains ro-compat features unknown to the running kernel.
This bug has been fixed on linux kernel (d0a58e8 xfs: disallow rw
remount on fs with unknown ro-compat features), this case is the
regression testcase of that bug.
Signed-off-by: Zorro Lang <zlang@redhat.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Current generic/095 warning whitelist include:
"xfs_file_dio_aio_read" and "xfs_file_read_iter"
But from the history we can see:
1) xfs_file_dio_aio_read() was splited from xfs_file_read_iter() by
below commit:
bbc5a74 xfs: split xfs_file_read_iter into buffered and direct I/O helpers
2) xfs_file_read_iter() came from xfs_file_aio_read() as below
commit:
b4f5d2c xfs: switch to ->read_iter()
Some Linux distro (e.g: RHEL-7) still use the old kernel with old
function xfs_file_aio_read().
Signed-off-by: Zorro Lang <zlang@redhat.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Define test groups for those tests which have _require_xfs_io_command
for punch, collapse, insert, and zero. This makes it easier to
exclude tests that use one of these fallocate commands. Or if you
want to specifically test for those fallocate commands you can do
this.
This obviates an out-of-tree xfstests patch I maintain which used an
XFS_IO_AVOID environment variable to suppress running tests that use
punch, collapse, insert, etc. This was rejected because of the
claim that it could be done using groups. So this commit is in
response to those upstream comments.
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Test that if we rename a file, without changing its parent
directory, create a new file that has the old name of the file we
renamed, doing an fsync against the file we renamed works correctly
and after a power failure both files exists.
This is motivated by an issue found in btrfs, introduced in the
linux kernel 4.8-rc2, that caused lockdep to report a deadlock
scenario. The btrfs issue is fixed by the patch titled:
"Btrfs: fix lockdep warning on deadlock against an inode's log mutex"
Tested against ext3/4, xfs, f2fs and reiserfs as well.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Check if SGID is cleared upon chmod / setfacl when the owner is not
in the owning group. As of today, the kernel fails to clear SGID in
setxattr (which is what acl_set_file is implemented on top of) in
that case; see this patch:
https://patchwork.kernel.org/patch/9290507/
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Besides fail_at_unmount, all EIO error handling can stop umount
hanging on IO error too.
This case test EIO/max_retries and EIO/retry_timeout_seconds as
below:
1) fail_at_unmount=0 && \
EIO/max_retries=1 && \
EIO/retry_timeout_seconds=0
2) fail_at_unmount=0 && \
EIO/max_retries=-1 && \
EIO/retry_timeout_seconds=1
Make sure when fail_at_unmount=0, umount won't hang there.
Signed-off-by: Zorro Lang <zlang@redhat.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Nearly 1/3 code is used to reset the xfs error handling attributes,
This part can be picked up, and used for other cases. So move them
to a new function _reset_xfs_sysfs_error_handling() in common/rc.
Signed-off-by: Zorro Lang <zlang@redhat.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
When blocksize is 512b, xfs/033 will run mkfs.xfs again to make sure
inode size is 512b. As below:
# devzero blows away 512byte blocks, so make 512byte inodes (at least)
_scratch_mkfs_xfs | _filter_mkfs 2>$tmp.mkfs
. $tmp.mkfs
[ $isize -lt 512 ] && \
_scratch_mkfs_xfs -isize=512 | _filter_mkfs >/dev/null 2>&1
But after "566ebd5 mkfs: default to CRC enabled filesystems",
xfsprogs enable CRC as default. If MKFS_OPTIONS is "-b size=512 -m
crc=0", the first "_scratch_mkfs_xfs" will make crc=0, but the
second will change "crc" to "1", because it'll waive the "-b
size=512 -m crc=0" options.
xfs/033 need to check xfs's sb_features, to use different .out
files. So we should import the mkfs output($tmp.mkfs file) again, if
we mkfs again.
Signed-off-by: Zorro Lang <zlang@redhat.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
The code in _require_metadata_journaling() currently rejects ext4
file systems that have been created without a journal. However, an
ext4 file system with a journal should also be rejected if the
journal is not loaded and made available for use at mount time.
Signed-off-by: Eric Whitney <enwlinux@gmail.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
In generic/135, benign warning message output from _scratch_mkfs on
stderr can cause a golden output mismatch, as can occur when testing
an ext4 bigalloc file system. It's the only generic test that does
not also redirect stderr when directing _scratch_mkfs' stdout to
/dev/null. Since error output from _scratch_mkfs doesn't seem
important to the overall intent of the test, there should be no harm
in making its stderr handling consistent.
Signed-off-by: Eric Whitney <enwlinux@gmail.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
If we don't have temp file used to test on xfsprogs 3.2.2,
_require_xfs_mkfs_validation will fail because mkfs.xfs can't create
this file. We need to skip tests before xfsprogs 4.7.0, so fix it.
This feature was introduced in xfsprogs 4.2.0 by commit 20cec860e16b
("mkfs.xfs: always use underlying fs sector size when mkfs'ing a
file")
Signed-off-by: Xiao Yang <yangx.jy@cn.fujitsu.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
glibc 2.24 defines Q_GETNEXTQUOTA but not `struct nextdqblk`. We
need to define it ourselves, but let's give it a hacky name that
won't cause another build error when the definition is added to
glibc.
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
This exercises the xfs_quota "state" command for every
combination of user, group, and project quota enablement
on both crc (with project quota inode on disk) and non-crc
(where project quota overlaps group quota on disk) filesystems.
It currently requires patches to the upstream kernel
("quota: fill in Q_XGETQSTAT inode information for inactive quotas")
and to xfsprogs ("xfs_quota: wire up XFS_GETQSTATV") to pass.
Rather than trying to make it run on older kernels without
crc capability, I restrict the test to crc-capable environments,
but test with and without crcs enabled (and hence with and without
project quota inode on disk).
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Bill O'Donnell <billodo@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
_qmount_option() did not properly filter the
"prjquota" mount option along with the "pquota"
variant, as it does for the other types.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Bill O'Donnell <billodo@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Starting from kernel commit 337684a1746f ("fs: return EPERM on
immutable inode"), EPERM is returned instead of EACCES on immutable
inodes. So update src/t_immutable.c to accept EPERM as a PASS.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Add the on-disk structures added by the realtime rmapbt.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Teach _check_xfs_filesystem to scrub mounted filesystems before
unmounting and fscking them. This is mostly to test the online scrub
tool...
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Test sharing blocks via reflink and dedupe between two different
mountpoints of the same filesystem. This shouldn't work, since we
don't allow cross-mountpoint functions.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Document all build and runtime requirements for xfstests
on Fedora/RHEL/CentOS boxes, as was done earlier for Ubuntu.
Minor other README edits as well.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Ensure that we can create a file with a single extent, reflink two
blocks out of the middle of that extent, and the resulting fiemap
reports two shared extents, instead of lazily reporting the entire
huge extent as shared.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
ext4 supports dynamic expansion of inode size via inode's
i_extra_size field. Test the code performing the expansion in the
kernel.
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
The existing _require_defrag function rejects ext4 file systems
mounted with the dax option or constructed without extents.
However, there are also a number of other ext4 file system
configuration cases that do not currently support online defrag,
including encryption, data journaling, and bigalloc. In the future,
online defrag functionality may be implemented for some of these,
and new configuration cases could be added that don't support it.
Rather than add a laundry list of mount and file system
configuration options to the existing _require_defrag function that
will need ongoing maintenance, use the available src/e4compact
program to directly determine whether the kernel supports ext4's
move extent ioctl on the test file system.
Signed-off-by: Eric Whitney <enwlinux@gmail.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Apparently the XFS attr_list_by_handle ioctl has never actually
copied the cursor contents back to user space, which means that
iteration has never worked. Add a test case for this and see
"xfs: in _attrlist_by_handle, copy the cursor back to userspace".
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Calling _mount doesn't work when we want to add mount options such
as realtime devices. Fix it by calling _scratch_options first then
mount with $SCRATCH_OPTIONS.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
This test uses the scratch device, so cycle that, not the test dev.
This is also a xfs_fsr test, so put it in the fsr group.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Zorro Lang
Theodore Ts'o
Filipe Manana
Andreas Gruenbacher
Eric Whitney
Xiao Yang
Omar Sandoval
Eric Sandeen
Eric Sandeen
Eryu Guan
Darrick J. Wong
Jan Kara