This test is motivated by an fsync issue discovered in btrfs.
The issue was that after punching a hole for a small range, which
affected only a partial page, an fsync operation would have no effect
at all. This was because for this particular case the btrfs hole
punching implementation did not update some btrfs specific inode
metadata that is required to determine if an fsync operation needs
to update the fsync log. For this to happen, it was also necessary
that in the transaction where the hole punching was performed, and
before the fsync operation, no other operation that modified the file
(or its metadata) was performed.
The btrfs issue was fixed by the following linux kernel patch:
Btrfs: add missing inode update when punching hole
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 we could lose file data, that was previously
fsync'ed successfully, if we end up adding a hard link to our
inode and then persist the fsync log later via an fsync of other
inode for example. This is similar to my previous test, except
that in this test the inode that ends up losing data was created
(with some data) in a transaction different from the one we made
an fsync.
The btrfs issue was fixed by the following linux kernel patch:
Btrfs: fix fsync data loss after adding hard link to inode
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 we could lose file data, that was previously
fsync'ed successfully, if we end up adding a hard link to our
inode and then persist the fsync log later via an fsync of other
inode for example.
The btrfs issue was fixed by the following linux kernel patch:
Btrfs: fix fsync data loss after adding hard link to inode
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
This patch moves the generic testcases defined in xfs into tests/generic/.
xfs/085 -> generic/052
xfs/086 -> generic/054
xfs/087 -> generic/055
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
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 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>
Stress btrfs' block group allocation and deallocation while running
fstrim in parallel. Part of the goal is also to get data block groups
deallocated so that new metadata block groups, using the same physical
device space ranges, get allocated while fstrim is running. This caused
several issues ranging from invalid memory accesses, kernel crashes,
metadata or data corruption, free space cache inconsistencies, free
space leaks and memory leaks.
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 verifies that replacing a xattr's value is an atomic
operation. This is motivated by an issue in btrfs where replacing
a xattr's value wasn't an atomic operation, it consisted of
removing the old value and then inserting the new value in a
btree. This made readers (getxattr and listxattrs) not getting
neither the old nor the new value during a short time window.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
perform AIO-DIO and fcntl(F_SETFL) concurently. Unaligned AIO likely
results in synchronization which makes the race window wider.
Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
This tests whether the file or directory overwritten by rename is properly
removed (nlink is zero).
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Dave Chinner <david@fromorbit.com>
This test is motivated by a bug found in btrfs when replaying a
directory from the fsync log. The issue was that if a directory
entry is both found in the persisted metadata and in the fsync
log, at log replay time the directory got set with a wrong i_size.
This had the consequence of not being able to rmdir empty
directories (failed with errno ENOTEMPTY).
This was fixed in btrfs with the following linux kernel patch:
Btrfs: fix directory recovery from fsync log
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
XFS allocates extra indirect blocks for delayed allocation extents at
write time. When delalloc extents are split, the existing indirect block
reservation was historically divided up evenly among the new extents
even though the overall requirement for two extents could exceed the
requirement for the original. Repeated delalloc extent splits ultimately
leads to extents with 0 indirect blocks and in turn leads to assert
failures in XFS.
Add a test to stress indirect block reservation for delayed allocation
extents. The test converts a single delalloc extent to many and operates
on the remaining extents to detect or trigger potential problems.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
XFS had a data corruption problem where writeback of pages to unwritten
extents would fail to run unwritten extent conversion at I/O completion.
This causes subsequent reads of written, but unconverted regions to
return zeroes. This occurs on sub-page block size filesystems when
writeback contends for the inode lock (e.g., with a file writer).
Add a test that creates the conditions to reproduce the data corruption
and detect it by looking for unwritten extents after all said extents
have been overwritten.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
This test exposed a problem with XFS where it failed to write back a
partial page correctly during a fcollapse operation. This left a
stray dirty buffer on the page, and hence invalidation of the page
then failed of the fcollapse returned an EBUSY error.
Make this a generic test so that we can ensure that all filesystems
handle the case correctly. Test case originally worked out and
written by Brian Foster.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
This test exposed a problem with mapped writes to the tail page of a
file in XFS and potentially ext4. Eric did all the hard work of
taking the bug report and generating the reproducable test case on
ext4. Make it a generic test so that we can ensure that all
filesystems handle the case correctly.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
This test exposed a problem with mapped writes to the tail page of a
file in XFS. Hence make it a generic test so taht we can ensure that
all fielsystems handle the case correctly.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
This test is motivated by a btrfs issue where a ranged fsync would
prevent a subsequent fsync from persisting any extents that were
dirty at the time of the first fsync but that were outside the range
of that first fsync (which should have been persisted by the second
fsync).
This bug in btrfs is fixed by the following linux kernel patch:
Btrfs: fix fsync data loss after a ranged fsync
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Check defrag utility on file with large number of fragments, from 20 to 2000.
2000 fragments means that b+tree will have at least two index blocks.
Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
By closing the file descriptor before calling io_destroy, you pretty
much guarantee that the last put on the ioctx will be done in interrupt
context (during I/O completion). This behavior has unearthed bugs in
the kernel in several different kernel versions, so let's add a test to
poke at it.
The original test case was provided by Matt Cross. He has graciously
relicensed it under the GPL v2 or later so that it can be included in
xfstests. I've modified the test a bit so that it would generate a
stable output format and to run for a fixed amount of time.
Signed-off-by: Jeff Moyer <jmoyer@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
The following kernel commit introduced a race condition that causes
getcwd(2) to return "/" instead of correct path
232d2d6 dcache: Translating dentry into pathname without taking rename_lock
Jan Stancek hit it once when building ltp and Mikulas Patocka could
hit it by running lvm2 test suite. Please refer to this thread
https://www.mail-archive.com/ltp-list@lists.sourceforge.net/msg17896.html
These commits fixed the bug
ede4ceb prepend_path() needs to reinitialize dentry/vfsmount/mnt on restarts
f650080 __dentry_path() fixes
Cc: Artem Savkov <asavkov@redhat.com>
Cc: Jan Stancek <jstancek@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Run 8 processes writing 1k files to seperate files in seperate dirs to
hit ENOSPC on small fs with little free space. Loop for 100 iterations.
Regression test for
34cf865 ext4: fix deadlock when writing in ENOSPC conditions
Signed-off-by: Eryu Guan <eguan@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Having just removed the largeacl test from the shared ACL test,
reintroduce the same test as an generic test so that we can
handle the different limits in supported ACL count appropriately
across different filesystems and different configurations within
filesystem types.
Filesystems have to add support to _acl_get_max to run
this test - the default behaviour right now is to throw a
notrun error like this:
generic/026 14s ... [not run] ext4 does not define maximum ACL count
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Namjae Jeon
Filipe Manana
Jaegeuk Kim
Dmitry Monakhov
Miklos Szeredi
Brian Foster
Dave Chinner
Dave Chinner
Jeff Moyer
Eryu Guan