David Disseldorp [Thu, 12 Jan 2017 17:18:48 +0000 (18:18 +0100)]
common: remove unused XFSPROGS_VERSION check
This was only used to check for mkfs.xfs -n version=ci support,
which is carried in xfsprogs >= 2.10. _require_xfs_mkfs_ciname() is
now used to explicitly check for the mkfs parameter.
Eryu Guan [Fri, 16 Dec 2016 04:14:14 +0000 (12:14 +0800)]
Revert "generic/38[3-6]: require project quota to be enabled on SCRATCH_DEV"
Commit 23f60ef304d ("generic/38[3-6]: require project quota to be
enabled on SCRATCH_DEV") introduced a regression that leads
_require_projquota, which uses src/feature to call a quotactl,
operate before the filesystem is mounted, and results in tests not
run on XFS and ext4.
Revert the problematic patch first, because don't want to lose
project quota test coverage on XFS and ext4. We can fix the false
failure on ext3/2 later,
Reported-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Eric Sandeen <sandeen@redhat.com> Signed-off-by: Eryu Guan <eguan@redhat.com>
Roman Pen [Wed, 11 Jan 2017 09:52:39 +0000 (10:52 +0100)]
generic: reproduce ext4 bugs in a shift extents logic
Regression test which targets two nasty ext4 bugs in a logic which
shifts extents:
1) 14d981f468a1 ("ext4: Include forgotten start block on fallocate insert range")
Test tries to insert many blocks at the same offset to reproduce
the following layout on ext4:
block #0 block #1
|ext0 ext1|ext2 ext3 ...|
^
insert of a new block
Because of an incorrect range first block is never reached,
thus ext1 is untouched, resulting to a hole at a wrong offset:
What we got:
block #0 block #1
|ext0 ext1| ext2 ext3 ...|
^
hole at a wrong offset
What we expect:
block #0 block #1
|ext0 ext1|ext2 ext3 ...|
^
hole at a correct offset
2) 2b3864b32403 ("ext4: do not polute the extents cache while shifting extents")
Extents status tree is filled in with outdated offsets while doing
extent shift, that leads to wrong data blocks. That's why md5sum
of a result file is being checked after each block insert.
Amir Goldstein [Mon, 9 Jan 2017 13:39:11 +0000 (15:39 +0200)]
xfs: test handling of invalid inode modes
Set all possible file type values for different types of files
and verify that xfs_repair detects the correct errors.
When setting invalid file type values (e.g. core.mode = 0170644),
all files are expected to have been junked by xfs_repair.
When setting valid file type values to non matching file types,
xfs_repair would either detect wrong format and junk the file, e.g.:
would have junked entry "DATA" in directory PARENT_INO
or detect a ftype mismatch error, e.g.:
would fix ftype mismatch (5/3) in directory/child PARENT_INO/FIFO_INO
If ftype feature is enabled, when setting file type to one of the
special types (i.e. FIFO(1), CHRDEV(2),BLKDEV(6),SOCKET(14)),
xfs_repair is expected to detect ftype mismatch error. Otherwise,
xfs_repair is not expected to detect ftype mismatch error.
Darrick J. Wong [Mon, 9 Jan 2017 20:55:18 +0000 (12:55 -0800)]
xfs/ext4: check negative inode size
Craft a malicious filesystem image with a negative inode size,
then try to trigger a kernel DoS by appending data to the file.
Ideally this should trigger verifier errors instead of hanging.
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>
Darrick J. Wong [Thu, 5 Jan 2017 01:05:08 +0000 (17:05 -0800)]
reflink: make error reporting consistent when simulating EIO
When we're using dm-error to simulate failed devices, we don't really
know if the write or the fdatasync is going to receive the EIO. For
tests that make a single (failed) write attempt and never retry, it's
sufficient to check that the file md5 doesn't change after recovery.
For tests that /do/ retry the write, we should capture the entire output
and just look for the word error instead of enshrining the exact perror
message (filename/function call and everything) in the golden output.
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>
Darrick J. Wong [Mon, 9 Jan 2017 20:53:35 +0000 (12:53 -0800)]
reflink: fix quota tests to work properly
Fix the reflink quota tests to su to the fsgqa user so that we actually
test enforcement of quotas. Seems that XFS enforces user quotas even
if root is writing to a user file, whereas everything else lets root
writes through. Also clean up some of the variable usage and
_require_user.
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>
Darrick J. Wong [Thu, 5 Jan 2017 01:04:55 +0000 (17:04 -0800)]
ocfs2/reflink: fix file block size reporting
Some of the reflink tests try to require a specific filesystem block
size so that they can test file block manipulation functions. That's
straightforward for most filesystems but ocfs2 throws in the additional
twist that data fork block mappings are stored in units of clusters, not
blocks, which causes these reflink tests to fail.
Therefore, introduce a new helper that retrieves the file minimum block
size and adapt the reflink tests to use that instead.
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>
Eric Sandeen [Mon, 9 Jan 2017 16:49:21 +0000 (10:49 -0600)]
xfs/010: explicitly allow corrupted write in xfs_db
xfs/010 wants to write corruption and test how xfs_repair
deals, but when:
xfs: forbid AG btrees with level == 0
is merged to userspace, this new test fails the write verifier
in xfs_db.
Add "-c" to allow the corrupted write, do the corruptions all
in one xfs_db command (so it doesn't have to re-read the
corrupted data on 2nd startup), and filter out the
"Allowing write of corrupted data and bad CRC"
output from the "write -c" command.
Signed-off-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Brian Foster <bfoster@redhat.com> Signed-off-by: Eryu Guan <eguan@redhat.com>
Brian Foster [Fri, 6 Jan 2017 19:39:30 +0000 (14:39 -0500)]
generic: racing getxattr requests against xattr add/remove
This test reproduces a bug in XFS where a getxattr of an existing
xattr returns failure due to a race with a setxattr that causes
inode attribute fork conversion.
Eryu Guan [Fri, 30 Dec 2016 03:34:19 +0000 (11:34 +0800)]
generic/158,304: filter dedupe error message
Kernel commit 22725ce4e4a0 ("vfs: fix isize/pos/len checks for reflink &
dedupe") added more checks on reflink and dedupe, rejected dedupe past
EOF early and explicitly, and causes generic/158 and generic/304 to fail.
Try dedupe from past EOF
-dedupe: Invalid argument
+XFS_IOC_FILE_EXTENT_SAME: Invalid argument
Try dedupe to past EOF, destination offset beyond EOF
Also there's an xfsprogs patch from Darrick ("xfs_io: prefix dedupe
command error messages consistently") to change all xfs_io dedupe
error message prefixes to "XFS_IOC_FILE_EXTENT_SAME".
So introduce a new _filter_dedupe_error, change all "dedupe" prefix
to XFS_IOC_FILE_EXTENT_SAME, make tests pass with both old/new
kernel & userspace.
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Eryu Guan <eguan@redhat.com>
Xiong Zhou [Wed, 4 Jan 2017 08:59:11 +0000 (16:59 +0800)]
overlay: copy up and remove from another namespace
Via unshare, copy up dir from lower dir then remove
them in another unshared namespace.
This would fail before v4.7 kernel with EPERM.
Kernel commit 3fe6e52f
ovl: override creds with the ones from the superblock mounter
fixed this issue, with this reproducer provided. Original
reproducer was written by commit author:
Antonio Murdaca <amurdaca@redhat.com>
Amir Goldstein [Wed, 21 Dec 2016 13:50:21 +0000 (15:50 +0200)]
generic: test correct d_type values
Verify correct d_type values of dir entries.
This test does NOT require that file system support the filetype
feature. It verifies that either all file types are reported as
DT_UNKNOWN or that all file types are reported correctly.
For fs for which we know how to test the filetype feature (xfs|ext*)
verify getting DT_UNKNOWN IFF filetype feature is disabled.
Special dir entries . and .. MAY be reported as DT_UNKNOWN IF
filetype feature is disabled (ext4), but MAY also be reported as
DT_DIR in this case (xfs).
Theodore Ts'o [Thu, 22 Dec 2016 19:53:46 +0000 (14:53 -0500)]
overlay: stress test changes to top and bottom layers simultaneously
Introduce a test which runs fsstress on the top and bottom overlayfs
directories simultaneously to find potential races that plagued
wrapfs derived file systems.
[eguan: add copyright info and fix minor code style issue]
Eric Sandeen [Sat, 24 Dec 2016 10:07:45 +0000 (18:07 +0800)]
generic: test GETNEXTQUOTA near INT_MAX
XFS kernel code had a bug where GETNEXTQUOTA-type
quotactls requesting an ID near UINT_MAX could overflow
and return 0 as the "next" active ID.
This test checks that by creating an active quota near
UINT_MAX, then asking for the next one after it.
The proper answer is ENOENT, but if we wrap we'll return
ID 0.
This also changes test-nextquota.c so that it checks
both GETNEXTQUOTA and XGETNEXTQUOTA even if one fails;
it stores the failure conditions and returns 1 if either
of them fails.
Eric Biggers [Thu, 15 Dec 2016 20:26:24 +0000 (12:26 -0800)]
generic: test enforcement of one encryption policy per tree
Add an xfstest which verifies that the filesystem forbids operations
that would violate the constraint that all files in an encrypted
directory tree use the same encryption policy.
Eric Biggers [Thu, 15 Dec 2016 20:26:20 +0000 (12:26 -0800)]
encrypt: add utilities for testing filesystem encryption
Add utility functions for testing filesystem-level encryption via
the common API currently supported by ext4 and f2fs, in development
for ubifs and planned for xfs. Setting and getting encryption
policies will use new commands being added to xfs_io, while adding
and removing encryption keys will use keyctl.
Eryu Guan [Sun, 4 Dec 2016 16:15:08 +0000 (00:15 +0800)]
fstests: teach _scratch_mkfs to handle mkfs option conflicts
Currently in _scratch_mkfs only xfs and ext4 could handle the mkfs
failure caused by conflicts between $MKFS_OPTIONS and mkfs options
specified by tests, because of _scratch_mkfs_xfs and
_scratch_mkfs_ext4. This is a very useful functionality that allows
tests to specify mkfs options safely and to test specific fs
configurations, without worrying about mkfs failures caused by these
options.
Now teach _scratch_mkfs to handle such mkfs option conflicts for
other filesystems too, i.e. mkfs again only with mkfs options
specified by tests. Also add the ability to filter unnecessary
messages from mkfs stderr.
Also update some btrfs tests to throw away _scratch_mkfs stdout,
because previously _scratch_mkfs did this for btrfs.
Reviewed-by: Brian Foster <bfoster@redhat.com> Signed-off-by: Eryu Guan <eguan@redhat.com>
Amir Goldstein [Tue, 13 Dec 2016 05:12:37 +0000 (07:12 +0200)]
overlay: test hardlink breakage on copy up
Introduce a new test to demonstrate a known issue with overlayfs:
- file A and B are hardlinked in lower
- modify A to trigger copy up
- file A is no longer a hardlink of file B
Eric Biggers [Tue, 13 Dec 2016 06:57:01 +0000 (22:57 -0800)]
generic/062: don't assume same readdir order after re-creating directory
generic/062 uses getfattr to dump xattrs for a directory tree, then
deletes and recreates that directory tree, then dumps the xattrs
again and compares the dump to the original. This was failing when
run on ext4 with encryption enabled because getfattr's output is in
readdir order, but ext4 encryption by design chooses unpredictable
encrypted filenames for each new directory, causing the readdir
order to change after backup and restore. It is not really a valid
assumption that the readdir order will always be the same, so update
the test to sort the filenames, removing this assumption.
Signed-off-by: Eric Biggers <ebiggers@google.com> Reviewed-by: Brian Foster <bfoster@redhat.com> Signed-off-by: Eryu Guan <eguan@redhat.com>
Amir Goldstein [Tue, 13 Dec 2016 06:18:29 +0000 (08:18 +0200)]
overlay: test unstable inode number
Introduce a new test to demonstrate a known issue with overlayfs:
- stat file A shows inode number X
- modify A to trigger copy up
- stat file A shows inode number Y != X
Also test if d_ino of readdir entries changes after copy up.
There is a POC patch by Miklos Szeredi that fixes this issue.
Not dealing with hardlinks here, because there is more to test
then stable inode number.
Hardlinks will get a test of their own.
Wang Xiaoguang [Wed, 2 Nov 2016 04:07:27 +0000 (12:07 +0800)]
generic: make 17[1-4] work well when btrfs compression is enabled
When enabling btrfs compression, original codes can not fill fs
correctly, here we introduce _fill_fs() in common/rc, which'll keep
creating and writing files until enospc error occurs. Note _fill_fs
is copied from tests/generic/256, but with some minor modifications.
Darrick J. Wong [Sun, 11 Dec 2016 21:53:22 +0000 (13:53 -0800)]
reflink: don't test disjoint block sharing sets
Unlike xfs/btrfs which store refcounting information as part of the
space metadata, ocfs2 implements block sharing (reflink) by creating
refcount btrees that are shared between subsets of files. Effectively,
this means that a ocfs2 can have multiple disjoint sets of files that
share blocks, which also means that blocks cannot be reflinked between
two disjoint refcounted-file-sets. generic/119 tests the ability to do
this, so we cannot run it for ocfs2. Create a _require helper to check
for this.
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>
Amir Goldstein [Thu, 8 Dec 2016 10:52:20 +0000 (12:52 +0200)]
common: implement _require_xfs_io_command "open"
-c "open $f" is broken in xfs_io <= 4.8. Using it results
in an endless loop and xfs_io exists with error EMFILE.
A fix for that bug makes "open" a 'one shot' command.
Along with this fix, a new -C flag was introduced to explicitly
request to execute a one shot command.
Check for -C flag support as an indication that -c "open $f"
is not broken.
Qu Wenruo [Fri, 9 Dec 2016 01:19:59 +0000 (09:19 +0800)]
btrfs: Use _require_btrfs_qgroup_report to replace open code
Introduce new _require_btrfs_qgroup_report function, which will
check the accessibility to "btrfs check --qgroup-report", then set a
global flag to info _check_scratch_fs() to do extra qgroup check.
Qu Wenruo [Thu, 8 Dec 2016 08:12:12 +0000 (16:12 +0800)]
common/btrfs: rename and enhance _require_btrfs to _require_btrfs_command
Rename _require_btrfs() to _require_btrfs_command() to avoid
confusion, as all other _require_btrfs_* has a quite clear suffix,
like _require_btrfs_mkfs_feature() or _require_btrfs_fs_feature().
Also enhance _require_btrfs_command() to accept 2nd level commands
or options.
Options will be determined by the first "-" char. This is quite
useful for case like "btrfs inspect-internal dump-tree" and "btrfs
check --qgroup-report".
[eguan: make local vars in _require_btrfs_command "local"]
Ross Zwisler [Wed, 7 Dec 2016 04:04:01 +0000 (21:04 -0700)]
build: fix build warnings and notify_others() bug
This patch addresses the following build warnings:
fsx.c: In function 'do_punch_hole':
fsx.c:940:3: warning: this 'if' clause does not guard... [-Wmisleading-indentation]
if (!quiet && testcalls > simulatedopcount)
^~
fsx.c:942:4: note: ...this statement, but the latter is misleadingly indented as if it is guarded by the 'if'
log4(OP_PUNCH_HOLE, offset, length, FL_SKIPPED);
^~~~
fsx.c:947:3: warning: this 'if' clause does not guard... [-Wmisleading-indentation]
if (!quiet && testcalls > simulatedopcount)
^~
fsx.c:949:4: note: ...this statement, but the latter is misleadingly indented as if it is guarded by the 'if'
log4(OP_PUNCH_HOLE, offset, length, FL_SKIPPED);
^~~~
fsx.c: In function 'do_zero_range':
fsx.c:995:3: warning: this 'if' clause does not guard... [-Wmisleading-indentation]
if (!quiet && testcalls > simulatedopcount)
^~
fsx.c:997:4: note: ...this statement, but the latter is misleadingly indented as if it is guarded by the 'if'
log4(OP_ZERO_RANGE, offset, length, FL_SKIPPED |
^~~~
[CC] growfiles
growfiles.c: In function 'notify_others':
growfiles.c:1458:6: warning: this 'if' clause does not guard... [-Wmisleading-indentation]
if ( Forker_pids[ind] != Pid )
^~
growfiles.c:1462:10: note: ...this statement, but the latter is misleadingly indented as if it is guarded by the 'if'
kill(Forker_pids[ind], SIGUSR2);
^~~~
The warnings in fsx.c were just spacing issues of the form:
if (length == 0) {
if (!quiet && testcalls > simulatedopcount)
prt("skipping zero length punch hole\n");
log4(OP_PUNCH_HOLE, offset, length, FL_SKIPPED);
return;
}
Where the log4() call just needs to be unindented. log4() calls
elsewhere in that same file are not protected with any sort of
'quiet' check, and commonly follow prt() calls which are. See
doread(), domapread(), etc.
The warning from growfiles.c was actually a bug. notify_others() is
looping through the Forker_pids[] array and sending SIGUSR2 to all
other processes. However, with the current logic it only *logs* the
kill for other processes, and kills all other processes plus the
Forker_pids[] entry that matches 'Pid'.
Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Lukas Czerner <lczerner@redhat.com> Cc: Allison Henderson <achender@vnet.ibm.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Nathan Scott <nathans@sgi.com> Reviewed-by: Eryu Guan <eguan@redhat.com> Signed-off-by: Eryu Guan <eguan@redhat.com>
Omar Sandoval [Tue, 6 Dec 2016 01:01:28 +0000 (17:01 -0800)]
generic/35[67]: disable swapfile tests on Btrfs
Btrfs doesn't support swapfiles (yet?), so generic/356 fails
erroneously, and generic/357 only passes by accident. Let's add a
_require_scratch_swapfile helper and add it to these tests.
Signed-off-by: Omar Sandoval <osandov@fb.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Eryu Guan <eguan@redhat.com>
Eric Biggers [Wed, 30 Nov 2016 22:31:42 +0000 (14:31 -0800)]
build: allow building with latest xfslibs on 32-bit Linux
The next version of xfslibs will require that its users enable
transparent largefile support. This broke building xfstests on
32-bit Linux because _FILE_OFFSET_BITS=64 was not getting defined
when testing for xfs headers. Fix this by defining
_FILE_OFFSET_BITS=64 when doing configure checks for the xfs
headers.
It was not necessary to define _FILE_OFFSET_BITS=64 for the actual
build because it's already done, in builddefs.in.
The problem could also be solved by adding AC_SYS_LARGEFILE to
configure.ac. It's debatable, but I chose not to do that because it
might give the impression that it affects the actual compilation of
the test programs and not just the configure checks. (Since not all
.c files include config.h first, a _FILE_OFFSET_BITS definition in
config.h wouldn't always take effect.)
Dave Chinner [Tue, 29 Nov 2016 21:32:32 +0000 (08:32 +1100)]
common: split XFS functions from common/rc
common/rc has become a dumping ground for common functions that
don't have a specific topic file. It's getting huge and difficiult
to manage, and people are duplicating functionality because they
can't easily find existing functions in the mess.
Let's start to make it a little easier to manage by splitting out
the XFS specific functions into common/xfs and source that from
common/rc automatically. Other filesytems can follow suit in
future, leaving us with only generic functionality in common/rc.
Signed-Off-By: Dave Chinner <dchinner@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Eryu Guan <eguan@redhat.com>
Dave Chinner [Wed, 30 Nov 2016 04:59:08 +0000 (15:59 +1100)]
generic/095: add iomap_dio_rw() to the dmesg filter
Oops, we did it again - we moved the XFS dio code warning to a new
function with the iomap direct IO rework. Update the filter to pick
up the new warning location.
Signed-Off-By: Dave Chinner <dchinner@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Eryu Guan <eguan@redhat.com>
Dave Chinner [Tue, 29 Nov 2016 02:48:09 +0000 (13:48 +1100)]
check: add brief result summaries
When running multiple sections and hundreds of tests in each config
section, it's hard to see what failed from the summary output
because of the hundreds of tests listed in the "tests run" and
"tests not run" output. Add a "-b" option for brief result summaries
that only output the tests that failed and the summary count of
tests failed.
Jan Kara [Tue, 22 Nov 2016 10:16:35 +0000 (11:16 +0100)]
ext4/001: Fixup file layout with DAX
Since commit 9538d1b69399 (ext4: avoid split extents for DAX
writes), currently sitting in ext4 tree, the block layout of files
has changed when the filesystem is mounted with -o dax and now
matches 'nodelalloc' variant instead of 'nozero'. Fixup the test
configuration.
Signed-off-by: Jan Kara <jack@suse.cz> Reviewed-by: Eryu Guan <eguan@redhat.com> Signed-off-by: Eryu Guan <eguan@redhat.com>
Amir Goldstein [Thu, 24 Nov 2016 20:12:12 +0000 (22:12 +0200)]
overlay: test ro/rw fd data inconsistecies
Introduce a new test to demonstrates a known issue with overlayfs:
- process A opens file F for read
- process B writes new data to file F
- process A reads old data from file F
This issue is about to be fixed with a patch set by Miklos Szeredi.
Jaegeuk Kim [Fri, 18 Nov 2016 19:36:51 +0000 (11:36 -0800)]
generic: check some inline truncation behaviors
This patch introduces a new testcase to test some small truncations
to check inline_data and its cached data are truncated correctly at
the same time.
The inline_data feature was introduced in ext4 and f2fs as follows.
ext4 : http://lwn.net/Articles/468678/
f2fs : http://lwn.net/Articles/573408/
The basic idea is embedding small-sized file's data into relatively
large inode space.
In ext4, up to 132 bytes of data can be stored in 256 bytes-sized inode.
In f2fs, up to 3.4KB of data can be embedded into 4KB-sized inode block.
Jaegeuk Kim [Mon, 21 Nov 2016 20:02:49 +0000 (12:02 -0800)]
generic: check inode metadata on f{data}sync after power-cut
This patch adds case to test fsync and fdatasync with power-cuts.
The rule to check is:
1) fsync should guarantee all the inode metadata after power-cut,
2) fdatasync should guarantee i_size and i_blocks at least after
power-cut.
Note that, in XFS, it allocates more blocks when doing writes, so it
should close the file before fsync in order to get actual i_blocks
after power-cut. Or, it needs to do truncate the file with a
specific size to turn it off at the beginning.
Omar Sandoval [Thu, 17 Nov 2016 00:29:34 +0000 (16:29 -0800)]
generic: concurrent non-overlapping direct I/O on the same extents
There have been a couple of logic bugs in `btrfs_get_extent()` which
could lead to spurious -EEXIST errors from read or write. This test
exercises those conditions by having two threads race to add an
extent to the extent map.
This is fixed by Linux commit 8dff9c853410 ("Btrfs: deal with
duplciates during extent_map insertion in btrfs_get_extent") and the
patch "Btrfs: deal with existing encompassing extent map in
btrfs_get_extent()"
(http://marc.info/?l=linux-btrfs&m=147873402311143&w=2).
Although the bug is Btrfs-specific, nothing about the test is.
Liu Bo [Wed, 16 Nov 2016 19:17:03 +0000 (11:17 -0800)]
generic/098: update test for truncating a file into the middle of a hole
This updates generic/098 by adding a sync option, i.e. 'sync' after
the second write, and with btrfs's NO_HOLES, we could still get
wrong isize after remount.
This gets fixed by the patch
'Btrfs: fix truncate down when no_holes feature is enabled'
Signed-off-by: Liu Bo <bo.li.liu@oracle.com> Reviewed-by: Eryu Guan <eguan@redhat.com> Signed-off-by: Eryu Guan <eguan@redhat.com>
Eryu Guan [Sat, 12 Nov 2016 18:12:33 +0000 (02:12 +0800)]
generic: fix $tmp path in tests
Some tests use "tmp/$$" as $tmp, and this would result in ENOENT
failure if $tmp is referenced in helpers or tests, because we have
no "tmp" dir in pwd.
Fix them to use "/tmp/$$" as all other tests do.
Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Eryu Guan <eguan@redhat.com>
Eryu Guan [Sat, 12 Nov 2016 17:22:26 +0000 (01:22 +0800)]
generic/38[3-6]: require project quota to be enabled on SCRATCH_DEV
generic/38[3-6] are made generic from xfs tests to test project
quota on other filesystems, e.g. ext4. But we should make sure
project quota is enabled in between of _scratch_enable_pquota and
_qmount. Otherwise mount will fail if there's no project quota
support on SCRATCH_DEV, e.g. ext2 and ext3.
Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Eryu Guan <eguan@redhat.com>
Many of the commands using fallocate had an extra % at the beggining
of the error messages, missed the 0x prefix for hex output and printed
the length where it should print the end of the region.
Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Eryu Guan <eguan@redhat.com>
Eric Sandeen [Sat, 5 Nov 2016 02:00:09 +0000 (21:00 -0500)]
xfs: testcase for kernelspace xfs_fsr extent handling flaw
This is a testcase for a bug which goes way back; googling
"xfs_trans_log_inode NULL pointer dereference" yields sporadic
reports over several years.
The test sets up several two-extent files with speculative
preallocation on them, and then runs xfs_fsr. The kernelside
code ignores the preallocation, and therefore sets up the
temporary inode incorrectly after the inode fork swap.
It is a "dangerous" test because the extent mishandling on
the temporary inode causes a null pointer dereference and
oops when the inode's i_itemp pointer gets overwritten
and we blow up in logging code that tries to use it.
Eryu Guan [Fri, 4 Nov 2016 16:16:10 +0000 (00:16 +0800)]
xfs/021: filter out detailed array element info from xfs_db
Starting from xfsprogs commit 384283555871 ("xfs_db: print one array
element per line"), xfs_db prints one array element per line. This
breaks the filter in xfs/021, which now fails as:
Zorro Lang [Tue, 8 Nov 2016 05:33:50 +0000 (13:33 +0800)]
fstests: test xfs_copy V5 XFS without -d option
Before xfsprogs commit a872b62 (xfs_copy: band-aids for CRC
filesystems), xfs_copy requires the "-d" option to copy a V5 XFS,
because it can't rewrite the UUID of V5 XFS properly.
Now xfs_copy already full support to copy a V5 XFS. But for above
old problem, xfstests use below patch to make sure xfs_copy always
use "-d" option to copy a V5 XFS:
8346e53 common: append -d option to XFS_COPY_PROG when testing v5 xfs
That cause xfstests miss the coverage of copying a V5 XFS without
"-d". For test this feature I did below things:
1. Changed init_rc(), add "-d" to $XFS_COPY_PROG if xfs_copy can't
copy a V5 XFS properly.
2. xfs/073 test V4 xfs forcibly by specify "-m crc=0" in case. I
think it's useless now, so remove it.
3. Changed xfs/032. If xfs_copy full support to copy a V5 XFS, test
with and without "-d" option, or only test with "-d" option.
Zorro Lang [Mon, 7 Nov 2016 06:30:12 +0000 (14:30 +0800)]
generic: multi-thread freeze/unfreeze test under fsstress
Multi-threads freeze/unfreeze maybe trigger some bugs, e.g: panic,
hang or data corruption etc. It does't check the return value of
freeze/unfreeze, but it tries to make sure fsstress won't run fails,
and no any other bugs happen.
[eguan: add comment on why we check _scratch_mount failure]
Eryu Guan [Sun, 6 Nov 2016 17:43:33 +0000 (01:43 +0800)]
fstests: make xfs_io mandatory
_test_inode_flag() and _test_inode_extsz() use "which $XFS_IO_PROG"
to check if xfs_io command is available. And "-i" option was added
to XFS_IO_PROG varibable by commit 54659ecdb575 ("fstests: run
xfs_io with -i option if supported"). So the command becomes "which
/usr/sbin/xfs_io -i", and it stops and waits for input from stdin,
because "-i" option of "which" means "Read aliases from stdin".
I've seen xfs/008 hang when testing with latest xfsprogs, where
xfs_io has "-i" support.
Fix it by removing the xfs_io command detections, and making xfs_io
mandatory in common/config.
Also fix the indentions in these functions, use tab instead of
spaces, while we're at it.
Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Eryu Guan <eguan@redhat.com>
Dave Chinner [Wed, 2 Nov 2016 21:15:25 +0000 (08:15 +1100)]
filter: xfs_io output has dropped "64" from error messages
Upstream xfs_io has been converted to always use LFS compliant
(i.e. 64 bit) pwrite() rather than pwrite64(). Similar changes have
been made for multiple syscalls that have "*64" variants. hence the
error output of all these commands has changed, such as "pwrite64:
..." to "pwrite: ....".
Make a filter to catch the *64 variants and strip it, and
convert all the golden output to use the non-*64 variant. This will
make all golden output matching work correctly regardless of what
version of xfs_io is in use.
Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Eryu Guan <eguan@redhat.com>
Try to discover the allocation size. For NFS, the st_blksize is
optimized for the network, and will usually be much larger than the
allocation size of the exported filesystem, which may trigger
preallocation strategies in the exported filesystem causing the seek
tests here to fail.
Eryu Guan [Sat, 22 Oct 2016 05:27:33 +0000 (13:27 +0800)]
xfs/054: only filter ROOT_INO at the beginning
In the test ROOT_INO is filtered out and/or replaced, but if
ROOT_INO is also 32, more "32"s are filtered and replaced than
expected. This happens to me when testing 512B block size XFS and 1k
block size CRC enabled XFS.
Fix it by filtering out only ROOT_INO at the beginning of a line,
and removing all "g" modifiers in sed expressions.
Also the ROOT_INO should be the root inode number of TEST_DIR not
SCRATCH_MNT.
Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Eryu Guan <eguan@redhat.com>
Amir Goldstein [Wed, 19 Oct 2016 08:52:33 +0000 (11:52 +0300)]
overlay/013: adjust test expectation to kernel v4.8
Commit 4d0c5ba (vfs: do get_write_access() on upper layer of overlayfs)
that was merged for kernel v4.8 changes the behavior of execve+truncate
in a subtle way.
When file is executed from lower and then truncated, the truncation is
allowed, becasue it only affects the size of the copied up file and does
not affect the VM mapping of the executed lower file.
Amir Goldstein [Wed, 19 Oct 2016 08:52:31 +0000 (11:52 +0300)]
t_truncate_self: describe self better in error message
Include the name of the program that is trying to trucate itself
to error messages to distinguish between output for different
test cases (e.g. test_lower and test_upper in overlay/013).
Xiao Yang [Wed, 19 Oct 2016 05:33:07 +0000 (13:33 +0800)]
xfs: remove the device parameter from __populate_check_xfs_dir
In commit c08ebd092 ("xfs: fix $XFS_DB_PROG usage"), the need for
callers to pass the device to populate into
__populate_check_xfs_dir() was removed. So we can now clean up all
the callers by removing the device parameter.
Signed-off-by: Xiao Yang <yangx.jy@cn.fujitsu.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Eryu Guan <eguan@redhat.com>
Darrick J. Wong [Mon, 17 Oct 2016 22:09:28 +0000 (15:09 -0700)]
fstests: fix fallocate commands that need the unshare switch
We re-added the UNSHARE flag to fallocate, so go make sure that all
the unshare tests actually check that the installed copy of xfs_io
supports the 'funshare' command and that the underlying filesystem
understands the flag, and change the tests themselves to use
funshare.
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>
Darrick J. Wong [Mon, 17 Oct 2016 22:09:21 +0000 (15:09 -0700)]
xfs: set the refcount btree cowflag when setting up cow
When we're setting up a fake cow extent in the refcountbt to test
cleanup of fake cow extents, set the cowflag in the record field
to reflect our new disk format of storing the staging extents in
the right side of the tree.
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>
projects / xfstests-dev.git / log