setup: add section support

Make the setup command section aware so that it is easy to test
whether the section config code is generating the correct
configurations or not.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

xfs: support realtime/log device setup changes in config sections

Currently changing the devices used by "USE_EXTERNAL" environmental
variable is not supported by the config section parsing. Add the
functionality so that we can use config sections to test external
device configs successfully.

This required tracking down a bug in _check_xfs_filesystem() which
was causing a log device to be passed to a test device without an
external log device. This was caused by an uninitialised variable in
the function. I also added full output file removals to the first
couple of generic tests that were failing, because that's where the
check failure output ends up in this case.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

fstests: canonicalize the mount points by removing the trailing "/"

removing all the trailing "/"s of mounting point name by readlink

Suggested-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Jia He <hejianet@gmail.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

fstests: btrfs, verify fitrim does not discard reserved device area

Verify that when a fitrim operation is made against a btrfs filesystem,
the ranges [0, 64Kb[ and [68Kb, 1Mb[ of the device are not discarded,
they remain with the content they had before the fitrim operation. These
regions of the device are reserved for a boot loader to use at its will.

In the 4.3 linux kernel we got a regression that allowed a fitrim
operation to discard these reserved ranges of the device, resulting in
the filesystem becoming unbootable after a fitrim. The issue is fixed
by the following patch (targeted for 4.5 and 4.3/4.4 stable releases):

 "Btrfs: fix fitrim discarding device area reserved for boot loader's use"

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

xfs: test XFS torn log write detection

XFS torn log write detection includes a mechanism to inject CRC errors
into log records at runtime and shutdown the fs accordingly. This
ensures that the CRC verification pass on the subsequent mount discovers
an invalid record near the head of the log and considers it a torn
write.

This test runs a workload with error injection enabled and verifies that
the subsequent mount is successful. The test repeats for several
iterations using a random frequency factor for the error event each
time.

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

f2fs: add Makefile for f2fs

Installation failed due to Makefile omission, so fix this.

Signed-off-by: Xiao Yang <yangx.jy@cn.fujitsu.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

shared: test mount needs_recovery extN partition with noload option

Once upon a time, ext4 encounter NULL pointer dereference under this
situaiton due to jump to a wrong label. Part of this commit fixed
this Oops:

744692d ext4: use ext4_get_block_write in buffer write

We can also run this test on ext2/3.

This case is based on a script from Monakhov Dmitriy @ openvz.

Cc: Monakhov Dmitriy <dmonakhov@openvz.org>
Signed-off-by: Xiong Zhou <xzhou@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

common/scsi_debug: fix removal of scsi_debug module

Occasionally scsi_debug cannot be removed because it's still in use and
causes xfs/279 to fail.

Now dryrun the removal by modprobe firstly then do the real rmmod.

Signed-off-by: Eryu Guan <eguan@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

fstests: fix cleanup of test btrfs/003

If the test fails after removing a device and before adding it back, it
attempts to add back the device in its _cleanup() function. However this
is broken because the device identifier is stored in a variable local to
the function _test_replace() and not in a global variable. So make the
variable global instead of local.

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

fstests: cleanup test btrfs/031

The test was using $SCRATCH_MNT as a mountpoint for $SCRATCH_DEV, which
is counter intuitive and not expected by the fstests framework - this
made the test fail after commit 27d077ec0bda (common: use mount/umount
helpers everywhere). So rewrite the test to use the scratch device for
all data and use a test specific directory inside $TEST_DIR to use as a
mount point for a cross mount of $SCRATCH_DEV.

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

fstests: cleanup test btrfs/029

The test was using $SCRATCH_MNT as a mountpoint for $SCRATCH_DEV, which
is counter intuitive and not expected by the fstests framework - this
made the test fail after commit 27d077ec0bda (common: use mount/umount
helpers everywhere). So rewrite the test to use the scratch device for
all data and use a test specific directory inside $TEST_DIR to use as a
mount point for a cross mount of $SCRATCH_DEV.

This test was also overriding $seqres.full, through the redirect ">"
operator, if a call to cp failed. Fix that by using instead the operator
">>". Also make the test use the function _mount() instead of calling
the mount program directly.

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

fsx: Add mechanism to replay failed operations

Create a $name.fsxops file next to $test.fsxlog.  When a test fails,
dump the operations in the log into that file in a simple, parseable
format like:

  fallocate 0x2e0f2 0xf04a 0x0 keep_size
  truncate 0x0 0x11e00 0x0 *
  write 0x73400 0x6c00 0x11e00
  skip punch_hole 0x71539913 0xdf76 0x7a000 close_open
  mapread 0x56000 0x16d08 0x7a000

Here, each operation is on a separate line. When the first word is
"skip", the operation will be skipped.  The next parameters are offset,
length, and the current file size, followed by optional flags like
keep_size and clode_open.  A trailing asterisk indicates that the
operation overlaps with the operation that has failed.

Add a --replay-ops option that allows to replay the operations recorded
in such a $name.fsxops file.  (The log can be modified to easily narrow
down which operations are causing the failure.)

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

fsx: Improve operation logging

Add FL_SKIPPED, FL_CLOSE_OPEN, and FL_KEEP_SIZE flags to the log
entries.  Use FL_SKIPPED to indicate that an operation was skipped.  Use
FL_CLOSE_OPEN to encode when an operation is followed by a close/open
cycle.  Use FL_KEEP_SIZE to indicate when the OP_ZERO_RANGE and
OP_FALLOCATE operations should keep the file size unchanged and put the
current file size into args[2] so that we can tell which operation was
actually called from the log.

After that, arg2 of log4 is always either unused or the current file size, so
remove it and unconditionally remember the current file size.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

fsx: Generate test parameters in test()

Generate all test parameters in test(), including keep_size.

The code is slightly more complicated than it could be to produce the
same sequence of operations for the same random seed.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

fsx: Report number of successful operations

Report the number of successful operations at the end of an fsx run.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

fsx: Fix hex numbers in operation dump

Some hex numbers are prefixed with "0x" and right-aligned with spaces,
leading to output like "0x beef".  Make that "0x0beef" instead.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

fsx: Small improvements and fixes

Move the run_fsx shell function into common/rc.  Fix it to avoid
duplicate output on errors.  Write the actual fsx parameters used into
$seqres.full instead of the BSIZE and PSIZE placeholders.

Include the symbolic fallocate mode in fsx error messages instead of the
numeric value.  Use fprintf(stderr, ...) instead of warn() when
including strerror(errno) doesn't make sense.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

common/punch: handle bmap output for unaligned extents

xfs/242 fails if the mapping flags show unaligned extents;
fix up the regexp to allow this, we really only care about
the unwritten flag.

Signed-off-by: eric Sandeen <sandeen@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

generic: create a dedupe group

Create a group for just the deduplication tests.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

reflink: test a big CoW operation

Test what happens when we send largeish buffers to CoW all at once.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

reflink: test CoW with blocksize < pagesize

Test CoW operations when blocksize < pagesize and the only reflink
block is in the middle of the page.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

reflink: more CoW tests for reflink and dedupe

These tests examine the behavior of advanced and tricky copy on write
situations.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

reflink: test ENOSPC when expanding btrees during reflink operations

See what happens when we ENOSPC while growing a btree on behalf of
some reflink operation.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

reflink: more tests

Add more tests for unaligned copy-on-write things, and explicitly
test the ability to pass "len == 0" to mean reflink/dedupe all
the way to the end of the file".

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

xfs/129: require loop

Missed a _require_loop for xfs/129.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

xfs/128: allow larger margin for disk space usage

The free blocks count can vary from our calculations by up to 8% on a
1k-block filesystem, so permit that.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

generic/15[78]: fix error messages in the golden output

Fix the error messages in the golden output for generic/15[78], which
examine the responses to invalid inputs as returned by the
clone/clone_range/extent_same ioctls.  Also fix a filtering omission.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

tests: don't leave cruft behind on the TEST_MNT

Don't leave cruft behind on the test device's filesystem, so as to
avoid filling it with debris.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

test-scripts: test migration scripts

Add two scripts: "nextid" finds the next available test ID number in a
group, and "mvtest" relocates a test, fixes the golden output, and
moves the group entry for that test.

v2: sorting group files should preserve group order; nextid should use
the same algorithm as new; move both tools to tools/.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>

generic/079: _notrun if fs doesn't support ioctl

Overlayfs directory inode doesn't support ioctl and reports
"Inappropriate ioctl", so grep for this error message and _notrun if the
message is found.

Signed-off-by: Eryu Guan <eguan@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

generic/062: don't create WHITEOUT_DEV

overlay doesn't support creating WHITEOUT_DEV (major and minor device ID
are both 0). Create null device instead.

Signed-off-by: Eryu Guan <eguan@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

overlayfs: rudimentary test support

Adding basic overlayfs support to fstests, it doesn't test anything
overlayfs specific, but runs existing tests on top of overlayfs. It's
following the path from Eric's patchset and Zab's review back in Mar.

A new fstype "overlay" is added, and TEST_DEV/SCRATCH_DEV are required
to be fs paths, and overlayfs is mounted at TEST_DIR/SCRATCH_MNT, so
tests can be run there.

To test overlayfs, setup config as something like the following

TEST_DEV=/mnt/ovl/test
TEST_DIR=/mnt/testarea/test
SCRATCH_DEV=/mnt/ovl/scratch
SCRATCH_MNT=/mnt/testarea/scratch

then run

./check -overlay -g auto

Signed-off-by: Eryu Guan <eguan@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

common: use mount/umount helpers everywhere

Replace every explicit mount/umount of scratch or test devices with
helper functions. This allows the next patch to add in hooks to these
functions in order to set up & tear down overlayfs on every mount/umount

(also adds _test_unmount(), which didn't exist prior)

[Eryu Guan rebased the patch agains latest master and replaced more
mount/umount with helpers]

Signed-off-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Eric Sandeen <esandeen@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

xfs: test user and group quota names beginning with digits

There's a known bug of xfsprogs, when a user or group name beinning
with digits, xfs_quota can't create 'limit' for it.

Signed-off-by: Zorro Lang <zlang@redhat.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

xfs: test project quota name beginning with digits

There's a known bug of xfsprogs, when a project name beinning with
digits, it can't be found by run xfs_quota 'quota -p -v ...' command.

Signed-off-by: Zorro Lang <zlang@redhat.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

generic/275: add to the 'enospc' group

Currently we're developing a new in-band deduplication feature for btrfs,
when enabing this feature, it will take much long time to hit the enospc
condition which 275 tries to create. I think 275 is also certain enospc
test and we should add it to 'enospc' group, then we can skip it easily by
excluding 'enospc' group when running fstests.

Signed-off-by: Xiaoguang Wang <wangxg.fnst@cn.fujitsu.com>

f2fs/001: test for -ENOSPC with inline operation

This patch introduces f2fs/001 to test -ENOSPC when inline operation was
activated.

Inline_data feature is described in:
  http://lwn.net/Articles/573408/

Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

common: add mkfs options for f2fs

This patch adds an interface to give mkfs option for f2fs.

Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

common: support f2fs for _scratch_mkfs_sized

The mkfs.f2fs has an option to build a certain sized filesystem by giving
the number of sectors.

So, this patch adds to use that.

Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

xfs/293: tighten up checks for documented xfs_io commands

Some commands (like "zero") are simple words which commonly
occur in the manpage text even if they aren't documented as
commands.

Grep for "   $COMMAND" instead of the bare word, because
the documented commands show up as indented.

This reveals that the "zero" command is not documented yet.
(It catches "help" too, because it's documented differently;
I'll fix that up in the manpage when I add "zero").

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

xfs/242: remove open-coded check for "zero" command

We have helpers for this, no need to open-code it.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

dmflakey: don't run for dax mounts

There is no sense running dmflakey tests with the dax mount option,
since dmflakey doesn't support it.  Mark these as _notrun so that it is
clear that this type of testing is not happenning for dax.

Suggested-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Jeff Moyer <jmoyer@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

generic/042: remove from the 'auto' group

This test fails 100% of the time for me on xfs and current git head, and
is not run for ext4 since ext4 does not support shutdown.  After talking
with bfoster, it isn't expected to succeed right now.  Since the auto
group is for tests that *are* expected to succeed, let's move this one
out.

Signed-off-by: Jeff Moyer <jmoyer@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

defrag: loosen input file requirements

We don't have perfect control of file allocation for these tests;
in some cases we may fail to adequately fragment a file prior to
defragmentation testing, and today that will fail the test.

Attack this on 2 fronts:

1) Explicitly allow fewer extents on one of the input files in
   generic/018 where the allocator has discretion.
2) _notrun rather than _fail if we don't create enough extents;
   this is a defrag test, not an allocator/fragmentation test,
   so just skip the test if we can't create an acceptable file
   for defrag testing.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Acked-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Dave Chinner <david@fromorbit.com>

xfs: make sure to _require_attrs when necessary

xfs/073 and xfs/295 forgot to _require_attrs.  Fix it.

Signed-off-by: Jeff Moyer <jmoyer@redhat.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs/106: work on non-4k page sized machines

This commit makes use of the new _filter_xfs_io_pages_modified filtering
function to print information in terms of page size units rather than file
offset.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs/103: work on non-4k block sized filesystems

This commit makes use of the new _filter_xfs_io_blocks_modified filtering
function to print information in terms of file blocks rather than file
offset.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs/098: work on non-4k block sized filesystems

This commit makes use of the new _filter_xfs_io_blocks_modified filtering
function to print information in terms of file blocks rather than file
offset.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs/097: work on non-4k block sized filesystems

This commit makes use of the new _filter_xfs_io_blocks_modified filtering
function to print information in terms of file blocks rather than file
offset.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs/095: work on non-4k block sized filesystems

This commit makes use of the new _filter_xfs_io_blocks_modified filtering
function to print information in terms of file blocks rather than file
offset.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs/094: work on non-4k block sized filesystems

This commit makes use of the new _filter_xfs_io_blocks_modified filtering
function to print information in terms of file blocks rather than file
offset.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs/052:  work on non-4k block sized filesystems

This commit makes use of the new _filter_xfs_io_blocks_modified filtering
function to print information in terms of file blocks rather than file
offset.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

filter: filter xfs_io's output in units of page size

The helpers introduced in this commit will be used to make btrfs tests that
assume 4k as the page size to work on non-4k page-sized systems as well.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs/096: work on non-4k block sized filesystems

This commit makes use of the new _filter_xfs_io_blocks_modified filtering
function to print information in terms of file blocks rather than file
offset.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs/056: to work on non-4k block sized filesystems

This commit makes use of the new _filter_xfs_io_blocks_modified and _filter_od
filtering functions to print information in terms of file blocks rather than
file offset.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs/055: work on non-4k block sized filesystems

This commit makes use of the new _filter_xfs_io_blocks_modified and _filter_od
filtering functions to print information in terms of file blocks rather than
file offset.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs/017: work on non-4k block sized filesystems

This commit makes use of the new _filter_xfs_io_blocks_modified filtering
function to print information in terms of file blocks rather than file
offset.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

filter: Filter xfs_io and od's output in units of FS block size

The helpers introduced in this commit will be used to make btrfs tests that
assume 4k as the block size to work on non-4k blocksized filesystem instances
as well.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

xfs: verify invalid metadata LSN detection

XFS v5 superblock fs' use metadata LSN tracking to determine when an
on-disk structure was last written to disk. This is used to ensure log
recovery operates correctly after an unclean shutdown. To work
correctly, the on-disk metadata LSNs must always remain behind the
current LSN with respect to the log.

Historically, xfs_repair had a problem where it incorrectly formats the
log to an LSN that is potentially behind existing metadata LSNs. As
such, xfs_repair and the kernel have been updated to prevent, detect and
recover from the problem. Add a test that intentionally formats the log
incorrectly and verifies that the fs fails to mount and that xfs_repair
detects the invalid metadata LSNs.

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

xfs: test xfsprogs log formatting infrastructure

The xfsprogs libxfs layer implements its own log formatting code to
support utilities that might need to format the log, such as mkfs,
repair, metadump, etc. This code is fairly independent from kernel log
writing code. Therefore, add a test that reformats the log from
userspace with various supported log stripe unit alignments and verifies
that the end result is a correctly formatted log.

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

common: mkfs.btrfs now requires --mixed for small filesystems

generic/077 fails on btrfs progs v4.3:
 # ./check generic/077
 FSTYP         -- btrfs
 PLATFORM      -- Linux/x86_64 lenovo 4.4.0-rc2_HEAD_1ec218373b8ebda821aec00bb156a9c94fad9cd4_
 MKFS_OPTIONS  -- /dev/sdb6
 MOUNT_OPTIONS -- /dev/sdb6 /var/ltf/tester/scratch_mnt

 generic/077 344s ... [failed, exit status 1] - output mismatch (see /var/lib/xfstests/results//generic/077.out.bad)
     --- tests/generic/077.out   2015-11-23 17:06:27.144983112 +0800
     +++ /var/lib/xfstests/results//generic/077.out.bad  2015-11-23 17:41:25.187062895 +0800
     @@ -1,7 +1,5 @@
      QA output created by 077
      *** create filesystem
     -*** set default ACL
     -*** populate filesystem, pass #1
     -*** populate filesystem, pass #2
     -*** all done
     +mkfs failed
     +(see /var/lib/xfstests/results//generic/077.full for details)
      *** unmount
 Ran: generic/077
 Failures: generic/077
 Failed 1 of 1 tests

Reason:
 btrfs progs v4.3 use non-mixed blockgroup for small volume as default,
 it need at least 100M to build a filesystem.

Fix:
 Force mixed mode for small-size fs, to make mkfs success.

Suggested-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Zhao Lei <zhaolei@cn.fujitsu.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

ext4/001: Select different output based on mount options

Depending on mount options, we handle unwritten extents somewhat
differently. So sometimes we end up zeroing out unwritten extent and
converting it to written one and sometimes we just split it. Choose
expected output based on mount options.

Signed-off-by: Jan Kara <jack@suse.com>
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Dave Chinner <david@fromorbit.com>

common: Improve _link_output_file to be more versatile

Currently _link_output_file() selects output file suffix based on the
current operating system. Make it more versatile by allowing selection
of output file suffix based on any feature string. The idea is that
in config file ($seq.cfg) there are several lines like:

feat1,feat2: suffix

The function is passed a feature string (or uses os_name,MOUNT_OPTIONS
if no argument is passed) and selects output file with a suffix for
which all features are present in the feature string. If there is no
matching line, output with 'default' suffix is selected.

Update all tests using _link_out_file to the new calling convention.

Signed-off-by: Jan Kara <jack@suse.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

common: Factor out function to get OS name

Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Dave Chinner <david@fromorbit.com>

generic/081: Support old version of lvm

generic/081 and 108 fails in RHEL 6.3, like:
 # ./check generic/081
 FSTYP         -- btrfs
 PLATFORM      -- Linux/x86_64 kerneldev 4.2.0-rc5_HEAD_d770e558e21961ad6cfdf0ff7df0eb5d7d4f0754_+
 MKFS_OPTIONS  -- /dev/vdd
 MOUNT_OPTIONS -- /dev/vdd /var/ltf/tester/scratch_mnt

 generic/081
 [failed, exit status 1] - output mismatch (see /var/lib/xfstests/results//generic/081.out.bad)
    --- tests/generic/081.out   2015-07-13 17:07:03.000000000 +0800
    +++ /var/lib/xfstests/results//generic/081.out.bad  2015-10-28 12:20:49.000000000 +0800
    @@ -1,2 +1,3 @@
     QA output created by 081
     Silence is golden
    +ERROR: checking status of /dev/mapper/vg_081-base_081: No such file or directory
 Ran: generic/081
 Failures: generic/081
 Failed 1 of 1 tests

Reason:
 Command of "lvm lvcreate --yes" failed because lvm in RHEL 6.3
 don't support '--yes' option.

Fix:
 Use yes pipe instead '--yes' option for lvm, to make the command
 support both new and old version of lvm.

Suggested-by: Dave Chinner <david@fromorbit.com>
Suggested-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Zhao Lei <zhaolei@cn.fujitsu.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

common: Add more checks to _check_dmesg

Teach _check_dmesg to look for improper RCU usage and circular locking
dependency messages. It's useful to check for these as they might point
to a real problem in the filesystem's implementation (or the current
implementation just confuses the checker, probably worth simplifying
the filesystem's implementation).

Currently the test btrfs/071 for example triggers such warnings:

[  912.924839] ===============================
[  912.925617] [ INFO: suspicious RCU usage. ]
[  912.926394] 4.3.0-rc5-btrfs-next-17+ #1 Not tainted
[  912.927274] -------------------------------
[  912.928364] fs/btrfs/volumes.c:1977 suspicious rcu_dereference_check() usage!
[  912.929626]
[  912.929626] other info that might help us debug this:
[  912.929626]
[  912.931197]
[  912.931197] rcu_scheduler_active = 1, debug_locks = 1
[  912.933822] 4 locks held by btrfs/6400:
[  912.934558]  #0:  (&fs_info->dev_replace.lock_finishing_cancel_unmount){+.+...}, at: [<ffffffffa046a193>] btrfs_dev_replace_finishing+0x3e/0x696 [btrfs]
[  912.948929]  #1:  (uuid_mutex){+.+.+.}, at: [<ffffffffa046a24a>] btrfs_dev_replace_finishing+0xf5/0x696 [btrfs]
[  912.950987]  #2:  (&fs_devs->device_list_mutex){+.+.+.}, at: [<ffffffffa046a263>] btrfs_dev_replace_finishing+0x10e/0x696 [btrfs]
[  912.953265]  #3:  (&fs_info->chunk_mutex){+.+...}, at: [<ffffffffa046a278>] btrfs_dev_replace_finishing+0x123/0x696 [btrfs]
(...)

[  912.987973] ======================================================
[  912.989242] [ INFO: possible circular locking dependency detected ]
[  912.990583] 4.3.0-rc5-btrfs-next-17+ #1 Not tainted
[  912.990801] -------------------------------------------------------
[  912.990801] btrfs/6400 is trying to acquire lock:
[  912.990801]  (&bdev->bd_mutex){+.+.+.}, at: [<ffffffff8119d202>] __blkdev_get+0xa3/0x3d9
[  912.990801]
[  912.990801] but task is already holding lock:
[  912.990801]  (&fs_info->chunk_mutex){+.+.+.}, at: [<ffffffffa046a278>] btrfs_dev_replace_finishing+0x123/0x696 [btrfs]
[  912.990801]
[  912.990801] which lock already depends on the new lock.
[  912.990801]
[  912.990801]
[  912.990801] the existing dependency chain (in reverse order) is:
(...)

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs/091: fix missing common/reflink include

Commit 20d7bfad2d38 ("reflink: add test support routines to a separate
file") moved the function _require_cp_reflink to the new file
common/reflink but forgot to make btrfs/091 source that file, leading
to the following failure:

$ ./check btrfs/091
FSTYP         -- btrfs
PLATFORM      -- Linux/x86_64 debian3 4.3.0-rc5-btrfs-next-17+
MKFS_OPTIONS  -- /dev/sdc
MOUNT_OPTIONS -- /dev/sdc /home/fdmanana/btrfs-tests/scratch_1

btrfs/091 1s ... - output mismatch (see .../results//btrfs/091.out.bad)
    --- tests/btrfs/091.out 2015-05-03 01:19:42.128976975 +0100
    +++ .../results/btrfs/091.out.bad 2015-11-18 15:56:35.332745132 +0000
    @@ -1,4 +1,5 @@
     QA output created by 091
    +./tests/btrfs/091: line 49: _require_cp_reflink: command not found
     wrote 262144/262144 bytes at offset 0
     XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
     65536 65536
    ...
    (Run 'diff -u tests/btrfs/091.out .../results/btrfs/091.out.bad'  \
        to see the entire diff)

So just make btrfs/091 source common/reflink in order to know the
definition of _require_cp_reflink.

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs: test direct IO write against compressed extent

Test that doing a direct IO write against a file range that contains one
prealloc extent and one compressed extent works correctly.

From the linux kernel 4.0 onwards, this either triggered an assertion
failure (leading to a BUG_ON) when CONFIG_BTRFS_ASSERT=y or resulted
in an arithmetic underflow of an inode's space reservation for write
operations.

That issue is fixed by the following linux kernel patch:

  "Btrfs: fix extent accounting for partial direct IO writes"

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

generic/269: Fix failure due to race killing fsstress job

When running the test generic/269 against a fast scratch device (loopback
device backed by a file on tmpfs) I often got a test failure due to the
fact that the fsstress job had already completed before the test attempted
to kill it, producing the following failure output:

generic/269 91s ... - output mismatch (see .../generic/269.out.bad)
    --- tests/generic/269.out 2014-11-17 20:59:50.974203000 +0000
    +++ /home/fdmanana/git/hub/xfstests/results//generic/269.out.bad 2015-11-13 15:41:59.669893035 +0000
    @@ -3,3 +3,4 @@
     Run fsstress

     Run dd writers in parallel
    +./tests/generic/269: line 59: kill: (13417) - No such process
    ...
    (Run 'diff -u tests/generic/269.out .../generic/269.out.bad'  to see the entire diff)

So fix this false failure by redirecting the standard output and error
from the kill into the seq file.

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

generic/285: fix the wrong result printed in full log

This issue was found by testing f2fs module with generic/285, and the
related bug has already been fixed since commit 2e023174a88d ("f2fs:
avoid data offset overflow when lseeking huge file"), but forgot to fix
it in xfstest suit.

The wrong result printed is printed in log below:
10. Test a huge file for offset overflow
10.01 SEEK_HOLE expected 65536 or 0, got 65536.                   succ
10.02 SEEK_HOLE expected 65536 or 0, got 65536.                   succ
10.03 SEEK_DATA expected 0 or 0, got 0.                           succ
10.04 SEEK_DATA expected 1 or 1, got 1.                           succ
10.05 SEEK_HOLE expected 0 or 0, got 0.                           FAIL
10.06 SEEK_DATA expected -65536 or -65536, got -65536.            succ
10.07 SEEK_DATA expected -65535 or -65535, got -65535.            succ
10.08 SEEK_DATA expected -65536 or -65536, got -65536.            FAIL

The result printed in the log shows that when some test cases failed, the
data we expected and got are the same that is not correct obviously.

The reason is that we cast the result from type off_t(64-bit) to type
long(32-bit) when doing huge file offset seeking tests in 32-bit machine.

This patch fixes the wrong printing issue.

Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs: test quota disable during quota rescan

This test case tests if we are able to disable quotas on a filesystem
while a quota rescan is running.  Up to now (4.3) this would result
in a kernel NULL pointer dereference.

Fixed by patch (btrfs: qgroup: fix quota disable during rescan).

Signed-off-by: Justin Maggard <jmaggard@netgear.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs: test unmount during quota rescan

This test case tests if we are able to unmount a filesystem while
a quota rescan is running.  Up to now (4.3) this would result
in a kernel NULL pointer dereference.

Fixed by patch (btrfs: qgroup: exit the rescan worker during umount).

Signed-off-by: Justin Maggard <jmaggard@netgear.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

ext4/001: Update output

After patch "ext4: Fix races of writeback with punch hole and zero
range" we don't flush range that's going to be zeroed out which results
in different final extent layout because some extents will be zeroed-out
instead of being split. Update the output file accordingly.

Signed-off-by: Jan Kara <jack@suse.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

_filter_fiemap: Avoid awk interval regexps

Older versions of awk do not accept interval regexps by default. Avoid
them in _filter_fiemap to keep better compatibility since it's pretty
trivial.

Signed-off-by: Jan Kara <jack@suse.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

common: add helper function _flakey_drop_and_remount

To avoid having many tests repeating the following pattern:

        _load_flakey_table $FLAKEY_DROP_WRITES
        _unmount_flakey

        _load_flakey_table $FLAKEY_ALLOW_WRITES
        _mount_flakey

add the helper function _flakey_drop_and_remount to remove
the existing duplicated code and serve as a shortcut.

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

fstests: generic test for fsync after hole punching

Test that a file fsync works after punching a hole for the same file
range multiple times, and that after log/journal replay the file's
content and layout are correct.

This test is motivated by a bug found in btrfs, which is fixed by
the following linux kernel patch:

  "Btrfs: fix hole punching when using the no-holes feature"

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

reflink: test what happens when we hit resource limits

Add a few horrible opt-in stress tests to see what happens if we try
to reflink the same block billions of times, and what happens if we
run out of space while reflinking a file.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>

reflink: test that CoW writes fail when we're out of space

Ensure that copy-on-writing a reflinked file when there's no free disk
space reflects the desired ENOSPC back to userspace during the write
call.  Tests the buffered IO, direct IO, and mmap write paths.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>

reflink: concurrent operations tests

Make sure that running reflink ops while other IO is ongoing doesn't
break the filesystem.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>

xfs: test xfs-specific reflink pieces

Check that the various XFS tools still work properly on reflinked XFSes.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>

reflink: test error conditions due to bad inputs

Check that we can feed bad inputs to reflink/dedupe and it'll reject
them.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>

reflink: test accuracy of free block counts

Check that the free block counts seem to be handled correctly in
the reflink operation and subsequent attempts to rewrite reflinked
copies.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>

reflink: test the various fallocate modes

Check that the variants of fallocate (allocate, punch, zero range,
collapse range, insert range) do the right thing when they're run
against a range of reflinked blocks.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>

reflink: test CoW behaviors of reflinked files

Ensure that CoW happens correctly with buffered, directio, and mmap writes.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>

reflink: basic tests of the reflink and dedupe ioctls

Test the operation of the btrfs (and now xfs) reflink and dedupe
ioctls at various file offsets and with matching and nonmatching
files.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>

reflink: add test support routines to a separate file

Put all the reflink/dedupe-related test support routines in a separate
file, then modify the existing reflink tests to use them.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs: move btrfs reflink tests to generic

Move the cp --reflink tests from btrfs/ to generic/ since xfs now
supports that ioctl.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <david@fromorbit.com>

generic/312: fix dev_blocks calculation

Current calculation fails when there are many same type of devices
with the testing device, eg sda1 sda2 ... sda10, sda11, sda12 ..
$(grep sda1 /proc/partitions) gets multiple numbers.

$ grep ram1 /proc/partitions
   1        1   10485760 ram1
   1       10   10485760 ram10
   1       11   10485760 ram11
   1       12   10485760 ram12
   1       13   10485760 ram13
   1       14   10485760 ram14
   1       15   10485760 ram15
$ grep -w ram1 /proc/partitions
   1        1   10485760 ram1

Fix this by adding the -w option to grep to match the block device
exactly.

Signed-off-by: Xiong Zhou <xzhou@redhat.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs/012: add a regression test for deleting ext2_saved

Btrfs now has changed to delete subvolume/snapshot asynchronously,
which means that after umount, if we've already deleted 'ext2_saved',
rollback can still be completed, which should not.

So this adds a regression test for this.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

ext4: move 30[1234] from the dangerous to the auto group

The ext4/301, ext4/302, ext4/303, and ext4/304 tests are not crashing
on 3.10.89, 3.14.53, 3.18.21, 4.1.8, and 4.3-rc2.  So promote these
tests from the dangerous to the auto group.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs: cloning of compressed inline extent after truncation

Test that truncating a file that consists of a compressed and inlined extent
to a smaller size and then cloning it into another file is not possible and
does not result in leaking stale data (data past the truncation offset) nor
losing data in the clone operation's destination file.

This btrfs issue is fixed by the linux kernel patch titled:

  "Btrfs: fix truncation of compressed and inlined extents"

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

common: properly quote strings tested with -n

Per the Advanced Bash Scripting Guide:

"The -n test requires that the string be quoted within the test brackets.
Using an unquoted string with ! -z, or even just the unquoted string
alone within test bracket normally works, however, this is an unsafe
practice. Always quote a tested string."

And indeed:

$ unset FOOBAR
$ [ -n $FOOBAR ] || echo nope
$ [ -n "$FOOBAR" ] || echo nope
nope

Ran into this on a box w/o the attr program installed, and passed
_require_attrs.  Quoting the string fixes this; fix it there
and other occurrences in common/* as well.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs/112: add missing _require_xfs_io_command for falloc

I forgot to add the requirement for the xfs_io command "falloc", which
the test makes use of. Also fixed a weird indentation (mix of spaces
and tabs) for one line of a comment.

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs/035: check for data loss

The test currently verifies that cloning one file with an inline extent
with a size of 10 bytes into a file with an inline extent that has a size
of 20 bytes succeeds. But this results in data loss, because the btrfs
clone operation drops the 20 bytes inline extent from the destination
inode and then copies the 10 bytes inline extent from the source file
into the destination file, resulting in data loss of the last 10 bytes
of data that the destination file had.

Fixing btrfs to correctly operate for this case (not resulting in data
loss) is actually a lot of work and brings a lot of complexity, specially
considering that any of the inline extents can be compressed. For the
moment there's a fix to make the clone operation return the errno
EOPNOTSUPP and not touch any of the inodes. This is the same approach
we do for other cases involving operation against inline extents, so
this just adds one more case that should have never been allowed.
Cloning inline extents is a rare operation and pointless, since it
involves copying them and not doing any actual deduplication or saving
space.

The btrfs patch for the linux kernel that prevents this data loss,
and fixes some file corruption cases, is titled:

  "Btrfs: fix file corruption and data loss after cloning inline extents"

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs: test for cloning of inline extents

Test several cases of cloning inline extents that used to lead to file
corruption or data loss.

These file corruption and data loss cases are fixed by the linux kernel
patch titled:

  "Btrfs: fix file corruption and data loss after cloning inline extents"

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

src/runas: Fixes and cleanups

The runas helper runs a command as another user and/or with different group
memberships.  Fix the following problems:

 * Use setgid instead of setegid and setuid instead of seteuid.
   Otherwise, the command will run with the original real UID
   and/or GID; those could be made the effective IDs again.

 * When only a GID is specified, remove all supplementary
   GIDs.  Otherwise, the command would remain in the same
   supplementary groups as runas -- which often is the root
   group.

 * Use execvp instead of execv which searches the PATH when
   necessary.  The runas helper is always called either with a
   '/' in the pathname or as "runas ... `which program`", so
   we obviously want PATH lookup, anyway.

 * There is no advantage in fork'ing and waiting for the child
   over directly exec'ing the command; the test cases already
   have to deal with commands which can be killed by signals.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs: test sending snapshots received from other filesystems

Test that sending a snapshot received from a different filesystem is
possible for both full and incremental send operations.

This used to work until the linux kernel release 4.2, but a commit [1] in
that release introduced a regression which did not allow this anymore.

The regression is fixed by the linux kernel patch titled:

  "btrfs: fix resending received snapshot with parent"

[1] 37b8d27de5d0 ("Btrfs: use received_uuid of parent during send")

Cc: Josef Bacik <jbacik@fb.com>
Cc: Robin Ruede <rruede+git@gmail.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs: send/receive across different filesystems

Test that sending and receiving snapshots across different filesystems
works for full and incremental send operations.

This used to fail before the linux kernel release 4.2. The linux kernel
commit that fixed this issue was the following:

  37b8d27de5d0 ("Btrfs: use received_uuid of parent during send")

Cc: Josef Bacik <jbacik@fb.com>
Cc: Robin Ruede <rruede+git@gmail.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

xfs/167: wait for fsstress to exit

sync doesn't guarantee all fsstress processes died, and sometimes it
ends up running fsck on a mounted fs.

Use wait to wait for fsstress to exit.

Signed-off-by: Eryu Guan <eguan@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

btrfs: send with shared and compressed extents

Test that a send operation works correctly with reflinked files (cloned
extents which multiple files point to) that have compressed extents.

This used to fail on btrfs, resulting in different file digests after
receiving the send stream, and got fixed by the linux kernel patch
titled:

  "Btrfs: send, fix file corruption due to incorrect cloning operations"

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>