actually hook up 198, make it generic

Not entirely sure how I managed this, but 198 was not in groups
so was never run - and wasn't in the makefile, so was never built.
Oops.

Fix that up, make it a generic test, and move it to the
aio-dio-regress subdir.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

don't fail if nslookup is not found

When nslookup is not available, execution would stop. This patch moves the
check for nslookup and the _get_fqdn function into the ./new script (since
this is the only place where it's used) and we don't fail any more but try
to find out our FQDN without nslookup.

Signed-off-by: Christian Kujau <lists@nerdbynature.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>

add missing golden output for test 215

Signed-off-by: Christoph Hellwig <hch@lst.de>

fix trivial typo in test 204 description

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

small common.config cleanup

Don't list various non-mandatory options near the check for the
mandatory ones, and make the test for the mandatory ones a proper if
statements.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

don't print scratch mkfs and mount options without scratch device

Also clean up the way the test configuration is printed.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

xfstests: add tests to check log size scaling

Add two tests that check that log size scaling works correctly for old
and new maximum log sizes.

Signed-off-by: Dave Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

xfstests: test c/mtime updates through mapped writes

Check that we correctly update the timestamps when writing to a file
through an mmap mapping.  Currently fails for XFS due a VFS bug but
succeeds for many other filesystems.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

Fix syntax error in common.rc

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>

fallocate + read/write tests, ext4 regression tests

New test to test basic mixed fallocate + read & write,
includes a couple regression tests for bugs that ext4
hit.  Uses xfs_io to generate fallocate calls, so requires
git xfsprogs and very recent glibc at this point.

Ext4 folks, this is hopefully a reasonable example of
how to add a new test.   :)

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

Move duplicated xfs_io filters to common.filter

Also fix up 132 output, which was misfiltered due
to a bug in the filtering.

Doing this because I need this same filter for the next
added test.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

basic fallocate boundary tests

This is just like test 072, but using fallocate instead
of the xfs ioctl.  Just very basic fallocate tests.

Also adds a "prealloc" group (./check -g prealloc)
and a _require_xfs_io_falloc check.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

fix common.config to allow SCRATCH_DEV and SCRATCH_MNT to be optional

Commit 3ae9f2f8 purports to allow SCRATCH_DEV and SCRATCH_MNT to be
optional, but tests in common.config will cause check to exit with an
error if these environment variables are not defined.  Fix this problem.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Christoph Hellwig <hch@lst.de>

fix _require_scratch test for extN, resierfs, gfs2, and btrfs

The extN, reiserfs, gfs2, and btrfs filesysytem types should use the
same check for a block device as XFS and UDF, and not the test for
NFS, which was checking for host:/foo/bar/baz when checking for a
scratch device.

Also, the NFS logic was also incorrect, in that it would allow a
zero-length SCRATCH_DEV to continue.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Felix Blyakher <felixb@sgi.com>

skip _check_test_fs unless a test is actually run

Running _check_test_fs can take a non-trivial amount of time, and if a
test has been skipped because it doesn't work on Linux, or it doesn't
work of for ext4, it's pointless to re-run _check_test_fs.  So move
the call to _check_test_fs so it is only run if a test is actually run.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Felix Blyakher <felixb@sgi.com>

Skip dump _cleanup if filesystem is not xfs

Many/most dump tests include common.dump before checking
supported fs, and this means that even if the dump tests
are skipped, the sleep & filesystem check in _cleanup
gets run, which makes it take rather a long time to
skip these tests for non-xfs filesystems.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>

Gracefully fail aio tests if they weren't built

configure & make succeeds w/o aio headers or libs,
so tests should handle that.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

Enable test 080 for Linux

Seems to work here... Also disable IRIX as supported,
since the comments & code say it doesn't work and
it short-circuits anyway.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

Enable still more tests for generic filesystems

Make the following tests _supported_fs generic:

088 - test out CAP_DAC_OVERRIDE and CAP_DAC_SEARCH code
089 - Emulate the way Linux mount manipulates /etc/mtab
113 - aio-stress (explicitly mark as generic)
126 - tests various file permission options
129 - looptests

These all pass on ext3, ext4, btrfs, and gfs2 as well
as xfs.

Also remove "generic" group from "groups," which was
accidentally added.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

allow more xfs_io tests to be generic

These tests just do generic reads & writes with xfs_io;
if we add a "-F" they will run on other filesystems.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>

Allow fsx tests to run on generic filesystems

There is already logic in 075 to gracefully skip nfs for fsx
invocations with -x (xfs-specific preallocation) - just extend
this to any non-xfs filesystem, and add to test 112 as well.

Later we can change this behavior to use fallocate and include
more filesystems but this gets some fsx coverage for now.

Test 127 doesn't seem to have anything xfs-specific, so mark
that as generic too.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>

Make "new" script insert appropriate license on tests

Don't default to SGI copyright, someone else may be
making this test...

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>

Add GPL license plate to SGI's files.

Signed-off-by: Felix Blyakher <felixb@sgi.com>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Felix Blyakher <felixb@sgi.com>

update aio-stress.c, with copyright information intact

Pulled from:
http://ltp.cvs.sourceforge.net/viewvc/ltp/ltp/testcases/kernel/io/ltp-aiodio/aio-stress.c
revision 1.8

There are a couple new options, but the main motivator is to
get the license text intact again.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>

add license boilerplate to 205

Signed-off-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

aio-dio-regress: fix aio-dio-extend-stat on s390

This patch:

o Changes the BUFSIZE to 4096 so that we can successfully perform direct
  I/O on devices that have a sector size of 4k, such as the virtual disks
  found on the s390 architecture.
o Removes an unused variable.
o Checks the proper field in the ioevent to determine if there was an
  error in the I/O submission.

Signed-off-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Zach Brown <zach.brown@oracle.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>

Revert "aio-dio-regress: fix aio-dio-extend-stat on s390"

This reverts commit 6b26794772a4acbe8df1cc48d7a169011268d97e.

A later version of this patch was submitted, and I committed
the wrong one.  -Eric

xfstests: fix syntax error in check

Bash 3.2 wants a space before the closing "]" in a test.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Felix Blyakher <felixb@sgi.com>

Check for xlog_assign_lsn in xfs headers

Checking for libxfs.h isn't enough; some debian installs
have libxfs.h but no xlog_assign_lsn, and the loggen build still
fails.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Tested-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>

Hi,

So, the xfs test suite does a mount, followed by running the test, then
an unmount after the test exits.  aio-dio-invalidate-failure spawns two
children, and will kill them off before it exits.  The problem is that
it doesn't wait for them to exit before returning, so the xfs test
harness ends up failing the umount as the mount point is still busy.

The fix is to simply wait for each of the children exits before
returning from the parent.

(Eric Sandeen: add one more waitpid to error case)

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

aio-dio-regress: fix aio-dio-extend-stat on s390

Hi,

The following patch queries the sector size of the underlying device so
that we can size the write buffer appropriately.  Without this patch, we
try to do an O_DIRECT write of 1KB to a device that has a sector size of
4KB.  This returns EINVAL, of course.  I also noticed that the test was
not checking the right fields for the return code.  It was checking the
original iocb's nbytes field instead of event->res.  So, I fixed that.

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

aio-dio-regress: align I/O buffers to 4k for 4k sector devices

(and virtual devices as found on s390)

Without this patch, these tests can randomly fail on s390 systems which
use a virtual block device with sector size of 4k.  Testing confirms
that this patch resolves the issue.

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

Add aio-dio-regress tests

Pull in the aio-dio-regress test suite from
http://git.kernel.org/?p=linux/kernel/git/zab/aio-dio-regress.git

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

don't run 122 if required headers aren't found

Now that libxfs etc is optional, 122 (which compiles on the fly)
is failing if those extra headers are missing.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>

Make the SCRATCH_DEV optional

Tests will fail gracefully if there is no SCRATCH_DEV defined, so change

the check script to allow it, and update README to reflect this.

This just makes it simpler to set up the tests, at the expense of
coverage.

Long term it might be nice to allow either SCRATCH_DEV xor TEST_DEV
to be missing, but for now this is simple.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>

enable generic filesystems to be checked

This includes a fair bit of rearranging to avoid code duplication,
but the goal is to allow 'fsck -n -t $FSTYP $device' to be run on
any generic filesystem.

Any FS for which this doesn't work will need it's own fsck routine.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: Christoph Hellwig <hch@lst.de>

libxfs.h is really only needed for the loggen test;
make it config-time optional.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@infradead.org>

xfstests: initial reiserfs, gfs2, btrfs support

This patch gets reiserfs, gfs2, and btrfs off the ground.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

Fix 206 expected output

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

Test trim of last small AG for large filesystem resizes

As reported at
http://article.gmane.org/gmane.comp.file-systems.xfs.general/29187
this trimming may cause an overflow in the new size calculation.

Patch to fix it, and testcase at
http://article.gmane.org/gmane.comp.file-systems.xfs.general/29193

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

xfstests: add GPL boilerplates to my testcases

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>

Report which tests did get run

fs type mismatches are now likely enough that it may be
nice to report which tests -were- run at the end...

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

Set default extN mount options

ext2/3/4 don't support acls & attrs w/o specific mount
options, so make those the default whenever we mount
these filesystems.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

Detect FS type to test based on TEST_DEV

This helps support generic / non-xfs fileystems.

Rather than defaulting to xfs or expecting a fs
type passed in, just look at what is on TEST_DEV
and use that for FSTYP by default.

We may wish to add a generic way to override this later.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>

Set up testdir for generic filesystems

Make the default filesystem testdir setup the same as
what xfs did; simply set up the testdir variable.

Update the comment for UDF & NFS, as to why they are
currently doing something different.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

069: make scratch mkfs quiet

The _scratch_mkfs call in test 069 was not redirecting
stderr to /dev/null; other mkfs's may be more noisy
on stderr.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

Support "generic" filesystem type

Create a new "generic" _supported_fs type for tests
which are not really filesystem-specific.  "generic"
tests do expect that acl & attr are supported though.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

Use xfs.h rather than libxfs.h

Use xfs/xfs.h when possible rather than xfs/libxfs.h,
now that libxfs.h isn't part of a normal xfsprogs-devel
install.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

Drop the xfs-specific MAXNAMELEN in favor of the Posix NAME_MAX

This makes it easier to lose libxfs.h later.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

xfstests: move GPL boilerplates to fix lsqa.pl

It turns out lsqa.pl nees the test number and description first in the
file, so move the GPL boilerplates below it.

Also remove acouple of cases where we have one full copyright line + gpl
boilerplate before the description and another copyright line after
the description.

Signed-off-by: Christoph Hellwig <hch@lst.de>

xfstests: only run 164 for xfs

Test 164 uses the xfs_io bmap subcommand which only works for XFS,
so don't offer it for nfs and udf.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

xfstests: add GPL boilerplates to my testcases

Signed-off-by: Christoph Hellwig <hch@lst.de>

Merge branch 'master' of git://oss.sgi.com/xfs/cmds/xfstests

Add GPL license plate to SGI's test files.

Signed-off-by: Felix Blyakher <felixb@sgi.com>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Felix Blyakher <felixb@sgi.com>

fsx: Add check for mmap last-page zero fill
Check to make sure the VM system provided 0's beyond the true end of
the file mapping (as required by mmap def in 1996 posix 1003.1)

This check was taken from the FreeBSD version of fsx.c.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Eric Sandeen <sandeen@sandeen.net>

fsx: Replace use of bcmp() with memcmp()
The bcmp() function comes from BSD 4.3, and was deprecated in
POSIX.1-2001.  It was removed entirely in POSIX.1-2008.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Eric Sandeen <sandeen@sandeen.net>

fsx: Replace use of bzero() with memset()
The bzero() function comes from BSD 4.3, and was deprecated in
POSIX.1-2001.  It was removed entirely in POSIX.1-2008.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Eric Sandeen <sandeen@sandeen.net>

fsx: Use %ll instead of %q in printf format statements
The ISO C approved way of printing long long variables is %ll.  The %q
length modifier was used in BSD 4.4 and Linux libc5, and its use is
discouraged as non-standards compliant.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Eric Sandeen <sandeen@sandeen.net>

fsx: Use SEEK_END instead of the BSD'ish L_XTND
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Eric Sandeen <sandeen@sandeen.net>

xfstests: tidy up 205

Test 205 wasn't marked executable, and had the wrong test
number in the comments.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Felix Blyakher <felixb@sgi.com>

xfstests 042: point xfs_fsr right at the fragmented file

Somehow 042 as it stands isn't actually defragging the fragmented
file we created (see 042.full post-run)

Probably need to look into that as well, but for now this should
make the call to fsr actually _do_ something.

This has let 2 fsr bugs slip through qa in the past few months.   :(

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@infradead.org>

xfstests: fix 130, 132 for inifinitely-fast disks

Due to the new quantum/holographic storage I'm testing, my
disks go infinitely fast, thereby breaking some filters:

-XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+3 KiB, 1 ops; 0.0000 sec (inf EiB/sec and inf ops/sec)

I'm no regexp expert but I think the below change will fix it.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Josef 'Jeff' Sipek <jeffpc@josefsipek.net>
Reviewed-by: Christoph Hellwig <hch@infradead.org>

Revert "patch more_set_prog_path"

This reverts commit b2fcff900de26dd0b742d2e77fbfc6b8dc13c83e.

This snuck in unintentionally due to my weak git-fu,
and was breaking things.

xfstests: filter out imaxpct value in test 096

On large filesystems, imaxpc drops from 25 to 5,
so filter out that difference.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>

xfstests: skip fsr tests if xfs_fsr utils not found

Similar to the xfsdump-related patch; gracefully fail
if no xfs_fsr is found.

I'll probably send another for the acl & attr userspace,
it's actually not always installed on minimal boxes.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Felix Blyakher <felixb@sgi.com>

xfstests: skip dump tests if dump utils not found

I was running xfstests on a NAS box which had no xfsdump,
and any tests which required xfsdump were failing in not-so-nice
ways:

+ xfsdump not found
038      [failed, exit status 1] - output mismatch (see 038.out.bad)

I think using _notrun is a lot better:

038      [not run] xfsdump not found

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@infradead.org>
Reviewed-by: Josef 'Jeff' Sipek" <jeffpc@josefsipek.net>
Reviewed-by: Felix Blyakher <felixb@sgi.com>

Merge branch 'master' of git://git.kernel.org/pub/scm/fs/xfs/xfstests-dev

patch more_set_prog_path

Merge branch 'master' of git://git.kernel.org/pub/scm/fs/xfs/xfstests-dev

xfstests: fix async io error handling in fsx

The result of async io returned in the event.res in addition
to the number of bytes read/written provides negated error
number. The broken libaio defines event.res as unsigned
while the same structure in the kernel defines it as signed.
The kernel indeed treats it as signed, and returns the
negated error number. Till libaio is fixed we provide
the signed long temp var.
Also set errno for each error condition in aio_rw, as the
caller is not aio aware and expects ret(-1)+errno by the
traditional libc convention.

Signed-off-by: Felix Blyakher <felixb@sgi.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>

xfstests: test 205, test writing to ENOSPC

Use larger files and different writing styles to fill a 100MB filesystem
to being full. In each case we should get very close to the filesystem
being full before getting ENOSPC. This tests different types of ENOSPC
failures to test 203 and requires more changes to pass.

Signed-off-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

xfstests: add test 204, a simple delayed allocation ENOSPC test

Using a small (100MB) filesystem and writing lots of single block files
can result in spurious ENOSPCs being reported. Reproduce this test case
so we can confirm that it gets fixed.

Signed-off-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

xfstests: Add a .gitignore file

Ignore generated files.

Signed-off-by: Josef 'Jeff' Sipek <jeffpc@josefsipek.net>
Signed-off-by: Josef 'Jeff' Sipek <jsipek@eecs.umich.edu>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Felix Blyakher <felixb@sgi.com>

xfstests: make the mode consistent for all the test scripts

Signed-off-by: Josef 'Jeff' Sipek <jeffpc@josefsipek.net>
Signed-off-by: Josef 'Jeff' Sipek <jsipek@eecs.umich.edu>
Signed-off-by: Christoph Hellwig <hch@lst.de>

reduce the number of processes forked

One of the big cpu time consumers when running xfsqa on UML
is forking of new processes. when looping lots of times,
using 'expr' to calculate the loop counter increment means
we fork at least once every loop. using shell builtins means
that we don't fork and many tests run substantially faster.

Some tests are even runnable with this modification. e.g. 110
went from taking 4500s to run down to 9s with the loop iterators
changed to avoid forking.

Signed-off-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

add test 203, xfs_io bmap reallocation

Test that we can get all extent/hole information for files with more
than 16 extents and 15 holes which require a reallocation based on
XFS_IOC_FSGETXATTR.

Based on a testcase from Tomasz Majkowski <moosh009@gmail.com>.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

xfstests: a couple of fixes for external logs

Fix a couple of issues when running xfsqa with external logs:

 - update the 096 golden output for the external log case
 - add a new _scratch_xfs_check similar to _scratch_xfs_logprint and
   _scratch_xfs_repair that take the log device into account and use it
   in test 134
 - use _scratch_xfs_repair in test 202 to fix it for external log
   devices

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

add "quick" group

Add a new group for test to quickly verify WIP patches.  I've started
it by only taking tests fro mthe auto group that take 20 seconds or less
to complete on the kvm virtual machine on my laptop.  The total run
of the quick group takes about 8 minutes for me.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

test 134 requires quotas

Add a  _require_quota call to make sure it shows up as not run and not
as failed if the kernel doesn't support quotas.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

remove tests 171, 172 and 173 from auto group

I don't have a single machine where these succeed, and due to the wait
the filestreams allocator works these tests might depend on the phase
on the moon to get their expected output.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

test 199: test features2 fixup for the regular case, too

Eric Sandeen noted that we should not only test the mount ro first, then
remount rw case in 199 but also the regular mount rw case.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Felix Blyakher <felixb@sgi.com>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

test 201: fix comment typos

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

xfstests: Fix *FLAGS passing and dependencies.

Pass *FLAGS in some targets.

Drop LIBHANDLE, LIBATTR and LIBACL from deps since there are in form
"-llibrary".

Signed-off-by: Arkadiusz Miśkiewicz <arekm@maven.pl>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

Use explict libtool CC tag

(sometimes libtool can't decide what tag is correct one
if omited).

Preserve CFLAGS/CPPFLAGS to allow:
CPPFLAGS="-I$HOME/here-is-xfsprogs-installed/include" \
LDFLAGS="-L$HOME/here-is-xfsprogs-installed/lib" \
./configure  ...

Signed-off-by: Arkadiusz Miśkiewicz <arekm@maven.pl>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

add test 202, xfs_repair handling of single-AG filesystems

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

fix 122 for removed btree types

With the the btree refactoring various btree trypes have been removed,
so remove them from the golden output for 122.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

add test 201: btree corruption without CONFIG_LBD

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <david@fromorbit.com>

run 092 unconditionally

Now that we support inode64 filesystem on 32-bit plaforms tests 092
should be run unconditionally.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

Don't run test 197 on 64-bit machines

If it always passes on 64 bits, this gives a false
sense of security.  Make it obvious.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

xfstests: add test 200, exercise read-only block devices and mounts

Make sure we do the right thing with blockdevices with a hard read-only
flag.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

xfstests: add test 199, bad_features2 correction

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

Add Makepkgs support to xfstests (tar & rpm only so far)

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

Create xfstests install target.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

test 198, testcase for aio+dio writes into sparse files

Older kernels were not doing completion properly, and converting
unwritten extents at the wrong offsets; this exposed stale data.

Xen w/ the blktap driver was hitting this.

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>

test 197, add a testcase for d_off truncation

Make sure our directory offsets fit into a 32 bit value.
Based on a report by John Stanley.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

test 196, add a testcase for renames across project boundaries

Make sure rename across project boundaries is rejected and doesn't
cause hangs.  Based on a report and testcase from Arkadiusz Miskiewicz.

Signed-off-by: Christoph Hellwig <hch@lst.de>

test 195, add a testcase for the nodump flag

Based on http://oss.sgi.com/bugzilla/show_bug.cgi?id=340.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

update for xfs_dinode_core_t removal

Strip away the core. prefixes in 186 to make sure the golden output
works for both old and new xfsprogs.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

fix compiler warnings

Include all nessecary headers, use the correct format strings and don't
redefine _GNU_SOURCE.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>

test 194, test tricky mapping/conversion around holes

Related to
http://oss.sgi.com/bugzilla/show_bug.cgi?id=801

Latest patch posted on that bug makes this testcase pass...

first 2 tests are simple buffred writ tests making sure stale
data isn't exposed, and hole-blocks aren't mapped.

2nd 2 tests are more related to the above bug, tricky testcase
uncovered by fsx on ppc64 which actually re-maps a block
which should be a hole, bringing stale data back into existence.

V2, don't use non-posix awk extensions

Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Christoph Hellwig <hch@lst.de>