Re: [PATCH 10/8] xfs: check that fs freeze minimizes required recovery

[Eryu Guan <eguan@redhat.com>]

On Wed, Jan 03, 2018 at 11:26:26AM -0800, Darrick J. Wong wrote:
> From: Darrick J. Wong <darrick.wong@oracle.com>
> 
> Make sure that a fs freeze operation cleans up as much of the filesystem
> so as to minimize the recovery required in a crash/remount scenario.  In
> particular we want to check that we don't leave CoW preallocations
> sitting around in the refcountbt, though this test looks for anything
> out of the ordinary on the frozen fs.
> 
> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
> ---
>  tests/xfs/903     |  107 +++++++++++++++++++++++++++++++++++++++++++++++++++++
>  tests/xfs/903.out |   10 +++++
>  tests/xfs/group   |    1 
>  3 files changed, 118 insertions(+)
>  create mode 100755 tests/xfs/903
>  create mode 100644 tests/xfs/903.out
> 
> diff --git a/tests/xfs/903 b/tests/xfs/903
> new file mode 100755
> index 0000000..1686356
> --- /dev/null
> +++ b/tests/xfs/903
> @@ -0,0 +1,107 @@
> +#! /bin/bash
> +# FS QA Test No. 903
> +#
> +# Test that frozen filesystems are relatively clean and not full of errors.
> +# Prior to freezing a filesystem, we want to minimize the amount of recovery
> +# that will have to happen if the system goes down while the fs is frozen.
> +# Therefore, start up fsstress and cycle through a few freeze/thaw cycles
> +# to ensure that nothing blows up when we try to do this.
> +#
> +# Unfortunately the log will probably still be dirty, so we can't do much
> +# about enforcing a clean repair -n run.
> +#
> +#-----------------------------------------------------------------------
> +# Copyright (c) 2000-2002 Silicon Graphics, Inc.  All Rights Reserved.
> +# Copyright (c) 2018 Oracle.  All Rights Reserved.
> +#
> +# This program is free software; you can redistribute it and/or
> +# modify it under the terms of the GNU General Public License as
> +# published by the Free Software Foundation.
> +#
> +# This program is distributed in the hope that it would be useful,
> +# but WITHOUT ANY WARRANTY; without even the implied warranty of
> +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> +# GNU General Public License for more details.
> +#
> +# You should have received a copy of the GNU General Public License
> +# along with this program; if not, write the Free Software Foundation,
> +# Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
> +#
> +#-----------------------------------------------------------------------
> +#
> +
> +seq=`basename $0`
> +seqres=$RESULT_DIR/$seq
> +echo "QA output created by $seq"
> +
> +here=`pwd`
> +tmp=/tmp/$$
> +status=1
> +trap "_cleanup; rm -f $tmp.*; exit \$status" 0 1 2 3 15
> +
> +_cleanup()
> +{
> +	# Make sure we thaw the fs before we unmount or else we remove the
> +	# mount without actually deactivating the filesystem(!)
> +	$XFS_IO_PROG -x -c "thaw" $SCRATCH_MNT 2> /dev/null
> +	echo "*** unmount"
> +	_scratch_unmount 2>/dev/null
> +}
> +
> +# get standard environment, filters and checks
> +. ./common/rc
> +. ./common/filter
> +
> +# real QA test starts here
> +_supported_fs xfs
> +_supported_os Linux
> +
> +_require_scratch
> +
> +# xfs_db will OOM kill the machine if you don't have huge amounts of RAM, so
> +# don't run this on large filesystems.
> +_require_no_large_scratch_dev

Looks like this is copied from some other test, but seems
_check_xfs_filesystem already skips _xfs_check if $LARGE_SCRATCH_DEV is
'yes', so we don't need this _require rule now.

> +
> +echo "*** init FS"
> +
> +rm -f $seqres.full
> +_scratch_unmount >/dev/null 2>&1

_require_scratch umounts it for you :)

> +echo "*** MKFS ***" >>$seqres.full
> +echo "" >>$seqres.full
> +_scratch_mkfs_xfs >>$seqres.full 2>&1 || _fail "mkfs failed"
> +_scratch_mount >>$seqres.full 2>&1 || _fail "mount failed"
> +
> +echo "*** test"
> +
> +for l in 0 1 2 3 4
> +do
> +	echo "    *** test $l"
> +	FSSTRESS_ARGS=`_scale_fsstress_args -d $SCRATCH_MNT -n 1000 $FSSTRESS_AVOID`
> +	$FSSTRESS_PROG  $FSSTRESS_ARGS >>$seqres.full
> +
> +	$XFS_IO_PROG -x -c 'freeze' $SCRATCH_MNT
> +
> +	# Log will probably be dirty after the freeze, record state
> +	echo "" >>$seqres.full
> +	echo "*** xfs_logprint ***" >>$seqres.full
> +	echo "" >>$seqres.full
> +	log=clean
> +	_scratch_xfs_logprint -tb 2>&1 | tee -a $seqres.full \
> +		| head | grep -q "<CLEAN>" || log=dirty
> +
> +	# Fail if repair complains and the log is clean
> +	echo "" >>$seqres.full
> +	echo "*** XFS_REPAIR -n ***" >>$seqres.full
> +	echo "" >>$seqres.full
> +	_scratch_xfs_repair -f -n >> $seqres.full 2>&1
> +
> +	if [ $? -ne 0 ] && [ "$log" = "clean" ]; then
> +		_fail "xfs_repair failed"
> +	fi

Hmm, I enlarged the loop count to 100 and didn't see a single CLEAN log,
I suspect this test is unlikely to fail..

Thanks,
Eryu

> +
> +	$XFS_IO_PROG -x -c 'thaw' $SCRATCH_MNT
> +done
> +
> +echo "*** done"
> +status=0
> +exit 0
> diff --git a/tests/xfs/903.out b/tests/xfs/903.out
> new file mode 100644
> index 0000000..378f0cb
> --- /dev/null
> +++ b/tests/xfs/903.out
> @@ -0,0 +1,10 @@
> +QA output created by 903
> +*** init FS
> +*** test
> +    *** test 0
> +    *** test 1
> +    *** test 2
> +    *** test 3
> +    *** test 4
> +*** done
> +*** unmount
> diff --git a/tests/xfs/group b/tests/xfs/group
> index e1b1582..23c26c2 100644
> --- a/tests/xfs/group
> +++ b/tests/xfs/group
> @@ -435,3 +435,4 @@
>  435 auto quick clone
>  436 auto quick clone fsr
>  708 auto quick other
> +903 mount auto quick stress