Re: [PATCH v2] xfstests: add regression test for running btrfs balance and defrag concurrently

[Dave Chinner <david@fromorbit.com>]

On Wed, Oct 30, 2013 at 10:12:52PM +0800, Liu Bo wrote:
> The test aims to trigger snapshot-aware defrag path in write endio by
> running balance, which is not expected and leads to a crash.
> 
> Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
> ---
> v2: remove useless words as required.
> 
>  tests/btrfs/021     |   90 +++++++++++++++++++++++++++++++++++++++++++++++++++
>  tests/btrfs/021.out |    2 +
>  tests/btrfs/group   |    1 +
>  3 files changed, 93 insertions(+), 0 deletions(-)
>  create mode 100755 tests/btrfs/021
>  create mode 100644 tests/btrfs/021.out
> 
> diff --git a/tests/btrfs/021 b/tests/btrfs/021
> new file mode 100755
> index 0000000..23eff47
> --- /dev/null
> +++ b/tests/btrfs/021
> @@ -0,0 +1,90 @@
> +#! /bin/bash
> +# FS QA Test No. 021
> +#
> +# A regression test of running btrfs balance and defrag concurrently.
> +#
> +# The test aims to trigger snapshot-aware defrag path in endio by
> +# running balance, which is not expected and leads to a crash.
> +#
> +#-----------------------------------------------------------------------
> +# Copyright (c) 2013 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	# failure is the default!
> +trap "_cleanup; exit \$status" 0 1 2 3 15
> +
> +_cleanup()
> +{
> +    cd /
> +    rm -f $tmp.*
> +}
> +
> +# get standard environment, filters and checks
> +. ./common/rc
> +. ./common/filter
> +
> +# real QA test starts here
> +run_test()
> +{
> +	$BTRFS_UTIL_PROG balance start $SCRATCH_MNT >> $seqres.full &
> +
> +	sleep 0.5
> +
> +	find $SCRATCH_MNT -type f -print0 | xargs -0 $BTRFS_UTIL_PROG filesystem defrag -f

Lines longer than 80 columns.

> +
> +	sync
> +	wait
> +}
> +
> +# Modify as appropriate.

Comment not needed.

> +_supported_fs btrfs
> +_supported_os Linux
> +_require_scratch
> +
> +rm -f $seqres.full
> +
> +_scratch_mkfs > /dev/null 2>&1
> +_scratch_mount
> +
> +
> +for ((j=0; j<100; j++))
> +do
> +	touch $SCRATCH_MNT/padding-$j
> +done
> +
> +for ((j=0; j<50; j++))
> +do
> +	for i in `seq 20 -1 1`

Please us a single syntax for for loops.

for j in `seq 0 50`; do
	for i in `seq 20 -1 1`; do
...

> +	do
> +		dd if=/dev/zero of=$SCRATCH_MNT/foo-$j bs=4k count=1 seek=$i oflag=direct conv=notrunc 2>/dev/null;
> +	done

Using xfs_io is preferable to dd.

		$XFS_IO_PROG -f -d -c "pwrite $(($i * 4096)) 4096" \
			$SCRATCH_MNT/foo-$j | _filter_xfs_io
> +done
> +
> +sync

Why do direct IO if you then run sync? If you are trying to fragment
the file, then please add a comment that you are doing a backwards
write to fragment it and how it interacts with the "touch loop"
above it. That way, if in future backwards writes are fixed to no
longer fragment files, we know that this test needs a new method of
fragmenting files...

> +# success, all done
> +echo "Silence is golden"

And with the use of the _filter_xfs_io, you get detection of write
failures due to xfs_io failures because the golden image match will
fail....

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com