Re: [PATCH v3] fstests/xfs: xfs_repair should junk empty attribute leaf blocks

[Eryu Guan <eguan@redhat.com>]

On Thu, Apr 13, 2017 at 02:23:21PM +0800, Zorro Lang wrote:
> There was a bug during log replay, the attr/attr3 leaf verifier
> reported corruption when encountering a leaf attribute with a
> count of 0 in the header, as below:
> 
> Metadata corruption detected at xfs_attr3_leaf block 0x480988/0x1000
> 
> commit f714016 from xfsprogs has fixed this bug. This test case
> will emulate this corruption by xfs_db and use xfs_repair to fix
> it.
> 
> Signed-off-by: Zorro Lang <zlang@redhat.com>
> ---
> 
> Hi,
> 
> V3 did below changes:
> 1) Only test on V4 XFS.
> 2) Use 'ablock 0' to instead of "a.bmx[0].startblock"
> 3) Use "dd ... | attr -s ..." to instead of that python command in
>    "setfattr ..."
> 4) Do "xfs_repair -n" to check if xfs_repair can find the corruption
>    we made.
> 5) Use xfs_db to check if xfs_repair has fixed the corruption properly.
>    I check it by trying to locate 'ablock 0' again. Should I turn to
>    check if core.naextents == 0?
> 
> Thanks,
> Zorro
> 
>  tests/xfs/288     | 109 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
>  tests/xfs/288.out |   2 +
>  tests/xfs/group   |   1 +
>  3 files changed, 112 insertions(+)
>  create mode 100755 tests/xfs/288
>  create mode 100644 tests/xfs/288.out
> 
> diff --git a/tests/xfs/288 b/tests/xfs/288
> new file mode 100755
> index 0000000..5e6167f
> --- /dev/null
> +++ b/tests/xfs/288
> @@ -0,0 +1,109 @@
> +#! /bin/bash
> +# FS QA Test 288
> +#
> +# When an attribute leaf block count is 0, xfs_repair should junk
> +# that leaf directly (as xfsprogs commit f714016).
> +#
> +#-----------------------------------------------------------------------
> +# Copyright (c) 2017 Red Hat Inc.  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
> +. ./common/attr
> +
> +# remove previous $seqres.full before test
> +rm -f $seqres.full
> +
> +# V4 XFS attr type is 'attr'
> +write_cmd="write"

This variable is not used in this test.

> +
> +# Modify as appropriate.
> +_supported_fs xfs
> +_supported_os Linux
> +_require_scratch
> +_require_attrs
> +
> +# Due to xfs_db's write -d option is still not stable, and there's no
> +# plan to support attr3. So only run this case on V4 XFS.
> +# Please update this if xfs_db get enough improvement in one day.
> +if [ -z "$XFS_MKFS_HAS_NO_META_SUPPORT" ]; then
> +	mkfs_opts="-m crc=0"
> +fi
> +$MKFS_XFS_PROG -f $mkfs_opts $SCRATCH_DEV | _filter_mkfs 2>$tmp.mkfs >/dev/null
> +. $tmp.mkfs

_scratch_mkfs "$mkfs_opts" | _filter_mkfs ...

If there's mkfs option conflicts between $MKFS_OPTIONS and local
$mkfs_opts, _scratch_mkfs will try mkfs again with user specified
$mkfs_opts only.

> +
> +_scratch_mount
> +
> +touch $SCRATCH_MNT/$seq.attrfile
> +inum=$(stat -c '%i' $SCRATCH_MNT/$seq.attrfile)
> +
> +# To get an attr block leaf, we need to extend attr format to extent
> +# or btree format at least, and the max inode size is half of filesystem
> +# block size, so write half of block size attr to make sure attr
> +# out of local format.
> +dd if=/dev/zero bs=$((dbsize/2)) count=1 2>/dev/null | \
> +	$ATTR_PROG -s testattr $SCRATCH_MNT/$seq.attrfile >/dev/null

I think we can use perl here, as perl is widely used in fstests and no
new dependency needed.

Thanks,
Eryu

> +
> +_scratch_unmount
> +# manually corrupt the XFS, by set the header count of attr to 0
> +_scratch_xfs_db -x -c "inode $inum" \
> +		   -c "ablock 0" \
> +		   -c "write hdr.count 0" >> $seqres.full
> +
> +# make sure xfs_repair can find above corruption. If it can't, that
> +# means we need to fix this bug on current xfs_repair
> +_scratch_xfs_repair -n >> $seqres.full 2>&1
> +if [ $? -eq 0 ];then
> +	_fail "xfs_repair can't find the corruption"
> +else
> +	# If xfs_repair can find this corruption, then this repair
> +	# should junk above leaf attribute and fix this XFS.
> +	_scratch_xfs_repair >> $seqres.full 2>&1
> +
> +	# Old xfs_repair maybe find and fix this corruption by
> +	# reset the first used heap value and the usedbytes cnt
> +	# in ablock 0. That's not what we want. So check if
> +	# xfs_repair has junked the whole ablock 0 by xfs_db.
> +	_scratch_xfs_db -x -c "inode $inum" -c "ablock 0" | \
> +		grep -q "no attribute data"
> +	if [ $? -ne 0 ]; then
> +		_fail "xfs_repair didn't junk the empty attr leaf"
> +	fi
> +fi
> +
> +echo "Silence is golden"
> +
> +# success, all done
> +status=0
> +exit
> diff --git a/tests/xfs/288.out b/tests/xfs/288.out
> new file mode 100644
> index 0000000..2958a5c
> --- /dev/null
> +++ b/tests/xfs/288.out
> @@ -0,0 +1,2 @@
> +QA output created by 288
> +Silence is golden
> diff --git a/tests/xfs/group b/tests/xfs/group
> index 75769f9..f27a7b6 100644
> --- a/tests/xfs/group
> +++ b/tests/xfs/group
> @@ -285,6 +285,7 @@
>  285 dangerous_fuzzers dangerous_scrub
>  286 dangerous_fuzzers dangerous_scrub dangerous_online_repair
>  287 auto dump quota quick
> +288 auto quick repair fuzzers
>  290 auto rw prealloc quick ioctl zero
>  291 auto repair
>  292 auto mkfs quick
> -- 
> 2.7.4
> 
> --
> To unsubscribe from this list: send the line "unsubscribe fstests" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html