Re: [PATCH 5/5] generic: add write vs fcollapse test

[Brian Foster <bfoster@redhat.com>]

On Wed, Sep 17, 2014 at 11:41:53AM +1000, Dave Chinner wrote:
> This test exposed a problem with XFS where it failed to write back a
> partial page correctly during a fcollapse operation. This left a
> stray dirty buffer on the page, and hence invalidation of the page
> then failed of the fcollapse returned an EBUSY error.
> 
> Make this a generic test so that we can ensure that all filesystems
> handle the case correctly. Test case originally worked out and
> written by Brian Foster.
> 
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
>  tests/generic/031     | 73 +++++++++++++++++++++++++++++++++++++++++++++++++++
>  tests/generic/031.out | 19 ++++++++++++++
>  tests/generic/group   |  1 +
>  3 files changed, 93 insertions(+)
>  create mode 100644 tests/generic/031
>  create mode 100644 tests/generic/031.out
> 
> diff --git a/tests/generic/031 b/tests/generic/031
> new file mode 100644
> index 0000000..7d615c0
> --- /dev/null
> +++ b/tests/generic/031
> @@ -0,0 +1,73 @@
> +#! /bin/bash
> +# FS QA Test No. generic/031
> +#
> +# Test non-aligned writes against fcollapse to ensure that partial pages are
> +# correctly written and aren't left behind causing invalidation or data
> +# corruption issues.
> +#
> +#-----------------------------------------------------------------------
> +# Copyright (c) 2014 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 $tmp.*
> +}
> +
> +# get standard environment, filters and checks
> +. ./common/rc
> +. ./common/filter
> +
> +# real QA test starts here
> +
> +# Modify as appropriate.
> +_supported_fs generic
> +_supported_os IRIX Linux
> +_require_scratch
> +_require_xfs_io_command "fcollapse"
> +
> +testfile=$SCRATCH_MNT/testfile
> +
> +_scratch_mkfs > /dev/null 2>&1
> +_scratch_mount
> +
> +$XFS_IO_PROG -f \
> +	-c "pwrite 185332 55756" \
> +	-c "fcollapse 28672 40960" \
> +	-c "pwrite 133228 63394" \
> +	-c "fcollapse 0 4096" \
> +$testfile | _filter_xfs_io
> +

A comment (or per-line comments as in the other tests) would be good
here to explain what's going on. E.g.:

# This sequence of operations exploits a known failure to handle partial
# page writeback on sub page sized fsb filesystems. This occurs when a
# page has a non-contiguous mix of dirty and clean blocks (e.g., dirty
# block, clean block, dirty block, ...).
#
# The first write and collapse creates a dirty range in the file,
# flushes and truncates the file down to an unaligned boundary. The
# truncate implicit in the collapse dirties a block somewhere in the 2nd
# half of the new EOF page. The second write creates more dirty data,
# but specifically only writes to the first few bytes of the EOF page.
# This and the previous truncate creates the mixed page state described
# above.
#
# The final collapse attempts to flush and invalidate the entire cached
# set for the file. If the writeback of the mixed page does not
# complete, the invalidate fails with -EBUSY upon hitting a dirty page
# and aborts the collapse.
$XFS_IO_PROG -f \
	-c "pwrite 185332 55756"	# write extend file
	-c "fcollapse 28672 40960"	# collapse to unaligned boundary
	-c "pwrite 133228 63394"	# dirty first part of new eof page
	-c "fcollapse 0 4096"		# try a collapse
$testfile | _filter_xfs_io

> +echo "==== Pre-Remount ==="
> +hexdump -C $testfile
> +_scratch_remount
> +echo "==== Post-Remount =="
> +hexdump -C $testfile
> +

Note that this test is a collapse failure moreso than the data
corruption error (e.g., collapse returns EBUSY), though the reason for
that occurrence (data sync failure) is certainly a data corruption
issue. The hexdump/remount checks are fine, I just want to clarify that
they aren't necessary and they will probably just reflect the fact that
the collapse failed. I think the comment is sufficient to provide that
context and avoid any confusion.

Other than that this looks pretty good, thanks for knocking these tests
out.

Brian

> +status=0
> +exit
> +
> diff --git a/tests/generic/031.out b/tests/generic/031.out
> new file mode 100644
> index 0000000..194bfa4
> --- /dev/null
> +++ b/tests/generic/031.out
> @@ -0,0 +1,19 @@
> +QA output created by 031
> +wrote 55756/55756 bytes at offset 185332
> +XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
> +wrote 63394/63394 bytes at offset 133228
> +XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
> +==== Pre-Remount ===
> +00000000  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
> +*
> +0001f860  00 00 00 00 00 00 00 00  00 00 00 00 cd cd cd cd  |................|
> +0001f870  cd cd cd cd cd cd cd cd  cd cd cd cd cd cd cd cd  |................|
> +*
> +0002fdc0
> +==== Post-Remount ==
> +00000000  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
> +*
> +0001f860  00 00 00 00 00 00 00 00  00 00 00 00 cd cd cd cd  |................|
> +0001f870  cd cd cd cd cd cd cd cd  cd cd cd cd cd cd cd cd  |................|
> +*
> +0002fdc0
> diff --git a/tests/generic/group b/tests/generic/group
> index ba1c913..dd6db3b 100644
> --- a/tests/generic/group
> +++ b/tests/generic/group
> @@ -33,6 +33,7 @@
>  028 auto quick
>  029 auto quick rw
>  030 auto quick rw
> +031 auto quick prealloc rw
>  053 acl repair auto quick
>  062 attr udf auto quick
>  068 other auto freeze dangerous stress
> -- 
> 2.0.0
> 
> --
> 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