| Message ID | 1471004010-52985-1-git-send-email-bfoster@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Fri, Aug 12, 2016 at 08:13:30AM -0400, Brian Foster wrote: > XFS had a bug that lead to a possible out-of-order log recovery > situation (e.g., replay a stale modification from the log over more > recent metadata in destination buffer). This resulted in false > corruption reports during log recovery and thus mount failure. > > This condition is caused by system crash or filesystem shutdown shortly > after a successful log recovery. Add a test to run a combined workload, > fs shutdown and log recovery loop known to reproduce the problem on > affected kernels. > > Signed-off-by: Brian Foster <bfoster@redhat.com> > --- > > This test reproduces the problem described and addressed in the > following patchset: > > http://oss.sgi.com/pipermail/xfs/2016-August/050840.html > > It runs anywhere from 50-100s in the couple of environments I've tested > on so far and reproduces the problem for me with 100% reliability. Note > that the bug only affects crc=1 kernels. Looks good overall, and tested with the above patchset applied and test passed without problems. Some minor issues inline > > Brian > > tests/xfs/999 | 87 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ > tests/xfs/999.out | 2 ++ > tests/xfs/group | 1 + > 3 files changed, 90 insertions(+) > create mode 100755 tests/xfs/999 > create mode 100644 tests/xfs/999.out > > diff --git a/tests/xfs/999 b/tests/xfs/999 > new file mode 100755 > index 0000000..f9dd7f7 > --- /dev/null > +++ b/tests/xfs/999 > @@ -0,0 +1,87 @@ > +#! /bin/bash > +# FS QA Test No. 999 > +# > +# Test XFS log recovery ordering on v5 superblock filesystems. XFS had a problem > +# where it would incorrectly replay older modifications from the log over more > +# recent versions of metadata due to failure to update metadata LSNs during log > +# recovery. This could result in false positive reports of corruption during log > +# recovery and permanent mount failure. > +# > +# To test this situation, run frequent shutdowns immediately after log recovery. > +# Ensure that log recovery does not recover stale modifications and cause > +# spurious corruption reports and/or mount failures. > +# > +#----------------------------------------------------------------------- > +# Copyright (c) 2016 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.* > + killall -9 fsstress > /dev/null 2>&1 We need a '_require_command "$KILLALL_PROGA" killall' and use $KILLALL_PROG in the test. > + _scratch_unmount > /dev/null 2>&1 > +} > + > +# get standard environment, filters and checks > +. ./common/rc > + > +# Modify as appropriate. > +_supported_fs xfs I'm wondering if this test can be made generic by adding a "_require_scratch_shutdown"? Like generic/042 to generic/051 Thanks, Eryu -- 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
On Mon, Aug 15, 2016 at 01:29:33PM +0800, Eryu Guan wrote: > On Fri, Aug 12, 2016 at 08:13:30AM -0400, Brian Foster wrote: > > XFS had a bug that lead to a possible out-of-order log recovery > > situation (e.g., replay a stale modification from the log over more > > recent metadata in destination buffer). This resulted in false > > corruption reports during log recovery and thus mount failure. > > > > This condition is caused by system crash or filesystem shutdown shortly > > after a successful log recovery. Add a test to run a combined workload, > > fs shutdown and log recovery loop known to reproduce the problem on > > affected kernels. > > > > Signed-off-by: Brian Foster <bfoster@redhat.com> > > --- > > > > This test reproduces the problem described and addressed in the > > following patchset: > > > > http://oss.sgi.com/pipermail/xfs/2016-August/050840.html > > > > It runs anywhere from 50-100s in the couple of environments I've tested > > on so far and reproduces the problem for me with 100% reliability. Note > > that the bug only affects crc=1 kernels. > > Looks good overall, and tested with the above patchset applied and test > passed without problems. Some minor issues inline > > > > > Brian > > > > tests/xfs/999 | 87 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ > > tests/xfs/999.out | 2 ++ > > tests/xfs/group | 1 + > > 3 files changed, 90 insertions(+) > > create mode 100755 tests/xfs/999 > > create mode 100644 tests/xfs/999.out > > > > diff --git a/tests/xfs/999 b/tests/xfs/999 > > new file mode 100755 > > index 0000000..f9dd7f7 > > --- /dev/null > > +++ b/tests/xfs/999 > > @@ -0,0 +1,87 @@ > > +#! /bin/bash > > +# FS QA Test No. 999 > > +# > > +# Test XFS log recovery ordering on v5 superblock filesystems. XFS had a problem > > +# where it would incorrectly replay older modifications from the log over more > > +# recent versions of metadata due to failure to update metadata LSNs during log > > +# recovery. This could result in false positive reports of corruption during log > > +# recovery and permanent mount failure. > > +# > > +# To test this situation, run frequent shutdowns immediately after log recovery. > > +# Ensure that log recovery does not recover stale modifications and cause > > +# spurious corruption reports and/or mount failures. > > +# > > +#----------------------------------------------------------------------- > > +# Copyright (c) 2016 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 $tmp.* > > + killall -9 fsstress > /dev/null 2>&1 > > We need a '_require_command "$KILLALL_PROGA" killall' and use > $KILLALL_PROG in the test. > Ok. > > + _scratch_unmount > /dev/null 2>&1 > > +} > > + > > +# get standard environment, filters and checks > > +. ./common/rc > > + > > +# Modify as appropriate. > > +_supported_fs xfs > > I'm wondering if this test can be made generic by adding a > "_require_scratch_shutdown"? Like generic/042 to generic/051 > Hmm, probably. I'll give it a try, thanks! Brian > Thanks, > Eryu -- 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
diff --git a/tests/xfs/999 b/tests/xfs/999 new file mode 100755 index 0000000..f9dd7f7 --- /dev/null +++ b/tests/xfs/999 @@ -0,0 +1,87 @@ +#! /bin/bash +# FS QA Test No. 999 +# +# Test XFS log recovery ordering on v5 superblock filesystems. XFS had a problem +# where it would incorrectly replay older modifications from the log over more +# recent versions of metadata due to failure to update metadata LSNs during log +# recovery. This could result in false positive reports of corruption during log +# recovery and permanent mount failure. +# +# To test this situation, run frequent shutdowns immediately after log recovery. +# Ensure that log recovery does not recover stale modifications and cause +# spurious corruption reports and/or mount failures. +# +#----------------------------------------------------------------------- +# Copyright (c) 2016 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.* + killall -9 fsstress > /dev/null 2>&1 + _scratch_unmount > /dev/null 2>&1 +} + +# get standard environment, filters and checks +. ./common/rc + +# Modify as appropriate. +_supported_fs xfs +_supported_os Linux + +_require_scratch + +rm -f $seqres.full + +echo "Silence is golden." + +_scratch_mkfs_xfs >> $seqres.full 2>&1 +_scratch_mount || _fail "mount failed" + +for i in $(seq 1 50); do + ($FSSTRESS_PROG -d $SCRATCH_MNT -n 999999 -p 4 >> $seqres.full &) \ + > /dev/null 2>&1 + + # purposely include 0 second sleeps to test shutdown immediately after + # recovery + sleep $((RANDOM % 3)) + $XFS_IO_PROG -xc shutdown $SCRATCH_MNT + + ps -e | grep fsstress > /dev/null 2>&1 + while [ $? == 0 ]; do + killall -9 fsstress > /dev/null 2>&1 + wait > /dev/null 2>&1 + ps -e | grep fsstress > /dev/null 2>&1 + done + + # quit if mount fails so we don't shutdown the host fs + _scratch_cycle_mount || _fail "cycle mount failed" +done + +# success, all done +status=0 +exit diff --git a/tests/xfs/999.out b/tests/xfs/999.out new file mode 100644 index 0000000..d254382 --- /dev/null +++ b/tests/xfs/999.out @@ -0,0 +1,2 @@ +QA output created by 999 +Silence is golden. diff --git a/tests/xfs/group b/tests/xfs/group index 6905a62..aad41b5 100644 --- a/tests/xfs/group +++ b/tests/xfs/group @@ -308,3 +308,4 @@ 325 auto quick clone 326 auto quick clone 327 auto quick clone +999 auto log metadata
XFS had a bug that lead to a possible out-of-order log recovery situation (e.g., replay a stale modification from the log over more recent metadata in destination buffer). This resulted in false corruption reports during log recovery and thus mount failure. This condition is caused by system crash or filesystem shutdown shortly after a successful log recovery. Add a test to run a combined workload, fs shutdown and log recovery loop known to reproduce the problem on affected kernels. Signed-off-by: Brian Foster <bfoster@redhat.com> --- This test reproduces the problem described and addressed in the following patchset: http://oss.sgi.com/pipermail/xfs/2016-August/050840.html It runs anywhere from 50-100s in the couple of environments I've tested on so far and reproduces the problem for me with 100% reliability. Note that the bug only affects crc=1 kernels. Brian tests/xfs/999 | 87 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ tests/xfs/999.out | 2 ++ tests/xfs/group | 1 + 3 files changed, 90 insertions(+) create mode 100755 tests/xfs/999 create mode 100644 tests/xfs/999.out