| Message ID | 20170316090851.21608-4-quwenruo@cn.fujitsu.com (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
On Thu, Mar 16, 2017 at 05:08:51PM +0800, Qu Wenruo wrote: > Btrfs balance will reloate date extent, but its hash is removed too late ^^^^^^^ relocate > at run_delayed_ref() time, which will cause extent ref increased > during balance, cause either find_data_references() gives WARN_ON() > or even run_delayed_refs() fails and cause transaction abort. > > Add such concurrency test for inband dedupe and data balance. > > Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com> > --- > tests/btrfs/202 | 109 ++++++++++++++++++++++++++++++++++++++++++++++++++++ > tests/btrfs/202.out | 3 ++ > tests/btrfs/group | 1 + > 3 files changed, 113 insertions(+) > create mode 100755 tests/btrfs/202 > create mode 100644 tests/btrfs/202.out > > diff --git a/tests/btrfs/202 b/tests/btrfs/202 > new file mode 100755 > index 0000000..60bb924 > --- /dev/null > +++ b/tests/btrfs/202 > @@ -0,0 +1,109 @@ > +#! /bin/bash > +# FS QA Test 203 > +# > +# Btrfs inband dedupe with balance concurrency test > +# > +# This can spot inband dedupe error which will increase delayed ref on > +# an data extent inside RO block group > +# > +#----------------------------------------------------------------------- > +# Copyright (c) 2017 Fujitsu. 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 / > + kill $populate_pid &> /dev/null > + kill $balance_pid &> /dev/null > + wait > + # Check later comment for reason > + $BTRFS_UTIL_PROG balance cancel $SCRATCH_MNT &> /dev/null > + rm -f $tmp.* I sometimes saw "btrfs balance start" hold $SCRATCH_MNT from umounting and result in fs inconsistency after test. Some debug codes show that the "btrfs balance start $SCRATCH_MNT" is still running. If I add what btrfs/061 does: while ps aux | grep "balance start" | grep -qv grep; do sleep 1 done test finishes fine. But btrfs-balance(8) says balance cancel should block and wait for the backgroud balance process to finish. Seems like a "balance cancel" bug? Otherwise these three patches look fine to me overall, except the wanted "dedupe feature check", and btrfs/200 is quick enough to fit in 'quick' group. I think the "dedupe feature check" should _notrun the tests if current btrfs has incompatible features set, e.g. compress and/or nodatacow. Thanks, Eryu -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" 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/btrfs/202 b/tests/btrfs/202 new file mode 100755 index 0000000..60bb924 --- /dev/null +++ b/tests/btrfs/202 @@ -0,0 +1,109 @@ +#! /bin/bash +# FS QA Test 203 +# +# Btrfs inband dedupe with balance concurrency test +# +# This can spot inband dedupe error which will increase delayed ref on +# an data extent inside RO block group +# +#----------------------------------------------------------------------- +# Copyright (c) 2017 Fujitsu. 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 / + kill $populate_pid &> /dev/null + kill $balance_pid &> /dev/null + wait + # Check later comment for reason + $BTRFS_UTIL_PROG balance cancel $SCRATCH_MNT &> /dev/null + rm -f $tmp.* +} + +# get standard environment, filters and checks +. ./common/rc +. ./common/filter +. ./common/reflink + +# remove previous $seqres.full before test +rm -f $seqres.full + +# real QA test starts here + +_supported_fs btrfs +_supported_os Linux +_require_scratch +_require_cp_reflink +_require_btrfs_command dedupe +_require_btrfs_fs_feature dedupe + +dedupe_bs=128k +file_size_in_kilo=4096 +init_file=$SCRATCH_MNT/foo +run_time=$((60 * $TIME_FACTOR)) + +_scratch_mkfs >> $seqres.full 2>&1 +_scratch_mount + +do_dedupe_balance_test() +{ + _run_btrfs_util_prog dedupe enable -b $dedupe_bs -s inmemory $SCRATCH_MNT + + # create the initial file and fill hash pool + $XFS_IO_PROG -f -c "pwrite -S 0x0 -b $dedupe_bs 0 $dedupe_bs" -c "fsync" \ + $init_file | _filter_xfs_io + + _btrfs_stress_balance $SCRATCH_MNT >/dev/null 2>&1 & + balance_pid=$! + + # Populate fs with all 0 data, to trigger enough in-band dedupe work + # to race with balance + _populate_fs -n 5 -f 10000000 -d 1 -r $SCRATCH_MNT \ + -s $file_size_in_kilo &> /dev/null & + populate_pid=$! + + sleep $run_time + + kill $populate_pid + kill $balance_pid + wait + + # Sometimes even we killed $balance_pid and wait returned, + # balance may still be running, use balance cancel to wait it. + # As this is just a workaround, we don't want it pollute seqres + # so call $BTRFS_UTIL_PROG directly + $BTRFS_UTIL_PROG balance cancel $SCRATCH_MNT &> /dev/null + + rm $SCRATCH_MNT/* -rf &> /dev/null + _run_btrfs_util_prog dedupe disable $SCRATCH_MNT +} + +do_dedupe_balance_test + +# success, all done +status=0 +exit diff --git a/tests/btrfs/202.out b/tests/btrfs/202.out new file mode 100644 index 0000000..ce03aa7 --- /dev/null +++ b/tests/btrfs/202.out @@ -0,0 +1,3 @@ +QA output created by 202 +wrote 131072/131072 bytes at offset 0 +XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) diff --git a/tests/btrfs/group b/tests/btrfs/group index 1856d7e..930057a 100644 --- a/tests/btrfs/group +++ b/tests/btrfs/group @@ -143,3 +143,4 @@ 138 auto compress 200 auto ib-dedupe 201 auto ib-dedupe +202 auto ib-dedupe balance
Btrfs balance will reloate date extent, but its hash is removed too late at run_delayed_ref() time, which will cause extent ref increased during balance, cause either find_data_references() gives WARN_ON() or even run_delayed_refs() fails and cause transaction abort. Add such concurrency test for inband dedupe and data balance. Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com> --- tests/btrfs/202 | 109 ++++++++++++++++++++++++++++++++++++++++++++++++++++ tests/btrfs/202.out | 3 ++ tests/btrfs/group | 1 + 3 files changed, 113 insertions(+) create mode 100755 tests/btrfs/202 create mode 100644 tests/btrfs/202.out