| Message ID | 1363710260-10225-1-git-send-email-list.btrfs@jan-o-sch.net (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
On 3/19/13 11:24 AM, Jan Schmidt wrote: > The backref resolver test for btrfs was run on a static file system so far. > Resolving backrefs on a busy file system is what happens in reality and that > is what should be checked by this test. > > I added a parameter such that the script can easily be changed to the former > behavior for development purposes. In practice I'm not sure anyone will ever tune it, but that's ok I think. > Furthermore, this increases two constants which make the test simply cycle a > few seconds longer, increasing the chance to hit on something suspicious in > case we broke something. Normally we don't change existing tests lest new failures look like regressions when they aren't, but hey, "btrfs is an experimental filesystem" so maybe it's ok in this case. ;) At some point when things are settled down, we wouldn't want to make a change like this. But for now it doesn't bother me. Looks ok to me: Reviewed-by: Eric Sandeen <sandeen@redhat.com> Thanks, -Eric > Signed-off-by: Jan Schmidt <list.btrfs@jan-o-sch.net> > --- > 276 | 29 ++++++++++++++++++++++++++--- > 1 files changed, 26 insertions(+), 3 deletions(-) > > diff --git a/276 b/276 > index 65da464..40b30cf 100755 > --- a/276 > +++ b/276 > @@ -33,9 +33,16 @@ echo "QA output created by $seq" > here=`pwd` > tmp=/tmp/$$ > status=1 > +noise_pid=0 > > _cleanup() > { > + if [ $noise_pid -ne 0 ]; then > + echo "background noise kill $noise_pid" >>$seq.full > + kill $noise_pid > + noise_pid=0 > + wait > + fi > echo "*** unmount" > umount $SCRATCH_MNT 2>/dev/null > rm -f $tmp.* > @@ -171,6 +178,7 @@ workout() > nfiles=$2 > procs=$3 > snap_name=$4 > + do_bg_noise=$5 > > umount $SCRATCH_DEV >/dev/null 2>&1 > echo "*** mkfs -dsize=$fsz" >>$seq.full > @@ -179,7 +187,7 @@ workout() > || _fail "size=$fsz mkfs failed" > run_check _scratch_mount > # -w ensures that the only ops are ones which cause write I/O > - run_check $FSSTRESS_PROG -d $SCRATCH_MNT -w -p $procs -n 1000 \ > + run_check $FSSTRESS_PROG -d $SCRATCH_MNT -w -p $procs -n 2000 \ > $FSSTRESS_AVOID > > run_check $BTRFS_UTIL_PROG subvol snap $SCRATCH_MNT \ > @@ -190,9 +198,10 @@ workout() > > # make some noise but ensure we're not touching existing data > # extents. > - run_check $FSSTRESS_PROG -d $SCRATCH_MNT -p $procs -n 2000 \ > + run_check $FSSTRESS_PROG -d $SCRATCH_MNT -p $procs -n 4000 \ > -z -f chown=3 -f link=1 -f mkdir=2 -f mknod=2 \ > -f rename=2 -f setxattr=1 -f symlink=2 > + > clean_dir="$SCRATCH_MNT/next" > mkdir $clean_dir > # now make more files to get a higher tree > @@ -201,6 +210,18 @@ workout() > run_check umount $SCRATCH_DEV >/dev/null 2>&1 > run_check _scratch_mount "-o atime" > > + if [ $do_bg_noise -ne 0 ]; then > + # make background noise while backrefs are being walked > + while /bin/true; do > + echo background fsstress >>$seq.full > + run_check $FSSTRESS_PROG -d $SCRATCH_MNT/bgnoise -n 999 > + echo background rm >>$seq.full > + rm -rf $SCRATCH_MNT/bgnoise/ > + done & > + noise_pid=`jobs -p %1` > + echo "background noise by $noise_pid" >>$seq.full > + fi > + > cnt=0 > errcnt=0 > dir="$SCRATCH_MNT/$snap_name/" > @@ -230,6 +251,7 @@ workout() > break > fi > done > + > if [ $errcnt -gt 0 ]; then > _fail "test failed: $errcnt error(s)" > fi > @@ -241,8 +263,9 @@ snap_name="snap1" > filesize=`expr 2000 \* 1024 \* 1024` > nfiles=4 > numprocs=1 > +do_bg_noise=1 > > -workout $filesize $nfiles $numprocs $snap_name > +workout $filesize $nfiles $numprocs $snap_name $do_bg_noise > > echo "*** done" > status=0 > -- 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
On 19.03.2013 18:09, Eric Sandeen wrote: >> Furthermore, this increases two constants which make the test simply cycle a >> few seconds longer, increasing the chance to hit on something suspicious in >> case we broke something. > > Normally we don't change existing tests lest new failures look like regressions > when they aren't, but hey, "btrfs is an experimental filesystem" so maybe it's > ok in this case. ;) At some point when things are settled down, we wouldn't > want to make a change like this. But for now it doesn't bother me. (justification) I thought about adding this modification as a separate test - and I have no strict objections against doing so. It's just that I hate duplicating code and I couldn't think of a good way to share all that code between two individual tests. Plus: We do need the new version, because it shows backref walking in fact is currently broken, while the old version would never fail where the new one wouldn't. Thanks for taking a look! -Jan -- 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
On Tue, Mar 19, 2013 at 08:17:57PM +0100, Jan Schmidt wrote: > On 19.03.2013 18:09, Eric Sandeen wrote: > >> Furthermore, this increases two constants which make the test simply cycle a > >> few seconds longer, increasing the chance to hit on something suspicious in > >> case we broke something. > > > > Normally we don't change existing tests lest new failures look like regressions > > when they aren't, but hey, "btrfs is an experimental filesystem" so maybe it's > > ok in this case. ;) At some point when things are settled down, we wouldn't > > want to make a change like this. But for now it doesn't bother me. > > (justification) I thought about adding this modification as a separate > test - and I have no strict objections against doing so. It's just that > I hate duplicating code and I couldn't think of a good way to share all > that code between two individual tests. The current way is to use a common.<blah> file to do it. However, using test templates is the way I want to do it in future - it will be perfect for these sorts of test variations: http://oss.sgi.com/archives/xfs/2013-03/msg00578.html Cheers, Dave.
Thanks for submitting this patch Jan, this patch has been committed.
Thanks
--Rich
commit cfe73f711eff55c374d9c0db8921089ec4ad8607
Author: Jan Schmidt <list.btrfs@jan-o-sch.net>
Date: Tue Mar 19 16:24:20 2013 +0000
xfstests: add background noise to test 276 (btrfs backref resolving)
--
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/276 b/276 index 65da464..40b30cf 100755 --- a/276 +++ b/276 @@ -33,9 +33,16 @@ echo "QA output created by $seq" here=`pwd` tmp=/tmp/$$ status=1 +noise_pid=0 _cleanup() { + if [ $noise_pid -ne 0 ]; then + echo "background noise kill $noise_pid" >>$seq.full + kill $noise_pid + noise_pid=0 + wait + fi echo "*** unmount" umount $SCRATCH_MNT 2>/dev/null rm -f $tmp.* @@ -171,6 +178,7 @@ workout() nfiles=$2 procs=$3 snap_name=$4 + do_bg_noise=$5 umount $SCRATCH_DEV >/dev/null 2>&1 echo "*** mkfs -dsize=$fsz" >>$seq.full @@ -179,7 +187,7 @@ workout() || _fail "size=$fsz mkfs failed" run_check _scratch_mount # -w ensures that the only ops are ones which cause write I/O - run_check $FSSTRESS_PROG -d $SCRATCH_MNT -w -p $procs -n 1000 \ + run_check $FSSTRESS_PROG -d $SCRATCH_MNT -w -p $procs -n 2000 \ $FSSTRESS_AVOID run_check $BTRFS_UTIL_PROG subvol snap $SCRATCH_MNT \ @@ -190,9 +198,10 @@ workout() # make some noise but ensure we're not touching existing data # extents. - run_check $FSSTRESS_PROG -d $SCRATCH_MNT -p $procs -n 2000 \ + run_check $FSSTRESS_PROG -d $SCRATCH_MNT -p $procs -n 4000 \ -z -f chown=3 -f link=1 -f mkdir=2 -f mknod=2 \ -f rename=2 -f setxattr=1 -f symlink=2 + clean_dir="$SCRATCH_MNT/next" mkdir $clean_dir # now make more files to get a higher tree @@ -201,6 +210,18 @@ workout() run_check umount $SCRATCH_DEV >/dev/null 2>&1 run_check _scratch_mount "-o atime" + if [ $do_bg_noise -ne 0 ]; then + # make background noise while backrefs are being walked + while /bin/true; do + echo background fsstress >>$seq.full + run_check $FSSTRESS_PROG -d $SCRATCH_MNT/bgnoise -n 999 + echo background rm >>$seq.full + rm -rf $SCRATCH_MNT/bgnoise/ + done & + noise_pid=`jobs -p %1` + echo "background noise by $noise_pid" >>$seq.full + fi + cnt=0 errcnt=0 dir="$SCRATCH_MNT/$snap_name/" @@ -230,6 +251,7 @@ workout() break fi done + if [ $errcnt -gt 0 ]; then _fail "test failed: $errcnt error(s)" fi @@ -241,8 +263,9 @@ snap_name="snap1" filesize=`expr 2000 \* 1024 \* 1024` nfiles=4 numprocs=1 +do_bg_noise=1 -workout $filesize $nfiles $numprocs $snap_name +workout $filesize $nfiles $numprocs $snap_name $do_bg_noise echo "*** done" status=0
The backref resolver test for btrfs was run on a static file system so far. Resolving backrefs on a busy file system is what happens in reality and that is what should be checked by this test. I added a parameter such that the script can easily be changed to the former behavior for development purposes. Furthermore, this increases two constants which make the test simply cycle a few seconds longer, increasing the chance to hit on something suspicious in case we broke something. Signed-off-by: Jan Schmidt <list.btrfs@jan-o-sch.net> --- 276 | 29 ++++++++++++++++++++++++++--- 1 files changed, 26 insertions(+), 3 deletions(-)