Re: [PATCH] xfstests/shared: dedup integrity test by duperemove

["Darrick J. Wong" <darrick.wong@oracle.com>]

On Wed, May 30, 2018 at 01:13:45AM +0800, Zorro Lang wrote:
> On Tue, May 29, 2018 at 09:30:23AM -0700, Darrick J. Wong wrote:
> > On Wed, May 30, 2018 at 12:13:04AM +0800, Zorro Lang wrote:
> > > On Tue, May 29, 2018 at 08:07:59AM -0700, Darrick J. Wong wrote:
> > > > On Mon, May 28, 2018 at 12:54:27PM +0800, Zorro Lang wrote:
> > > > > Duperemove is a tool for finding duplicated extents and submitting
> > > > > them for deduplication, and it supports XFS. This case trys to
> > > > > verify the integrity of XFS after running duperemove.
> > > > > 
> > > > > Signed-off-by: Zorro Lang <zlang@redhat.com>
> > > > > ---
> > > > > 
> > > > > Hi,
> > > > > 
> > > > > There's not many softwares support XFS dedup now, duperemove is a rare one.
> > > > > So I write this case by using duperemove.
> > > > > 
> > > > > I use fsstress to make many files and data randomly, I don't know if there're
> > > > > better things I can use? Because fsstress only write '0xff' into files, maybe
> > > > > I should add an option to make fsstress can write random character?
> > > > 
> > > > Heh.  But you probably don't want totally random contents because then
> > > > dupremove doesn't do much.
> > > 
> > > No matter how random contents I get, I will copy once :)
> > 
> > I suppose so.  Once fsstress pokes reflink enough there ought to be a
> > fair number of easy targets for dedupe... on the other hand I think it's
> > a useful test for "literally everything in this fs is identical, dedupe
> > everything" :)
> 
> Do you mean dd (with same char) a big file to fill the whole scratch_dev, then
> dedupe?

I was mostly thinking one test where the entire fs is dedupable (because
we only ever write 0xff or whatever), and a second one where we make
duperemove hunt for things.

> Hmm... how about iteration test as below?
> 
> # fsstress -d $mnt/dir0 ...
> # for ((i=1; i<100; i++));do
>     cp -a dir$((i-1)) dir$i
>     find $dir$i -type f -exec md5sum {} \; > $TEST_DIR/${seq}md5.sum$i
>     duperemove -dr $mnt/
>     md5sum -c $TEST_DIR/${seq}md5.sum$i
>   done
> # scratch_cycle_mount
> # for ((i=1; i<100; i++));do
>     md5sum -c $TEST_DIR/${seq}md5.sum$i
>   done
> 
> But this will cost lots of test time. For save time, need to reduce files count
> and size.

Constrict the fs size?

> Hmm... if we have different dedupe test ways, maybe we can write not only one
> cases.
> 
> > 
> > > > 
> > > > > 
> > > > > Please tell me, if you have better ideas:)
> > > > > 
> > > > > PS: This case test passed on XFS(with reflink=1) and btrfs. And the duperemove
> > > > > can reclaim some space in the test, see below:
> > > > > 
> > > > >   Before duperemove
> > > > >     Filesystem                 1K-blocks    Used Available Use% Mounted on
> > > > >     /dev/mapper/xxxx-xfscratch 31441920K 583692K 30858228K   2% /mnt/scratch
> > > > > 
> > > > >   After duperemove
> > > > >     Filesystem                 1K-blocks    Used Available Use% Mounted on
> > > > >     /dev/mapper/xxxx-xfscratch 31441920K 345728K 31096192K   2% /mnt/scratch
> > > > > 
> > > > > Thanks,
> > > > > Zorro
> > > > > 
> > > > >  common/config        |  1 +
> > > > >  tests/shared/008     | 88 ++++++++++++++++++++++++++++++++++++++++++++++++++++
> > > > >  tests/shared/008.out |  2 ++
> > > > >  tests/shared/group   |  1 +
> > > > >  4 files changed, 92 insertions(+)
> > > > >  create mode 100755 tests/shared/008
> > > > >  create mode 100644 tests/shared/008.out
> > > > > 
> > > > > diff --git a/common/config b/common/config
> > > > > index 02c378a9..def559c1 100644
> > > > > --- a/common/config
> > > > > +++ b/common/config
> > > > > @@ -207,6 +207,7 @@ export SQLITE3_PROG="`set_prog_path sqlite3`"
> > > > >  export TIMEOUT_PROG="`set_prog_path timeout`"
> > > > >  export SETCAP_PROG="`set_prog_path setcap`"
> > > > >  export GETCAP_PROG="`set_prog_path getcap`"
> > > > > +export DUPEREMOVE_PROG="`set_prog_path duperemove`"
> > > > >  
> > > > >  # use 'udevadm settle' or 'udevsettle' to wait for lv to be settled.
> > > > >  # newer systems have udevadm command but older systems like RHEL5 don't.
> > > > > diff --git a/tests/shared/008 b/tests/shared/008
> > > > > new file mode 100755
> > > > > index 00000000..dace5429
> > > > > --- /dev/null
> > > > > +++ b/tests/shared/008
> > > > > @@ -0,0 +1,88 @@
> > > > > +#! /bin/bash
> > > > > +# FS QA Test 008
> > > > > +#
> > > > > +# Dedup integrity test by duperemove
> > > > > +#
> > > > > +#-----------------------------------------------------------------------
> > > > > +# Copyright (c) 2018 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/reflink
> > > > > +
> > > > > +# remove previous $seqres.full before test
> > > > > +rm -f $seqres.full
> > > > > +
> > > > > +# real QA test starts here
> > > > > +
> > > > > +# duperemove only supports btrfs and xfs (with reflink feature).
> > > > > +# Add other filesystems if it supports more later.
> > > > > +_supported_fs xfs btrfs
> > > > > +_supported_os Linux
> > > > 
> > > > _require_command "$DUPEREMOVE_PROG" duperemove ?
> > > 
> > > Yes, it would be better to use this template, not check
> > > [ "$DUPEREMOVE_PROG" = "" ].
> > > 
> > > > 
> > > > > +_require_scratch_reflink
> > > > 
> > > > _require_scratch_dedupe
> > > 
> > > Yes, I should check XFS_IOC_FILE_EXTENT_SAME, not XFS_IOC_CLONE*.
> > > 
> > > > 
> > > > > +
> > > > > +[ "$DUPEREMOVE_PROG" = "" ] && _notrun "duperemove not found"
> > > > > +_scratch_mkfs > $seqres.full 2>&1
> > > > > +_scratch_mount >> $seqres.full 2>&1
> > > > > +
> > > > > +testdir=$SCRATCH_MNT/test-$seq
> > > > > +mkdir $testdir
> > > > > +
> > > > > +fsstress_opts="-w -r -f mknod=0"
> > > > > +# Create some files and make a duplicate
> > > > > +$FSSTRESS_PROG $fsstress_opts -d $testdir \
> > > > > +	       -n $((500 * LOAD_FACTOR)) -p 10 >/dev/null 2>&1
> > > > > +duptestdir=${testdir}.dup
> > > > > +cp -a $testdir $duptestdir
> > > > > +
> > > > > +# Make some difference in two directories
> > > > > +$FSSTRESS_PROG $fsstress_opts -d $testdir -n 200 -p 5 >/dev/null 2>&1
> > > > > +$FSSTRESS_PROG $fsstress_opts -d $duptestdir -n 200 -p 5 >/dev/null 2>&1
> > > > > +
> > > > > +# Record all files' md5 checksum
> > > > > +find $testdir -type f -exec md5sum {} \; > $TEST_DIR/${seq}md5.sum
> > > > > +find $duptestdir -type f -exec md5sum {} \; > $TEST_DIR/dup${seq}md5.sum
> > > > > +
> > > > > +# Dedup
> > > > > +echo "== Duperemove output ==" >> $seqres.full
> > > > > +$DUPEREMOVE_PROG -dr $SCRATCH_MNT/ >>$seqres.full 2>&1
> > > > > +
> > > > > +# Verify all files' integrity
> > > > > +md5sum -c --quiet $TEST_DIR/${seq}md5.sum
> > > > > +md5sum -c --quiet $TEST_DIR/dup${seq}md5.sum
> > > > 
> > > > Can we _scratch_mount_cycle and md5sum -c again so that we test that the
> > > > pagecache contents don't mutate and a fresh read from the disk also
> > > > doesn't show mutations?
> > > 
> > > If so, is the md5sum data safe? Should I do cycle_mount before get md5 checksum?
> > > What 'fresh read' do you mean, from above duperemove processes? Or you hope to
> > > read all files once before cycle_mount?
> > 
> > Since this is dedupe, the md5sum should never change.  The existing
> > md5sum -c check makes sure that the dedupe operation doesn't
> > mishandle/corrupt the page cache such that it suddenly starts returning
> > incorrect contents; and the post-cycle md5sum -c check I propose would
> > flush the page cache and make sure that the on-disk contents also have
> > not changed.
> 
> Make sense, I'll do a cycle_mount and md5sum -c check again.

Ok.

--D

> > 
> > --D
> > 
> > > Thanks,
> > > Zorro
> > > 
> > > > 
> > > > --D
> > > > 
> > > > > +
> > > > > +echo "Silence is golden"
> > > > > +
> > > > > +status=0
> > > > > +exit
> > > > > diff --git a/tests/shared/008.out b/tests/shared/008.out
> > > > > new file mode 100644
> > > > > index 00000000..dd68d5a4
> > > > > --- /dev/null
> > > > > +++ b/tests/shared/008.out
> > > > > @@ -0,0 +1,2 @@
> > > > > +QA output created by 008
> > > > > +Silence is golden
> > > > > diff --git a/tests/shared/group b/tests/shared/group
> > > > > index b3663a03..de7fe79f 100644
> > > > > --- a/tests/shared/group
> > > > > +++ b/tests/shared/group
> > > > > @@ -10,6 +10,7 @@
> > > > >  005 dangerous_fuzzers
> > > > >  006 auto enospc
> > > > >  007 dangerous_fuzzers
> > > > > +008 auto quick dedupe
> > > > >  032 mkfs auto quick
> > > > >  272 auto enospc rw
> > > > >  289 auto quick
> > > > > -- 
> > > > > 2.14.3
> > > > > 
> > > > > --
> > > > > To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> > > > > the body of a message to majordomo@vger.kernel.org
> > > > > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> > > > --
> > > > To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> > > > the body of a message to majordomo@vger.kernel.org
> > > > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> > > --
> > > To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> > > the body of a message to majordomo@vger.kernel.org
> > > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> > --
> > 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
> --
> 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