Message ID | 149808226863.8924.5799576767468365376.stgit@birch.djwong.org (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
On Wed, Jun 21, 2017 at 02:57:48PM -0700, Darrick J. Wong wrote: > From: Darrick J. Wong <darrick.wong@oracle.com> > > Test a regression in XFS where we blow out a transaction reservation if > we create a big file, share every other block, and delete the first > file. There's nothing particularly fs-specific about this stress test, > so put it in generic. > > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> This test took me 3019s to finish with v4.12-rc7 kernel, and another host "hung" at "Delete file1" (it had been more than 1 hour, and I lost my patience and hard-reboot the host). Are they expected results? If the bug is still existed in latest upstream kernel, I tend to merge it after the fix landing in linus tree. If v4.12-rc7 doesn't suffer from this bug, the test time should be reduced. And another minor nit below. > --- > tests/generic/931 | 94 +++++++++++++++++++++++++++++++++++++++++++++++++ > tests/generic/931.out | 6 +++ > tests/generic/group | 1 + > 3 files changed, 101 insertions(+) > create mode 100755 tests/generic/931 > create mode 100644 tests/generic/931.out > > > diff --git a/tests/generic/931 b/tests/generic/931 > new file mode 100755 > index 0000000..afadf81 > --- /dev/null > +++ b/tests/generic/931 > @@ -0,0 +1,94 @@ > +#! /bin/bash > +# FS QA Test No. 931 > +# > +# See how well we handle deleting a file with a million refcount extents. > +# > +#----------------------------------------------------------------------- > +# Copyright (c) 2017, Oracle and/or its affiliates. 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 -rf "$tmp".* $testdir/file1 'rm -rf' looks a bit scary, and we're only deleting regular files not directories, 'rm -f' should be sufficient. Thanks, Eryu > +} > + > +# get standard environment, filters and checks > +. ./common/rc > +. ./common/filter > +. ./common/attr > +. ./common/reflink > + > +# real QA test starts here > +_supported_os Linux > +_require_scratch_reflink > +_require_cp_reflink > +_require_test_program "punch-alternating" > + > +rm -f "$seqres.full" > + > +echo "Format and mount" > +_scratch_mkfs > "$seqres.full" 2>&1 > +_scratch_mount >> "$seqres.full" 2>&1 > + > +testdir="$SCRATCH_MNT/test-$seq" > +mkdir "$testdir" > + > +# Setup for one million blocks, but we'll accept stress testing down to > +# 2^17 blocks... that should be plenty for anyone. > +fnr=20 > +free_blocks=$(stat -f -c '%a' "$testdir") > +blksz=$(_get_block_size "$testdir") > +space_avail=$((free_blocks * blksz)) > +calc_space() { > + blocks_needed=$(( 2 ** (fnr + 1) )) > + space_needed=$((blocks_needed * blksz * 5 / 4)) > +} > +calc_space > +while test $space_needed -gt $space_avail; do > + fnr=$((fnr - 1)) > + calc_space > +done > +test $fnr -lt 17 && _notrun "Insufficient space for stress test; would only create $blocks_needed extents ($space_needed/$space_avail blocks)." > + > +echo "Create a many-block file" > +echo "creating $blocks_needed blocks..." >> "$seqres.full" > +$XFS_IO_PROG -f -c "pwrite -S 0x61 -b 4194304 0 $((2 ** (fnr + 1) * blksz))" "$testdir/file1" >> "$seqres.full" > + > +echo "Reflinking file" > +_cp_reflink $testdir/file1 $testdir/file2 > + > +echo "Punch file2" > +echo "Punching file2..." >> "$seqres.full" > +"$here/src/punch-alternating" "$testdir/file2" >> "$seqres.full" > +echo "...done" >> "$seqres.full" > +_scratch_cycle_mount > + > +echo "Delete file1" > +rm -rf $testdir/file1 > + > +# success, all done > +status=0 > +exit > diff --git a/tests/generic/931.out b/tests/generic/931.out > new file mode 100644 > index 0000000..c7b724e > --- /dev/null > +++ b/tests/generic/931.out > @@ -0,0 +1,6 @@ > +QA output created by 931 > +Format and mount > +Create a many-block file > +Reflinking file > +Punch file2 > +Delete file1 > diff --git a/tests/generic/group b/tests/generic/group > index ab1e9d3..b0d1844 100644 > --- a/tests/generic/group > +++ b/tests/generic/group > @@ -443,3 +443,4 @@ > 438 auto > 439 auto quick punch > 440 auto quick encrypt > +931 auto quick clone > -- 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 Thu, Jun 29, 2017 at 05:36:14PM +0800, Eryu Guan wrote: > On Wed, Jun 21, 2017 at 02:57:48PM -0700, Darrick J. Wong wrote: > > From: Darrick J. Wong <darrick.wong@oracle.com> > > > > Test a regression in XFS where we blow out a transaction reservation if > > we create a big file, share every other block, and delete the first > > file. There's nothing particularly fs-specific about this stress test, > > so put it in generic. > > > > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> > > This test took me 3019s to finish with v4.12-rc7 kernel, and another > host "hung" at "Delete file1" (it had been more than 1 hour, and I lost > my patience and hard-reboot the host). Are they expected results? No. The take-forever-or-crash behavior should be fixed by "xfs: try to avoid blowing out the transaction reservation when bunmaping a shared extent" in 4.13. Feel free to hang on to this one until -rc1. :) > If the bug is still existed in latest upstream kernel, I tend to merge > it after the fix landing in linus tree. If v4.12-rc7 doesn't suffer from > this bug, the test time should be reduced. <shrug> This is what I saw just now: FSTYP -- xfs (debug) PLATFORM -- Linux/x86_64 birch-mtr0 4.12.0-rc6-dgc MKFS_OPTIONS -- -f -m reflink=1,rmapbt=1, -i sparse=1, /dev/pmem1 MOUNT_OPTIONS -- /dev/pmem1 /opt generic/931 21s Ran: generic/931 Passed all 1 tests Though if it takes forever for everyone else, please kick this one out of auto/quick. At least in theory, before the patch the test will either blow out a transaction reservation and hang the system, or if it does succeed it'll have done so by scraping long and hard for log space. That is probably why it takes 3000+ seconds on your test box, unless you were also testing xfs for-next. > And another minor nit below. > > > --- > > tests/generic/931 | 94 +++++++++++++++++++++++++++++++++++++++++++++++++ > > tests/generic/931.out | 6 +++ > > tests/generic/group | 1 + > > 3 files changed, 101 insertions(+) > > create mode 100755 tests/generic/931 > > create mode 100644 tests/generic/931.out > > > > > > diff --git a/tests/generic/931 b/tests/generic/931 > > new file mode 100755 > > index 0000000..afadf81 > > --- /dev/null > > +++ b/tests/generic/931 > > @@ -0,0 +1,94 @@ > > +#! /bin/bash > > +# FS QA Test No. 931 > > +# > > +# See how well we handle deleting a file with a million refcount extents. > > +# > > +#----------------------------------------------------------------------- > > +# Copyright (c) 2017, Oracle and/or its affiliates. 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 -rf "$tmp".* $testdir/file1 > > 'rm -rf' looks a bit scary, and we're only deleting regular files not > directories, 'rm -f' should be sufficient. Yes. Will you fix it on the way in or should I resend? --D > Thanks, > Eryu > > > +} > > + > > +# get standard environment, filters and checks > > +. ./common/rc > > +. ./common/filter > > +. ./common/attr > > +. ./common/reflink > > + > > +# real QA test starts here > > +_supported_os Linux > > +_require_scratch_reflink > > +_require_cp_reflink > > +_require_test_program "punch-alternating" > > + > > +rm -f "$seqres.full" > > + > > +echo "Format and mount" > > +_scratch_mkfs > "$seqres.full" 2>&1 > > +_scratch_mount >> "$seqres.full" 2>&1 > > + > > +testdir="$SCRATCH_MNT/test-$seq" > > +mkdir "$testdir" > > + > > +# Setup for one million blocks, but we'll accept stress testing down to > > +# 2^17 blocks... that should be plenty for anyone. > > +fnr=20 > > +free_blocks=$(stat -f -c '%a' "$testdir") > > +blksz=$(_get_block_size "$testdir") > > +space_avail=$((free_blocks * blksz)) > > +calc_space() { > > + blocks_needed=$(( 2 ** (fnr + 1) )) > > + space_needed=$((blocks_needed * blksz * 5 / 4)) > > +} > > +calc_space > > +while test $space_needed -gt $space_avail; do > > + fnr=$((fnr - 1)) > > + calc_space > > +done > > +test $fnr -lt 17 && _notrun "Insufficient space for stress test; would only create $blocks_needed extents ($space_needed/$space_avail blocks)." > > + > > +echo "Create a many-block file" > > +echo "creating $blocks_needed blocks..." >> "$seqres.full" > > +$XFS_IO_PROG -f -c "pwrite -S 0x61 -b 4194304 0 $((2 ** (fnr + 1) * blksz))" "$testdir/file1" >> "$seqres.full" > > + > > +echo "Reflinking file" > > +_cp_reflink $testdir/file1 $testdir/file2 > > + > > +echo "Punch file2" > > +echo "Punching file2..." >> "$seqres.full" > > +"$here/src/punch-alternating" "$testdir/file2" >> "$seqres.full" > > +echo "...done" >> "$seqres.full" > > +_scratch_cycle_mount > > + > > +echo "Delete file1" > > +rm -rf $testdir/file1 > > + > > +# success, all done > > +status=0 > > +exit > > diff --git a/tests/generic/931.out b/tests/generic/931.out > > new file mode 100644 > > index 0000000..c7b724e > > --- /dev/null > > +++ b/tests/generic/931.out > > @@ -0,0 +1,6 @@ > > +QA output created by 931 > > +Format and mount > > +Create a many-block file > > +Reflinking file > > +Punch file2 > > +Delete file1 > > diff --git a/tests/generic/group b/tests/generic/group > > index ab1e9d3..b0d1844 100644 > > --- a/tests/generic/group > > +++ b/tests/generic/group > > @@ -443,3 +443,4 @@ > > 438 auto > > 439 auto quick punch > > 440 auto quick encrypt > > +931 auto quick clone > > > -- > 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
On Thu, Jun 29, 2017 at 09:07:46AM -0700, Darrick J. Wong wrote: > On Thu, Jun 29, 2017 at 05:36:14PM +0800, Eryu Guan wrote: > > On Wed, Jun 21, 2017 at 02:57:48PM -0700, Darrick J. Wong wrote: > > > From: Darrick J. Wong <darrick.wong@oracle.com> > > > > > > Test a regression in XFS where we blow out a transaction reservation if > > > we create a big file, share every other block, and delete the first > > > file. There's nothing particularly fs-specific about this stress test, > > > so put it in generic. > > > > > > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> > > > > This test took me 3019s to finish with v4.12-rc7 kernel, and another > > host "hung" at "Delete file1" (it had been more than 1 hour, and I lost > > my patience and hard-reboot the host). Are they expected results? > > No. The take-forever-or-crash behavior should be fixed by "xfs: try to > avoid blowing out the transaction reservation when bunmaping a shared > extent" in 4.13. Feel free to hang on to this one until -rc1. :) Thanks! I'll take it in after 4.13-rc1 then :) > > > If the bug is still existed in latest upstream kernel, I tend to merge > > it after the fix landing in linus tree. If v4.12-rc7 doesn't suffer from > > this bug, the test time should be reduced. > > <shrug> This is what I saw just now: > > FSTYP -- xfs (debug) > PLATFORM -- Linux/x86_64 birch-mtr0 4.12.0-rc6-dgc > MKFS_OPTIONS -- -f -m reflink=1,rmapbt=1, -i sparse=1, /dev/pmem1 > MOUNT_OPTIONS -- /dev/pmem1 /opt > > generic/931 21s > Ran: generic/931 > Passed all 1 tests > > Though if it takes forever for everyone else, please kick this one out > of auto/quick. At least in theory, before the patch the test will > either blow out a transaction reservation and hang the system, or if it > does succeed it'll have done so by scraping long and hard for log space. > That is probably why it takes 3000+ seconds on your test box, unless > you were also testing xfs for-next. I'm actually testing for-next branch now, will confirm the test time on for-next kernel. > > > And another minor nit below. > > > > > --- > > > tests/generic/931 | 94 +++++++++++++++++++++++++++++++++++++++++++++++++ > > > tests/generic/931.out | 6 +++ > > > tests/generic/group | 1 + > > > 3 files changed, 101 insertions(+) > > > create mode 100755 tests/generic/931 > > > create mode 100644 tests/generic/931.out > > > > > > > > > diff --git a/tests/generic/931 b/tests/generic/931 > > > new file mode 100755 > > > index 0000000..afadf81 > > > --- /dev/null > > > +++ b/tests/generic/931 > > > @@ -0,0 +1,94 @@ > > > +#! /bin/bash > > > +# FS QA Test No. 931 > > > +# > > > +# See how well we handle deleting a file with a million refcount extents. > > > +# > > > +#----------------------------------------------------------------------- > > > +# Copyright (c) 2017, Oracle and/or its affiliates. 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/$$ > > > +status=1 # failure is the default! > > > +trap "_cleanup; exit \$status" 0 1 2 3 15 > > > + > > > +_cleanup() > > > +{ > > > + cd / > > > + rm -rf "$tmp".* $testdir/file1 > > > > 'rm -rf' looks a bit scary, and we're only deleting regular files not > > directories, 'rm -f' should be sufficient. > > Yes. Will you fix it on the way in or should I resend? I can fix it. Thanks, Eryu -- To unsubscribe from this list: send the line "unsubscribe fstests" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/tests/generic/931 b/tests/generic/931 new file mode 100755 index 0000000..afadf81 --- /dev/null +++ b/tests/generic/931 @@ -0,0 +1,94 @@ +#! /bin/bash +# FS QA Test No. 931 +# +# See how well we handle deleting a file with a million refcount extents. +# +#----------------------------------------------------------------------- +# Copyright (c) 2017, Oracle and/or its affiliates. All Rights Reserved. +# +# This program is free software; you can redistribute it and/or +# modify it under the terms of the GNU General Public License as +# published by the Free Software Foundation. +# +# This program is distributed in the hope that it would be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program; if not, write the Free Software Foundation, +# Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA +#----------------------------------------------------------------------- + +seq=`basename "$0"` +seqres="$RESULT_DIR/$seq" +echo "QA output created by $seq" + +here=`pwd` +tmp=/tmp/$$ +status=1 # failure is the default! +trap "_cleanup; exit \$status" 0 1 2 3 15 + +_cleanup() +{ + cd / + rm -rf "$tmp".* $testdir/file1 +} + +# get standard environment, filters and checks +. ./common/rc +. ./common/filter +. ./common/attr +. ./common/reflink + +# real QA test starts here +_supported_os Linux +_require_scratch_reflink +_require_cp_reflink +_require_test_program "punch-alternating" + +rm -f "$seqres.full" + +echo "Format and mount" +_scratch_mkfs > "$seqres.full" 2>&1 +_scratch_mount >> "$seqres.full" 2>&1 + +testdir="$SCRATCH_MNT/test-$seq" +mkdir "$testdir" + +# Setup for one million blocks, but we'll accept stress testing down to +# 2^17 blocks... that should be plenty for anyone. +fnr=20 +free_blocks=$(stat -f -c '%a' "$testdir") +blksz=$(_get_block_size "$testdir") +space_avail=$((free_blocks * blksz)) +calc_space() { + blocks_needed=$(( 2 ** (fnr + 1) )) + space_needed=$((blocks_needed * blksz * 5 / 4)) +} +calc_space +while test $space_needed -gt $space_avail; do + fnr=$((fnr - 1)) + calc_space +done +test $fnr -lt 17 && _notrun "Insufficient space for stress test; would only create $blocks_needed extents ($space_needed/$space_avail blocks)." + +echo "Create a many-block file" +echo "creating $blocks_needed blocks..." >> "$seqres.full" +$XFS_IO_PROG -f -c "pwrite -S 0x61 -b 4194304 0 $((2 ** (fnr + 1) * blksz))" "$testdir/file1" >> "$seqres.full" + +echo "Reflinking file" +_cp_reflink $testdir/file1 $testdir/file2 + +echo "Punch file2" +echo "Punching file2..." >> "$seqres.full" +"$here/src/punch-alternating" "$testdir/file2" >> "$seqres.full" +echo "...done" >> "$seqres.full" +_scratch_cycle_mount + +echo "Delete file1" +rm -rf $testdir/file1 + +# success, all done +status=0 +exit diff --git a/tests/generic/931.out b/tests/generic/931.out new file mode 100644 index 0000000..c7b724e --- /dev/null +++ b/tests/generic/931.out @@ -0,0 +1,6 @@ +QA output created by 931 +Format and mount +Create a many-block file +Reflinking file +Punch file2 +Delete file1 diff --git a/tests/generic/group b/tests/generic/group index ab1e9d3..b0d1844 100644 --- a/tests/generic/group +++ b/tests/generic/group @@ -443,3 +443,4 @@ 438 auto 439 auto quick punch 440 auto quick encrypt +931 auto quick clone