[adding ext4 list]
On Wed, Nov 15, 2017 at 04:58:33PM +0800, Chao Yu wrote:
> f2fs can skip isize updating in fsync(), since during mount, f2fs tries
> to recovery isize according to valid block address or preallocated flag
> in last fsynced dnode block.
>
> However, fallocate() breaks our rule with setting FALLOC_FL_KEEP_SIZE
> flag, since it can preallocated block cross EOF, once the file is fsynced,
> in POR, we will recover isize incorrectly based on these fallocated
> blocks.
>
> This patch adds a new testcase to test fallocate, in order to verify
> whether filesystem will do correct recovery on isize.
>
> Signed-off-by: Chao Yu <[email protected]>
This test fails on ext4 as
==== falloc -k 1024 test with fdatasync ====
+Before: "b: 8216 s: 4202496"
+After : "b: 8208 s: 4202496"
==== falloc -k 4096 test with fdatasync ====
+Before: "b: 8216 s: 4202496"
+After : "b: 8208 s: 4202496"
==== falloc -k 104857600 test with fdatasync ====
+Before: "b: 213008 s: 4202496"
+After : "b: 8208 s: 4202496"
So the block counts are changed after fdatasync & fs shutdown in the
fallocate(2) KEEP_SIZE case. Looks like a real ext4 bug, fdatasync
failed to flush the preallocated blocks beyond EOF to disk, but I want
to confirm first with ext4 people. If it turns out to be a case issue,
we'd like to update test case accordingly before merging it.
> ---
> v2:
> - change copyright to Huawei Co.
> - refactor description of testcase.
> - set status correctly.
> tests/generic/468 | 127 ++++++++++++++++++++++++++++++++++++++++++++++++++
> tests/generic/468.out | 13 ++++++
> tests/generic/group | 1 +
> 3 files changed, 141 insertions(+)
> create mode 100755 tests/generic/468
> create mode 100644 tests/generic/468.out
>
> diff --git a/tests/generic/468 b/tests/generic/468
> new file mode 100755
> index 000000000000..242589a35e34
> --- /dev/null
> +++ b/tests/generic/468
> @@ -0,0 +1,127 @@
> +#! /bin/bash
> +# FS QA Test 468
> +#
> +# This testcase is a fallocate variant of generic/392, it expands to test
> +# block preallocation functionality of fallocate.
> +# In this case, we are trying to execute:
> +# 1. fallocate {,-k}
> +# 2. f{data,}sync
> +# 3. power-cuts
> +# 4. recovery filesystem during mount
> +# 5. check inode's metadata
> +#
> +# In the case of fsync, filesystem should recover all the inode metadata, while
> +# recovering i_blocks and i_size at least for fdatasync, so this testcase excepts
> +# that inode metadata will be unchanged after recovery.
> +#
> +#-----------------------------------------------------------------------
> +# Copyright (c) 2017 Huawei. All Rights Reserved.
> +# Author: Chao Yu <[email protected]>
> +#
> +# 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
> +
> +# remove previous $seqres.full before test
> +rm -f $seqres.full
> +
> +# real QA test starts here
> +
> +# Modify as appropriate.
> +_supported_fs generic
> +_supported_os Linux
> +_require_test
I'd removed this _require_test call, as test device is not used. I'll
fix it on commit if we don't need v3 (due to the ext4 failure).
> +
> +_require_scratch
> +_require_scratch_shutdown
> +_require_xfs_io_command "falloc" "-k"
> +
> +_scratch_mkfs >/dev/null 2>&1
> +_require_metadata_journaling $SCRATCH_DEV
> +_scratch_mount
> +
> +testfile=$SCRATCH_MNT/testfile
> +
> +# check inode metadata after shutdown
> +check_inode_metadata()
> +{
> + sync_mode=$1
> +
> + # fsync or fdatasync
> + if [ $sync_mode = "fsync" ]; then
> + stat_opt='-c "b: %b s: %s a: %x m: %y c: %z"'
> + else
> + stat_opt='-c "b: %b s: %s"'
> + fi
> +
> + before=`stat "$stat_opt" $testfile`
> +
> + $XFS_IO_PROG -c "$sync_mode" $testfile | _filter_xfs_io
> + src/godown $SCRATCH_MNT | tee -a $seqres.full
> + _scratch_cycle_mount
> +
> + after=`stat "$stat_opt" $testfile`
> +
> + if [ "$before" != "$after" ]; then
> + echo "Before: $before"
> + echo "After : $after"
> + status=1; # this is a failure!
> + fi
> + echo "Before: $before" >> $seqres.full
> + echo "After : $after" >> $seqres.full
> + rm $testfile
> +}
> +
> +# fallocate XX KB with f{data}sync, followed by power-cut
> +test_falloc()
> +{
> + echo "==== falloc $2$3 test with $1 ====" | tee -a $seqres.full
> + $XFS_IO_PROG -f -c "truncate 4202496" \
> + -c "pwrite 0 4202496" \
> + -c "fsync" \
> + -c "falloc $2 4202496 $3"\
> + $testfile >/dev/null
> + check_inode_metadata $1
> +}
> +
> +for i in fsync fdatasync; do
> + test_falloc $i "" 1024
> + test_falloc $i "" 4096
> + test_falloc $i "" 104857600
> + test_falloc $i "-k " 1024
> + test_falloc $i "-k " 4096
> + test_falloc $i "-k " 104857600
> +done
> +
> +status = 0
This is wrong in bash, should be "status=0", no spaces between "=" :)
I can fix it on commit too.
Thanks,
Eryu
> +exit
> diff --git a/tests/generic/468.out b/tests/generic/468.out
> new file mode 100644
> index 000000000000..b3a28d5ea07c
> --- /dev/null
> +++ b/tests/generic/468.out
> @@ -0,0 +1,13 @@
> +QA output created by 468
> +==== falloc 1024 test with fsync ====
> +==== falloc 4096 test with fsync ====
> +==== falloc 104857600 test with fsync ====
> +==== falloc -k 1024 test with fsync ====
> +==== falloc -k 4096 test with fsync ====
> +==== falloc -k 104857600 test with fsync ====
> +==== falloc 1024 test with fdatasync ====
> +==== falloc 4096 test with fdatasync ====
> +==== falloc 104857600 test with fdatasync ====
> +==== falloc -k 1024 test with fdatasync ====
> +==== falloc -k 4096 test with fdatasync ====
> +==== falloc -k 104857600 test with fdatasync ====
> diff --git a/tests/generic/group b/tests/generic/group
> index 9183950c8dfc..ea2dc04d931e 100644
> --- a/tests/generic/group
> +++ b/tests/generic/group
> @@ -470,3 +470,4 @@
> 465 auto rw quick aio
> 466 auto quick rw
> 467 auto quick exportfs
> +468 shutdown auto quick metadata
> --
> 2.15.0.55.gc2ece9dc4de6
>
On Wed, Nov 15, 2017 at 07:37:28PM +0800, Eryu Guan wrote:
> [adding ext4 list]
>
> On Wed, Nov 15, 2017 at 04:58:33PM +0800, Chao Yu wrote:
> > f2fs can skip isize updating in fsync(), since during mount, f2fs tries
> > to recovery isize according to valid block address or preallocated flag
> > in last fsynced dnode block.
> >
> > However, fallocate() breaks our rule with setting FALLOC_FL_KEEP_SIZE
> > flag, since it can preallocated block cross EOF, once the file is fsynced,
> > in POR, we will recover isize incorrectly based on these fallocated
> > blocks.
> >
> > This patch adds a new testcase to test fallocate, in order to verify
> > whether filesystem will do correct recovery on isize.
> >
> > Signed-off-by: Chao Yu <[email protected]>
>
> This test fails on ext4 as
>
> ==== falloc -k 1024 test with fdatasync ====
> +Before: "b: 8216 s: 4202496"
> +After : "b: 8208 s: 4202496"
> ==== falloc -k 4096 test with fdatasync ====
> +Before: "b: 8216 s: 4202496"
> +After : "b: 8208 s: 4202496"
> ==== falloc -k 104857600 test with fdatasync ====
> +Before: "b: 213008 s: 4202496"
> +After : "b: 8208 s: 4202496"
>
> So the block counts are changed after fdatasync & fs shutdown in the
> fallocate(2) KEEP_SIZE case. Looks like a real ext4 bug, fdatasync
> failed to flush the preallocated blocks beyond EOF to disk, but I want
> to confirm first with ext4 people. If it turns out to be a case issue,
> we'd like to update test case accordingly before merging it.
I think I've found the problem, ext4 fallocate(2) path is missing a
ext4_update_inode_fsync_trans(handle, inode, 1) call, as what commit
("67a7d5f561f4 ext4: fix fdatasync(2) after extent manipulation
operations") did.
I'll send an ext4 patch soon.
Thanks,
Eryu
On 2017/11/15 19:37, Eryu Guan wrote:
> [adding ext4 list]
>
> On Wed, Nov 15, 2017 at 04:58:33PM +0800, Chao Yu wrote:
>> f2fs can skip isize updating in fsync(), since during mount, f2fs tries
>> to recovery isize according to valid block address or preallocated flag
>> in last fsynced dnode block.
>>
>> However, fallocate() breaks our rule with setting FALLOC_FL_KEEP_SIZE
>> flag, since it can preallocated block cross EOF, once the file is fsynced,
>> in POR, we will recover isize incorrectly based on these fallocated
>> blocks.
>>
>> This patch adds a new testcase to test fallocate, in order to verify
>> whether filesystem will do correct recovery on isize.
>>
>> Signed-off-by: Chao Yu <[email protected]>
>
> This test fails on ext4 as
>
> ==== falloc -k 1024 test with fdatasync ====
> +Before: "b: 8216 s: 4202496"
> +After : "b: 8208 s: 4202496"
> ==== falloc -k 4096 test with fdatasync ====
> +Before: "b: 8216 s: 4202496"
> +After : "b: 8208 s: 4202496"
> ==== falloc -k 104857600 test with fdatasync ====
> +Before: "b: 213008 s: 4202496"
> +After : "b: 8208 s: 4202496"
>
> So the block counts are changed after fdatasync & fs shutdown in the
> fallocate(2) KEEP_SIZE case. Looks like a real ext4 bug, fdatasync
> failed to flush the preallocated blocks beyond EOF to disk, but I want
> to confirm first with ext4 people. If it turns out to be a case issue,
> we'd like to update test case accordingly before merging it.
>
>> ---
>> v2:
>> - change copyright to Huawei Co.
>> - refactor description of testcase.
>> - set status correctly.
>> tests/generic/468 | 127 ++++++++++++++++++++++++++++++++++++++++++++++++++
>> tests/generic/468.out | 13 ++++++
>> tests/generic/group | 1 +
>> 3 files changed, 141 insertions(+)
>> create mode 100755 tests/generic/468
>> create mode 100644 tests/generic/468.out
>>
>> diff --git a/tests/generic/468 b/tests/generic/468
>> new file mode 100755
>> index 000000000000..242589a35e34
>> --- /dev/null
>> +++ b/tests/generic/468
>> @@ -0,0 +1,127 @@
>> +#! /bin/bash
>> +# FS QA Test 468
>> +#
>> +# This testcase is a fallocate variant of generic/392, it expands to test
>> +# block preallocation functionality of fallocate.
>> +# In this case, we are trying to execute:
>> +# 1. fallocate {,-k}
>> +# 2. f{data,}sync
>> +# 3. power-cuts
>> +# 4. recovery filesystem during mount
>> +# 5. check inode's metadata
>> +#
>> +# In the case of fsync, filesystem should recover all the inode metadata, while
>> +# recovering i_blocks and i_size at least for fdatasync, so this testcase excepts
>> +# that inode metadata will be unchanged after recovery.
>> +#
>> +#-----------------------------------------------------------------------
>> +# Copyright (c) 2017 Huawei. All Rights Reserved.
>> +# Author: Chao Yu <[email protected]>
>> +#
>> +# 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
>> +
>> +# remove previous $seqres.full before test
>> +rm -f $seqres.full
>> +
>> +# real QA test starts here
>> +
>> +# Modify as appropriate.
>> +_supported_fs generic
>> +_supported_os Linux
>> +_require_test
>
> I'd removed this _require_test call, as test device is not used. I'll
> fix it on commit if we don't need v3 (due to the ext4 failure).
Oh, that's right, let me know if we need v3. ;)
>
>> +
>> +_require_scratch
>> +_require_scratch_shutdown
>> +_require_xfs_io_command "falloc" "-k"
>> +
>> +_scratch_mkfs >/dev/null 2>&1
>> +_require_metadata_journaling $SCRATCH_DEV
>> +_scratch_mount
>> +
>> +testfile=$SCRATCH_MNT/testfile
>> +
>> +# check inode metadata after shutdown
>> +check_inode_metadata()
>> +{
>> + sync_mode=$1
>> +
>> + # fsync or fdatasync
>> + if [ $sync_mode = "fsync" ]; then
>> + stat_opt='-c "b: %b s: %s a: %x m: %y c: %z"'
>> + else
>> + stat_opt='-c "b: %b s: %s"'
>> + fi
>> +
>> + before=`stat "$stat_opt" $testfile`
>> +
>> + $XFS_IO_PROG -c "$sync_mode" $testfile | _filter_xfs_io
>> + src/godown $SCRATCH_MNT | tee -a $seqres.full
>> + _scratch_cycle_mount
>> +
>> + after=`stat "$stat_opt" $testfile`
>> +
>> + if [ "$before" != "$after" ]; then
>> + echo "Before: $before"
>> + echo "After : $after"
>> + status=1; # this is a failure!
>> + fi
>> + echo "Before: $before" >> $seqres.full
>> + echo "After : $after" >> $seqres.full
>> + rm $testfile
>> +}
>> +
>> +# fallocate XX KB with f{data}sync, followed by power-cut
>> +test_falloc()
>> +{
>> + echo "==== falloc $2$3 test with $1 ====" | tee -a $seqres.full
>> + $XFS_IO_PROG -f -c "truncate 4202496" \
>> + -c "pwrite 0 4202496" \
>> + -c "fsync" \
>> + -c "falloc $2 4202496 $3"\
>> + $testfile >/dev/null
>> + check_inode_metadata $1
>> +}
>> +
>> +for i in fsync fdatasync; do
>> + test_falloc $i "" 1024
>> + test_falloc $i "" 4096
>> + test_falloc $i "" 104857600
>> + test_falloc $i "-k " 1024
>> + test_falloc $i "-k " 4096
>> + test_falloc $i "-k " 104857600
>> +done
>> +
>> +status = 0
>
> This is wrong in bash, should be "status=0", no spaces between "=" :)
> I can fix it on commit too.
Thanks for correcting my coding style issue.
Thanks,
>
> Thanks,
> Eryu
>
>> +exit
>> diff --git a/tests/generic/468.out b/tests/generic/468.out
>> new file mode 100644
>> index 000000000000..b3a28d5ea07c
>> --- /dev/null
>> +++ b/tests/generic/468.out
>> @@ -0,0 +1,13 @@
>> +QA output created by 468
>> +==== falloc 1024 test with fsync ====
>> +==== falloc 4096 test with fsync ====
>> +==== falloc 104857600 test with fsync ====
>> +==== falloc -k 1024 test with fsync ====
>> +==== falloc -k 4096 test with fsync ====
>> +==== falloc -k 104857600 test with fsync ====
>> +==== falloc 1024 test with fdatasync ====
>> +==== falloc 4096 test with fdatasync ====
>> +==== falloc 104857600 test with fdatasync ====
>> +==== falloc -k 1024 test with fdatasync ====
>> +==== falloc -k 4096 test with fdatasync ====
>> +==== falloc -k 104857600 test with fdatasync ====
>> diff --git a/tests/generic/group b/tests/generic/group
>> index 9183950c8dfc..ea2dc04d931e 100644
>> --- a/tests/generic/group
>> +++ b/tests/generic/group
>> @@ -470,3 +470,4 @@
>> 465 auto rw quick aio
>> 466 auto quick rw
>> 467 auto quick exportfs
>> +468 shutdown auto quick metadata
>> --
>> 2.15.0.55.gc2ece9dc4de6
>>