Re: [PATCH] generic/371: run write(2) and fallocate(2) in parallel

[Wang Xiaoguang]

hello,

On 07/21/2016 06:41 PM, Eryu Guan wrote:

On Thu, Jul 21, 2016 at 03:30:25PM +0800, Wang Xiaoguang wrote:

Currently in btrfs, there is something wrong with fallocate(2)'s data
space reservation, it'll temporarily occupy more data space thant it
really needs, which in turn will impact other operations' data request.

In this test case, it runs write(2) and fallocate(2) in parallel and the
total needed data space for these two operations don't exceed whole fs
free data space, to see whether we will get any unexpected ENOSPC error.

Signed-off-by: Wang Xiaoguang 
---
  tests/generic/371 | 119 ++
  tests/generic/371.out |   2 +
  tests/generic/group   |   1 +
  3 files changed, 122 insertions(+)
  create mode 100755 tests/generic/371
  create mode 100644 tests/generic/371.out

diff --git a/tests/generic/371 b/tests/generic/371
new file mode 100755
index 000..b85327a
--- /dev/null
+++ b/tests/generic/371
@@ -0,0 +1,119 @@
+#! /bin/bash
+# FS QA Test 371
+#
+# Run write(2) and fallocate(2) in parallel and the total needed data space
+# for these operations don't exceed whole fs free data space, to see whether
+# we will get any unexpected ENOSPC error.
+#
+#---
+# Copyright (c) 2016 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 /
+   rm -f $tmp.*
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+
+# remove previous $seqres.full before test
+rm -f $seqres.full
+
+# Modify as appropriate.
+_supported_fs generic
+_supported_os Linux
+_require_scratch

Need '_require_xfs_io_command "falloc"', otherwise test fails on ext3/2,
because they don't support fallocate(2).

OK, I see.




+
+_scratch_mkfs_sized $((256 * 1024 * 1024)) >> $seqres.full 2>&1
+_scratch_mount
+
+testfile1=$SCRATCH_MNT/testfile1
+testfile2=$SCRATCH_MNT/testfile2
+
+write_work()
+{
+   rm -f $testfile1
+   while [ 1 ]; do
+   $XFS_IO_PROG -f -c "pwrite 0 80M" $testfile1 >>$seqres.full 2>&1
+   grep "No space left on device" $seqres.full >/dev/null
+   if [ $? -eq 0 ]; then
+   echo "unexpected ENOSPC error occurs"
+   exit 1
+   fi

No need to grep for error message, just redirect stdout to /dev/null and
any error will appear in stderr, which will break golden image.

And $seqres.full is where we dump logs for debug purpose, tests should
dump output they need to somewhere like $tmp.

I see.




+   rm -f $testfile1
+   done
+}
+
+fallocate_work()
+{
+   rm -f  $testfile2
+   while [ 1 ]; do
+   $XFS_IO_PROG -f -c "falloc 0 80M" $testfile2 >>$seqres.full 2>&1
+   grep "No space left on device" $seqres.full >/dev/null
+   if [ $? -eq 0 ]; then
+   echo "unexpected ENOSPC error occurs"
+   exit 1
+   fi
+   rm -f $testfile2
+   done
+}
+
+run_time=$((180 * $TIME_FACTOR))

180s is too long time, I can reproduce it in around 10s on my test vm,
just loop for 100 times for each operation (pwrite and falloc)


+write_work &
+write_work_pid=$!
+fallocate_work &
+fallocate_work_pid=$!
+
+for ((elapsed_time = 0; elapsed_time < run_time; elapsed_time += 5)); do
+   kill -0 $write_work_pid >/dev/null 2>&1
+   if [ $? -ne 0 ]; then
+   kill $fallocate_work_pid >/dev/null 2>&1
+   break
+   fi
+
+   kill -0 $fallocate_work_pid >/dev/null 2>&1
+   if [ $? -ne 0 ]; then
+   kill $write_work_pid >/dev/null 2>&1
+   break
+   fi
+   sleep 5
+done
+
+kill $fallocate_work_pid >/dev/null 2>&1
+kill $write_work_pid >/dev/null 2>&1
+wait
+
+# wait un-finished xfs_io
+while ps aux | grep "xfs_io" | grep -qv grep; do
+   sleep 1
+done

And this seems unnecessarily complicated

So I'd write it as:

echo "Silence is golden"
for ((i=0; i<100; i++)); do
 $XFS_IO_PROG -fc 

Re: [PATCH] generic/371: run write(2) and fallocate(2) in parallel

[Dave Chinner]

On Thu, Jul 21, 2016 at 06:41:00PM +0800, Eryu Guan wrote:
> On Thu, Jul 21, 2016 at 03:30:25PM +0800, Wang Xiaoguang wrote:
> > +
> > +run_time=$((180 * $TIME_FACTOR))
> 
> 180s is too long time, I can reproduce it in around 10s on my test vm,
> just loop for 100 times for each operation (pwrite and falloc)

Or simply do:

run_time=$((20 * $TIME_FACTOR * $TIME_FACTOR))

So that the time factor scales up quickly when you want to run long
running tests.

> > --- a/tests/generic/group
> > +++ b/tests/generic/group
> > @@ -373,3 +373,4 @@
> >  368 auto quick richacl
> >  369 auto quick richacl
> >  370 auto quick richacl
> > +371 auto enospc prealloc stress
> 
> So we can add 'quick' group and remove 'stress'.

I'd leave the test in stress if we can run it based on time as per
my above suggestion.

Cheers,

Dave.
-- 
Dave Chinner
da...@fromorbit.com
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] generic/371: run write(2) and fallocate(2) in parallel

[Eryu Guan]

On Thu, Jul 21, 2016 at 03:30:25PM +0800, Wang Xiaoguang wrote:
> Currently in btrfs, there is something wrong with fallocate(2)'s data
> space reservation, it'll temporarily occupy more data space thant it
> really needs, which in turn will impact other operations' data request.
> 
> In this test case, it runs write(2) and fallocate(2) in parallel and the
> total needed data space for these two operations don't exceed whole fs
> free data space, to see whether we will get any unexpected ENOSPC error.
> 
> Signed-off-by: Wang Xiaoguang 
> ---
>  tests/generic/371 | 119 
> ++
>  tests/generic/371.out |   2 +
>  tests/generic/group   |   1 +
>  3 files changed, 122 insertions(+)
>  create mode 100755 tests/generic/371
>  create mode 100644 tests/generic/371.out
> 
> diff --git a/tests/generic/371 b/tests/generic/371
> new file mode 100755
> index 000..b85327a
> --- /dev/null
> +++ b/tests/generic/371
> @@ -0,0 +1,119 @@
> +#! /bin/bash
> +# FS QA Test 371
> +#
> +# Run write(2) and fallocate(2) in parallel and the total needed data space
> +# for these operations don't exceed whole fs free data space, to see whether
> +# we will get any unexpected ENOSPC error.
> +#
> +#---
> +# Copyright (c) 2016 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 /
> + rm -f $tmp.*
> +}
> +
> +# get standard environment, filters and checks
> +. ./common/rc
> +. ./common/filter
> +
> +# remove previous $seqres.full before test
> +rm -f $seqres.full
> +
> +# Modify as appropriate.
> +_supported_fs generic
> +_supported_os Linux
> +_require_scratch

Need '_require_xfs_io_command "falloc"', otherwise test fails on ext3/2,
because they don't support fallocate(2).

> +
> +_scratch_mkfs_sized $((256 * 1024 * 1024)) >> $seqres.full 2>&1
> +_scratch_mount
> +
> +testfile1=$SCRATCH_MNT/testfile1
> +testfile2=$SCRATCH_MNT/testfile2
> +
> +write_work()
> +{
> + rm -f $testfile1
> + while [ 1 ]; do
> + $XFS_IO_PROG -f -c "pwrite 0 80M" $testfile1 >>$seqres.full 2>&1
> + grep "No space left on device" $seqres.full >/dev/null
> + if [ $? -eq 0 ]; then
> + echo "unexpected ENOSPC error occurs"
> + exit 1
> + fi

No need to grep for error message, just redirect stdout to /dev/null and
any error will appear in stderr, which will break golden image.

And $seqres.full is where we dump logs for debug purpose, tests should
dump output they need to somewhere like $tmp.

> + rm -f $testfile1
> + done
> +}
> +
> +fallocate_work()
> +{
> + rm -f  $testfile2
> + while [ 1 ]; do
> + $XFS_IO_PROG -f -c "falloc 0 80M" $testfile2 >>$seqres.full 2>&1
> + grep "No space left on device" $seqres.full >/dev/null
> + if [ $? -eq 0 ]; then
> + echo "unexpected ENOSPC error occurs"
> + exit 1
> + fi
> + rm -f $testfile2
> + done
> +}
> +
> +run_time=$((180 * $TIME_FACTOR))

180s is too long time, I can reproduce it in around 10s on my test vm,
just loop for 100 times for each operation (pwrite and falloc)

> +write_work &
> +write_work_pid=$!
> +fallocate_work &
> +fallocate_work_pid=$!
> +
> +for ((elapsed_time = 0; elapsed_time < run_time; elapsed_time += 5)); do
> + kill -0 $write_work_pid >/dev/null 2>&1
> + if [ $? -ne 0 ]; then
> + kill $fallocate_work_pid >/dev/null 2>&1
> + break
> + fi
> +
> + kill -0 $fallocate_work_pid >/dev/null 2>&1
> + if [ $? -ne 0 ]; then
> + kill $write_work_pid >/dev/null 2>&1
> + break
> + fi
> + sleep 5
> +done
> +
> +kill $fallocate_work_pid >/dev/null 2>&1
> +kill $write_work_pid >/dev/null 2>&1
> +wait
> +
> +# wait un-finished xfs_io
> +while ps aux | grep "xfs_io" | grep -qv grep; do
> + sleep 1
> +done

And this seems 

[PATCH] generic/371: run write(2) and fallocate(2) in parallel

[Wang Xiaoguang]

Currently in btrfs, there is something wrong with fallocate(2)'s data
space reservation, it'll temporarily occupy more data space thant it
really needs, which in turn will impact other operations' data request.

In this test case, it runs write(2) and fallocate(2) in parallel and the
total needed data space for these two operations don't exceed whole fs
free data space, to see whether we will get any unexpected ENOSPC error.

Signed-off-by: Wang Xiaoguang 
---
 tests/generic/371 | 119 ++
 tests/generic/371.out |   2 +
 tests/generic/group   |   1 +
 3 files changed, 122 insertions(+)
 create mode 100755 tests/generic/371
 create mode 100644 tests/generic/371.out

diff --git a/tests/generic/371 b/tests/generic/371
new file mode 100755
index 000..b85327a
--- /dev/null
+++ b/tests/generic/371
@@ -0,0 +1,119 @@
+#! /bin/bash
+# FS QA Test 371
+#
+# Run write(2) and fallocate(2) in parallel and the total needed data space
+# for these operations don't exceed whole fs free data space, to see whether
+# we will get any unexpected ENOSPC error.
+#
+#---
+# Copyright (c) 2016 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 /
+   rm -f $tmp.*
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+
+# remove previous $seqres.full before test
+rm -f $seqres.full
+
+# Modify as appropriate.
+_supported_fs generic
+_supported_os Linux
+_require_scratch
+
+_scratch_mkfs_sized $((256 * 1024 * 1024)) >> $seqres.full 2>&1
+_scratch_mount
+
+testfile1=$SCRATCH_MNT/testfile1
+testfile2=$SCRATCH_MNT/testfile2
+
+write_work()
+{
+   rm -f $testfile1
+   while [ 1 ]; do
+   $XFS_IO_PROG -f -c "pwrite 0 80M" $testfile1 >>$seqres.full 2>&1
+   grep "No space left on device" $seqres.full >/dev/null
+   if [ $? -eq 0 ]; then
+   echo "unexpected ENOSPC error occurs"
+   exit 1
+   fi
+   rm -f $testfile1
+   done
+}
+
+fallocate_work()
+{
+   rm -f  $testfile2
+   while [ 1 ]; do
+   $XFS_IO_PROG -f -c "falloc 0 80M" $testfile2 >>$seqres.full 2>&1
+   grep "No space left on device" $seqres.full >/dev/null
+   if [ $? -eq 0 ]; then
+   echo "unexpected ENOSPC error occurs"
+   exit 1
+   fi
+   rm -f $testfile2
+   done
+}
+
+run_time=$((180 * $TIME_FACTOR))
+write_work &
+write_work_pid=$!
+fallocate_work &
+fallocate_work_pid=$!
+
+for ((elapsed_time = 0; elapsed_time < run_time; elapsed_time += 5)); do
+   kill -0 $write_work_pid >/dev/null 2>&1
+   if [ $? -ne 0 ]; then
+   kill $fallocate_work_pid >/dev/null 2>&1
+   break
+   fi
+
+   kill -0 $fallocate_work_pid >/dev/null 2>&1
+   if [ $? -ne 0 ]; then
+   kill $write_work_pid >/dev/null 2>&1
+   break
+   fi
+   sleep 5
+done
+
+kill $fallocate_work_pid >/dev/null 2>&1
+kill $write_work_pid >/dev/null 2>&1
+wait
+
+# wait un-finished xfs_io
+while ps aux | grep "xfs_io" | grep -qv grep; do
+   sleep 1
+done
+
+echo "Silence is golden"
+status=0
+exit
diff --git a/tests/generic/371.out b/tests/generic/371.out
new file mode 100644
index 000..22ec8a2
--- /dev/null
+++ b/tests/generic/371.out
@@ -0,0 +1,2 @@
+QA output created by 371
+Silence is golden
diff --git a/tests/generic/group b/tests/generic/group
index 97ecb65..3d4a802 100644
--- a/tests/generic/group
+++ b/tests/generic/group
@@ -373,3 +373,4 @@
 368 auto quick richacl
 369 auto quick richacl
 370 auto quick richacl
+371 auto enospc prealloc stress
-- 
2.9.0



--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html