diff mbox series

[f2fs-dev] f2fs/005: add testcase to check checkpoint disabling functionality

Message ID 20241012101501.377899-1-chao@kernel.org (mailing list archive)
State Superseded
Headers show
Series [f2fs-dev] f2fs/005: add testcase to check checkpoint disabling functionality | expand

Commit Message

Chao Yu Oct. 12, 2024, 10:15 a.m. UTC 
This patch introduce a regression test to check whether f2fs handles
dirty inode correctly when checkpoint is disabled in a corner case,
it may hang umount before the bug is fixed.

Cc: Qi Han <hanqi@vivo.com>
Signed-off-by: Chao Yu <chao@kernel.org>
---
 tests/f2fs/005     | 59 ++++++++++++++++++++++++++++++++++++++++++++++
 tests/f2fs/005.out |  1 +
 2 files changed, 60 insertions(+)
 create mode 100755 tests/f2fs/005
 create mode 100644 tests/f2fs/005.out

Comments

Zorro Lang Oct. 25, 2024, 4:11 a.m. UTC | #1 
On Sat, Oct 12, 2024 at 06:15:01PM +0800, Chao Yu wrote:
> This patch introduce a regression test to check whether f2fs handles
> dirty inode correctly when checkpoint is disabled in a corner case,
> it may hang umount before the bug is fixed.
> 
> Cc: Qi Han <hanqi@vivo.com>
> Signed-off-by: Chao Yu <chao@kernel.org>
> ---
>  tests/f2fs/005     | 59 ++++++++++++++++++++++++++++++++++++++++++++++
>  tests/f2fs/005.out |  1 +
>  2 files changed, 60 insertions(+)
>  create mode 100755 tests/f2fs/005
>  create mode 100644 tests/f2fs/005.out
> 
> diff --git a/tests/f2fs/005 b/tests/f2fs/005
> new file mode 100755
> index 00000000..4faf1bb9
> --- /dev/null
> +++ b/tests/f2fs/005
> @@ -0,0 +1,59 @@
> +#! /bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +# Copyright (c) 2024 Oppo.  All Rights Reserved.
> +#
> +# FS QA Test No. f2fs/005
> +#
> +# This is a regression test to check whether f2fs handles dirty
> +# inode correctly when checkpoint is disabled, it may hang umount
> +# before the bug is fixed.

Is this a duplicated test of another patch from you?

https://lore.kernel.org/fstests/20241025034413.5s6ecilfgqcoeudt@dell-per750-06-vm-08.rhts.eng.pek2.redhat.com/T/#t

Or just similar?

If this's a different test ...(see below)

> +#
> +. ./common/preamble
> +_begin_fstest auto quick
> +
> +_cleanup()
> +{
> +	rm -f $img
> +	rmdir $mnt
> +	_scratch_unmount >> $seqres.full
> +	cd /
> +	rm -r -f $tmp.*
> +}
> +
> +_require_scratch
> +_scratch_mkfs >> $seqres.full
> +_scratch_mount >> $seqres.full
> +
> +img=$SCRATCH_MNT/f2fs.img
> +mnt=$SCRATCH_MNT/f2fs.mnt
> +testfile=$mnt/testfile
> +tmpfile=$mnt/tmpfile
> +tmpdir=$mnt/tmpdir
> +
> +mkdir $mnt
> +dd if=/dev/zero of=$img bs=1M count=50 2>/dev/null
> +$MKFS_F2FS_PROG -f $img >/dev/null 2>&1
> +sync
> +
> +# use mode=lfs to let f2fs always triggers OPU
> +mount -t $FSTYP -o loop,mode=lfs,checkpoint=disable:10%,noinline_dentry $img $mnt

You can use _scratch_mkfs_sized, don't need a loop device.
And without the loop device, above _cleanup is not needed either.

> +
> +dd if=/dev/zero of=$testfile bs=1M count=5 2>/dev/null
> +mkdir $tmpdir
> +touch $tmpfile
> +sync
> +
> +# it dirties tmpdir inode by updating ctime,
> +# but it doesn't moving the inode to gdirty_list.
> +mv $tmpfile $tmpdir
> +
> +# it runs out of free segment
> +dd if=/dev/zero of=$testfile bs=1M count=5 conv=notrunc conv=fsync 2>/dev/null
> +
> +mount -o remount,checkpoint=enable $mnt
> +
> +# it may hang umount if tmpdir is still dirty during evict()
> +umount $mnt
> +
> +status=0
> +exit
> diff --git a/tests/f2fs/005.out b/tests/f2fs/005.out
> new file mode 100644
> index 00000000..caa3c880
> --- /dev/null
> +++ b/tests/f2fs/005.out
> @@ -0,0 +1 @@
> +QA output created by 005

If nothing output, please:

echo "Silence is golden"

Thanks,
Zorro

> -- 
> 2.40.1
>
Chao Yu Oct. 28, 2024, 1:46 p.m. UTC | #2 
On 2024/10/25 12:11, Zorro Lang wrote:
> On Sat, Oct 12, 2024 at 06:15:01PM +0800, Chao Yu wrote:
>> This patch introduce a regression test to check whether f2fs handles
>> dirty inode correctly when checkpoint is disabled in a corner case,
>> it may hang umount before the bug is fixed.
>>
>> Cc: Qi Han <hanqi@vivo.com>
>> Signed-off-by: Chao Yu <chao@kernel.org>
>> ---
>>   tests/f2fs/005     | 59 ++++++++++++++++++++++++++++++++++++++++++++++
>>   tests/f2fs/005.out |  1 +
>>   2 files changed, 60 insertions(+)
>>   create mode 100755 tests/f2fs/005
>>   create mode 100644 tests/f2fs/005.out
>>
>> diff --git a/tests/f2fs/005 b/tests/f2fs/005
>> new file mode 100755
>> index 00000000..4faf1bb9
>> --- /dev/null
>> +++ b/tests/f2fs/005
>> @@ -0,0 +1,59 @@
>> +#! /bin/bash
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Copyright (c) 2024 Oppo.  All Rights Reserved.
>> +#
>> +# FS QA Test No. f2fs/005
>> +#
>> +# This is a regression test to check whether f2fs handles dirty
>> +# inode correctly when checkpoint is disabled, it may hang umount
>> +# before the bug is fixed.
> 
> Is this a duplicated test of another patch from you?
> 
> https://lore.kernel.org/fstests/20241025034413.5s6ecilfgqcoeudt@dell-per750-06-vm-08.rhts.eng.pek2.redhat.com/T/#t
> 
> Or just similar?
> 
> If this's a different test ...(see below)

It's a different one.

> 
>> +#
>> +. ./common/preamble
>> +_begin_fstest auto quick
>> +
>> +_cleanup()
>> +{
>> +	rm -f $img
>> +	rmdir $mnt
>> +	_scratch_unmount >> $seqres.full
>> +	cd /
>> +	rm -r -f $tmp.*
>> +}
>> +
>> +_require_scratch
>> +_scratch_mkfs >> $seqres.full
>> +_scratch_mount >> $seqres.full
>> +
>> +img=$SCRATCH_MNT/f2fs.img
>> +mnt=$SCRATCH_MNT/f2fs.mnt
>> +testfile=$mnt/testfile
>> +tmpfile=$mnt/tmpfile
>> +tmpdir=$mnt/tmpdir
>> +
>> +mkdir $mnt
>> +dd if=/dev/zero of=$img bs=1M count=50 2>/dev/null
>> +$MKFS_F2FS_PROG -f $img >/dev/null 2>&1
>> +sync
>> +
>> +# use mode=lfs to let f2fs always triggers OPU
>> +mount -t $FSTYP -o loop,mode=lfs,checkpoint=disable:10%,noinline_dentry $img $mnt
> 
> You can use _scratch_mkfs_sized, don't need a loop device.
> And without the loop device, above _cleanup is not needed either.
> 
>> +
>> +dd if=/dev/zero of=$testfile bs=1M count=5 2>/dev/null
>> +mkdir $tmpdir
>> +touch $tmpfile
>> +sync
>> +
>> +# it dirties tmpdir inode by updating ctime,
>> +# but it doesn't moving the inode to gdirty_list.
>> +mv $tmpfile $tmpdir
>> +
>> +# it runs out of free segment
>> +dd if=/dev/zero of=$testfile bs=1M count=5 conv=notrunc conv=fsync 2>/dev/null
>> +
>> +mount -o remount,checkpoint=enable $mnt
>> +
>> +# it may hang umount if tmpdir is still dirty during evict()
>> +umount $mnt
>> +
>> +status=0
>> +exit
>> diff --git a/tests/f2fs/005.out b/tests/f2fs/005.out
>> new file mode 100644
>> index 00000000..caa3c880
>> --- /dev/null
>> +++ b/tests/f2fs/005.out
>> @@ -0,0 +1 @@
>> +QA output created by 005
> 
> If nothing output, please:
> 
> echo "Silence is golden"

Will update this patch according to your comments, thank you!

Thanks,

> 
> Thanks,
> Zorro
> 
>> -- 
>> 2.40.1
>>
>
diff mbox series

Patch

diff --git a/tests/f2fs/005 b/tests/f2fs/005
new file mode 100755
index 00000000..4faf1bb9
--- /dev/null
+++ b/tests/f2fs/005
@@ -0,0 +1,59 @@ 
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2024 Oppo.  All Rights Reserved.
+#
+# FS QA Test No. f2fs/005
+#
+# This is a regression test to check whether f2fs handles dirty
+# inode correctly when checkpoint is disabled, it may hang umount
+# before the bug is fixed.
+#
+. ./common/preamble
+_begin_fstest auto quick
+
+_cleanup()
+{
+	rm -f $img
+	rmdir $mnt
+	_scratch_unmount >> $seqres.full
+	cd /
+	rm -r -f $tmp.*
+}
+
+_require_scratch
+_scratch_mkfs >> $seqres.full
+_scratch_mount >> $seqres.full
+
+img=$SCRATCH_MNT/f2fs.img
+mnt=$SCRATCH_MNT/f2fs.mnt
+testfile=$mnt/testfile
+tmpfile=$mnt/tmpfile
+tmpdir=$mnt/tmpdir
+
+mkdir $mnt
+dd if=/dev/zero of=$img bs=1M count=50 2>/dev/null
+$MKFS_F2FS_PROG -f $img >/dev/null 2>&1
+sync
+
+# use mode=lfs to let f2fs always triggers OPU
+mount -t $FSTYP -o loop,mode=lfs,checkpoint=disable:10%,noinline_dentry $img $mnt
+
+dd if=/dev/zero of=$testfile bs=1M count=5 2>/dev/null
+mkdir $tmpdir
+touch $tmpfile
+sync
+
+# it dirties tmpdir inode by updating ctime,
+# but it doesn't moving the inode to gdirty_list.
+mv $tmpfile $tmpdir
+
+# it runs out of free segment
+dd if=/dev/zero of=$testfile bs=1M count=5 conv=notrunc conv=fsync 2>/dev/null
+
+mount -o remount,checkpoint=enable $mnt
+
+# it may hang umount if tmpdir is still dirty during evict()
+umount $mnt
+
+status=0
+exit
diff --git a/tests/f2fs/005.out b/tests/f2fs/005.out
new file mode 100644
index 00000000..caa3c880
--- /dev/null
+++ b/tests/f2fs/005.out
@@ -0,0 +1 @@ 
+QA output created by 005