Question: Why tests/generic/482 and tests/generic/487 can't pass both at the same time.

Question: Why tests/generic/482 and tests/generic/487 can't pass both at the same time.

[蔡晓斌]

Hi, I have a question when I ran the xfstests, please help me to solve this question, tks.


In my environment, I ues the machine with architecture of aarch64, and then setting the env in local.config file like below:




# Ideally define at least these 4 to match your environment
# The first 2 are required.
# See README for other variables which can be set.
#
# Note: SCRATCH_DEV >will< get overwritten!


export TEST_DEV=/dev/mmcblk0p12
export TEST_DIR=/mnt/xfstests/dir
export FSTYP=ext4
export SCRATCH_DEV=/dev/mmcblk0p13
export SCRATCH_MNT=/mnt/xfstests/dir-mnt
export SCRATCH_LOGDEV=/dev/mmcblk0p14
export USE_EXTERNAL=yes
export DUMP_CORRUPT_FS=1
export KEEP_DMESG=yes
export USE_KMEMLEK=yes
export LOGWRITES_DEV=/dev/mmcblk0p15
export MIN_FSSIZE=10485760 # 10MB
export DIFF_LENGTH=0


And execute the command "./check -l -T tests/generic/[0-9][0-9][0-9]" to run the xfstests. When it ran out, I find that "tests/generic/482" was failed, and "tests/generic/487" was successed. 


So I check the source code for those two tests, and I found that used different function in them, like below :


tests/generic/482 :


# real QA test starts here


# Modify as appropriate.
_supported_fs generic


_require_no_logdev
_require_command "$KILLALL_PROG" killall


--------------------------------------------------------------


tests/generic/487 : 



# real QA test starts here
_require_scratch_nocheck


sflag='-s'
case $FSTYP in
        btrfs)
                _notrun "btrfs has a specialized test for this"
                ;;
        *)
                ;;
esac


_require_logdev
_require_dm_target error


Then continue to check the source code :

common/rc :



# this test needs a logdev
#
_require_logdev()
{
    [ -z "$SCRATCH_LOGDEV" -o ! -b "$SCRATCH_LOGDEV" ] && \
        _notrun "This test requires a valid \$SCRATCH_LOGDEV"
    [ "$USE_EXTERNAL" != yes ] && \
        _notrun "This test requires USE_EXTERNAL to be enabled"


    # ensure its not mounted
    $UMOUNT_PROG $SCRATCH_LOGDEV 2>/dev/null
}


# This test requires that an external log device is not in use
#
_require_no_logdev()
{
[ "$USE_EXTERNAL" = "yes" ] && [ -n "$SCRATCH_LOGDEV" ] && \
_notrun "Test not compatible with external logs, skipped this test"
}


I find that these two functions behave as mutually exclusive in use, so I don't understand why xfstests can't pass both tests at the same time. Please help me to solve this question, thank you again. 


Looking forward to your reply.

Re: Question: Why tests/generic/482 and tests/generic/487 can't pass both at the same time.

[Zorro Lang]

On Mon, Mar 25, 2024 at 10:47:22AM +0800, 蔡晓斌 wrote:
> 
> Hi, I have a question when I ran the xfstests, please help me to solve this question, tks.
> 
> 
> In my environment, I ues the machine with architecture of aarch64, and then setting the env in local.config file like below:
> 
> 
> 
> 
> # Ideally define at least these 4 to match your environment
> # The first 2 are required.
> # See README for other variables which can be set.
> #
> # Note: SCRATCH_DEV >will< get overwritten!
> 
> 
> export TEST_DEV=/dev/mmcblk0p12
> export TEST_DIR=/mnt/xfstests/dir
> export FSTYP=ext4
> export SCRATCH_DEV=/dev/mmcblk0p13
> export SCRATCH_MNT=/mnt/xfstests/dir-mnt
> export SCRATCH_LOGDEV=/dev/mmcblk0p14
> export USE_EXTERNAL=yes
> export DUMP_CORRUPT_FS=1
> export KEEP_DMESG=yes
> export USE_KMEMLEK=yes
> export LOGWRITES_DEV=/dev/mmcblk0p15
> export MIN_FSSIZE=10485760 # 10MB
> export DIFF_LENGTH=0
> 
> 
> And execute the command "./check -l -T tests/generic/[0-9][0-9][0-9]" to run the xfstests. When it ran out, I find that "tests/generic/482" was failed, and "tests/generic/487" was successed. 
> 
> 
> So I check the source code for those two tests, and I found that used different function in them, like below :
> 
> 
> tests/generic/482 :
> 
> 
> # real QA test starts here
> 
> 
> # Modify as appropriate.
> _supported_fs generic
> 
> 
> _require_no_logdev
> _require_command "$KILLALL_PROG" killall
> 
> 
> --------------------------------------------------------------
> 
> 
> tests/generic/487 : 
> 
> 
> 
> # real QA test starts here
> _require_scratch_nocheck
> 
> 
> sflag='-s'
> case $FSTYP in
>         btrfs)
>                 _notrun "btrfs has a specialized test for this"
>                 ;;
>         *)
>                 ;;
> esac
> 
> 
> _require_logdev
> _require_dm_target error
> 
> 
> Then continue to check the source code :
> 
> common/rc :
> 
> 
> 
> # this test needs a logdev
> #
> _require_logdev()
> {
>     [ -z "$SCRATCH_LOGDEV" -o ! -b "$SCRATCH_LOGDEV" ] && \
>         _notrun "This test requires a valid \$SCRATCH_LOGDEV"
>     [ "$USE_EXTERNAL" != yes ] && \
>         _notrun "This test requires USE_EXTERNAL to be enabled"
> 
> 
>     # ensure its not mounted
>     $UMOUNT_PROG $SCRATCH_LOGDEV 2>/dev/null
> }
> 
> 
> # This test requires that an external log device is not in use
> #
> _require_no_logdev()
> {
> [ "$USE_EXTERNAL" = "yes" ] && [ -n "$SCRATCH_LOGDEV" ] && \
> _notrun "Test not compatible with external logs, skipped this test"
> }
> 
> 
> I find that these two functions behave as mutually exclusive in use, so I don't understand why xfstests can't pass both tests at the same time. Please help me to solve this question, thank you again. 

Not sure what's the problem you hit. g/482 and g/487 are totally different two
test cases.

The g/482 calls _require_no_logdev due to it needs dmlogwrites device, and:

  commit 50f6d46410367586ab151b357e67393599eebd1c
  Author: Darrick J. Wong <djwong@kernel.org>
  Date:   Tue Jul 26 12:48:51 2022 -0700

      dmlogwrites: skip generic tests when external logdev in use

but the g/487 calls _require_logdev due to it needs an external log device, to
avoid the effection from the dm-error of data device.

The g/482 shouldn't fail if there's a SCRATCH_LOGDEV, it should _notrun. If it
failed you'd better to check that failure.

About how to run g/482 and g/487, as one requires logdev, the other one requires
no_log, so you should have two config sections at least, one have SCRATCH_LOGDEV,
the other one unset it. Then run two kinds of tests (with or without log device).

Thanks,
Zorro

> 
> 
> Looking forward to your reply.
> 

Re:Re: Question: Why tests/generic/482 and tests/generic/487 can't pass both at the same time.

[蔡晓斌]

At 2024-03-26 22:16:13, "Zorro Lang" <zlang@redhat.com> wrote:
>On Mon, Mar 25, 2024 at 10:47:22AM +0800, 蔡晓斌 wrote:
>> 
>> Hi, I have a question when I ran the xfstests, please help me to solve this question, tks.
>> 
>> 
>> In my environment, I ues the machine with architecture of aarch64, and then setting the env in local.config file like below:
>> 
>> 
>> 
>> 
>> # Ideally define at least these 4 to match your environment
>> # The first 2 are required.
>> # See README for other variables which can be set.
>> #
>> # Note: SCRATCH_DEV >will< get overwritten!
>> 
>> 
>> export TEST_DEV=/dev/mmcblk0p12
>> export TEST_DIR=/mnt/xfstests/dir
>> export FSTYP=ext4
>> export SCRATCH_DEV=/dev/mmcblk0p13
>> export SCRATCH_MNT=/mnt/xfstests/dir-mnt
>> export SCRATCH_LOGDEV=/dev/mmcblk0p14
>> export USE_EXTERNAL=yes
>> export DUMP_CORRUPT_FS=1
>> export KEEP_DMESG=yes
>> export USE_KMEMLEK=yes
>> export LOGWRITES_DEV=/dev/mmcblk0p15
>> export MIN_FSSIZE=10485760 # 10MB
>> export DIFF_LENGTH=0
>> 
>> 
>> And execute the command "./check -l -T tests/generic/[0-9][0-9][0-9]" to run the xfstests. When it ran out, I find that "tests/generic/482" was failed, and "tests/generic/487" was successed. 
>> 
>> 
>> So I check the source code for those two tests, and I found that used different function in them, like below :
>> 
>> 
>> tests/generic/482 :
>> 
>> 
>> # real QA test starts here
>> 
>> 
>> # Modify as appropriate.
>> _supported_fs generic
>> 
>> 
>> _require_no_logdev
>> _require_command "$KILLALL_PROG" killall
>> 
>> 
>> --------------------------------------------------------------
>> 
>> 
>> tests/generic/487 : 
>> 
>> 
>> 
>> # real QA test starts here
>> _require_scratch_nocheck
>> 
>> 
>> sflag='-s'
>> case $FSTYP in
>>         btrfs)
>>                 _notrun "btrfs has a specialized test for this"
>>                 ;;
>>         *)
>>                 ;;
>> esac
>> 
>> 
>> _require_logdev
>> _require_dm_target error
>> 
>> 
>> Then continue to check the source code :
>> 
>> common/rc :
>> 
>> 
>> 
>> # this test needs a logdev
>> #
>> _require_logdev()
>> {
>>     [ -z "$SCRATCH_LOGDEV" -o ! -b "$SCRATCH_LOGDEV" ] && \
>>         _notrun "This test requires a valid \$SCRATCH_LOGDEV"
>>     [ "$USE_EXTERNAL" != yes ] && \
>>         _notrun "This test requires USE_EXTERNAL to be enabled"
>> 
>> 
>>     # ensure its not mounted
>>     $UMOUNT_PROG $SCRATCH_LOGDEV 2>/dev/null
>> }
>> 
>> 
>> # This test requires that an external log device is not in use
>> #
>> _require_no_logdev()
>> {
>> [ "$USE_EXTERNAL" = "yes" ] && [ -n "$SCRATCH_LOGDEV" ] && \
>> _notrun "Test not compatible with external logs, skipped this test"
>> }
>> 
>> 
>> I find that these two functions behave as mutually exclusive in use, so I don't understand why xfstests can't pass both tests at the same time. Please help me to solve this question, thank you again. 
>
>Not sure what's the problem you hit. g/482 and g/487 are totally different two
>test cases.
>
>The g/482 calls _require_no_logdev due to it needs dmlogwrites device, and:
>
>  commit 50f6d46410367586ab151b357e67393599eebd1c
>  Author: Darrick J. Wong <djwong@kernel.org>
>  Date:   Tue Jul 26 12:48:51 2022 -0700
>
>      dmlogwrites: skip generic tests when external logdev in use
>
>but the g/487 calls _require_logdev due to it needs an external log device, to
>avoid the effection from the dm-error of data device.
>
>The g/482 shouldn't fail if there's a SCRATCH_LOGDEV, it should _notrun. If it
>failed you'd better to check that failure.
>
>About how to run g/482 and g/487, as one requires logdev, the other one requires
>no_log, so you should have two config sections at least, one have SCRATCH_LOGDEV,
>the other one unset it. Then run two kinds of tests (with or without log device).
>
>Thanks,
>Zorro

>


Hi Zorro, thank you for resolving my question, I know what to do next.

>> 
>> 
>> Looking forward to your reply.
>>