Message-ID: <ff6b4e2f-dbd3-479b-a522-a1ae4837b3df@gmail.com>
Date: Thu, 6 Feb 2025 23:32:43 +0530
From: "Nirjhar Roy (IBM)" <nirjhar.roy.lists@...il.com>
To: Zorro Lang <zlang@...hat.com>, "Darrick J. Wong" <djwong@...nel.org>
Cc: fstests@...r.kernel.org, linux-ext4@...r.kernel.org,
 linux-xfs@...r.kernel.org, ritesh.list@...il.com, ojaswin@...ux.ibm.com,
 zlang@...nel.org
Subject: Re: [PATCH v2] check: Fix fs specfic imports when
 $FSTYPE!=$OLD_FSTYPE


On 2/1/25 12:05, Zorro Lang wrote:
> On Fri, Jan 31, 2025 at 08:24:57AM -0800, Darrick J. Wong wrote:
>> On Fri, Jan 31, 2025 at 06:49:50PM +0530, Nirjhar Roy (IBM) wrote:
>>> On 1/29/25 21:32, Darrick J. Wong wrote:
>>>> On Wed, Jan 29, 2025 at 04:48:10PM +0530, Nirjhar Roy (IBM) wrote:
>>>>> On 1/28/25 23:39, Darrick J. Wong wrote:
>>>>>> On Tue, Jan 28, 2025 at 05:00:22AM +0000, Nirjhar Roy (IBM) wrote:
>>>>>>> Bug Description:
>>>>>>>
>>>>>>> _test_mount function is failing with the following error:
>>>>>>> ./common/rc: line 4716: _xfs_prepare_for_eio_shutdown: command not found
>>>>>>> check: failed to mount /dev/loop0 on /mnt1/test
>>>>>>>
>>>>>>> when the second section in local.config file is xfs and the first section
>>>>>>> is non-xfs.
>>>>>>>
>>>>>>> It can be easily reproduced with the following local.config file
>>>>>>>
>>>>>>> [s2]
>>>>>>> export FSTYP=ext4
>>>>>>> export TEST_DEV=/dev/loop0
>>>>>>> export TEST_DIR=/mnt1/test
>>>>>>> export SCRATCH_DEV=/dev/loop1
>>>>>>> export SCRATCH_MNT=/mnt1/scratch
>>>>>>>
>>>>>>> [s1]
>>>>>>> export FSTYP=xfs
>>>>>>> export TEST_DEV=/dev/loop0
>>>>>>> export TEST_DIR=/mnt1/test
>>>>>>> export SCRATCH_DEV=/dev/loop1
>>>>>>> export SCRATCH_MNT=/mnt1/scratch
>>>>>>>
>>>>>>> ./check selftest/001
>>>>>>>
>>>>>>> Root cause:
>>>>>>> When _test_mount() is executed for the second section, the FSTYPE has
>>>>>>> already changed but the new fs specific common/$FSTYP has not yet
>>>>>>> been done. Hence _xfs_prepare_for_eio_shutdown() is not found and
>>>>>>> the test run fails.
>>>>>>>
>>>>>>> Fix:
>>>>>>> Remove the additional _test_mount in check file just before ". commom/rc"
>>>>>>> since ". commom/rc" is already sourcing fs specific imports and doing a
>>>>>>> _test_mount.
>>>>>>>
>>>>>>> Fixes: 1a49022fab9b4 ("fstests: always use fail-at-unmount semantics for XFS")
>>>>>>> Signed-off-by: Nirjhar Roy (IBM) <nirjhar.roy.lists@...il.com>
>>>>>>> ---
>>>>>>>     check | 12 +++---------
>>>>>>>     1 file changed, 3 insertions(+), 9 deletions(-)
>>>>>>>
>>>>>>> diff --git a/check b/check
>>>>>>> index 607d2456..5cb4e7eb 100755
>>>>>>> --- a/check
>>>>>>> +++ b/check
>>>>>>> @@ -784,15 +784,9 @@ function run_section()
>>>>>>>     			status=1
>>>>>>>     			exit
>>>>>>>     		fi
>>>>>>> -		if ! _test_mount
>>>>>> Don't we want to _test_mount the newly created filesystem still?  But
>>>>>> perhaps after sourcing common/rc ?
>>>>>>
>>>>>> --D
>>>>> common/rc calls init_rc() in the end and init_rc() already does a
>>>>> _test_mount. _test_mount after sourcing common/rc will fail, won't it? Does
>>>>> that make sense?
>>>>>
>>>>> init_rc()
>>>>> {
>>>>>       # make some further configuration checks here
>>>>>       if [ "$TEST_DEV" = ""  ]
>>>>>       then
>>>>>           echo "common/rc: Error: \$TEST_DEV is not set"
>>>>>           exit 1
>>>>>       fi
>>>>>
>>>>>       # if $TEST_DEV is not mounted, mount it now as XFS
>>>>>       if [ -z "`_fs_type $TEST_DEV`" ]
>>>>>       then
>>>>>           # $TEST_DEV is not mounted
>>>>>           if ! _test_mount
>>>>>           then
>>>>>               echo "common/rc: retrying test device mount with external set"
>>>>>               [ "$USE_EXTERNAL" != "yes" ] && export USE_EXTERNAL=yes
>>>>>               if ! _test_mount
>>>>>               then
>>>>>                   echo "common/rc: could not mount $TEST_DEV on $TEST_DIR"
>>>>>                   exit 1
>>>>>               fi
>>>>>           fi
>>>>>       fi
>>>>> ...
>>>> ahahahaha yes it does.
>>>>
>>>> /commit message reading comprehension fail, sorry about that.
>>>>
>>>> Though now that you point it out, should check elide the init_rc call
>>>> about 12 lines down if it re-sourced common/rc ?
>>> Yes, it should. init_rc() is getting called twice when common/rc is getting
>>> re-sourced. Maybe I can do like
>>>
>>>
>>> if $RECREATE_TEST_DEV || [ "$OLD_FSTYP" != "$FSTYP" ]; then
>>>
>>>      <...>
>>>
>>>      . common/rc # changes in this patch
>>>
>>>      <...>
>>>
>>> elif [ "$OLD_TEST_FS_MOUNT_OPTS" != "$TEST_FS_MOUNT_OPTS" ]; then
>>>
>>>      ...
>>>
>>>      init_rc() # explicitly adding an init_rc() for this condition
>>>
>>> else
>>>
>>>      init_rc() # # explicitly adding an init_rc() for all other conditions.
>>> This will prevent init_rc() from getting called twice during re-sourcing
>>> common/rc
>>>
>>> fi
>>>
>>> What do you think?
>> Sounds fine as a mechanical change, but I wonder, should calling init_rc
>> be explicit?  There are not so many places that source common/rc:
>>
>> $ git grep 'common/rc'
>> check:362:if ! . ./common/rc; then
>> check:836:              . common/rc
>> common/preamble:52:     . ./common/rc
>> soak:7:. ./common/rc
>> tests/generic/749:18:. ./common/rc
>>
>> (I filtered out the non-executable matches)
>>
>> I think the call in generic/749 is unnecessary and I don't know what
>> soak does.  But that means that one could insert an explicit call to
>> init_rc at line 366 and 837 in check and at line 53 in common/preamble,
>> and we can clean up one more of those places where sourcing a common/
>> file actually /does/ something quietly under the covers.
>>
>> (Unless the maintainer is ok with the status quo...?)
> I think people just hope to import the helpers in common/rc mostly, don't
> want to run init_rc again. Maybe we can make sure the init_rc is only run
> once each time?
>
> E.g.
>
>    if [ _INIT_RC != "done" ];then
> 	init_rc
> 	_INIT_RC="done"
>    fi
>
> Or any better idea.

Yes, this idea looks good too. However, after thinking a bit more, I 
like Darrick's idea to remove the call to init_rc from common/rc and 
explicitly calling them explicitly whenever necessary makes more sense. 
This also makes the interface/reason to source common/rc more 
meaningful, and also not making common/rc do something via init_rc 
silently. What do you think?

--NR

>
> Thanks,
> Zorro
>
>> --D
>>
>>>> Reviewed-by: "Darrick J. Wong" <djwong@...nel.org>
>>>>
>>>> --D
>>>>
>>>>> ...
>>>>>
>>>>> --NR
>>>>>
>>>>>
>>>>>
>>>>>>> -		then
>>>>>>> -			echo "check: failed to mount $TEST_DEV on $TEST_DIR"
>>>>>>> -			status=1
>>>>>>> -			exit
>>>>>>> -		fi
>>>>>>> -		# TEST_DEV has been recreated, previous FSTYP derived from
>>>>>>> -		# TEST_DEV could be changed, source common/rc again with
>>>>>>> -		# correct FSTYP to get FSTYP specific configs, e.g. common/xfs
>>>>>>> +		# Previous FSTYP derived from TEST_DEV could be changed, source
>>>>>>> +		# common/rc again with correct FSTYP to get FSTYP specific configs,
>>>>>>> +		# e.g. common/xfs
>>>>>>>     		. common/rc
>>>>>>>     		_prepare_test_list
>>>>>>>     	elif [ "$OLD_TEST_FS_MOUNT_OPTS" != "$TEST_FS_MOUNT_OPTS" ]; then
>>>>>>> -- 
>>>>>>> 2.34.1
>>>>>>>
>>>>>>>
>>>>> -- 
>>>>> Nirjhar Roy
>>>>> Linux Kernel Developer
>>>>> IBM, Bangalore
>>>>>
>>> -- 
>>> Nirjhar Roy
>>> Linux Kernel Developer
>>> IBM, Bangalore
>>>
>>>
-- 
Nirjhar Roy
Linux Kernel Developer
IBM, Bangalore

Powered by blists - more mailing lists