Re: [PATCH 5.10 0/2] Fix xfs/179 for 5.10 stable

[Guoqing Jiang <guoqing.jiang@linux.dev>]

On 8/7/23 17:11, Greg KH wrote:
> On Mon, Aug 07, 2023 at 10:39:44AM +0800, Guoqing Jiang wrote:
>>
>> On 8/4/23 23:47, Darrick J. Wong wrote:
>>> On Thu, Aug 03, 2023 at 05:36:50PM +0800, Guoqing Jiang wrote:
>>>> Hi,
>>>>
>>>> With the two patches applied, xfs/179 can pass in 5.10.188. Otherwise I got
>>>>
>>>> [root@localhost xfstests]# ./check xfs/179
>>>> FSTYP         -- xfs (non-debug)
>>>> PLATFORM      -- Linux/x86_64 localhost 5.10.188-default #14 SMP Thu Aug 3 15:23:19 CST 2023
>>>> MKFS_OPTIONS  -- -f /dev/loop1
>>>> MOUNT_OPTIONS -- -o context=system_u:object_r:root_t:s0 /dev/loop1 /mnt/scratch
>>>>
>>>> xfs/179 1s ... [failed, exit status 1]- output mismatch (see /root/xfstests/results//xfs/179.out.bad)
>>>>       --- tests/xfs/179.out	2023-07-13 16:12:27.000000000 +0800
>>>>       +++ /root/xfstests/results//xfs/179.out.bad	2023-08-03 16:55:38.173787911 +0800
>>>>       @@ -8,3 +8,5 @@
>>>>        Check scratch fs
>>>>        Remove reflinked files
>>>>        Check scratch fs
>>>>       +xfs_repair fails
>>>>       +(see /root/xfstests/results//xfs/179.full for details)
>>>>       ...
>>>>       (Run 'diff -u /root/xfstests/tests/xfs/179.out /root/xfstests/results//xfs/179.out.bad'  to see the entire diff)
>>>>
>>>> HINT: You _MAY_ be missing kernel fix:
>>>>         b25d1984aa88 xfs: estimate post-merge refcounts correctly
>>>>
>>>> Ran: xfs/179
>>>> Failures: xfs/179
>>>> Failed 1 of 1 tests
>>>>
>>>> Please review if they are approriate for 5.10 stable.
>>> Seems fine to me, but ... there is no maintainer for 5.10; is your
>>> employer willing to support this LTS kernel?
>> Hi Darrick,
>>
>> Thanks for your review! I think Amir is the maintainer for 5.10 😉. I can
>> help
>> if needed since our kernel is heavily based on 5.10 stable. We also run
>> tests
>> against 5.10 stable, that is why I send fixes patches for it.
>>
>> Hi Greg,
>>
>> Could you consider add the two to your list? Thank you!
> Sorry, but as these would only be in th 5.10.y release, and not in any
> newer stable kernel, you would have a regression if you moved to a newer
> stable kernel branch, right?

Fair enough.

> Because of that, no, I can't take this, nor should you want me to, as
> you would have a regression if you upgraded, right?
>
> I'll be glad to do so if we have backports for all relevant stable
> kernels.
>

I found Ted had mentioned the two commits ([1]), so it probably make sense
to port them to 5.10 as well.

[1]. https://lore.kernel.org/linux-xfs/20230802031039.GC358316@mit.edu/

Thanks,
Guoqing