> On Mar 20, 2024, at 2:47 AM, Amir Goldstein <amir73il@xxxxxxxxx> wrote:
>
> On Wed, Mar 20, 2024 at 12:32 AM Jan Kara <jack@xxxxxxx> wrote:
>>
>> On Tue 19-03-24 18:10:15, Amir Goldstein wrote:
>>> On Tue, Mar 19, 2024 at 5:57 PM Chuck Lever III <chuck.lever@xxxxxxxxxx> wrote:
>>>>> On Mar 19, 2024, at 11:32 AM, Sasha Levin <sashal@xxxxxxxxxx> wrote:
>>>>> On Tue, Mar 19, 2024 at 04:26:34AM +0200, Amir Goldstein wrote:
>>>>>> Sasha,
>>>>>> Something is off.
>>>>>> This is a new feature.
>>>>>> Not sure how it got selected for stable and dragged a *lot* of
>>>>>> infrastructure code changes with it.
>>>>>> Can you explain why triggered this backports or is it just "AI"?
>>>>>
>>>>> Hey Amir,
>>>>>
>>>>> The patches you've pointed out are part of a series backported by Chuck
>>>>> for the benefit of nfsd.
>>>>>
>>>>> In general, we don't object to new functionality as long as:
>>>>>
>>>>> 1. It helps reduce divergence of later fixes from upstream.
>>>>> 2. It's well tested.
>>>>
>>>> Amir, this is why I asked you about how you test fanotify.
>>>>
>>>> The goal of my backport was to address issues with the NFSD
>>>> filecache, and unfortunately, a lot of it depends on fixes
>>>> and features in fs/notify.
>>>
>>> OK, I wonder which features filecache depends on?
>>> I can't believe that it depends on any of the final
>>> "wire up fanotify XXX" commits.
>>>
>>> Anyway, I do not have an objection to backporting those features,
>>> just wanted to know if there was a reason.
>>> In house, we are using the 5.15 LTS kernel with some of those
>>> features backported.
>>>
>>> Jan, WDYT?
>>
>> So if somebody (Chuck in this case) actually consciously backported stuff
>> and tested everything works then I have no objection...
>>
>
> OK. fun!
>
> Chuck,
>
> If we are going to backport those features to 5.15.y, let's do it properly.
> Please follow up with update of man pages and LTS tests where relevant.
> I will reply with notes to specific patches.
>
> For this one, need to update fanotify_mark(2) once commits with the specific
> 5.15.y version after commits are merged, i.e.:
>
> FAN_FS_ERROR (since Linux 5.16 and 5.15.???)
Note that I'm planning similar updates for v5.10.y.
And thank you for your time and your review!
> LTP test fanotify22 tests FAN_FS_ERROR.
> This test indicates that there is a bug fix that needs to be backported:
>
> .tags = (const struct tst_tag[]) {
> {"linux-git", "124e7c61deb2"},
>
> I guess this was not backported? so I wonder how come you have not
> observed the issue with fanotify22.
> Maybe the test does not cover it??
Yes, fanotify22 failed for me, and I applied that patch and
retested without failure. Where are you seeing that issue?
Check the nfsd-5.15.y branch in my tree against what is
pending for linux-5.15.y.
--
Chuck Lever
