Re: rcu stalls during fstests runs for xfs

[Matthieu Baerts <matttbe@kernel.org>]

Hi Shinichiro,

Sorry to jump in, but I *think* our CI for the MPTCP subsystem is
hitting the same issue.

On 30/01/2026 12:16, Shinichiro Kawasaki wrote:
> On Jan 29, 2026 / 15:19, Paul E. McKenney wrote:
> [...]
>>>>> I have seen the static-key pattern called out by Dave Chinner when running
>>>>> KASAN on large systems.  We worked around this by disabling KASAN's use
>>>>> of static keys.  In case you were running KASAN in these tests.
>>>>
>>>> As to KASAN, yes, I enable it in my test runs. I find three static-keys under
>>>> mm/kasan/*. I will think if they can be disabled in my test runs. Thanks.
>>>
>>> There is a set of Kconfig options that disables static branches.  If you
>>> cannot find them quickly, please let me know and I can look them up.
> 
> Thank you. But now I know the fix series by Thomas is available. I prioritize
> the evaluation of the fix series. Later on, I will try disabling the static-keys
> if it is required.
> 
>>
>> And Thomas Gleixner posted an alleged fix to the CID issue here:
>>
>> https://lore.kernel.org/lkml/20260129210219.452851594@kernel.org/
>>
>> Please let him know whether or not it helps.
> 
> Good to see this fix candidate series, thanks :) I have set up the patches and
> started my regular test runs. So far, the hangs have been observed once or twice
> a week. To confirm the effect of the fix series, I think two weeks runs will be
> required. Once I get the result, will share it on this thread and with Thomas.

I know it is only one week now, but did you see any effects so far? On
my side, I applied the v2 series -- which has been applied in
tip/sched/urgent -- but I still have issues, and it looks like it is
even more frequent. Maybe what I see is different. If you no longer see
the issues on your side after one week, I'm going to start a new thread
with my issues not to mix them.

Note that in my case, the issue is visible on a system where nested VMs
are used, with and without KASAN (enabled via debug.config), just after
having started a VSOCK listening socket via socat.

Cheers,
Matt
-- 
Sponsored by the NGI0 Core fund.