Re: [PATCH] rcu: Enable runtime reset of the RCU stall panic count

[Joel Fernandes <joelagnelf@nvidia.com>]

> On Oct 6, 2026, at 10:56 PM, Cunlong Li <shenxiaogll@gmail.com> wrote:
> 
> On Tue, Oct 06, 2026 at 05:27:51PM -0400, Joel Fernandes wrote:
>> 
>> 
>> On 10/6/2026 9:17 AM, Cunlong Li wrote:
>>>> It sounds like you do know that it is likely transient - but based on this description and the below, it sounds like you have not yet found a use of this patch.
>>>> 
>>>> 
>>> To be precise, we worry that some of the stalls were transient, so a
>>> panic reboot would affect the users.  But we have no hard evidence
>>> either way: every incident ended in a forced reset, and the dmesg was
>>> lost.  Going forward, we plan to enable the NMI and kdump to keep
>>> investigating the cause of the RCU stalls. We will also monitor dmesg
>>> to confirm whether transient RCU stalls actually occur.
>> 
>> 
>> Right, I just worry that if you're not seeing recoverable/transient stalls and
>> you start using this patch, then you might be causing side-effects that don't
>> really help your problem and might be masking the real issue. It would be good
>> to debug this further, and fix the right way IMO.
>> 
>> If you show that this patch indeed helps your problem, we can consider it for
>> 7.5, but there's little bit of time. :-)
> 
> Hi Joel,
> 
> Understood, thanks.  We will keep monitoring dmesg for the root cause.
> Once we observe self-recovering RCU stalls, we will resubmit the patch
> with the evidence.

Thank you!!!

Joel

> 
> Thanks,
> Cunlong
> 
>> 
>> Thanks,
>> 
>> Joel Fernandes
>> 
>>