Re: [PATCH v13 01/17] preempt: Track NMI nesting to separate per-CPU counter

[Joel Fernandes <joelagnelf@nvidia.com>]

On 10/14/2025 4:25 AM, Peter Zijlstra wrote:
> On Mon, Oct 13, 2025 at 05:27:32PM -0400, Joel Fernandes wrote:
>>
>>
>> On 10/13/2025 4:00 PM, Peter Zijlstra wrote:
>>> On Mon, Oct 13, 2025 at 11:48:03AM -0400, Lyude Paul wrote:
>>>> From: Joel Fernandes <joelagnelf@nvidia.com>
>>>>
>>>> Move NMI nesting tracking from the preempt_count bits to a separate per-CPU
>>>> counter (nmi_nesting). This is to free up the NMI bits in the preempt_count,
>>>> allowing those bits to be repurposed for other uses.  This also has the benefit
>>>> of tracking more than 16-levels deep if there is ever a need.
>>>>
>>>> Suggested-by: Boqun Feng <boqun.feng@gmail.com>
>>>> Signed-off-by: Joel Fernandes <joelaf@google.com>
>>>> Signed-off-by: Lyude Paul <lyude@redhat.com>
>>>> ---
>>>>  include/linux/hardirq.h   | 17 +++++++++++++----
>>>>  kernel/softirq.c          |  2 ++
>>>>  rust/kernel/alloc/kvec.rs |  5 +----
>>>>  rust/kernel/cpufreq.rs    |  3 +--
>>>>  4 files changed, 17 insertions(+), 10 deletions(-)
>>>>
>>>> diff --git a/include/linux/hardirq.h b/include/linux/hardirq.h
>>>> index d57cab4d4c06f..177eed1de35cc 100644
>>>> --- a/include/linux/hardirq.h
>>>> +++ b/include/linux/hardirq.h
>>>> @@ -10,6 +10,8 @@
>>>>  #include <linux/vtime.h>
>>>>  #include <asm/hardirq.h>
>>>>  
>>>> +DECLARE_PER_CPU(unsigned int, nmi_nesting);
>>>
>>> Urgh, and it isn't even in the same cacheline as the preempt_count :/
>>
>> Great point. I will move this to DECLARE_PER_CPU_CACHE_HOT()
>> so it's co-located with preempt_count and run some tests. Let me know if that
>> works for you, thanks!
> 
> Well, I hate how on entry we then end up incrementing both. How terrible
> would it be to make __preempt_count u64 instead?

Would that break 32-bit x86? I have to research this more. This was what I
initially thought of doing but ISTR some challenges. I'd like to think that was
my imagination, but I will revisit it and see what it takes.

Thanks!

 - Joel