Re: [PATCHv9 2/3] irq: use a struct for the kstat_irqs in the interrupt descriptor

[Thomas Gleixner]

On Fri, Feb 23 2024 at 15:18, Bitao Hu wrote:
> On 2024/2/22 21:22, Thomas Gleixner wrote:
>>> -	if (desc->kstat_irqs) {
>>> -		for_each_online_cpu(j)
>>> -			any_count |= data_race(*per_cpu_ptr(desc->kstat_irqs, j));
>>> -	}
>>> +	if (desc->kstat_irqs)
>>> +		any_count = data_race(desc->tot_count);
>> 
>> This is an unrelated change and needs to be split out into a separate
>> patch with a proper changelog which explains why this is equivalent.
>> 
>
> Alright, I will remove this change witch is not related to the purpose
> of this patch.
>
> I guess that the purpose of suggesting this change in your V1 response
> was to speedup the 'show_interrupts'. However, after reviewing the
> usage of 'desc->tot_count' in 'unsigned int kstat_irqs(unsigned int 
> irq)', I think the change might be as follows:
>
> diff --git a/kernel/irq/proc.c b/kernel/irq/proc.c
> index 623b8136e9af..53b8d6edd7ac 100644
> --- a/kernel/irq/proc.c
> +++ b/kernel/irq/proc.c
> @@ -489,8 +489,13 @@ int show_interrupts(struct seq_file *p, void *v)
>                  goto outsparse;
>
>          if (desc->kstat_irqs) {
> -               for_each_online_cpu(j)
> -                       any_count |= 
> data_race(per_cpu(desc->kstat_irqs->cnt, j));
> +               if (!irq_settings_is_per_cpu_devid(desc) &&
> +                   !irq_settings_is_per_cpu(desc) &&
> +                   !irq_is_nmi(desc))
> +                       any_count = data_race(desc->tot_count);
> +               else
> +                       for_each_online_cpu(j)
> +                               any_count |= 
> data_race(per_cpu(desc->kstat_irqs->cnt, j));
>          }
>
>          if ((!desc->action || irq_desc_is_chained(desc)) && !any_count)
>
> Is my idea correct?

Yes.