On Tue, Jun 10, 2025, Dapeng Mi wrote:
> On 5/30/2025 6:19 AM, Sean Christopherson wrote:
> > @@ -51,7 +51,7 @@ void pmu_init(void)
> > }
> > pmu.gp_counter_width = PMC_DEFAULT_WIDTH;
> > pmu.gp_counter_mask_length = pmu.nr_gp_counters;
> > - pmu.gp_counter_available = (1u << pmu.nr_gp_counters) - 1;
> > + pmu.arch_event_available = (1u << pmu.nr_gp_counters) - 1;
>
> "available architectural events" and "available GP counters" are two
> different things. I know this would be changed in later patch 09/16, but
> it's really confusing. Could we merge the later patch 09/16 into this patch?
Ya. I was trying to not mix too many things in one patch, but looking at this
again, I 100% agree that squashing 7-9 into one patch is better overall.
> > @@ -463,7 +463,7 @@ static void check_counters_many(void)
> > int i, n;
> >
> > for (i = 0, n = 0; n < pmu.nr_gp_counters; i++) {
> > - if (!pmu_gp_counter_is_available(i))
> > + if (!pmu_arch_event_is_available(i))
> > continue;
>
> The intent of check_counters_many() is to verify all available GP and fixed
> counters can count correctly at the same time. So we should select another
> available event to verify the counter instead of skipping the counter if an
> event is not available.
Agreed, but I'm going to defer that for now, this series already wanders in too
many directions. Definitely feel free to post a patch.
