lkml.org 
[lkml]   [2022]   [Aug]   [25]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [RFC v2] perf: Rewrite core context handling
On Thu, Aug 25, 2022 at 11:09:05AM +0530, Ravi Bangoria wrote:
> > -static inline int __pmu_filter_match(struct perf_event *event)
> > -{
> > - struct pmu *pmu = event->pmu;
> > - return pmu->filter_match ? pmu->filter_match(event) : 1;
> > -}
> > -
> > -/*
> > - * Check whether we should attempt to schedule an event group based on
> > - * PMU-specific filtering. An event group can consist of HW and SW events,
> > - * potentially with a SW leader, so we must check all the filters, to
> > - * determine whether a group is schedulable:
> > - */
> > -static inline int pmu_filter_match(struct perf_event *event)
> > -{
> > - struct perf_event *sibling;
> > -
> > - if (!__pmu_filter_match(event))
> > - return 0;
> > -
> > - for_each_sibling_event(sibling, event) {
> > - if (!__pmu_filter_match(sibling))
> > - return 0;
> > - }
> > -
> > - return 1;
> > -}
> > -
> > static inline int
> > event_filter_match(struct perf_event *event)
> > {
> > return (event->cpu == -1 || event->cpu == smp_processor_id()) &&
> > - perf_cgroup_match(event) && pmu_filter_match(event);
> > + perf_cgroup_match(event);
>
> There are many callers of event_filter_match() which might not endup calling
> visit_groups_merge(). I hope this is intentional change?

I thought I did, but lets go through them again.

event_filter_match() is called from:

- __perf_event_enable(); here we'll end up in ctx_sched_in() which
will dutifully skip the pmu in question.

(fwiw, this is one of those sites where ctx_sched_{out,in}() could do
with a @pmu argument.

- merge_sched_in(); this is after the new callsite in
visit_groups_merge().

- perf_adjust_freq_unthrottle_context(); if the pmu was skipped in
visit_groups_merge() then ->state != ACTIVE and we'll bail out.

- perf_iterate_ctx() / perf_iterate_sb_cpu(); these are for generating
side-band events, and arguably not delivering them when running on
the 'wrong' CPU wasn't right to begin with.


So I tihnk we're good. Hmm?

\
 
 \ /
  Last update: 2022-08-25 11:19    [W:0.101 / U:0.396 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site