Re: [PATCH 1/5] libperf: Populate system-wide evsel maps

[Namhyung Kim]

Hi Ian,

On Thu, Oct 6, 2022 at 11:44 AM Ian Rogers <irogers@xxxxxxxxxx> wrote:
>
> On Mon, Oct 3, 2022 at 1:46 PM Namhyung Kim <namhyung@xxxxxxxxxx> wrote:
> >
> > Setting proper cpu and thread maps for system wide evsels regardless of
> > user requested cpu in __perf_evlist__propagate_maps().  Those evsels
> > need to be active on all cpus always.  Do it in the libperf so that we
> > can guarantee it has proper maps.
> >
> > Reviewed-by: Adrian Hunter <adrian.hunter@xxxxxxxxx>
> > Signed-off-by: Namhyung Kim <namhyung@xxxxxxxxxx>
> > ---
> >  tools/lib/perf/evlist.c | 15 +++++++++------
> >  1 file changed, 9 insertions(+), 6 deletions(-)
> >
> > diff --git a/tools/lib/perf/evlist.c b/tools/lib/perf/evlist.c
> > index 6b1bafe267a4..187129652ab6 100644
> > --- a/tools/lib/perf/evlist.c
> > +++ b/tools/lib/perf/evlist.c
> > @@ -40,11 +40,11 @@ static void __perf_evlist__propagate_maps(struct perf_evlist *evlist,
> >          * We already have cpus for evsel (via PMU sysfs) so
> >          * keep it, if there's no target cpu list defined.
> >          */
>
> This comment is for the 'else if' case and so is a little out of place
> before the system wide test.

Right, I'll add a comment for system-wide evsels too.

>
> > -       if (!evsel->own_cpus ||
> > -           (!evsel->system_wide && evlist->has_user_cpus) ||
> > -           (!evsel->system_wide &&
> > -            !evsel->requires_cpu &&
> > -            perf_cpu_map__empty(evlist->user_requested_cpus))) {
> > +       if (evsel->system_wide) {
> > +               perf_cpu_map__put(evsel->cpus);
> > +               evsel->cpus = perf_cpu_map__new(NULL);
>
> Looking at perf_cpu_map__new, will this mean that in system wide mode
> every event/evsel will now read /sys/devices/system/cpu/online here?
> We may want to cache the cpumap to avoid this.

Yeah, it's redundant.  I thought it's ok since the system-wide evsels
are not common.  We can consider caching when it becomes a problem.

Thanks,
Namhyung


>
> > +       } else if (!evsel->own_cpus || evlist->has_user_cpus ||
> > +                  (!evsel->requires_cpu && perf_cpu_map__empty(evlist->user_requested_cpus))) {
> >                 perf_cpu_map__put(evsel->cpus);
> >                 evsel->cpus = perf_cpu_map__get(evlist->user_requested_cpus);
> >         } else if (evsel->cpus != evsel->own_cpus) {
> > @@ -52,7 +52,10 @@ static void __perf_evlist__propagate_maps(struct perf_evlist *evlist,
> >                 evsel->cpus = perf_cpu_map__get(evsel->own_cpus);
> >         }
> >
> > -       if (!evsel->system_wide) {
> > +       if (evsel->system_wide) {
> > +               perf_thread_map__put(evsel->threads);
> > +               evsel->threads = perf_thread_map__new_dummy();
> > +       } else {
> >                 perf_thread_map__put(evsel->threads);
> >                 evsel->threads = perf_thread_map__get(evlist->threads);
> >         }
> > --
> > 2.38.0.rc1.362.ged0d419d3c-goog
> >