lkml.org 
[lkml]   [2021]   [Apr]   [25]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: [PATCH v4 3/4] perf-stat: introduce config stat.bpf-counter-events
Date


> On Apr 25, 2021, at 12:09 PM, Song Liu <songliubraving@fb.com> wrote:
>
>
>
>> On Apr 25, 2021, at 8:16 AM, Jiri Olsa <jolsa@redhat.com> wrote:
>>
>> On Sun, Apr 25, 2021 at 11:38:43AM -0300, Arnaldo Carvalho de Melo wrote:
>>> Em Wed, Apr 21, 2021 at 12:18:23PM +0200, Jiri Olsa escreveu:
>>>> On Tue, Apr 20, 2021 at 09:21:32PM +0000, Song Liu wrote:
>>>>>
>>>>>
>>>>>> On Apr 20, 2021, at 10:31 AM, Jiri Olsa <jolsa@redhat.com> wrote:
>>>>>>
>>>>>> On Mon, Apr 19, 2021 at 01:36:48PM -0700, Song Liu wrote:
>>>>>>
>>>>>> SNIP
>>>>>>
>>>>>>> if (stat_config.initial_delay < 0) {
>>>>>>> @@ -784,11 +790,11 @@ static int __run_perf_stat(int argc, const char **argv, int run_idx)
>>>>>>> if (affinity__setup(&affinity) < 0)
>>>>>>> return -1;
>>>>>>>
>>>>>>> - if (target__has_bpf(&target)) {
>>>>>>> - evlist__for_each_entry(evsel_list, counter) {
>>>>>>> - if (bpf_counter__load(counter, &target))
>>>>>>> - return -1;
>>>>>>> - }
>>>>>>> + evlist__for_each_entry(evsel_list, counter) {
>>>>>>> + if (bpf_counter__load(counter, &target))
>>>>>>> + return -1;
>>>>>>> + if (!evsel__is_bpf(counter))
>>>>>>> + all_counters_use_bpf = false;
>>>>>>
>>>>>> could be done in bpf_counter__load, check below:
>>>>>>
>>>>>>> }
>>>>>>>
>>>>>>> evlist__for_each_cpu (evsel_list, i, cpu) {
>>>>>>> diff --git a/tools/perf/util/bpf_counter.c b/tools/perf/util/bpf_counter.c
>>>>>>> index 5de991ab46af9..33b1888103dfa 100644
>>>>>>> --- a/tools/perf/util/bpf_counter.c
>>>>>>> +++ b/tools/perf/util/bpf_counter.c
>>>>>>> @@ -790,7 +790,8 @@ int bpf_counter__load(struct evsel *evsel, struct target *target)
>>>>>>> {
>>>>>>> if (target->bpf_str)
>>>>>>> evsel->bpf_counter_ops = &bpf_program_profiler_ops;
>>>>>>> - else if (target->use_bpf)
>>>>>>> + else if (target->use_bpf ||
>>>>>>> + evsel__match_bpf_counter_events(evsel->name))
>>>>>>> evsel->bpf_counter_ops = &bperf_ops;
>>>>>>
>>>>>> with:
>>>>>> else
>>>>>> all_counters_use_bpf = false;
>>>>>>
>>>>>> I was also thinking of oving it to evlist, but it's sat specific,
>>>>>> so I think it's good as static.. thanks for changing the implementation
>>>>>
>>>>> Hmm... then we need to somehow make all_counters_use_bpf visible in
>>>>> bpf_counter.c, which won't be very clean. Also, since this is stat
>>>>> specific, I guess it is better to keep it inside builtin-stat.c?
>>>>> The runtime overhead should be minimal.
>>>>
>>>> ah it's different file :) then it's better as it is, sorry
>>>
>>> Is this a Reviewed-by?
>>
>> there's still the matter of disable callback:
>> https://lore.kernel.org/lkml/YH8Pw4m0w6DuuEXo@krava/
>>
>> it looks like now it could wrong value if we don't disable it
>
> Sorry for the delay. I will add the disable part and send v5 soon.

I just sent v5 with the disable() callback, and a minor fix in 3/5. I haven't
figured out the best fallback mechanism. Will work on that in the coming week.

Thanks,
Song

\
 
 \ /
  Last update: 2021-04-25 23:47    [W:0.062 / U:0.604 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site