linux-kernel - Re: [PATCH] perf tools: Skip BPF sideband event for userspace profiling

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAP-5=fVh2tTD22sFrPt37OJAhOsggvt2AVwf45p_pxTa=pnVKA@mail.gmail.com>
Date: Wed, 26 Feb 2025 12:59:33 -0800
From: Ian Rogers <irogers@...gle.com>
To: Namhyung Kim <namhyung@...nel.org>
Cc: Arnaldo Carvalho de Melo <acme@...nel.org>, Kan Liang <kan.liang@...ux.intel.com>, 
	Jiri Olsa <jolsa@...nel.org>, Adrian Hunter <adrian.hunter@...el.com>, 
	Peter Zijlstra <peterz@...radead.org>, Ingo Molnar <mingo@...nel.org>, 
	LKML <linux-kernel@...r.kernel.org>, linux-perf-users@...r.kernel.org, 
	James Clark <james.clark@...aro.org>, Song Liu <song@...nel.org>
Subject: Re: [PATCH] perf tools: Skip BPF sideband event for userspace profiling

On Wed, Feb 26, 2025 at 12:30 PM Namhyung Kim <namhyung@...nel.org> wrote:
>
> The BPF sideband information is tracked using a separate thread and
> evlist.  But it's only useful for profiling kernel and we can skip it
> when users profile their application only.

nit: It may be worth noting that profiling an application implicitly
excludes the kernel samples.

> It seems it already fails to open the sideband event in that case.
> Let's remove the noise in the verbose output anyway.
>
> Cc: Song Liu <song@...nel.org>
> Signed-off-by: Namhyung Kim <namhyung@...nel.org>

Reviewed-by: Ian Rogers <irogers@...gle.com>

I wonder if the second evlist could be avoided similar to Howard's
off-CPU sample events?
I also wonder if we should make the evlist responsible for BPF and
dummy/sideband events. Having unnecessary events increases the list
size iterated over when creating sideband data, and so has a runtime
cost. Having the logic separated in places like builtin-top and
builtin-record feels suboptimal.

Thanks,
Ian

> ---
>  tools/perf/builtin-record.c |  3 +++
>  tools/perf/builtin-top.c    |  3 +++
>  tools/perf/util/evlist.c    | 14 ++++++++++++++
>  tools/perf/util/evlist.h    |  1 +
>  4 files changed, 21 insertions(+)
>
> diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
> index 0e45bd64185403ae..cc61f5d6c599039c 100644
> --- a/tools/perf/builtin-record.c
> +++ b/tools/perf/builtin-record.c
> @@ -2535,6 +2535,9 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)
>                 goto out_free_threads;
>         }
>
> +       if (!evlist__needs_bpf_sb_event(rec->evlist))
> +               opts->no_bpf_event = true;
> +
>         err = record__setup_sb_evlist(rec);
>         if (err)
>                 goto out_free_threads;
> diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
> index 6440b5c1757d92ce..c284a384542ff822 100644
> --- a/tools/perf/builtin-top.c
> +++ b/tools/perf/builtin-top.c
> @@ -1833,6 +1833,9 @@ int cmd_top(int argc, const char **argv)
>                 goto out_delete_evlist;
>         }
>
> +       if (!evlist__needs_bpf_sb_event(top.evlist))
> +               top.record_opts.no_bpf_event = true;
> +
>  #ifdef HAVE_LIBBPF_SUPPORT
>         if (!top.record_opts.no_bpf_event) {
>                 top.sb_evlist = evlist__new();
> diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
> index f0dd174e2debdbe8..43adf6b3d855771a 100644
> --- a/tools/perf/util/evlist.c
> +++ b/tools/perf/util/evlist.c
> @@ -2594,3 +2594,17 @@ bool evlist__has_bpf_output(struct evlist *evlist)
>
>         return false;
>  }
> +
> +bool evlist__needs_bpf_sb_event(struct evlist *evlist)
> +{
> +       struct evsel *evsel;
> +
> +       evlist__for_each_entry(evlist, evsel) {
> +               if (evsel__is_dummy_event(evsel))
> +                       continue;
> +               if (!evsel->core.attr.exclude_kernel)
> +                       return true;
> +       }
> +
> +       return false;
> +}
> diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
> index adddb1db1ad2b25d..edcbf1c10e92f0c4 100644
> --- a/tools/perf/util/evlist.h
> +++ b/tools/perf/util/evlist.h
> @@ -435,5 +435,6 @@ void evlist__check_mem_load_aux(struct evlist *evlist);
>  void evlist__warn_user_requested_cpus(struct evlist *evlist, const char *cpu_list);
>  void evlist__uniquify_name(struct evlist *evlist);
>  bool evlist__has_bpf_output(struct evlist *evlist);
> +bool evlist__needs_bpf_sb_event(struct evlist *evlist);
>
>  #endif /* __PERF_EVLIST_H */
> --
> 2.48.1.658.g4767266eb4-goog
>