Received-SPF: pass (google.com: domain of linux-kernel+bounces-99898-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Precedence: bulk
MIME-Version: 1.0
References: <20240302001139.604829-1-weilin.wang@intel.com> <20240302001139.604829-5-weilin.wang@intel.com>
In-Reply-To: <20240302001139.604829-5-weilin.wang@intel.com>
From: Namhyung Kim <namhyung@kernel.org>
Date: Mon, 11 Mar 2024 23:47:44 -0700
Message-ID: <CAM9d7ciVnWkaLMwHcJwC58KYDTW9t4SQevEGnqKcNZmJ7XF57Q@mail.gmail.com>
Subject: Re: [RFC PATCH v3 4/6] perf stat: Create another thread for sample
 data processing
To: weilin.wang@intel.com
Cc: Ian Rogers <irogers@google.com>, Peter Zijlstra <peterz@infradead.org>, 
	Ingo Molnar <mingo@redhat.com>, Arnaldo Carvalho de Melo <acme@kernel.org>, 
	Alexander Shishkin <alexander.shishkin@linux.intel.com>, Jiri Olsa <jolsa@kernel.org>, 
	Adrian Hunter <adrian.hunter@intel.com>, Kan Liang <kan.liang@linux.intel.com>, 
	linux-perf-users@vger.kernel.org, linux-kernel@vger.kernel.org, 
	Perry Taylor <perry.taylor@intel.com>, Samantha Alt <samantha.alt@intel.com>, 
	Caleb Biggers <caleb.biggers@intel.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Fri, Mar 1, 2024 at 4:11=E2=80=AFPM <weilin.wang@intel.com> wrote:
>
> From: Weilin Wang <weilin.wang@intel.com>
>
> Another thread is required to synchronize between perf stat and perf reco=
rd
> when we pass data through pipe.
>
> Signed-off-by: Weilin Wang <weilin.wang@intel.com>
> ---
>  tools/perf/builtin-stat.c | 58 +++++++++++++++++++++++----------------
>  1 file changed, 34 insertions(+), 24 deletions(-)
>
> diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
> index 3e5865572266..2372175d3408 100644
> --- a/tools/perf/builtin-stat.c
> +++ b/tools/perf/builtin-stat.c
> @@ -777,7 +777,6 @@ static int process_sample_event(struct perf_tool *too=
l,
>         const char *evname;
>         struct tpebs_retire_lat *t;
>
> -       pr_debug("entering function %s\n ", __func__);
>         evname =3D evsel__name(evsel);
>
>         pr_debug("[%03d] ", sample->cpu);
> @@ -808,9 +807,9 @@ static int process_feature_event(struct perf_session =
*session,
>         return 0;
>  }
>
> -static int __cmd_script(struct child_process *cmd __maybe_unused)
> +static void *__cmd_script(void *arg __maybe_unused)
>  {
> -       int err =3D 0;
> +       struct child_process *cmd =3D arg;
>         struct perf_session *session;
>         struct perf_data data =3D {
>                 .mode =3D PERF_DATA_MODE_READ,
> @@ -826,29 +825,15 @@ static int __cmd_script(struct child_process *cmd _=
_maybe_unused)
>                 .attr            =3D perf_event__process_attr,
>                 },
>         };
> -       struct tpebs_event *e;
> -
> -       list_for_each_entry(e, &stat_config.tpebs_events, nd) {
> -               struct tpebs_retire_lat *new =3D malloc(sizeof(struct tpe=
bs_retire_lat));
> -
> -               if (!new)
> -                       return -1;
> -               new->name =3D strdup(e->name);
> -               new->tpebs_name =3D strdup(e->tpebs_name);
> -               new->count =3D 0;
> -               new->sum =3D 0;
> -               list_add_tail(&new->nd, &stat_config.tpebs_results);
> -       }
>
> -       kill(cmd->pid, SIGTERM);
>         session =3D perf_session__new(&data, &script.tool);
>         if (IS_ERR(session))
> -               return PTR_ERR(session);
> +               return NULL;
>         script.session =3D session;
> -       err =3D perf_session__process_events(session);
> +       perf_session__process_events(session);
>         perf_session__delete(session);
>
> -       return err;
> +       return NULL;
>  }
>
>  static int __run_perf_stat(int argc, const char **argv, int run_idx)
> @@ -868,15 +853,37 @@ static int __run_perf_stat(int argc, const char **a=
rgv, int run_idx)
>         int err;
>         bool second_pass =3D false;
>         struct child_process cmd;
> +       pthread_t thread_script;
>
>         //Prepare perf record for sampling event retire_latency before fo=
rk and prepare workload
>         if (stat_config.tpebs_event_size > 0) {
>                 int ret;
>
> +               struct tpebs_event *e;
>                 pr_debug("perf stat pid =3D %d\n", getpid());
>                 ret =3D prepare_perf_record(&cmd);
>                 if (ret)
>                         return ret;
> +
> +               list_for_each_entry(e, &stat_config.tpebs_events, nd) {
> +                       struct tpebs_retire_lat *new =3D malloc(sizeof(st=
ruct tpebs_retire_lat));
> +
> +                       if (!new)
> +                               return -1;

This would leave the record command.  I think you'd better
do this before prepare_perf_record().


> +                       new->name =3D strdup(e->name);
> +                       new->tpebs_name =3D strdup(e->tpebs_name);
> +                       new->count =3D 0;
> +                       new->sum =3D 0;
> +                       list_add_tail(&new->nd, &stat_config.tpebs_result=
s);
> +               }
> +
> +               if (pthread_create(&thread_script, NULL, __cmd_script, &c=
md)) {
> +                       kill(cmd.pid, SIGTERM);
> +                       close(cmd.out);
> +                       pr_err("Could not create thread to process sample=
 data.\n");
> +                       return -1;
> +               }
> +               sleep(2);

Why?

Thanks,
Namhyung


>         }
>
>         if (forks) {
> @@ -1087,12 +1094,15 @@ static int __run_perf_stat(int argc, const char *=
*argv, int run_idx)
>         if (stat_config.tpebs_event_size > 0) {
>                 int ret;
>
> -               pr_debug("pid =3D %d\n", getpid());
> -               pr_debug("cmd.pid =3D %d\n", cmd.pid);
> +               pr_debug("Workload finished, finishing record\n");
> +               pr_debug("Perf stat pid =3D %d, Perf record pid =3D %d\n"=
, getpid(), cmd.pid);
>
> -               ret =3D __cmd_script(&cmd);
> +               kill(cmd.pid, SIGTERM);
> +               pthread_join(thread_script, NULL);
>                 close(cmd.out);
> -               pr_debug("%d\n", ret);
> +               ret =3D finish_command(&cmd);
> +               if (ret !=3D -ERR_RUN_COMMAND_WAITPID_SIGNAL)
> +                       return ret;
>         }
>
>         if (stat_config.walltime_run_table)
> --
> 2.43.0
>