From: Namhyung Kim <namhyung@kernel.org>
To: Levi Yun <yeoreum.yun@arm.com>
Cc: peterz@infradead.org, mingo@redhat.com, acme@kernel.org,
mark.rutland@arm.com, alexander.shishkin@linux.intel.com,
jolsa@kernel.org, irogers@google.com, james.clark@linaro.org,
asmadeus@codewreck.org, linux-perf-users@vger.kernel.org,
linux-kernel@vger.kernel.org, nd@arm.com
Subject: Re: [PATCH v4 1/2] perf stat: Close cork_fd when create_perf_stat_counter() failed
Date: Tue, 24 Sep 2024 12:29:10 -0700 [thread overview]
Message-ID: <ZvMTBryjl5xTYr8W@google.com> (raw)
In-Reply-To: <20240913135907.648734-2-yeoreum.yun@arm.com>
Hello,
On Fri, Sep 13, 2024 at 02:59:06PM +0100, Levi Yun wrote:
> When create_perf_stat_counter() failed, it doesn't close workload.cork_fd
> open in evlist__prepare_workload(). This could make too many open file
> error while __run_perf_stat() repeats.
>
> Introduce evlist__cancel_workload to close workload.cork_fd and
> wait workload.child_pid until exit to clear child process
> when create_perf_stat_counter() is failed with COUNTER_FATAL.
>
> Signed-off-by: Levi Yun <yeoreum.yun@arm.com>
> Reviewed-by: James Clark <james.clark@linaro.org>
> ---
> tools/perf/builtin-stat.c | 4 ++++
> tools/perf/util/evlist.c | 14 +++++++++++++-
> tools/perf/util/evlist.h | 1 +
> 3 files changed, 18 insertions(+), 1 deletion(-)
>
> diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
> index 661832756a24..954eb37ce7b8 100644
> --- a/tools/perf/builtin-stat.c
> +++ b/tools/perf/builtin-stat.c
> @@ -763,6 +763,8 @@ static int __run_perf_stat(int argc, const char **argv, int run_idx)
>
> switch (stat_handle_error(counter)) {
> case COUNTER_FATAL:
> + if (forks)
> + evlist__cancel_workload(evsel_list);
> return -1;
> case COUNTER_RETRY:
> goto try_again;
> @@ -804,6 +806,8 @@ static int __run_perf_stat(int argc, const char **argv, int run_idx)
>
> switch (stat_handle_error(counter)) {
> case COUNTER_FATAL:
> + if (forks)
> + evlist__cancel_workload(evsel_list);
> return -1;
I don't think you covered all the places it can return before starting
the workload. You'd better have an error handling code at the end with
a label and then goto there.
Thanks,
Namhyung
> case COUNTER_RETRY:
> goto try_again_reset;
> diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
> index 3a719edafc7a..51a221679c92 100644
> --- a/tools/perf/util/evlist.c
> +++ b/tools/perf/util/evlist.c
> @@ -46,6 +46,7 @@
> #include <sys/mman.h>
> #include <sys/prctl.h>
> #include <sys/timerfd.h>
> +#include <sys/wait.h>
>
> #include <linux/bitops.h>
> #include <linux/hash.h>
> @@ -1465,7 +1466,7 @@ int evlist__prepare_workload(struct evlist *evlist, struct target *target, const
> * For cancelling the workload without actually running it,
> * the parent will just close workload.cork_fd, without writing
> * anything, i.e. read will return zero and we just exit()
> - * here.
> + * here (See evlist__cancel_workload()).
> */
> if (ret != 1) {
> if (ret == -1)
> @@ -1546,6 +1547,17 @@ int evlist__start_workload(struct evlist *evlist)
> return 0;
> }
>
> +void evlist__cancel_workload(struct evlist *evlist)
> +{
> + int status;
> +
> + if (evlist->workload.cork_fd > 0) {
> + close(evlist->workload.cork_fd);
> + evlist->workload.cork_fd = -1;
> + waitpid(evlist->workload.pid, &status, WNOHANG);
> + }
> +}
> +
> int evlist__parse_sample(struct evlist *evlist, union perf_event *event, struct perf_sample *sample)
> {
> struct evsel *evsel = evlist__event2evsel(evlist, event);
> diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
> index cb91dc9117a2..12f929ffdf92 100644
> --- a/tools/perf/util/evlist.h
> +++ b/tools/perf/util/evlist.h
> @@ -184,6 +184,7 @@ int evlist__prepare_workload(struct evlist *evlist, struct target *target,
> const char *argv[], bool pipe_output,
> void (*exec_error)(int signo, siginfo_t *info, void *ucontext));
> int evlist__start_workload(struct evlist *evlist);
> +void evlist__cancel_workload(struct evlist *evlist);
>
> struct option;
>
> --
> LEVI:{C3F47F37-75D8-414A-A8BA-3980EC8A46D7}
>
next prev parent reply other threads:[~2024-09-24 19:29 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-09-13 13:59 [PATCH v4 0/2] Minor fixes error handling of perf stat Levi Yun
2024-09-13 13:59 ` [PATCH v4 1/2] perf stat: Close cork_fd when create_perf_stat_counter() failed Levi Yun
2024-09-24 19:29 ` Namhyung Kim [this message]
2024-09-25 10:38 ` Yeo Reum Yun
2024-09-13 13:59 ` [PATCH v4 2/2] perf stat: Stop repeating when ref_perf_stat() returns -1 Levi Yun
2024-09-24 7:13 ` [PATCH v4 0/2] Minor fixes error handling of perf stat Yeo Reum Yun
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=ZvMTBryjl5xTYr8W@google.com \
--to=namhyung@kernel.org \
--cc=acme@kernel.org \
--cc=alexander.shishkin@linux.intel.com \
--cc=asmadeus@codewreck.org \
--cc=irogers@google.com \
--cc=james.clark@linaro.org \
--cc=jolsa@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-perf-users@vger.kernel.org \
--cc=mark.rutland@arm.com \
--cc=mingo@redhat.com \
--cc=nd@arm.com \
--cc=peterz@infradead.org \
--cc=yeoreum.yun@arm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.