[PATCH v2] perf ftrace latency: Do not read trace files when BPF is used

[PATCH v2] perf ftrace latency: Do not read trace files when BPF is used

[Namhyung Kim]

I've realized that it didn't set up the tracing files when BPF is used
so poll() just returns immediately.  It ends up with calling poll()
unnecessarily in a loop.

BPF still needs the loop to wait for the target process exiting or a
signal from users.  Let's use a semaphore instead.

Suggested-by: Ian Rogers <irogers@google.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
---
 tools/perf/builtin-ftrace.c | 28 ++++++++++++++++++++++------
 1 file changed, 22 insertions(+), 6 deletions(-)

diff --git a/tools/perf/builtin-ftrace.c b/tools/perf/builtin-ftrace.c
index 4f881a40c311ae52..6017493ff179758b 100644
--- a/tools/perf/builtin-ftrace.c
+++ b/tools/perf/builtin-ftrace.c
@@ -6,23 +6,23 @@
  * Copyright (c) 2020  Changbin Du <changbin.du@gmail.com>, significant enhancement.
  */
 
-#include "builtin.h"
-
+#include <ctype.h>
 #include <errno.h>
-#include <unistd.h>
-#include <signal.h>
-#include <stdlib.h>
 #include <fcntl.h>
 #include <inttypes.h>
 #include <math.h>
 #include <poll.h>
-#include <ctype.h>
+#include <semaphore.h>
+#include <signal.h>
+#include <stdlib.h>
+#include <unistd.h>
 #include <linux/capability.h>
 #include <linux/err.h>
 #include <linux/string.h>
 #include <linux/zalloc.h>
 #include <sys/stat.h>
 
+#include "builtin.h"
 #include "debug.h"
 #include <subcmd/pager.h>
 #include <subcmd/parse-options.h>
@@ -45,6 +45,7 @@
 
 static volatile sig_atomic_t workload_exec_errno;
 static volatile sig_atomic_t done;
+static sem_t sig_sem;
 
 static struct stats latency_stats;  /* for tracepoints */
 
@@ -53,6 +54,7 @@ static char tracing_instance[PATH_MAX];	/* Trace instance directory */
 static void sig_handler(int sig __maybe_unused)
 {
 	done = true;
+	sem_post(&sig_sem);
 }
 
 /*
@@ -68,6 +70,7 @@ static void ftrace__workload_exec_failed_signal(int signo __maybe_unused,
 {
 	workload_exec_errno = info->si_value.sival_int;
 	done = true;
+	sem_post(&sig_sem);
 }
 
 static bool check_ftrace_capable(void)
@@ -1146,6 +1149,11 @@ static int __cmd_latency(struct perf_ftrace *ftrace)
 
 	line[0] = '\0';
 	while (!done) {
+		if (ftrace->target.use_bpf) {
+			sem_wait(&sig_sem);
+			break;
+		}
+
 		if (poll(&pollfd, 1, -1) < 0)
 			break;
 
@@ -1852,6 +1860,8 @@ int cmd_ftrace(int argc, const char **argv)
 	INIT_LIST_HEAD(&ftrace.nograph_funcs);
 	INIT_LIST_HEAD(&ftrace.event_pair);
 
+	sem_init(&sig_sem, 0, 0);
+
 	signal(SIGINT, sig_handler);
 	signal(SIGUSR1, sig_handler);
 	signal(SIGCHLD, sig_handler);
@@ -2015,5 +2025,11 @@ int cmd_ftrace(int argc, const char **argv)
 	delete_filter_func(&ftrace.nograph_funcs);
 	delete_filter_func(&ftrace.event_pair);
 
+	signal(SIGINT, SIG_DFL);
+	signal(SIGUSR1, SIG_DFL);
+	signal(SIGCHLD, SIG_DFL);
+	signal(SIGPIPE, SIG_DFL);
+
+	sem_destroy(&sig_sem);
 	return ret;
 }
-- 
2.55.0.897.gb25b4bd76c-goog

Re: [PATCH v2] perf ftrace latency: Do not read trace files when BPF is used

[sashiko-bot]

> I've realized that it didn't set up the tracing files when BPF is used
> so poll() just returns immediately.  It ends up with calling poll()
> unnecessarily in a loop.
> 
> BPF still needs the loop to wait for the target process exiting or a
> signal from users.  Let's use a semaphore instead.
> 
> Suggested-by: Ian Rogers <irogers@google.com>
> Cc: Steven Rostedt <rostedt@goodmis.org>
> Cc: Masami Hiramatsu <mhiramat@kernel.org>
> Signed-off-by: Namhyung Kim <namhyung@kernel.org>

Sashiko has reviewed this patch and found no issues. It looks great!

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260831063938.756610-1-namhyung@kernel.org?part=1

Re: [PATCH v2] perf ftrace latency: Do not read trace files when BPF is used

[Ian Rogers]

On Sun, Aug 30, 2026 at 11:39 PM Namhyung Kim <namhyung@kernel.org> wrote:
>
> I've realized that it didn't set up the tracing files when BPF is used
> so poll() just returns immediately.  It ends up with calling poll()
> unnecessarily in a loop.
>
> BPF still needs the loop to wait for the target process exiting or a
> signal from users.  Let's use a semaphore instead.
>
> Suggested-by: Ian Rogers <irogers@google.com>
> Cc: Steven Rostedt <rostedt@goodmis.org>
> Cc: Masami Hiramatsu <mhiramat@kernel.org>
> Signed-off-by: Namhyung Kim <namhyung@kernel.org>

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

Thanks,
Ian

> ---
>  tools/perf/builtin-ftrace.c | 28 ++++++++++++++++++++++------
>  1 file changed, 22 insertions(+), 6 deletions(-)
>
> diff --git a/tools/perf/builtin-ftrace.c b/tools/perf/builtin-ftrace.c
> index 4f881a40c311ae52..6017493ff179758b 100644
> --- a/tools/perf/builtin-ftrace.c
> +++ b/tools/perf/builtin-ftrace.c
> @@ -6,23 +6,23 @@
>   * Copyright (c) 2020  Changbin Du <changbin.du@gmail.com>, significant enhancement.
>   */
>
> -#include "builtin.h"
> -
> +#include <ctype.h>
>  #include <errno.h>
> -#include <unistd.h>
> -#include <signal.h>
> -#include <stdlib.h>
>  #include <fcntl.h>
>  #include <inttypes.h>
>  #include <math.h>
>  #include <poll.h>
> -#include <ctype.h>
> +#include <semaphore.h>
> +#include <signal.h>
> +#include <stdlib.h>
> +#include <unistd.h>
>  #include <linux/capability.h>
>  #include <linux/err.h>
>  #include <linux/string.h>
>  #include <linux/zalloc.h>
>  #include <sys/stat.h>
>
> +#include "builtin.h"
>  #include "debug.h"
>  #include <subcmd/pager.h>
>  #include <subcmd/parse-options.h>
> @@ -45,6 +45,7 @@
>
>  static volatile sig_atomic_t workload_exec_errno;
>  static volatile sig_atomic_t done;
> +static sem_t sig_sem;
>
>  static struct stats latency_stats;  /* for tracepoints */
>
> @@ -53,6 +54,7 @@ static char tracing_instance[PATH_MAX];       /* Trace instance directory */
>  static void sig_handler(int sig __maybe_unused)
>  {
>         done = true;
> +       sem_post(&sig_sem);
>  }
>
>  /*
> @@ -68,6 +70,7 @@ static void ftrace__workload_exec_failed_signal(int signo __maybe_unused,
>  {
>         workload_exec_errno = info->si_value.sival_int;
>         done = true;
> +       sem_post(&sig_sem);
>  }
>
>  static bool check_ftrace_capable(void)
> @@ -1146,6 +1149,11 @@ static int __cmd_latency(struct perf_ftrace *ftrace)
>
>         line[0] = '\0';
>         while (!done) {
> +               if (ftrace->target.use_bpf) {
> +                       sem_wait(&sig_sem);
> +                       break;
> +               }
> +
>                 if (poll(&pollfd, 1, -1) < 0)
>                         break;
>
> @@ -1852,6 +1860,8 @@ int cmd_ftrace(int argc, const char **argv)
>         INIT_LIST_HEAD(&ftrace.nograph_funcs);
>         INIT_LIST_HEAD(&ftrace.event_pair);
>
> +       sem_init(&sig_sem, 0, 0);
> +
>         signal(SIGINT, sig_handler);
>         signal(SIGUSR1, sig_handler);
>         signal(SIGCHLD, sig_handler);
> @@ -2015,5 +2025,11 @@ int cmd_ftrace(int argc, const char **argv)
>         delete_filter_func(&ftrace.nograph_funcs);
>         delete_filter_func(&ftrace.event_pair);
>
> +       signal(SIGINT, SIG_DFL);
> +       signal(SIGUSR1, SIG_DFL);
> +       signal(SIGCHLD, SIG_DFL);
> +       signal(SIGPIPE, SIG_DFL);
> +
> +       sem_destroy(&sig_sem);
>         return ret;
>  }
> --
> 2.55.0.897.gb25b4bd76c-goog
>

Re: [PATCH v2] perf ftrace latency: Do not read trace files when BPF is used

[Arnaldo Carvalho de Melo]

On Fri, Sep 04, 2026 at 10:36:08AM -0700, Ian Rogers wrote:
> On Sun, Aug 30, 2026 at 11:39 PM Namhyung Kim <namhyung@kernel.org> wrote:
> >
> > I've realized that it didn't set up the tracing files when BPF is used
> > so poll() just returns immediately.  It ends up with calling poll()
> > unnecessarily in a loop.
> >
> > BPF still needs the loop to wait for the target process exiting or a
> > signal from users.  Let's use a semaphore instead.
> >
> > Suggested-by: Ian Rogers <irogers@google.com>
> > Cc: Steven Rostedt <rostedt@goodmis.org>
> > Cc: Masami Hiramatsu <mhiramat@kernel.org>
> > Signed-off-by: Namhyung Kim <namhyung@kernel.org>
> 
> Reviewed-by: Ian Rogers <irogers@google.com>

Thanks, applied to perf-tools-next, for v7.4.

- Arnaldo