[PATCH v4 2/9] perf record --off-cpu: Add --off-cpu-thresh

[Howard Chu]

Add the --off-cpu-thresh argument to specify the off-cpu time threshold.
If the off-cpu time exceeds this threshold, dump the off-cpu data
directly.

Suggested-by: Ian Rogers <irogers@xxxxxxxxxx>
Signed-off-by: Howard Chu <howardchu95@xxxxxxxxx>
---
 tools/perf/builtin-record.c            | 26 ++++++++++++++++++++++++++
 tools/perf/util/bpf_off_cpu.c          |  2 ++
 tools/perf/util/bpf_skel/off_cpu.bpf.c |  2 ++
 tools/perf/util/record.h               |  1 +
 4 files changed, 31 insertions(+)

diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index 72345d1e54b0..60c6fe7b4804 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -3133,6 +3133,28 @@ static int record__parse_mmap_pages(const struct option *opt,
 	return ret;
 }
 
+static int record__parse_off_cpu_thresh(const struct option *opt,
+					const char *str,
+					int unset __maybe_unused)
+{
+	struct record_opts *opts = opt->value;
+	char *endptr;
+	u64 off_cpu_thresh;
+
+	if (!str)
+		return -EINVAL;
+
+	off_cpu_thresh = strtoul(str, &endptr, 10);
+
+	/* threshold isn't string "0", yet strtoull() returns 0, parsing failed. */
+	if (*endptr || (off_cpu_thresh == 0 && strcmp(str, "0")))
+		return -EINVAL;
+	else
+		opts->off_cpu_thresh = off_cpu_thresh;
+
+	return 0;
+}
+
 void __weak arch__add_leaf_frame_record_opts(struct record_opts *opts __maybe_unused)
 {
 }
@@ -3326,6 +3348,7 @@ static struct record record = {
 		.ctl_fd              = -1,
 		.ctl_fd_ack          = -1,
 		.synth               = PERF_SYNTH_ALL,
+		.off_cpu_thresh      = OFF_CPU_THRESH_DEFAULT,
 	},
 	.tool = {
 		.sample		= process_sample_event,
@@ -3560,6 +3583,9 @@ static struct option __record_options[] = {
 	OPT_BOOLEAN(0, "off-cpu", &record.off_cpu, "Enable off-cpu analysis"),
 	OPT_STRING(0, "setup-filter", &record.filter_action, "pin|unpin",
 		   "BPF filter action"),
+	OPT_CALLBACK(0, "off-cpu-thresh", &record.opts, "us",
+		     "Dump off-cpu samples if off-cpu time reaches this threshold. The unit is microseconds. (default: 500000)",
+		     record__parse_off_cpu_thresh),
 	OPT_END()
 };
 
diff --git a/tools/perf/util/bpf_off_cpu.c b/tools/perf/util/bpf_off_cpu.c
index 6af36142dc5a..1e0e454bfb5e 100644
--- a/tools/perf/util/bpf_off_cpu.c
+++ b/tools/perf/util/bpf_off_cpu.c
@@ -272,6 +272,8 @@ int off_cpu_prepare(struct evlist *evlist, struct target *target,
 		}
 	}
 
+	skel->bss->offcpu_thresh = opts->off_cpu_thresh * 1000ull;
+
 	err = off_cpu_bpf__attach(skel);
 	if (err) {
 		pr_err("Failed to attach off-cpu BPF skeleton\n");
diff --git a/tools/perf/util/bpf_skel/off_cpu.bpf.c b/tools/perf/util/bpf_skel/off_cpu.bpf.c
index d877a0a9731f..cca1b6990a57 100644
--- a/tools/perf/util/bpf_skel/off_cpu.bpf.c
+++ b/tools/perf/util/bpf_skel/off_cpu.bpf.c
@@ -96,6 +96,8 @@ const volatile bool uses_cgroup_v1 = false;
 
 int perf_subsys_id = -1;
 
+__u64 sample_id, sample_type, offcpu_thresh;
+
 /*
  * Old kernel used to call it task_struct->state and now it's '__state'.
  * Use BPF CO-RE "ignored suffix rule" to deal with it like below:
diff --git a/tools/perf/util/record.h b/tools/perf/util/record.h
index a6566134e09e..3c11416e6627 100644
--- a/tools/perf/util/record.h
+++ b/tools/perf/util/record.h
@@ -79,6 +79,7 @@ struct record_opts {
 	int	      synth;
 	int	      threads_spec;
 	const char    *threads_user_spec;
+	u64	      off_cpu_thresh;
 };
 
 extern const char * const *record_usage;
-- 
2.45.2