[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20240726102826.787004-4-howardchu95@gmail.com>
Date: Fri, 26 Jul 2024 18:28:24 +0800
From: Howard Chu <howardchu95@...il.com>
To: namhyung@...nel.org
Cc: irogers@...gle.com,
acme@...nel.org,
adrian.hunter@...el.com,
jolsa@...nel.org,
kan.liang@...ux.intel.com,
linux-perf-users@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: [PATCH v3 3/5] perf record off-cpu: processing of embedded sample
In the first evsel__parse_sample, we parse bpf output as raw samples.
After getting the raw_data, parse it the second time to get the embedded
samples.
Also, because we rely on evlist to dump direct off-cpu samples, if we
evsel__open() a bpf output event on a specific pid, it will fail. So pid = -1 for
every bpf output event.
Signed-off-by: Howard Chu <howardchu95@...il.com>
Suggested-by: Ian Rogers <irogers@...gle.com>
Suggested-by: Arnaldo Carvalho de Melo <acme@...nel.org>
---
tools/perf/builtin-script.c | 2 +-
tools/perf/util/evsel.c | 16 ++++++++++++++--
tools/perf/util/evsel.h | 13 +++++++++++++
tools/perf/util/session.c | 23 ++++++++++++++++++++++-
4 files changed, 50 insertions(+), 4 deletions(-)
diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c
index c16224b1fef3..c142fb44817f 100644
--- a/tools/perf/builtin-script.c
+++ b/tools/perf/builtin-script.c
@@ -653,7 +653,7 @@ static int perf_session__check_output_opt(struct perf_session *session)
evlist__for_each_entry(session->evlist, evsel) {
not_pipe = true;
- if (evsel__has_callchain(evsel)) {
+ if (evsel__has_callchain(evsel) || evsel__embed_has_callchain(evsel)) {
use_callchain = true;
break;
}
diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index bc603193c477..6c64ec475b80 100644
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -298,6 +298,7 @@ void evsel__init(struct evsel *evsel,
evsel->pmu_name = NULL;
evsel->group_pmu_name = NULL;
evsel->skippable = false;
+ evsel->sample_type_embed = 0;
}
struct evsel *evsel__new_idx(struct perf_event_attr *attr, int idx)
@@ -440,6 +441,7 @@ struct evsel *evsel__clone(struct evsel *orig)
evsel->weak_group = orig->weak_group;
evsel->use_config_name = orig->use_config_name;
evsel->pmu = orig->pmu;
+ evsel->sample_type_embed = orig->sample_type_embed;
if (evsel__copy_config_terms(evsel, orig) < 0)
goto out_err;
@@ -2282,6 +2284,10 @@ static int evsel__open_cpu(struct evsel *evsel, struct perf_cpu_map *cpus,
test_attr__ready();
+ /* BPF output event can only be system-wide, off-cpu filters tasks in BPF */
+ if (evsel__is_bpf_output(evsel))
+ pid = -1;
+
/* Debug message used by test scripts */
pr_debug2_peo("sys_perf_event_open: pid %d cpu %d group_fd %d flags %#lx",
pid, perf_cpu_map__cpu(cpus, idx).cpu, group_fd, evsel->open_flags);
@@ -2592,6 +2598,14 @@ int evsel__parse_sample(struct evsel *evsel, union perf_event *event,
*/
union u64_swap u;
+ array = event->sample.array;
+
+ /* use raw_data passed in to read embedded data */
+ if (evsel__has_embed(evsel) && evsel__is_bpf_output(evsel) && data->raw_data) {
+ array = data->raw_data;
+ type = evsel->sample_type_embed;
+ }
+
memset(data, 0, sizeof(*data));
data->cpu = data->pid = data->tid = -1;
data->stream_id = data->id = data->time = -1ULL;
@@ -2607,8 +2621,6 @@ int evsel__parse_sample(struct evsel *evsel, union perf_event *event,
return perf_evsel__parse_id_sample(evsel, event, data);
}
- array = event->sample.array;
-
if (perf_event__check_size(event, evsel->sample_size))
return -EFAULT;
diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h
index 80b5f6dd868e..0d25e82c6154 100644
--- a/tools/perf/util/evsel.h
+++ b/tools/perf/util/evsel.h
@@ -184,6 +184,9 @@ struct evsel {
};
/* Is the tool's fd for /proc/pid/stat or /proc/stat. */
bool pid_stat;
+
+ /* for samples embedded in BPF output */
+ __u64 sample_type_embed;
};
struct perf_missing_features {
@@ -469,6 +472,11 @@ static inline bool evsel__is_bpf_output(struct evsel *evsel)
return evsel__match(evsel, SOFTWARE, SW_BPF_OUTPUT);
}
+static inline bool evsel__has_embed(struct evsel *evsel)
+{
+ return evsel->sample_type_embed != 0;
+}
+
static inline bool evsel__is_clock(const struct evsel *evsel)
{
return evsel__match(evsel, SOFTWARE, SW_CPU_CLOCK) ||
@@ -525,6 +533,11 @@ static inline bool evsel__has_callchain(const struct evsel *evsel)
evsel->synth_sample_type & PERF_SAMPLE_CALLCHAIN;
}
+static inline bool evsel__embed_has_callchain(const struct evsel *evsel)
+{
+ return evsel->sample_type_embed & PERF_SAMPLE_CALLCHAIN;
+}
+
static inline bool evsel__has_br_stack(const struct evsel *evsel)
{
/*
diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
index 5596bed1b8c8..d87d85a3e21e 100644
--- a/tools/perf/util/session.c
+++ b/tools/perf/util/session.c
@@ -1528,6 +1528,23 @@ static int evlist__deliver_sample(struct evlist *evlist, struct perf_tool *tool,
/* We know evsel != NULL. */
u64 sample_type = evsel->core.attr.sample_type;
u64 read_format = evsel->core.attr.read_format;
+ struct perf_sample sample_embed;
+
+ if (evsel__is_bpf_output(evsel) && evsel__has_embed(evsel) &&
+ sample->raw_data && sample->raw_size - sizeof(__u32) > sizeof(struct perf_event_header)) {
+ int err;
+
+ sample_embed.raw_data = sample->raw_data;
+
+ err = evsel__parse_sample(evsel, event, &sample_embed);
+ if (err) {
+ pr_err("Can't parse BPF-embedded sample, err = %d\n", err);
+ return err;
+ }
+
+ sample_type = evsel->sample_type_embed;
+ sample = &sample_embed;
+ }
/* Standard sample delivery. */
if (!(sample_type & PERF_SAMPLE_READ))
@@ -1639,8 +1656,12 @@ static int perf_session__deliver_event(struct perf_session *session,
const char *file_path)
{
struct perf_sample sample;
- int ret = evlist__parse_sample(session->evlist, event, &sample);
+ int ret;
+
+ /* set to NULL so we don't accidentally parse BPF-embedded sample */
+ sample.raw_data = NULL;
+ ret = evlist__parse_sample(session->evlist, event, &sample);
if (ret) {
pr_err("Can't parse sample, err = %d\n", ret);
return ret;
--
2.45.2
Powered by blists - more mailing lists