[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20240807153843.3231451-8-howardchu95@gmail.com>
Date: Wed, 7 Aug 2024 23:38:41 +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 v4 7/9] perf record --off-cpu: Parse BPF output embedded data
Add a sample_type_embed member to the struct evsel, along with a couple
of helper functions.
In session.c, we parse BPF output embedded samples in a two-step
process.
Initial Parsing: Treat the sample as a regular BPF-output event.
Secondary Parsing: Extract data from raw_data and parse it according to
the sample_type_embed specification. Since the second step relies on the
raw_data obtained in the first step, we must avoid zero-initializing the
sample data after the first step.
Suggested-by: Ian Rogers <irogers@...gle.com>
Suggested-by: Arnaldo Carvalho de Melo <acme@...nel.org>
Signed-off-by: Howard Chu <howardchu95@...il.com>
---
tools/perf/builtin-script.c | 5 +++--
tools/perf/util/evsel.c | 36 ++++++++++++++++++++++++++++--------
tools/perf/util/evsel.h | 13 +++++++++++++
tools/perf/util/session.c | 16 +++++++++++++++-
4 files changed, 59 insertions(+), 11 deletions(-)
diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c
index c16224b1fef3..4145e61b8d38 100644
--- a/tools/perf/builtin-script.c
+++ b/tools/perf/builtin-script.c
@@ -653,7 +653,8 @@ 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;
}
@@ -2295,7 +2296,7 @@ static void process_event(struct perf_script *script,
else if (PRINT_FIELD(BRSTACKOFF))
perf_sample__fprintf_brstackoff(sample, thread, attr, fp);
- if (evsel__is_bpf_output(evsel) && PRINT_FIELD(BPF_OUTPUT))
+ if (evsel__is_bpf_output(evsel) && !evsel__has_embed(evsel) && PRINT_FIELD(BPF_OUTPUT))
perf_sample__fprintf_bpf_output(sample, fp);
perf_sample__fprintf_insn(sample, attr, thread, machine, fp, al);
diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index c1aac09ad0a5..a5a5b3ff16e1 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;
@@ -2589,6 +2591,7 @@ int evsel__parse_sample(struct evsel *evsel, union perf_event *event,
u16 max_size = event->header.size;
const void *endp = (void *)event + max_size;
u64 sz;
+ bool ip_in_callchain = false;
/*
* used for cross-endian analysis. See git commit 65014ab3
@@ -2596,14 +2599,27 @@ int evsel__parse_sample(struct evsel *evsel, union perf_event *event,
*/
union u64_swap u;
- memset(data, 0, sizeof(*data));
- data->cpu = data->pid = data->tid = -1;
- data->stream_id = data->id = data->time = -1ULL;
- data->period = evsel->core.attr.sample_period;
- data->cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK;
- data->misc = event->header.misc;
- data->data_src = PERF_MEM_DATA_SRC_NONE;
- data->vcpu = -1;
+ /*
+ * For sample data embedded in BPF output, don't clear the sample we read in the first pass,
+ * and read the embedded data from raw_data in the second pass.
+ */
+ if (evsel__has_embed(evsel) && evsel__is_bpf_output(evsel) && data->raw_data) {
+ type = evsel->sample_type_embed;
+ array = data->raw_data;
+ /* if callchain is embedded in BPF output, so is ip */
+ if (type & PERF_SAMPLE_CALLCHAIN)
+ ip_in_callchain = true;
+ } else { /* for normal samples, clear to zero before reading */
+ array = event->sample.array;
+ memset(data, 0, sizeof(*data));
+ data->cpu = data->pid = data->tid = -1;
+ data->stream_id = data->id = data->time = -1ULL;
+ data->period = evsel->core.attr.sample_period;
+ data->cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK;
+ data->misc = event->header.misc;
+ data->data_src = PERF_MEM_DATA_SRC_NONE;
+ data->vcpu = -1;
+ }
if (event->header.type != PERF_RECORD_SAMPLE) {
if (!evsel->core.attr.sample_id_all)
@@ -2732,6 +2748,10 @@ int evsel__parse_sample(struct evsel *evsel, union perf_event *event,
data->callchain = (struct ip_callchain *)array++;
if (data->callchain->nr > max_callchain_nr)
return -EFAULT;
+
+ if (ip_in_callchain && data->callchain->nr > 1)
+ data->ip = data->callchain->ips[1];
+
sz = data->callchain->nr * sizeof(u64);
OVERFLOW_CHECK(array, sz, max_size);
array = (void *)array + sz;
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..18aa5e0263b7 100644
--- a/tools/perf/util/session.c
+++ b/tools/perf/util/session.c
@@ -1529,6 +1529,16 @@ static int evlist__deliver_sample(struct evlist *evlist, struct perf_tool *tool,
u64 sample_type = evsel->core.attr.sample_type;
u64 read_format = evsel->core.attr.read_format;
+ /* parse sample the second time to get embedded data from raw_data */
+ if (evsel__is_bpf_output(evsel) && evsel__has_embed(evsel) && sample->raw_data) {
+ int err = evsel__parse_sample(evsel, event, sample);
+
+ if (err) {
+ pr_err("failed to parse BPF ouput embedded data, err = %d\n", err);
+ return err;
+ }
+ }
+
/* Standard sample delivery. */
if (!(sample_type & PERF_SAMPLE_READ))
return tool->sample(tool, event, sample, evsel, machine);
@@ -1639,8 +1649,12 @@ 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;
+
+ /* avoid reading embedded data from raw_data */
+ 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