mirror of
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
synced 2025-12-27 14:41:22 -05:00
Scan the software PMU first rather than last as it is the least likely
to fail the probe. Specifying the software PMU by name was enabled by
commit 9957d8c801 ("perf jevents: Add common software event
json"). For hardware events, add core PMU names when getting events to
probe so that not all PMUs are scanned. For example, when legacy
events support wildcards and for the event "cycles:u" on x86, we want
to only scan the "cpu" PMU and not all uncore PMUs for the event too.
Tested-by: Thomas Richter <tmricht@linux.ibm.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: James Clark <james.clark@linaro.org>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
209 lines
4.1 KiB
C
209 lines
4.1 KiB
C
/* SPDX-License-Identifier: GPL-2.0 */
|
|
|
|
#include "perf-sys.h"
|
|
#include "util/cloexec.h"
|
|
#include "util/evlist.h"
|
|
#include "util/evsel.h"
|
|
#include "util/parse-events.h"
|
|
#include "util/perf_api_probe.h"
|
|
#include <perf/cpumap.h>
|
|
#include <errno.h>
|
|
|
|
typedef void (*setup_probe_fn_t)(struct evsel *evsel);
|
|
|
|
static int perf_do_probe_api(setup_probe_fn_t fn, struct perf_cpu cpu, const char *str)
|
|
{
|
|
struct evlist *evlist;
|
|
struct evsel *evsel;
|
|
unsigned long flags = perf_event_open_cloexec_flag();
|
|
int err = -EAGAIN, fd;
|
|
static pid_t pid = -1;
|
|
|
|
evlist = evlist__new();
|
|
if (!evlist)
|
|
return -ENOMEM;
|
|
|
|
if (parse_event(evlist, str))
|
|
goto out_delete;
|
|
|
|
evsel = evlist__first(evlist);
|
|
|
|
while (1) {
|
|
fd = sys_perf_event_open(&evsel->core.attr, pid, cpu.cpu, -1, flags);
|
|
if (fd < 0) {
|
|
if (pid == -1 && errno == EACCES) {
|
|
pid = 0;
|
|
continue;
|
|
}
|
|
goto out_delete;
|
|
}
|
|
break;
|
|
}
|
|
close(fd);
|
|
|
|
fn(evsel);
|
|
|
|
fd = sys_perf_event_open(&evsel->core.attr, pid, cpu.cpu, -1, flags);
|
|
if (fd < 0) {
|
|
if (errno == EINVAL)
|
|
err = -EINVAL;
|
|
goto out_delete;
|
|
}
|
|
close(fd);
|
|
err = 0;
|
|
|
|
out_delete:
|
|
evlist__delete(evlist);
|
|
return err;
|
|
}
|
|
|
|
static bool perf_probe_api(setup_probe_fn_t fn)
|
|
{
|
|
struct perf_pmu *pmu;
|
|
struct perf_cpu_map *cpus;
|
|
struct perf_cpu cpu;
|
|
int ret = 0;
|
|
|
|
cpus = perf_cpu_map__new_online_cpus();
|
|
if (!cpus)
|
|
return false;
|
|
cpu = perf_cpu_map__cpu(cpus, 0);
|
|
perf_cpu_map__put(cpus);
|
|
|
|
ret = perf_do_probe_api(fn, cpu, "software/cpu-clock/u");
|
|
if (!ret)
|
|
return true;
|
|
|
|
pmu = perf_pmus__scan_core(/*pmu=*/NULL);
|
|
if (pmu) {
|
|
const char *try[] = {"cycles", "instructions", NULL};
|
|
char buf[256];
|
|
int i = 0;
|
|
|
|
while (ret == -EAGAIN && try[i]) {
|
|
snprintf(buf, sizeof(buf), "%s/%s/u", pmu->name, try[i++]);
|
|
ret = perf_do_probe_api(fn, cpu, buf);
|
|
if (!ret)
|
|
return true;
|
|
}
|
|
}
|
|
return false;
|
|
}
|
|
|
|
static void perf_probe_sample_identifier(struct evsel *evsel)
|
|
{
|
|
evsel->core.attr.sample_type |= PERF_SAMPLE_IDENTIFIER;
|
|
}
|
|
|
|
static void perf_probe_comm_exec(struct evsel *evsel)
|
|
{
|
|
evsel->core.attr.comm_exec = 1;
|
|
}
|
|
|
|
static void perf_probe_context_switch(struct evsel *evsel)
|
|
{
|
|
evsel->core.attr.context_switch = 1;
|
|
}
|
|
|
|
static void perf_probe_text_poke(struct evsel *evsel)
|
|
{
|
|
evsel->core.attr.text_poke = 1;
|
|
}
|
|
|
|
static void perf_probe_build_id(struct evsel *evsel)
|
|
{
|
|
evsel->core.attr.build_id = 1;
|
|
}
|
|
|
|
static void perf_probe_cgroup(struct evsel *evsel)
|
|
{
|
|
evsel->core.attr.cgroup = 1;
|
|
}
|
|
|
|
bool perf_can_sample_identifier(void)
|
|
{
|
|
return perf_probe_api(perf_probe_sample_identifier);
|
|
}
|
|
|
|
bool perf_can_comm_exec(void)
|
|
{
|
|
return perf_probe_api(perf_probe_comm_exec);
|
|
}
|
|
|
|
bool perf_can_record_switch_events(void)
|
|
{
|
|
return perf_probe_api(perf_probe_context_switch);
|
|
}
|
|
|
|
bool perf_can_record_text_poke_events(void)
|
|
{
|
|
return perf_probe_api(perf_probe_text_poke);
|
|
}
|
|
|
|
bool perf_can_record_cpu_wide(void)
|
|
{
|
|
struct perf_event_attr attr = {
|
|
.type = PERF_TYPE_SOFTWARE,
|
|
.config = PERF_COUNT_SW_CPU_CLOCK,
|
|
.exclude_kernel = 1,
|
|
};
|
|
struct perf_cpu_map *cpus;
|
|
struct perf_cpu cpu;
|
|
int fd;
|
|
|
|
cpus = perf_cpu_map__new_online_cpus();
|
|
if (!cpus)
|
|
return false;
|
|
|
|
cpu = perf_cpu_map__cpu(cpus, 0);
|
|
perf_cpu_map__put(cpus);
|
|
|
|
fd = sys_perf_event_open(&attr, -1, cpu.cpu, -1, 0);
|
|
if (fd < 0)
|
|
return false;
|
|
close(fd);
|
|
|
|
return true;
|
|
}
|
|
|
|
/*
|
|
* Architectures are expected to know if AUX area sampling is supported by the
|
|
* hardware. Here we check for kernel support.
|
|
*/
|
|
bool perf_can_aux_sample(void)
|
|
{
|
|
struct perf_event_attr attr = {
|
|
.size = sizeof(struct perf_event_attr),
|
|
.exclude_kernel = 1,
|
|
/*
|
|
* Non-zero value causes the kernel to calculate the effective
|
|
* attribute size up to that byte.
|
|
*/
|
|
.aux_sample_size = 1,
|
|
};
|
|
int fd;
|
|
|
|
fd = sys_perf_event_open(&attr, -1, 0, -1, 0);
|
|
/*
|
|
* If the kernel attribute is big enough to contain aux_sample_size
|
|
* then we assume that it is supported. We are relying on the kernel to
|
|
* validate the attribute size before anything else that could be wrong.
|
|
*/
|
|
if (fd < 0 && errno == E2BIG)
|
|
return false;
|
|
if (fd >= 0)
|
|
close(fd);
|
|
|
|
return true;
|
|
}
|
|
|
|
bool perf_can_record_build_id(void)
|
|
{
|
|
return perf_probe_api(perf_probe_build_id);
|
|
}
|
|
|
|
bool perf_can_record_cgroup(void)
|
|
{
|
|
return perf_probe_api(perf_probe_cgroup);
|
|
}
|