perf tools: Move routines that probe for perf API features to separate file

Trying to disentangle this a bit further, unfortunately it uses
parse_events(), its interesting to have it separated anyway, so do it.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
This commit is contained in:
Arnaldo Carvalho de Melo 2020-05-05 11:49:08 -03:00
parent 354575c00d
commit 40c7d2460e
11 changed files with 186 additions and 158 deletions

View File

@ -23,6 +23,7 @@
#include "../../util/event.h"
#include "../../util/evlist.h"
#include "../../util/evsel.h"
#include "../../util/perf_api_probe.h"
#include "../../util/evsel_config.h"
#include "../../util/pmu.h"
#include "../../util/cs-etm.h"

View File

@ -25,6 +25,7 @@
#include "../../../util/pmu.h"
#include "../../../util/debug.h"
#include "../../../util/auxtrace.h"
#include "../../../util/perf_api_probe.h"
#include "../../../util/record.h"
#include "../../../util/target.h"
#include "../../../util/tsc.h"

View File

@ -34,6 +34,7 @@
#include "util/tsc.h"
#include "util/parse-branch-options.h"
#include "util/parse-regs-options.h"
#include "util/perf_api_probe.h"
#include "util/llvm-utils.h"
#include "util/bpf-loader.h"
#include "util/trigger.h"

View File

@ -88,6 +88,7 @@ perf-y += counts.o
perf-y += stat.o
perf-y += stat-shadow.o
perf-y += stat-display.o
perf-y += perf_api_probe.o
perf-y += record.o
perf-y += srcline.o
perf-y += srccode.o

View File

@ -33,6 +33,7 @@
#include "evsel.h"
#include "evsel_config.h"
#include "symbol.h"
#include "util/perf_api_probe.h"
#include "util/synthetic-events.h"
#include "thread_map.h"
#include "asm/bug.h"

View File

@ -23,6 +23,7 @@
#include "asm/bug.h"
#include "bpf-event.h"
#include "util/string2.h"
#include "util/perf_api_probe.h"
#include <signal.h>
#include <unistd.h>
#include <sched.h>

View File

@ -173,10 +173,6 @@ void evlist__close(struct evlist *evlist);
struct callchain_param;
void perf_evlist__set_id_pos(struct evlist *evlist);
bool perf_can_sample_identifier(void);
bool perf_can_record_switch_events(void);
bool perf_can_record_cpu_wide(void);
bool perf_can_aux_sample(void);
void perf_evlist__config(struct evlist *evlist, struct record_opts *opts,
struct callchain_param *callchain);
int record_opts__config(struct record_opts *opts);

View File

@ -33,6 +33,7 @@
#include "tsc.h"
#include "intel-pt.h"
#include "config.h"
#include "util/perf_api_probe.h"
#include "util/synthetic-events.h"
#include "time-utils.h"

View File

@ -0,0 +1,164 @@
/* 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, int 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_events(evlist, str, NULL))
goto out_delete;
evsel = evlist__first(evlist);
while (1) {
fd = sys_perf_event_open(&evsel->core.attr, pid, 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, -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)
{
const char *try[] = {"cycles:u", "instructions:u", "cpu-clock:u", NULL};
struct perf_cpu_map *cpus;
int cpu, ret, i = 0;
cpus = perf_cpu_map__new(NULL);
if (!cpus)
return false;
cpu = cpus->map[0];
perf_cpu_map__put(cpus);
do {
ret = perf_do_probe_api(fn, cpu, try[i++]);
if (!ret)
return true;
} while (ret == -EAGAIN && try[i]);
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;
}
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_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;
int cpu, fd;
cpus = perf_cpu_map__new(NULL);
if (!cpus)
return false;
cpu = cpus->map[0];
perf_cpu_map__put(cpus);
fd = sys_perf_event_open(&attr, -1, 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;
}

View File

@ -0,0 +1,14 @@
/* SPDX-License-Identifier: GPL-2.0 */
#ifndef __PERF_API_PROBE_H
#define __PERF_API_PROBE_H
#include <stdbool.h>
bool perf_can_aux_sample(void);
bool perf_can_comm_exec(void);
bool perf_can_record_cpu_wide(void);
bool perf_can_record_switch_events(void);
bool perf_can_sample_identifier(void);
#endif // __PERF_API_PROBE_H

View File

@ -10,163 +10,10 @@
#include <subcmd/parse-options.h>
#include <perf/cpumap.h>
#include "cloexec.h"
#include "util/perf_api_probe.h"
#include "record.h"
#include "../perf-sys.h"
typedef void (*setup_probe_fn_t)(struct evsel *evsel);
static int perf_do_probe_api(setup_probe_fn_t fn, int 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_events(evlist, str, NULL))
goto out_delete;
evsel = evlist__first(evlist);
while (1) {
fd = sys_perf_event_open(&evsel->core.attr, pid, 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, -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)
{
const char *try[] = {"cycles:u", "instructions:u", "cpu-clock:u", NULL};
struct perf_cpu_map *cpus;
int cpu, ret, i = 0;
cpus = perf_cpu_map__new(NULL);
if (!cpus)
return false;
cpu = cpus->map[0];
perf_cpu_map__put(cpus);
do {
ret = perf_do_probe_api(fn, cpu, try[i++]);
if (!ret)
return true;
} while (ret == -EAGAIN && try[i]);
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;
}
bool perf_can_sample_identifier(void)
{
return perf_probe_api(perf_probe_sample_identifier);
}
static 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_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;
int cpu, fd;
cpus = perf_cpu_map__new(NULL);
if (!cpus)
return false;
cpu = cpus->map[0];
perf_cpu_map__put(cpus);
fd = sys_perf_event_open(&attr, -1, 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;
}
/*
* perf_evsel__config_leader_sampling() uses special rules for leader sampling.
* However, if the leader is an AUX area event, then assume the event to sample