[PATCH v5 0/5] perf: Add new option '--workload-config' to set workload sched_policy/prio/cpumask

[PATCH v5 0/5] perf: Add new option '--workload-config' to set workload sched_policy/prio/cpumask

[Changbin Du]

[I still think this is a handy option, so do some improvment and resend it.]

To get consistent benchmarking results, sometimes we need to set the
sched_policy/priority/cpumask of the workload to reduce system noise.

For example, CPU binding is required on big.little system.

  $ perf stat -- taskset -c 0 ls

However, the events of 'taskset' itself are also counted here. To get more
accurate result, this should be avoided.

To get away of the middleman, this adds a new option '--workload-config' to
do the same jobs for stat and record commands.

  --workload-config <[sched_policy=policy][,sched_prio=priority][,cpu-list=list]>
        setup target workload (the <command>) attributes:

                sched_policy: other|fifo|rr|batch|idle
                sched_prio: scheduling priority for fifo|rr, nice value for other
                cpu-list: CPU affinity. e.g. 1-3:5 is processors #1, #2, #3 and #5

For example,

  $ sudo perf stat --workload-config sched_policy=fifo,sched_prio=40,cpu-list=0-3:7 -- ls

Above command will make 'ls' run on CPU #0-#3 and #7 with fifo scheduler and
realtime priority is 40.

v5:
  o rename '--workload-attr' as '--workload-config' (sounds better) .
  o transform to key-value pair style option.

v4:
  - add a test case for perf-stat. (suggeted by Ian Rogers)
  - fix warning found by 0-DAY.
v3:
  - replace taskset with --workload-attr option in documents and tests.
v2:
  - Use cpu list spec instead of cpu mask number.
  - Update documents.

Changbin Du (5):
  perf cpumap: Add __perf_cpu_map__new and perf_cpu_map__2_cpuset
  perf: util: support string type option for perf_parse_sublevel_options
  perf: add new option '--workload-config' to set workload
    sched_policy/prio/cpumask
  perf: replace taskset with --workload-config option
  perf test: add test case for --workload-config option

 tools/lib/perf/cpumap.c                       |  45 +++++++-
 tools/lib/perf/include/perf/cpumap.h          |   4 +
 tools/lib/perf/libperf.map                    |   2 +
 tools/perf/Documentation/intel-hybrid.txt     |   2 +-
 tools/perf/Documentation/perf-record.txt      |   7 ++
 tools/perf/Documentation/perf-stat.txt        |   8 +-
 tools/perf/builtin-record.c                   |  27 +++++
 tools/perf/builtin-stat.c                     |  19 +++
 tools/perf/tests/cpumap.c                     |  23 ++++
 tools/perf/tests/shell/stat.sh                |  19 +++
 .../tests/shell/stat_bpf_counters_cgrp.sh     |   2 +-
 tools/perf/tests/shell/test_arm_coresight.sh  |   2 +-
 tools/perf/tests/shell/test_data_symbol.sh    |   2 +-
 tools/perf/tests/shell/test_intel_pt.sh       |   2 +-
 tools/perf/util/evlist.c                      | 108 ++++++++++++++++++
 tools/perf/util/evlist.h                      |   3 +
 tools/perf/util/parse-sublevel-options.c      |  12 +-
 tools/perf/util/parse-sublevel-options.h      |   7 ++
 tools/perf/util/target.h                      |   9 ++
 19 files changed, 291 insertions(+), 12 deletions(-)

-- 
2.25.1

[PATCH v5 1/5] perf cpumap: Add __perf_cpu_map__new and perf_cpu_map__2_cpuset

[Changbin Du]

This adds two new api which will be used later.
  - __perf_cpu_map__new: accept a specified separator instead of ','.
  - perf_cpu_map__2_cpuset: convert perf_cpu_map to cpu_set_t.

Signed-off-by: Changbin Du <changbin.du@huawei.com>
---
 tools/lib/perf/cpumap.c              | 45 ++++++++++++++++++++++++++--
 tools/lib/perf/include/perf/cpumap.h |  4 +++
 tools/lib/perf/libperf.map           |  2 ++
 tools/perf/tests/cpumap.c            | 23 ++++++++++++++
 4 files changed, 71 insertions(+), 3 deletions(-)

diff --git a/tools/lib/perf/cpumap.c b/tools/lib/perf/cpumap.c
index 2a5a29217374..23e907078b28 100644
--- a/tools/lib/perf/cpumap.c
+++ b/tools/lib/perf/cpumap.c
@@ -1,4 +1,5 @@
 // SPDX-License-Identifier: GPL-2.0-only
+#define _GNU_SOURCE
 #include <perf/cpumap.h>
 #include <stdlib.h>
 #include <linux/refcount.h>
@@ -7,6 +8,7 @@
 #include <stdio.h>
 #include <string.h>
 #include <unistd.h>
+#include <sched.h>
 #include <ctype.h>
 #include <limits.h>
 
@@ -201,7 +203,7 @@ static struct perf_cpu_map *cpu_map__read_all_cpu_map(void)
 	return cpus;
 }
 
-struct perf_cpu_map *perf_cpu_map__new(const char *cpu_list)
+struct perf_cpu_map *__perf_cpu_map__new(const char *cpu_list, char sep)
 {
 	struct perf_cpu_map *cpus = NULL;
 	unsigned long start_cpu, end_cpu = 0;
@@ -225,7 +227,7 @@ struct perf_cpu_map *perf_cpu_map__new(const char *cpu_list)
 		p = NULL;
 		start_cpu = strtoul(cpu_list, &p, 0);
 		if (start_cpu >= INT_MAX
-		    || (*p != '\0' && *p != ',' && *p != '-'))
+		    || (*p != '\0' && *p != sep && *p != '-'))
 			goto invalid;
 
 		if (*p == '-') {
@@ -233,7 +235,7 @@ struct perf_cpu_map *perf_cpu_map__new(const char *cpu_list)
 			p = NULL;
 			end_cpu = strtoul(cpu_list, &p, 0);
 
-			if (end_cpu >= INT_MAX || (*p != '\0' && *p != ','))
+			if (end_cpu >= INT_MAX || (*p != '\0' && *p != sep))
 				goto invalid;
 
 			if (end_cpu < start_cpu)
@@ -278,6 +280,11 @@ struct perf_cpu_map *perf_cpu_map__new(const char *cpu_list)
 	return cpus;
 }
 
+struct perf_cpu_map *perf_cpu_map__new(const char *cpu_list)
+{
+	return __perf_cpu_map__new(cpu_list, ',');
+}
+
 static int __perf_cpu_map__nr(const struct perf_cpu_map *cpus)
 {
 	return RC_CHK_ACCESS(cpus)->nr;
@@ -479,3 +486,35 @@ struct perf_cpu_map *perf_cpu_map__intersect(struct perf_cpu_map *orig,
 	free(tmp_cpus);
 	return merged;
 }
+
+/* The caller is responsible for freeing returned cpu_set_t with CPU_FREE(). */
+cpu_set_t *perf_cpu_map__2_cpuset(struct perf_cpu_map *cpus, size_t *cpuset_size)
+{
+	cpu_set_t *cpusetp;
+	int max_cpu;
+	struct perf_cpu cpu;
+	int idx;
+
+	if (perf_cpu_map__has_any_cpu(cpus))
+		return NULL;
+
+	max_cpu = perf_cpu_map__max(cpus).cpu;
+	if (max_cpu < 0)
+		return NULL;
+
+	cpusetp = CPU_ALLOC(max_cpu + 1);
+	if (cpusetp == NULL)
+		return NULL;
+
+	*cpuset_size = CPU_ALLOC_SIZE(max_cpu + 1);
+	CPU_ZERO_S(*cpuset_size, cpusetp);
+
+	perf_cpu_map__for_each_cpu(cpu, idx, cpus) {
+		if (cpu.cpu == -1)
+			continue;
+
+		CPU_SET_S(cpu.cpu, *cpuset_size, cpusetp);
+	}
+
+	return cpusetp;
+}
diff --git a/tools/lib/perf/include/perf/cpumap.h b/tools/lib/perf/include/perf/cpumap.h
index e38d859a384d..1a0498f92dbe 100644
--- a/tools/lib/perf/include/perf/cpumap.h
+++ b/tools/lib/perf/include/perf/cpumap.h
@@ -3,6 +3,7 @@
 #define __LIBPERF_CPUMAP_H
 
 #include <perf/core.h>
+#include <sched.h>
 #include <stdio.h>
 #include <stdbool.h>
 
@@ -23,6 +24,7 @@ struct perf_cpu_map;
  */
 LIBPERF_API struct perf_cpu_map *perf_cpu_map__dummy_new(void);
 LIBPERF_API struct perf_cpu_map *perf_cpu_map__default_new(void);
+LIBPERF_API struct perf_cpu_map *__perf_cpu_map__new(const char *cpu_list, char sep);
 LIBPERF_API struct perf_cpu_map *perf_cpu_map__new(const char *cpu_list);
 LIBPERF_API struct perf_cpu_map *perf_cpu_map__read(FILE *file);
 LIBPERF_API struct perf_cpu_map *perf_cpu_map__get(struct perf_cpu_map *map);
@@ -46,6 +48,8 @@ LIBPERF_API bool perf_cpu_map__equal(const struct perf_cpu_map *lhs,
  */
 LIBPERF_API bool perf_cpu_map__has_any_cpu(const struct perf_cpu_map *map);
 
+LIBPERF_API cpu_set_t *perf_cpu_map__2_cpuset(struct perf_cpu_map *cpus, size_t *cpuset_size);
+
 #define perf_cpu_map__for_each_cpu(cpu, idx, cpus)		\
 	for ((idx) = 0, (cpu) = perf_cpu_map__cpu(cpus, idx);	\
 	     (idx) < perf_cpu_map__nr(cpus);			\
diff --git a/tools/lib/perf/libperf.map b/tools/lib/perf/libperf.map
index 190b56ae923a..fe0946e34471 100644
--- a/tools/lib/perf/libperf.map
+++ b/tools/lib/perf/libperf.map
@@ -5,6 +5,7 @@ LIBPERF_0.0.1 {
 		perf_cpu_map__default_new;
 		perf_cpu_map__get;
 		perf_cpu_map__put;
+		__perf_cpu_map__new;
 		perf_cpu_map__new;
 		perf_cpu_map__read;
 		perf_cpu_map__nr;
@@ -12,6 +13,7 @@ LIBPERF_0.0.1 {
 		perf_cpu_map__empty;
 		perf_cpu_map__max;
 		perf_cpu_map__has;
+		perf_cpu_map__2_cpuset;
 		perf_thread_map__new_array;
 		perf_thread_map__new_dummy;
 		perf_thread_map__set_pid;
diff --git a/tools/perf/tests/cpumap.c b/tools/perf/tests/cpumap.c
index 7730fc2ab40b..ae5e5337ea4f 100644
--- a/tools/perf/tests/cpumap.c
+++ b/tools/perf/tests/cpumap.c
@@ -1,5 +1,6 @@
 // SPDX-License-Identifier: GPL-2.0
 #include "tests.h"
+#include <sched.h>
 #include <stdio.h>
 #include "cpumap.h"
 #include "event.h"
@@ -247,12 +248,34 @@ static int test__cpu_map_equal(struct test_suite *test __maybe_unused, int subte
 	return TEST_OK;
 }
 
+static int test__cpu_map_convert(struct test_suite *test __maybe_unused, int subtest __maybe_unused)
+{
+	struct perf_cpu_map *any = perf_cpu_map__dummy_new();
+	struct perf_cpu_map *cpus = perf_cpu_map__new("1-2");
+	cpu_set_t *cpu_set;
+	size_t setsize;
+
+	cpu_set = perf_cpu_map__2_cpuset(any, &setsize);
+	TEST_ASSERT_VAL("not equal", cpu_set == NULL);
+	CPU_FREE(cpu_set);
+
+	cpu_set = perf_cpu_map__2_cpuset(cpus, &setsize);
+	TEST_ASSERT_VAL("cpus", cpu_set != NULL);
+	TEST_ASSERT_VAL("bad cpuset", !CPU_ISSET_S(0, setsize, cpu_set));
+	TEST_ASSERT_VAL("bad cpuset", CPU_ISSET_S(1, setsize, cpu_set));
+	TEST_ASSERT_VAL("bad cpuset", CPU_ISSET_S(2, setsize, cpu_set));
+	CPU_FREE(cpu_set);
+
+	return TEST_OK;
+}
+
 static struct test_case tests__cpu_map[] = {
 	TEST_CASE("Synthesize cpu map", cpu_map_synthesize),
 	TEST_CASE("Print cpu map", cpu_map_print),
 	TEST_CASE("Merge cpu map", cpu_map_merge),
 	TEST_CASE("Intersect cpu map", cpu_map_intersect),
 	TEST_CASE("Equal cpu map", cpu_map_equal),
+	TEST_CASE("Convert cpu map", cpu_map_convert),
 	{	.name = NULL, }
 };
 
-- 
2.25.1

[PATCH v5 2/5] perf: util: support string type option for perf_parse_sublevel_options

[Changbin Du]

Add string type option to get non-int value.

Signed-off-by: Changbin Du <changbin.du@huawei.com>
---
 tools/perf/util/parse-sublevel-options.c | 12 +++++++++---
 tools/perf/util/parse-sublevel-options.h |  7 +++++++
 2 files changed, 16 insertions(+), 3 deletions(-)

diff --git a/tools/perf/util/parse-sublevel-options.c b/tools/perf/util/parse-sublevel-options.c
index a841d17ffd57..d08a1ccc9616 100644
--- a/tools/perf/util/parse-sublevel-options.c
+++ b/tools/perf/util/parse-sublevel-options.c
@@ -34,10 +34,16 @@ static int parse_one_sublevel_option(const char *str,
 		return -1;
 	}
 
-	if (vstr)
-		v = atoi(vstr);
+	if (vstr) {
+		/* The value of option either is a integer or string. */
+		if (opt->value_ptr) {
+			v = atoi(vstr);
+			*opt->value_ptr = v;
+		} else {
+			*opt->str_ptr = strdup(vstr);
+		}
+	}
 
-	*opt->value_ptr = v;
 	free(s);
 	return 0;
 }
diff --git a/tools/perf/util/parse-sublevel-options.h b/tools/perf/util/parse-sublevel-options.h
index 578b18ef03bb..d536ebe43b58 100644
--- a/tools/perf/util/parse-sublevel-options.h
+++ b/tools/perf/util/parse-sublevel-options.h
@@ -3,7 +3,14 @@
 
 struct sublevel_option {
 	const char *name;
+
+	/*
+	 * Only one of below can be non-null. So we simply support
+	 * two types: integer and string. For string, the caller is
+	 * responsible for freeing allocated memory after use.
+	 */
 	int *value_ptr;
+	char **str_ptr;
 };
 
 int perf_parse_sublevel_options(const char *str, struct sublevel_option *opts);
-- 
2.25.1

[PATCH v5 3/5] perf: add new option '--workload-config' to set workload sched_policy/prio/cpumask

[Changbin Du]

To get consistent benchmarking results, sometimes we need to set the
sched_policy/priority/cpumask of the workload to reduce system noise.

For example, CPU binding is required on big.little system.

  $ perf stat -- taskset -c 0 ls

However, the events of 'taskset' itself are also counted here. To get more
accurate result, this should be avoided.

To get away of the middleman, this adds a new option '--workload-config' to
do the same jobs for stat and record commands.

  --workload-config <[sched_policy=policy][,sched_prio=priority][,cpu-list=list]>
        setup target workload (the <command>) attributes:

		sched_policy: other|fifo|rr|batch|idle
                sched_prio: scheduling priority for fifo|rr, nice value for other
                cpu-list: CPU affinity. e.g. 1-3:5 is processors #1, #2, #3 and #5

For example,

  $ sudo perf stat --workload-config sched_policy=fifo,sched_prio=40,cpu-list=0-3:7 -- ls

Above command will make 'ls' run on CPU #0-#3 and #7 with fifo scheduler and
realtime priority is 40.

Cc: kernel test robot <yujie.liu@intel.com>
Signed-off-by: Changbin Du <changbin.du@huawei.com>

---
v2: Use cpu list spec instead of cpu mask number.
v3:
  o rename '--workload-attr' as '--workload-config'
  o transform to key-value style option
---
 tools/perf/Documentation/perf-record.txt |   7 ++
 tools/perf/Documentation/perf-stat.txt   |   6 ++
 tools/perf/builtin-record.c              |  27 ++++++
 tools/perf/builtin-stat.c                |  19 ++++
 tools/perf/util/evlist.c                 | 108 +++++++++++++++++++++++
 tools/perf/util/evlist.h                 |   3 +
 tools/perf/util/target.h                 |   9 ++
 7 files changed, 179 insertions(+)

diff --git a/tools/perf/Documentation/perf-record.txt b/tools/perf/Documentation/perf-record.txt
index d5217be012d7..da4692751e17 100644
--- a/tools/perf/Documentation/perf-record.txt
+++ b/tools/perf/Documentation/perf-record.txt
@@ -821,6 +821,13 @@ filtered through the mask provided by -C option.
 	only, as of now.  So the applications built without the frame
 	pointer might see bogus addresses.
 
+--workload-config <[sched_policy=policy][,sched_prio=priority][,cpu-list=list]>::
+	setup target workload (the <command>) attributes:
+
+	  sched_policy: other|fifo|rr|batch|idle
+	  sched_prio: scheduling priority for fifo|rr, nice value for other
+	  cpu-list: CPU affinity. e.g. 1-3:5 is processors #1, #2, #3 and #5
+
 include::intel-hybrid.txt[]
 
 SEE ALSO
diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt
index 8f789fa1242e..b2038f7e236a 100644
--- a/tools/perf/Documentation/perf-stat.txt
+++ b/tools/perf/Documentation/perf-stat.txt
@@ -262,6 +262,12 @@ disable events during measurements:
  wait -n ${perf_pid}
  exit $?
 
+--workload-config <[sched_policy=policy][,sched_prio=priority][,cpu-list=list]>::
+	setup target workload (the <command>) attributes:
+
+	  sched_policy: other|fifo|rr|batch|idle
+	  sched_prio: scheduling priority for fifo|rr, nice value for other
+	  cpu-list: CPU affinity. e.g. 1-3:5 is processors #1, #2, #3 and #5
 
 --pre::
 --post::
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index 34bb31f08bb5..20799a1e60f6 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -3277,6 +3277,17 @@ static int parse_record_synth_option(const struct option *opt,
 	return 0;
 }
 
+static int record_parse_workload_attr_opt(const struct option *opt,
+					  const char *arg,
+					  int unset __maybe_unused)
+{
+	struct record_opts *opts = opt->value;
+
+	return evlist__parse_workload_config(arg, &opts->target.workload.sched_policy,
+					     &opts->target.workload.sched_priority,
+					     &opts->target.workload.cpu_map);
+}
+
 /*
  * XXX Ideally would be local to cmd_record() and passed to a record__new
  * because we need to have access to it in record__exit, that is called
@@ -3297,6 +3308,8 @@ static struct record record = {
 		.target		     = {
 			.uses_mmap   = true,
 			.default_per_cpu = true,
+			.workload.sched_policy = -1,
+			.workload.sched_priority = 0,
 		},
 		.mmap_flush          = MMAP_FLUSH_DEFAULT,
 		.nr_threads_synthesize = 1,
@@ -3321,6 +3334,12 @@ static struct record record = {
 const char record_callchain_help[] = CALLCHAIN_RECORD_HELP
 	"\n\t\t\t\tDefault: fp";
 
+const char record_workload_config_help[] =
+	"setup target workload (the <command>) attributes:\n\n"
+	HELP_PAD "sched_policy: other|fifo|rr|batch|idle\n"
+	HELP_PAD "sched_prio: scheduling priority for fifo|rr, nice value for other\n"
+	HELP_PAD "cpu-list: CPU affinity. e.g. 1-3:5 is processors #1, #2, #3 and #5";
+
 static bool dry_run;
 
 static struct parse_events_option_args parse_events_option_args = {
@@ -3535,6 +3554,10 @@ static struct option __record_options[] = {
 			    "write collected trace data into several data files using parallel threads",
 			    record__parse_threads),
 	OPT_BOOLEAN(0, "off-cpu", &record.off_cpu, "Enable off-cpu analysis"),
+	OPT_CALLBACK(0, "workload-config", &record.opts,
+		     "[sched_policy=policy][,sched_prio=priority][,cpu-list=list]",
+		     record_workload_config_help,
+		     &record_parse_workload_attr_opt),
 	OPT_END()
 };
 
@@ -4221,6 +4244,10 @@ int cmd_record(int argc, const char **argv)
 	record__free_thread_masks(rec, rec->nr_threads);
 	rec->nr_threads = 0;
 	evlist__close_control(rec->opts.ctl_fd, rec->opts.ctl_fd_ack, &rec->opts.ctl_fd_close);
+	if (rec->opts.target.workload.cpu_map) {
+		perf_cpu_map__put(rec->opts.target.workload.cpu_map);
+		rec->opts.target.workload.cpu_map = NULL;
+	}
 	return err;
 }
 
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 07b48f6df48e..a7a3a788e7d9 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -108,6 +108,8 @@ static bool all_counters_use_bpf = true;
 
 static struct target target = {
 	.uid	= UINT_MAX,
+	.workload.sched_policy = -1,
+	.workload.sched_priority = 0,
 };
 
 #define METRIC_ONLY_LEN 20
@@ -1160,6 +1162,14 @@ static int parse_cache_level(const struct option *opt,
 	return 0;
 }
 
+static int parse_workload_attr_opt(const struct option *opt __maybe_unused, const char *arg,
+				   int unset __maybe_unused)
+{
+	return evlist__parse_workload_config(arg, &target.workload.sched_policy,
+					     &target.workload.sched_priority,
+					     &target.workload.cpu_map);
+}
+
 static struct option stat_options[] = {
 	OPT_BOOLEAN('T', "transaction", &transaction_run,
 		    "hardware transaction statistics"),
@@ -1220,6 +1230,10 @@ static struct option stat_options[] = {
 	OPT_BOOLEAN(0, "append", &append_file, "append to the output file"),
 	OPT_INTEGER(0, "log-fd", &output_fd,
 		    "log output to fd, instead of stderr"),
+	OPT_CALLBACK(0, "workload-config", &stat_config,
+		     "[sched_policy=policy][,sched_prio=priority][,cpu-list=list]",
+		     record_workload_config_help,
+		     &parse_workload_attr_opt),
 	OPT_STRING(0, "pre", &pre_cmd, "command",
 			"command to run prior to the measured command"),
 	OPT_STRING(0, "post", &post_cmd, "command",
@@ -2893,5 +2907,10 @@ int cmd_stat(int argc, const char **argv)
 	metricgroup__rblist_exit(&stat_config.metric_events);
 	evlist__close_control(stat_config.ctl_fd, stat_config.ctl_fd_ack, &stat_config.ctl_fd_close);
 
+	if (target.workload.cpu_map) {
+		perf_cpu_map__put(target.workload.cpu_map);
+		target.workload.cpu_map = NULL;
+	}
+
 	return status;
 }
diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index 7ef43f72098e..7ad7a4fed282 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -33,6 +33,7 @@
 #include "util/bpf-filter.h"
 #include "util/stat.h"
 #include "util/util.h"
+#include "util/parse-sublevel-options.h"
 #include <signal.h>
 #include <unistd.h>
 #include <sched.h>
@@ -46,6 +47,7 @@
 #include <sys/mman.h>
 #include <sys/prctl.h>
 #include <sys/timerfd.h>
+#include <sys/resource.h>
 
 #include <linux/bitops.h>
 #include <linux/hash.h>
@@ -1398,6 +1400,109 @@ int evlist__open(struct evlist *evlist)
 	return err;
 }
 
+int evlist__parse_workload_config(const char *str, int *sched_policy, int *sched_priority,
+				  struct perf_cpu_map **cpu_map)
+{
+	char *policy_str = NULL;
+	int priority = -1;
+	char *cpu_list = NULL;
+	int ret;
+	struct sublevel_option workload_conf_opts[] = {
+		{ .name = "sched_policy",	.str_ptr = &policy_str},
+		{ .name = "sched_prio",		.value_ptr = &priority},
+		{ .name = "cpu-list",		.str_ptr = &cpu_list},
+		{ .name = NULL, }
+	};
+
+	ret = perf_parse_sublevel_options(str, workload_conf_opts);
+	if (ret)
+		return ret;
+
+	/* sched policy, default to 'other'. */
+	if (!policy_str || !strncmp(policy_str, "other", sizeof("other")))
+		*sched_policy = SCHED_OTHER;
+	else if (!strncmp(policy_str, "fifo", sizeof("fifo")))
+		*sched_policy = SCHED_FIFO;
+	else if (!strncmp(policy_str, "rr", sizeof("rr")))
+		*sched_policy = SCHED_RR;
+	else if (!strncmp(policy_str, "batch", sizeof("batch")))
+		*sched_policy = SCHED_BATCH;
+	else if (!strncmp(policy_str, "idle", sizeof("idle")))
+		*sched_policy = SCHED_IDLE;
+	else {
+		pr_err("workload_attr: unknown sched policy %s\n", policy_str);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	/* check sched priority and set default value */
+	if (*sched_policy == SCHED_FIFO || *sched_policy == SCHED_RR) {
+		if (priority == -1)
+			priority = 99; /* default to lowest priority */
+		else if (priority < 1 || priority > 99) {
+			pr_err("workload_attr: invalid priority %d for fifo and rr, allowed [1,99]\n",
+				priority);
+			ret = -EINVAL;
+			goto out;
+		}
+	} else if (*sched_policy == SCHED_OTHER && priority == -1)
+		priority = 0;
+	*sched_priority = priority;
+
+	/* allowed cpu list */
+	*cpu_map = __perf_cpu_map__new(cpu_list, ':');
+	if (!*cpu_map) {
+		pr_err("workload_attr: failed to get cpus map from %s\n", cpu_list);
+		ret = -EINVAL;
+	}
+
+out:
+	free(policy_str);
+	free(cpu_list);
+	return ret;
+}
+
+static int configurate_workload(struct target *target)
+{
+	struct sched_param param;
+	int policy = target->workload.sched_policy;
+	int priority = target->workload.sched_priority;
+
+	if (policy >= 0) {
+		param.sched_priority = (policy == SCHED_FIFO || policy == SCHED_RR) ?
+					priority : 0;
+		if (sched_setscheduler(0, policy, &param) != 0) {
+			pr_err("failed to set the sched policy %d: %s\n", policy, strerror(errno));
+			return -1;
+		}
+
+		if (policy == SCHED_OTHER) {
+			if (setpriority(PRIO_PROCESS, 0, priority) != 0) {
+				pr_err("failed to set the nice value %d: %s\n", priority, strerror(errno));
+				return -1;
+			}
+		}
+	}
+
+	if (target->workload.cpu_map) {
+		size_t cpuset_size = -1;
+		cpu_set_t *cpu_set;
+
+		cpu_set = perf_cpu_map__2_cpuset(target->workload.cpu_map, &cpuset_size);
+		if (!cpu_set)
+			return -1;
+
+		if (sched_setaffinity(0, cpuset_size, cpu_set) != 0) {
+			pr_err("failed to set the sched affinity: %s\n", strerror(errno));
+			CPU_FREE(cpu_set);
+			return -1;
+		}
+		CPU_FREE(cpu_set);
+	}
+
+	return 0;
+}
+
 int evlist__prepare_workload(struct evlist *evlist, struct target *target, const char *argv[],
 			     bool pipe_output, void (*exec_error)(int signo, siginfo_t *info, void *ucontext))
 {
@@ -1464,6 +1569,9 @@ int evlist__prepare_workload(struct evlist *evlist, struct target *target, const
 			exit(ret);
 		}
 
+		if (configurate_workload(target) != 0)
+			exit(-1);
+
 		execvp(argv[0], (char **)argv);
 
 		if (exec_error) {
diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
index 664c6bf7b3e0..540e17d0d9fe 100644
--- a/tools/perf/util/evlist.h
+++ b/tools/perf/util/evlist.h
@@ -15,6 +15,7 @@
 #include <pthread.h>
 #include <signal.h>
 #include <unistd.h>
+#include <sched.h>
 
 struct pollfd;
 struct thread_map;
@@ -180,6 +181,8 @@ void evlist__set_id_pos(struct evlist *evlist);
 void evlist__config(struct evlist *evlist, struct record_opts *opts, struct callchain_param *callchain);
 int record_opts__config(struct record_opts *opts);
 
+int evlist__parse_workload_config(const char *str, int *sched_policy, int *sched_priority,
+				  struct perf_cpu_map **cpu_set);
 int evlist__prepare_workload(struct evlist *evlist, struct target *target,
 			     const char *argv[], bool pipe_output,
 			     void (*exec_error)(int signo, siginfo_t *info, void *ucontext));
diff --git a/tools/perf/util/target.h b/tools/perf/util/target.h
index d582cae8e105..78b7e7ab1c7b 100644
--- a/tools/perf/util/target.h
+++ b/tools/perf/util/target.h
@@ -4,6 +4,7 @@
 
 #include <stdbool.h>
 #include <sys/types.h>
+#include <sched.h>
 
 struct target {
 	const char   *pid;
@@ -19,6 +20,12 @@ struct target {
 	bool	     use_bpf;
 	int	     initial_delay;
 	const char   *attr_map;
+
+	struct {
+		int	 sched_policy;
+		int	 sched_priority;
+		struct perf_cpu_map *cpu_map;
+	} workload;
 };
 
 enum target_errno {
@@ -103,4 +110,6 @@ static inline bool target__uses_dummy_map(struct target *target)
 	return use_dummy;
 }
 
+extern const char record_workload_config_help[];
+
 #endif /* _PERF_TARGET_H */
-- 
2.25.1

[PATCH v5 4/5] perf: replace taskset with --workload-config option

[Changbin Du]

Replace the taskset with our new option --workload-config.

Signed-off-by: Changbin Du <changbin.du@huawei.com>
---
 tools/perf/Documentation/intel-hybrid.txt        | 2 +-
 tools/perf/Documentation/perf-stat.txt           | 2 +-
 tools/perf/tests/shell/stat_bpf_counters_cgrp.sh | 2 +-
 tools/perf/tests/shell/test_arm_coresight.sh     | 2 +-
 tools/perf/tests/shell/test_data_symbol.sh       | 2 +-
 tools/perf/tests/shell/test_intel_pt.sh          | 2 +-
 6 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/tools/perf/Documentation/intel-hybrid.txt b/tools/perf/Documentation/intel-hybrid.txt
index e7a776ad25d7..f2b9d623a656 100644
--- a/tools/perf/Documentation/intel-hybrid.txt
+++ b/tools/perf/Documentation/intel-hybrid.txt
@@ -132,7 +132,7 @@ displayed. The percentage is the event's running time/enabling time.
 One example, 'triad_loop' runs on cpu16 (atom core), while we can see the
 scaled value for core cycles is 160,444,092 and the percentage is 0.47%.
 
-perf stat -e cycles \-- taskset -c 16 ./triad_loop
+perf stat -e cycles --workload-config cpu-list=16 \-- ./triad_loop
 
 As previous, two events are created.
 
diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt
index b2038f7e236a..fc22d72560e1 100644
--- a/tools/perf/Documentation/perf-stat.txt
+++ b/tools/perf/Documentation/perf-stat.txt
@@ -464,7 +464,7 @@ on workload with changing phases.
 
 To interpret the results it is usually needed to know on which
 CPUs the workload runs on. If needed the CPUs can be forced using
-taskset.
+--workload-config option.
 
 --td-level::
 Print the top-down statistics that equal the input level. It allows
diff --git a/tools/perf/tests/shell/stat_bpf_counters_cgrp.sh b/tools/perf/tests/shell/stat_bpf_counters_cgrp.sh
index e75d0780dc78..c8737cb32a63 100755
--- a/tools/perf/tests/shell/stat_bpf_counters_cgrp.sh
+++ b/tools/perf/tests/shell/stat_bpf_counters_cgrp.sh
@@ -60,7 +60,7 @@ check_system_wide_counted()
 
 check_cpu_list_counted()
 {
-	check_cpu_list_counted_output=$(perf stat -C 0,1 --bpf-counters --for-each-cgroup ${test_cgroups} -e cpu-clock -x, taskset -c 1 sleep 1  2>&1)
+	check_cpu_list_counted_output=$(perf stat -C 0,1 --bpf-counters --for-each-cgroup ${test_cgroups} -e cpu-clock -x, --workload-config cpu-list=1 -- sleep 1  2>&1)
 	if echo ${check_cpu_list_counted_output} | grep -q -F "<not "; then
 		echo "Some CPU events are not counted"
 		if [ "${verbose}" = "1" ]; then
diff --git a/tools/perf/tests/shell/test_arm_coresight.sh b/tools/perf/tests/shell/test_arm_coresight.sh
index f1bf5621160f..e5be16fa7171 100755
--- a/tools/perf/tests/shell/test_arm_coresight.sh
+++ b/tools/perf/tests/shell/test_arm_coresight.sh
@@ -38,7 +38,7 @@ record_touch_file() {
 	echo "Recording trace (only user mode) with path: CPU$2 => $1"
 	rm -f $file
 	perf record -o ${perfdata} -e cs_etm/@$1/u --per-thread \
-		-- taskset -c $2 touch $file > /dev/null 2>&1
+		--workload-attr cpu-list=$2 -- touch $file > /dev/null 2>&1
 }
 
 perf_script_branch_samples() {
diff --git a/tools/perf/tests/shell/test_data_symbol.sh b/tools/perf/tests/shell/test_data_symbol.sh
index 69bb6fe86c50..d2b58c78a18b 100755
--- a/tools/perf/tests/shell/test_data_symbol.sh
+++ b/tools/perf/tests/shell/test_data_symbol.sh
@@ -50,7 +50,7 @@ echo "Recording workload..."
 # specific CPU and test in per-CPU mode.
 is_amd=$(grep -E -c 'vendor_id.*AuthenticAMD' /proc/cpuinfo)
 if (($is_amd >= 1)); then
-	perf mem record -o ${PERF_DATA} -C 0 -- taskset -c 0 $TEST_PROGRAM &
+	perf mem record -o ${PERF_DATA} -C 0 --workload-config cpu-list=0 -- $TEST_PROGRAM &
 else
 	perf mem record --all-user -o ${PERF_DATA} -- $TEST_PROGRAM &
 fi
diff --git a/tools/perf/tests/shell/test_intel_pt.sh b/tools/perf/tests/shell/test_intel_pt.sh
index 3a8b9bffa022..46f53aece135 100755
--- a/tools/perf/tests/shell/test_intel_pt.sh
+++ b/tools/perf/tests/shell/test_intel_pt.sh
@@ -110,7 +110,7 @@ test_system_wide_side_band()
 	can_cpu_wide 1 || return $?
 
 	# Record on CPU 0 a task running on CPU 1
-	perf_record_no_decode -o "${perfdatafile}" -e intel_pt//u -C 0 -- taskset --cpu-list 1 uname
+	perf_record_no_decode -o "${perfdatafile}" -e intel_pt//u -C 0 --workload-config cpu-list=1 -- uname
 
 	# Should get MMAP events from CPU 1 because they can be needed to decode
 	mmap_cnt=$(perf script -i "${perfdatafile}" --no-itrace --show-mmap-events -C 1 2>/dev/null | grep -c MMAP)
-- 
2.25.1

[PATCH v5 5/5] perf test: add test case for --workload-config option

[Changbin Du]

Ensure the sched attributes are applied to workload.

Signed-off-by: Changbin Du <changbin.du@huawei.com>
---
 tools/perf/tests/shell/stat.sh | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/tools/perf/tests/shell/stat.sh b/tools/perf/tests/shell/stat.sh
index 3f1e67795490..7b320a25505a 100755
--- a/tools/perf/tests/shell/stat.sh
+++ b/tools/perf/tests/shell/stat.sh
@@ -16,6 +16,24 @@ test_default_stat() {
   echo "Basic stat command test [Success]"
 }
 
+test_stat_workload_config() {
+  echo "stat with --workload-config test"
+  if ! perf stat --workload-config cpu-list=1 -- bash -c 'taskset -pc $$' 2>&1 | grep -E -q "current affinity list: 1"
+  then
+    echo "stat with --workload-config test [Failed]"
+    err=1
+    return
+  fi
+
+  if ! perf stat --workload-config sched_policy=other,sched_prio=10 -- bash -c 'ps -o pid,cls,ni,cmd -p $$' 2>&1 | grep -E -q "TS\s+10"
+  then
+    echo "stat with --workload-config test [Failed]"
+    err=1
+    return
+  fi
+  echo "stat with --workload-config test [Success]"
+}
+
 test_stat_record_report() {
   echo "stat record and report test"
   if ! perf stat record -o - true | perf stat report -i - 2>&1 | \
@@ -147,6 +165,7 @@ test_cputype() {
 }
 
 test_default_stat
+test_stat_workload_config
 test_stat_record_report
 test_stat_record_script
 test_stat_repeat_weak_groups
-- 
2.25.1