perf tools: Add support for exclusive groups/events

Peter suggested that using the exclusive mode in perf could avoid some
problems with bad scheduling of groups. Exclusive is implemented in the
kernel, but wasn't exposed by the perf tool, so hard to use without
custom low level API users.

Add support for marking groups or events with :e for exclusive in the
perf tool.  The implementation is basically the same as the existing
pinned attribute.

Committer testing:

  # perf test "parse event"
   6: Parse event definition strings                                  : Ok
  # perf test -v "parse event" |& grep :u*e
  running test 56 'instructions:uep'
  running test 57 '{cycles,cache-misses,branch-misses}:e'
  #
  #
  # grep "model name" -m1 /proc/cpuinfo
  model name	: AMD Ryzen 9 3900X 12-Core Processor
  #
  # perf stat -a -e '{cycles,cache-misses,branch-misses}:e' sleep 1

   Performance counter stats for 'system wide':

       <not counted>      cycles                                                        (0.00%)
       <not counted>      cache-misses                                                  (0.00%)
       <not counted>      branch-misses                                                 (0.00%)

         1.001269893 seconds time elapsed

  Some events weren't counted. Try disabling the NMI watchdog:
  	echo 0 > /proc/sys/kernel/nmi_watchdog
  	perf stat ...
  	echo 1 > /proc/sys/kernel/nmi_watchdog
  # echo 0 > /proc/sys/kernel/nmi_watchdog
  # perf stat -a -e '{cycles,cache-misses,branch-misses}:e' sleep 1

   Performance counter stats for 'system wide':

       1,298,663,141      cycles
          30,962,215      cache-misses
           5,325,150      branch-misses

         1.001474934 seconds time elapsed

  #
  # The output for asking for precise events on AMD needs to improve, it
  # supposedly works only for system wide or per CPU
  #
  # perf stat -a -e '{cycles,cache-misses,branch-misses}:uep' sleep 1
  Error:
  The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (cycles).
  /bin/dmesg | grep -i perf may provide additional information.

  # perf stat -a -e '{cycles,cache-misses,branch-misses}:ue' sleep 1

   Performance counter stats for 'system wide':

         746,363,126      cycles
          16,881,611      cache-misses
           2,871,259      branch-misses

         1.001636066 seconds time elapsed

  #

Signed-off-by: Andi Kleen <ak@linux.intel.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/20201014144255.22699-1-andi@firstfloor.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

4 files changed, 68 insertions, 4 deletions

diff --git a/tools/perf/Documentation/perf-list.txt b/tools/perf/Documentation/perf-list.txt
index 10ed539a8859..4c7db1da8fcc 100644
--- a/tools/perf/Documentation/perf-list.txt
+++ b/tools/perf/Documentation/perf-list.txt
@@ -58,6 +58,7 @@ counted. The following modifiers exist:
  S - read sample value (PERF_SAMPLE_READ)
  D - pin the event to the PMU
  W - group is weak and will fallback to non-group if not schedulable,
+ e - group or event are exclusive and do not share the PMU
 
 The 'p' modifier can be used for specifying how precise the instruction
 address should be. The 'p' modifier can be specified multiple times:
diff --git a/tools/perf/tests/parse-events.c b/tools/perf/tests/parse-events.c
index aae0fd9045c1..611512f22b34 100644
--- a/tools/perf/tests/parse-events.c
+++ b/tools/perf/tests/parse-events.c
@@ -557,6 +557,7 @@ static int test__checkevent_pmu_events(struct evlist *evlist)
 	TEST_ASSERT_VAL("wrong exclude_hv", evsel->core.attr.exclude_hv);
 	TEST_ASSERT_VAL("wrong precise_ip", !evsel->core.attr.precise_ip);
 	TEST_ASSERT_VAL("wrong pinned", !evsel->core.attr.pinned);
+	TEST_ASSERT_VAL("wrong exclusive", !evsel->core.attr.exclusive);
 
 	return 0;
 }
@@ -575,6 +576,7 @@ static int test__checkevent_pmu_events_mix(struct evlist *evlist)
 	TEST_ASSERT_VAL("wrong exclude_hv", evsel->core.attr.exclude_hv);
 	TEST_ASSERT_VAL("wrong precise_ip", !evsel->core.attr.precise_ip);
 	TEST_ASSERT_VAL("wrong pinned", !evsel->core.attr.pinned);
+	TEST_ASSERT_VAL("wrong exclusive", !evsel->core.attr.exclusive);
 
 	/* cpu/pmu-event/u*/
 	evsel = evsel__next(evsel);
@@ -587,6 +589,7 @@ static int test__checkevent_pmu_events_mix(struct evlist *evlist)
 	TEST_ASSERT_VAL("wrong exclude_hv", evsel->core.attr.exclude_hv);
 	TEST_ASSERT_VAL("wrong precise_ip", !evsel->core.attr.precise_ip);
 	TEST_ASSERT_VAL("wrong pinned", !evsel->core.attr.pinned);
+	TEST_ASSERT_VAL("wrong exclusive", !evsel->core.attr.pinned);
 
 	return 0;
 }
@@ -1277,6 +1280,49 @@ static int test__pinned_group(struct evlist *evlist)
 	return 0;
 }
 
+static int test__checkevent_exclusive_modifier(struct evlist *evlist)
+{
+	struct evsel *evsel = evlist__first(evlist);
+
+	TEST_ASSERT_VAL("wrong exclude_user", !evsel->core.attr.exclude_user);
+	TEST_ASSERT_VAL("wrong exclude_kernel", evsel->core.attr.exclude_kernel);
+	TEST_ASSERT_VAL("wrong exclude_hv", evsel->core.attr.exclude_hv);
+	TEST_ASSERT_VAL("wrong precise_ip", evsel->core.attr.precise_ip);
+	TEST_ASSERT_VAL("wrong exclusive", evsel->core.attr.exclusive);
+
+	return test__checkevent_symbolic_name(evlist);
+}
+
+static int test__exclusive_group(struct evlist *evlist)
+{
+	struct evsel *evsel, *leader;
+
+	TEST_ASSERT_VAL("wrong number of entries", 3 == evlist->core.nr_entries);
+
+	/* cycles - group leader */
+	evsel = leader = evlist__first(evlist);
+	TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
+	TEST_ASSERT_VAL("wrong config",
+			PERF_COUNT_HW_CPU_CYCLES == evsel->core.attr.config);
+	TEST_ASSERT_VAL("wrong group name", !evsel->group_name);
+	TEST_ASSERT_VAL("wrong leader", evsel->leader == leader);
+	TEST_ASSERT_VAL("wrong exclusive", evsel->core.attr.exclusive);
+
+	/* cache-misses - can not be pinned, but will go on with the leader */
+	evsel = evsel__next(evsel);
+	TEST_ASSERT_VAL("wrong type", PERF_TYPE_HARDWARE == evsel->core.attr.type);
+	TEST_ASSERT_VAL("wrong config",
+			PERF_COUNT_HW_CACHE_MISSES == evsel->core.attr.config);
+	TEST_ASSERT_VAL("wrong exclusive", !evsel->core.attr.exclusive);
+
+	/* branch-misses - ditto */
+	evsel = evsel__next(evsel);
+	TEST_ASSERT_VAL("wrong config",
+			PERF_COUNT_HW_BRANCH_MISSES == evsel->core.attr.config);
+	TEST_ASSERT_VAL("wrong exclusive", !evsel->core.attr.exclusive);
+
+	return 0;
+}
 static int test__checkevent_breakpoint_len(struct evlist *evlist)
 {
 	struct evsel *evsel = evlist__first(evlist);
@@ -1765,7 +1811,17 @@ static struct evlist_test test__events[] = {
 		.name  = "cycles:k",
 		.check = test__sym_event_dc,
 		.id    = 55,
-	}
+	},
+	{
+		.name  = "instructions:uep",
+		.check = test__checkevent_exclusive_modifier,
+		.id    = 56,
+	},
+	{
+		.name  = "{cycles,cache-misses,branch-misses}:e",
+		.check = test__exclusive_group,
+		.id    = 57,
+	},
 };
 
 static struct evlist_test test__events_pmu[] = {
diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index 3bcdf084df91..3b273580fb84 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -1775,6 +1775,7 @@ struct event_modifier {
 	int sample_read;
 	int pinned;
 	int weak;
+	int exclusive;
 };
 
 static int get_event_modifier(struct event_modifier *mod, char *str,
@@ -1790,6 +1791,7 @@ static int get_event_modifier(struct event_modifier *mod, char *str,
 	int precise_max = 0;
 	int sample_read = 0;
 	int pinned = evsel ? evsel->core.attr.pinned : 0;
+	int exclusive = evsel ? evsel->core.attr.exclusive : 0;
 
 	int exclude = eu | ek | eh;
 	int exclude_GH = evsel ? evsel->exclude_GH : 0;
@@ -1833,6 +1835,8 @@ static int get_event_modifier(struct event_modifier *mod, char *str,
 			sample_read = 1;
 		} else if (*str == 'D') {
 			pinned = 1;
+		} else if (*str == 'e') {
+			exclusive = 1;
 		} else if (*str == 'W') {
 			weak = 1;
 		} else
@@ -1866,6 +1870,7 @@ static int get_event_modifier(struct event_modifier *mod, char *str,
 	mod->sample_read = sample_read;
 	mod->pinned = pinned;
 	mod->weak = weak;
+	mod->exclusive = exclusive;
 
 	return 0;
 }
@@ -1879,7 +1884,7 @@ static int check_modifier(char *str)
 	char *p = str;
 
 	/* The sizeof includes 0 byte as well. */
-	if (strlen(str) > (sizeof("ukhGHpppPSDIW") - 1))
+	if (strlen(str) > (sizeof("ukhGHpppPSDIWe") - 1))
 		return -1;
 
 	while (*p) {
@@ -1921,8 +1926,10 @@ int parse_events__modifier_event(struct list_head *list, char *str, bool add)
 		evsel->precise_max         = mod.precise_max;
 		evsel->weak_group	   = mod.weak;
 
-		if (evsel__is_group_leader(evsel))
+		if (evsel__is_group_leader(evsel)) {
 			evsel->core.attr.pinned = mod.pinned;
+			evsel->core.attr.exclusive = mod.exclusive;
+		}
 	}
 
 	return 0;
diff --git a/tools/perf/util/parse-events.l b/tools/perf/util/parse-events.l
index 3ca5fd2829ca..9db5097317f4 100644
--- a/tools/perf/util/parse-events.l
+++ b/tools/perf/util/parse-events.l
@@ -210,7 +210,7 @@ name_tag	[\'][a-zA-Z_*?\[\]][a-zA-Z0-9_*?\-,\.\[\]:=]*[\']
 name_minus	[a-zA-Z_*?][a-zA-Z0-9\-_*?.:]*
 drv_cfg_term	[a-zA-Z0-9_\.]+(=[a-zA-Z0-9_*?\.:]+)?
 /* If you add a modifier you need to update check_modifier() */
-modifier_event	[ukhpPGHSDIW]+
+modifier_event	[ukhpPGHSDIWe]+
 modifier_bp	[rwx]{1,3}
 
 %%