Skip to contentSkip to conversation

perf test: Add event group test

PATCH In review Ravi Bangoria 6 participants 11 messages Dec 6, 2022 – Dec 7, 2022 perf events

lore
Share & tools
lore
#0 lore raw

v3 cover perf test: Add event group test

Multiple events in a group can belong to one or more pmus, however there are some limitations to it. One of the limitation is, perf doesn't allow creating a group of events from different hw pmus. Write a simple test to create various combinations of hw, sw and uncore pmu events and verify group creation succeeds or fails as expected.

v2: https://lore.kernel.org/r/[email protected]
v2->v3:
- Define a set of uncore pmus that supports more than 3 events. This will prevent false negative results by not picking random uncore pmu. Test will be skipped if no uncore pmu found.

Ravi Bangoria (2):
perf tool: Move pmus list variable to new a file perf test: Add event group test

 tools/perf/tests/Build          |   1 +
 tools/perf/tests/builtin-test.c |   1 +
 tools/perf/tests/event_groups.c | 127 ++++++++++++++++++++++++++++++++
 tools/perf/tests/tests.h        |   1 +
 tools/perf/util/Build           |   1 +
 tools/perf/util/pmu.c           |   2 +-
 tools/perf/util/pmus.c          |   5 ++
 tools/perf/util/pmus.h          |   9 +++
 8 files changed, 146 insertions(+), 1 deletion(-)
 create mode 100644 tools/perf/tests/event_groups.c
 create mode 100644 tools/perf/util/pmus.c
 create mode 100644 tools/perf/util/pmus.h
Signature · 1 line
2.38.1
#1 lore raw

PATCH v3 1/2 perf tool: Move pmus list variable to new a file

pmus list variable is defined as static variable under pmu.c file. Introduce new file pmus.c and migrate this variable to it. Also make it non static so that it can be accessed from outside.

---
 tools/perf/util/Build  | 1 +
 tools/perf/util/pmu.c  | 2 +-
 tools/perf/util/pmus.c | 5 +++++
 tools/perf/util/pmus.h | 9 +++++++++
 4 files changed, 16 insertions(+), 1 deletion(-)
 create mode 100644 tools/perf/util/pmus.c
 create mode 100644 tools/perf/util/pmus.h
tools/perf/util/Build +1 -0
index ab37f588ee8b..d04802bfa23f 100644
@@ -73,6 +73,7 @@ perf-y += trace-event-parse.o
 perf-y += parse-events-flex.o
 perf-y += parse-events-bison.o
 perf-y += pmu.o
+perf-y += pmus.o
 perf-y += pmu-flex.o
 perf-y += pmu-bison.o
 perf-y += pmu-hybrid.o
tools/perf/util/pmu.c +1 -1
index e9a4f31926bf..f5e10f41c042 100644
@@ -22,6 +22,7 @@
 #include "debug.h"
 #include "evsel.h"
 #include "pmu.h"
+#include "pmus.h"
 #include "parse-events.h"
 #include "print-events.h"
 #include "header.h"
@@ -58,7 +59,6 @@ struct perf_pmu_format {
 int perf_pmu_parse(struct list_head *list, char *name);
 extern FILE *perf_pmu_in;
 
-static LIST_HEAD(pmus);
 static bool hybrid_scanned;
 
 /*
tools/perf/util/pmus.c new file +5 -0
new file mode 100644
index 000000000000..7f3b93c4d229
@@ -0,0 +1,5 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/list.h>
+#include <pmus.h>
+
+LIST_HEAD(pmus);
tools/perf/util/pmus.h new file +9 -0
new file mode 100644
index 000000000000..5ec12007eb5c
@@ -0,0 +1,9 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef __PMUS_H
+#define __PMUS_H
+
+extern struct list_head pmus;
+
+#define perf_pmus__for_each_pmu(pmu) list_for_each_entry(pmu, &pmus, list)
+
+#endif /* __PMUS_H */
Signature · 1 line
2.38.1
#2 lore raw

PATCH v3 2/2 perf test: Add event group test

In reply to Ravi Bangoria

Multiple events in a group can belong to one or more pmus, however there are some limitations to it. One of the limitation is, perf doesn't allow creating a group of events from different hw pmus. Write a simple test to create various combinations of hw, sw and uncore pmu events and verify group creation succeeds or fails as expected.

---
 tools/perf/tests/Build          |   1 +
 tools/perf/tests/builtin-test.c |   1 +
 tools/perf/tests/event_groups.c | 127 ++++++++++++++++++++++++++++++++
 tools/perf/tests/tests.h        |   1 +
 4 files changed, 130 insertions(+)
 create mode 100644 tools/perf/tests/event_groups.c
tools/perf/tests/Build +1 -0
index 11b69023011b..658b5052c24d 100644
@@ -67,6 +67,7 @@ perf-y += expand-cgroup.o
 perf-y += perf-time-to-tsc.o
 perf-y += dlfilter-test.o
 perf-y += sigtrap.o
+perf-y += event_groups.o
 
 $(OUTPUT)tests/llvm-src-base.c: tests/bpf-script-example.c tests/Build
 	$(call rule_mkdir)
tools/perf/tests/builtin-test.c +1 -0
index 4c6ae59a4dfd..ddd8262bfa26 100644
@@ -110,6 +110,7 @@ static struct test_suite *generic_tests[] = {
 	&suite__perf_time_to_tsc,
 	&suite__dlfilter,
 	&suite__sigtrap,
+	&suite__event_groups,
 	NULL,
 };
 
tools/perf/tests/event_groups.c new file +127 -0

127 lines not on this page · raw message

tools/perf/tests/tests.h +1 -0
index e15f24cfc909..fb4b5ad4dd0f 100644
@@ -147,6 +147,7 @@ DECLARE_SUITE(expand_cgroup_events);
 DECLARE_SUITE(perf_time_to_tsc);
 DECLARE_SUITE(dlfilter);
 DECLARE_SUITE(sigtrap);
+DECLARE_SUITE(event_groups);
 
 /*
  * PowerPC and S390 do not support creation of instruction breakpoints using the
Signature · 1 line
2.38.1
Acked-by on v3 2/2
#3 lore raw
Quoted 64 lines · Ravi Bangoria

Multiple events in a group can belong to one or more pmus, however there are some limitations to it. One of the limitation is, perf doesn't allow creating a group of events from different hw pmus. Write a simple test to create various combinations of hw, sw and uncore pmu events and verify group creation succeeds or fails as expected.

---
  tools/perf/tests/Build          |   1 +
  tools/perf/tests/builtin-test.c |   1 +
  tools/perf/tests/event_groups.c | 127 ++++++++++++++++++++++++++++++++
  tools/perf/tests/tests.h        |   1 +
  4 files changed, 130 insertions(+)

create mode 100644 tools/perf/tests/event_groups.c

tools/perf/tests/Build +1 -0
index 11b69023011b..658b5052c24d 100644
@@ -67,6 +67,7 @@ perf-y += expand-cgroup.o
  perf-y += perf-time-to-tsc.o
  perf-y += dlfilter-test.o
  perf-y += sigtrap.o
+perf-y += event_groups.o
 
  $(OUTPUT)tests/llvm-src-base.c: tests/bpf-script-example.c tests/Build
  	$(call rule_mkdir)
tools/perf/tests/builtin-test.c +1 -0
index 4c6ae59a4dfd..ddd8262bfa26 100644
@@ -110,6 +110,7 @@ static struct test_suite *generic_tests[] = {
  	&suite__perf_time_to_tsc,
  	&suite__dlfilter,
  	&suite__sigtrap,
+	&suite__event_groups,
  	NULL,
  };
 
tools/perf/tests/event_groups.c new file +18 -0
new file mode 100644
index 000000000000..612c0444aaa8
@@ -0,0 +1,127 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <string.h>
+#include <unistd.h>
+#include <stdio.h>
+#include "linux/perf_event.h"
+#include "tests.h"
+#include "debug.h"
+#include "pmu.h"
+#include "pmus.h"
+#include "header.h"
+#include "../perf-sys.h"
+
+/* hw: cycles, sw: context-switch, uncore: [arch dependent] */
+static int types[] = {0, 1, -1};
+static unsigned long configs[] = {0, 3, 0};
+
+#define NR_UNCORE_PMUS 5
+

tools/perf/tests/event_groups.c line 28

Quoted 10 lines
tools/perf/tests/event_groups.c new file +10 -0
+/* Uncore pmus that support more than 3 counters */
+static struct uncore_pmus {
+	const char *name;
+	__u64 config;
+} uncore_pmus[NR_UNCORE_PMUS] = {
+	{ "amd_l3", 0x0 },
+	{ "amd_df", 0x0 },
+	{ "uncore_imc_0", 0x1 },         /* Intel */
+	{ "core_imc", 0x318 },           /* PowerPC: core_imc/CPM_STCX_FIN/ */
+	{ "hv_24x7", 0x22000000003 },    /* PowerPC: hv_24x7/CPM_STCX_FIN/ */

For Powerpc event config values

Quoted 111 lines
tools/perf/tests/event_groups.c new file +99 -0
+};
+
+static int event_open(int type, unsigned long config, int group_fd)
+{
+	struct perf_event_attr attr;
+
+	memset(&attr, 0, sizeof(struct perf_event_attr));
+	attr.type = type;
+	attr.size = sizeof(struct perf_event_attr);
+	attr.config = config;
+	/*
+	 * When creating an event group, typically the group leader is
+	 * initialized with disabled set to 1 and any child events are
+	 * initialized with disabled set to 0. Despite disabled being 0,
+	 * the child events will not start until the group leader is
+	 * enabled.
+	 */
+	attr.disabled = group_fd == -1 ? 1 : 0;
+
+	return sys_perf_event_open(&attr, -1, 0, group_fd, 0);
+}
+
+static int setup_uncore_event(void)
+{
+	struct perf_pmu *pmu;
+	int i;
+
+	if (list_empty(&pmus))
+		perf_pmu__scan(NULL);
+
+	perf_pmus__for_each_pmu(pmu) {
+		for (i = 0; i < NR_UNCORE_PMUS; i++) {
+			if (!strcmp(uncore_pmus[i].name, pmu->name)) {
+				pr_debug("Using %s for uncore pmu event\n", pmu->name);
+				types[2] = pmu->type;
+				configs[2] = uncore_pmus[i].config;
+				return 0;
+			}
+		}
+	}
+	return -1;
+}
+
+static int run_test(int i, int j, int k)
+{
+	int erroneous = ((((1 << i) | (1 << j) | (1 << k)) & 5) == 5);
+	int group_fd, sibling_fd1, sibling_fd2;
+
+	group_fd = event_open(types[i], configs[i], -1);
+	if (group_fd == -1)
+		return -1;
+
+	sibling_fd1 = event_open(types[j], configs[j], group_fd);
+	if (sibling_fd1 == -1) {
+		close(group_fd);
+		return erroneous ? 0 : -1;
+	}
+
+	sibling_fd2 = event_open(types[k], configs[k], group_fd);
+	if (sibling_fd2 == -1) {
+		close(sibling_fd1);
+		close(group_fd);
+		return erroneous ? 0 : -1;
+	}
+
+	close(sibling_fd2);
+	close(sibling_fd1);
+	close(group_fd);
+	return erroneous ? -1 : 0;
+}
+
+static int test__event_groups(struct test_suite *text __maybe_unused, int subtest __maybe_unused)
+{
+	int i, j, k;
+	int ret;
+	int r;
+
+	ret = setup_uncore_event();
+	if (ret || types[2] == -1)
+		return TEST_SKIP;
+
+	ret = TEST_OK;
+	for (i = 0; i < 3; i++) {
+		for (j = 0; j < 3; j++) {
+			for (k = 0; k < 3; k++) {
+				r = run_test(i, j, k);
+				if (r)
+					ret = TEST_FAIL;
+
+				pr_debug("0x%x 0x%lx, 0x%x 0x%lx, 0x%x 0x%lx: %s\n",
+					 types[i], configs[i], types[j], configs[j],
+					 types[k], configs[k], r ? "Fail" : "Pass");
+			}
+		}
+	}
+	return ret;
+}
+
+DEFINE_SUITE("Event groups", event_groups);
tools/perf/tests/tests.h +1 -0
index e15f24cfc909..fb4b5ad4dd0f 100644
@@ -147,6 +147,7 @@ DECLARE_SUITE(expand_cgroup_events);
  DECLARE_SUITE(perf_time_to_tsc);
  DECLARE_SUITE(dlfilter);
  DECLARE_SUITE(sigtrap);
+DECLARE_SUITE(event_groups);
 
  /*
   * PowerPC and S390 do not support creation of instruction breakpoints using the
#4 lore raw
In reply to Ravi Bangoria
Quoted 104 lines · Ravi Bangoria

Multiple events in a group can belong to one or more pmus, however there are some limitations to it. One of the limitation is, perf doesn't allow creating a group of events from different hw pmus. Write a simple test to create various combinations of hw, sw and uncore pmu events and verify group creation succeeds or fails as expected.

---

tools/perf/tests/Build          |   1 +
tools/perf/tests/builtin-test.c |   1 +
tools/perf/tests/event_groups.c | 127 ++++++++++++++++++++++++++++++++
tools/perf/tests/tests.h        |   1 +
4 files changed, 130 insertions(+)

create mode 100644 tools/perf/tests/event_groups.c

tools/perf/tests/Build +0 -0
index 11b69023011b..658b5052c24d 100644
@@ -67,6 +67,7 @@ perf-y += expand-cgroup.o

perf-y += perf-time-to-tsc.o
perf-y += dlfilter-test.o
perf-y += sigtrap.o
+perf-y += event_groups.o

$(OUTPUT)tests/llvm-src-base.c: tests/bpf-script-example.c tests/Build

	$(call rule_mkdir)
tools/perf/tests/builtin-test.c +1 -0
index 4c6ae59a4dfd..ddd8262bfa26 100644
@@ -110,6 +110,7 @@ static struct test_suite *generic_tests[] = {
 	&suite__perf_time_to_tsc,
 	&suite__dlfilter,
 	&suite__sigtrap,
+	&suite__event_groups,
 	NULL,
 };
 
tools/perf/tests/event_groups.c new file +54 -0
new file mode 100644
index 000000000000..612c0444aaa8
@@ -0,0 +1,127 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <string.h>
+#include <unistd.h>
+#include <stdio.h>
+#include "linux/perf_event.h"
+#include "tests.h"
+#include "debug.h"
+#include "pmu.h"
+#include "pmus.h"
+#include "header.h"
+#include "../perf-sys.h"
+
+/* hw: cycles, sw: context-switch, uncore: [arch dependent] */
+static int types[] = {0, 1, -1};
+static unsigned long configs[] = {0, 3, 0};
+
+#define NR_UNCORE_PMUS 5
+
+/* Uncore pmus that support more than 3 counters */
+static struct uncore_pmus {
+	const char *name;
+	__u64 config;
+} uncore_pmus[NR_UNCORE_PMUS] = {
+	{ "amd_l3", 0x0 },
+	{ "amd_df", 0x0 },
+	{ "uncore_imc_0", 0x1 },         /* Intel */
+	{ "core_imc", 0x318 },           /* PowerPC: core_imc/CPM_STCX_FIN/ */
+	{ "hv_24x7", 0x22000000003 },    /* PowerPC: hv_24x7/CPM_STCX_FIN/ */
+};
+
+static int event_open(int type, unsigned long config, int group_fd)
+{
+	struct perf_event_attr attr;
+
+	memset(&attr, 0, sizeof(struct perf_event_attr));
+	attr.type = type;
+	attr.size = sizeof(struct perf_event_attr);
+	attr.config = config;
+	/*
+	 * When creating an event group, typically the group leader is
+	 * initialized with disabled set to 1 and any child events are
+	 * initialized with disabled set to 0. Despite disabled being 0,
+	 * the child events will not start until the group leader is
+	 * enabled.
+	 */
+	attr.disabled = group_fd == -1 ? 1 : 0;
+
+	return sys_perf_event_open(&attr, -1, 0, group_fd, 0);
+}
+
+static int setup_uncore_event(void)
+{
+	struct perf_pmu *pmu;
+	int i;

tools/perf/tests/event_groups.c line 64

Quoted 10 lines
tools/perf/tests/event_groups.c new file +10 -0
+
+	if (list_empty(&pmus))
+		perf_pmu__scan(NULL);
+
+	perf_pmus__for_each_pmu(pmu) {
+		for (i = 0; i < NR_UNCORE_PMUS; i++) {
+			if (!strcmp(uncore_pmus[i].name, pmu->name)) {
+				pr_debug("Using %s for uncore pmu event\n", pmu->name);
+				types[2] = pmu->type;
+				configs[2] = uncore_pmus[i].config;

Hi Ravi,

Observed failure while running the test on powerpc. It is because the uncore PMU ie hv_24x7 needs performance monitoring to be enabled in powerpc. So to handle such cases, can we add an “event_open" check before proceeding with the test. Below is the change on top of “tmp.perf/core” .

From 8b33fb900c26beafc28f75b6f64631f8fdd045c2 Mon Sep 17 00:00:00 2001
From: Athira Rajeev <[email protected]>
Date: Tue, 6 Dec 2022 20:17:25 +0530
Subject: [PATCH] perf test: Update event group check for support of uncore
event

Event group test checks group creation for combinations of hw, sw and uncore PMU events. Some of the uncore pmu event requires performance enablement explicitly. Example, hv_24x7 event in powerpc. Hence add a check to see if event_open succeeds before proceeding.

Fixes 5c88101b797d ("perf test: Add event group test for events in multiple PMUs")
---
 tools/perf/tests/event_groups.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
tools/perf/tests/event_groups.c +5 -1
index 612c0444aaa8..ad52e1da259a 100644
@@ -51,7 +51,7 @@ static int event_open(int type, unsigned long config, int group_fd)
 static int setup_uncore_event(void)
 {
 	struct perf_pmu *pmu;
-	int i;
+	int i, fd;
 
 	if (list_empty(&pmus))
 		perf_pmu__scan(NULL);
@@ -62,6 +62,10 @@ static int setup_uncore_event(void)
 				pr_debug("Using %s for uncore pmu event\n", pmu->name);
 				types[2] = pmu->type;
 				configs[2] = uncore_pmus[i].config;
+				fd = event_open(types[2], configs[2], -1);
+				if (fd < 0)
+					return -1;
+				close(fd);
 				return 0;
 			}
 		}
Signature · 7 lines
2.31.1


Ravi, Please share your comments on this. If this looks fine, I can post this as separate fix patch.

Thanks
Athira
Quoted 78 lines
tools/perf/tests/event_groups.c new file +63 -0
+				return 0;
+			}
+		}
+	}
+	return -1;
+}
+
+static int run_test(int i, int j, int k)
+{
+	int erroneous = ((((1 << i) | (1 << j) | (1 << k)) & 5) == 5);
+	int group_fd, sibling_fd1, sibling_fd2;
+
+	group_fd = event_open(types[i], configs[i], -1);
+	if (group_fd == -1)
+		return -1;
+
+	sibling_fd1 = event_open(types[j], configs[j], group_fd);
+	if (sibling_fd1 == -1) {
+		close(group_fd);
+		return erroneous ? 0 : -1;
+	}
+
+	sibling_fd2 = event_open(types[k], configs[k], group_fd);
+	if (sibling_fd2 == -1) {
+		close(sibling_fd1);
+		close(group_fd);
+		return erroneous ? 0 : -1;
+	}
+
+	close(sibling_fd2);
+	close(sibling_fd1);
+	close(group_fd);
+	return erroneous ? -1 : 0;
+}
+
+static int test__event_groups(struct test_suite *text __maybe_unused, int subtest __maybe_unused)
+{
+	int i, j, k;
+	int ret;
+	int r;
+
+	ret = setup_uncore_event();
+	if (ret || types[2] == -1)
+		return TEST_SKIP;
+
+	ret = TEST_OK;
+	for (i = 0; i < 3; i++) {
+		for (j = 0; j < 3; j++) {
+			for (k = 0; k < 3; k++) {
+				r = run_test(i, j, k);
+				if (r)
+					ret = TEST_FAIL;
+
+				pr_debug("0x%x 0x%lx, 0x%x 0x%lx, 0x%x 0x%lx: %s\n",
+					 types[i], configs[i], types[j], configs[j],
+					 types[k], configs[k], r ? "Fail" : "Pass");
+			}
+		}
+	}
+	return ret;
+}
+
+DEFINE_SUITE("Event groups", event_groups);
tools/perf/tests/tests.h +1 -0
index e15f24cfc909..fb4b5ad4dd0f 100644
@@ -147,6 +147,7 @@ DECLARE_SUITE(expand_cgroup_events);
 DECLARE_SUITE(perf_time_to_tsc);
 DECLARE_SUITE(dlfilter);
 DECLARE_SUITE(sigtrap);
+DECLARE_SUITE(event_groups);
 
 /*
 * PowerPC and S390 do not support creation of instruction breakpoints using the
Signature · 1 line
2.38.1
#5 lore raw

Hi Athira,

Quoted 28 lines
Quoted 4 lines
(quoted diff) +4 -0
+static int setup_uncore_event(void)
+{
+	struct perf_pmu *pmu;
+	int i;

Review

Quoted 10 lines
(quoted diff) +10 -0
+
+	if (list_empty(&pmus))
+		perf_pmu__scan(NULL);
+
+	perf_pmus__for_each_pmu(pmu) {
+		for (i = 0; i < NR_UNCORE_PMUS; i++) {
+			if (!strcmp(uncore_pmus[i].name, pmu->name)) {
+				pr_debug("Using %s for uncore pmu event\n", pmu->name);
+				types[2] = pmu->type;
+				configs[2] = uncore_pmus[i].config;

Hi Ravi,

Observed failure while running the test on powerpc. It is because the uncore PMU ie hv_24x7 needs performance monitoring to be enabled in powerpc. So to handle such cases, can we add an “event_open" check before proceeding with the test. Below is the change on top of “tmp.perf/core” .

From 8b33fb900c26beafc28f75b6f64631f8fdd045c2 Mon Sep 17 00:00:00 2001
From: Athira Rajeev <[email protected]>
Date: Tue, 6 Dec 2022 20:17:25 +0530
Subject: [PATCH] perf test: Update event group check for support of uncore
event

Quoted 3 lines

Event group test checks group creation for combinations of hw, sw and uncore PMU events. Some of the uncore pmu event requires performance enablement explicitly.

You need to open an event to activate hv_24x7 pmu?

Quoted 24 lines

Example, hv_24x7
event in powerpc. Hence add a check to see if event_open succeeds before proceeding.

Fixes 5c88101b797d ("perf test: Add event group test for events in multiple PMUs")
---
 tools/perf/tests/event_groups.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
tools/perf/tests/event_groups.c +1 -1
index 612c0444aaa8..ad52e1da259a 100644
@@ -51,7 +51,7 @@ static int event_open(int type, unsigned long config, int group_fd)
 static int setup_uncore_event(void)
 {
 	struct perf_pmu *pmu;
-	int i;
+	int i, fd;
 
 	if (list_empty(&pmus))
 		perf_pmu__scan(NULL);

tools/perf/tests/event_groups.c static int setup_uncore_event(void) · line 64

Quoted 3 lines
tools/perf/tests/event_groups.c +0 -0
@@ -62,6 +62,10 @@ static int setup_uncore_event(void)
 				pr_debug("Using %s for uncore pmu event\n", pmu->name);
 				types[2] = pmu->type;
 				configs[2] = uncore_pmus[i].config;

Sure. Just add a comment here to explain why are we opening a standalone event here.

tools/perf/tests/event_groups.c static int setup_uncore_event(void) · line 71

Quoted 7 lines
tools/perf/tests/event_groups.c +4 -0
+				fd = event_open(types[2], configs[2], -1);
+				if (fd < 0)
+					return -1;
+				close(fd);
 				return 0;
 			}
 		}

Thanks,
Ravi

#6 lore raw
Quoted 29 lines · Ravi Bangoria

Hi Athira,

Quoted 27 lines
Quoted 4 lines
(quoted diff) +4 -0
+static int setup_uncore_event(void)
+{
+	struct perf_pmu *pmu;
+	int i;

Review

Quoted 10 lines
(quoted diff) +10 -0
+
+	if (list_empty(&pmus))
+		perf_pmu__scan(NULL);
+
+	perf_pmus__for_each_pmu(pmu) {
+		for (i = 0; i < NR_UNCORE_PMUS; i++) {
+			if (!strcmp(uncore_pmus[i].name, pmu->name)) {
+				pr_debug("Using %s for uncore pmu event\n", pmu->name);
+				types[2] = pmu->type;
+				configs[2] = uncore_pmus[i].config;

Hi Ravi,

Observed failure while running the test on powerpc. It is because the uncore PMU ie hv_24x7 needs performance monitoring to be enabled in powerpc. So to handle such cases, can we add an “event_open" check before proceeding with the test. Below is the change on top of “tmp.perf/core” .

From 8b33fb900c26beafc28f75b6f64631f8fdd045c2 Mon Sep 17 00:00:00 2001
From: Athira Rajeev <[email protected]>
Date: Tue, 6 Dec 2022 20:17:25 +0530
Subject: [PATCH] perf test: Update event group check for support of uncore
event

Quoted 4 lines
Quoted 3 lines

Event group test checks group creation for combinations of hw, sw and uncore PMU events. Some of the uncore pmu event requires performance enablement explicitly.

You need to open an event to activate hv_24x7 pmu?

hv_24x7 pmu supports events which can also provide system-wide resource data and partition should have permissions to access those, hence the check.

Maddy

Quoted 40 lines
Quoted 25 lines

Example, hv_24x7
event in powerpc. Hence add a check to see if event_open succeeds before proceeding.

Fixes 5c88101b797d ("perf test: Add event group test for events in multiple PMUs")
---
  tools/perf/tests/event_groups.c | 6 +++++-
  1 file changed, 5 insertions(+), 1 deletion(-)
tools/perf/tests/event_groups.c +1 -1
index 612c0444aaa8..ad52e1da259a 100644
@@ -51,7 +51,7 @@ static int event_open(int type, unsigned long config, int group_fd)
  static int setup_uncore_event(void)
  {
  	struct perf_pmu *pmu;
-	int i;
+	int i, fd;
 
  	if (list_empty(&pmus))
  		perf_pmu__scan(NULL);

tools/perf/tests/event_groups.c static int setup_uncore_event(void) · line 64

Quoted 3 lines
tools/perf/tests/event_groups.c +0 -0
@@ -62,6 +62,10 @@ static int setup_uncore_event(void)
  				pr_debug("Using %s for uncore pmu event\n", pmu->name);
  				types[2] = pmu->type;
  				configs[2] = uncore_pmus[i].config;

Sure. Just add a comment here to explain why are we opening a standalone event here.

tools/perf/tests/event_groups.c static int setup_uncore_event(void) · line 71

Quoted 7 lines
tools/perf/tests/event_groups.c +4 -0
+				fd = event_open(types[2], configs[2], -1);
+				if (fd < 0)
+					return -1;
+				close(fd);
  				return 0;
  			}
  		}

Thanks,
Ravi

#7 lore raw
Quoted 7 lines
Quoted 4 lines
Quoted 3 lines

Event group test checks group creation for combinations of hw, sw and uncore PMU events. Some of the uncore pmu event requires performance enablement explicitly.

You need to open an event to activate hv_24x7 pmu?

hv_24x7 pmu supports events which can also provide system-wide resource data and partition should have permissions to access those, hence the check.

Ah ok. Got it.

Thanks,
Ravi

reviewer Acked-by on the v3 series
#8 lore raw
In reply to Ravi Bangoria
Quoted 8 lines · Ravi Bangoria

Multiple events in a group can belong to one or more pmus, however there are some limitations to it. One of the limitation is, perf doesn't allow creating a group of events from different hw pmus. Write a simple test to create various combinations of hw, sw and uncore pmu events and verify group creation succeeds or fails as expected.

Quoted 5 lines

v2: https://lore.kernel.org/r/[email protected]
v2->v3:
- Define a set of uncore pmus that supports more than 3 events. This will prevent false negative results by not picking random uncore pmu. Test will be skipped if no uncore pmu found.

Thanks!
Ian

Quoted 20 lines

Ravi Bangoria (2):
perf tool: Move pmus list variable to new a file perf test: Add event group test

 tools/perf/tests/Build          |   1 +
 tools/perf/tests/builtin-test.c |   1 +
 tools/perf/tests/event_groups.c | 127 ++++++++++++++++++++++++++++++++
 tools/perf/tests/tests.h        |   1 +
 tools/perf/util/Build           |   1 +
 tools/perf/util/pmu.c           |   2 +-
 tools/perf/util/pmus.c          |   5 ++
 tools/perf/util/pmus.h          |   9 +++
 8 files changed, 146 insertions(+), 1 deletion(-)
 create mode 100644 tools/perf/tests/event_groups.c
 create mode 100644 tools/perf/util/pmus.c
 create mode 100644 tools/perf/util/pmus.h

--
2.38.1

Acked-by on the v3 series
#9 lore raw
In reply to Ravi Bangoria
Quoted 7 lines · Ravi Bangoria

Multiple events in a group can belong to one or more pmus, however there are some limitations to it. One of the limitation is, perf doesn't allow creating a group of events from different hw pmus. Write a simple test to create various combinations of hw, sw and uncore pmu events and verify group creation succeeds or fails as expected.

Quoted 8 lines

v2: https://lore.kernel.org/r/[email protected]
v2->v3:
- Define a set of uncore pmus that supports more than 3 events. This will prevent false negative results by not picking random uncore pmu. Test will be skipped if no uncore pmu found.

Ravi Bangoria (2):
perf tool: Move pmus list variable to new a file

a new file

Other than the above typo, the patch series looks good to me.

Thanks,
Kan

Quoted 15 lines

perf test: Add event group test

 tools/perf/tests/Build          |   1 +
 tools/perf/tests/builtin-test.c |   1 +
 tools/perf/tests/event_groups.c | 127 ++++++++++++++++++++++++++++++++
 tools/perf/tests/tests.h        |   1 +
 tools/perf/util/Build           |   1 +
 tools/perf/util/pmu.c           |   2 +-
 tools/perf/util/pmus.c          |   5 ++
 tools/perf/util/pmus.h          |   9 +++
 8 files changed, 146 insertions(+), 1 deletion(-)
 create mode 100644 tools/perf/tests/event_groups.c
 create mode 100644 tools/perf/util/pmus.c
 create mode 100644 tools/perf/util/pmus.h
#10 lore raw

Em Tue, Dec 06, 2022 at 09:47:57AM -0500, Liang, Kan escreveu:

Quoted 14 lines · Ravi Bangoria
Quoted 13 lines

Multiple events in a group can belong to one or more pmus, however there are some limitations to it. One of the limitation is, perf doesn't allow creating a group of events from different hw pmus. Write a simple test to create various combinations of hw, sw and uncore pmu events and verify group creation succeeds or fails as expected.

v2: https://lore.kernel.org/r/[email protected]
v2->v3:
- Define a set of uncore pmus that supports more than 3 events. This will prevent false negative results by not picking random uncore pmu. Test will be skipped if no uncore pmu found.

Quoted 8 lines
Quoted 2 lines

Ravi Bangoria (2):
perf tool: Move pmus list variable to new a file

a new file

Other than the above typo, the patch series looks good to me.

Fixed that up, added your Acked-by, thanks.

- Arnaldo