Skip to contentSkip to conversation

perf list: Skip the invalid hybrid pmu

PATCH Superseded Jin Yao 3 participants 5 messages Jun 10, 2021 – Jul 7, 2021 perf events

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Jin Yao

author
#0 lore raw

PATCH 1/1 perf list: Skip the invalid hybrid pmu

On hybrid platform, such as Alderlake, if atom CPUs are offlined, the kernel still exports the sysfs path '/sys/devices/cpu_atom/' for 'cpu_atom' pmu but the file '/sys/devices/cpu_atom/cpus' is empty, which indicates this is an invalid pmu.

The perf-list needs to check and skip the invalid hybrid pmu.

Before:

  # perf list
  ...
  branch-instructions OR cpu_atom/branch-instructions/ [Kernel PMU event]
  branch-instructions OR cpu_core/branch-instructions/ [Kernel PMU event]
  branch-misses OR cpu_atom/branch-misses/           [Kernel PMU event]
  branch-misses OR cpu_core/branch-misses/           [Kernel PMU event]
  bus-cycles OR cpu_atom/bus-cycles/                 [Kernel PMU event]
  bus-cycles OR cpu_core/bus-cycles/                 [Kernel PMU event]
  ...

The cpu_atom events are still displayed even if atom CPUs are offlined.

After:

  # perf list
  ...
  branch-instructions OR cpu_core/branch-instructions/ [Kernel PMU event]
  branch-misses OR cpu_core/branch-misses/           [Kernel PMU event]
  bus-cycles OR cpu_core/bus-cycles/                 [Kernel PMU event]
  ...

Now only cpu_core events are displayed.

---
 tools/perf/util/pmu-hybrid.c | 11 +++++++++++
 tools/perf/util/pmu-hybrid.h |  2 ++
 tools/perf/util/pmu.c        |  3 +++
 3 files changed, 16 insertions(+)
tools/perf/util/pmu-hybrid.c +11 -0
index f51ccaac60ee..fcc1182f8fe5 100644
@@ -87,3 +87,14 @@ char *perf_pmu__hybrid_type_to_pmu(const char *type)
 	free(pmu_name);
 	return NULL;
 }
+
+bool perf_pmu__is_invalid_hybrid(const char *name)
+{
+	if (strncmp(name, "cpu_", 4))
+		return false;
+
+	if (perf_pmu__hybrid_mounted(name))
+		return false;
+
+	return true;
+}
tools/perf/util/pmu-hybrid.h +2 -0
index 2b186c26a43e..8261a312c854 100644
@@ -30,4 +30,6 @@ static inline int perf_pmu__hybrid_pmu_num(void)
 	return num;
 }
 
+bool perf_pmu__is_invalid_hybrid(const char *name);
+
 #endif /* __PMU_HYBRID_H */
tools/perf/util/pmu.c +3 -0
index 88c8ecdc60b0..281670e9c4bd 100644
@@ -1604,6 +1604,9 @@ void print_pmu_events(const char *event_glob, bool name_only, bool quiet_flag,
 	pmu = NULL;
 	j = 0;
 	while ((pmu = perf_pmu__scan(pmu)) != NULL) {
+		if (perf_pmu__is_invalid_hybrid(pmu->name))
+			continue;
+
 		list_for_each_entry(alias, &pmu->aliases, list) {
 			char *name = alias->desc ? alias->name :
 				format_alias(buf, sizeof(buf), pmu, alias);
Signature · 1 line
2.17.1
#1 lore raw
Quoted 4 lines · Jin Yao

On hybrid platform, such as Alderlake, if atom CPUs are offlined, the kernel still exports the sysfs path '/sys/devices/cpu_atom/' for 'cpu_atom' pmu but the file '/sys/devices/cpu_atom/cpus' is empty, which indicates this is an invalid pmu.

The patch looks good, but it would probably make sense to fix the kernel to avoid this too.

Of course the tools patch would still be needed for older kernels.

-Andi

author
#2 lore raw

Hi Arnaldo, hi Jiri,

Quoted 8 lines · Andi Kleen
Quoted 4 lines · Jin Yao

On hybrid platform, such as Alderlake, if atom CPUs are offlined, the kernel still exports the sysfs path '/sys/devices/cpu_atom/' for 'cpu_atom' pmu but the file '/sys/devices/cpu_atom/cpus' is empty, which indicates this is an invalid pmu.

Quoted 7 lines

The patch looks good, but it would probably make sense to fix the kernel to avoid this too.

Of course the tools patch would still be needed for older kernels.

-Andi

Any comments for this patch?

Thanks
Jin Yao

#3 lore raw
In reply to Jin Yao
Quoted 74 lines · Jin Yao

On hybrid platform, such as Alderlake, if atom CPUs are offlined, the kernel still exports the sysfs path '/sys/devices/cpu_atom/' for 'cpu_atom' pmu but the file '/sys/devices/cpu_atom/cpus' is empty, which indicates this is an invalid pmu.

The perf-list needs to check and skip the invalid hybrid pmu.

Before:

  # perf list
  ...
  branch-instructions OR cpu_atom/branch-instructions/ [Kernel PMU event]
  branch-instructions OR cpu_core/branch-instructions/ [Kernel PMU event]
  branch-misses OR cpu_atom/branch-misses/           [Kernel PMU event]
  branch-misses OR cpu_core/branch-misses/           [Kernel PMU event]
  bus-cycles OR cpu_atom/bus-cycles/                 [Kernel PMU event]
  bus-cycles OR cpu_core/bus-cycles/                 [Kernel PMU event]
  ...

The cpu_atom events are still displayed even if atom CPUs are offlined.

After:

  # perf list
  ...
  branch-instructions OR cpu_core/branch-instructions/ [Kernel PMU event]
  branch-misses OR cpu_core/branch-misses/           [Kernel PMU event]
  bus-cycles OR cpu_core/bus-cycles/                 [Kernel PMU event]
  ...

Now only cpu_core events are displayed.

---
 tools/perf/util/pmu-hybrid.c | 11 +++++++++++
 tools/perf/util/pmu-hybrid.h |  2 ++
 tools/perf/util/pmu.c        |  3 +++
 3 files changed, 16 insertions(+)
tools/perf/util/pmu-hybrid.c +11 -0
index f51ccaac60ee..fcc1182f8fe5 100644
@@ -87,3 +87,14 @@ char *perf_pmu__hybrid_type_to_pmu(const char *type)
 	free(pmu_name);
 	return NULL;
 }
+
+bool perf_pmu__is_invalid_hybrid(const char *name)
+{
+	if (strncmp(name, "cpu_", 4))
+		return false;
+
+	if (perf_pmu__hybrid_mounted(name))
+		return false;
+
+	return true;
+}
tools/perf/util/pmu-hybrid.h +2 -0
index 2b186c26a43e..8261a312c854 100644
@@ -30,4 +30,6 @@ static inline int perf_pmu__hybrid_pmu_num(void)
 	return num;
 }
 
+bool perf_pmu__is_invalid_hybrid(const char *name);
+
 #endif /* __PMU_HYBRID_H */
tools/perf/util/pmu.c +0 -0
index 88c8ecdc60b0..281670e9c4bd 100644

tools/perf/util/pmu.c void print_pmu_events(const char *event_glob, bool name_only, bool quiet_flag, · line 1608

Quoted 5 lines
tools/perf/util/pmu.c +2 -0
@@ -1604,6 +1604,9 @@ void print_pmu_events(const char *event_glob, bool name_only, bool quiet_flag,
 	pmu = NULL;
 	j = 0;
 	while ((pmu = perf_pmu__scan(pmu)) != NULL) {
+		if (perf_pmu__is_invalid_hybrid(pmu->name))
+			continue;

hum why not detect it in pmu_lookup early on
and not add that pmu at all?

thanks,
jirka

Quoted 7 lines
tools/perf/util/pmu.c +1 -0
+
 		list_for_each_entry(alias, &pmu->aliases, list) {
 			char *name = alias->desc ? alias->name :
 				format_alias(buf, sizeof(buf), pmu, alias);
Signature · 1 line
2.17.1
author
#4 lore raw

Hi Jiri,

Quoted 13 lines · Jiri Olsa

tools/perf/util/pmu.c void print_pmu_events(const char *event_glob, bool name_only, bool quiet_flag, · line 1608

Quoted 10 lines
tools/perf/util/pmu.c +2 -0
index 88c8ecdc60b0..281670e9c4bd 100644
@@ -1604,6 +1604,9 @@ void print_pmu_events(const char *event_glob, bool name_only, bool quiet_flag,
  	pmu = NULL;
  	j = 0;
  	while ((pmu = perf_pmu__scan(pmu)) != NULL) {
+		if (perf_pmu__is_invalid_hybrid(pmu->name))
+			continue;

hum why not detect it in pmu_lookup early on
and not add that pmu at all?

Quoted 3 lines

thanks,
jirka

Yes, it's a good idea that detects it in pmu_lookup, thanks!

Thanks
Jin Yao