Skip to contentSkip to conversation

perf tool_pmu: Fix aggregation on duration_time

PATCH Applied Ian Rogers 3 participants 3 messages Apr 23, 2025 – Apr 24, 2025 perf events

lore
Share & tools
lore

Ian Rogers

author reviewer
#0 lore raw

PATCH 1/1 perf tool_pmu: Fix aggregation on duration_time

evsel__count_has_error fails counters when the enabled or running time are 0. The duration_time event reads 0 when the cpu_map_idx != 0 to avoid aggregating time over CPUs. Change the enable and running time to always have a ratio of 100% so that evsel__count_has_error won't fail.

Before:

```
$ sudo /tmp/perf/perf stat --per-core -a -M UNCORE_FREQ sleep 1

 Performance counter stats for 'system wide':

S0-D0-C0              1      2,615,819,485      UNC_CLOCK.SOCKET                 #     2.61 UNCORE_FREQ
S0-D0-C0              2      <not counted>      duration_time

       1.002111784 seconds time elapsed
```

After:

```
$ perf stat --per-core -a -M UNCORE_FREQ sleep 1

 Performance counter stats for 'system wide':

S0-D0-C0              1        758,160,296      UNC_CLOCK.SOCKET                 #     0.76 UNCORE_FREQ
S0-D0-C0              2      1,003,438,246      duration_time

       1.002486017 seconds time elapsed
```

Note: the metric reads the value a different way and isn't impacted.

Fixes 240505b2d0ad ("perf tool_pmu: Factor tool events into their own PMU")
---
 tools/perf/util/tool_pmu.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
tools/perf/util/tool_pmu.c +7 -1
index 97b327d1ce4a..727a10e3f990 100644
@@ -486,8 +486,14 @@ int evsel__tool_pmu_read(struct evsel *evsel, int cpu_map_idx, int thread)
 		delta_start *= 1000000000 / ticks_per_sec;
 	}
 	count->val    = delta_start;
-	count->ena    = count->run = delta_start;
 	count->lost   = 0;
+	/*
+	 * The values of enabled and running must make a ratio of 100%. The
+	 * exact values don't matter as long as they are non-zero to avoid
+	 * issues with evsel__count_has_error.
+	 */
+	count->ena++;
+	count->run++;
 	return 0;
 }
 
Signature · 1 line
2.49.0.805.g082f7c87e0-goog
reviewer Reviewed-by on v1 1/1
#1 lore raw
Quoted 49 lines · Ian Rogers

evsel__count_has_error fails counters when the enabled or running time are 0. The duration_time event reads 0 when the cpu_map_idx != 0 to avoid aggregating time over CPUs. Change the enable and running time to always have a ratio of 100% so that evsel__count_has_error won't fail.

Before:

```
$ sudo /tmp/perf/perf stat --per-core -a -M UNCORE_FREQ sleep 1

  Performance counter stats for 'system wide':

S0-D0-C0              1      2,615,819,485      UNC_CLOCK.SOCKET                 #     2.61 UNCORE_FREQ
S0-D0-C0              2      <not counted>      duration_time

        1.002111784 seconds time elapsed
```

After:

```
$ perf stat --per-core -a -M UNCORE_FREQ sleep 1

  Performance counter stats for 'system wide':

S0-D0-C0              1        758,160,296      UNC_CLOCK.SOCKET                 #     0.76 UNCORE_FREQ
S0-D0-C0              2      1,003,438,246      duration_time

        1.002486017 seconds time elapsed
```

Note: the metric reads the value a different way and isn't impacted.

Fixes 240505b2d0ad ("perf tool_pmu: Factor tool events into their own PMU")
---
  tools/perf/util/tool_pmu.c | 8 +++++++-
  1 file changed, 7 insertions(+), 1 deletion(-)
tools/perf/util/tool_pmu.c +0 -1
index 97b327d1ce4a..727a10e3f990 100644
@@ -486,8 +486,14 @@ int evsel__tool_pmu_read(struct evsel *evsel, int cpu_map_idx, int thread)
  		delta_start *= 1000000000 / ticks_per_sec;
  	}
  	count->val    = delta_start;
-	count->ena    = count->run = delta_start;
  	count->lost   = 0;

tools/perf/util/tool_pmu.c int evsel__tool_pmu_read(struct evsel *evsel, int cpu_map_idx, int thread) · line 499

Quoted 10 lines
tools/perf/util/tool_pmu.c +7 -0
+	/*
+	 * The values of enabled and running must make a ratio of 100%. The
+	 * exact values don't matter as long as they are non-zero to avoid
+	 * issues with evsel__count_has_error.
+	 */
+	count->ena++;
+	count->run++;
  	return 0;
  }
 
maintainer Applied
#2 lore raw
Quoted 10 lines · James Clark
Quoted 10 lines

+++ b/tools/perf/util/tool_pmu.c

tools/perf/util/tool_pmu.c +3 -1
@@ -486,8 +486,14 @@ int evsel__tool_pmu_read(struct evsel *evsel, int cpu_map_idx, int thread)
  		delta_start *= 1000000000 / ticks_per_sec;
  	}
  	count->val    = delta_start;
-	count->ena    = count->run = delta_start;
  	count->lost   = 0;
+	/*
+	 * The values of enabled and running must make a ratio of 100%. The
+	 * exact values don't matter as long as they are non-zero to avoid
Quoted 8 lines

tools/perf/util/tool_pmu.c int evsel__tool_pmu_read(struct evsel *evsel, int cpu_map_idx, int thread) · line 498

Quoted 6 lines
tools/perf/util/tool_pmu.c +4 -0
+	 * issues with evsel__count_has_error.
+	 */
+	count->ena++;
+	count->run++;
  	return 0;
  }

Thanks, applied to perf-tools-next,

- Arnaldo