Skip to contentSkip to conversation

perf topdown: Don't require config1==0 for a slots event

PATCH New Zide Chen 2 participants 2 messages Sep 17 perf events

lore
Share & tools
lore

Zide Chen

author
#0 lore raw

PATCH 1/1 perf topdown: Don't require config1==0 for a slots event

arch_is_topdown_slots() requires config1 == 0, so it misidentified "cpu/slots,metrics_clear=1/" as not being the slots event, and an additional slots event gets inserted, which fails event scheduling since the group now needs two fixed counter 3.

  $ perf stat -e "{cpu/slots,metrics_clear=1/,cpu/topdown-retiring/}" -- sleep 1
  WARNING: events were regrouped to match PMUs
       <not counted>      slots
     <not supported>      cpu/slots,metrics_clear=1/
       <not counted>      cpu/topdown-retiring/

Drop the config1 == 0 check from arch_is_topdown_slots(). This is
safe: config1 has no bearing on the event's identifying config value (TOPDOWN_SLOTS, 0x0400).

The attr.type == PERF_TYPE_RAW check is what actually matters: only the core PMU is registered with PERF_TYPE_RAW, and on x86, 0x400 is guaranteed to be slots event on core CPU.

arch_is_topdown_metrics() keeps its own config1 == 0 check, since genuine metrics events have no "metrics_clear" and config1 remains 0.

Add a test verifying that an explicit "slots,metrics_clear=1" event in a group does not trigger an extra slots event injection.

Fixes 5b546de9cc17 ("perf topdown: Use attribute to see an event is a topdown metic or slots")

---
v2: include stdio.h and string.h for musl compatibility.

---
 tools/perf/arch/x86/tests/topdown.c | 30 +++++++++++++++++++++++++++++
 tools/perf/arch/x86/util/topdown.c  |  3 +--
 2 files changed, 31 insertions(+), 2 deletions(-)
tools/perf/arch/x86/tests/topdown.c +30 -0
index 2b6f47ce4932..520075329c37 100644
@@ -1,5 +1,7 @@
 // SPDX-License-Identifier: GPL-2.0
 #include <errno.h>
+#include <stdio.h>
+#include <string.h>
 #include "arch-tests.h"
 #include "../util/topdown.h"
 #include "debug.h"
@@ -229,10 +231,38 @@ static int test__x86_topdown_slots_injection(struct test_suite *test __maybe_unu
 	return TEST_OK;
 }
 
+/*
+ * An explicit "slots,metrics_clear=1" event is still the slots event and
+ * must not cause an extra slots event to be injected into the group.
+ */
+static int test__x86_topdown_metrics_clear(struct test_suite *test __maybe_unused,
+					   int subtest __maybe_unused)
+{
+	struct perf_pmu *pmu;
+	char event_str[128];
+	int ret;
+
+	if (!topdown_sys_has_perf_metrics())
+		return TEST_OK;
+
+	pmu = perf_pmus__find_by_type(PERF_TYPE_RAW);
+	if (!pmu || !perf_pmu__has_format(pmu, "metrics_clear"))
+		return TEST_OK;
+
+	snprintf(event_str, sizeof(event_str),
+		 "{%s/slots,metrics_clear=1/,%s/topdown-retiring/}",
+		 pmu->name, pmu->name);
+	ret = test_sort(event_str, 2, 1);
+	TEST_ASSERT_EQUAL("explicit metrics_clear slots event isn't duplicated", ret, TEST_OK);
+
+	return TEST_OK;
+}
+
 static struct test_case x86_topdown_tests[] = {
 	TEST_CASE("topdown events", x86_topdown),
 	TEST_CASE("topdown sorting", x86_topdown_sorting),
 	TEST_CASE("topdown slots injection", x86_topdown_slots_injection),
+	TEST_CASE("topdown metrics_clear no extra slots", x86_topdown_metrics_clear),
 	{ .name = NULL, }
 };
 
tools/perf/arch/x86/util/topdown.c +1 -2
index bafd285119d7..64bc64e335a2 100644
@@ -37,8 +37,7 @@ bool topdown_sys_has_perf_metrics(void)
 bool arch_is_topdown_slots(const struct evsel *evsel)
 {
 	return evsel->core.attr.type == PERF_TYPE_RAW &&
-	       evsel->core.attr.config == TOPDOWN_SLOTS &&
-	       evsel->core.attr.config1 == 0;
+	       evsel->core.attr.config == TOPDOWN_SLOTS;
 }
 
 bool arch_is_topdown_metrics(const struct evsel *evsel)
Signature · 1 line
2.55.0
#1 lore raw
Sashiko has reviewed this patch and found no issues. It looks great!
Quoted 19 lines

arch_is_topdown_slots() requires config1 == 0, so it misidentified "cpu/slots,metrics_clear=1/" as not being the slots event, and an additional slots event gets inserted, which fails event scheduling since the group now needs two fixed counter 3.

  $ perf stat -e "{cpu/slots,metrics_clear=1/,cpu/topdown-retiring/}" -- sleep 1
  WARNING: events were regrouped to match PMUs
       <not counted>      slots
     <not supported>      cpu/slots,metrics_clear=1/
       <not counted>      cpu/topdown-retiring/

Drop the config1 == 0 check from arch_is_topdown_slots(). This is
safe: config1 has no bearing on the event's identifying config value (TOPDOWN_SLOTS, 0x0400).

The attr.type == PERF_TYPE_RAW check is what actually matters: only the core PMU is registered with PERF_TYPE_RAW, and on x86, 0x400 is guaranteed to be slots event on core CPU.

Quoted 8 lines

arch_is_topdown_metrics() keeps its own config1 == 0 check, since genuine metrics events have no "metrics_clear" and config1 remains 0.

Add a test verifying that an explicit "slots,metrics_clear=1" event in a group does not trigger an extra slots event injection.

Fixes 5b546de9cc17 ("perf topdown: Use attribute to see an event is a topdown metic or slots")

Sashiko has reviewed this patch and found no issues. It looks great!

Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1