Skip to contentSkip to conversation

perf tool_pmu: Add skip check in tool_pmu__event_to_str()

PATCH Applied [email protected] 3 participants 3 messages Feb 7, 2025 – Feb 12, 2025 perf events

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 perf tool_pmu: Add skip check in tool_pmu__event_to_str()

From: Kan Liang <[email protected]>

Some topdown related metrics may fail on hybrid machines.

$perf stat -M tma_frontend_bound
Cannot resolve IDs for tma_frontend_bound:
cpu_atom@TOPDOWN_FE_BOUND.ALL@ / (8 * cpu_atom@CPU_CLK_UNHALTED.CORE@)

In the find_tool_events(), the tool_pmu__event_to_str() is used to compare the tool_events. It only checks the event name, no PMU or arch. So the tool_events[TOOL_PMU__EVENT_SLOTS] is set to true, because the p-core Topdown metrics has "slots" event.
The tool_events is shared. So when parsing the e-core metrics, the "slots" is automatically added.

The "slots" event as a tool event should only be available on arm64. It has a different meaning on X86. The tool_pmu__skip_event() intends handle the case. Apply it for tool_pmu__event_to_str() as well.

There is a lack of sanity check in the expr__get_id(). Add the check.

Fixes 069057239a67 ("perf tool_pmu: Move expr literals to tool_pmu")
---
 tools/perf/util/expr.c     | 2 ++
 tools/perf/util/tool_pmu.c | 3 ++-
 2 files changed, 4 insertions(+), 1 deletion(-)
tools/perf/util/expr.c +2 -0
index c221dcce6666..6413537442aa 100644
@@ -215,6 +215,8 @@ int expr__add_ref(struct expr_parse_ctx *ctx, struct metric_ref *ref)
 int expr__get_id(struct expr_parse_ctx *ctx, const char *id,
 		 struct expr_id_data **data)
 {
+	if (!ctx || !id)
+		return -1;
 	return hashmap__find(ctx->ids, id, data) ? 0 : -1;
 }
 
tools/perf/util/tool_pmu.c +2 -1
index 4fb097578479..3a68debe7143 100644
@@ -62,7 +62,8 @@ int tool_pmu__num_skip_events(void)
 
 const char *tool_pmu__event_to_str(enum tool_pmu_event ev)
 {
-	if (ev > TOOL_PMU__EVENT_NONE && ev < TOOL_PMU__EVENT_MAX)
+	if ((ev > TOOL_PMU__EVENT_NONE && ev < TOOL_PMU__EVENT_MAX) &&
+	    !tool_pmu__skip_event(tool_pmu__event_names[ev]))
 		return tool_pmu__event_names[ev];
 
 	return NULL;
Signature · 1 line
2.38.1
reviewer Reviewed-by on v1 1/1
#1 lore raw
Quoted 20 lines · On Fri, Feb 7, 2025 at 7:28 AM <[email protected]> wrote:

From: Kan Liang <[email protected]>

Some topdown related metrics may fail on hybrid machines.

$perf stat -M tma_frontend_bound
Cannot resolve IDs for tma_frontend_bound:
cpu_atom@TOPDOWN_FE_BOUND.ALL@ / (8 * cpu_atom@CPU_CLK_UNHALTED.CORE@)

In the find_tool_events(), the tool_pmu__event_to_str() is used to compare the tool_events. It only checks the event name, no PMU or arch. So the tool_events[TOOL_PMU__EVENT_SLOTS] is set to true, because the p-core Topdown metrics has "slots" event.
The tool_events is shared. So when parsing the e-core metrics, the "slots" is automatically added.

The "slots" event as a tool event should only be available on arm64. It has a different meaning on X86. The tool_pmu__skip_event() intends handle the case. Apply it for tool_pmu__event_to_str() as well.

Quoted 5 lines

There is a lack of sanity check in the expr__get_id(). Add the check.

Fixes 069057239a67 ("perf tool_pmu: Move expr literals to tool_pmu")

Thanks Kan!

Ian

maintainer Applied
#2 lore raw
In reply to [email protected]
Quoted 6 lines · On Fri, 07 Feb 2025 07:28:44 -0800, [email protected] wrote:

Some topdown related metrics may fail on hybrid machines.

$perf stat -M tma_frontend_bound
Cannot resolve IDs for tma_frontend_bound:
cpu_atom@TOPDOWN_FE_BOUND.ALL@ / (8 * cpu_atom@CPU_CLK_UNHALTED.CORE@)

Quoted 8 lines

In the find_tool_events(), the tool_pmu__event_to_str() is used to compare the tool_events. It only checks the event name, no PMU or arch. So the tool_events[TOOL_PMU__EVENT_SLOTS] is set to true, because the p-core Topdown metrics has "slots" event.
The tool_events is shared. So when parsing the e-core metrics, the "slots" is automatically added.

[...]

Applied to perf-tools-next, thanks!

Best regards,
Namhyung