Skip to contentSkip to conversation

perf script: Fix output type for dynamically allocated core PMU's

PATCH Applied Thomas Falcon 3 participants 3 messages Mar 5, 2025 – Mar 6, 2025 perf events

lore
Share & tools
lore
#0 lore raw

PATCH v6 1/1 perf script: Fix output type for dynamically allocated core PMU's

This patch was originally posted here:

https://lore.kernel.org/all/[email protected]/

I have rebased on top of Arnaldo's patch here:

https://lore.kernel.org/all/Z2XCi3PgstSrV0SE@x1/

The original commit message:
"
perf script output may show different fields on different core PMU's that exist on heterogeneous platforms. For example,

perf record -e "{cpu_core/mem-loads-aux/,cpu_core/event=0xcd,\
umask=0x01,ldlat=3,name=MEM_UOPS_RETIRED.LOAD_LATENCY/}:upp"\
-c10000 -W -d -a -- sleep 1

perf script:

chromium-browse 46572 [002] 544966.882384: 10000 cpu_core/MEM_UOPS_RETIRED.LOAD_LATENCY/: 7ffdf1391b0c 10268100142 \
|OP LOAD|LVL L1 hit|SNP None|TLB L1 or L2 hit|LCK No|BLK N/A 5 7 0 7fad7c47425d [unknown] (/usr/lib64/libglib-2.0.so.0.8000.3)

perf record -e cpu_atom/event=0xd0,umask=0x05,ldlat=3,\
name=MEM_UOPS_RETIRED.LOAD_LATENCY/upp -c10000 -W -d -a -- sleep 1

perf script:

gnome-control-c 534224 [023] 544951.816227: 10000 cpu_atom/MEM_UOPS_RETIRED.LOAD_LATENCY/: 7f0aaaa0aae0 [unknown] (/usr/lib64/libglib-2.0.so.0.8000.3)

Some fields, such as data_src, are not included by default.

The cause is that while one PMU may be assigned a type such as PERF_TYPE_RAW, other core PMU's are dynamically allocated at boot time. If this value does not match an existing PERF_TYPE_X value, output_type(perf_event_attr.type) will return OUTPUT_TYPE_OTHER.

Instead search for a core PMU with a matching perf_event_attr type and, if one is found, return PERF_TYPE_RAW to match output of other core PMU's.
"

---
v2: restrict pmu lookup to platforms with more than one core pmu
v3: only scan core pmu list
v4: rebase on top of Arnaldo's patch
v5: update based on Namhyung's feedback here
https://lore.kernel.org/lkml/[email protected]/
v6: update based on Ian's feedback here
https://lore.kernel.org/linux-perf-users/CAP-5=fVwrG5ZViysO7qKYUw0326WFfa7nOzwe70drYfHq=AXhA@mail.gmail.com/
---

 tools/perf/builtin-script.c | 16 +++++++++++++---
 1 file changed, 13 insertions(+), 3 deletions(-)
tools/perf/builtin-script.c +13 -3
index d797cec4f054..253c4ca395e1 100644
@@ -400,10 +400,20 @@ static inline int output_type(unsigned int type)
 
 static inline int evsel__output_type(struct evsel *evsel)
 {
-	if (evsel->script_output_type == OUTPUT_TYPE_UNSET)
-		evsel->script_output_type = output_type(evsel->core.attr.type);
+	int type = evsel->script_output_type;
 
-	return evsel->script_output_type;
+	if (type == OUTPUT_TYPE_UNSET) {
+		type = output_type(evsel->core.attr.type);
+		if (type == OUTPUT_TYPE_OTHER) {
+			struct perf_pmu *pmu = evsel__find_pmu(evsel);
+
+			if (pmu && pmu->is_core)
+				type = PERF_TYPE_RAW;
+		}
+		evsel->script_output_type = type;
+	}
+
+	return type;
 }
 
 static bool output_set_by_user(void)
Signature · 1 line
2.48.1
reviewer Reviewed-by on v6 1/1
#1 lore raw
Quoted 37 lines · Thomas Falcon

This patch was originally posted here:

https://lore.kernel.org/all/[email protected]/

I have rebased on top of Arnaldo's patch here:

https://lore.kernel.org/all/Z2XCi3PgstSrV0SE@x1/

The original commit message:
"
perf script output may show different fields on different core PMU's that exist on heterogeneous platforms. For example,

perf record -e "{cpu_core/mem-loads-aux/,cpu_core/event=0xcd,\
umask=0x01,ldlat=3,name=MEM_UOPS_RETIRED.LOAD_LATENCY/}:upp"\
-c10000 -W -d -a -- sleep 1

perf script:

chromium-browse 46572 [002] 544966.882384: 10000 cpu_core/MEM_UOPS_RETIRED.LOAD_LATENCY/: 7ffdf1391b0c 10268100142 \
|OP LOAD|LVL L1 hit|SNP None|TLB L1 or L2 hit|LCK No|BLK N/A 5 7 0 7fad7c47425d [unknown] (/usr/lib64/libglib-2.0.so.0.8000.3)

perf record -e cpu_atom/event=0xd0,umask=0x05,ldlat=3,\
name=MEM_UOPS_RETIRED.LOAD_LATENCY/upp -c10000 -W -d -a -- sleep 1

perf script:

gnome-control-c 534224 [023] 544951.816227: 10000 cpu_atom/MEM_UOPS_RETIRED.LOAD_LATENCY/: 7f0aaaa0aae0 [unknown] (/usr/lib64/libglib-2.0.so.0.8000.3)

Some fields, such as data_src, are not included by default.

The cause is that while one PMU may be assigned a type such as PERF_TYPE_RAW, other core PMU's are dynamically allocated at boot time. If this value does not match an existing PERF_TYPE_X value, output_type(perf_event_attr.type) will return OUTPUT_TYPE_OTHER.

Quoted 8 lines

Instead search for a core PMU with a matching perf_event_attr type and, if one is found, return PERF_TYPE_RAW to match output of other core PMU's.
"

Thanks,
Ian

Quoted 44 lines

---
v2: restrict pmu lookup to platforms with more than one core pmu
v3: only scan core pmu list
v4: rebase on top of Arnaldo's patch
v5: update based on Namhyung's feedback here
https://lore.kernel.org/lkml/[email protected]/
v6: update based on Ian's feedback here
https://lore.kernel.org/linux-perf-users/CAP-5=fVwrG5ZViysO7qKYUw0326WFfa7nOzwe70drYfHq=AXhA@mail.gmail.com/
---

 tools/perf/builtin-script.c | 16 +++++++++++++---
 1 file changed, 13 insertions(+), 3 deletions(-)
tools/perf/builtin-script.c +13 -4
index d797cec4f054..253c4ca395e1 100644
@@ -400,10 +400,20 @@ static inline int output_type(unsigned int type)
 
 static inline int evsel__output_type(struct evsel *evsel)
 {
-       if (evsel->script_output_type == OUTPUT_TYPE_UNSET)
-               evsel->script_output_type = output_type(evsel->core.attr.type);
+       int type = evsel->script_output_type;
 
-       return evsel->script_output_type;
+       if (type == OUTPUT_TYPE_UNSET) {
+               type = output_type(evsel->core.attr.type);
+               if (type == OUTPUT_TYPE_OTHER) {
+                       struct perf_pmu *pmu = evsel__find_pmu(evsel);
+
+                       if (pmu && pmu->is_core)
+                               type = PERF_TYPE_RAW;
+               }
+               evsel->script_output_type = type;
+       }
+
+       return type;
 }
 
 static bool output_set_by_user(void)
--

2.48.1