[PATCH 3/3] perf list: Don't deduplicate core PMUs when listing events

James Clark posted 3 patches 11 months, 1 week ago
[PATCH 3/3] perf list: Don't deduplicate core PMUs when listing events
Posted by James Clark 11 months, 1 week ago
Commit 7afbf90ea2e2 ("perf pmu: Don't de-duplicate core PMUs") fixed a
display mismatch related to deduplication within a single PMU, but it
didn't fix the case where deduplicated PMUs aren't listed at all.

Fix it by using the same function which takes is_core into account,
except in the use_core_pmus block where it's always going to be true.
Before this change, -v would be required to get the same behavior for
core PMUs. Now it's no longer required:

Before:
 $ perf list | grep br_indirect_spec -A 1
 br_indirect_spec
    [Branch speculatively executed,indirect branch. Unit: armv8_cortex_a53]

After:
 $ perf list | grep br_indirect_spec -A 2
    [Branch speculatively executed,indirect branch. Unit: armv8_cortex_a53,
     armv8_cortex_a57]

Signed-off-by: James Clark <james.clark@linaro.org>
---
 tools/perf/util/pmu.c  | 5 +++--
 tools/perf/util/pmu.h  | 2 ++
 tools/perf/util/pmus.c | 8 +++++---
 3 files changed, 10 insertions(+), 5 deletions(-)

diff --git a/tools/perf/util/pmu.c b/tools/perf/util/pmu.c
index 57450c73fb63..caff0d309012 100644
--- a/tools/perf/util/pmu.c
+++ b/tools/perf/util/pmu.c
@@ -834,9 +834,10 @@ static int is_sysfs_pmu_core(const char *name)
  *
  * @skip_duplicate_pmus: False in verbose mode so all uncore PMUs are visible
  */
-static size_t pmu_deduped_name_len(const struct perf_pmu *pmu, const char *name,
-				   bool skip_duplicate_pmus)
+size_t pmu_deduped_name_len(const struct perf_pmu *pmu, const char *name,
+			    bool skip_duplicate_pmus)
 {
+	name = name ?: "";
 	return skip_duplicate_pmus && !pmu->is_core
 		? pmu_name_len_no_suffix(name)
 		: strlen(name);
diff --git a/tools/perf/util/pmu.h b/tools/perf/util/pmu.h
index b93014cc3670..ce6a394a695d 100644
--- a/tools/perf/util/pmu.h
+++ b/tools/perf/util/pmu.h
@@ -297,5 +297,7 @@ struct perf_pmu *perf_pmus__find_core_pmu(void);
 
 const char *perf_pmu__name_from_config(struct perf_pmu *pmu, u64 config);
 bool perf_pmu__is_fake(const struct perf_pmu *pmu);
+size_t pmu_deduped_name_len(const struct perf_pmu *pmu, const char *name,
+			    bool skip_duplicate_pmus);
 
 #endif /* __PMU_H */
diff --git a/tools/perf/util/pmus.c b/tools/perf/util/pmus.c
index cb1b14ade25b..1acc27af4d02 100644
--- a/tools/perf/util/pmus.c
+++ b/tools/perf/util/pmus.c
@@ -358,12 +358,14 @@ static struct perf_pmu *perf_pmus__scan_skip_duplicates(struct perf_pmu *pmu)
 	if (!pmu) {
 		pmu_read_sysfs(PERF_TOOL_PMU_TYPE_ALL_MASK);
 		pmu = list_prepare_entry(pmu, &core_pmus, list);
-	} else
-		last_pmu_name_len = pmu_name_len_no_suffix(pmu->name ?: "");
+	} else {
+		last_pmu_name_len = pmu_deduped_name_len(pmu, pmu->name,
+							 /*skip_duplicate_pmus=*/true);
+	}
 
 	if (use_core_pmus) {
 		list_for_each_entry_continue(pmu, &core_pmus, list) {
-			int pmu_name_len = pmu_name_len_no_suffix(pmu->name ?: "");
+			int pmu_name_len = strlen(pmu->name ?: "");
 
 			if (last_pmu_name_len == pmu_name_len &&
 			    !strncmp(last_pmu_name, pmu->name ?: "", pmu_name_len))

-- 
2.34.1
Re: [PATCH 3/3] perf list: Don't deduplicate core PMUs when listing events
Posted by Ian Rogers 11 months, 1 week ago
On Tue, Mar 4, 2025 at 5:50 AM James Clark <james.clark@linaro.org> wrote:
>
> Commit 7afbf90ea2e2 ("perf pmu: Don't de-duplicate core PMUs") fixed a
> display mismatch related to deduplication within a single PMU, but it
> didn't fix the case where deduplicated PMUs aren't listed at all.
>
> Fix it by using the same function which takes is_core into account,
> except in the use_core_pmus block where it's always going to be true.
> Before this change, -v would be required to get the same behavior for
> core PMUs. Now it's no longer required:
>
> Before:
>  $ perf list | grep br_indirect_spec -A 1
>  br_indirect_spec
>     [Branch speculatively executed,indirect branch. Unit: armv8_cortex_a53]
>
> After:
>  $ perf list | grep br_indirect_spec -A 2
>     [Branch speculatively executed,indirect branch. Unit: armv8_cortex_a53,
>      armv8_cortex_a57]
>
> Signed-off-by: James Clark <james.clark@linaro.org>
> ---
>  tools/perf/util/pmu.c  | 5 +++--
>  tools/perf/util/pmu.h  | 2 ++
>  tools/perf/util/pmus.c | 8 +++++---
>  3 files changed, 10 insertions(+), 5 deletions(-)
>
> diff --git a/tools/perf/util/pmu.c b/tools/perf/util/pmu.c
> index 57450c73fb63..caff0d309012 100644
> --- a/tools/perf/util/pmu.c
> +++ b/tools/perf/util/pmu.c
> @@ -834,9 +834,10 @@ static int is_sysfs_pmu_core(const char *name)
>   *
>   * @skip_duplicate_pmus: False in verbose mode so all uncore PMUs are visible
>   */
> -static size_t pmu_deduped_name_len(const struct perf_pmu *pmu, const char *name,
> -                                  bool skip_duplicate_pmus)
> +size_t pmu_deduped_name_len(const struct perf_pmu *pmu, const char *name,
> +                           bool skip_duplicate_pmus)

nit: I think the name should be perf_pmu__deduped_name_len for
consistency with the other non-static functions.

>  {
> +       name = name ?: "";

nit: Should this just use pmu->name ?

>         return skip_duplicate_pmus && !pmu->is_core
>                 ? pmu_name_len_no_suffix(name)
>                 : strlen(name);
> diff --git a/tools/perf/util/pmu.h b/tools/perf/util/pmu.h
> index b93014cc3670..ce6a394a695d 100644
> --- a/tools/perf/util/pmu.h
> +++ b/tools/perf/util/pmu.h
> @@ -297,5 +297,7 @@ struct perf_pmu *perf_pmus__find_core_pmu(void);
>
>  const char *perf_pmu__name_from_config(struct perf_pmu *pmu, u64 config);
>  bool perf_pmu__is_fake(const struct perf_pmu *pmu);
> +size_t pmu_deduped_name_len(const struct perf_pmu *pmu, const char *name,
> +                           bool skip_duplicate_pmus);
>
>  #endif /* __PMU_H */
> diff --git a/tools/perf/util/pmus.c b/tools/perf/util/pmus.c
> index cb1b14ade25b..1acc27af4d02 100644
> --- a/tools/perf/util/pmus.c
> +++ b/tools/perf/util/pmus.c
> @@ -358,12 +358,14 @@ static struct perf_pmu *perf_pmus__scan_skip_duplicates(struct perf_pmu *pmu)
>         if (!pmu) {
>                 pmu_read_sysfs(PERF_TOOL_PMU_TYPE_ALL_MASK);
>                 pmu = list_prepare_entry(pmu, &core_pmus, list);
> -       } else
> -               last_pmu_name_len = pmu_name_len_no_suffix(pmu->name ?: "");
> +       } else {
> +               last_pmu_name_len = pmu_deduped_name_len(pmu, pmu->name,
> +                                                        /*skip_duplicate_pmus=*/true);
> +       }
>
>         if (use_core_pmus) {
>                 list_for_each_entry_continue(pmu, &core_pmus, list) {
> -                       int pmu_name_len = pmu_name_len_no_suffix(pmu->name ?: "");
> +                       int pmu_name_len = strlen(pmu->name ?: "");
>
>                         if (last_pmu_name_len == pmu_name_len &&
>                             !strncmp(last_pmu_name, pmu->name ?: "", pmu_name_len))

Can this code be removed given there shouldn't be core PMUs with
identical names? ie:
```
if (use_core_pmus) {
    list_for_each_entry_continue(pmu, &core_pmus, list)
        return pmu;

    pmu = NULL;
    pmu = list_prepare_entry(pmu, &other_pmus, list);
}
```

Thanks,
Ian

>
> --
> 2.34.1
>
Re: [PATCH 3/3] perf list: Don't deduplicate core PMUs when listing events
Posted by James Clark 11 months, 1 week ago

On 05/03/2025 9:51 pm, Ian Rogers wrote:
> On Tue, Mar 4, 2025 at 5:50 AM James Clark <james.clark@linaro.org> wrote:
>>
>> Commit 7afbf90ea2e2 ("perf pmu: Don't de-duplicate core PMUs") fixed a
>> display mismatch related to deduplication within a single PMU, but it
>> didn't fix the case where deduplicated PMUs aren't listed at all.
>>
>> Fix it by using the same function which takes is_core into account,
>> except in the use_core_pmus block where it's always going to be true.
>> Before this change, -v would be required to get the same behavior for
>> core PMUs. Now it's no longer required:
>>
>> Before:
>>   $ perf list | grep br_indirect_spec -A 1
>>   br_indirect_spec
>>      [Branch speculatively executed,indirect branch. Unit: armv8_cortex_a53]
>>
>> After:
>>   $ perf list | grep br_indirect_spec -A 2
>>      [Branch speculatively executed,indirect branch. Unit: armv8_cortex_a53,
>>       armv8_cortex_a57]
>>
>> Signed-off-by: James Clark <james.clark@linaro.org>
>> ---
>>   tools/perf/util/pmu.c  | 5 +++--
>>   tools/perf/util/pmu.h  | 2 ++
>>   tools/perf/util/pmus.c | 8 +++++---
>>   3 files changed, 10 insertions(+), 5 deletions(-)
>>
>> diff --git a/tools/perf/util/pmu.c b/tools/perf/util/pmu.c
>> index 57450c73fb63..caff0d309012 100644
>> --- a/tools/perf/util/pmu.c
>> +++ b/tools/perf/util/pmu.c
>> @@ -834,9 +834,10 @@ static int is_sysfs_pmu_core(const char *name)
>>    *
>>    * @skip_duplicate_pmus: False in verbose mode so all uncore PMUs are visible
>>    */
>> -static size_t pmu_deduped_name_len(const struct perf_pmu *pmu, const char *name,
>> -                                  bool skip_duplicate_pmus)
>> +size_t pmu_deduped_name_len(const struct perf_pmu *pmu, const char *name,
>> +                           bool skip_duplicate_pmus)
> 
> nit: I think the name should be perf_pmu__deduped_name_len for
> consistency with the other non-static functions.
> 

Will change.

>>   {
>> +       name = name ?: "";
> 
> nit: Should this just use pmu->name ?
> 

I can keep this part at the callsite in 
perf_pmus__scan_skip_duplicates() to avoid any confusion about this 
function accessing pmu->name or not. The only reason this function takes 
a separate name parameter is to allow it to work with struct 
pmu_event_info elsewhere as well.

>>          return skip_duplicate_pmus && !pmu->is_core
>>                  ? pmu_name_len_no_suffix(name)
>>                  : strlen(name);
>> diff --git a/tools/perf/util/pmu.h b/tools/perf/util/pmu.h
>> index b93014cc3670..ce6a394a695d 100644
>> --- a/tools/perf/util/pmu.h
>> +++ b/tools/perf/util/pmu.h
>> @@ -297,5 +297,7 @@ struct perf_pmu *perf_pmus__find_core_pmu(void);
>>
>>   const char *perf_pmu__name_from_config(struct perf_pmu *pmu, u64 config);
>>   bool perf_pmu__is_fake(const struct perf_pmu *pmu);
>> +size_t pmu_deduped_name_len(const struct perf_pmu *pmu, const char *name,
>> +                           bool skip_duplicate_pmus);
>>
>>   #endif /* __PMU_H */
>> diff --git a/tools/perf/util/pmus.c b/tools/perf/util/pmus.c
>> index cb1b14ade25b..1acc27af4d02 100644
>> --- a/tools/perf/util/pmus.c
>> +++ b/tools/perf/util/pmus.c
>> @@ -358,12 +358,14 @@ static struct perf_pmu *perf_pmus__scan_skip_duplicates(struct perf_pmu *pmu)
>>          if (!pmu) {
>>                  pmu_read_sysfs(PERF_TOOL_PMU_TYPE_ALL_MASK);
>>                  pmu = list_prepare_entry(pmu, &core_pmus, list);
>> -       } else
>> -               last_pmu_name_len = pmu_name_len_no_suffix(pmu->name ?: "");
>> +       } else {
>> +               last_pmu_name_len = pmu_deduped_name_len(pmu, pmu->name,
>> +                                                        /*skip_duplicate_pmus=*/true);
>> +       }
>>
>>          if (use_core_pmus) {
>>                  list_for_each_entry_continue(pmu, &core_pmus, list) {
>> -                       int pmu_name_len = pmu_name_len_no_suffix(pmu->name ?: "");
>> +                       int pmu_name_len = strlen(pmu->name ?: "");
>>
>>                          if (last_pmu_name_len == pmu_name_len &&
>>                              !strncmp(last_pmu_name, pmu->name ?: "", pmu_name_len))
> 
> Can this code be removed given there shouldn't be core PMUs with
> identical names? ie:
> ```
> if (use_core_pmus) {
>      list_for_each_entry_continue(pmu, &core_pmus, list)
>          return pmu;
> 
>      pmu = NULL;
>      pmu = list_prepare_entry(pmu, &other_pmus, list);
> }
> ```
> 
> Thanks,
> Ian
> 

Even better, yes it probably can.

Thanks
James

>>
>> --
>> 2.34.1
>>