2024-01-24 23:42:26

by Ian Rogers

Subject: [PATCH v1] perf pmu: Treat the msr pmu as software

The msr PMU is a software one, meaning msr events may be grouped
with events in a hardware context. As the msr PMU isn't marked as a
software PMU by perf_pmu__is_software, groups with the msr PMU in
are broken and the msr events placed in a different group. This
may lead to multiplexing errors where a hardware event isn't
counted while the msr event, such as tsc, is. Fix all of this by
marking the msr PMU as software, which agrees with the driver.

Before:
```
$ perf stat -e '{slots,tsc}' -a true
WARNING: events were regrouped to match PMUs

Performance counter stats for 'system wide':

1,750,335 slots
4,243,557 tsc

0.001456717 seconds time elapsed
```

After:
```
$ perf stat -e '{slots,tsc}' -a true
Performance counter stats for 'system wide':

12,526,380 slots
3,415,163 tsc

0.001488360 seconds time elapsed
```

Fixes: 251aa040244a ("perf parse-events: Wildcard most "numeric" events")
Signed-off-by: Ian Rogers <[email protected]>
---
The fixes tag is close enough rather than being fully accurate. The
regression was introduced earlier by the automatic event regrouping.
---
tools/perf/util/pmu.c | 12 +++++++++++-
1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/tools/perf/util/pmu.c b/tools/perf/util/pmu.c
index 3c9609944a2f..88b9aa7d3a27 100644
--- a/tools/perf/util/pmu.c
+++ b/tools/perf/util/pmu.c
@@ -1760,6 +1760,12 @@ bool pmu__name_match(const struct perf_pmu *pmu, const char *pmu_name)

bool perf_pmu__is_software(const struct perf_pmu *pmu)
{
+ const char *known_sw_pmus[] = {
+ "kprobe",
+ "msr",
+ "uprobe",
+ };
+
if (pmu->is_core || pmu->is_uncore || pmu->auxtrace)
return false;
switch (pmu->type) {
@@ -1771,7 +1777,11 @@ bool perf_pmu__is_software(const struct perf_pmu *pmu)
case PERF_TYPE_BREAKPOINT: return true;
default: break;
}
- return !strcmp(pmu->name, "kprobe") || !strcmp(pmu->name, "uprobe");
+ for (size_t i = 0; i < ARRAY_SIZE(known_sw_pmus); i++) {
+ if (!strcmp(pmu->name, known_sw_pmus[i]))
+ return true;
+ }
+ return false;
}

FILE *perf_pmu__open_file(const struct perf_pmu *pmu, const char *name)
--
2.43.0.429.g432eaa2c6b-goog

2024-01-25 19:58:08

by Liang, Kan

Subject: Re: [PATCH v1] perf pmu: Treat the msr pmu as software

On 2024-01-24 6:42 p.m., Ian Rogers wrote:
> The msr PMU is a software one, meaning msr events may be grouped
> with events in a hardware context. As the msr PMU isn't marked as a
> software PMU by perf_pmu__is_software, groups with the msr PMU in
> are broken and the msr events placed in a different group. This
> may lead to multiplexing errors where a hardware event isn't
> counted while the msr event, such as tsc, is. Fix all of this by
> marking the msr PMU as software, which agrees with the driver.

Yes, the MSR PMU is in SW context, perf_sw_context.

Reviewed-by: Kan Liang <[email protected]>

Thanks,
Kan

>
> Before:
> ```
> $ perf stat -e '{slots,tsc}' -a true
> WARNING: events were regrouped to match PMUs
>
> Performance counter stats for 'system wide':
>
> 1,750,335 slots
> 4,243,557 tsc
>
> 0.001456717 seconds time elapsed
> ```
>
> After:
> ```
> $ perf stat -e '{slots,tsc}' -a true
> Performance counter stats for 'system wide':
>
> 12,526,380 slots
> 3,415,163 tsc
>
> 0.001488360 seconds time elapsed
> ```
>
> Fixes: 251aa040244a ("perf parse-events: Wildcard most "numeric" events")
> Signed-off-by: Ian Rogers <[email protected]>
> ---
> The fixes tag is close enough rather than being fully accurate. The
> regression was introduced earlier by the automatic event regrouping.
> ---
> tools/perf/util/pmu.c | 12 +++++++++++-
> 1 file changed, 11 insertions(+), 1 deletion(-)
>
> diff --git a/tools/perf/util/pmu.c b/tools/perf/util/pmu.c
> index 3c9609944a2f..88b9aa7d3a27 100644
> --- a/tools/perf/util/pmu.c
> +++ b/tools/perf/util/pmu.c
> @@ -1760,6 +1760,12 @@ bool pmu__name_match(const struct perf_pmu *pmu, const char *pmu_name)
>
> bool perf_pmu__is_software(const struct perf_pmu *pmu)
> {
> + const char *known_sw_pmus[] = {
> + "kprobe",
> + "msr",
> + "uprobe",
> + };
> +
> if (pmu->is_core || pmu->is_uncore || pmu->auxtrace)
> return false;
> switch (pmu->type) {
> @@ -1771,7 +1777,11 @@ bool perf_pmu__is_software(const struct perf_pmu *pmu)
> case PERF_TYPE_BREAKPOINT: return true;
> default: break;
> }
> - return !strcmp(pmu->name, "kprobe") || !strcmp(pmu->name, "uprobe");
> + for (size_t i = 0; i < ARRAY_SIZE(known_sw_pmus); i++) {
> + if (!strcmp(pmu->name, known_sw_pmus[i]))
> + return true;
> + }
> + return false;
> }
>
> FILE *perf_pmu__open_file(const struct perf_pmu *pmu, const char *name)

2024-01-27 18:17:01

by Namhyung Kim

Subject: Re: [PATCH v1] perf pmu: Treat the msr pmu as software

On Wed, 24 Jan 2024 15:42:00 -0800, Ian Rogers wrote:
> The msr PMU is a software one, meaning msr events may be grouped
> with events in a hardware context. As the msr PMU isn't marked as a
> software PMU by perf_pmu__is_software, groups with the msr PMU in
> are broken and the msr events placed in a different group. This
> may lead to multiplexing errors where a hardware event isn't
> counted while the msr event, such as tsc, is. Fix all of this by
> marking the msr PMU as software, which agrees with the driver.
>
> [...]

Applied to perf-tools-next, thanks!

Best regards,
--
Namhyung Kim <[email protected]>