Hi Ian:
On 2023/8/19 1:19, Ian Rogers wrote:
> PMU caps are written as HEADER_PMU_CAPS or for the special case of the
> PMU "cpu" as HEADER_CPU_PMU_CAPS. As the PMU "cpu" is special, and not
> any "core" PMU, the logic had become broken and core PMUs not called
> "cpu" were not having their caps written. This affects ARM and s390
> non-hybrid PMUs.
>
> Simplify the PMU caps writing logic to scan one fewer time and to be
> more explicit in its behavior.
>
> Reported-by: Wei Li <[email protected]>
> Fixes: 178ddf3bad98 ("perf header: Avoid hybrid PMU list in write_pmu_caps")
> Signed-off-by: Ian Rogers <[email protected]>
> ---
> tools/perf/util/header.c | 31 ++++++++++++++++---------------
> 1 file changed, 16 insertions(+), 15 deletions(-)
>
> diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c
> index 52fbf526fe74..13c71d28e0eb 100644
> --- a/tools/perf/util/header.c
> +++ b/tools/perf/util/header.c
> @@ -1605,8 +1605,15 @@ static int write_pmu_caps(struct feat_fd *ff,
> int ret;
>
> while ((pmu = perf_pmus__scan(pmu))) {
> - if (!pmu->name || !strcmp(pmu->name, "cpu") ||
> - perf_pmu__caps_parse(pmu) <= 0)
> + if (!strcmp(pmu->name, "cpu")) {
So you removed the check of 'pmu->name', does this check really redundant? since
we can find such checks in many places in the perf code. If not, i think it is
necessary for strcmp().
> + /*
> + * The "cpu" PMU is special and covered by
> + * HEADER_CPU_PMU_CAPS. Note, core PMUs are
> + * counted/written here for ARM, s390 and Intel hybrid.
> + */
> + continue;
> + }
> + if (perf_pmu__caps_parse(pmu) <= 0)
> continue;
> nr_pmu++;
> }
> @@ -1619,23 +1626,17 @@ static int write_pmu_caps(struct feat_fd *ff,
> return 0;
>
> /*
> - * Write hybrid pmu caps first to maintain compatibility with
> - * older perf tool.
> + * Note older perf tools assume core PMUs come first, this is a property
> + * of perf_pmus__scan.
> */
> - if (perf_pmus__num_core_pmus() > 1) {
> - pmu = NULL;
> - while ((pmu = perf_pmus__scan_core(pmu))) {
> - ret = __write_pmu_caps(ff, pmu, true);
> - if (ret < 0)
> - return ret;
> - }
> - }
> -
> pmu = NULL;
> while ((pmu = perf_pmus__scan(pmu))) {
> - if (pmu->is_core || !pmu->nr_caps)
> + if (!strcmp(pmu->name, "cpu")) {
same here
Thanks,
Wei
> + /* Skip as above. */
> + continue;
> + }
> + if (perf_pmu__caps_parse(pmu) <= 0)
> continue;
> -
> ret = __write_pmu_caps(ff, pmu, true);
> if (ret < 0)
> return ret;
Em Sat, Aug 19, 2023 at 12:16:09PM +0800, liwei (GF) escreveu:
> Hi Ian:
>
> On 2023/8/19 1:19, Ian Rogers wrote:
> > PMU caps are written as HEADER_PMU_CAPS or for the special case of the
> > PMU "cpu" as HEADER_CPU_PMU_CAPS. As the PMU "cpu" is special, and not
> > any "core" PMU, the logic had become broken and core PMUs not called
> > "cpu" were not having their caps written. This affects ARM and s390
> > non-hybrid PMUs.
> >
> > Simplify the PMU caps writing logic to scan one fewer time and to be
> > more explicit in its behavior.
> >
> > Reported-by: Wei Li <[email protected]>
> > Fixes: 178ddf3bad98 ("perf header: Avoid hybrid PMU list in write_pmu_caps")
> > Signed-off-by: Ian Rogers <[email protected]>
> > ---
> > tools/perf/util/header.c | 31 ++++++++++++++++---------------
> > 1 file changed, 16 insertions(+), 15 deletions(-)
> >
> > diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c
> > index 52fbf526fe74..13c71d28e0eb 100644
> > --- a/tools/perf/util/header.c
> > +++ b/tools/perf/util/header.c
> > @@ -1605,8 +1605,15 @@ static int write_pmu_caps(struct feat_fd *ff,
> > int ret;
> >
> > while ((pmu = perf_pmus__scan(pmu))) {
> > - if (!pmu->name || !strcmp(pmu->name, "cpu") ||
> > - perf_pmu__caps_parse(pmu) <= 0)
> > + if (!strcmp(pmu->name, "cpu")) {
>
> So you removed the check of 'pmu->name', does this check really redundant? since
> we can find such checks in many places in the perf code. If not, i think it is
> necessary for strcmp().
Indeed, when sorting in tools/perf/util/pmus.c in cmp_sevent() we have:
/* Order by PMU name. */
if (as->pmu != bs->pmu) {
a_pmu_name = a_pmu_name ?: (as->pmu->name ?: "");
b_pmu_name = b_pmu_name ?: (bs->pmu->name ?: "");
ret = strcmp(a_pmu_name, b_pmu_name);
if (ret)
return ret;
}
And even if in this specific case, for some reason, we could guarantee
that pmu->name isn't NULL, then removing that check should be best left
for a separate patch with an explanation as to why that is safe.
Having it as:
while ((pmu = perf_pmus__scan(pmu))) {
- if (!pmu->name || !strcmp(pmu->name, "cpu") ||
- perf_pmu__caps_parse(pmu) <= 0)
+ if (!pmu->name || !strcmp(pmu->name, "cpu")) {
even eases a bit reviewing, as we see we're just removing that
perf_pmu__caps_parse(pmu) line.
Ian?
- Arnaldo
> > + /*
> > + * The "cpu" PMU is special and covered by
> > + * HEADER_CPU_PMU_CAPS. Note, core PMUs are
> > + * counted/written here for ARM, s390 and Intel hybrid.
> > + */
> > + continue;
> > + }
> > + if (perf_pmu__caps_parse(pmu) <= 0)
> > continue;
> > nr_pmu++;
> > }
> > @@ -1619,23 +1626,17 @@ static int write_pmu_caps(struct feat_fd *ff,
> > return 0;
> >
> > /*
> > - * Write hybrid pmu caps first to maintain compatibility with
> > - * older perf tool.
> > + * Note older perf tools assume core PMUs come first, this is a property
> > + * of perf_pmus__scan.
> > */
> > - if (perf_pmus__num_core_pmus() > 1) {
> > - pmu = NULL;
> > - while ((pmu = perf_pmus__scan_core(pmu))) {
> > - ret = __write_pmu_caps(ff, pmu, true);
> > - if (ret < 0)
> > - return ret;
> > - }
> > - }
> > -
> > pmu = NULL;
> > while ((pmu = perf_pmus__scan(pmu))) {
> > - if (pmu->is_core || !pmu->nr_caps)
> > + if (!strcmp(pmu->name, "cpu")) {
>
> same here
>
> Thanks,
> Wei
>
> > + /* Skip as above. */
> > + continue;
> > + }
> > + if (perf_pmu__caps_parse(pmu) <= 0)
> > continue;
> > -
> > ret = __write_pmu_caps(ff, pmu, true);
> > if (ret < 0)
> > return ret;
--
- Arnaldo