[PATCH v1 1/2] perf metrics: Avoid segv if default metricgroup isn't set

[PATCH v1 1/2] perf metrics: Avoid segv if default metricgroup isn't set

[Ian Rogers]

A metric is default by having "Default" within its groups. The default
metricgroup name needn't be set and this can result in segv in
default_metricgroup_cmp and perf_stat__print_shadow_stats_metricgroup
that assume it has a value when there is a Default metric group. To
avoid the segv initialize the value to "".

Fixes: 1c0e47956a8e ("perf metrics: Sort the Default metricgroup")
Signed-off-by: Ian Rogers <irogers@google.com>
---
 tools/perf/util/metricgroup.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/perf/util/metricgroup.c b/tools/perf/util/metricgroup.c
index 0484736d9fe4..ca3e0404f187 100644
--- a/tools/perf/util/metricgroup.c
+++ b/tools/perf/util/metricgroup.c
@@ -225,7 +225,7 @@ static struct metric *metric__new(const struct pmu_metric *pm,
 
 	m->pmu = pm->pmu ?: "cpu";
 	m->metric_name = pm->metric_name;
-	m->default_metricgroup_name = pm->default_metricgroup_name;
+	m->default_metricgroup_name = pm->default_metricgroup_name ?: "";
 	m->modifier = NULL;
 	if (modifier) {
 		m->modifier = strdup(modifier);
-- 
2.43.0.rc2.451.g8631bc7472-goog

[PATCH v1 2/2] perf vendor events arm64: Fix default AmpereOne metrics

[Ian Rogers]

Add default metric group name for TopdownL1 metrics.

Fixes: 59faeaf80d02 ("perf vendor events arm64: Fix for AmpereOne metrics")
Signed-off-by: Ian Rogers <irogers@google.com>
---
 tools/perf/pmu-events/arch/arm64/ampere/ampereone/metrics.json | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/tools/perf/pmu-events/arch/arm64/ampere/ampereone/metrics.json b/tools/perf/pmu-events/arch/arm64/ampere/ampereone/metrics.json
index e2848a9d4848..afcdad58ef89 100644
--- a/tools/perf/pmu-events/arch/arm64/ampere/ampereone/metrics.json
+++ b/tools/perf/pmu-events/arch/arm64/ampere/ampereone/metrics.json
@@ -231,6 +231,7 @@
         "MetricName": "slots_lost_misspeculation_fraction",
         "MetricExpr": "100 * ((OP_SPEC - OP_RETIRED) / (CPU_CYCLES * #slots))",
         "BriefDescription": "Fraction of slots lost due to misspeculation",
+        "DefaultMetricgroupName": "TopdownL1",
         "MetricGroup": "Default;TopdownL1",
         "ScaleUnit": "1percent of slots"
     },
@@ -238,6 +239,7 @@
         "MetricName": "retired_fraction",
         "MetricExpr": "100 * (OP_RETIRED / (CPU_CYCLES * #slots))",
         "BriefDescription": "Fraction of slots retiring, useful work",
+        "DefaultMetricgroupName": "TopdownL1",
         "MetricGroup": "Default;TopdownL1",
 	"ScaleUnit": "1percent of slots"
     },
-- 
2.43.0.rc2.451.g8631bc7472-goog

Re: [PATCH v1 1/2] perf metrics: Avoid segv if default metricgroup isn't set

[Ilkka Koskinen]

On Mon, 4 Dec 2023, Ian Rogers wrote:
> A metric is default by having "Default" within its groups. The default
> metricgroup name needn't be set and this can result in segv in
> default_metricgroup_cmp and perf_stat__print_shadow_stats_metricgroup
> that assume it has a value when there is a Default metric group. To
> avoid the segv initialize the value to "".
>
> Fixes: 1c0e47956a8e ("perf metrics: Sort the Default metricgroup")
> Signed-off-by: Ian Rogers <irogers@google.com>

Thanks! I was going to look for the bug but got pulled to other 
tasks. The patch looks good to me and I tested it successfully on 
AmpereOne.

   Reviewed-and-tested-by: Ilkka Koskinen <ilkka@os.amperecomputing.com>


Cheers, Ilkka

> ---
> tools/perf/util/metricgroup.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/tools/perf/util/metricgroup.c b/tools/perf/util/metricgroup.c
> index 0484736d9fe4..ca3e0404f187 100644
> --- a/tools/perf/util/metricgroup.c
> +++ b/tools/perf/util/metricgroup.c
> @@ -225,7 +225,7 @@ static struct metric *metric__new(const struct pmu_metric *pm,
>
> 	m->pmu = pm->pmu ?: "cpu";
> 	m->metric_name = pm->metric_name;
> -	m->default_metricgroup_name = pm->default_metricgroup_name;
> +	m->default_metricgroup_name = pm->default_metricgroup_name ?: "";
> 	m->modifier = NULL;
> 	if (modifier) {
> 		m->modifier = strdup(modifier);
> -- 
> 2.43.0.rc2.451.g8631bc7472-goog
>
>

Re: [PATCH v1 1/2] perf metrics: Avoid segv if default metricgroup isn't set

[Namhyung Kim]

Hello,

On Mon, Dec 4, 2023 at 2:45 PM Ilkka Koskinen
<ilkka@os.amperecomputing.com> wrote:
>
>
>
> On Mon, 4 Dec 2023, Ian Rogers wrote:
> > A metric is default by having "Default" within its groups. The default
> > metricgroup name needn't be set and this can result in segv in
> > default_metricgroup_cmp and perf_stat__print_shadow_stats_metricgroup
> > that assume it has a value when there is a Default metric group. To
> > avoid the segv initialize the value to "".
> >
> > Fixes: 1c0e47956a8e ("perf metrics: Sort the Default metricgroup")
> > Signed-off-by: Ian Rogers <irogers@google.com>
>
> Thanks! I was going to look for the bug but got pulled to other
> tasks. The patch looks good to me and I tested it successfully on
> AmpereOne.
>
>    Reviewed-and-tested-by: Ilkka Koskinen <ilkka@os.amperecomputing.com>

Looks like it needs to go through perf-tools for v6.7.
Ian, do you think this one is enough?

Thanks,
Namhyung

Re: [PATCH v1 1/2] perf metrics: Avoid segv if default metricgroup isn't set

[Ian Rogers]

On Mon, Dec 4, 2023 at 7:33 PM Namhyung Kim <namhyung@kernel.org> wrote:
>
> Hello,
>
> On Mon, Dec 4, 2023 at 2:45 PM Ilkka Koskinen
> <ilkka@os.amperecomputing.com> wrote:
> >
> >
> >
> > On Mon, 4 Dec 2023, Ian Rogers wrote:
> > > A metric is default by having "Default" within its groups. The default
> > > metricgroup name needn't be set and this can result in segv in
> > > default_metricgroup_cmp and perf_stat__print_shadow_stats_metricgroup
> > > that assume it has a value when there is a Default metric group. To
> > > avoid the segv initialize the value to "".
> > >
> > > Fixes: 1c0e47956a8e ("perf metrics: Sort the Default metricgroup")
> > > Signed-off-by: Ian Rogers <irogers@google.com>
> >
> > Thanks! I was going to look for the bug but got pulled to other
> > tasks. The patch looks good to me and I tested it successfully on
> > AmpereOne.
> >
> >    Reviewed-and-tested-by: Ilkka Koskinen <ilkka@os.amperecomputing.com>
>
> Looks like it needs to go through perf-tools for v6.7.
> Ian, do you think this one is enough?

From a user's pov the json fix is nicer as otherwise perf stat output
for the relevant metrics lacks a heading on the left. This fix is
smaller. I'm easy about which to take :-)

Thanks,
Ian

> Thanks,
> Namhyung

Re: [PATCH v1 1/2] perf metrics: Avoid segv if default metricgroup isn't set

[Arnaldo Carvalho de Melo]

Em Mon, Dec 04, 2023 at 07:33:18PM -0800, Namhyung Kim escreveu:
> On Mon, Dec 4, 2023 at 2:45 PM Ilkka Koskinen <ilkka@os.amperecomputing.com> wrote:
> > On Mon, 4 Dec 2023, Ian Rogers wrote:
> > > A metric is default by having "Default" within its groups. The default
> > > metricgroup name needn't be set and this can result in segv in
> > > default_metricgroup_cmp and perf_stat__print_shadow_stats_metricgroup
> > > that assume it has a value when there is a Default metric group. To
> > > avoid the segv initialize the value to "".

> > > Fixes: 1c0e47956a8e ("perf metrics: Sort the Default metricgroup")
> > > Signed-off-by: Ian Rogers <irogers@google.com>

> > Thanks! I was going to look for the bug but got pulled to other
> > tasks. The patch looks good to me and I tested it successfully on
> > AmpereOne.

> >    Reviewed-and-tested-by: Ilkka Koskinen <ilkka@os.amperecomputing.com>
 
> Looks like it needs to go through perf-tools for v6.7.
> Ian, do you think this one is enough?

So I had this on my local perf-tools-next, removed now, I also have some
other fixes by Ian on the tmp.perf-tools-next, please let me know what
you guys decide to have in perf-tools for v6.7 so that I can remove it
from there.

- Arnaldo

Re: [PATCH v1 1/2] perf metrics: Avoid segv if default metricgroup isn't set

[Namhyung Kim]

Hi Arnaldo,

On Tue, Dec 5, 2023 at 7:51 AM Arnaldo Carvalho de Melo <acme@kernel.org> wrote:
>
> Em Mon, Dec 04, 2023 at 07:33:18PM -0800, Namhyung Kim escreveu:
> > On Mon, Dec 4, 2023 at 2:45 PM Ilkka Koskinen <ilkka@os.amperecomputing.com> wrote:
> > > On Mon, 4 Dec 2023, Ian Rogers wrote:
> > > > A metric is default by having "Default" within its groups. The default
> > > > metricgroup name needn't be set and this can result in segv in
> > > > default_metricgroup_cmp and perf_stat__print_shadow_stats_metricgroup
> > > > that assume it has a value when there is a Default metric group. To
> > > > avoid the segv initialize the value to "".
>
> > > > Fixes: 1c0e47956a8e ("perf metrics: Sort the Default metricgroup")
> > > > Signed-off-by: Ian Rogers <irogers@google.com>
>
> > > Thanks! I was going to look for the bug but got pulled to other
> > > tasks. The patch looks good to me and I tested it successfully on
> > > AmpereOne.
>
> > >    Reviewed-and-tested-by: Ilkka Koskinen <ilkka@os.amperecomputing.com>
>
> > Looks like it needs to go through perf-tools for v6.7.
> > Ian, do you think this one is enough?
>
> So I had this on my local perf-tools-next, removed now, I also have some
> other fixes by Ian on the tmp.perf-tools-next, please let me know what
> you guys decide to have in perf-tools for v6.7 so that I can remove it
> from there.

I think we need this one and the Ampere default-metric-group fix.

https://lore.kernel.org/r/20231201021550.1109196-2-ilkka@os.amperecomputing.com/

Also perf list JSON fix can go to v6.7.

https://lore.kernel.org/r/20231129213428.2227448-2-irogers@google.com/

Thanks,
Namhyung

Re: [PATCH v1 1/2] perf metrics: Avoid segv if default metricgroup isn't set

[Arnaldo Carvalho de Melo]

Em Tue, Dec 05, 2023 at 09:24:42AM -0800, Namhyung Kim escreveu:
> Hi Arnaldo,
> 
> On Tue, Dec 5, 2023 at 7:51 AM Arnaldo Carvalho de Melo <acme@kernel.org> wrote:
> >
> > Em Mon, Dec 04, 2023 at 07:33:18PM -0800, Namhyung Kim escreveu:
> > > On Mon, Dec 4, 2023 at 2:45 PM Ilkka Koskinen <ilkka@os.amperecomputing.com> wrote:
> > > > On Mon, 4 Dec 2023, Ian Rogers wrote:
> > > > > A metric is default by having "Default" within its groups. The default
> > > > > metricgroup name needn't be set and this can result in segv in
> > > > > default_metricgroup_cmp and perf_stat__print_shadow_stats_metricgroup
> > > > > that assume it has a value when there is a Default metric group. To
> > > > > avoid the segv initialize the value to "".
> >
> > > > > Fixes: 1c0e47956a8e ("perf metrics: Sort the Default metricgroup")
> > > > > Signed-off-by: Ian Rogers <irogers@google.com>
> >
> > > > Thanks! I was going to look for the bug but got pulled to other
> > > > tasks. The patch looks good to me and I tested it successfully on
> > > > AmpereOne.
> >
> > > >    Reviewed-and-tested-by: Ilkka Koskinen <ilkka@os.amperecomputing.com>
> >
> > > Looks like it needs to go through perf-tools for v6.7.
> > > Ian, do you think this one is enough?
> >
> > So I had this on my local perf-tools-next, removed now, I also have some
> > other fixes by Ian on the tmp.perf-tools-next, please let me know what
> > you guys decide to have in perf-tools for v6.7 so that I can remove it
> > from there.
> 
> I think we need this one and the Ampere default-metric-group fix.
> 
> https://lore.kernel.org/r/20231201021550.1109196-2-ilkka@os.amperecomputing.com/
> 
> Also perf list JSON fix can go to v6.7.
> 
> https://lore.kernel.org/r/20231129213428.2227448-2-irogers@google.com/

Ok, removed both, please augment the later description to:

"perf list: Fix JSON segfault by setting the used skip_fuplicate_pmus callback"

The original description was vague, improving it a bit like that helps
when just looking at the output of "git log --oneline".

Thanks,

- Arnaldo

Re: [PATCH v1 1/2] perf metrics: Avoid segv if default metricgroup isn't set

[Namhyung Kim]

On Tue, Dec 5, 2023 at 10:48 AM Arnaldo Carvalho de Melo
<acme@kernel.org> wrote:
>
> Em Tue, Dec 05, 2023 at 09:24:42AM -0800, Namhyung Kim escreveu:
> > Hi Arnaldo,
> >
> > On Tue, Dec 5, 2023 at 7:51 AM Arnaldo Carvalho de Melo <acme@kernel.org> wrote:
> > >
> > > Em Mon, Dec 04, 2023 at 07:33:18PM -0800, Namhyung Kim escreveu:
> > > > On Mon, Dec 4, 2023 at 2:45 PM Ilkka Koskinen <ilkka@os.amperecomputing.com> wrote:
> > > > > On Mon, 4 Dec 2023, Ian Rogers wrote:
> > > > > > A metric is default by having "Default" within its groups. The default
> > > > > > metricgroup name needn't be set and this can result in segv in
> > > > > > default_metricgroup_cmp and perf_stat__print_shadow_stats_metricgroup
> > > > > > that assume it has a value when there is a Default metric group. To
> > > > > > avoid the segv initialize the value to "".
> > >
> > > > > > Fixes: 1c0e47956a8e ("perf metrics: Sort the Default metricgroup")
> > > > > > Signed-off-by: Ian Rogers <irogers@google.com>
> > >
> > > > > Thanks! I was going to look for the bug but got pulled to other
> > > > > tasks. The patch looks good to me and I tested it successfully on
> > > > > AmpereOne.
> > >
> > > > >    Reviewed-and-tested-by: Ilkka Koskinen <ilkka@os.amperecomputing.com>
> > >
> > > > Looks like it needs to go through perf-tools for v6.7.
> > > > Ian, do you think this one is enough?
> > >
> > > So I had this on my local perf-tools-next, removed now, I also have some
> > > other fixes by Ian on the tmp.perf-tools-next, please let me know what
> > > you guys decide to have in perf-tools for v6.7 so that I can remove it
> > > from there.
> >
> > I think we need this one and the Ampere default-metric-group fix.
> >
> > https://lore.kernel.org/r/20231201021550.1109196-2-ilkka@os.amperecomputing.com/
> >
> > Also perf list JSON fix can go to v6.7.
> >
> > https://lore.kernel.org/r/20231129213428.2227448-2-irogers@google.com/
>
> Ok, removed both, please augment the later description to:
>
> "perf list: Fix JSON segfault by setting the used skip_fuplicate_pmus callback"
>
> The original description was vague, improving it a bit like that helps
> when just looking at the output of "git log --oneline".

Done and pushed to tmp.perf-tools!

Thanks,
Namhyung

Re: [PATCH v1 1/2] perf metrics: Avoid segv if default metricgroup isn't set

[Ian Rogers]

On Tue, Dec 5, 2023 at 11:11 AM Namhyung Kim <namhyung@kernel.org> wrote:
>
> On Tue, Dec 5, 2023 at 10:48 AM Arnaldo Carvalho de Melo
> <acme@kernel.org> wrote:
> >
> > Em Tue, Dec 05, 2023 at 09:24:42AM -0800, Namhyung Kim escreveu:
> > > Hi Arnaldo,
> > >
> > > On Tue, Dec 5, 2023 at 7:51 AM Arnaldo Carvalho de Melo <acme@kernel.org> wrote:
> > > >
> > > > Em Mon, Dec 04, 2023 at 07:33:18PM -0800, Namhyung Kim escreveu:
> > > > > On Mon, Dec 4, 2023 at 2:45 PM Ilkka Koskinen <ilkka@os.amperecomputing.com> wrote:
> > > > > > On Mon, 4 Dec 2023, Ian Rogers wrote:
> > > > > > > A metric is default by having "Default" within its groups. The default
> > > > > > > metricgroup name needn't be set and this can result in segv in
> > > > > > > default_metricgroup_cmp and perf_stat__print_shadow_stats_metricgroup
> > > > > > > that assume it has a value when there is a Default metric group. To
> > > > > > > avoid the segv initialize the value to "".
> > > >
> > > > > > > Fixes: 1c0e47956a8e ("perf metrics: Sort the Default metricgroup")
> > > > > > > Signed-off-by: Ian Rogers <irogers@google.com>
> > > >
> > > > > > Thanks! I was going to look for the bug but got pulled to other
> > > > > > tasks. The patch looks good to me and I tested it successfully on
> > > > > > AmpereOne.
> > > >
> > > > > >    Reviewed-and-tested-by: Ilkka Koskinen <ilkka@os.amperecomputing.com>
> > > >
> > > > > Looks like it needs to go through perf-tools for v6.7.
> > > > > Ian, do you think this one is enough?
> > > >
> > > > So I had this on my local perf-tools-next, removed now, I also have some
> > > > other fixes by Ian on the tmp.perf-tools-next, please let me know what
> > > > you guys decide to have in perf-tools for v6.7 so that I can remove it
> > > > from there.
> > >
> > > I think we need this one and the Ampere default-metric-group fix.
> > >
> > > https://lore.kernel.org/r/20231201021550.1109196-2-ilkka@os.amperecomputing.com/
> > >
> > > Also perf list JSON fix can go to v6.7.
> > >
> > > https://lore.kernel.org/r/20231129213428.2227448-2-irogers@google.com/
> >
> > Ok, removed both, please augment the later description to:
> >
> > "perf list: Fix JSON segfault by setting the used skip_fuplicate_pmus callback"
> >
> > The original description was vague, improving it a bit like that helps
> > when just looking at the output of "git log --oneline".
>
> Done and pushed to tmp.perf-tools!

Thanks Namhyung, there was a typo in Arnaldo's commit message
"s/fuplicate/duplicate/" could you fix?

Ian

> Thanks,
> Namhyung

Re: [PATCH v1 1/2] perf metrics: Avoid segv if default metricgroup isn't set

[Namhyung Kim]

On Tue, Dec 5, 2023 at 11:15 AM Ian Rogers <irogers@google.com> wrote:
>
> On Tue, Dec 5, 2023 at 11:11 AM Namhyung Kim <namhyung@kernel.org> wrote:
> >
> > On Tue, Dec 5, 2023 at 10:48 AM Arnaldo Carvalho de Melo
> > <acme@kernel.org> wrote:

> > > "perf list: Fix JSON segfault by setting the used skip_fuplicate_pmus callback"
> > >
> > > The original description was vague, improving it a bit like that helps
> > > when just looking at the output of "git log --oneline".
> >
> > Done and pushed to tmp.perf-tools!
>
> Thanks Namhyung, there was a typo in Arnaldo's commit message
> "s/fuplicate/duplicate/" could you fix?

Oops, fixed now.  Thanks for checking.
Namhyung

Re: [PATCH v1 1/2] perf metrics: Avoid segv if default metricgroup isn't set

[Arnaldo Carvalho de Melo]

Em Mon, Dec 04, 2023 at 02:44:54PM -0800, Ilkka Koskinen escreveu:
> 
> 
> On Mon, 4 Dec 2023, Ian Rogers wrote:
> > A metric is default by having "Default" within its groups. The default
> > metricgroup name needn't be set and this can result in segv in
> > default_metricgroup_cmp and perf_stat__print_shadow_stats_metricgroup
> > that assume it has a value when there is a Default metric group. To
> > avoid the segv initialize the value to "".
> > 
> > Fixes: 1c0e47956a8e ("perf metrics: Sort the Default metricgroup")
> > Signed-off-by: Ian Rogers <irogers@google.com>
> 
> Thanks! I was going to look for the bug but got pulled to other tasks. The
> patch looks good to me and I tested it successfully on AmpereOne.
> 
>   Reviewed-and-tested-by: Ilkka Koskinen <ilkka@os.amperecomputing.com>

Thanks, applied to perf-tools-next.

- Arnaldo