linux-kernel - Re: [PATCH linux-next] perf tests: Handle s390 metrics in perf all metrics test

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <CAP-5=fUcNNcuFEyreGa0dpqfJzZtkDdp4yJXgZR0+LN_Mj5RaQ@mail.gmail.com>
Date: Wed, 19 Nov 2025 09:06:38 -0800
From: Ian Rogers <irogers@...gle.com>
To: Thomas Richter <tmricht@...ux.ibm.com>
Cc: linux-kernel@...r.kernel.org, linux-s390@...r.kernel.org, 
	linux-perf-users@...r.kernel.org, acme@...nel.org, namhyung@...nel.org, 
	agordeev@...ux.ibm.com, gor@...ux.ibm.com, sumanthk@...ux.ibm.com, 
	hca@...ux.ibm.com, japo@...ux.ibm.com
Subject: Re: [PATCH linux-next] perf tests: Handle s390 metrics in perf all
 metrics test

On Wed, Nov 19, 2025 at 2:48 AM Thomas Richter <tmricht@...ux.ibm.com> wrote:
>
> This test case fails on s390 because some counters metrics
> are not available and the metric defined on them fail.
> Add s390x specific list of metrics to be skipped.
>
> Add an extra line to display which metric has been skipped.
> It is shown with the verbose option turned on.
>
> Also return "Ok" instead of "Skipped" when no errors
> have been detected, but some metric have been skipped.
> The seems more logical otherwise "Skipped" means all
> metric tests have been skipped instead of just a few.
>
> Signed-off-by: Thomas Richter <tmricht@...ux.ibm.com>

Thanks Thomas, I believe the intent with these metrics is they don't
fail but report unsupported for their counters. There's been some
discussion here:
https://lore.kernel.org/lkml/aRi9xnwdLh3Dir9f@google.com/
with Namhyung and James pointing out the issue on AMD and ARM, and me
being confused we Intel isn't failing similarly for events like itlb
accesses that Intel doesn't support. I'll try to find time to dig into
the issue.

Thanks,
Ian

> ---
>  tools/perf/tests/shell/stat_all_metrics.sh | 8 +++++++-
>  1 file changed, 7 insertions(+), 1 deletion(-)
>
> diff --git a/tools/perf/tests/shell/stat_all_metrics.sh b/tools/perf/tests/shell/stat_all_metrics.sh
> index a7edf01b3943..b86b36a49228 100755
> --- a/tools/perf/tests/shell/stat_all_metrics.sh
> +++ b/tools/perf/tests/shell/stat_all_metrics.sh
> @@ -7,6 +7,10 @@ ParanoidAndNotRoot()
>    [ "$(id -u)" != 0 ] && [ "$(cat /proc/sys/kernel/perf_event_paranoid)" -gt $1 ]
>  }
>
> +# Ignore metric which are not supported on s390x
> +[ "$(uname -m)" = "s390x" ] && ignore="|branch_miss_rate|l1d_miss_rate|llc_miss_rate|\
> +               dtlb_miss_rate|itlb_miss_rate|l1i_miss_rate|l1_prefetch_miss_rate"
> +
>  test_prog="sleep 0.01"
>  system_wide_flag="-a"
>  if ParanoidAndNotRoot 0
> @@ -27,9 +31,10 @@ for m in $(perf list --raw-dump metrics); do
>    fi
>    if [[ "$result" =~ "Cannot resolve IDs for" || "$result" =~ "No supported events found" ]]
>    then
> -    if [[ "$m" == @(l1_prefetch_miss_rate|stalled_cycles_per_instruction) ]]
> +    if [[ "$m" == @(l1_prefetch_miss_rate|stalled_cycles_per_instruction$ignore) ]]
>      then
>        # Default metrics that may use unsupported events.
> +      echo "Skipped metric $m"
>        continue
>      fi
>      echo "Metric contains missing events"
> @@ -106,4 +111,5 @@ for m in $(perf list --raw-dump metrics); do
>    err=1
>  done
>
> +[ "$err" -eq 2 ] && err=0
>  exit "$err"
> --
> 2.51.1
>