Re: tmp.perf/core segfaulting on metrics test

[Namhyung Kim]

Hi Arnaldo,

On Fri, Dec 2, 2022 at 11:11 AM Arnaldo Carvalho de Melo
<acme@xxxxxxxxxx> wrote:
>
> While trying to get back at processing patches for Linux 6.2, what I
> have at tmp.perf/core branch is failing like below, can someone try to
> help in diagnosing this?

I think you missed this

  https://lore.kernel.org/r/20221130111521.334152-2-james.clark@xxxxxxx

>
> People really should try to run 'perf test' before pushing things
> upstream, but maybe this is related to a recent merge with
> torvalds/master, I'll check, but as a heads up and if someone wants to
> help...

Right, but the metric test is special since it contains hardware-
specific metrics.  But I'll double check with perf test before sending
patches.

Thanks,
Namhyung


>
> [root@quaco ~]# perf test 99
>  99: perf all metricgroups test                                      : FAILED!
> [root@quaco ~]# perf test -v 99
>  99: perf all metricgroups test                                      :
> --- start ---
> test child forked, pid 1494614
> Testing Backend
>
>  Performance counter stats for 'system wide':
>
>          9,978,757      CPU_CLK_UNHALTED.THREAD_ANY      #     56.7 %  tma_core_bound
> /home/acme/libexec/perf-core/tests/shell/stat_all_metricgroups.sh: line 7: 1494617 Segmentation fault      (core dumped) perf stat -M "$m" -a true
> test child finished with -1
> ---- end ----
> perf all metricgroups test: FAILED!
> [root@quaco ~]#
>
>
> --
>
> - Arnaldo