* Test case '105: perf stat tests' fails for s390.
@ 2026-03-09 13:04 Thomas Richter
2026-03-09 18:23 ` Ian Rogers
0 siblings, 1 reply; 3+ messages in thread
From: Thomas Richter @ 2026-03-09 13:04 UTC (permalink / raw)
To: ian Rogers; +Cc: linux-perf-use., Jan Polensky, Sumanth Korikkar
Ian,
test case '105: perf stat tests' fails for s390 (linux 7.0.0rc2) . It boils down to this command
# ./perf stat -C 0-31 -a true
Error:
No supported events found.
The first bad commit is this one:
commit cee275edcdb1 ("perf metricgroup: Don't early exit if no CPUID table exists")
Before this commit the output of above command is:
Output before that commit:
./perf stat -C 0-31 -a true
Performance counter stats for 'system wide':
105 context-switches # 3579.8 cs/sec cs_per_second
29.33 msec cpu-clock # 7.8 CPUs CPUs_utilized
34 cpu-migrations # 1159.2 migrations/sec migrations_per_second
52 page-faults # 1772.9 faults/sec page_faults_per_second
5,026,832 cpu-cycles # 0.2 GHz cycles_frequency
3,029,266 instructions # 0.6 instructions insn_per_cycle
0.000851608 seconds time elapsed
Since this is executed on an s390 LPAR, hardware event CPU_CYCLES is available, so it
should be used. This is something I will look into.
Also some other test cases fail after the bad commit:
./perf test 2>&1 | grep FAILED
99: perf stat CSV output linter : FAILED!
102: perf stat JSON output linter : FAILED!
104: perf stat STD output linter : FAILED!
105: perf stat tests : FAILED!
133: perftool-testsuite_report : FAILED!
Before the bad commit the output is
./perf test 2>&1 | grep FAILED
133: perftool-testsuite_report : FAILED!
Any ideas where to look too narrow down the debuggung are very welcome.
Thanks a lot.
--
Thomas Richter, Dept 3303, IBM s390 Linux Development, Boeblingen, Germany
--
IBM Deutschland Research & Development GmbH
Vorsitzender des Aufsichtsrats: Wolfgang Wendt
Geschäftsführung: David Faller
Sitz der Gesellschaft: Böblingen / Registergericht: Amtsgericht Stuttgart, HRB 243294
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: Test case '105: perf stat tests' fails for s390.
2026-03-09 13:04 Test case '105: perf stat tests' fails for s390 Thomas Richter
@ 2026-03-09 18:23 ` Ian Rogers
2026-03-11 5:52 ` Thomas Richter
0 siblings, 1 reply; 3+ messages in thread
From: Ian Rogers @ 2026-03-09 18:23 UTC (permalink / raw)
To: Thomas Richter; +Cc: linux-perf-use., Jan Polensky, Sumanth Korikkar
On Mon, Mar 9, 2026 at 6:04 AM Thomas Richter <tmricht@linux.ibm.com> wrote:
>
> Ian,
>
> test case '105: perf stat tests' fails for s390 (linux 7.0.0rc2) . It boils down to this command
> # ./perf stat -C 0-31 -a true
> Error:
> No supported events found.
>
> The first bad commit is this one:
> commit cee275edcdb1 ("perf metricgroup: Don't early exit if no CPUID table exists")
>
> Before this commit the output of above command is:
> Output before that commit:
> ./perf stat -C 0-31 -a true
>
> Performance counter stats for 'system wide':
>
> 105 context-switches # 3579.8 cs/sec cs_per_second
> 29.33 msec cpu-clock # 7.8 CPUs CPUs_utilized
> 34 cpu-migrations # 1159.2 migrations/sec migrations_per_second
> 52 page-faults # 1772.9 faults/sec page_faults_per_second
> 5,026,832 cpu-cycles # 0.2 GHz cycles_frequency
> 3,029,266 instructions # 0.6 instructions insn_per_cycle
>
> 0.000851608 seconds time elapsed
>
> Since this is executed on an s390 LPAR, hardware event CPU_CYCLES is available, so it
> should be used. This is something I will look into.
>
> Also some other test cases fail after the bad commit:
> ./perf test 2>&1 | grep FAILED
> 99: perf stat CSV output linter : FAILED!
> 102: perf stat JSON output linter : FAILED!
> 104: perf stat STD output linter : FAILED!
> 105: perf stat tests : FAILED!
> 133: perftool-testsuite_report : FAILED!
>
> Before the bad commit the output is
> ./perf test 2>&1 | grep FAILED
> 133: perftool-testsuite_report : FAILED!
>
> Any ideas where to look too narrow down the debuggung are very welcome.
Hi Thomas, sorry for the breakage! I think the issue is a missed
commit in the v7.0 PR, specifically:
https://web.git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/commit/tools/perf/util/metricgroup.c?h=perf-tools-next&id=c5a244bf17caf2de22f9e100832b75f72b31d3e6
This has also shown up as breaking stable backports:
https://lore.kernel.org/lkml/072e2a07-5c6f-47b5-9695-0a3ffe854ac8@kernel.org/
Thanks,
Ian
> Thanks a lot.
>
> --
> Thomas Richter, Dept 3303, IBM s390 Linux Development, Boeblingen, Germany
> --
> IBM Deutschland Research & Development GmbH
>
> Vorsitzender des Aufsichtsrats: Wolfgang Wendt
>
> Geschäftsführung: David Faller
>
> Sitz der Gesellschaft: Böblingen / Registergericht: Amtsgericht Stuttgart, HRB 243294
>
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: Test case '105: perf stat tests' fails for s390.
2026-03-09 18:23 ` Ian Rogers
@ 2026-03-11 5:52 ` Thomas Richter
0 siblings, 0 replies; 3+ messages in thread
From: Thomas Richter @ 2026-03-11 5:52 UTC (permalink / raw)
To: Ian Rogers; +Cc: linux-perf-use., Jan Polensky, Sumanth Korikkar
On 3/9/26 19:23, Ian Rogers wrote:
> On Mon, Mar 9, 2026 at 6:04 AM Thomas Richter <tmricht@linux.ibm.com> wrote:
>>
>> Ian,
>>
>> test case '105: perf stat tests' fails for s390 (linux 7.0.0rc2) . It boils down to this command
>> # ./perf stat -C 0-31 -a true
>> Error:
>> No supported events found.
>>
>> The first bad commit is this one:
>> commit cee275edcdb1 ("perf metricgroup: Don't early exit if no CPUID table exists")
>>
>> Before this commit the output of above command is:
>> Output before that commit:
>> ./perf stat -C 0-31 -a true
>>
>> Performance counter stats for 'system wide':
>>
>> 105 context-switches # 3579.8 cs/sec cs_per_second
>> 29.33 msec cpu-clock # 7.8 CPUs CPUs_utilized
>> 34 cpu-migrations # 1159.2 migrations/sec migrations_per_second
>> 52 page-faults # 1772.9 faults/sec page_faults_per_second
>> 5,026,832 cpu-cycles # 0.2 GHz cycles_frequency
>> 3,029,266 instructions # 0.6 instructions insn_per_cycle
>>
>> 0.000851608 seconds time elapsed
>>
>> Since this is executed on an s390 LPAR, hardware event CPU_CYCLES is available, so it
>> should be used. This is something I will look into.
>>
>> Also some other test cases fail after the bad commit:
>> ./perf test 2>&1 | grep FAILED
>> 99: perf stat CSV output linter : FAILED!
>> 102: perf stat JSON output linter : FAILED!
>> 104: perf stat STD output linter : FAILED!
>> 105: perf stat tests : FAILED!
>> 133: perftool-testsuite_report : FAILED!
>>
>> Before the bad commit the output is
>> ./perf test 2>&1 | grep FAILED
>> 133: perftool-testsuite_report : FAILED!
>>
>> Any ideas where to look too narrow down the debuggung are very welcome.
>
> Hi Thomas, sorry for the breakage! I think the issue is a missed
> commit in the v7.0 PR, specifically:
> https://web.git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools-next.git/commit/tools/perf/util/metricgroup.c?h=perf-tools-next&id=c5a244bf17caf2de22f9e100832b75f72b31d3e6
> This has also shown up as breaking stable backports:
> https://lore.kernel.org/lkml/072e2a07-5c6f-47b5-9695-0a3ffe854ac8@kernel.org/
>
> Thanks,
> Ian
>
Ian, Thanks very much!
This is the fix I needed, in case its needed:
Tested-by: Thomas Richter <tmricht@linux.ibm.com>
--
Thomas Richter, Dept 3303, IBM s390 Linux Development, Boeblingen, Germany
--
IBM Deutschland Research & Development GmbH
Vorsitzender des Aufsichtsrats: Wolfgang Wendt
Geschäftsführung: David Faller
Sitz der Gesellschaft: Böblingen / Registergericht: Amtsgericht Stuttgart, HRB 243294
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2026-03-11 5:52 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-03-09 13:04 Test case '105: perf stat tests' fails for s390 Thomas Richter
2026-03-09 18:23 ` Ian Rogers
2026-03-11 5:52 ` Thomas Richter
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox