[PATCH] tst-{pthread|sched}-affinity-inheritance: Only use online CPUs in affinity mask.

[Stefan Liebler]

I've just send a V2:
[PATCH v2] tst-{pthread|sched}-affinity-inheritance: Only use in-mask
CPUs in affinity mask.
https://inbox.sourceware.org/libc-alpha/20260123135642.1328020-1-stli@linux.ibm.com/T/#u

On 1/22/26 18:27, Florian Weimer wrote:
> * Stefan Liebler:
> 
>> On s390x, the two tests are failing on a lpar with linux 6.18 booted with nosmt:
>> FAIL:  misc/tst-sched-affinity-inheritance
>> FAIL:  nptl/tst-pthread-affinity-inheritance
>>
>> Both tests were failing because they are claiming that the CPU affinity mask was
>> not correctly inherited in the forked process/thread.
>> The error already happened before forking, as the mask was set up to use
>> CPUs 0-14 in set_cpu_mask().  But according to e.g. lscpu or the line
>> "Cpus_allowed_list" in /proc/self/status, only the following CPUs were online:
>> 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30
>> while 1, 3, 5, ... were configured, but not online.
>>
>> Now the test first gets the current CPU mask at startup in do_test() and
>> set_cpu_mask() only selects those online CPUs.
> 
> The “online” part is a bit confusing to me because I associated that
> with /sys/devices/system/cpu/online.
> 
> Maybe “in-mask CPUs” is a good short alternative?
I've adjusted the comments in V2.

> 
> To be clear, checking the startup mask is the right thing to do.  I
> think going to /sys and trying to pick up CPUs from there would be
> confusing.
> 
>> diff --git a/nptl/tst-skeleton-affinity-inheritance.c b/nptl/tst-skeleton-affinity-inheritance.c
>> index e1f328ae26..0910727cc3 100644
>> --- a/nptl/tst-skeleton-affinity-inheritance.c
>> +++ b/nptl/tst-skeleton-affinity-inheritance.c
>> @@ -45,6 +45,7 @@ struct test_param
>>    int nproc;
>>    int nproc_configured;
>>    cpu_set_t *set;
>> +  cpu_set_t *set_at_startup;
>>    size_t size;
>>    bool entry;
>>  };
>> @@ -58,12 +59,53 @@ set_cpu_mask (struct test_param *param, bool entry)
>>    if (entry && cpus > 1)
>>      cpus--;
>>  
>> +  /* Only use CPUs which were online at start of the test process.  */
>>    CPU_ZERO_S (param->size, param->set);
>> -  while (cpus > 0)
>> -    CPU_SET_S (--cpus, param->size, param->set);
>> +  for (int cur = 0; cur < param->nproc_configured && cpus > 0; cur ++)
>> +    {
>> +      if (CPU_ISSET_S (cur, param->size, param->set_at_startup))
>> +	{
>> +	  CPU_SET_S (cur, param->size, param->set);
>> +	  cpus --;
>> +	}
> 
> Extra spaces: “cur ++“, “cpus --”
Done in V2
> 
> And I think the check against param->nproc_configured is not needed for
> this loop.
Done in V2. Also in the loop in verify_my_affinity().
> 
>> +static void
>> +verify_my_affinity (int nproc_configured, size_t size,
>> +		    const cpu_set_t *expected_set)
>> +{
>> +  cpu_set_t *set = CPU_ALLOC (nproc_configured);
>> +  cpu_set_t *xor_set = CPU_ALLOC (nproc_configured);
>> +
>> +  if (set == NULL || xor_set== NULL)
>> +    FAIL_EXIT1 ("verify_my_affinity: Failed to allocate cpuset: %m\n");
> 
> Missing space: “xor_set==”
Done in V2
> 
> Overall it looks okay to me, but I haven't tested it yet.
> 
> Thanks,
> Florian
> 
Thanks,
Stefan