Re: [PATCH 2/2] sched/topology: optimize topology_span_sane()

[Valentin Schneider]

On 02/08/24 10:57, Yury Norov wrote:
> The function may call cpumask_equal with tl->mask(cpu) == tl->mask(i),
> even when cpu != i.

For which architecture have you observed this? AFAIA all implementations of
tl->sched_domain_mask_f are built on a per-CPU cpumask.

e.g. for the default topology, we have:

  cpu_smt_mask() -> topology_sibling_cpumask()

which is implemented as:

  arch/loongarch/include/asm/topology.h:35:#define topology_sibling_cpumask(cpu)		(&cpu_sibling_map[cpu])
  arch/mips/include/asm/topology.h:18:#define topology_sibling_cpumask(cpu)		(&cpu_sibling_map[cpu])
  arch/powerpc/include/asm/topology.h:139:#define topology_sibling_cpumask(cpu)	(per_cpu(cpu_sibling_map, cpu))
  arch/s390/include/asm/topology.h:31:#define topology_sibling_cpumask(cpu)	  (&cpu_topology[cpu].thread_mask)
  arch/sparc/include/asm/topology_64.h:50:#define topology_sibling_cpumask(cpu)		(&per_cpu(cpu_sibling_map, cpu))
  arch/x86/include/asm/topology.h:186:#define topology_sibling_cpumask(cpu)		(per_cpu(cpu_sibling_map, cpu))
  include/linux/arch_topology.h:91:#define topology_sibling_cpumask(cpu)	(&cpu_topology[cpu].thread_sibling)
  include/linux/topology.h:218:#define topology_sibling_cpumask(cpu)		cpumask_of(cpu)

and ditto for cpu_coregroup_mask() & cpu_cpu_mask().