Re: [PATCH v3 1/2] sched/topology: Don't try to build empty sched domains

[Dietmar Eggemann <dietmar.eggemann@xxxxxxx>]

On 15/10/2019 17:42, Valentin Schneider wrote:

[...]

> diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
> index c52bc91f882b..a859e5539440 100644
> --- a/kernel/cgroup/cpuset.c
> +++ b/kernel/cgroup/cpuset.c
> @@ -817,6 +817,11 @@ static int generate_sched_domains(cpumask_var_t **domains,
>  		struct cpuset *a = csa[i];
>  		int apn = a->pn;
>  
> +		if (cpumask_empty(a->effective_cpus)) {
> +			ndoms--;
> +			continue;
> +		}
> +
>  		for (j = 0; j < csn; j++) {
>  			struct cpuset *b = csa[j];
>  			int bpn = b->pn;
> @@ -859,6 +864,9 @@ static int generate_sched_domains(cpumask_var_t **domains,
>  			continue;
>  		}
>  
> +		if (cpumask_empty(a->effective_cpus))
> +			continue;
> +

Can you not just prevent that a cpuset pointer (cp) is added to the
cpuset array (csa[]) in case cpumask_empty(cp->effective_cpus)?

@@ -798,9 +800,14 @@ static int generate_sched_domains(cpumask_var_t
**domains, cpumask_subset(cp->cpus_allowed, top_cpuset.effective_cpus))
                        continue;

-   if (is_sched_load_balance(cp))
+   if (is_sched_load_balance(cp) && !cpumask_empty(cp->effective_cpus))
            csa[csn++] = cp;

>  		dp = doms[nslot];
>  
>  		if (nslot == ndoms) {

[...]