Re: [PATCH v14 7/7] timers: Exclude isolated cpus from timer migration

[Gabriele Monaco]

On Wed, 2025-11-12 at 17:04 +0100, Frederic Weisbecker wrote:
> Le Tue, Nov 04, 2025 at 11:47:39AM +0100, Gabriele Monaco a écrit :
> 
> 
> This duplicates a lot tmigr_isolated_exclude_cpumask().
> Would this work?
> 
> static int __init tmigr_init_isolation(void)
> {
> 	cpumask_var_t cpumask __free(free_cpumask_var) = CPUMASK_VAR_NULL;
> 
> 	static_branch_enable(&tmigr_exclude_isolated);
> 
> 	if (!housekeeping_enabled(HK_TYPE_DOMAIN))
> 		return 0;
> 	if (!alloc_cpumask_var(&cpumask, GFP_KERNEL))
> 		return -ENOMEM;
> 
> 	cpumask_andnot(cpumask, cpu_possible_mask,
> housekeeping_cpumask(HK_TYPE_DOMAIN));
> 
> 	return tmigr_isolated_exclude_cpumask(cpumask);
> }
> 
> If so please add my Reviewed-by to the next version.

Good point, it should work, I'm going to test it tomorrow.

What comes to my mind is that tmigr_isolated_exclude_cpumask expects the hotplug
lock to be held, so I should probably just hold it before calling it:

static int __init tmigr_init_isolation(void)
{
	cpumask_var_t cpumask __free(free_cpumask_var) = CPUMASK_VAR_NULL;

	static_branch_enable(&tmigr_exclude_isolated);

	if (!housekeeping_enabled(HK_TYPE_DOMAIN))
		return 0;
	if (!alloc_cpumask_var(&cpumask, GFP_KERNEL))
		return -ENOMEM;

	cpumask_andnot(cpumask, cpu_possible_mask,
housekeeping_cpumask(HK_TYPE_DOMAIN));
+	guard(cpus_read_lock)();
	return tmigr_isolated_exclude_cpumask(cpumask);
}

Thanks,
Gabriele