On 6 October 2020 22:01:18 BST, Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote:
>On Mon, Oct 05 2020 at 16:28, David Woodhouse wrote:
>>
>> #else /* CONFIG_SMP */
>>
>> +#define irq_default_affinity (NULL)
>
>...
>
>> static int alloc_descs(unsigned int start, unsigned int cnt, int
>node,
>> const struct irq_affinity_desc *affinity,
>> + const struct cpumask *default_affinity,
>> struct module *owner)
>> {
>> struct irq_desc *desc;
>> int i;
>>
>> /* Validate affinity mask(s) */
>> + if (!default_affinity || cpumask_empty(default_affinity))
>> + return -EINVAL;
>
>How is that supposed to work on UP?
Hm, good point.
>Aside of that I really hate to have yet another argument just
>because.
Yeah, I was trying to avoid having to allocate a whole array of irq_affinity_desc just to fill them all in with the same default.
But perhaps I need to treat the "affinity_max" like we do cpu_online_mask, and allow affinity to be set even to offline/unreachable CPUs at this point. Then we can be more relaxed about the default affinities.
--
Sent from my Android device with K-9 Mail. Please excuse my brevity.
