Re: [PATCH] irq: sparse irqs, fix #2

[ebiederm@xmission.com (Eric W. Biederman)]

Andrew Morton <akpm@linux-foundation.org> writes:

> On Thu, 14 Aug 2008 15:36:52 +0200 Ingo Molnar <mingo@elte.hu> wrote:
>
>> +static inline cpumask_t vector_allocation_domain(int cpu)
>> +{
>> +        /* Careful. Some cpus do not strictly honor the set of cpus
>> +         * specified in the interrupt destination when using lowest
>> +         * priority interrupt delivery mode.
>> +         *
>> +         * In particular there was a hyperthreading cpu observed to
>> +         * deliver interrupts to the wrong hyperthread when only one
>> +         * hyperthread was specified in the interrupt desitination.
>> +         */
>> +        cpumask_t domain = { { [0] = APIC_ALL_CPUS, } };
>> +        return domain;
>> +}
>
> I haven't looked at callers of this, but...
>
> Does it need to be allocated on the stack?  Local cpumask_t's are a
> size problem.  Can we build this in .rodata at compile time instead?
>
> Is this the caller?

Yes.

>
> +	for_each_cpu_mask(cpu, mask) {
> +		cpumask_t domain, new_mask;
> +		int new_cpu;
> +		int vector;
> +
> +		domain = vector_allocation_domain(cpu);
> +		cpus_and(new_mask, domain, cpu_online_map);
>
> If so we could perhaps do
>
>
> static noinline const cpumask_t *vector_allocation_domain(int cpu)
> {
>         /* Careful. Some cpus do not strictly honor the set of cpus
>          * specified in the interrupt destination when using lowest
>          * priority interrupt delivery mode.
>          *
>          * In particular there was a hyperthreading cpu observed to
>          * deliver interrupts to the wrong hyperthread when only one
>          * hyperthread was specified in the interrupt desitination.
>          */
>         static const cpumask_t domain = { { [0] = APIC_ALL_CPUS, } };
>         return &domain;
> }
>
>
> ...
>
> +	for_each_cpu_mask(cpu, mask) {
> +		cpumask_t domain, new_mask;
> +		int new_cpu;
> +		int vector;
> +
> +		__cpus_and(new_mask, vector_allocation_domain(cpu),
> +				&cpu_online_map);
>
> otoh, perhaps this new function is one implementation of
> genapic.vector_allocation_domain(), in which case the inlining was
> unneeded and misleading.

Likely.  Why these things live in header files...

> I give up.  Have a little think about the stack bloat, please.
>
> btw, whoever wrote that function is in need of a tab key.

Unfortunate gradual accreation of functionality.

vector_allocation_domain could perhaps be better named.  Round up
this cpu to the set of cpus I need to allocate a vector on.

Eric