Re: RFC: IRQ affinity (aka interrupt routing)

[Kengo NAKAHARA <k-nakahara%iij.ad.jp@localhost>]

Hi,

Thank you for your idea.

(2014/08/27 10:09), Matt Thomas wrote:
> As I've been reading this discussion, it seems very x86 centric.
>
> I've thinking about adding
>
> void intr_distribute(void *ih, const kcpuset_t *newset, kcpuset_t *oldset)
>
> for my ports that can do MP.  This could be used to obtain the current
> set of cpus setup to receive interrupt for <ih> or set a new sets of
> cpus.  To set an interrupt across all CPUs, you could use
>
> intr_distribibute(ih, &kcpuset_running, NULL);
>
> By default only the boot CPU would be setup to get interrupts when
> the interrupt was established.

It seems good, except return value. IRQ affinity may fail (e.g. when
all cpus are set "nointr" flag), so return value should not be void.
I use the API in MI code like this,
====================
intrctl_ioctl(..., void *data, ...)
{
    switch(cmd) {
    case IOC_INTR_AFFINITY:
        ih = intr_handler(data->intrid);
        if (ih == NULL )
            return EINVAL;

        kcpuset_create(&intr_cpuset, true);
        kcpuset_set(intr_cpuset, data->cpuid);
        error = intr_distribute(ih, intr_cpuset, NULL);
        break;
    }

    return error;
}
====================
Could you comment this design?


BTW, how do you think about MSI/MSI-X proposal?
    - yours
      http://mail-index.netbsd.org/tech-kern/2011/08/05/msg011130.html
    - dyoung's
      http://mail-index.netbsd.org/tech-kern/2014/06/06/msg017209.html
    - mine
      http://mail-index.netbsd.org/tech-kern/2014/07/10/msg017336.html

Thanks,

--
//////////////////////////////////////////////////////////////////////
Internet Initiative Japan Inc.

Device Engineering Section,
Core Product Development Department,
Product Division,
Technology Unit

Kengo NAKAHARA <k-nakahara%iij.ad.jp@localhost>