From: Martin Pieuchot <mpi@openbsd.org>
Subject: Re: clockintr_queue: add "cq_cpu", a back-pointer to the dispatch CPU
To: Scott Cheloha <scottcheloha@gmail.com>
Cc: tech@openbsd.org
Date: Sat, 24 Feb 2024 16:58:02 +0100

On 23/02/24(Fri) 20:31, Scott Cheloha wrote:
> Thank you for your comments on the tickless timeout roadmap.
> I have an idea of where things may get tricky, but now have
> ideas for managing it.
> 
> --
> 
> Part (1) is to extend struct intrclock with a new function, ic_send().
> ic_send() will allow arbitrary CPUs to poke the dispatch CPU for a
> given clockintr_queue to tell that CPU to reprogram its interrupt
> clock.  On the backend, ic_send() will usually be implemented with an
> IPI.

Please send a full diff that can be tested so we can give comments on
something that works.

Note that clockintr_bind(9) is still not in tree.  Please put it in.