Re: cpu_rmap maps CPUs to wrong interrupts after reprogramming affinities

[Jakub Kicinski]

On Fri, 13 Dec 2024 10:18:30 -0800 Caleb Sander wrote:
> I can see a few possible ways to address this:
> - Store the current affinity masks for all the IRQs in struct cpu_rmap
> so the next closest IRQ can be computed when a CPU's closest IRQ is
> invalidated. This would significantly increase the size of struct
> cpu_rmap.

Ahmed is actively working on something along those lines:
https://lore.kernel.org/all/20241218165843.744647-1-ahmed.zaki@xxxxxxxxx/
hopefully we can leverage it?