Re: clarification on explicit preemption points in Fiasco IPC

> From: Sergio Ruocco <[email protected]>
> Subject: clarification on explicit preemption points in Fiasco IPC
> As far as I know, Fiasco is entirely preemptable. Then why sprinkling
> the IPC path with Enable IRQs/NOP NOP/Disable IRQs ?
>
> Thanks in advance,
>
>         Sergio

I have neither seen Fiasco code nor have any prior experience with it;
but, my guess would be that during the IPC critical path the irqs are
kept disabled. So, possibly to reduce latency irqs are given a chance
to run at safe points within this path.

If you expect that they're already enabled during that path, then
confirm by checking that sti() and cli() are no-ops as well. They
*must* be no-ops, otherwise particularly cli() would incorrectly
disable the irqs.

Thanks,
Bahadir