On 23.08.22 23:38, Richard Henderson wrote:
> First, as pointed out by David; second by inspection.
>
> I really wish there were a better way to structure this,
> but alas, I don't see any alternatives that aren't just
> different but similar amounts of ugly.
>
The only feasible way would be having a arch-specific callback from
inside the probe code that would, similarly to tlb_fill code for !USER
store these values in the cpu environment -- then we could similarly
just look them up after the probe access.
--
Thanks,
David / dhildenb