Re: [PATCH v2 2/3] x86/smp: Allow forcing the mwait hint for play dead loop

From: Dave Hansen
Date: Wed Nov 06 2024 - 09:47:00 EST


On 11/6/24 00:14, Artem Bityutskiy wrote:
> The non-architectural part of it is basically this.
>
> The algorithm assumes that in the sub-state part of the mwait hint (4 least
> significant bits) always starts with 0 and then next sub-state is always 1, and
> there are no gaps. This is not documented in Intel SDM. This happens to work on
> majority of Intel platforms, but this is not architectural.

I wouldn't call that a "non-architectural algorithm". It's a bug, plain
and simple.