Re: [PATCH v4 2/4] x86/spec-ctrl: defer context-switch IBPB until guest entry

[Roger Pau Monné <roger.pau@xxxxxxxxxx>]

On Mon, Dec 18, 2023 at 02:58:01PM +0100, Jan Beulich wrote:
> On 18.12.2023 13:39, Roger Pau Monné wrote:
> > On Tue, Feb 14, 2023 at 05:11:05PM +0100, Jan Beulich wrote:
> >> In order to avoid clobbering Xen's own predictions, defer the barrier as
> >> much as possible. Merely mark the CPU as needing a barrier issued the
> >> next time we're exiting to guest context.
> > 
> > While I understand that doing the flush in the middle of the guest
> > context might not be ideal, as it's my understanding we also
> 
> s/guest context/context switch/?

Indeed, sorry.

> > needlessly flush Xen predictions, I'm unsure whether this makes any
> > difference in practice, and IMO just makes the exit to guest paths
> > more complex.
> 
> I need to redirect this question to Andrew, who suggested that doing so
> can be expected to make a difference. When we were discussing this, I
> could easily see it might make a difference, but I cannot provide hard
> proof.

That's fine, but with the added complexity in the return to guests
paths I think we need some kind of justification for such a change.
If it was the other way around I could easily see it as a benefits if
just for code clarity reasons.

Thanks, Roger.