Re: [PATCH 00/16] KVM: arm64: Rework timer offsetting for fun and profit

[Marc Zyngier <maz@kernel.org>]

Hi Simon,

On Tue, 21 Feb 2023 16:28:36 +0000,
"Veith, Simon" <sveith@amazon.de> wrote:
> 
> Hello Marc,
> 
> On Thu, 2023-02-16 at 14:21 +0000, Marc Zyngier wrote:
> > This series aims at satisfying multiple goals:
> > 
> > - allow a VMM to atomically restore a timer offset for a whole VM
> >   instead of updating the offset each time a vcpu get its counter
> >   written
> > 
> > - allow a VMM to save/restore the physical timer context, something
> >   that we cannot do at the moment due to the lack of offsetting
> > 
> > - provide a framework that is suitable for NV support, where we get
> >   both global and per timer, per vcpu offsetting
> 
> Thank you so much for following up on my (admittedly very basic) patch
> with your own proposal!

No worries. There is nothing like "nerd sniping"... ;-)

> > This has been moderately tested with nVHE, VHE and NV. I do not have
> > access to CNTPOFF-aware HW, so the jury is still out on that one
> 
> Same here about CNTPOFF -- I gave it a quick spin on Graviton2 and
> Graviton3, and neither chip claims the ARM64_HAS_ECV_CNTPOFF capability
> from your patch.

Nah, G2/G3 are old stuff (v8.2 and v8.3+ respectively). You need at
least a v8.6 system, something that is made of the best Unobtainium
money can buy. M2 has ECV, but without CNTPOFF, which is just silly.

> I am working on testing your series with our userspace and will report
> back.

Thanks a lot, that'd be super helpful. Please don't get too attached
to the userspace interface though, as it is likely to change (Oliver
has an interesting suggestion to simplify it, but I need to convince
myself that it doesn't break migration of the physical timer).

Thanks,

	M.

-- 
Without deviation from the norm, progress is not possible.