[PATCH v3 00/17] KVM: arm64: Allow using VHE in the nVHE hypervisor

Marc Zyngier maz at kernel.org
Wed Jul 10 01:51:24 PDT 2024


On Wed, 10 Jul 2024 07:45:52 +0100,
Tangnianyao <tangnianyao at huawei.com> wrote:
> 
> Hi Marz,
> 
> I'm trying to learn pKVM and have a question.
> 
> Why pKVM developed on E2H=0 firstly? It tried to avoid host access
> guest memory with stage2 translation, and it seems not necessarily
> rely on HCR_EL2.E2H=0.

Because we don't want to leave ARMv8.0 systems behind. They are still
a large portion of the arm64 HW running Linux.

> Is hVHE an alternative plan of pKVM ? To allow pKVM run on E2H res1 system ?

Supporting E2H RES1 implementations is indeed one of the motivations.
But there is a lot more to it, such as being able to use the two TTBRs
to perform address space isolation inside the hypervisor.

I presented this[1] at KVM Forum two years ago, which explains what we
could do (some of which is already in progress).

	M.

[1] https://static.sched.com/hosted_files/kvmforum2022/b4/NYSM-NYD-KVM-2022.pdf

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



More information about the linux-arm-kernel mailing list