Re: KVM NV + SVE host OS warning

[Marc Zyngier <maz@kernel.org>]

On Wed, 08 Oct 2025 16:22:31 +0100,
Marc Zyngier <maz@kernel.org> wrote:
> 
> On Wed, 08 Oct 2025 14:43:29 +0100,
> Jan Kotas <jank@cadence.com> wrote:
> > >>
> > >> Is it possible to trap/emulate HCR_EL2, and make sure E2H is always 1,
> > >> in cases when only KVM_ARM_VCPU_HAS_EL2 is set?
> > >> Would that help?
> > > 
> > > There are no traps for HCR_EL2.
> > > 
> > >> As a test, I tried forcing this bit before running VCPUs, but
> > >> reading it via MRS got a value without it being set.
> > > 
> > > Where did you read that from?
> > 
> > I checked the X1 register, after: mrs x1, hcr_el2 in __check_hvhe.
> 
> That's too late. At this point, the write to HCR_EL2 will have already
> occurred,
> 
> > 
> > 
> > > If you apply the following change to your guest, does it start
> > > behaving?
> > > 
> > > M.
> > > 
> > > diff --git a/arch/arm64/include/asm/el2_setup.h b/arch/arm64/include/asm/el2_setup.h
> > > index 46033027510cc..392c9f4016f2e 100644
> > > --- a/arch/arm64/include/asm/el2_setup.h
> > > +++ b/arch/arm64/include/asm/el2_setup.h
> > > @@ -34,7 +34,7 @@
> > > mrs_s x1, SYS_ID_AA64MMFR4_EL1
> > > sbfx x1, x1, #ID_AA64MMFR4_EL1_E2H0_SHIFT, #ID_AA64MMFR4_EL1_E2H0_WIDTH
> > > cmp x1, #0
> > > - b.ge .LnVHE_\@
> > > +// b.ge .LnVHE_\@
> > > 
> > > orr x0, x0, #HCR_E2H
> > > .LnVHE_\@:
> > > 
> > 
> > With this change Guest OS starts the boot process.
> > 
> > [ 0.000000] CPU features: detected: Virtualization Host Extensions
> > [ 6.998696] SMP: Total of 4 processors activated.
> > [ 7.257045] CPU: All CPU(s) started at EL2
> > [ 185.781062] SVE: maximum available vector length 16 bytes per vector
> > [ 186.104159] SVE: default vector length 16 bytes per vector
> 
> Great. That confirms my suspicion that we cannot advertise VHE on CPUs
> that do not have FEAT_FGT. Oh well. I'll post a patch to forbid VHE
> guests on these machines.

As an alternative, and in an effort to keep at least VHE guests
running on V2 (and other machines suffering from the same situation),
I have posted a potential workaround at [1] for the guest to reliably
detect the situation.

Could you please test it with your setup and report whether this
allows you to boot a VHE guest? You will need to patch the guest.

Note that this still won't let you recursively virtualise EL2 (the HW
can do it, but KVM in the guest doesn't know it), but that's better
than nothing.

Thanks,

	M.

[1] https://lore.kernel.org/r/20251009121239.29370-1-maz@kernel.org

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