On Fri, Feb 21, 2020 at 11:29:49AM +0100, Pavel Machek wrote:
> Hi!
>
> > Hardcode the EPT page-walk level for L2 to be 4 levels, as KVM's MMU
> > currently also hardcodes the page walk level for nested EPT to be 4
> > levels. The L2 guest is all but guaranteed to soft hang on its first
> > instruction when L1 is using EPT, as KVM will construct 4-level page
> > tables and then tell hardware to use 5-level page tables.
>
> I don't get it. 7/191 reverts the patch, then 9/191 reverts the
> revert. Can we simply drop both 7 and 9, for exactly the same result?
>
> (Patch 8 is a unused file, so it does not change the picture).
Patch 07 is reverting this patch from the same unused file,
arch/x86/kvm/vmx/vmx.c[*]. The reason patch 07 looks like a normal diff is
that a prior patch in 4.19.105 created the unused file (which is what's
reverted by patch 08 here).
Patch 09 reintroduces the fix for the correct file, arch/x86/kvm/vmx.c.
[*] In upstream, vmx.c now lives in arch/x86/kvm/vmx/, but in 4.19 and
earlier it lives in arch/x86/kvm/.
>
> Best regards,
> Pavel
>
> > +++ b/arch/x86/kvm/vmx.c
> > @@ -5302,6 +5302,9 @@ static void vmx_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
> >
> > static int get_ept_level(struct kvm_vcpu *vcpu)
> > {
> > + /* Nested EPT currently only supports 4-level walks. */
> > + if (is_guest_mode(vcpu) && nested_cpu_has_ept(get_vmcs12(vcpu)))
> > + return 4;
> > if (cpu_has_vmx_ept_5levels() && (cpuid_maxphyaddr(vcpu) > 48))
> > return 5;
> > return 4;
> > --
> > 2.20.1
> >
> >
>
> --
> (english) http://www.livejournal.com/~pavelmachek
> (cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
