Re: [PATCH v8 00/26] Enable CET Virtualization

[Sean Christopherson]

On Mon, Jan 08, 2024, Weijiang Yang wrote:
> On 1/6/2024 12:21 AM, Sean Christopherson wrote:
> > On Fri, Jan 05, 2024, Weijiang Yang wrote:
> > > On 1/5/2024 8:54 AM, Sean Christopherson wrote:
> > > > On Fri, Jan 05, 2024, Rick P Edgecombe wrote:
> > > > > > For CALL/RET (and presumably any branch instructions with IBT?) other
> > > > > > instructions that are directly affected by CET, the simplest thing would
> > > > > > probably be to disable those in KVM's emulator if shadow stacks and/or IBT
> > > > > > are enabled, and let KVM's failure paths take it from there.
> > > > > Right, that is what I was wondering might be the normal solution for
> > > > > situations like this.
> > > > If KVM can't emulate something, it either retries the instruction (with some
> > > > decent logic to guard against infinite retries) or punts to userspace.
> > > What kind of error is proper if KVM has to punt to userspace?
> > KVM_INTERNAL_ERROR_EMULATION.  See prepare_emulation_failure_exit().
> > 
> > > Or just inject #UD into guest on detecting this case?
> > No, do not inject #UD or do anything else that deviates from architecturally
> > defined behavior.
> 
> Thanks!
> But based on current KVM implementation and patch 24, seems that if CET is exposed
> to guest, the emulation code or shadow paging mode couldn't be activated at the same time:

No, requiring unrestricted guest only disables the paths where KVM *delibeately*
emulates the entire guest code stream.  In no way, shape, or form does it prevent
KVM from attempting to emulate arbitrary instructions.

> In vmx.c,
> hardware_setup(void):
> if (!cpu_has_vmx_unrestricted_guest() || !enable_ept)
>         enable_unrestricted_guest = 0;
> 
> in vmx_set_cr0():
> [...]
>         if (enable_unrestricted_guest)
>                 hw_cr0 |= KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST;
>         else {
>                 hw_cr0 |= KVM_VM_CR0_ALWAYS_ON;
>                 if (!enable_ept)
>                         hw_cr0 |= X86_CR0_WP;
> 
>                 if (vmx->rmode.vm86_active && (cr0 & X86_CR0_PE))
>                         enter_pmode(vcpu);
> 
>                 if (!vmx->rmode.vm86_active && !(cr0 & X86_CR0_PE))
>                         enter_rmode(vcpu);
>         }
> [...]
> 
> And in patch 24:
> 
> +   if (!cpu_has_load_cet_ctrl() || !enable_unrestricted_guest ||
> +       !cpu_has_vmx_basic_no_hw_errcode()) {
> +       kvm_cpu_cap_clear(X86_FEATURE_SHSTK);
> +       kvm_cpu_cap_clear(X86_FEATURE_IBT);
> +   }
> 
> Not sure if I missed anything.
> 
>