Sean Christopherson <sean.j.christopherson@xxxxxxxxx> writes:
> On Sat, Jan 18, 2020 at 10:42:31PM +0100, Paolo Bonzini wrote:
>> IMHO the features should stay available in case the guest chooses not to
>> use eVMCS. A guest that uses eVMCS should know which features it cannot
>> use and not enable them.
>
> Makes sense, wasn't thinking about the scenario where the guest doesn't
> support eVMCS or doesn't want to use it for whatever reason.
>
> Rather than update vmx->nested.msrs or filter vmx_get_msr(), what about
> manually adding eVMCS consistency checks on the disallowed bits and handle
> SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES as a one-off case by simply
> clearing it from the eVMCS?
Unfortunately, this doesn't work because ... Windows. Not only Hyper-V
enables SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES, it actually expects it
to work (somehow) so when I do
diff --git a/arch/x86/kvm/vmx/evmcs.c b/arch/x86/kvm/vmx/evmcs.c
index 72359709cdc1..e6c30eec2817 100644
--- a/arch/x86/kvm/vmx/evmcs.c
+++ b/arch/x86/kvm/vmx/evmcs.c
@@ -361,11 +361,5 @@ int nested_enable_evmcs(struct kvm_vcpu *vcpu,
if (evmcs_already_enabled)
return 0;
- vmx->nested.msrs.pinbased_ctls_high &= ~EVMCS1_UNSUPPORTED_PINCTRL;
- vmx->nested.msrs.entry_ctls_high &= ~EVMCS1_UNSUPPORTED_VMENTRY_CTRL;
- vmx->nested.msrs.exit_ctls_high &= ~EVMCS1_UNSUPPORTED_VMEXIT_CTRL;
- vmx->nested.msrs.secondary_ctls_high &= ~EVMCS1_UNSUPPORTED_2NDEXEC;
- vmx->nested.msrs.vmfunc_controls &= ~EVMCS1_UNSUPPORTED_VMFUNC;
-
return 0;
}
diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
index bb8afe0c5e7f..cd1f5a1c884b 100644
--- a/arch/x86/kvm/vmx/nested.c
+++ b/arch/x86/kvm/vmx/nested.c
@@ -1590,7 +1590,7 @@ static int copy_enlightened_to_vmcs12(struct vcpu_vmx *vmx)
evmcs->pin_based_vm_exec_control;
vmcs12->vm_exit_controls = evmcs->vm_exit_controls;
vmcs12->secondary_vm_exec_control =
- evmcs->secondary_vm_exec_control;
+ evmcs->secondary_vm_exec_control & ~SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES;
}
if (unlikely(!(evmcs->hv_clean_fields &
Hyper-V 2016 with > 1 vCPU fails to boot :-(
--
Vitaly
