Re: [PATCH v2 4/5] KVM: x86: Add support for legacy VMware backdoors in nested setups

[Sean Christopherson <seanjc@google.com>]

On Tue, Apr 22, 2025, Zack Rusin wrote:
> diff --git a/arch/x86/kvm/svm/nested.c b/arch/x86/kvm/svm/nested.c
> index 04c375bf1ac2..74c472e51479 100644
> --- a/arch/x86/kvm/svm/nested.c
> +++ b/arch/x86/kvm/svm/nested.c
> @@ -22,6 +22,7 @@
>  #include <asm/debugreg.h>
>  
>  #include "kvm_emulate.h"
> +#include "kvm_vmware.h"
>  #include "trace.h"
>  #include "mmu.h"
>  #include "x86.h"
> @@ -1517,6 +1518,11 @@ int nested_svm_exit_special(struct vcpu_svm *svm)
>  			 svm->vcpu.arch.apf.host_apf_flags)
>  			/* Trap async PF even if not shadowing */
>  			return NESTED_EXIT_HOST;
> +#ifdef CONFIG_KVM_VMWARE
> +		else if ((exit_code == (SVM_EXIT_EXCP_BASE + GP_VECTOR)) &&
> +			 kvm_vmware_wants_backdoor_to_l0(vcpu, to_svm(vcpu)->vmcb->save.cpl))
> +			return NESTED_EXIT_HOST;
> +#endif

Either provide a stub or do

		else if (IS_ENABLED(CONFIG_KVM_VMWARE) && ...)

Don't do both.  And definitely don't add a stub and #ifdef (some) callers.  I'd
say just drop the #ifdef and rely on the kvm_vmware_wants_backdoor_to_l0() stub
to get the compiler to optimize out the entire elif.

> @@ -6386,6 +6387,11 @@ static bool nested_vmx_l0_wants_exit(struct kvm_vcpu *vcpu,
>  			return true;
>  		else if (is_ve_fault(intr_info))
>  			return true;
> +#ifdef CONFIG_KVM_VMWARE
> +		else if (is_gp_fault(intr_info) &&
> +			 kvm_vmware_wants_backdoor_to_l0(vcpu, vmx_get_cpl(vcpu)))
> +			return true;
> +#endif

Same thing here.