Re: [PATCH v12 22/24] KVM: nVMX: Enable CET support for nested guest

[Chao Gao <chao.gao@intel.com>]

>@@ -4479,6 +4543,9 @@ static bool is_vmcs12_ext_field(unsigned long field)
> 	case GUEST_IDTR_BASE:
> 	case GUEST_PENDING_DBG_EXCEPTIONS:
> 	case GUEST_BNDCFGS:
>+	case GUEST_S_CET:
>+	case GUEST_SSP:
>+	case GUEST_INTR_SSP_TABLE:
> 		return true;
> 	default:
> 		break;
>@@ -4529,6 +4596,10 @@ static void sync_vmcs02_to_vmcs12_rare(struct kvm_vcpu *vcpu,
> 	vmcs12->guest_pending_dbg_exceptions =
> 		vmcs_readl(GUEST_PENDING_DBG_EXCEPTIONS);
> 
>+	cet_vmcs_fields_get(&vmx->vcpu, &vmcs12->guest_s_cet,
>+			    &vmcs12->guest_ssp,
>+			    &vmcs12->guest_ssp_tbl);
>+
> 	vmx->nested.need_sync_vmcs02_to_vmcs12_rare = false;
> }
> 
>@@ -4760,6 +4831,10 @@ static void load_vmcs12_host_state(struct kvm_vcpu *vcpu,
> 	if (vmcs12->vm_exit_controls & VM_EXIT_CLEAR_BNDCFGS)
> 		vmcs_write64(GUEST_BNDCFGS, 0);
> 
>+	if (vmcs12->vm_exit_controls & VM_EXIT_LOAD_CET_STATE)
>+		cet_vmcs_fields_set(vcpu, vmcs12->host_s_cet, vmcs12->host_ssp,
>+				    vmcs12->host_ssp_tbl);
>+

Xin wrote a new test [*] and found a bug here. If VM_EXIT_LOAD_CET_STATE is not
set, the guest values should be retained after the nested vm-exit.

so here should be

	if (vmcs12->vm_exit_controls & VM_EXIT_LOAD_CET_STATE)
		cet_vmcs_fields_set(vcpu, vmcs12->host_s_cet, vmcs12->host_ssp,
				    vmcs12->host_ssp_tbl);
	else
		cet_vmcs_fields_set(vcpu, vmcs12->guest_s_cet, vmcs12->guest_ssp,
				    vmcs12->guest_ssp_tbl);

This creates a dependency that vmcs12->guest_s_cet/ssp/ssp_tbl must be
up-to-date here. So, vmcs12->guest_s_cet/ssp/ssp_tbl should be synced from
vmcs02 on each nested VM-exit rather than lazily in sync_vmcs02_to_vmcs12_rare().
Specifically, the cet_vmcs_fields_get() in sync_vmcs02_to_vmcs12_rare() should
be moved to sync_vmcs02_to_vmcs12(), and is_vmcs12_ext_field() should return
false for CET VMCS fields.

Note that Xin's test differs from the test I wrote [**] in L2 guest behavior.
His test writes to the S_CET MSR, while my test reads the S_CET MSR, which
is why this bug escaped from my CET test.

[*]: https://github.com/xinli-intel/kvm-unit-tests/commit/f1df81c3189a3328adb47c7dd6cd985830fe738f
[**]: https://lore.kernel.org/kvm/20250626073459.12990-9-minipli@grsecurity.net/

below diff can fix this issue:

diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
index 7c88fedc27c7..eadd659ae22f 100644
--- a/arch/x86/kvm/vmx/nested.c
+++ b/arch/x86/kvm/vmx/nested.c
@@ -4596,9 +4596,6 @@ static bool is_vmcs12_ext_field(unsigned long field)
	case GUEST_IDTR_BASE:
	case GUEST_PENDING_DBG_EXCEPTIONS:
	case GUEST_BNDCFGS:
-	case GUEST_S_CET:
-	case GUEST_SSP:
-	case GUEST_INTR_SSP_TABLE:
		return true;
	default:
		break;
@@ -4649,10 +4646,6 @@ static void sync_vmcs02_to_vmcs12_rare(struct kvm_vcpu *vcpu,
	vmcs12->guest_pending_dbg_exceptions =
		vmcs_readl(GUEST_PENDING_DBG_EXCEPTIONS);
 
-	cet_vmcs_fields_get(&vmx->vcpu, &vmcs12->guest_s_cet,
-			    &vmcs12->guest_ssp,
-			    &vmcs12->guest_ssp_tbl);
-
	vmx->nested.need_sync_vmcs02_to_vmcs12_rare = false;
 }
 
@@ -4759,6 +4752,10 @@ static void sync_vmcs02_to_vmcs12(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12)
 
	if (vmcs12->vm_exit_controls & VM_EXIT_SAVE_IA32_EFER)
		vmcs12->guest_ia32_efer = vcpu->arch.efer;
+
+	cet_vmcs_fields_get(&vmx->vcpu, &vmcs12->guest_s_cet,
+			    &vmcs12->guest_ssp,
+			    &vmcs12->guest_ssp_tbl);
 }
 
 /*
@@ -4884,9 +4881,17 @@ static void load_vmcs12_host_state(struct kvm_vcpu *vcpu,
	if (vmcs12->vm_exit_controls & VM_EXIT_CLEAR_BNDCFGS)
		vmcs_write64(GUEST_BNDCFGS, 0);
 
+	/*
+	 * Load CET state from host state if VM_EXIT_LOAD_CET_STATE is set.
+	 * otherwise CET state should be retained across VM-exit, i.e.,
+	 * guest values should be propagated from vmcs12 to vmcs01.
+	 */
	if (vmcs12->vm_exit_controls & VM_EXIT_LOAD_CET_STATE)
		cet_vmcs_fields_set(vcpu, vmcs12->host_s_cet, vmcs12->host_ssp,
				    vmcs12->host_ssp_tbl);
+	else
+		cet_vmcs_fields_set(vcpu, vmcs12->guest_s_cet, vmcs12->guest_ssp,
+				    vmcs12->guest_ssp_tbl);
 
	if (vmcs12->vm_exit_controls & VM_EXIT_LOAD_IA32_PAT) {
		vmcs_write64(GUEST_IA32_PAT, vmcs12->host_ia32_pat);