diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
index 3472916657e1..7d0d95f40cd2 100644
--- a/arch/x86/kvm/svm/svm.c
+++ b/arch/x86/kvm/svm/svm.c
@@ -1054,6 +1054,11 @@ static void svm_recalc_instruction_intercepts(struct kvm_vcpu *vcpu)
 	 * No need to toggle any of the vgif/vls/etc. enable bits here, as they
 	 * are set when the VMCB is initialized and never cleared (if the
 	 * relevant intercepts are set, the enablements are meaningless anyway).
+	 *
+	 * FIXME: When #GP is not intercepted, a #GP on these instructions (e.g.
+	 * due to CPL > 0) could be injected by hardware before the instruction
+	 * is intercepted, leading to #GP taking precedence over #UD from the
+	 * guest's perspective.
 	 */
 	if (!(vcpu->arch.efer & EFER_SVME)) {
 		svm_set_intercept(svm, INTERCEPT_VMLOAD);
@@ -2294,6 +2299,9 @@ static int gp_interception(struct kvm_vcpu *vcpu)
 		if (!is_guest_mode(vcpu))
 			return svm_invoke_exit_handler(vcpu, svm_exit_code);
 
+		if (nested_svm_check_permissions(vcpu))
+			return 1;
+
 		if (!page_address_valid(vcpu, kvm_register_read(vcpu, VCPU_REGS_RAX)))
 			goto reinject;