KVM: SVM: Handle nested selective_cr0 intercept correctly
[pandora-kernel.git] / arch / x86 / kvm / svm.c
index b4aac5c..631d2e5 100644 (file)
@@ -1043,6 +1043,27 @@ static void svm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
 {
        struct vcpu_svm *svm = to_svm(vcpu);
 
+       if (is_nested(svm)) {
+               /*
+                * We are here because we run in nested mode, the host kvm
+                * intercepts cr0 writes but the l1 hypervisor does not.
+                * But the L1 hypervisor may intercept selective cr0 writes.
+                * This needs to be checked here.
+                */
+               unsigned long old, new;
+
+               /* Remove bits that would trigger a real cr0 write intercept */
+               old = vcpu->arch.cr0 & SVM_CR0_SELECTIVE_MASK;
+               new = cr0 & SVM_CR0_SELECTIVE_MASK;
+
+               if (old == new) {
+                       /* cr0 write with ts and mp unchanged */
+                       svm->vmcb->control.exit_code = SVM_EXIT_CR0_SEL_WRITE;
+                       if (nested_svm_exit_handled(svm) == NESTED_EXIT_DONE)
+                               return;
+               }
+       }
+
 #ifdef CONFIG_X86_64
        if (vcpu->arch.efer & EFER_LME) {
                if (!is_paging(vcpu) && (cr0 & X86_CR0_PG)) {