Re: [PATCH] kvm: x86: svm: remove SVM_EXIT_READ_CR* intercepts

["Radim Krčmář" <rkrcmar@redhat.com>]

2015-03-12 17:17-0400, Bandan Das:
> Joel Schopp <joel.schopp@amd.com> writes:
> > @@ -2968,29 +2964,10 @@ static int cr_interception(struct vcpu_svm *svm)
> >  			kvm_queue_exception(&svm->vcpu, UD_VECTOR);
> >  			return 1;
> >  		}
> > -	} else { /* mov from cr */
> > -		[reads of CR 0..8]
> > +	} else { /* mov from cr, should never trap in svm */
> > +		WARN(1, "unhandled read from CR%d", cr);
> > +		kvm_queue_exception(&svm->vcpu, UD_VECTOR);
> > +		return 1;
> 
> Can we end up here if a nested hypervisor sets cr read interception ?

No.  If the nested hypervisor sets intercept bits, we're going to detect
them in 'handle_exit -> nested_svm_exit_handled -> nested_svm_intercept'
and enter L1 before the cr_interception handler.