Right. The guest's RFLAGS.IF may be clear, but "an external interrupt causes a VM exit if the 'external-interrupt exiting' VM-execution control is 1." On Mon, Nov 27, 2017 at 10:27 AM, Paolo Bonzini <pbonzini@xxxxxxxxxx> wrote: > On 27/11/2017 19:13, Jim Mattson wrote: >> I don't understand the concern regarding CPUID5_ECX_INTERRUPT_BREAK. >> Even if the CPU has this feature, can't the guest bypass it by >> disabling interrupts and invoking MWAIT with bit 0 of ECX clear? > > Hmm, you're right. In fact, an external interrupt should cause a VM > exit if the “external-interrupt exiting” VM-execution control is 1, even > if the VM is executing MWAIT. > > We should write a unit test for that though. > > Thanks, > > Paolo
