Re: [PATCH 2/3] kvm: x86: do not use KVM_REQ_EVENT for APICv interrupt injection

From: Paolo Bonzini
Date: Wed Sep 28 2016 - 08:14:27 EST




On 28/09/2016 14:06, Wu, Feng wrote:
>> Then the
>> > guest is entered with PIR.ON, but the PI interrupt is not pending and
>> > hence the interrupt is never delivered to the guest.
> Why "never", at least, the interrupt should be delivered to the guest in the next
> vm-entry, right? I mean vm-entry -> vm-exit -> _vm-entry_ (interrupts will be
> delivered at this vm-entery).

Sure, but you could in principle have a case where the vmexit never
happens (right now nohz_full CPUs have a 1 Hz timer tick, but that's
just a limitation of isolcpus). When that happens, the interrupt might
never be delivered because it is not recorded in IRR.

Also, if the guest issues an EOI, the pending posted interrupt may be
reordered incorrectly with a lower-priority interrupt already in IRR.

But I'll re-check the wording in the commit message before posting the
non-RFC version.

Paolo