On Wed, Sep 29, 2010 at 09:17:14AM +0800, Sheng Yang wrote:
> > > + else {
> > > + enable_irq(irq);
> > > + if (assigned_dev->guest_msix_entries[index].flags &
> > > + KVM_ASSIGNED_MSIX_PENDING)
> > > + schedule_work(&assigned_dev->interrupt_work);
> > > + }
> >
> > Hmm, won't this lose interrupts which were sent while bit was pending?
> > It is also pretty heavy if as you say guests touch the mask a lot.
> > I think we must keep the interrupt disabled, just set a bit
> > and delay interrupt injection until vector is unmasked
> > or deleted. The interface to do this will need more thought:
> > e.g. how can userspace clear this bit then?
>
> I think it's fine. Because we didn't modify pending bit here, and the interrupt
> handler would schedule an work to check it regardless of if the IRQ is disable. By
> this meaning, no interrupt would be lose.
AFAICS unmasking the host irq will trigger pending message. If thats
correct, it should be fine.
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
