On Thu, Jul 29, 2010 at 11:49:31AM +0300, Avi Kivity wrote: > On 07/28/2010 08:22 PM, Marcelo Tosatti wrote: > > > >>>>>that needs reinjection (or an exception). > >>>>That can happen event now, no? A pending exception, interrupt comes > >>>>along, injection picks up the exception but leaves the interrupt. > >>>> > >>>>Now the situation can be more complicated: > >>>> > >>>>- pending exception > >>>>- injection > >>>>- interrupt, sets KVM_REQ_EVENT > >>>>- notices KVM_REQ_EVENT > >>>>- drops KVM_REQ_EVENT, cancels exception (made pending again) > >>>>- goes back > >>>>- injection (injects exception again, interrupt is pending) > >>>> > >>>>as far as I can tell, this is all fine. > >>>But you cleared KVM_REQ_EVENT. Which means you're not going to inject > >>>the pending interrupt on the next entry. > >>Doh. So we need to set KVM_REQ_EVENT again, after the final check > >>for vcpu->requests, to make sure we redo injection again. > >> > >>So we can make inject_pending_event() return true if there's more in > >>the queue, and if it did, re-raise KVM_REQ_EVENT just before entry? > >Yeah, that would do it. > > On second and third thoughts, that is unneeded. If an interrupt or > nmi is still pending after event injection, we will request an > interrupt or nmi window which will set KVM_REQ_EVENT. An exception > cannot be pending after an event injection since it is the highest > priority event. > > Yes? Yep. Userspace irqchip is still broken though. Can't see whats wrong with svm. -- 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
