On 30/07/2015 08:21, Steve Rutherford wrote:
> */
> int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v)
> {
> - if (!irqchip_in_kernel(v->kvm))
> + if (!pic_in_kernel(v->kvm))
> return v->arch.interrupt.pending;
>
> if (kvm_cpu_has_extint(v))
> @@ -75,7 +88,7 @@ int kvm_cpu_has_injectable_intr(struct kvm_vcpu *v)
> */
> int kvm_cpu_has_interrupt(struct kvm_vcpu *v)
> {
> - if (!irqchip_in_kernel(v->kvm))
> + if (!pic_in_kernel(v->kvm))
> return v->arch.interrupt.pending;
>
> if (kvm_cpu_has_extint(v))
> @@ -103,7 +123,7 @@ int kvm_cpu_get_interrupt(struct kvm_vcpu *v)
> {
> int vector;
>
> - if (!irqchip_in_kernel(v->kvm))
> + if (!pic_in_kernel(v->kvm) && v->arch.interrupt.pending)
> return v->arch.interrupt.nr;
>
> vector = kvm_cpu_get_extint(v);
I have one more doubt about these three hunks.
v->arch.interrupt should not be used at all with split irqchip. In
particular:
- kvm_cpu_has_injectable_intr should go through kvm_cpu_has_extint and
query pending_userspace_extint
- same for kvm_cpu_has_interrupt
- kvm_cpu_get_interrupt should go through kvm_cpu_get_extint and
return/clear v->arch.pending_external_vector.
So I think !irqchip_in_kernel(v->kvm) is the right test. In
particular, with pic_in_kernel, kvm_cpu_has_extint's irqchip_split case
is dead. I am then not sure how you reach this code in x86.c:
/* kvm_cpu_has_injectable_intr doesn't take extints into account? */
} else if (kvm_cpu_has_injectable_intr(vcpu)) {
/*
* Because interrupts can be injected asynchronously, we are
* calling check_nested_events again here to avoid a race
condition.
* See https://lkml.org/lkml/2014/7/2/60 for discussion about
this
* proposal and current concerns. Perhaps we should be setting
* KVM_REQ_EVENT only on certain events and not unconditionally?
*/
if (is_guest_mode(vcpu) && kvm_x86_ops->check_nested_events) {
r = kvm_x86_ops->check_nested_events(vcpu, req_int_win);
if (r != 0)
return r;
}
if (kvm_x86_ops->interrupt_allowed(vcpu)) {
/*
* kvm_cpu_get_interrupt does take extints into account
* because of the " && v->arch.interrupt.pending", but
* you won't get here unless you have an APIC interrupt
* pending!
*/
kvm_queue_interrupt(vcpu, kvm_cpu_get_interrupt(vcpu),
false);
kvm_x86_ops->set_irq(vcpu);
}
}
Paolo
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
