On Thu, 2020-11-26 at 19:00 +0100, Paolo Bonzini wrote:
> static int kvm_vcpu_ready_for_interrupt_injection(struct kvm_vcpu
> *vcpu)
> {
> + WARN_ON(pic_in_kernel(vcpu->kvm));
> +
But what if I *want* to inject Xen event channel interrupts while the
actual PIC is in the kernel? :)
Not that I'll have to once the kernel is fixed and I can enable my
shiny new userspace PIC/PIT/IOAPIC code, I suppose....
> return kvm_arch_interrupt_allowed(vcpu) &&
> - !kvm_cpu_has_interrupt(vcpu) &&
> !kvm_event_needs_reinjection(vcpu) &&
> - kvm_cpu_accept_dm_intr(vcpu);
> + (!lapic_in_kernel(vcpu)
> + || (kvm_apic_accept_pic_intr(vcpu)
> + && !pending_userspace_extint(v));
> }
>
I'll give this version a spin...
static int kvm_vcpu_ready_for_interrupt_injection(struct kvm_vcpu *vcpu)
{
return kvm_arch_interrupt_allowed(vcpu) &&
!kvm_event_needs_reinjection(vcpu) &&
(!lapic_in_kernel(vcpu)
|| (kvm_apic_accept_pic_intr(vcpu)
&& vcpu->arch.pending_external_vector == -1));
}
Attachment:
smime.p7s
Description: S/MIME cryptographic signature
