On Mon, May 26, 2014 at 04:56:54PM +0200, Paolo Bonzini wrote:
> Il 26/05/2014 16:28, Michael S. Tsirkin ha scritto:
> >> static inline void apic_clear_isr(int vec, struct kvm_lapic *apic)
> >> {
> >>- if (__apic_test_and_clear_vector(vec, apic->regs + APIC_ISR))
> >>+ struct kvm_vcpu *vcpu;
> >>+ if (!__apic_test_and_clear_vector(vec, apic->regs + APIC_ISR))
> >>+ return;
> >>+
> >>+ vcpu = apic->vcpu;
> >>+
> >>+ /*
> >>+ * We do get here for APIC virtualization enabled if the guest
> >>+ * uses the Hyper-V APIC enlightenment. In this case we may need
> >>+ * to trigger a new interrupt delivery by writing the SVI field;
> >>+ * on the other hand isr_count and highest_isr_cache are unused
> >>+ * and must be left alone.
> >>+ */
> >>+ if (unlikely(kvm_apic_vid_enabled(vcpu->kvm)))
> >>+ kvm_x86_ops->hwapic_isr_update(vcpu->kvm,
> >>+ apic_find_highest_isr(apic));
> >
> >I note that an indirect call on data path is mostly unnecessary here:
> > static int vmx_vm_has_apicv(struct kvm *kvm)
> > {
> > return enable_apicv && irqchip_in_kernel(kvm);
> > }
> >is all it does, and irqchip_in_kernel also has an rmb within it, which
> >is somewhat expensive on x86: and there's no way to reach this code
> >with irqchip disabled, correct?
>
> smp_rmb is just a compiler barrier, it's not expensive. The indirect call
> is probably more expensive.
>
> That said, other places in the paths (kvm_cpu_has_injectable_intr,
> kvm_cpu_get_interrupt) already call kvm_apic_vid_enabled, so this patch
> doesn't introduce something new.
>
> >How about adding a bool flag in kvm_vcpu_arch, and testing that?
>
> Yes, that's possible. It's also possible to test
> kvm_x86_ops->hwapic_isr_update != NULL and zero the field in vmx.c's
> hardware_setup.
> Can you make a patch?
>
> Paolo
On top of this one? Sure.
--
MST
--
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
