On Thu, Jul 17, 2014 at 02:04:11PM +0200, Paolo Bonzini wrote:
>Il 17/07/2014 13:28, Paolo Bonzini ha scritto:
>> Il 17/07/2014 13:03, Wanpeng Li ha scritto:
>>> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
>>> index 4ae5ad8..a704f71 100644
>>> --- a/arch/x86/kvm/vmx.c
>>> +++ b/arch/x86/kvm/vmx.c
>>> @@ -8697,6 +8697,9 @@ static void nested_vmx_vmexit(struct kvm_vcpu *vcpu,
>>> u32 exit_reason,
>>> if ((exit_reason == EXIT_REASON_EXTERNAL_INTERRUPT)
>>> && nested_exit_intr_ack_set(vcpu)) {
>>> int irq = kvm_cpu_get_interrupt(vcpu);
>>> +
>>> + if (irq < 0 && kvm_apic_vid_enabled(vcpu->kvm))
>>> + irq = kvm_lapic_find_highest_irr(vcpu);
>>> WARN_ON(irq < 0);
>>> vmcs12->vm_exit_intr_info = irq |
>>> INTR_INFO_VALID_MASK | INTR_TYPE_EXT_INTR;
>>
>> I wonder if this should be kvm_apic_has_interrupt, so that the PPR
>> register is taken into consideration?
>
>
>And actually, I think the acknowledging should include the three steps to
>set-ISR/update-PPR/clear-IRR. (With APICv update PPR is not strictly
>necessary, but it doesn't hurt either).
>
>You cannot let the processor do these because it would deliver the interrupt
>through the IDT, but you still must do it in the hypervisor.
>
>This gives this patch:
>
>diff --git a/arch/x86/kvm/irq.c b/arch/x86/kvm/irq.c
>index bd0da43..a1ec6a5 100644
>--- a/arch/x86/kvm/irq.c
>+++ b/arch/x86/kvm/irq.c
>@@ -108,7 +108,7 @@ int kvm_cpu_get_interrupt(struct kvm_vcpu *v)
>
> vector = kvm_cpu_get_extint(v);
>
>- if (kvm_apic_vid_enabled(v->kvm) || vector != -1)
>+ if (vector != -1)
> return vector; /* PIC */
>
> return kvm_get_apic_interrupt(v); /* APIC */
>diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
>index 3855103..6cbc7af 100644
>--- a/arch/x86/kvm/lapic.c
>+++ b/arch/x86/kvm/lapic.c
>@@ -360,10 +360,20 @@ static inline void apic_clear_irr(int vec, struct
>kvm_lapic *apic)
>
> static inline void apic_set_isr(int vec, struct kvm_lapic *apic)
> {
>- /* Note that we never get here with APIC virtualization enabled. */
>+ if (!__apic_test_and_set_vector(vec, apic->regs + APIC_ISR)) {
>+ /*
>+ * With APIC virtualization enabled, all caching is disabled
>+ * because the processor can modify ISR under the hood. Instead
>+ * just set SVI.
>+ */
>+ if (kvm_apic_vid_enabled(vcpu->kvm)) {
>+ kvm_x86_ops->hwapic_isr_update(vcpu->kvm, vec);
>+ return;
>+ }
>
>- if (!__apic_test_and_set_vector(vec, apic->regs + APIC_ISR))
> ++apic->isr_count;
>+ }
>+
> BUG_ON(apic->isr_count > MAX_APIC_VECTOR);
> /*
> * ISR (in service register) bit is set when injecting an interrupt.
>@@ -1627,11 +1637,16 @@ int kvm_get_apic_interrupt(struct kvm_vcpu *vcpu)
> int vector = kvm_apic_has_interrupt(vcpu);
> struct kvm_lapic *apic = vcpu->arch.apic;
>
>- /* Note that we never get here with APIC virtualization enabled. */
>-
> if (vector == -1)
> return -1;
>
>+ /*
>+ * We get here even with APIC virtualization enabled, if doing
>+ * nested virtualization and L1 runs with the "acknowledge interrupt
>+ * on exit" mode. Then we cannot inject the interrupt via RVI,
>+ * because the process would deliver it through the IDT.
>+ */
>+
> apic_set_isr(vector, apic);
> apic_update_ppr(apic);
> apic_clear_irr(vector, apic);
>
>
>I think the right way to do it must be something like this; you cannot
>do it just in nested_vmx_vmexit. Testing is welcome since I don't have
>easy access to APICv-capable hardware (it would take a few days).
I will test it tomorrow, it's late today for me. ;-)
Regards,
Wanpeng Li
>
>Paolo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
