On Tue, Jan 20, 2015 at 5:54 PM, Paolo Bonzini <pbonz...@redhat.com> wrote:
>
>
> On 20/01/2015 09:48, Wincy Van wrote:
>> +static int vmx_deliver_nested_posted_interrupt(struct kvm_vcpu *vcpu,
>> + int vector)
>> +{
>> + int r = 0;
>> + struct vmcs12 *vmcs12;
>> +
>> + /*
>> + * Since posted intr delivery is async,
>> + * we must aquire a spin-lock to avoid
>> + * the race of vmcs12.
>> + */
>> + spin_lock(&to_vmx(vcpu)->nested.vmcs12_lock);
>> + vmcs12 = get_vmcs12(vcpu);
>> + if (!is_guest_mode(vcpu) || !vmcs12) {
>> + r = -1;
>> + goto out;
>> + }
>
> is_guest_mode should be checked first outside the lock, to avoid
> affecting the non-nested fast path. You can then recheck it later
> inside the lock.
Agreed, will do.
>
> Another way to avoid the spinlock: in prepare_vmcs02 or a similar place,
> you can save vmcs12->posted_intr_nv in a new field
> vmx->nested.posted_intr_nv; just set it to -1 if
> !nested_cpu_has_posted_intr(vmcs12). In vmclear, again you just set the
> field to -1, and here you can do
>
> if (!is_guest_mode(vcpu) ||
> vector != to_vmx(vcpu)->nested.posted_intr_nv) {
> r = -1;
> goto out;
> }
>
> You don't need to access vmcs12, and while there is a race, it's okay
> because there is no pointer access.
That's a good idea. I will apply it to the next version.
>
>>
>> + if (vcpu->mode == IN_GUEST_MODE)
>> + apic->send_IPI_mask(get_cpu_mask(vcpu->cpu),
>> + POSTED_INTR_VECTOR);
>
> Please add a comment that PIR and ON have been set by the L1 hypervisor.
Will do.
>
> I'll do a full review the other patches as soon as possible.
>
Thank you, I will send v3 after it is done.
Wincy
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
