On 21/06/2016 13:35, Marcelo Tosatti wrote:
> On Tue, Jun 21, 2016 at 09:50:54AM +0200, Paolo Bonzini wrote:
>>
>>
>> On 20/06/2016 15:05, Marcelo Tosatti wrote:
>>>
>>> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
>>> index ea306ad..89be6e9 100644
>>> --- a/arch/x86/kvm/x86.c
>>> +++ b/arch/x86/kvm/x86.c
>>> @@ -2991,6 +2991,7 @@ static int kvm_vcpu_ioctl_set_lapic(struct kvm_vcpu *vcpu,
>>> {
>>> kvm_apic_post_state_restore(vcpu, s);
>>> update_cr8_intercept(vcpu);
>>> + vcpu->arch.apic->lapic_timer.expired_tscdeadline = 0;
>>>
>>
>> I think this is not correct. You have programmed the host timer to an
>> early value when kvm_apic_post_state_restore called start_apic_timer.
>>
>> I think that:
>>
>> 1) post_state_restore should cancel the timer and clear
>> lapic_timer.pending before writing the registers, not afterwards.
>> lapic_timer.expired_tscdeadline can be cleared at the same time.
>>
>> 2) kvm_write_tsc should do the same and restart the timer afterwards.
>>
>> Thanks,
>
> Actually, the max should deal with the problem Alan is having with
> vmload as well (one delay of tenths of microseconds, or 0, per vmload
> should be irrelevant).
Yes, it does. Though I think it's a bug that we don't clear .pending on
KVM_SET_LAPIC, and I'm happy if you do a v2 of this patch to fix it. :)
Paolo
--
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
