Re: [PATCH v3] KVM: LAPIC: Fix lapic timer injection delay

[Paolo Bonzini <pbonzini@redhat.com>]

On 28/06/2017 15:55, Wanpeng Li wrote:
>>>       if ((atomic_read(&apic->lapic_timer.pending) &&
>>>               !apic_lvtt_period(apic)) ||
>>> -             kvm_x86_ops->set_hv_timer(apic->vcpu, tscdeadline)) {
>>> +             (ret = kvm_x86_ops->set_hv_timer(apic->vcpu, tscdeadline))) {
>>>               if (apic->lapic_timer.hv_timer_in_use)
>>>                       cancel_hv_timer(apic);
>>> +             if (ret == 1) {
>>> +                     apic_timer_expired(apic);
>>> +                     return true;
>>> +             }
>> The preemption timer can also be used for modes other than TSC deadline.
>>
>> In periodic mode, your patch would miss a call to
>> advance_periodic_target_expiration, which is only called by
>> kvm_lapic_expired_hv_timer.
> Actually I considered this before, however, I referred to apic timer
> periodic mode which is emulated by hrtimer

Periodic mode can also be emulated by preemption timer... it was added
by some Wanpeng Li in commit 8003c9ae204e ("KVM: LAPIC: add APIC Timer
periodic/oneshot mode VMX preemption timer support", 2016-11-02), do you
know him? ;)

> , there is no hrtimer start
> for the next period in start_sw_period(). If it is also buggy?

start_sw_period always goes through the hrtimer for periodic timer:

        if (apic_lvtt_oneshot(apic) &&
            ktime_after(ktime_get(),
                        apic->lapic_timer.target_expiration)) {
                apic_timer_expired(apic);
                return;
        }

        hrtimer_start(&apic->lapic_timer.timer,
                apic->lapic_timer.target_expiration,
                HRTIMER_MODE_ABS_PINNED);

(the direct call to apic_timer_expired is conditonal to
apic_lvtt_oneshot).  This way, apic_timer_fn takes care of advancing the
hrtimer deadline and returning HRTIMER_RESTART.

Paolo