Re: [PATCH v3 1/2] KVM: LAPIC: Tune lapic_timer_advance_ns smoothly

[Paolo Bonzini]

On 12/09/19 02:34, Wanpeng Li wrote:
>>> -        timer_advance_ns -= min((u32)ns,
>>> -            timer_advance_ns / LAPIC_TIMER_ADVANCE_ADJUST_STEP);
>>> +        timer_advance_ns -= ns;

Looking more closely, this assignment...

>>>    } else {
>>>    /* too late */
>>>        ns = advance_expire_delta * 1000000ULL;
>>>        do_div(ns, vcpu->arch.virtual_tsc_khz);
>>> -        timer_advance_ns += min((u32)ns,
>>> -            timer_advance_ns / LAPIC_TIMER_ADVANCE_ADJUST_STEP);
>>> +        timer_advance_ns += ns;

... and this one are dead code now.  However...

>>>    }
>>>
>>> +    timer_advance_ns = (apic->lapic_timer.timer_advance_ns *
>>> +        (LAPIC_TIMER_ADVANCE_ADJUST_STEP - 1) + advance_expire_delta) /
>>> +        LAPIC_TIMER_ADVANCE_ADJUST_STEP;

... you should instead remove this new assignment and just make the
assignments above just

	timer_advance -= ns / LAPIC_TIMER_ADVANCE_ADJUST_STEP;

and

	timer_advance -= ns / LAPIC_TIMER_ADVANCE_ADJUST_STEP;

In fact this whole last assignment is buggy, since advance_expire_delta
is in TSC units rather than nanoseconds.

>>>    if (abs(advance_expire_delta) < LAPIC_TIMER_ADVANCE_ADJUST_DONE)
>>>        apic->lapic_timer.timer_advance_adjust_done = true;
>>>    if (unlikely(timer_advance_ns > 5000)) {
>> This looks great.  But instead of patch 2, why not remove
>> timer_advance_adjust_done altogether?
> It can fluctuate w/o stop.

Possibly because of the wrong calculation of timer_advance_ns?

Paolo