Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH v3 1/2] KVM: LAPIC: Tune lapic_timer_advance_ns smoothly
To:     Wanpeng Li <wanpeng.li@hotmail.com>
Cc:     Wanpeng Li <kernellwp@gmail.com>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "kvm@vger.kernel.org" <kvm@vger.kernel.org>,
        =?UTF-8?B?UmFkaW0gS3LEjW3DocWZ?= <rkrcmar@redhat.com>
References: <1566980342-22045-1-git-send-email-wanpengli@tencent.com>
 <a1c6c974-a6f2-aa71-aa2e-4c987447f419@redhat.com>
 <TY2PR02MB4160421A8C88D96C8BCB971180B00@TY2PR02MB4160.apcprd02.prod.outlook.com>
From:   Paolo Bonzini <pbonzini@redhat.com>
Openpgp: preference=signencrypt
Message-ID: <8054e73d-1e09-0f98-4beb-3caa501f2ac7@redhat.com>
Date:   Thu, 12 Sep 2019 14:37:46 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101
 Thunderbird/60.8.0
MIME-Version: 1.0
In-Reply-To: <TY2PR02MB4160421A8C88D96C8BCB971180B00@TY2PR02MB4160.apcprd02.prod.outlook.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

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