DMARC-Filter: OpenDMARC Filter v1.3.2 mx1.redhat.com A9A74C0828D0
Subject: Re: [RFC] KVM: optimize the kvm_vcpu_on_spin
To: David Hildenbrand <david@redhat.com>,
        "Longpeng (Mike)" <longpeng2@huawei.com>
Cc: rkrcmar@redhat.com, agraf@suse.com, borntraeger@de.ibm.com,
        cohuck@redhat.com, christoffer.dall@linaro.org, marc.zyngier@arm.com,
        james.hogan@imgtec.com, kvm@vger.kernel.org,
        linux-kernel@vger.kernel.org, weidong.huang@huawei.com,
        arei.gonglei@huawei.com, wangxinxin.wang@huawei.com,
        longpeng.mike@gmail.com
References: <1501309377-195256-1-git-send-email-longpeng2@huawei.com>
 <bcbdafa9-349f-345e-dde1-28bfa1cf5826@redhat.com>
 <597F1DAE.4020809@huawei.com>
 <0677ed6e-280a-d2f3-d873-1daf99b39551@redhat.com>
From: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <3572e95a-a5eb-748b-25c8-b7e128cbbe1b@redhat.com>
Date: Mon, 31 Jul 2017 15:20:11 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101
 Thunderbird/52.2.1
MIME-Version: 1.0
In-Reply-To: <0677ed6e-280a-d2f3-d873-1daf99b39551@redhat.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 909
Lines: 20

On 31/07/2017 14:27, David Hildenbrand wrote:
>> I'm not sure whether the operation of get the vcpu's priority-level is
>> expensive on all architectures, so I record it in kvm_sched_out() for
>> minimal the extra cycles cost in kvm_vcpu_on_spin().
>>
> as you only care for x86 right now either way, you can directly optimize
> here for the good (here: x86) case (keeping changes and therefore
> possible bugs minimal).

I agree with Cornelia that this is inconsistent, so you shouldn't update
me->in_kernmode in kvm_vcpu_on_spin.  However, get_cpl requires
vcpu_load on Intel x86, so Mike's patch is necessary (vmx_get_cpl ->
vmx_read_guest_seg_ar -> vmcs_read32).

Alternatively, we can add a new callback kvm_x86_ops->sched_out to x86
KVM, and call vmx_get_cpl from the Intel implementation (vmx_sched_out).
 This will cache the result until the next sched_in, so that
kvm_vcpu_on_spin can use it.

Paolo