From mboxrd@z Thu Jan  1 00:00:00 1970
From: "Zhang Haoyu" <zhanghy@sangfor.com>
Subject: RE: [PATCH] KVM: x86: reset RVI upon system reset
Date: Thu, 11 Dec 2014 16:15:35 +0800
Message-ID: <201412111615355608149@sangfor.com>
References: <1415156023-1349-1-git-send-email-wei.w.wang@intel.com>,
 <5459C020.20103@intel.com>,
 <286AC319A985734F985F78AFA26841F77F3FCD@shsmsx102.ccr.corp.intel.com>,
 <5459DA93.6060104@intel.com>,
 <286AC319A985734F985F78AFA26841F77F404A@shsmsx102.ccr.corp.intel.com>,
 <5459E794.6020500@intel.com>,
 <5459F5AB.1050302@redhat.com>,
 <A9667DDFB95DB7438FA9D7D576C3D87E0ABC9BA1@SHSMSX104.ccr.corp.intel.com>
Mime-Version: 1.0
Content-Type: text/plain;
	charset="gb2312"
Content-Transfer-Encoding: 7bit
To: "Zhang, Yang Z" <yang.z.zhang@intel.com>,
	"Paolo Bonzini" <pbonzini@redhat.com>,
	"Chen, Tiejun" <tiejun.chen@intel.com>,
	"Wang, Wei W" <wei.w.wang@intel.com>,
	"kvm@vger.kernel.org" <kvm@vger.kernel.org>
Return-path: <kvm-owner@vger.kernel.org>
Received: from smtp.sangfor.com ([202.131.75.66]:56137 "EHLO mail.sangfor.cn"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S964919AbaLKIPx (ORCPT <rfc822;kvm@vger.kernel.org>);
	Thu, 11 Dec 2014 03:15:53 -0500
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

Then?

>> On 05/11/2014 10:02, Chen, Tiejun wrote:
>>>> I think both are ok.
>>>> If we zero max_irr in vmx_set_rvi(), we still need this check:
>>>> if ((is_guest_mode(vcpu) && nested_exit_on_intr(vcpu)) || max_irr
>>>> ==
>>>> -1)
>>> 
>>> No, I don't think we need to add this.
>> 
>> You don't, because the code will look like:
>> 
>>         if (is_guest_mode(vcpu) && nested_exit_on_intr(vcpu))
>>                 return;
>>         if (!is_guest_mode(vcpu)) {
>>                 vmx_set_rvi(max_irr);
>>                 return;
>>         }
>>         
>>         if (max_irr == -1)
>>                 return;
>> and thus vmx_set_rvi() is never reached if is_guest_mode(vcpu) &&
>> !nested_exit_on_intr(vcpu).
>
>I don't think the above code is perfect. Since hwapic_irr_update() is a hot point, it's better to move the first check after the second check. In this case, Wei's patch looks more reasonable.
>
>> 
>> I applied the lapic.c part of Wei's patch, and the vmx.c part of Tiejun's patch.
>> 
>> Paolo
>
>
>Best regards,