From mboxrd@z Thu Jan  1 00:00:00 1970
From: "Longpeng (Mike)" <longpeng2@huawei.com>
Subject: Re: [help] host kernel panic in kvm's wakeup_handler()
Date: Fri, 26 May 2017 18:53:06 +0800
Message-ID: <59280912.2000303@huawei.com>
References: <592504AE.6040306@huawei.com> <20170523223419.59f7d465@t450s.home> <5925144B.2030207@huawei.com> <11b3f9a3-7c89-639e-84e0-219c9e376a66@redhat.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 7bit
Cc: Alex Williamson <alex.williamson@redhat.com>,
        kvm <kvm@vger.kernel.org>,
        =?UTF-8?B?UmFkaW0gS3LEjW3DocWZ?= <rkrcmar@redhat.com>,
        "Huangweidong (C)" <weidong.huang@huawei.com>,
        Gonglei <arei.gonglei@huawei.com>,
        "wangxin (U)" <wangxinxin.wang@huawei.com>
To: Paolo Bonzini <pbonzini@redhat.com>
Return-path: <kvm-owner@vger.kernel.org>
Received: from szxga02-in.huawei.com ([45.249.212.188]:6831 "EHLO
        szxga02-in.huawei.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
        with ESMTP id S965647AbdEZKyG (ORCPT <rfc822;kvm@vger.kernel.org>);
        Fri, 26 May 2017 06:54:06 -0400
In-Reply-To: <11b3f9a3-7c89-639e-84e0-219c9e376a66@redhat.com>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>



On 2017/5/26 18:40, Paolo Bonzini wrote:

> 
> 
> On 24/05/2017 07:04, Longpeng (Mike) wrote:
>>>> it crashed at *1ec1* and %rax get a wrong value(0xdead000000100100) at *1e92*,
>>>> it seems the *blocked_vcpu_on_cpu* list is corrupted, but kvm only access this
>>>> list in pre_block/post_block/wakeup_handler, and these three functions seems good.
>>>>
>>>> kvm version is 4.4-stable.
>>>>
>>>> Do you have any ideas? Any suggestion would be greatly appreciated, thanks!
>>>>
>>> Is this only seen with posted interrupt support enabled?  Booting with
>>> intremap=nopost on the kernel commandline would disable it.  Thanks,
>>
>> We tested with PI support enabled, but we not sure if it only occurs with PI
>> enabled yet.
> 
> This code should not run at all with PI disabled, since the handler is
> only reachable through an IRTE.
> 
> As you said, the list manipulation in those function is fairly simple.
> If you have a reproducer, you can try running it with CONFIG_LIST_DEBUG
> and see what you get.
> 


OK.

We have already started test for a long time, but didn't panic yet.

Thanks :)

> Thanks,
> 
> Paolo
> 
> .
> 


-- 
Regards,
Longpeng(Mike)