From mboxrd@z Thu Jan  1 00:00:00 1970
From: Peng Fan <van.freenix@gmail.com>
Subject: Re: Question about KVM on Big.Little SoC
Date: Thu, 8 Sep 2016 09:46:01 +0800
Message-ID: <20160908014559.GC9097@linux-7smt.suse>
References: <20160906080005.GB5493@linux-7smt.suse>
 <57CE7A6A.1010508@arm.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Cc: andre.przywara@arm.com, christoffer.dall@linaro.org,
        kvm@vger.kernel.org
To: Marc Zyngier <marc.zyngier@arm.com>
Return-path: <kvm-owner@vger.kernel.org>
Received: from mail-pa0-f67.google.com ([209.85.220.67]:33720 "EHLO
        mail-pa0-f67.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
        with ESMTP id S1750991AbcIHBqO (ORCPT <rfc822;kvm@vger.kernel.org>);
        Wed, 7 Sep 2016 21:46:14 -0400
Received: by mail-pa0-f67.google.com with SMTP id h5so1664719pao.0
        for <kvm@vger.kernel.org>; Wed, 07 Sep 2016 18:46:14 -0700 (PDT)
Content-Disposition: inline
In-Reply-To: <57CE7A6A.1010508@arm.com>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

Hi Marc,
On Tue, Sep 06, 2016 at 09:12:26AM +0100, Marc Zyngier wrote:
>Peng,
>
>On 06/09/16 09:00, Peng Fan wrote:
>> Hi Andre,
>> 
>> I came across your patch which try to fix kvm on big.little.
>> https://lists.cs.columbia.edu/pipermail/kvmarm/2013-April/003989.html
>> 
>> I am trying KVM on a big.little SoC with 4 A53 and 2 A72.
>> I wonder whether kvm supports running on big.little SoC now?
>> 
>> I am checking arch/{arm,arm64}/kvm code, but I think you may give a quick answer.
>> 
>> And If kvm supports running on big.Little SoC, is it ok for me to pass "-cpu host"
>> to qemu? Can guest vcpu be scheduled freely on big and little SoCs?
>> 
>> Would you please kindly give me some clues about where is the piece code to handle
>> kvm on big.little?
>
>The current situation is that:
>
>- Yes, KVM works on big-little systems (it always had)
>- The only useful option you can pass to qemu is "-cpu host"
>- There is no guarantee that QEMU will even start, unless you constraint
>its affinity to one single type of CPUs (use taskset).
>- If you use taskset, there will be no scheduling between types of cores
>- If you do not, scheduling will happen, but that's a fairly unsafe
>thing to do (the guest doesn't know it is being migrated between CPUs
>that may have different behaviours...).

Thanks for your explaination.
I think I should use 'taskset [cpumask] qemu-system-aarch64 ....' to start
guest os -:)

>
>Hope this helps,

Yeah. This helps. I am trying using taskset.

Thanks,
Peng.

>
>	M.
>-- 
>Jazz is not dead. It just smells funny...