From mboxrd@z Thu Jan  1 00:00:00 1970
From: Peter Xu <peterx@redhat.com>
Subject: Re: [kvm-unit-tests PATCH 01/14] x86: vm: allow multiple init for vm
 setup
Date: Thu, 20 Oct 2016 19:01:07 +0800
Message-ID: <20161020110107.GL15168@pxdev.xzpeter.org>
References: <1476448852-30062-1-git-send-email-peterx@redhat.com>
 <1476448852-30062-2-git-send-email-peterx@redhat.com>
 <20161020081707.6ewyxtkaxuoslz6u@kamzik.brq.redhat.com>
 <20161020082408.GI15168@pxdev.xzpeter.org>
 <20161020084137.csjiv3dq6uuikupy@kamzik.brq.redhat.com>
 <20161020085522.GK15168@pxdev.xzpeter.org>
 <20161020093906.fn2tzg2flr4n3wpa@kamzik.brq.redhat.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=utf-8
Cc: kvm@vger.kernel.org, jan.kiszka@web.de, agordeev@redhat.com,
        rkrcmar@redhat.com, pbonzini@redhat.com
To: Andrew Jones <drjones@redhat.com>
Return-path: <kvm-owner@vger.kernel.org>
Received: from mx1.redhat.com ([209.132.183.28]:56586 "EHLO mx1.redhat.com"
        rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
        id S1750906AbcJTLBM (ORCPT <rfc822;kvm@vger.kernel.org>);
        Thu, 20 Oct 2016 07:01:12 -0400
Content-Disposition: inline
In-Reply-To: <20161020093906.fn2tzg2flr4n3wpa@kamzik.brq.redhat.com>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

On Thu, Oct 20, 2016 at 11:39:06AM +0200, Andrew Jones wrote:
> On Thu, Oct 20, 2016 at 04:55:22PM +0800, Peter Xu wrote:
> > On Thu, Oct 20, 2016 at 10:41:37AM +0200, Andrew Jones wrote:
> > 
> > [...]
> > 
> > > Let's do these patches separate from the series and maybe change setup_idt
> > > too? It'd be interesting to see if the assert in setup_idt would fire,
> > > i.e. if any users are relying on it being tolerant to multiple calls, and
> > > then find out why.
> > 
> > The problem should be: smp_init() is calling setup_idt(). So if we
> > change the init stuff in setup_idt() into an assertion, any test
> > program that calls both smp_init() and setup_idt() would possibly fail
> > the assertion. Actually I see most test cases are using:
> > 
> > 	setup_vm();
> > 	smp_init();
> > 	setup_idt();
> > 
> > to setup a basic environment, so I guess all of these use cases would
> > fail. In that sense, I'd slightly prefer keep setup_idt() as it is.
> 
> Well I guess the fix is just to remove the unnecessary setup_idt calls
> from all unit tests that do smp_init(). Anyway, I won't fight too hard
> for this cleanup, but I certainly prefer not to propagate the sloppiness
> further into the other setups.

It won't be hard to do that. Let me cook one for this.

Thanks,

-- peterx