From mboxrd@z Thu Jan  1 00:00:00 1970
From: Takuya Yoshikawa <yoshikawa.takuya@oss.ntt.co.jp>
Subject: Re: [PATCH RFC v2 6/6] KVM: introduce a new API for getting dirty
 bitmaps
Date: Tue, 20 Apr 2010 20:44:50 +0900
Message-ID: <4BCD93B2.2040003@oss.ntt.co.jp>
References: <20100420195349.dab60b1d.yoshikawa.takuya@oss.ntt.co.jp> <20100420200353.2d2a6dec.yoshikawa.takuya@oss.ntt.co.jp> <480E8E1E-32BD-429E-96C8-5AA69A8BDDF3@suse.de> <4BCD90FE.9060300@oss.ntt.co.jp> <04434D88-5274-43B7-B9B5-A2FE952A6CCE@suse.de>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Cc: avi@redhat.com, mtosatti@redhat.com, kvm@vger.kernel.org,
	kvm-ia64@vger.kernel.org, kvm-ppc@vger.kernel.org,
	fernando@oss.ntt.co.jp
To: Alexander Graf <agraf@suse.de>
Return-path: <kvm-owner@vger.kernel.org>
Received: from serv2.oss.ntt.co.jp ([222.151.198.100]:48507 "EHLO
	serv2.oss.ntt.co.jp" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1753723Ab0DTLld (ORCPT <rfc822;kvm@vger.kernel.org>);
	Tue, 20 Apr 2010 07:41:33 -0400
In-Reply-To: <04434D88-5274-43B7-B9B5-A2FE952A6CCE@suse.de>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

(2010/04/20 20:33), Alexander Graf wrote:
>>>>
>>>> -#define KVM_API_VERSION 12
>>>> +#define KVM_API_VERSION 13
>>>
>>> Is there a way to keep both interfaces around for some time at least? I'd prefer the API version not to change if not _really_ necessary.
>>>
>>> To enable the new dirty mapping you could for example use KVM_CAP_ENABLE_CAP :-).
>>
>> Thanks, I did not know what is the appropriate way for this kind of change.
>>
>> I just read the comments in the kvm.h and thought I had to update the number whenever
>> I added some entries to kvm.h .
>
> The rule of thumb is:
>
> KVM_API_VERSION change ->  complete breakage. No way to run older userspace on newer kernels.
> new CAP ->  optional feature
>
> If I read correctly you're changing quite a few existing interfaces, so old userspace wouldn't work anymore. Hence I'm proposing using activating the new way of dirty logging optionally.
>
> Alex
>

The fact is, I am trying to keep the existing interfaces completely(KVM_API_VERSION part is my mistake).

For x86, I have checked that current qemu works without any change.
I mean we can use get_dirty_log() as is.

And as you propose, we can determine which api(swith_* or get_*) we use optionally.