From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-kernel-owner@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S933922AbcBQKDY (ORCPT <rfc822;w@1wt.eu>);
	Wed, 17 Feb 2016 05:03:24 -0500
Received: from smtp47.i.mail.ru ([94.100.177.107]:51882 "EHLO smtp47.i.mail.ru"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S933366AbcBQKDW (ORCPT <rfc822;linux-kernel@vger.kernel.org>);
	Wed, 17 Feb 2016 05:03:22 -0500
Subject: Re: [PATCH v5 1/4] x86/signal/64: Add a comment about sigcontext->fs
 and gs
To: Ingo Molnar <mingo@kernel.org>, Andy Lutomirski <luto@kernel.org>
References: <cover.1455664054.git.luto@kernel.org>
 <baa78f3c84106fa5acbc319377b1850602f5deec.1455664054.git.luto@kernel.org>
 <20160217072122.GA19718@gmail.com>
Cc: x86@kernel.org, Denys Vlasenko <dvlasenk@redhat.com>,
        Cyrill Gorcunov <gorcunov@gmail.com>,
        Pavel Emelyanov <xemul@parallels.com>, Brian Gerst <brgerst@gmail.com>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        Oleg Nesterov <oleg@redhat.com>, Borislav Petkov <bp@alien8.de>
From: Stas Sergeev <stsp@list.ru>
X-Enigmail-Draft-Status: N1110
Message-ID: <56C44560.6010708@list.ru>
Date: Wed, 17 Feb 2016 13:03:12 +0300
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101
 Thunderbird/38.5.0
MIME-Version: 1.0
In-Reply-To: <20160217072122.GA19718@gmail.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
X-Mras: Ok
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

17.02.2016 10:21, Ingo Molnar пишет:
> 
> * Andy Lutomirski <luto@kernel.org> wrote:
> 
>> +	 * If the kernel ever adds explicit fs, gs, fsbase, and gsbase
>> +	 * save/restore, it will most likely need to be opt-in and use
>> +	 * different context slots.
> 
> Btw., that's not necessarily true: it could also be made opt-out, and a 
> modify_ldt() or any other cleanly identifiable legacy usage/signature that is 
> associated with DOSEMU might trigger the opt-out automatically as well.
But there are the new versions of dosemu that still use
"modify_ldt() or any other cleanly identifiable legacy usage/signature"
and yet wants a new functionality.
Please don't go for such an unreliable heuristic.

> I.e. behaviorally it would still keep the ABI and modern default behavior could 
> still be whatever we want to make it.
You can allow glibc to deal with the opt-ins.
It has symbol versioning system AFAIK exactly for that.