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 S932126AbbFQLPF (ORCPT <rfc822;w@1wt.eu>);
	Wed, 17 Jun 2015 07:15:05 -0400
Received: from mail-wg0-f46.google.com ([74.125.82.46]:34119 "EHLO
	mail-wg0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1752887AbbFQLO4 (ORCPT
	<rfc822;linux-kernel@vger.kernel.org>);
	Wed, 17 Jun 2015 07:14:56 -0400
Date: Wed, 17 Jun 2015 13:14:51 +0200
From: Ingo Molnar <mingo@kernel.org>
To: Andy Lutomirski <luto@kernel.org>
Cc: x86@kernel.org, linux-kernel@vger.kernel.org,
        =?iso-8859-1?Q?Fr=E9d=E9ric?= Weisbecker <fweisbec@gmail.com>,
        Rik van Riel <riel@redhat.com>, Oleg Nesterov <oleg@redhat.com>,
        Denys Vlasenko <vda.linux@googlemail.com>,
        Borislav Petkov <bp@alien8.de>, Kees Cook <keescook@chromium.org>,
        Brian Gerst <brgerst@gmail.com>
Subject: Re: [RFC/INCOMPLETE 00/13] x86: Rewrite exit-to-userspace code
Message-ID: <20150617111451.GA9335@gmail.com>
References: <cover.1434485184.git.luto@kernel.org>
 <20150617103226.GA30325@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20150617103226.GA30325@gmail.com>
User-Agent: Mutt/1.5.23 (2014-03-12)
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org


* Ingo Molnar <mingo@kernel.org> wrote:

> Basically there would be a single C function we'd call, which returns a 
> condition (or fixes up its return address on the stack directly) to determine 
> between the SYSRET and IRET return paths.

This we could do by returning the syscall result in RAX, and the SYSRET/IRET 
choice in RDX - that's the natural return parameter for 128-bit return values in 
the 64-bit C function ABI, and it's clobbered so it's available 'for free'.

We could do something similar for the IRQ entry/return code as well: there's no 
reason why IRQ flag tracking has to be maintained in assembly. We could move all 
but the IRQ stack switching code to C.

We can safely flip around the IRQ stack setting with the enter_from_user_mode 
call, so that IRQ stack switching becomes part of the register saving and kernel 
mode preparatory preamble.

This would allow further optimizations in the IRQ code as well: for example we 
could inline enter_from_user_mode() and prepare_exit_to_usermode().

Thanks,

	Ingo