From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752220Ab1CKH6j (ORCPT ); Fri, 11 Mar 2011 02:58:39 -0500 Received: from moutng.kundenserver.de ([212.227.17.10]:65051 "EHLO moutng.kundenserver.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751181Ab1CKH6i (ORCPT ); Fri, 11 Mar 2011 02:58:38 -0500 From: Arnd Bergmann To: Nobuhiro Iwamatsu Subject: Re: [PATCH 1/2] tty: serial: Fix build on architecture that does not have ioport Date: Fri, 11 Mar 2011 08:58:29 +0100 User-Agent: KMail/1.13.5 (Linux/2.6.38-rc6+; KDE/4.5.1; x86_64; ; ) Cc: gregkh@suse.de, linux-kernel@vger.kernel.org References: <1299822897-7047-1-git-send-email-nobuhiro.iwamatsu.yj@renesas.com> In-Reply-To: <1299822897-7047-1-git-send-email-nobuhiro.iwamatsu.yj@renesas.com> MIME-Version: 1.0 Content-Type: Text/Plain; charset="iso-8859-15" Content-Transfer-Encoding: 7bit Message-Id: <201103110858.29140.arnd@arndb.de> X-Provags-ID: V02:K0:7jLblq/C9qZvH1SLSr1U3gxTtQuwXs0Qa8sChmny2zR 0FjXd0tkkwbNfWfiNLcIb1SMJnBGG1H0h7Enm7rMtvwV6yTtBC M9D97nCMSp/QOvN1J7BhKjWiYi3A21bP/cfioRsD+knd5zi9m5 r9luWr7vGFdk02KEkVQ7qUeJbvetm0fy6dMnTxuVv22hXHO5WK OHDCfsGJKXNA540tW1yEA== Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Friday 11 March 2011 06:54:56 Nobuhiro Iwamatsu wrote: > Some CPU's do not have ioport. Therefore, these do not have inX/outX. > Because they define CONFIG_NO_IOPORT, These CPU's use this and do not > use inX/outX. > > Signed-off-by: Nobuhiro Iwamatsu I like this patch, I had played with something similar before but never got it to the state where I was happy with submitting it. I'd really like to get us to the point where platforms without PCI support do don't need to define inb/outb and can still use all other drivers. There are two small improvements that I would like to suggest to this patch: * use #ifdef CONFIG_SERIAL_8250_HUB6 to configure out hub6 support. It's already disabled on most systems, so we can save a bit binary code without making the driver uglier. * Introduce a new CONFIG_SERIAL_8250_IOPORT symbol and make the normal io_serial_* code depend on that. It can be a silent Kconfig symbol that automatically gets selected in the appropriate cases. I find this clearer than the double negation in "not no ioport". The code that is used in both HUB6 and regular cases can be under this symbol. Arnd