From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753616AbXDKSeH (ORCPT ); Wed, 11 Apr 2007 14:34:07 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1753628AbXDKSeH (ORCPT ); Wed, 11 Apr 2007 14:34:07 -0400 Received: from mail.suse.de ([195.135.220.2]:36134 "EHLO mx1.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753616AbXDKSeE (ORCPT ); Wed, 11 Apr 2007 14:34:04 -0400 Date: Wed, 11 Apr 2007 20:33:38 +0200 From: Olaf Dabrunz To: John Stoffel Cc: Gerd Hoffmann , linux-kernel@vger.kernel.org, Andrew Morton , Paul Mundt , Ralf Baechle , Andi Kleen , Alan Cox , Richard Henderson , Ivan Kokshaysky , Paul Mackerras , Benjamin Herrenschmidt , Jeremy Fitzhardinge Subject: Re: [patch v2] Fixes and cleanups for earlyprintk aka boot console. Message-ID: <20070411183338.GB29059@suse.de> Mail-Followup-To: John Stoffel , Gerd Hoffmann , linux-kernel@vger.kernel.org, Andrew Morton , Paul Mundt , Ralf Baechle , Andi Kleen , Alan Cox , Richard Henderson , Ivan Kokshaysky , Paul Mackerras , Benjamin Herrenschmidt , Jeremy Fitzhardinge References: <20070315154638.944768000@suse.de> <17913.45970.178029.519745@smtp.charter.net> <20070329163714.GA29059@suse.de> <17938.30190.281832.393043@smtp.charter.net> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: <17938.30190.281832.393043@smtp.charter.net> User-Agent: Mutt/1.5.9i Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On 03-Apr-07, John Stoffel wrote: > > Sorry, I've been away for the past 10+ days and not reading email at > all. NP. Happens to me too (as you can see). > Olaf> Try changing the order of the console= parameters: > > Olaf> kernel ... console=ttyS1,96008N1 console=tty0 > > This isn't good for me, since it then drops some output I want to save > via the serial console to the regular console. It would be nice if > all output went to BOTH consoles, though input should only be accepted > on the first console listed. That is actually the underlying problem: the Linux kernel does not provide the means to duplicate data streams in the kernel to several devices. This has to be done in user-space. The main reason for this is that error handling, locking and such things would need to be different if streams had a fan-in or a fan-out. It tends to make things more complicated. But probably it is an interesting project to explore and define the sematics of devices with fan-in and fan-out. I can image that user-space programs would be able to add fan-in or fan-out file-descriptors to an opened file-descriptor by using an ioctl(), but then they would need to be prepared to deal with different error handling and behaviour of such a "file". Some devices (as the console device) could then have a pre-defined fan-in or fan-out set by default, if opened with a special flag. A program using this flag would then say: I am prepared to deal with fan-in / fan-out semantics. -- Olaf Dabrunz (od/odabrunz), SUSE Linux Products GmbH, Nürnberg