From mboxrd@z Thu Jan 1 00:00:00 1970 From: Arnd Bergmann Subject: Re: [PATCH 04/27] asm-generic: add legacy I/O header files Date: Fri, 1 May 2009 15:56:38 +0200 Message-ID: <200905011556.39437.arnd@arndb.de> References: <200905011529.04774.arnd@arndb.de> <20090501144420.5a3e9c6e@lxorguk.ukuu.org.uk> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Return-path: Received: from moutng.kundenserver.de ([212.227.126.171]:54273 "EHLO moutng.kundenserver.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1760364AbZEAN4s (ORCPT ); Fri, 1 May 2009 09:56:48 -0400 In-Reply-To: <20090501144420.5a3e9c6e@lxorguk.ukuu.org.uk> Content-Disposition: inline Sender: linux-arch-owner@vger.kernel.org List-ID: To: Alan Cox Cc: Russell King , Michal Simek , linux-arch@vger.kernel.org, Remis Lima Baima , linux-kernel@vger.kernel.org On Friday 01 May 2009, Alan Cox wrote: > They shouldn't define it - we want those invalid cases to error. Stick > CLOCK_TICK_RATE in the depths of the x86 arch includes and then we can > fix the other cases properly. Either the device has an independant clock > of the same frequency (some watchdogs do this), it relies upon a > motherboard timer in which case the platform can define CLOCK_TICK_RATE > (or a better name for it), or it doesn't have any meaning in which case > we *want* it not to compile. I just noticed that linux/jiffies.h does not compile without a definition of CLOCK_TICK_RATE, so that's probably why every architecture defines it. If I find a way to remove that dependency, I can simply leave out the CLOCK_TICK_RATE from the generic header. I have not yet understood the relation of jiffies to CLOCK_TICK_RATE, but my suspicion is that it's not good for architectures that do not have a clock tick. Arnd <><