From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S266691AbUGVD1u (ORCPT ); Wed, 21 Jul 2004 23:27:50 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S266686AbUGVD1u (ORCPT ); Wed, 21 Jul 2004 23:27:50 -0400 Received: from sccrmhc11.comcast.net ([204.127.202.55]:3040 "EHLO sccrmhc11.comcast.net") by vger.kernel.org with ESMTP id S266691AbUGVD1t (ORCPT ); Wed, 21 Jul 2004 23:27:49 -0400 Subject: Re: reserve legacy io regions on powermac From: Albert Cahalan To: linux-kernel mailing list Cc: olh@suse.de, benh@kernel.crashing.org, geert@linux-m68k.org Content-Type: text/plain Organization: Message-Id: <1090457945.1231.711.camel@cube> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.2.4 Date: 21 Jul 2004 20:59:06 -0400 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org I think a great many drivers could be cleaned up by making IO fail if the proper IO address cookies haven't been obtained. This might be done with byteswapping, XOR, addition, putting a checksum in the top 2 bytes of a 64-bit cookie, or simply tracking where an ioremap has been done. Then the read and write operations can check this. In general, make IO fail if a driver doesn't play by the rules. Perhaps some of the low memory-mapped stuff on x86 could be moved.