From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Fri, 3 May 2002 17:37:44 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Fri, 3 May 2002 17:37:43 -0400 Received: from e31.co.us.ibm.com ([32.97.110.129]:60294 "EHLO e31.co.us.ibm.com") by vger.kernel.org with ESMTP id ; Fri, 3 May 2002 17:37:43 -0400 Date: Fri, 03 May 2002 15:35:52 -0700 From: "Martin J. Bligh" To: root@chaos.analogic.com, Jeff Dike cc: linux-kernel@vger.kernel.org Subject: Re: Virtual address space exhaustion (was Discontigmem virt_to_page() ) Message-ID: <222870000.1020465352@flay> In-Reply-To: X-Mailer: Mulberry/2.1.2 (Linux/x86) MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org > No. It's not stupid. Unix defines a kind of operating system that > has certain characteristics and/or attributes. Process/kernel shared > address space is one of them. It's a name that has historical > signifigance. Yes it is stupid. This is a small implementation detail, and has no real importance whatsoever. People have done this in the past (Dynix/PTX did it) will do so in the future. Nor does the kernel address space have to be global and shared across all tasks as stated earlier in this thread. What makes it Unix is the interface it presents to the world, and how it behaves, not the little details of how it's implemented inside. M. PS. I've been told Solaris x86 can do 4Gb for each of kernel and user space, though I've no first hand experience with that OS.