From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S965864AbXDBUvo (ORCPT ); Mon, 2 Apr 2007 16:51:44 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S965862AbXDBUvo (ORCPT ); Mon, 2 Apr 2007 16:51:44 -0400 Received: from netops-testserver-4-out.sgi.com ([192.48.171.29]:52648 "EHLO relay.sgi.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S965858AbXDBUvn (ORCPT ); Mon, 2 Apr 2007 16:51:43 -0400 Date: Mon, 2 Apr 2007 13:51:41 -0700 (PDT) From: Christoph Lameter X-X-Sender: clameter@schroedinger.engr.sgi.com To: Martin Bligh cc: Dave Hansen , Andi Kleen , linux-kernel@vger.kernel.org, linux-arch@vger.kernel.org, linux-mm@kvack.org, KAMEZAWA Hiroyuki , Andy Whitcroft Subject: Re: [PATCH 1/4] x86_64: Switch to SPARSE_VIRTUAL In-Reply-To: <461169CF.6060806@google.com> Message-ID: References: <20070401071024.23757.4113.sendpatchset@schroedinger.engr.sgi.com> <20070401071029.23757.78021.sendpatchset@schroedinger.engr.sgi.com> <200704011246.52238.ak@suse.de> <1175544797.22373.62.camel@localhost.localdomain> <461169CF.6060806@google.com> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Mon, 2 Apr 2007, Martin Bligh wrote: > > Its just the opposite. The vmemmap code is so efficient that we can remove > > lots of other code and gops of these alternate implementations. On x86_64 > > its even superior to FLATMEM since FLATMEM still needs a memory reference > > for the mem_map area. So if we make SPARSE standard for all configurations > > then there is no need anymore for FLATMEM DISCONTIG etc etc. Can we not > > cleanup all this mess? Get rid of all the gazillions of #ifdefs please? This > > would ease code maintenance significantly. I hate having to constantly > > navigate my way through all the alternatives. > > The original plan when this was first merged was pretty much that - > for sparsemem to replace discontigmem once it was well tested. Seems > to have got stalled halfway through ;-( But you made big boboo in SPARSEMEM. Virtual memmap is a textbook case that was not covered. Instead this horrible stuff that involves calling functions in VM primitives. We could have been there years ago... > Not sure we'll get away with replacing flatmem for all arches, but > we could at least get rid of discontigmem, it seems. I think we could start with x86_64 and ia64. Both will work fine with SPARSE VIRTUAL (and SGIs concerns about performance are addressed) and we could remove the other alternatives. That is going to throw out lots of stuff. Then proceed to other arches Could the SPARSEMEM folks take this over this patch? You have more resources and I am all alone on this. I will post another patchset today that also includes an IA64 implementation.