From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1758819Ab0BYCXn (ORCPT ); Wed, 24 Feb 2010 21:23:43 -0500 Received: from hera.kernel.org ([140.211.167.34]:59147 "EHLO hera.kernel.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1758760Ab0BYCXl (ORCPT ); Wed, 24 Feb 2010 21:23:41 -0500 Message-ID: <4B85E164.1080104@kernel.org> Date: Thu, 25 Feb 2010 11:33:08 +0900 From: Tejun Heo User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.1.5) Gecko/20091130 SUSE/3.0.0-1.1.1 Thunderbird/3.0 MIME-Version: 1.0 To: Yinghai Lu CC: Peter Zijlstra , Ingo Molnar , Christoph Lameter , "H. Peter Anvin" , Stephen Rothwell , Thomas Gleixner , Andrew Morton , Linus Torvalds , Jesse Barnes , linux-kernel@vger.kernel.org, linux-pci@vger.kernel.org, Pekka Enberg Subject: Re: [PATCH 16/35] x86: make 64 bit use early_res instead of bootmem before slab References: <1265793639-15071-1-git-send-email-yinghai@kernel.org> <1265793639-15071-17-git-send-email-yinghai@kernel.org> <20100215010833.15f9e09a.sfr@canb.auug.org.au> <4B7B4378.1070206@kernel.org> <1267052387.12790.10.camel@laptop> <4B85B65E.503@kernel.org> <4B85B724.9050304@kernel.org> <4B85DB79.4010708@kernel.org> <4B85DCAF.5030304@kernel.org> In-Reply-To: <4B85DCAF.5030304@kernel.org> X-Enigmail-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-4.2.3 (hera.kernel.org [127.0.0.1]); Thu, 25 Feb 2010 02:23:10 +0000 (UTC) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hello, On 02/25/2010 11:13 AM, Yinghai Lu wrote: > ok, will add free_early_partial to handle it. Great. > i got impression, when we are moving slab ealier. you said that > will check to make percpu setup to use slab allocator instead. I don't recall exactly but it probably was 4k remapping everything into vmalloc area and didn't need large contiguous area. The problem with the former approach was that it added a lot of extra TLB pressure. The current allocator avoids that by overlaying the static percpu areas onto the linear address space which requires punching holes in an contiguous area. Thanks. -- tejun