From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1759556AbYEPSxQ (ORCPT ); Fri, 16 May 2008 14:53:16 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1762378AbYEPSul (ORCPT ); Fri, 16 May 2008 14:50:41 -0400 Received: from netops-testserver-3-out.sgi.com ([192.48.171.28]:40366 "EHLO relay.sgi.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1762348AbYEPSuj (ORCPT ); Fri, 16 May 2008 14:50:39 -0400 Date: Fri, 16 May 2008 11:50:38 -0700 (PDT) From: Christoph Lameter X-X-Sender: clameter@schroedinger.engr.sgi.com To: =?ISO-8859-1?Q?St=E9phane_ANCELOT?= cc: Pekka Enberg , linux-kernel@vger.kernel.org Subject: Re: detecting kernel mem leak In-Reply-To: <482D4CC8.2050501@free.fr> Message-ID: References: <1209462094.4816ed4e93977@imp.free.fr> <84144f020804290546h666983aci77fbb08eb507e777@mail.gmail.com> <4821C526.8030901@free.fr> <4829566B.7090804@free.fr> <482D4CC8.2050501@free.fr> MIME-Version: 1.0 Content-Type: MULTIPART/MIXED; BOUNDARY="-1700579579-2072485358-1210963838=:3113" Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org This message is in MIME format. The first part should be readable text, while the remaining parts are likely unreadable without MIME-aware tools. ---1700579579-2072485358-1210963838=:3113 Content-Type: TEXT/PLAIN; charset=ISO-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE On Fri, 16 May 2008, St=E9phane ANCELOT wrote: > In the kernel, there is SLAB or SLUB , my kernel is at this time configur= ed > for SLAB allocator. SLAB does not support the slabinfo tool. It only supports /proc/slabinfo. > it is documented SLUB minimizes cache line usage. > Do you think I have to switch to SLUB ? If you want to use the slabinfo tool then yes. > 2) regarding memory debugging, your reply and some mesages told it was no= rmal > the memory was growing (with ext3 buffer_heads...) and released on demand= =2E > This sounds to me it becomes VERY VERY difficult telling if my system is > STABLE or NOT. Is there a way to bypass it ? This the basic design of memory handling in Linux. Why would the use of=20 memory mean that your system is unstable? > I assume I have to do some kind of small program trying to allocate almos= t the > full remaining memory available at startup to empty caches ? There is a way to drop caches. See what you can do with=20 /proc/sys/vm/drop_caches f.e. echo 1 >/proc/sys/vm/drop_caches echo 2 >/proc/sys/vm/drop_caches Should free most of memory. ---1700579579-2072485358-1210963838=:3113--