From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Tue, 2 Jul 2002 08:16:19 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Tue, 2 Jul 2002 08:16:18 -0400 Received: from mail.stellartec.com ([65.107.16.99]:45067 "EHLO nt_server.stellartec.com") by vger.kernel.org with ESMTP id ; Tue, 2 Jul 2002 08:16:18 -0400 Reply-To: From: sseeger@stellartec.com (Steven Seeger) To: Subject: 2.4.18-mips buffer cache problem Date: Tue, 2 Jul 2002 05:20:44 -0700 Message-ID: <013e01c221c2$e4074950$3501a8c0@wssseeger> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 (Normal) X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook CWS, Build 9.0.2416 (9.0.2911.0) X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2600.0000 Importance: Normal Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Hey guys. I've got 2.4.18 running on an NEC Osprey development board here using the NEC VR4181 processor. The problem I'm having must be arch specific, however if I knew where in the Kenrel exactly to look on the Linux side I would stand a better chance of figuring out what's going on. When I deal with files on either the CF card or /dev/rom, the buffer cache fills up. That's what's supposed to happen. However, Linux never seems to free these buffers. Once I've say, cat * > /dev/null in enough directories the system memory will all be used up and it won't ever free it. If I unmount the cf card, it frees up a lot of what was buffered up on the card so it seems the arch specific memory stuff works, but why won't it free it up otherwise? The files cached from the /dev/rom device (root fs) fills up memory and never frees. I have another working kernel that doesn't have this problem and I can fill up system memory with cache files and then run a little program that takes up a ton of memory and that program will keep running, but it page faults on this new kernel since it can't get the memory it needs from the cache. I don't see why the OOM won't kill it first, eitehr. Can somebody tell me where in the Kernel it says "hey, I'm running low on memory I should free up some of the memory in the file cache" stuff? If I know where that happens, I can try troubleshooting the problem. I'm sure it'll point to some specific mips problem. Thanks! Steve