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, 9 Nov 2001 20:27:11 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Fri, 9 Nov 2001 20:27:01 -0500 Received: from neon-gw-l3.transmeta.com ([63.209.4.196]:11532 "EHLO neon-gw.transmeta.com") by vger.kernel.org with ESMTP id ; Fri, 9 Nov 2001 20:26:44 -0500 To: linux-kernel@vger.kernel.org From: torvalds@transmeta.com (Linus Torvalds) Subject: Re: paging Oops in 2.4.1{4,5-pre1} Date: Sat, 10 Nov 2001 01:23:09 +0000 (UTC) Organization: Transmeta Corporation Message-ID: <9shvht$7i5$1@penguin.transmeta.com> In-Reply-To: <200111100038.fAA0cbU13195@danapple.com> X-Trace: palladium.transmeta.com 1005355596 29032 127.0.0.1 (10 Nov 2001 01:26:36 GMT) X-Complaints-To: news@transmeta.com NNTP-Posting-Date: 10 Nov 2001 01:26:36 GMT Cache-Post-Path: palladium.transmeta.com!unknown@penguin.transmeta.com X-Cache: nntpcache 2.4.0b5 (see http://www.nntpcache.org/) Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org In article <200111100038.fAA0cbU13195@danapple.com>, Daniel I. Applebaum wrote: > >I have a problem with kernels 2.4.14 and 2.4.15-pre1. As soon as my >system needs to page, it generates the following Oops. I can >duplicate this error at will by running a few memory-intensive >processes, such as 4-5 simultaneously compiles, StarOffice, and >Netscape. If I duplicate the test, but running with no swap, just >RAM, then I get the expected "Out of Memory: Killed process..." >errors. It is jumping to la-la land, apparently from "do_swap_page()". The interesting part there is that do_swap_page() doesn't even follow any suspicious function pointers or anything.. Can you do a gdb vmlinux and send me the output of "disassemble do_swap_page", along with a copy of the oops (the latter just because I don't keep archives of linux-kernel, so I don't want to have to search for the oops again). Oh, and only the first oops tends to be the really interesting one - after the kernel has oopsed once, kernel data structures are quite possibly corrupt, and subsequent oopses are suspect. That do_swap_page one _was_ the first oops, right? Linus