From owner-freebsd-hackers Thu Feb 2 15:08:35 1995 Return-Path: hackers-owner Received: (from root@localhost) by freefall.cdrom.com (8.6.9/8.6.6) id PAA20054 for hackers-outgoing; Thu, 2 Feb 1995 15:08:35 -0800 Received: from dkuug.dk (dkuug.dk [193.88.44.89]) by freefall.cdrom.com (8.6.9/8.6.6) with SMTP id PAA20043 for ; Thu, 2 Feb 1995 15:08:30 -0800 Received: from login.dknet.dk by dkuug.dk with SMTP id AA20746 (5.65c8/IDA-1.4.4j for ); Fri, 3 Feb 1995 00:08:00 +0100 Received: by login.dknet.dk (4.1/SMI-4.1DKnet00) id AA12315; Fri, 3 Feb 95 00:03:18 +0100 From: sos@login.dknet.dk (S|ren Schmidt) Message-Id: <9502022303.AA12315@login.dknet.dk> Subject: Re: I'm having a very perplexing problem To: kaleb@x.org (Kaleb Keithley) Date: Fri, 3 Feb 95 0:03:18 MET Cc: hackers@freefall.cdrom.com In-Reply-To: <9502022241.AA06393@fedora.x.org>; from "Kaleb Keithley" at Feb 2, 95 5:41 pm X-Charset: ASCII X-Char-Esc: 29 Sender: hackers-owner@FreeBSD.org Precedence: bulk It does sound very much like a memory problem, besides 100ns RAM is way to slow for a modern machine, 70na is more like it. It could also suggest you are having bad cache RAM chips, or at least some that are not fast enough. > In the process of juggling some memory between systems I managed to > kill the motherboard in my FreeBSD (1.1.5.1) system. After replacing > the motherboard began to get some programs, that never had a problem > before, now segv apparently at random, e.g. /bin/sh, /sbin/fsck, > and gcc's cc1. > > What's strange is that fsck would check /dev/rswd0a fine, but segv > on /dev/rswd0e (my /usr partition). More recently I went to rebuild > my kernel, and gcc runs fine, but cc1 segvs repeatedly when it tries > to compile rtsock.c. > > I was able to boot single user, remount my root partition, and replace > /bin/sh and /sbin/fsck with the versions from the kernel or cpio > install diskette. The replacements don't dump core. (And I know that > they're not the same as the "real" /bin/sh and /sbin/fsck.) Replacing > gcc's cc1 is more problematic only because I had installed 2.6.3 at one > point. I can of course revert to the 2.5.8 files that were originally > installed. > > One thought I had was that I had 70ns RAM in one bank and 100ns RAM > in the second bank. (I used to have 16 meg, but this new board only has > room for 8 SIMMs. :-() Thinking that this might somehow be the cause, > I put 100ns RAM in all the slots. This was after discovering /bin/sh > and /sbin/fsck were "bad" but before discovering the problem with cc1, > so I'm inclined to believe that it's not the memory speed difference > that is what's causing the problem. > > Both boards are apparently from the same Taiwan motherboard company; same > logo on the boxes anyway. The old board was a five year old ISA with an > OPTI chipset and AMI BIOS. The new board is an ISA-VLB with a chipset I > don't remember and Award BIOS. The old board had the ability to set memory > wait states in the CMOS setup -- the new board does not. > > This "feels" like a memory problem, but I don't know why, for instance, > one version of fsck would successfully fsck rwd0a but dump core on rwd0e, > when it had worked fine before changing the motherboard; while another > version of fsck on the new motherboard works fine on both partitions. > > Here's how my disk is partitioned: > > 6 partitions: > # size offset fstype [fsize bsize cpg] > a: 30240 164304 4.2BSD 512 4096 16 # (Cyl. 163 - 192) > b: 68544 194544 swap # (Cyl. 193 - 260) > c: 506016 164304 unused 0 0 # (Cyl. 163 - 664) > d: 671328 0 unused 0 0 # (Cyl. 0 - 665) > e: 387072 263088 4.2BSD 512 4096 16 # (Cyl. 261 - 644) > f: 20160 650160 4.2BSD 512 4096 16 # (Cyl. 645 - 664) > > Any ideas? > > -- > > Kaleb KEITHLEY > -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- Soren Schmidt (sos@FreeBSD.org | sos@login.dknet.dk) FreeBSD Core Team So much code to hack -- so little time ..