*BSD News Article 35031


Return to BSD News archive

Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!bunyip.cc.uq.oz.au!munnari.oz.au!news.Hawaii.Edu!ames!agate!howland.reston.ans.net!europa.eng.gtefsd.com!MathWorks.Com!news2.near.net!news.delphi.com!BIX.com!arog
From: arog@BIX.com (arog on BIX)
Newsgroups: comp.os.386bsd.misc
Subject: Re: Crash! Bang... Oops
Date: 19 Aug 94 05:35:14 GMT
Organization: Delphi Internet Services Corporation
Lines: 121
Message-ID: <arog.777274514@BIX.com>
References: <3311uk$t08@csugrad.cs.vt.edu>
NNTP-Posting-Host: bix.com

jaitken@csugrad.cs.vt.edu (Jeff Aitken) writes:

>I'm not sure what happened here, but I'll summarize as best I can:

>I moved recently (back to school).  Everything worked fine before I
>packed up.

>Once I got to school, I added 8MB of RAM (to bring my total to 16MB)

>Then I booted FreeBSD, and everything seemed to work ok

>Then I booted DOS, and things started to go haywire.  First, himem
>started complaining about memory being bad.  Well, OK, lets run a memory
>exerciser on it.  I assume FreeBSD has one, so I attempt to boot it.  I
>get something like this 

>Automatic reboot in progress

>/dev/rwd1a 
>/dev/rwd1e

>pid 30: rm: exit on signal 11

>seg fault - core dumped

>pid 4: sh: exit on signal 11

>Then I reboot, and there's a rm.core and sh.core file in /  (I booted
>into single user, BTW, because /dev/rwd1a was modified).  I try to rm
>rm.core.  System crashes, tries to sync disks (and fails), then dumps a
>memory image.  It then tries to reboot, but now cannot find /dev/wd1a.
>I cannot fsck it manually because fsck is not on the bootable floppy!
>(Or if it is I cannot find it)  When I try to mount either /dev/wd1a or
>/dev/wd1e, I get "/dev/wd1x on /mnt: Bogus super block"

>Hmmn, looks like the crash was a good one.  I yanked the extra 8MB of
>memory, and get the same results.  Does anyone have any idea what I can
>do to remedy this, other than re-install everything?  I really don't
>want to reconfigure everything.  I forget, does the installation muck
>with /usr?  Or does it just create a usable / ??

>Thanks!

>Jeff
>-- 
>Jeff Aitken
>jaitken@vt.edu

Jeff, I'll assume that you are at FreeBSD-1.1R or -1.1.5.1, the
level that I've been working with. 

fsck lives in sbin and should be run on the character (or 'raw')
device. Boot to single user with the -s at the boot prompt. 
[ah, like I'm in tutorial mode, so I'll be reciting the stuff
that everyone knows, just as I did earlier tonight on BIX
in one of these on Xenix filesystems.]

At this point, the root filesystem should be read-only. This
may well explain some of the problems in rming the *core files.

Run /sbin/fsck /dev/rwd1a and at /dev/rw1e . If the problem
is one of something mucked up in the structure of the fs, then
this has a reasonable chance of fixing it. 

Also, take a look at how things are entered into the machine's
cmos ram in terms of the hard.disks. I'll suggest strongly
that you go back to documentation to confirm that this is
set to the correct values. 

As I noted some while back and recieved conformation of
in mail, ST-506 and IDE drives that are not configured
in the cmos of the machine can be mounted by FreeBSD. 
I regard this a very desirable feature and hope that,
if its caused by a 'bug', that it will be retained in
the future releases. 

If you cannot get things to work with wd1 listed in cmos, 
set that drive to 'none' and try again. Its something of
a 'long shot' but it also serves to reduce the chance that
its something strange that direction. If it persists,
when this is tried, open the system and 'rock' the socketed
chips in their sockets and seat them all down... memory,
keyboard, anything that is in a socket. After all, a problem
at this level would affect both 'dos' and BSD.

In that the problem first manifested itself as you booted to
'dos', I'd suggest booting from a dos floppy and running
chkdsk at the dos drives and looking at the partition info
with the dos fdisk. You might also get a copy of pfdisk
(which runs from 'dos' and acts as an equivalent of BSD's
disklabel) and see if it reports the info as it should be. 
I'd probably elect to write the labels again from it
as yet.another.place that an invisable glitch might be
shot.

Last, make a second copy of your BSD filesystem disk,
as you did for the original install. From single user
or multi-user, this can be mounted and should mount
as read/write. Assuming that the problem fs doesnot 
contain /usr/bin and the other files needed to run
the editors [and I'll make an editorial comment here
that 'standard' BSD or not, there really needs to be
a line.editor in /bin for dealing with problems at this
level so that the need to use cat for file creation
is reduced]... ah, editors are there, then you can
edit the .profile on the mounted filesystem disk,
rm the install and *.gz in its root, and move the things
that are needed for a panic.disk in.

If it is the /usr filesystem thats munged, then you'll
need to use cat - > .profile to create a new one... and
I'd suggest mving the original one to a new name and
making lots of notes as to path and such for the new one.

.... I hope that helps and helps enough others to justify
	the use of bandwidth.

........................................................
Alan  Ogden, Moderator of 'nos' on BIX (and lots of other things too)
arog@BIX.com