*BSD News Article 11056


Return to BSD News archive

Received: by minnie.vk1xwt.ampr.org with NNTP
	id AA1205 ; Tue, 23 Feb 93 14:30:07 EST
Newsgroups: comp.unix.bsd
Path: sserve!manuel.anu.edu.au!munnari.oz.au!spool.mu.edu!howland.reston.ans.net!usc!cs.utexas.edu!swrinde!gatech!destroyer!cs.ubc.ca!unixg.ubc.ca!ucla-mic!pita!scott
From: scott@pita.cns.ucla.edu (Scott Burris)
Subject: Re: Adaptec 1542B: panic:kmem_malloc: kmem_map too small
Message-ID: <scott.729547054@pita>
Nntp-Posting-Host: pita.cns.ucla.edu
References: <1993Feb10.180517.1108@thunder.mcrcim.mcgill.edu> <3029@bigfoot.first.gmd.de>
Date: 12 Feb 93 12:05:09 PST
Lines: 25

gt@prosun.first.gmd.de (Gerd Truschinski) writes:

>I get the same error >panic: kmem_malloc: kmem_map too small< sporadicly under
>heavy disc-I/O, i.e. compiling gcc from two directories or one compile and
>one 'find / -name ...'. After the reboot I am able to work normaly. No
>disc-crash. I have the patch-kit 0.1 without the wd8013-patch.

My understanding is that this is a kernel memory fragmentation problem.
There is a patch in the patchkit which helps BUT DOES NOT SOLVE this problem.
The original SCSI driver seems to be particularly prone to triggering this.
Julian's SCSI driver is much better, although I have crashed the system with
this panic when taring a 700MB directory tree to DAT tape.  In normal
operations I do not see this panic with his driver.

I have experimented a little with the calculations in the VM code at startup
time with limited success.  My understanding is that release 0.2 will address
the kernel memory allocator problems.

BTW, this is all on a 486/33 with 16M of memory.

-- 
----------
Scott Burris
UCLA Campus Network Services
cnetslb@oac.ucla.edu (310) 206-4860 - OR - scott@pita.cns.ucla.edu