Subject: Kernel Memory exhaustion
To: None <port-pmax@NetBSD.ORG>
From: Bill Harris <billh@airmail.net>
List: port-pmax
Date: 08/19/1997 16:09:53
I'm running the Jul 20 snapshot and I seem to be running into the
same old virtual kernel memory exhaustion problem. Basically, if
you load the machine up, with even some simple processes like 'cp'
the kernel panics and the system reboots.
Simply executing the following command will do the machine in
nicely and predictably.
cp /netbsd /usr/tmp/junk
cat /netbsd >>/usr/tmp/junk
cd /usr/tmp
for i in 1 2 3 ... 25
do
cp junk junk.$i&
cp junk junk.1.$i&
cp junk junk.2.$i&
done
This 75 simple copies, which should load the machine
a bit, causes it to panic and reboot with our kernel memory exhaustion
problem.
I'm using the mostly stock DS5000_240 config, pertinent info pasted
below. I'm certain this is relative to the NKMEMCLUSTERS settings,
but what should they be set to?
This is so easy to duplicate and seems like a resonable load that
a machine might have, that there ought to be a solution.
Bill
-- # Support for specific models of DECstation
#options DS3100 # PMAX (kn01) DECstation 2100
#options DS5000_25 # MAXINE (kn02ca/xine) support
#options DS5000_100 # 3MIN (kn02ba/kmin) support
#options DS5000_200 # 3MAX (kn02) support
options DS5000_240 # 3MAXPLUS (kn03) support
# Standard system options
#options DIAGNOSTIC # extra kernel debugging checks
#options DEBUG # extra kernel debugging support
options "COMPAT_43" # compatibility with 4.3BSD
options KTRACE # system call tracing support
options "NKMEMCLUSTERS=1024" # 4K pages in kernel malloc pool
#options KGDB # support for kernel gdb
#options "KGDBRATE=19200" # kernel gdb port rate
#options "KGDBDEV=15*256+0" # device for kernel gdb
options NTP # network time protocol
#options UCONSOLE # users can redirect console
----------------------------------------------------------------------
Bill Harris KC5YZQ
Internet America (SYSTEMS)