Subject: CATS crash
To: None <port-arm32@netbsd.org>
From: Stephen B Streater <netbsd@surprise.demon.co.uk>
List: port-arm32
Date: 11/14/1998 10:41:15
I've been tracking down my CATS crash, which is easily
provocable. It seems to happen after I've had lots of
page faults:

...
_malloc
_amap_alloc
uvm_map
sys_obreak
_syscall

The error message is:
"panic: malloc: out of space in kmem_map"
caused by, it seems, uvm_km_kmemalloc failing.

The comment in the source says there may be enough memory,
but in the wrong sized buckets (kernel has 6MB to play with),
and writing the code to move it around wasn't a high priority
at the time. The kernel memory seems to be wired, so if I can
catch it before the machine seizes up I should be able to see
how much wired memory there is from "top".

I going to see if I can tell which of these events occurred:
a) 6MB not enough memory for kernel;
b) not enough memory in this bucket but other memory elsewhere;
c) memory leak in kernel meaning running out of memory is inevitable.

I'm only on chapter 2 of my Unix System Administrators Bible, so
it may take a bit of time to get all these answers.

PS I'm on netbsd_13f.

-- 
Stephen B Streater