Two years after I first saw this, Will Robinson appears to be back on our
164SX-based web server:

stratos 137 # grep "WILL ROBINSON" messages* 
messages:Apr 22 15:03:15 stratos /netbsd: DANGER WILL ROBINSON: FEN SET IN
cpu_fork!
messages:Apr 23 00:39:29 stratos /netbsd: DANGER WILL ROBINSON: FEN SET IN
cpu_fork!
messages:Apr 23 00:39:30 stratos /netbsd: DANGER WILL ROBINSON: FEN SET IN
cpu_fork!
messages.0:Apr 22 00:43:40 stratos /netbsd: DANGER WILL ROBINSON: FEN SET IN
cpu_fork!
messages.1:Apr 18 18:43:55 stratos /netbsd: DANGER WILL ROBINSON: FEN SET IN
cpu_fork!
messages.2:Apr 16 15:44:20 stratos /netbsd: DANGER WILL ROBINSON: FEN SET IN
cpu_fork!

Typically, this is  followed by a panic:

stratos 142 # grep "panic:" messages*
messages:Apr 23 00:56:35 stratos /netbsd: panic: uvm_fault_unwire: unwiring
non-wired memory
messages.0:Apr 22 00:56:24 stratos /netbsd: panic: uvm_fault_unwire:
unwiring non-wired memory
messages.1:Apr 18 18:36:46 stratos /netbsd: panic: uvm_fault_unwire:
unwiring non-wired memory

It looks like this is happening at times of high memory usages -- for
example, the Analog web reports run just before 1 AM, and will use up a pile
of memory even while the web server is still active.  (Analog memory
footprint is quite LARGE.)

Would the Will Robinson message indicate that there is problem with some of
the memory DIMMs -- or just that there aren't enough of them?  The system
has 384 MB, but each httpd process (with mod_perl etc) eats up 10 MB+.

Tom Haapanen
tomh@motorsport.com