I have three Sun 3/60's at my house that I've been playing with NetBSD on
for some time. All of them are maxed-out, RAM-wise (go get those 30-pin 1x9
SIMMs now - the PC weenies are dumping them to go to 72-pin SIMMs for a
song ($20-$22 per MB) as they upgrade to bloated W95 systems).

Two of the Suns (chronos.clock.org, ticking.clock.org) have disks, and the
third one I netboot (quartz.clock.org) and serve off of one of the others
(chronos.clock.org). The "everything core dumps" problem shows up only
occasionally on chronos, but more often than not on quartz. It is clearly
some shared library problem - if you try to telnet into a system in such a
state, you get a very odd "file does not exist" error with some trash
characters before the TCP connection goes away - I'm sure this is an
exec(2) error from inetd(8).

This problem has been with us for a long time - it happened to me prior to
the 1.1 release and prior to the DMA upgrades to the "si" driver (for
which, I might add, I am *quite* grateful - the performance win is
considerable - I run with si_options set to 3, since I had problems with
{de,re}select on the older set of the disks I have).

And people wonder why, despite the advantages of shared libraries (lower
RAM consumption, ability to update library routines system-wide, etc.) that
I am (and always have been) deeply suspicious of them as a critical
single-point for system-wide failure...

I will capture one of these errors and report to the list.

Erik Fair