red-handed

[Greg Lindahl <lindahl cs virginia edu>]

I'd been having some odd lockups of my PC164LX machines... they'd be
fine, and then they'd be pingable but you can't log in, and the
console is unresponsive.

I got on one right before it crashed and netstat -to showed the timers
weren't changing. Then later I got on aother right before it crashed,
and realized that "date" wasn't changing. Then I made a script which
ran date on each machine every 10 seconds. Machines that crash
definitely have stopped clocks.

Given that I saw this under 2.0.35 and 2.2.5-16 (rh 6.0's kernel), I
don't think it's a kernel bug. The only hardware oddity that I know of
is that when you reboot these guys, the cmos date is always wrong. I'm
going to replace all the batteries... could that possibly cause this
symptom?

I haven't had this much fun with clocks since I realized that flipping
the front-panel switch on a PDP-11 that turns off the clock breaks
Kermit, and little else.

-- g