5 day lockup on Densitron

[Clifton Royston]

On Fri, May 06, 2005 at 08:58:58AM -0400, Lowell Gilbert wrote:
> "Richard J. Valenta" <RJV at WEBLINKMO.COM> writes:
> > At work we had a machine which had in it a 200mhz pentium - the machine
> > was an all-in-one flat panel made by a company named Densitron for
> > industrial use.  On one side of it was a the flat/touch panel, and the
> > idea was that it was a complete machine that could be installed into a
> > wall - how cool is that?
...
> > So I restarted it, and logged in as root, ran top - and let it go for a
> > few days.  Alsways a lockup, I've now done this a few times and it seems
> > to happen at 5 days, 1 hour and roughly 40 minutes.  Without fail, after
> > 5 days I'll come downstairs and find it locked up at 5 days, 1 hour,
> > etc...
> > 
> > I'm at a loss as to why, I'm no expert with the crontab, so maybe I'm
> > missing something.  But I'm looking for ideas, thanks.
> 
> This is pretty weird, all right.  The time period isn't a round number
> in any units I can work out, in base 10 or 2.  It's probably some kind
> of coincidence, so try to think of common factors. 

  If this was designed as an industrial PC, intended to run specialized
software, it may have some specialized hardware.  What you describe
could conceivably be the result of a special counter or RTC chip
running as a "watchdog timer" with a count-down from boot time, and
generating some kind of special interrupt when that countdown reaches
0.  Watchdog devices are sometimes set up to require the application
software to "stroke" the timer periodically (reset it in software) with
the intent to force a reset of the system (usually a reboot) after
such-and-such a period of time if not stroked.  Sometimes they're even
tied to the system hardware reset line; the idea is to prevent a
software lockup from taking the system down indefinitely, though you're
seeing the reverse effect.

  This is all speculation, and it might sound like the idea is
completely wacky, but I have worked on hardware with watchdogs, though
 it's been a long time and it wasn't PC hardware.  If this is the case
(a big if) you'd need to know exactly what the hardware is, and what
it's set up to do on countdown, before you can begin figuring out how
to fix it.

  Here are a couple off-the-cuff ideas.  One possible way to check for
a watchdog - boot it off of something else, e.g. a DOS disk such as a
Win 98 startup disk, then leave it sitting at the command prompt for 5
days etc. and see if it stops responding, reboots or does something
else weird.  It won't prove anything if it doesn't - memory usage and
interrupt setup will be very different in that scenario - but it might
point you in some interesting direction.  Also, scour the BIOS menu
settings (I assume it has a BIOS?) for anything that sounds like it
might relate.  I'm not sure what else to tell you, other than to check
with the manufacturer if possible.

  -- Clifton

-- 
          Clifton Royston  --  cliftonr at tikitechnologies.com 
         Tiki Technologies Lead Programmer/Software Architect
"I'm gonna tell my son to grow up pretty as the grass is green
And whip-smart as the English Channel's wide..."
                                            -- 'Whip-Smart', Liz Phair