RE: Watchdog Reset & other issues

According to docs.sun.com a watchdog reset is:

When the system hardware detects an error from which it cannot recover.
(This is known as a Watchdog Reset.)
http://docs.sun.com/db/doc/806-2906-10/6jc3f0a9a?q=watchdog+openboot&a=view

I started having this same issue when I had just changed a configuration on
a SS20. I had just added a 8MB VSIMM to prepare for the Aurora 1.0 install.
I had never seen this error, so I immediately removed the VSIMM thinking it
was bad or not seated properly... you may want to check this BTW. Removing
the VSIMM did not fix the problem. But it was weird... intermittent almost.
It would be working fine one minute then bamm, watchdog reset. Then I
couldn't boot the kernel, every time it would drop out with a watchdog
reset. Leave the machine for awhile, come back and everything would be fine.

After lots of troubleshooting, it turns out the VSIMM was ok but one of the
hard drives had gone bad. Replaced the drive and haven't seen the watchdog
yet.

So check your VSIMM for proper seating in the system and then try it without
the VSIMM or your old graphics card. Double check any other replaceable
components. It could just be a bad VSIMM.

Good luck,

Peter

-----Original Message-----
From: [email protected]
[mailto:[email protected]]On Behalf Of Greg Pierce
Sent: Thursday, February 06, 2003 9:50 PM
To: [email protected]
Subject: RE: [aurora-sparc-devel] Watchdog Reset & other issues


> Hello all,
>
> I have a SS20.  I had been running Debian Sid, at which time I had it up
for
> extended durations with few problems.
>
> I recently got an 8 meg VSIMM (never opened, still in the sun box :) ).
> I figured I'd try Aurora since I've used Red Hat is the past & liked the
new
> desktop.
>
> 1.  I noticed that the performance seems to be a little anemic, but I
chalk
> this up to the new 2.4 kernel on a hypersparc.
>      I've heard comments that I should step back into 2.2 if it's an issue
> (can someone give me an option?).
>
> 2.  The other day I came home to a white screen with an open boot prompt
> saying "Watchdog Reset".
>      Is this a kernel issue also?  It didn't happen on 2.2, but I didn't
> have the new VSIMM then either.
>
> What is the watchdog reset exactly?  Is it likely to continue to happen?
> Any thoughts are appreciated.

I am not familiar with the precise implementation of the watchdog on the
Sparc 20, but in general the concept is that of a timer. The watchdog
timer is reset at power-on and increments every so-many system clocks
(the exact rate is determined by the system frequency divided by a
hardware prescaler). Eventually, the timer will overflow and generate
a hardware reset. It is up to the software executing on the system to
periodically reset the timer before it overflows (known as "giving the
dog a pat" or "patting the dog"). Typically the OS is responsible for
this task at the kernel level. The purpose of the watchdog is to reset
the system if a complete lockup occurs. This is useful for systems used
as routers or some server applications (i.e. telco, DNS servers, etc.).

Watchdog timers are always able to be deactivated. Most likely it is
possible to do so in the Openboot PROM - it could also possibly be a
hardware jumper, or something that can only be done by the OS (very
unlikely). I would suggest searching for documentation on your system
at Suns web site, or doing a Google search. No doubt it is a fairly
simple matter to deactivate the 'dog on your system.


--

Greg

"I would never make a bet against time travel. The person I am
betting with may be from the future and know the answer."

				-- Stephen Hawking

_______________________________________________
aurora-sparc-devel mailing list
[email protected]
http://lists.linuxpower.org/mailman/listinfo/aurora-sparc-devel