Re: NON IRQ DEADLOCK in 2.0.31

Rob Hagopian (hagopiar@vuser.vu.union.edu)
Fri, 14 Nov 1997 01:47:20 -0500 (EST)

I've actually had some problems with the software watchdog:
A user has a 150MB email, tries to open it in pine, pine goes hogwild
with memory (swap), the computer slows to a crawl, the watchdog process
doesn't get spawned fast enough, the computer resets.
What do you do when your session is interrupted? You try it again. This
went on for 3-4 reboots (not a fun task) before I tracked it down. Yes, I
know, process limits... still, a slow machine is much better than a dead
machine.
I seem to recall DEADLOCKS halt both processors or some such thing
(while(1)?), if panic is set, why not just jump to reset?
-Rob H.

On Fri, 14 Nov 1997, Andy Poling wrote:

> Yes: a watchdog timer. The kernel appears to support two or three different
> watchdog cards these days.
>
> The watchdog hardware works like a countdown timer. As long as the system
> is healthy, a user process keeps resetting the countdown timer before it can
> reach zero. If the system crashes or freezes, the timer reaches zero, and
> the watchdog card forces a hardware reset by twiddling the reset line (or
> PWROK as I recall) on the motherboard.
>
> They're relatively cheap ISA cards, and, with the hernel support, easy to
> install and use...
>
> There's also a "software watchdog" available if you build it into the
> kernel. It's less likely to successfully reboot the system in the case of
> a true crash... but the price sure is right.
>
> -Andy
>
> Global Auctions
> http://www.globalauctions.com
>