Subject: PM 7500 Final report
To: None <port-macppc@netbsd.org>
From: Jeff Laughlin <jlaughli@vtc.edu>
List: port-macppc
Date: 02/22/2002 00:43:31
I'm going back to school soon so I'm probably not going to have as much 
time to experiment with the system. So this may be my last post for a 
while. Thanks to everyone for the help I've received.

The "DEFAULT CATCH" errors have really irritated me, but I think I've at 
least figured out a work-around. It seems that the errors ONLY occur on a 
warm-boot. Not always, either, but more often then not. So the simple 
work-around I've found to be quite reliable is to turn the computer off for 
5 minutes, then boot. Seems to work every time that way.

Now as to WHY it works, well I have two theories. One is heat. Obviously 
anytime a computer works cold but not warm, and overheating component is 
almost certainly to blame. But this time I'm not so sure. There are only 
three components in this system which make a deal of heat. The 604e CPU, 
the 4 gig Barracuda, and the power supply. I spent a whole day trying 
various combinations of fans and cooling schemes to chill all of these 
components down as much as possible. The harddrive and the CPU didn't seem 
to care whether they were warm or cold. No noticeable difference. I even 
tried an 8 watt AC fan I pulled out of an ancient IBM dot matrix printer 
(ibm.overengineering.com). That thing sure blows. But no change. Now the 
power supply... I've always been suspicious of this power supply. 
Occasionally the fan will change pitch by a few notes. Odd. I put a 
multimeter on it and watched the voltages as I booted and rebooted and 
messed around. The only slightly odd thing I noticed was that when I 
crashed X, the 12v supply dropped to 11.85. That was odd because the lowest 
I'd seen it drop before was 11.90, and that it dropped when the box froze 
hard seemed strange. But X has been crashy. So I ruled out voltage 
fluctuations as the cause of DEFAULT CATCH. When I opened the case up all 
the way and put the IBM fan and another 3 inch fan on it's grill so that 
they were sucking through it like a bastard, I was able to reboot about two 
dozen times with no DEFAULT CATCH! I thought I had it, I was psyched! I 
mounted the IBM fan to the back grill of the box with some screws and 
applied some duct tape to ensure the fan would suck exclusively through the 
power supply. Well, after booting and running for five minutes I rebooted 
and got default catch. I cry :'(. There is no way with that fan that any 
properly functioning power supply could be overheating. It's still possible 
that it's malfunctioning. Maybe when I get a chance I'll cob an old PC AT 
supply in there somehow. But except for X, the system seemed stable as long 
as I didn't reboot. Grrrrr...

The other possible cause of the DEFAULT CATCH error is the state of the 
system. Some part of the system enters a state where OFW cannot boot. I 
have no clue what part that might be, though. The RAM shouldn't take 5 
minutes to completely loose it's contents. Could it have something to do 
with the NVRAM? But why would that be changed during normal operation? I 
suppose it's possible there could be residual data in the RAM up to five 
minutes later, obviously nothing valid but something that pisses off OFW. 
If this is the case wouldn't it be possible to write a small FORTH program 
to zero-out the affected area of ram? I wish I had more time to work on this.

Overall the system seems stable, except for X. I've been using the blackbox 
window manager, so it's possible that is the cause of the freezes. Though 
it seems odd that something like blackbox or X could cause the kernel to 
hang. I can't even SSH into it when it freezes in X. I haven't tried 
getting in via serial. I suspect it's just that OFW 1.0.5 sucks. Everything 
related to it's video support seems to be flaky. So I probably just won't 
use X. Or maybe I'll try using TWM for a while and see if that has the same 
problems.

I have to do a project on a programming language that I don't already know 
this semester. I think I'll ask to do it on FORTH. Then maybe I can work on 
these OFW issues.

73, happy hacking
-Jeff