Eike Dehling wrote:
> On other notice, i think i might have found (some) light; as mentioned 
> in an email linked in a message i got privately, i edited alpha's 
> interrupt.c, and made machine-checks non-fatal, basically i print them 
> but ignore them for the rest (removed the last line 'panic()' in the 
> 'machine_check()' function). Now i can see the messages on the console, 
> since the machine won't lock up anymore in a state where xfree has taken 
> control of the gfx-card.
> 
> [...]
> 
> I'll try finding out what causes these machine checks; i added the 
> "option 'NoInt10'" to XFree86Config already, but it didn't help.

I further 'enhanced' interrupt.c, to make it send a signal after a 
machine check, and i did a debug build of XFree86. Now gdb revealed 
where the problem was. (it's where the mga driver reads it's settings 
from the cards BIOS).

 From looking at the code, i think it might be trying to read from a 
bogus address, though i'm sure; i don't know xfree code.

My question, is there anyone who knows this code? I'd love to have this 
stuff working, but i don't really posses the knowledge to fix a broken 
xfree86 driver (except maybe with by a quick hack. :) )

Is this the right place for this problem it perhaps better to ask the 
person who wrote the driver about this; since it's not really netbsd 
related?

Greetings,

Eike.