dolecek@ics.muni.cz said:
> Even through I have ddb.onpanic=0, the kernel will drop into DDB
> prompt when page fault or something like that happens.

Perhaps you want to fix this?

> Since I am on different screen, I
> can't reboot the box cleanly or see what went wrong and have to
> hit the red button.

You can switch screens even if you are within DDB.
(as long as you stay away from X server controlled screens)

> doing this right
> would need to make Debugger() MI, rename the current Debugger() to,
> say, cpu_Debugger() and calling it after the screen is sanified.

The cpu_ prefix might be a good idea for namespace sanity.
The wscons call in Debugger() is questionable for some reasons:
-Interactions with the X server. The switch call will try to
 switch away from an X screen. This is a good thing in
 principle, but it causes a lot of kernel status change
 (device closes, vm) before the ddb prompt is reached.
 I'd prefer to get a ddb prompt immediately (even if it
 is invisible); for a screen cleanup we could add an md ddb
 command. This would also deal better with mi/md separation.
-unnecessary complexity - it is unnecessary because you can switch
 manually
-another overhead stack frame in the traceback

best regards
Matthias