According to Ted Lemon:
>
>It would be *truly* winning if the kernel could be told whether the
>console is in a graphics or character-cell mode, where the frame
>buffer is, how big pixels are, and so on, and if this could be
>settable via a privileged ioctl, so that if I'm running X and suddenly
>dump out into ddb, the kernel can Do The Right Thing.
>

Hmmm would be doable on architectures that have a /dev entry for the
frame buffer since you should be able to take the console away from X
in a sane manner.  For the i386 architecture it is harder, I suppose a
quick hack solution would be to jam "default"ish values into the VGA
registers OR use the method that pcvt uses to take the console away
from X nicely (dunno if this relies on X cooperating or not) OR write
a /dev/vga driver that records the register values when they are set.
For the last one I know that Interactive Unix did this - they had a
vga driver and you could pass a struct of register values in via an
ioctl that would update the display card registers, the upside was
that you could fetch back the current register config because it was
held in the kernel.  The real problem is that unless you just go for
bog standard VGA you need to have all sorts of drivers - perhaps we
could have a loadable module for the driver or a hybrid, a skeleton
driver that provides the basic functionality and a loadable part that
can exploit special features on the card but trying to work out what
you want it that would be a bitch and you would have to software
emulate things on the cards that don't have the features
.... hmmm... maybe not a good idea after all.

-- 
Brett Lymn, Computer Systems Administrator, British Aerospace Australia
===============================================================================
  What do you get when you cross a cantaloup with a dog?        Melancholy :-P