Index: [thread] [date] [subject] [author]
  From: Jon M. Taylor <taylorj@ecs.csus.edu>
  To  : ggi-develop@eskimo.com
  Date: Sun, 14 Feb 1999 21:30:43 -0800 (PST)

Re: Possible to rmmod kgicon?

On Sat, 13 Feb 1999, ERDI CHEN wrote:

> Hi All,
> 
> Does any one know if it is technically possible to make kgicon removable
> at runtime?  

	Yes, I do it every day at work actually.

> Can this be done by modifying kgicon to restore the console
> to its original state and release all resources, without any change to the
> kernel?  

	Already done, it just isn't being used on most existing KGI 
drivers.

> Such a feature would make driver programming much easier as I
> don't need to restart the system to test any changes.

	Sure does.  That is why I use it at work.  I have the card I'm
developing for in the AGP slot and an old SVGA PCI card.  I have one
monitor with two switchable inputs, but you can also use two monitors of
course.  The BIOS is set to use the PCI card as the primary display, on
which I load the VGA driver.  Then I can load the new driver, switch one
VC over to its control with 'con2fbmap 2 1', and test it out.  I can then 
remap the console back to the VGA driver, unload the new driver, fix 
something, recompile it, reload it, map the VC back, and so on.

	When you do this, it forces you to fix all the bugs and
unimplemented stuff in the card init and shutdown code.  This promotes
much greater stability in the driver as a whole.  Multiheading especially 
forces you to get the IO range mappings and IO methods right.  In my 
experience, that is where a lot of problems come from.

Jon
 
---
'Cloning and the reprogramming of DNA is the first serious step in 
becoming one with God.'
	- Scientist G. Richard Seed

Index: [thread] [date] [subject] [author]