On Thu, Mar 27, 2003 at 12:21:58AM +0100, Christoph Egger wrote: > > > Bah. We don't relocate resources unless we need to. So that ball's as > > > much in your court as it is in ours. > > > > Any X reassignment of PCI resources is unsafe in Linux 2.4, and going to > > get more so. The DGA comemnt applies here the other way around - you don't > > currently have a way to get the kernel to do the work for you. We need to > > address that. X doesn't know enough hw issues or locking > > That's the point where KGI comes into the place. Most of you, I guess, > remember > on the heating discussion on the linux kernel ml in 1997/1998. > > In the meantime, it has been redesigned and rewritten from scratch. So, all > what you (still) believe to know about KGI is very likely out of date. > > KGI is splitted in two parts. The kernel driver knows all stuff to access I still don't understand why you simply did not use the fbdev framework for this ? This is i believe one of the reasons GGI failed, because it tried to redesign everything, without taking into account any of the stuff that pre-existed at that time. > hardware and to map the graphic resources (i.e. accelerator) safely to the > userspace - BTW: ioctls are only used for setting up video modes and other > initialisation stuff and nothing else. How is that different from fbdevs ? Friendly, Sven Luther