|
From: James S. <jsi...@tr...> - 2001-11-21 20:12:34
|
> >In fact imageblit is from a newer implementation, and meant for 2.5.x. > > Geert, > Does this imply that in the newer version there is no high level > drawing functions that are touching my mmio and memory pointers. It > is all contained inside the driver? I certainly hope so. TIme to explain the new api a little bit. The new api is designed to seperate the console code from the framebuffer code. This allows you two things, a simpler api to work with and second you can run fbdev by itself on embedded devices. To do this the first thing to go was the fbcon-cfb* stuff. Instead the driver writer provides 3 basic accel functions which if you use the fbcon layer uses these functions to preform console functions. I abstracted it to be based around a graphics cards abilties to encourge people to implement and plus it does hide more the hardware features. It will avoid the problems you are having. > To add to the point these pointers should not even be visible to > the outside-of-driver world. It someone can see a virtual pointer, > at some point they will try to use it. In some drivers this would > cause bad things to happen. Another nice featue planned for 2.5.X is that the console system will be "shutdown" when you explictly open /dev/fb. This will prevent any conflicts between the console system and using graphics hardware in userland. As for userland programming hardware. Well that is DRI domain and it is a touchy subject I like to avoid. |