Hey folks,

I'm hoping somebody out there can provide some information on this.

I'm doing some driver development and I'm trying to get a handle on how the X video drivers work. The linux framebuffer drivers are pretty straightforward. Vesafb, for example. It initializes and probes and whatnot the framebuffer device to get started, then uses the generic drawing functions cfb_fillrect, cfb_copyarea, and cfb_imageblit to manipulate the screen.

According to documentation the X fbdev driver (xf86-video-fbdev) is a video driver that uses the linux framebuffer device. What exactly does that mean? Does it mean that the contents of the X framebuffer is the same as whats in /dev/fb0? If I don't use the X fbdev driver is the X framebuffer stored somewhere else physically?

Also, what functions does the X driver use to manipulate the screen? From some tinkering, I've determined that it does NOT use the generic functions above. The xf86-video-fbdev source doesn't contain any screen drawing functions.

Any help would be appreciated.

Thanks,
Nathan

Hi -

There's a really good explanation, by Geert Uytterhoeven, in the kernel source, e.g.:

/usr/src/linux/Documentation/fb/framebuffer.txt

Here are some other links:

http://en.wikipedia.org/wiki/Linux_framebuffer
http://tldp.org/HOWTO/Framebuffer-HOWTO.html
... and ...
http://www.linux-fbdev.org/

Basically, /dev/vesafb is what you're in during system boot ... and text mode ... and any of the "<Alt-Fx>" text mode virtual consoles.

On the other hand, when you're in GUI mode, you're using whatever X driver best matches your video H/W. For example, the PC I'm using now is running the "nvidia" X driver.

'Hope that helps .. PSM

paulsm4,

Thanks for the reply. I've read through the kernel documentation and the fbdev and fbdevhw man files. I guess I'm looking for a guide on the internals of the X video drivers and video system. When X wants to put some pixels on the screen, what is the lowest level function called in X before the request is sent to the hardware?

Nathan

Hi -

In answer to "difference between Linux fb and X11 video driver...", please refer back to my previous post:As far as the architecture of an X11 video driver, here's a good link:

http://www.xfree86.org/current/DESIGN.html

On most systems you have a X-server in use so use those functions or QT if you want to make bit manipulation. Then the X-Server has not to be killed. QT also has a openGL implementation if the X-Server driver supports it. The driver only says X11 how to handle the features (does settings) of you graphics card. You can also write directly to your card but you won't be able to do flicker free painting when the X-server is running. A basically CPU command from the X-server is f.e.
and the descriptor is set to your graphics memory.

Hey, now thats some good stuff right there. My ultimate goal here is to be able to detect when X is writing to the screen in some way more efficent than a constant scan of the framebuffer.

Much grass.

Nathan

Hello,everyone!:-)