[Xorg] DRI merging

[Jon Smirl]

--- Eric Anholt <eta at lclark.edu> wrote:
> As far as I understood, Mesa-solo required the linux framebuffer, which
> I don't have.  I seriously don't want to be the one to start work on the
> mode-setting library.  As far as pbuffers, once we've got the glue
> necessray I'll fix up the SiS driver for it, which should be quick.  But
> all of this will take time to be done, and they aren't areas I've
> already worked in.  Render, now, I should be able to fix this patch up
> in another day or maybe two, and make myself and normal users happy in
> the short term.

I added mode setting support to DRM while you were gone. That caused a big stink
and it is not in there now. There is a meeting scheduled at OLS to discuss the
issues between DRM and fbdev.

The complicated part of mode setting is what do you do when there are multiple
heads involved? If multiple head mode setting stays in fbdev then fbdev needs to
do memory management but DRM does complicated memory management that is not in
fbdev.

Another large point of contention is whether mode setting should be done
completely in the kernel including DDC support. The alternate solution is to
read DDC from user space and compute the mode info there. Then just pass the
final register values into the driver.

I also added a hot plug event to DRM to trigger a user space reset program than
ran in vm86 mode, but that's been pulled too. Read the dri-devel and fbdev mail
archives, there are hundreds of messages on this subject. Other topics that were
discussed included printk with DRM and rework of the kernel VT subsystem.

If I could get reset, mode setting, and cursor support into DRM mesa-solo
wouldn't need framebuffer anymore. 

=====
Jon Smirl
jonsmirl at yahoo.com


	
		
__________________________________
Do you Yahoo!?
Friends.  Fun.  Try the all-new Yahoo! Messenger.
http://messenger.yahoo.com/