r127 w/ KMS on PPC?

[Jerome Glisse]

On Thu, Apr 22, 2010 at 01:46:46PM +0100, Alex Buell wrote:
> On Wed, 2010-04-21 at 20:38 -0400, Alex Deucher wrote:
> > On Wed, Apr 21, 2010 at 6:26 PM, Alex Buell <alex.buell at munted.org.uk> wrote:
> > > On Thu, 2010-04-22 at 07:55 +1000, Dave Airlie wrote:
> > >> > I'd like to step up for that, it'd let me learn a lot. I have
> > >> already
> > >> > extracted the r128 and its DRI from the 2.6.32 kernel tree and built
> > >> > them as modules and they both work just fine, with the binary
> > >> firmware.
> > >> > What's involved in making KMS work on PPC? I'm sure I need the KMS
> > >> > helpers and the TTM module as well right?
> > >>
> > >> For r128, it could actually re-use a lot of the radeon KMS code, in
> > >> fact nearly all of it.
> > >>
> > >> Alex Deucher started doing a UMS merge of radeon/r128 at one point
> > >> http://cgit.freedesktop.org/~agd5f/xf86-video-ati/log/?h=r128-support
> > >>
> > >> the main things would be adding the differences in the bios parsing,
> > >> and modesetting paths to KMS first.
> > >
> > > I'll start digging into the Radeon, see what I can learn about KMS
> > > integration. For now I'll build the r128 out of the tree, simplifies
> > > things a bit. Is that OK?
> > >
> > > Any further reading would be tremendously helpful.
> > >
> > 
> > It should be pretty easy to get kms going with r128 by integrating it
> > into the radeon drm and ddx.  The hard part will be porting the r128
> > 3D driver to kms.  Also, kms support for r128-based macs will be a bit
> > trickier than x86 because the boards don't have an x86 video bios.
> > You may have to look at the xf86-video-r128 ddx and atyr128 kernel fb
> > driver for properly setting up the cards without an x86 bios.  Let me
> > know if you have any questions.
> 
> Also, I was wondering why these chipset specific _drm.h files are kept
> in include/drm/ i.e. r128_drm.h, radeon_drm.h and mga_drm.h instead of
> being within the drivers/gpu/drm/*/ directories? 

Because we install them along other kernel header file (ie
they are included in kernel-header package of distribution)
Maybe we should move non public header out of include/drm but
i don't think it's really needed.

Cheers,
Jerome