On Saturday 11 September 2004 01:52, David Eger wrote:

On Fri, Sep 10, 2004 at 01:28:57PM +0800, Antonino A. Daplas wrote:

quoted

On Friday 10 September 2004 10:23, Benjamin Herrenschmidt wrote:

quoted

Recent changes upstream are breaking fbdev on pmacs.

I haven't had time to go deep into that (but I suspect Linus sees it
too on his own g5 unless he removed offb from his .config).

From what I see, it seems that offb is kicking in by default, reserves
the mmio regions, and then whatever chip driver loads can't access
them.

offb is supposed to be a "fallback" driver in case no fbdev is taking
over, it should also be "forced" in with video=ofonly kernel command
line. This logic has been broken.

I dearly *hope* this is what I'm seeing with recent kernels, though I
have my doubts....  What I *do* know is that with recent bk snapshots
(post bk-1.2115) I have the following:  radeonfb seems to load, but when I
try to load X, my y-resolution seems to be half of what it ought to be...

X-fbdev, or X + native radeon driver?  Do still get a framebuffer console?
Any other fb apps you have that's also broken?

Not sure why that happens. Note that the fbdev changes in the latest bk are
mostly driver specific changes and the initialization code cleanup which I'm
addressing at the moment.

Tony