[PATCH] offb.c fix

[PATCH] offb.c fix

[Benjamin Herrenschmidt]

[-- Attachment #1: Type: text/plain, Size: 855 bytes --]

Enclosed is a small patch to offb.c that fixes a problem with some ATI
displays in 16 
(and potentially 32) bpp. The kluge used for accessing the colormap
should be used only for 8 bits, it broke 16 bits on my PowerBook G3 Series.

Geert, could you explain me briefly where I should put the static
colormap that bootx gives me ? I mean, in 8 bits, especially for non-ATI
display (without the hack), the colormap is fixed and currently set to
the logo colors. I plan to set more useful colors but currently offb
makes no use of the colormap provided by bootx. Implementing this would
allow userland stuffs to know which colors were set by bootx.

I'm not sure of what I should do with this cmap, where I should store it
in the fb device structures...

-- 
           E-Mail: <mailto:bh40@calva.net>
BenH.      Web   : <http://calvaweb.calvacom.fr/bh40/>



[-- Attachment #2: offb.patch.gz --]
[-- Type: application/octet-stream, Size: 749 bytes --]

[PATCH] new offb.c fix

[Benjamin Herrenschmidt]

[-- Attachment #1: Type: text/plain, Size: 1304 bytes --]

Enclosed is the latest version of the patch. I updated more pieces of
code here or there so that offb.c can now theorically work in directcolor
if the cmap_adr field is present.

However, I still disable this feature and uses the ATI kludge only in 8
bits since there is a difference in the way the color registers must be
programmed between different models of ATI chips, and this broke for
example 16 bits on PowerBook G3 Series (LT-G)

It still missing code for getting the default palette set by bootx but
this is less urgent. This fix was more important since I would like
everyone to be able to boot in "no video driver" mode, at least for the
installer. That also means that the installer could be packaged with a
kernel containing _only_ offb to save space.


Note: Geert, while testing in direct colors, I had a funny problem in 32
bits: everything was fine except the console cursor which had a different
color on each virtual console. Since I remember reading about similar
bugs in early versions of other fb devices, I guess it's a common
mistake. Do you have a clue about what's going on ? It's not very
important since we are always in truecolor with this version, but I'm
curious ;-)

-- 
           E-Mail: <mailto:bh40@calva.net>
BenH.      Web   : <http://calvaweb.calvacom.fr/bh40/>



[-- Attachment #2: offb.patch.gz --]
[-- Type: application/octet-stream, Size: 1333 bytes --]