Radeon M6 Issues - WAS (Regression with radeon support (2005))

[Adam Done <emlists@gmail.com>]

Benjamin Herrenschmidt wrote:

>On Fri, 2005-06-03 at 08:43 +0200, Jesus Climent wrote:
>  
>
>>Hi
>>
>>I am experiencing a regression with 2.6.12-rc5 under my iBook G3 2.2.
>>
>>The problems are as follows:
>>
>>1. When switching from console to X, the X picture is redrawn color layer by
>>color layer, like an old 3d modeling program, recalculating the surfaces and
>>showing them as they are calculated. It takes around 1 second or 2 to get the
>>screen back.
>>
>>    
>>
...

>
>Hrm... The actual "delay" thing is to work around a chip bug on M6
>chips. ATI didn't give me a better workaround than stuffing a delay
>after each PLL register accesses which can be a bit slow.
>
>
>  
>
...

>Ben.
>  
>

I am having the same issues since last year and with every new kernel I
have been trying to take care of this issue my self but have not been
able to find a solution.  After looking over the lists and seeing others
have experienced the same issue as I, I am looking to see if there was a
solution?


My latest working kernel without the delay, color change and blanking to
a white screen in X before I get a working X display is:

    vmlinux-2.6.11.9 on a TI PB 633



My yaboot.conf reads the following with my current kernel compile:

    ###  TEST Kernels

    image=vmlinux-2.6.15.4.pb2
            label=test1
            root=/dev/hda8
            append="video=ofonly"

    image=vmlinux-2.6.15.4.pb2
            label=test2
            root=/dev/hda8
            append="video=radeonfb"




But still I have the same issues.

video=ofonly gives me nasty nuclear blowout on my screen as it looks
like the lcd is on acid
video=radeonfb gives the same results as when I do not include any
append in the kernel.

I read Ben you have a patch but I have no idea as to where it may be. 
Is that correct and if so.  Is it usable for current kernels?

-Adam