Re: radeonfb patch for 2.4.22...

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

> 
> I have a small problem with radeonfb in 2.4.22-pre5 (+manually created
> radeonfb.h file). During boot, when the console is switched over to
> the frame buffer device, the screen becomes corrupted. Mostly by white
> squares in a grid pattern and some squares with other colors. Between
> the squares, normal characters can be seen, but each character is
> duplicated. Here is a picture: (not very sharp unfortunately)
> 
>         http://w1.894.telia.com/~u89404340/radeonfb.jpg
> 
> Text added after the switch is not corrupted, so eventually the
> corruption is scrolled off the screen and after that the framebuffer
> appears to be working correctly.

It's a known artifact caused by my latest stuffs, mostly because
I setup the display earlier than expected by the fbcon core, at
which point the console buffer contains junk. I'm working on a fix
though I can't reproduce on pmac.

> 2.4.22-pre3 does not have this problem. I haven't found a patch for
> the vanilla 0.1.8 version, so I don't know if that version also has
> this problem. I think someone has reported a similar problem in 2.5.x,
> but I don't remember the details.
> 
> Here are some messages from the kernel log:
> 
> Jul 14 23:08:44 best kernel: radeonfb: ref_clk=2700, ref_div=12, xclk=18300 from BIOS
> Jul 14 23:08:44 best kernel: radeonfb: panel ID string: Samsung LTN150P1-L02    
> Jul 14 23:08:44 best kernel: radeonfb: detected LCD panel size from BIOS: 1400x1050
> Jul 14 23:08:44 best kernel: Console: switching to colour frame buffer device 175x65
> Jul 14 23:08:44 best kernel: radeonfb: ATI Radeon M7 LW DDR SGRAM 64 MB
> Jul 14 23:08:44 best kernel: radeonfb: DVI port LCD monitor connected
> Jul 14 23:08:44 best kernel: radeonfb: CRT port no monitor connected
-- 
Benjamin Herrenschmidt <benh@kernel.crashing.org>