[PATCH] Fix ConnectorTable crash in radeon_output.c

[Michel Dänzer]

On Mit, 2012-05-02 at 08:48 +0200, Hans Verkuil wrote:
> On Tue April 3 2012 08:54:53 Michel Dänzer wrote:
> > On Fre, 2012-03-30 at 19:37 +0200, Hans Verkuil wrote:
> > > On Friday, February 24, 2012 15:15:40 Michel Dänzer wrote:
> > > > 
> > > > FWIW, I think most likely the problem is that acceleration is disabled,
> > 
> > The Xorg.0.log file confirms this.
> > 
> > 
> > > > probably because the GPU microcode from the firmware-linux-nonfree
> > > > package is missing.
> > > 
> > > I have that package installed, but it doesn't seem to be loading any firmware.
> > 
> > Please provide the dmesg output.
> 
> Here it is!

[...]

> [    0.000000] Using Sam440ep machine description

[...]

> [    2.542438] [drm] Loading R200 Microcode

This is where it's loading the firmware aka microcode.


> [    2.713029] [drm:r100_ring_test] *ERROR* radeon: ring test failed (scratch(0x15E4)=0xCAFEDEAD)
> [    2.738769] [drm:r100_cp_init] *ERROR* radeon: cp isn't working (-22).
> [    2.758263] radeon 0000:00:0c.0: failed initializing CP (-22).
> [    2.775678] radeon 0000:00:0c.0: Disabling GPU acceleration

This is why acceleration is disabled: the test of basic acceleration
functionality is failing.

AFAICT from my e-mail archive, the 440 has >32 bit physical addresses
and is not cache coherent, neither of which is supported by the radeon
driver yet.

For the latter, you could try if hacking the driver as described in
http://lists.ozlabs.org/pipermail/linuxppc-dev/2012-April/097668.html
helps. For the former, Benjamin Herrenschmidt (CC'd) might be able to
point you to some patches I keep hearing about but haven't seen yet. 


-- 
Earthling Michel Dänzer           |                   http://www.amd.com
Libre software enthusiast         |          Debian, X and DRI developer