Radeon/PowerPC: No video output after update 6.12.1 -> 6.12.99; 6.12.192

[Alex Deucher]

On Fri, Apr 23, 2010 at 2:00 PM, Albrecht Dreß <albrecht.dress at arcor.de> wrote:
> Am 22.04.10 19:23 schrieb(en) Albrecht Dreß:
>>
>> Sorry for the long delay, but it took siome time to find a working vga
>> monitor...
>>
>> Am 14.04.10 20:52 schrieb(en) Albrecht Dreß:
>>>
>>> Am 13.04.10 23:01 schrieb(en) Alex Deucher:
>>>>
>>>> Do either of these options help?
>>>> Option "ConnectorTable" "100,2,1,2,96,1,0,1"
>>>> Option "ConnectorTable" "100,2,2,2,96,1,0,1"
>>>
>>> They don't make any difference, same effects as before  - greenish
>>> shades, vertical blue line.  Switching to the text console works (as I can
>>> type "reboot", so the kb input goes the right way), but the display just
>>> flickers shortly, and then the shade/line does appear there, too.
>>>
>>> I also tried some other options (like 'Option "BusType" "PCI"' and
>>> 'DefaultDepth 8'), but they also don't make any difference - same effects,
>>> again.
>>>
>>>> Do you have a VGA monitor by any chance you could test with?
>>
>> Alex, this was apparently the right hint!  Here are my observations:
>>
>> Plugging *only* the vga monitor in, and disconnecting the ADC one, the Mac
>> refuses to boot.  It switches on, and after ~5 seconds (system test?) it
>> automatically powers off (sounds like a Notebook's card?).  The only chance
>> to boot is plugging in both monitors.  I first booted into OS X, and had
>> both monitors synchronised.
>>
>> I then booted into Linux, and now X comes up with the radeon driver,
>> showing the X screen on the VGA monitor, and the well-known green shades on
>> the ADC one.  This behaviour is independent of any 'ConnectorTable' option;
>> there is no difference between starting without, and with any of your two
>> settings above.
>>
>> An other observation: The VGA monitor has a smaller resolution (old
>> Belinea 1024x768), but X apparently used the size of the Formac attached to
>> the ADC (1280x1024).
>
> ... and now I also tested all those "MacModel" options, always rebooting
> between two tests:
> - ibook and powerbook-* produce no output/shades on the ADC connected tft;
> - imac-g5-isight seems to be out-of-sync for the tft;
> - mini-* and emac *do* actually produce a display on the tft, but smaller
> than the actual size (1152x864 instead of 1280x1024), with the right border
> being black, and a stripe from the top copied to the bottom.  Same happens
> with emac, but clipped to an even smaller size.
>
> A log from the mini test is attached.
>
> Is that helpful in any way?

Mac's don't have connector tables in their bioses like x86 boards, so
we have to hard code them based on how they are wired up per board.
That MacModel option just hardcodes connector tables the same way the
ConnectorTable option does.  It's just easier to specify the mac model
than to specify a ConnectorTable string.  I doubt they will be useful
in your case, since they are just arbitrary combinations of encoders,
connectors, and ddc channels corresponding to a specific machine which
may or may not be similar to your particular board.  The default
connector config for your system looks to be correct as far as I can
tell.  I'm not sure why it's not working.  Any chance you could bisect
between the last working version and something more current to see
what commit broke things?  Or dump the regs using radeontool in the
working and non-working versions of the driver?

Alex

>
> Thanks, Albrecht.
>