Mailinglist Archive: radeonhd (408 mails)

[Alex Deucher <alexdeucher@xxxxxxxxx>]

2009/6/26 Dariem Pérez Herrera <dariemp@xxxxxx>:
> Hi,
> Some time ago, I wrote about issues with GPU RS600. Now, with new updates
> (linux-2.6.30, latest mesa, xorg and radeonhd from git repos) the same
> problems persist. "MC not idle" is the tipical error message in log. My card
> (Radeon Xpress 1250) works fine with "radeon" driver (although with no
> acceleration and no SVIDEO output detection). I've been trying to hack a
> little into the source code of radeonhd and I found incorrect values for
> certain registers addresses for RS600. I'm attaching a patch. Before this
> changes, my notebook totally froze when starting X, leaving me with the only
> option to shut it down through power button. Now I just get a black screen,
> but the keyboard keeps responsive, so now I can restart it properly. That
> happened before, because the function to detect idle MC couldn't detect it,
> because of the incorrectly set registers, and now it works fine, telling me
> the error at the proper time, but still....
> I also attached my Xorg.0.log with -logverbose 7. I can see there are many
> functions been called which names suggest they are made for RS690, not
> RS600, so it is possible that MC cannot go idle because of this (incorrectly
> use of RS690 register addresses instead RS600 addresses). I'm not familiar
> with radeonhd source code, so trying to fix MC make me lost into the code
> ocean. Any mentorship or documentation available so I can try?

The radeon driver should have the correct bits for indirect MC access
on RS690/740 and RS600.  Note that RS690/740 and RS600 are different
WRT MC setup.  See:
RADEONINMC()
RADEONOUTMC()
radeon_get_mc_idle()
in radeon_driver.c for the relevant regs and bits.

Alex
--
To unsubscribe, e-mail: radeonhd+unsubscribe@xxxxxxxxxxxx
For additional commands, e-mail: radeonhd+help@xxxxxxxxxxxx