radeon driver: checkpoint

[Benjamin Herrenschmidt]

> That's ok. I think the problem is that RADEONPreInitVRAM misses a return 
> value. I'm not sure what the results of that exactly are as it works for 
> me, but I'd guess it's undefined which might cause the server to abort, 
> in case the result is non-zero. (Wondering why I don't get any compiler 
> warning...)
> 
> > (**) RADEON(0): RADEONFreeScreen
> Which is exatly what's happening here.

Yup, I'm surprised by the absence of a warning too...

> Some other unrelated remark, I think that the ChipFamily detect for 
> setting HDP_APER_CNTL is wrong, it should probably include at least 
> CHIP_FAMILY_RV410 too (which is newer than either R420 and RV380). Not 
> sure about igps (rs400 is definitely a lot newer), maybe it doesn't 
> matter for them.

I added it. It shouldn't matter for IGPs indeed.

I also fixed on the the messages, though I don't see what is wrong with
the the one about Xv DMA ...

Ben.