> > It worked fine until the latest framebuffer merges (of course without > > CONFIG_FB_ATY_GENERIC_LCD, which did not exist before the merges). > > In 2.4, CONFIG_FB_ATY_GENERIC_LCD depends on ia32.
We should fix that indeed. I found other problems with the "new" atyfb from James tree (which is what is in 2.6.3-ben at the moment). I fixed some (maybe not pushed yet) but it's unclear what's really going on (the wait_for_idle calls I added shouldn't be necessary, I lockup without them, and still had once lockup in console with them, though I can't reproduce now). It seems something dodgy is going on with the accel code that got added to this driver in 2.6. Ben.
