From: Leif D. <lde...@re...> - 2003-04-22 22:36:49
|
On Tue, 22 Apr 2003, Ian Romanick wrote: > Leif Delgass wrote: > > Disregard my comment about mga and __MUST_HAVE_AGP. It does define it, I > > was grepping in the wrong directory. At any rate, building the drm without > > CONFIG_AGP should work again now for r128, radeon, and sis (if you > > uncomment sis.o in MODULE_LIST). > > It shouldn't. There are still PCI G200s and PCI G450s out there. I see > no reason why those cards shouldn't work on systems that don't have AGP > available. :) Does that mean you're volunteering to add PCI support? ;) I think the PCI changes necessary for mach64 could come in handy there. There's a microcode and "primary" DMA region that would probably need to be allocated with pci_alloc_consistent. Speaking of which, I've been looking a bit at what needs to be done to add mapping support to the PCI interface Jose proposed. One of the problems as I understand it is that virt_to_page() is not a valid operation on an address returned by pci_alloc_consistent(). I've seen a proposal for a pci_consistent_to_page() function for this purpose, but I don't think that's happened yet. -- Leif Delgass http://www.retinalburn.net |