> > There are a growing number of 64 bit PCI implementations. Alpha has quite
> > a chunk of this, and so does UltraSparc.
> >
> > Is there a framework in place or a notion of how the Linux PCI stuff ought
> > to cope with this yet?
>
> There is something in the sparc tree, which is used also for sun4c and sun4d
> (where there is only 64MB of DMA per board, so one has to remap DMA areas
> during runtime).
> Besically, each driver before it starts doing dma uses some function which
> translates virtual address + length to dma address and allocates it, then
> after the dma the driver has to release it again.
> Martin Mares and myself are considering this as part of the new buses
> interface we plan for 2.3. The Ultra port supports even now huge amounts of
> memory, but uses bounce buffers for that. I plan to change it in early 2.3.
>
Good. There's a moral equivalent for this in some alpha implementations
(rawhide comes to mind). I have two different Qlogic cards that
could be ready to start taking full 64 bit addresses.
-matt
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/