EXA for radeon experimental patch

[Benjamin Herrenschmidt]

On Thu, 2005-09-01 at 23:54 +0200, Lars Knoll wrote:
> On Thursday 01 September 2005 23:16, Michel Dänzer wrote:
> > On Thu, 2005-09-01 at 11:35 +0200, Lars Knoll wrote:
> > > Even if the card doesn't have scatter/gather support, drm allows you to
> > > allocate a piece of consistent physical ram, and mmap it in the server.
> > > The handle you get is the physical address, so you should be able to use
> > > that to implement PCI dma transfers.
> >
> > Are you sure about that, in particular on a 64 bit system? The handle
> > may coincide with the physical address in a lot of cases, but I didn't
> > think that was guaranteed.
> 
> It's probably not guaranteed, but according to what I could see in the drm 
> code that's what it is (also on 64 bit systems, the handle is a ulong). Lots 
> of the drm code is still a mystery to me. To program the HW I need physical 
> addresses, but I don't really see how to get them. The only exception is 
> memory that is bound into the GART.

Yes, you are not supposed to use anything but GART memory at this point
I suppose, unless the DRM provides other means to you, but you certainly
shouldn't assume that the handle is a physical address. That may/will
change.

Of course, your nv-specific DRM driver could provide some additional
capabilities that you could use, and if they end up beeing useful enough
in the long run, could be moved to the main DRM.

Ben.