At 9:43 AM -0400 4/26/02, Reinier Jonker wrote:
>A few weeks ago, there was some discussion an hardware DMA
>implementation on AV Quadra's.
>Since I own a Macintosh Quadra 840AV, I'm quite interested in the
>current state of work.
>How's progress?
>
>It might be an idea to send something in the port-mac68k mailinglist.
>
>I'd also be happy to help, but I'm affraid my lack of OS programming
>experience is a serious problem :-)

It's coming along. Allen's contributed the bus_dma code which, amongst
other things, gives drivers physically contiguous regions to DMA to/from.

I made a DMA version of the SCSI driver but it was using bounce buffers
rather than real DMA. I'm working on a generic DMA driver that uses bus_dma
so that we can eliminate the bounce buffers and get the best possible
performance.

I just worked on the generic DMA driver today. It's coming along but I
haven't gotten time to test it yet and I still need to futz around with the
hardware/ROM code a little bit more to find out where the SCSI DMA
interrupt is. I actually have a gut feeling that all the DMA's come in
through what we call the Ethernet DMA interrupt today but I have to
experiment to actually find out if this is true.

Once the generic DMA driver is done I just have to fix up my SCSI code to
use that instead of the bounce buffer stuff (not too hard) and I may have
to fix up the ethernet code to avoid any conflicts, if it turns out there
are any.

So that's where we are right now. Technically, I think it's in the bag. The
real stumbling block is allocating the time to do the necessary things.


----------------------------------------------
 Michael Zucca - mrz5149@acm.org
----------------------------------------------
 "I'm too old to use Emacs." -- Rod MacDonald
----------------------------------------------