Hi Mark.

In message <Pine.SGI.3.95.970203033052.14622D-100000@physig4.ph.kcl.ac.uk>
          Mark Brinicombe <amb@physig.ph.kcl.ac.uk> wrote:

> Ok DMA will not be working as DMA support is not inplace in the existing
> drivers. This was due to the fact that the interrupt problem needed to be
> fixed before DMA could be implemented.

Oh no, I have got this terribly wrong a long time ago when these issues were
discussed in length for the first time.
I always thought the code for DMA and interrupts was in place and just didn't
work.

> However now that interrupts are fixed Scott and myself have been doing a
> bit of playing with DMA on the Cumana and Powertec cards.
> The major problem we have had to sort out is the fact that the IOMD DMA
> information in the TRM is not actually correct and DMA in practice behaves
> slightly differently. This is taken some time to suss out.
> The current sticking point in in the generic sfas code. I need to add
> support for cache cleaning and syncing for DMA transfers.
> 
> DMA is getting very close ....

Fine. I'm not going to be impatient because code still has to be written. I
always thought it was only a matter of bugfixing to get the SCSI drivers going
faster.
There are quite a lot of people who would benefit from this though so I hope
it's given a high priority.

Btw, can it be that a new version of csc.o is needed once again because of the
latest changes to the bus handling code? I think it is because all the other
drivers where changed. Can you please put it online then?

So long, Markus