At 7:31 Uhr +0100 29.01.1997, Michael R. Zucca wrote:

>While semaphores are good, there is still going to be some code that just
>must be executed with interrupts off.
>
>I'm not so sure why SCSI is such a bad culprit. I thought most II's and later
>had DMA scsi. Let the controller interrupt you when a DMA transfer is finished
>and lock/release the buffer using a semaphore. Maybe I'm just over-simplifing.

AFAIK, the only 68k Macs that do SCSI by DMA are the IIfx and the AV
machines (Allen, Scott?).

All others do some kind of pseudo-DMA where the CPU monitors the
ncr53c{80,96}'s DMA handshake lines.



	hauke

---
"It's never straight up and down"     (DEVO)