18-bit Device on a MicroVAX II

[Dan Ts'o]

In article <349 at rhesus.primate.wisc.edu> bin at rhesus.primate.wisc.edu (Brain in Neutral) writes:
>We have built a Parellel DMA Interface in house.  We
>have been unable to read anything from the device although
>I have successful accessed the different registers on the
>card.(I have verified this with a scope)  I get an interupt
>and the word count goes to 0.
>
>I am trying to determine if the Microvax II can handle an
>18 bit device.  Is it necessary to have 22 bit addresses?
>The board does not have any more room on it to add the 
>wires that would be needed for 22 bit addressing.

	The Ultrix-32 / 4.3BSD IO code treats the Qbus as an 18-bit Unibus
and thus 18-bit Qbus peripherals work fine. That is, as far as Qbus DMA
devices are concerned, they never need to worry that they will ever be
handed a 22-bit Qbus memory address to which they need to transfer data
to or from. The IO subsystem maps the physical memory only into the first
256kbytes of Qbus memory. I believe VMS on the Microvax works the same
way. I believe that the Microvax I does have problems in this area as I
think it does not implement scatter/gather IO mapping.

	It is possible to put Qbus devices on the Microvax II that appear
as part of the 4Mbyte of 22-bit Qbus space. See the QVSS driver for an
example. I have implemented a driver for a Data Translation frame buffer/
grabber which allows the user program to directly access the frame memory
which appears as 1Mbyte of Qbus memory.

	I hope, however, that the device you built uses BBS7 to access
the device registers rather than a Qbus 18-bit (or 22-bit) address.

	Let me know if you have any more questions...

				Cheers,
				Dan Ts'o		212-570-7671
				Dept. Neurobiology	dan at rna.rockefeller.edu
				Rockefeller Univ.	...cmcl2!rna!dan
				1230 York Ave.		rna!dan at nyu.edu
				NY, NY 10021		tso at rockefeller.edu