DMA on MVII

[Don E. Davis]

References:


Ultrix-32m has a /sys/vaxuba directory and supports kernal routines
such as ubago() and ubadone() even though the MVII is a Q bus rather
than a Unibus machine.  This makes it nice for those of us who
have to port drivers from VAX 780's.

Or at least it should.  I've encountered a strange problem during
DMA.  Every now and then, a 4 byte DMA will fail.  I'll get an
interrupt from the MDB DR11W board saying the DMA has completed,
but the contents of the target memory location will not have been
changed (even though the board CSR says READY).  

One of the things I monitored during debugging was
the address generated by the ubago() routine to my dgo() function.
For several transfers in a row, it will give me an address
of, say, 167734 (where 167 is a map register and 734 is the page
offset).  Then the address will change to 210734 and the DMA will 
fail.  The change of address always seems to precede the failure
of the DMA.  I am locking pages in memory with vslock().  

Has anyone seen anything like this before?  
-- 

					Don Davis
					JHU/APL
				...decvax!harpo!seismo!umcp-cs!aplcen!aplvax!ded
				...rlgvax!cvl!umcp-cs!aplcen!aplvax!ded