DEUNA problems: 4.2bsd, VAX 11/780-5

[Steven Bellovin]

Have anyone seen, or heard of, UBA map problems with the DEUNA driver?  We just
installed a DEUNA as the only DMA device on a 780-5's UBA, and we're
having some trouble.  It's hard to boot, since the probe routine often (but
not always) sets the IVMR (invalid map register) bit in the UBA Status register
for either the DEUNA probe, or sometimes the probe of the next device; either
way, one device or the other is often skipped by autoconf.c.  I've tried
lengthening the DELAY call (since it is a faster machine than DELAY was
originally intended for), and I've added a DEUNA RSET after the GET PCBB
command's delay, but to no avail.  Once, we've gotten the IVMR as a separate
uba0 error when we tried going multi-user; the FMER (failing map entry register)
pointed to map 0, and FUBAR was pointing right at the DEUNA csr, though the
VAX manual says nothing about what FUBAR should be on IVMR.

As best I read the code and the DEUNA manuals, GET PCBB should not do any
DMA unless some other port commands are sent, but the DEUNA appears to be
doing just that.  Besides, autoconf appears to validate map register 0.  We've
seen the problem with two different DEUNAs, and the UBA passes its diagnostics.

We're running 4.2bsd; we have moved the non-disk autconf code to before
the disk autoconf code, to keep our 3Com boards happy.  We have three other
DEUNAs, one on a 780-5, and we've never had any such problem.


		--Steve Bellovin
		{ucbvax,ihnp4,decvax, most of BTL}!ulysses!smb