> 1) It appears to happen only with one disk, an old Quantum 210S (SUN0207).
> Other disks, either with or without the Quantum present, do not seem to
> have the problem.

Having installed the latest snapshot, I had also noticed the problem.  My
3/80 has a pair of Quantum 425's in it.  I hadn't got around to looking at
the paritular piece of code that causes this.  I also noticed that I get
exactly the same symptoms when I un-mount a partition.  I've not tried
accessing the disk after the problem occurs.

I've suspected for a little while that there is something in the Quantum
SCSI firmware that our SCSI code doesn't get on with.  I have SUN0207
which is actually a Quantum 210S and I can't get an IPC nor the 3/80 to
work with it - every time a write completes, I get the esp errors and have
to wait for 2 timeouts before I can access any disk again.  Anything I've
written to the disk is there and reading is fine.  Strange.  When I sent
a message to the port-sparc list, I received lots of suggestions that it
was the termination or SCSI cabling.  I changed them.  I'm 99% sure that
SunOS worked on the same hardware though (not sure if that's relevant).

One other thing - the 3/80 used to run 1.3.3 with the same pair of 425's.
No problems at all.  Did we get the Sparc esp driver for 1.4?

J

PS.   While I'm here ... with a slightly trimmed down GENERIC3X, I get
'panic: malloc: out of space in kmem_map' errors when trying to build the
world.  Does anyone else see these?  If so, is it worth bumping up the
default?  (It's compiling a kernel with larger NKMEMCLUSTERS at the moment.)
-- 
                    My other computer also runs NetBSD
                          http://www.netbsd.org/