I closed PR kern/5654 (boot hangs due to scsi timeout) - was hardware.
I swapped the controllers and I have my 4G disk running now.

However it looks like the 1542B was worse than I thought - appeared to
be just the on-board terminators stuffed.  That might still be the
case, but the machine panic'ed a couple of times today with a "bad
dir", and sd0a appears quite comprehensively stuffed.

Running fsck shows so many errors its hard to believe a couple of
panic's could do it.

After the 2nd panic, when I boot single user and hit return for
/bin/sh I get the error:

aha0: aha_scsi_cmd, error 22 loading dma map

and its reset time.

I can run /bin/ksh fortunately, though attempts to run some other
programs also result in the dma map error.

The dead box was of course my bastion for the Internet link.  I've
setup the filters and split named and a few other things on zen so I
can get the link up to get this msg out... but there is a lot to
reconfigure to do without the bastion box permanently.

I'm confident that the 1542B is correctly configured, the bus is
terminated ok - removing the external passive terminator results in
the controller signally a lack of termination and failing to find sd0
at all.  The controller was running with parity checking enabled, and
seemed happy last night.  This evening I disabled parity checking with
not much change in behaviour.

The rest of the hardware has been quite reliable for several months,
though I've probably disturbed all the dust in switching the
controllers.

Can anyone give me the 30 second summary of the above dma error, and
whether it is likely to mean that the controller is terminally
ill, or just that the filesystem has been seriously corrupted?

Thanks.

--sjg