interfacing an Exabyte to a 4D/25

[Dave Olson]

In <3895 at umbc3.UMBC.EDU> strow at umbc5.umbc.edu (Dr. Larrabee Strow (PHYS)) writes:

| I am trying to hook an Exabyte tape drive up to a SGI 4D/25 running
| IRIX 3.2.  The tape drive uses address 7. When the system is booted 
| the power up diags fail with:
| 
| >    sc0:  Unexpected transfer phase.  State=4b  Phase=33 
| >    scsi (0,7,0) transfer aborted (Hardware error)   
| >    Device 7 failed DMA test
| >
| >    Diagnostics failed 

The problem here is with 'bad' firmware in the Exabyte.  The power up
diagnostics on the PI execute the senddiag command.  The Exabyte does
not accept this command, but unlike every other SCSI device I have dealt
with, it goes to a message in as soon as it sees the first command byte
(well, apparrently the 3rd in your case).  The solution is to get the
latest rev firmware for the Exabyte, or to suppress teh diagnostic.  Try
setting the bootmode to C (a capital c) in the PROM monitor.  Unless you
have very early PROM's, this will suppress most of the SCSI diags at boot
time, including the senddiag command.

| If I allow the bootup to continue, the system becomes essentially useless,
| pandora doesn't load, and I keep getting scsi errors from the hard drive,
| number 1.

This sounds as though you may have termination problems, as well.  Be
absolutely sure that none of the internal devices have terminators on them,
and if you have any external SCSI devices, only the last one on the cable
should be terminated.

| The supplier had us try changing dip switch number 2 on the MX board but
| the power up diags did not change.

Switch 2 is for parity, and should be enabled.  Turning on switch 1 will
shorten the power-on tests, which has alleviated some problems on some
machines (that is the way we ship it for the PI).  You should also have
switch 3 turned on, to force the Exabyte to disconnect only on a word
boundary, otherwise you will get error messages about unaligned DMA.

| Hopefully we just need to know the proper switch setting on the MX board -
| although the supplier should be able to help here they have no suggestions.

I don't think the switch settings are the problem, it sounds more like
a cable or termination problem.
--

	Dave Olson

Life would be so much easier if we could just look at the source code.