My UDB's disk drive has recently stopped working (it doesn't even spin
up anymore). No problem, I thought, it was crap, anyway, I will just get
some decent disk in there. Of course, no can do --- the power supply is
too weak.
So, my idea was to share the SCSI disk in my PC with the UDB. The setup
would be something like
_________________ ____________________
I V I
NCR NCR disk
UDB Pentium
with termination on the UDB NCR and on the disk. The UDB's NCR is supposed
to get SCSI id 5, the Pentium will grab 7, and the disk is at 0. Of course,
the UDB and the Pentium will use seperate partitions on the disk, otherwise
it would be sure to end in disaster.
Well, after unsuccessfully trying to convince the ARC console to leave alone
SCSI ID 7, I switched to SRM (and boy, I like it :-), and SRM has no problems
doing as it is told.
I managed to install a (minimal) RedHat 4.0 on the UDB's partition (while
the Pentium was sitting at the CMOS setup screen, thus leaving alone the
SCSI bus).
I _can_ boot both and use both machines at the same time (the Pentium gets
slightly annoyed at the SCSI bus being reset when the UDB boots, but
nothing bad happens). BUT:
Every now and again, the UDB will stop working with a message to the effect
of (from memory)
ncr53c8xx: aborting command due to timeout
and then a PID and some more info that doesn't seem to helpful. Unfortunately,
not only will it not recover from such a timeout, but it also seems to kill
the SCSI bus --- the Pentium can't access the disk anymore, either. BAD!
OK, I _do_ have an idea --- if I understand it correctly, the Pentium has
higher priority on the bus than the UDB. So it might happen that the UDB
wants to talk to the disk, and as the Pentium is using the bus at the
time, won't get through. This is probably unanticipated by the driver
writers, and might in extreme cases lead to a timeout. Does that sound
reasonable?
Now, my questions:
a) Any idea where I could change the timeout period to some really high
value? And why does a timeout hang the SCSI bus?
b) Is the suggested setup possible at all, using the cheap NCRs?
c) I am using the 2.0.21 kernel with the BSD-53c8xx driver on the Pentium,
and the default RedHat 4.0 Noname kernel on the UDB. Any known
problems with the SCSI drivers in either (if so, I can recompile
kernels for both platforms without trouble).
d) On a related note --- is there a way to make the SRM console autoboot
my MILO?
Any help appreciated --- that UDB is waiting to go into action again!
TIA
Bernie
-- To unsubscribe: mail -s unsubscribe axp-list-request@redhat.com < /dev/null
Copyright © 1995-1997 Red Hat Software. Legal notices