SUMMARY Re: "Illegal Request" messages from dump to Wangtek QIC-150

From: Bruce_Hamilton.LAX1B@xerox.com
Date: Thu Oct 17 1991 - 00:19:38 CDT


[I haven't tried the suggestions yet, but I'm willing to believe that SCSI
flakiness is the culprit.]

MY ORIGINAL QUESTION:

SS2, SunOS 4.1.1, Wangtek QIC-150 (in a SCSI shoebox from Andataco with a 669MB
disk).

Has anybody ever experienced the following console error messages when trying
to dump to a Wangtek QIC-150 or similar device? Dump happily enumerates the
directories, then bombs as soon as it tries any device operations. The errors
seem to come and go -- I've been unable to correlate them with my actions. I
HAVE done a couple of successful dumps and restores, but unfortunately, right
now I'm getting these error messages about 90% of the time.

Although the message from dump is something like "write error 5 feet into
tape", in fact no media operations were ever performed -- the media light never
went on. It's like the controller is in some screwy state (any way to reset
it?). The first console message below appears when dump first aborts, then the
second one when I tell dump to quit rather than retry.

st0: Error for command 'write', Error Level: 'Fatal'
            Block: 0 File Number: 0
            Sense Key: Illegal Request
st0: Error for command 'write file mark', Error Level: 'Fatal'
            Block: 0 File Number: 0
            Sense Key: Not Ready

Here's an example of the command sequence I'm using to invoke the dump:

  mt rewind
  sleep 5
  dump 0ufcs /dev/nrst0 1500 /disk2

----------------------------------------------------------------
THE REPLIES:
----------------------------------------------------------------
From: Jim Guyton <guyton@rand.org>

I've had lots of fun with SCSI over the last few years. Here's
a checklist for you ...

  0) Triple check the cables. Replace all/most of the scsi bus
     cables and see if the problem goes away. Make 'em shorter if
     you can.

  1) Is BOTH ends of your SCSI bus terminated? Hint: the innards of
     the SS/1 and SS/2 have room on the bus for two internal disks and
     that counts as one end of the bus! Hint #2: you can run with the
     bus not-properly terminated for a long time w/out knowing it.
     I've purchased internal scsi terminators for us, but it's usually
     only needed when the cable length gets a tad long.

  2) Have you tried turning off DISCONNECT-RECONNECT in the kernel?

  3) If (1) and (2) aren't enough, I've got a nifty little SCSI disk mode
     page editor that I've been hacking on (and off; mostly off) over the
     last year that will let you examine (and change) the mode registers
     in the disk. It's of limited use w/out either my handholding you
     through it, or unless you've tried to read a SCSI manual and/or want
     to learn all about it ...

----------------------------------------------------------------
From: etnibsd!vsh@uunet.uu.net (Steve Harris)

Have you tried doing (just to see what would happen):

        dump 0ufcs - 1500 /disk2 | dd of=/dev/nrst0 bs=512k

What about:

        dd if=some_huge_file of=/dev/nrst0 bs=512k

i.e., is the problem with the tape-drive/controller/tape, or is
it some interaction of dump and the drive?

----------------------------------------------------------------



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:18 CDT