Sun 2/170 Hardware (?) problem.

[Michael Burgett]

I am having a problem with the following configuration:

Hardware:
  Sun 2/170, 2MB Memory, xylogics 450 disk controller, CPC Tapemaster,
  3Com ethernet, Sun 2 Monochrome Video

  Disks:
	0 - Fuj 2322 SMD drive.
	1 - Fuj 2322 or 2312 drive

Software:
  SunOS 3.2

Symptoms:
  Drive 1 gets (many) errors, usually of the retry type.  Popular errors are:

xy1c: Write(or Read) Retry: (Memory Address Error) blk #xxxxx, abs blk #yyyyyy

and 

xy1c: Write(or Read) Retry: (Disk Sequencer Error) blk #xxxxx, abs blk #yyyyyy

which are annoying, but appear fairly harmless.  Occasionally I get the
dreaded Cylinder/Track error (or some such) which is usually the harbinger
of doom for some file(s) on the drive.

What I've tried:

  Replace both the data and command cables.
  Swapped controllers.
  Had my memory boards tested in another system.
  Tried 2 different drives as drive 1 (the 2312 and the 2322).
  Reformatted and tested the drives with diag. (This never shows any errors!)
  Changed cabling and termination so drive 0 and 1 have each had a turn at the
    end of the command cable. (Yes, only the drive at the end of the cable has
   termination.)
  Run each drive off of the others power supply.
  Connected the signal ground on the two drives.
  Re-arranged the cards in the 2/170 card cage.
  Running each drive off of a separate controller. (Although I'm going to try
    this again, just to make sure.)
  re-creating vmunix (Binary system, very limited compilation/linking of 
    kernel).
  Swearing, threatening and cajoling. ( .01 :-) )

Other Things I've noticed:

  blk # and abs blk # are the same. (Of course they are, I'm using the 'c'
    partition.)
  No matter which drive I have on as # 1, at least some of the blk #s listed
    as errors are the same. (20 and 80 are *very* popular.)
  all of the diag stuff (dmatest, sformat, test, scan, seek) runs without
    a single error.

Things that are not options:

  running without the second drive.
  upgrading to a later version of the OS.
  Paying $$ to have this fixed (By Sun or third party.  This machine sits in
    my home, and is used for mail/news and hacking/playing with C code.  I
    simply cannot afford to pay someone to come in and look at it.)

Please respond via email to {bdcsys | fiver}!guux!burgett as this is quite
obviously not an item of general interest.  If anyone is having a similar
problem, please email me and we can compare notes, and I will pass along
any clues and tips that I get.  Thanks. Mike Burgett