From: Paul Kranenburg <pk@cs.few.eur.nl>
   Date: Thu, 27 Mar 1997 22:47:43 +0100 (MET)

   [...explanation and pseudocode removed...]

Ok, I SUP'ed the changes and had some time to test them. (I guess my
combination of equipment is good at producing the bug so I'll
volunteer myself as gunieapig...)

On a positive note, I _did_ manage to copy two files (4Meg and 5Meg)
that previously I had never been able to copy, and a quick verify
using diff confirmed they copied correctly. Now for the bad news...

It still produces errors, in much a similar way, but of different
type. On writing, I even managed to panic the kernel twice.

The errors I got:

esp0: illegal command: 0x1 (state 4, phase 3, prevphase 3)

and later

esp0: illegal command: 0x12 (state 4, phase 7, prevphase 3)

both of these would hang the machine, near as I can tell - the console
would still echo characters, but nothing is happening.
Amongst the non-fatal errors (fixed by umounting and remounting) were

esp0: reselect without IDENTIFY, sending DEVICE RESET
esp0: identify failed

along with the usual unit attention, data = 00 ... stuff

Upon attempting to write large files, however, I have not managed to
do so yet since I will either get the "reselect without IDENTIFY" or a
kernel panic. The first time this happened I was not fast enough to
stop the reboot before writing it own, but the next time I did:

mode = 0100600, inum = 15, fs = /mnt
panic: ffs_valloc: dup alloc

Along with a traceback. Unfortunately, savecore claims there is no
core dump, so I can't give you any info that may produce (I am
admittantly very novice at this).

I hope this info will help...

Joachim.
BTW, my setup is a SS1+ with a Quantum PD700S at scsi 2 and a IOMEGA
ZIP at scsi 6. Accessing the Quantum has never produced any errors
whatsoever.