troublesome peripheral, kernel panics and pending commands

Hello,

I'm trying to use a SCSi scanner (Epson FilmScan 200) with latest SANE
and got a strange behaviour from the SCSI layer. I tried with two different adapters (initio and aic7xxx):

scsiinfo says:
Inquiry command
---------------

Relative Address                   0
Wide bus 32                        0
Wide bus 16                        0
Synchronous neg.                   0
Linked Commands                    0
Command Queueing                   0
SftRe                              0
Device Type                        3
Peripheral Qualifier               0
Removable?                         0
Device Type Modifier               0
ISO Version                        0
ECMA Version                       0
ANSI Version                       1
AENC                               0
TrmIOP                             0
Response Data Format               0
Vendor:                    EPSON
Product:                   FilmScan 200
Revision level:            1.01

recognition:

i91u: PCI Base=0xC000, IRQ=18, BIOS=0xFF000, SCSI ID=7
i91u: Reset SCSI Bus ...
scsi4 : Initio INI-9X00U/UW SCSI device driver; Revision: 1.04a
Vendor: EPSON Model: FilmScan 200 Rev: 1.01
Type: Processor ANSI SCSI revision: 01
4:0:4:0: Attached scsi generic sg4 type 3

the SANE driver does a SCSi inquiry:

[epson2] attach: sending INQUIRY
[epson2] >EPSON FilmScan 200 1.01<

and then sends out two bytes

[epson2] epson_send: size = 2
[epson2] buf[0] 1b .
[epson2] buf[1] 40 @

here is where it hangs. the driver expects to read a byte, but there seems to lie the problem.. the driver halts and I have to ctrl-c .

after I while, I can hear a reset from the scanner and the kernel hangs completely withtout leaving anything useful on the console or in the logs.

sometimes, if I do

rmmod sg
rmmod initio

before the reset, I get this message:

Initio INI-9X00U/UW SCSI device driver; Revision: 1.04a did not call scsi_unregister

 [<e0dd89e2>] exit_this_scsi_driver+0x8a/0xc7 [initio]
 [<c012e390>] sys_delete_module+0x19a/0x1c5
 [<c013da1a>] sys_munmap+0x3a/0x56
 [<c01026b5>] syscall_call+0x7/0xb

moments later the kernel panics
with Fatal exception in interrupt (included in the logs)

I tried enabling scsi logging, but since the system is based on SCSI disks, I have some troubles filtering the output:

after the ctrl-c:

sg_poll: sg4, res=0x104
sg_release: sg4
sg_fasync: sg4, mode=0
sg_remove_sfp: worrisome, 1 writes pending

so i've attached it. hope someone can help.

 Best regards,

 Alessandro Zummo,
  Tower Technologies - Turin, Italy

  http://www.towertech.it