SCSI problems

["Dieter BSD" <dieterbsd%engineer.com@localhost>]

Alpha 164lx
NetBSD 3.0.1
QLogic SCSI controllers (wide, diff)

Once upon a a time this machine was running fine with these controllers
and disks.  The disks were moved to a UDB/Multia box and worked fine there.
UDB/Multia box died, I moved the disks back to the 164lx.  I would like to
(1) get the data off the disks, (2) install NetBSD 5 on one of them to see
if an unrelated problem has been fixed.  But...

isp0 at pci2 dev 4 function 0: QLogic 1020 Fast Wide SCSI HBA
isp0: interrupting at eb164 irq 3
scsibus0 at isp0: 16 targets, 8 luns per target

sd1: sync (100.00ns offset 8), 16-bit (20.000MB/s) transfers, tagged queueing
sd2: sync (100.00ns offset 8), 16-bit (20.000MB/s) transfers, tagged queueing

isp0: Auto Request Sense failed for 0.1.0
isp0: BOTCHED cmd for 0.1.0 cmd 0x25 datalen 8
sd1(isp0:0:1:0): generic HBA error
sd1: sync (132.00ns offset 8), 16-bit (15.150MB/s) transfers, tagged queueing
isp0: Auto Request Sense failed for 0.1.0
isp0: BOTCHED cmd for 0.1.0 cmd 0x23 datalen 12
sd1(isp0:0:1:0): generic HBA error

132.00ns ?  15.150MB/s ?
Smells like data corruption?  Maybe the controller card went bad?

I tried 2 other controller cards with identical results.

# disklabel sd1
disklabel: /dev/rsd1c: Device not configured
# disklabel sd2
disklabel: /dev/rsd2c: Device not configured

isp0: Auto Request Sense failed for 0.1.0
isp0: BOTCHED cmd for 0.1.0 cmd 0x25 datalen 8
sd1(isp0:0:1:0): generic HBA error
sd1: sync (132.00ns offset 8), 16-bit (15.150MB/s) transfers, tagged queueing
isp0: Auto Request Sense failed for 0.1.0
isp0: BOTCHED cmd for 0.1.0 cmd 0x23 datalen 12
sd1(isp0:0:1:0): generic HBA error
isp0: Auto Request Sense failed for 0.2.0
isp0: BOTCHED cmd for 0.2.0 cmd 0x25 datalen 8
sd2(isp0:0:2:0): generic HBA error
sd2: sync (132.00ns offset 8), 16-bit (15.150MB/s) transfers, tagged queueing
isp0: Auto Request Sense failed for 0.2.0
isp0: BOTCHED cmd for 0.2.0 cmd 0x23 datalen 12
sd2(isp0:0:2:0): generic HBA error

None of the reasons for "Device not configured" listed in
http://www.netbsd.org/docs/kernel/
appear to apply. Am I getting "Device not configured" because the
commands are failing?  The disks are configured into the kernel:

strings /netbsd | grep "at scsibus0 target"
_CFG_sd0  at scsibus0 target  0 lun 0
_CFG_sd1  at scsibus0 target  1 lun 0
_CFG_sd2  at scsibus0 target  2 lun 0
_CFG_sd3  at scsibus0 target  3 lun 0
_CFG_sd4  at scsibus0 target  4 lun 0
_CFG_sd5  at scsibus0 target  5 lun 0
_CFG_sd6  at scsibus0 target  6 lun 0
_CFG_sd8  at scsibus0 target  8 lun 0
_CFG_sd9  at scsibus0 target  9 lun 0
_CFG_sd10 at scsibus0 target 10 lun 0
_CFG_sd11 at scsibus0 target 11 lun 0
_CFG_sd12 at scsibus0 target 12 lun 0
_CFG_sd13 at scsibus0 target 13 lun 0
_CFG_sd14 at scsibus0 target 14 lun 0
_CFG_sd15 at scsibus0 target 15 lun 0

It consistantly says 100.00ns 20.000MB/s at boot.
IIRC 100ns 20MB/s is correct. It consistantly says
132.00ns 15.150MB/s after boot.

3 cards failing the same way is rather unlikely.
2 disks failing the same way is rather unlikely.

What could be causing this?