>Number:         1243
>Category:       port-amiga
>Synopsis:       A3000 scsi hangs switching from/to sync
>Confidential:   yes
>Severity:       serious
>Priority:       medium
>Responsible:    gnats-admin (GNATS administrator)
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed Jul 19 05:05:01 1995
>Last-Modified:
>Originator:     Ignatios Souvatzis
>Organization:
	private site
>Release:        950710
>Environment:
System: NetBSD beverly 1.0A NetBSD 1.0A (BEVERLY) #117: Wed Jul 19 08:56:51 MET DST 1995 root@beverly:/usr/src/sys/arch/amiga/compile/BEVERLY amiga


>Description:
This is an A3000 with 12 MB of fastmem.

Installed disks are:

ahsc0 at mainbus0
scsibus0 at ahsc0
ahsc0 targ 0 lun 0: <QUANTUM, PD210S, 528_> SCSI2 0/direct fixed
sd0 at scsibus0: 199MB, 1156 cyl, 7 head, 50 sec, 512 bytes/sec
ahsc0 targ 1 lun 0: <MAXTOR, 7245-SCSI, 1357> SCSI1 0/direct fixed
sd1 at scsibus0: 234MB, 1944 cyl, 4 head, 61 sec, 512 bytes/sec
ahsc0 targ 2 lun 0: <QUANTUM, FIREBALL1080S, 1P08> SCSI2 0/direct fixed
sd2 at scsibus0: 1042MB, 3835 cyl, 4 head, 139 sec, 512 bytes/sec
ahsc0 targ 4 lun 0: <ARCHIVE, VIPER 150  21247, -005> SCSI1 1/sequential removable
st0 at scsibus0: rogue, drive empty

Problems:

a) When booting netbsd with sbic_inhibit_sync=0, the scsi bus hangs
in the busy state right after sd2 has gone sync with period 208ns
offset=8.

b) When booting netbsd with sbic_inhibit_sync=1 from AmigaOS with the disks 
running in sync mode, the scsibus hangs in the busy state right after the
"scsibus0 at ahsc0" line.

Checks I did: terminators are ok. Last time I looked, the 33c93 was jumpered
to the 14.28 MHz clock (ahsc.c assumes 143 * 100 kHz).

I operated the same disks with sync transfer mode from AmigaOS, so I know
it is possible with my hardware/terminator/cable/power supply combination.

>How-To-Repeat:
	see above.
>Fix:

The well known workaround is to always do everything async on the
SCSI bus. This gets annoying... I'd expect older disks to misbehave
(in fact, the Maxtor unexpectedly disconnects and gets reselected
by the hack in the driver code), but I'd expect an out of the box
disk no older in design than a few months to work right. Somebody
with a scsi2 spec should find out some day whats going wrong.

>Audit-Trail:
>Unformatted:
Ignatios Souvatzis