Subject: Re: Problems with A530 and -current
To: Jeffrey Ray Thieleke <email@example.com>
From: Michael L. Hitch <firstname.lastname@example.org>
Date: 04/16/1995 10:31:04
On Apr 15, 8:03pm, Jeffrey Ray Thieleke wrote:
> probe(gtsc0:1:0): non-media hardware failure, data = 00 00 00 00 43 00 00
> 00 00 00
> gtsc0 targ 1 lun 0: (QUANTUM, ELS170S, 4.07) SCSI2 0/direct fixed
> sd1 at scsibus0: 163 MB, 1536 cyl, 4 head, 54 sec, 512 bytes/sec
> * The biggest difference between 1.0 and -current's startup diagnostics is
> the "gtsc0: target 0 now synchronous" and "probe(gtsc0:1:0): non-media
> hardware failure" parts.
I think 1.0 had synchronous transfer disabled by default, but -current
now defaults to enabled. The "non-media hardware failure" error is
probably coming from the attempt to negotiate synchronous transfers on
sd1. The error data seems to indicate a "Message Error". I think I've
seen that once in a while with my GVP adapter, but I'm not sure what
> * I don't know if this is due to the "non-media hardware failure" or a
> difference between 1.0 and -current, but when I attempted to mount my
> AmigaDOS partitions, it fails, giving me a 'read only media' error.
It's a difference between 1.0 and -current. The adosfs now checks and
verifies the mount more strictly, and since it's currently a read only
filesystem, it requires that the mount specify the read-only option. The
1.0 version of mount_ados does not do this, so an updated mount_ados
is required and you have to specify that the mount is read-only.
> * When I reboot (Control-LAmiga-RAmiga), my A530 appears to be locked up.
> The drive access light stays on, but nothing happens. It takes an
> actual physical power down to fix things.
I don't know what might be causing this. One possibility might be that
the disk driver is still doing synchronous transfers, but the GVP driver
isn't. The Amiga reset should be doing a reset on the SCSI bus, which
should be causing the drive to reset and do asynchronous transfers.
> I wonder if the problem is due to -current trying to set sd0 to be
> accesses synchronously. Both the sd0 and sd1 have the sync option turned
> off in their RDB, but -current still says that "target 0 now synchronous".
NetBSD doesn't look at the synchronous transfer option in the RDB
[perhaps it should?].
> Perhaps my A530 SCSI controller can not operate synchronously, and when
> it is set to do so, it locks up (although this only happens on reboot).
If you can access sd0 ("dd if=/dev/rsd0c if=/dev/null count=100"),
then the NetBSD driver should be doing synchronous transfers. If the
drive does not get a SCSI bus reset when the Amiga is reset, that might
cause problems, because the GVP driver would be trying to do asynchronous
transfers while the drive was still doing synchronous.
> Is there some why to make -current access the drives only asynchronously,
> or does someone know what my lockup problem could be?
Binpatch _sbic_inhibit_sync to a non-zero value (or change the source
in sbic.c). That will disable the negotation for synchronous transfers.
Michael L. Hitch INTERNET: email@example.com
Office of Systems and Computing Services
Montana State University Bozeman, MT USA