Subject: Re: Synchronous scsi: Not handled well
To: RiscBSD Mailing List <port-arm32@NetBSD.ORG>
From: Robert Black <r.black@ic.ac.uk>
List: port-arm32
Date: 05/27/1997 15:26:38
On May 24,  2:24pm, Kjetil B. Thomassen wrote:
> Subject: Synchronous scsi: Not handled well
> I have an Acorn SCSI-1 card as delivered with an Acorn R140.
> When I connected an OEM (Apple) Quantum Maverick 540S to this
> card and booted RiscBSD, it set this up with synchronous transfers.
>
> Then when booting, the computer would hang at different points
> in the boot process, both with auto-build kernels and 4871.
>
> I moved the same drive to the Power-tec SCSI-2, and I haven't
> had any problems since.
>
> This indicates that RiscBSD may have problems with synchronous
> SCSI on my old 600 with ARM610.
>
> Has anyone else experienced the same problems?
>
> I believe that the reason for using synchronous SCSI with this
> combination of drive and host adapter is because that will give
> the best performance. The Acorn card is limited to around 1.5 MBytes/s
> with asynchronous SCSI, whereas it can handle 4 MBytes/s with
> synchronous SCSI. I was therefore delighted that synchronous
> SCSI was chosen with this combination since this would give a
> higher transfer rate than asynchronous.

Right, in order to get these transfer rates the on-card DMA will have to be
implemented. There are quite a few bugs (apparently) in the Acorn driver and I
have spent most of the weekend trying to trace some of them (for some reason
trying to talk to Jaz can hang/panic the machine or generate spurious error
messages - this is an Acorn SCSI-1 specific problem). I really hope this
doesn't mean I'm going to have to learn the SCSI spec by heart...

Cheers

Rob

PS And now some good news on the console front. I have a virtual terminal
working (as of 8am this morning). The significance of this is that the virtual
terminal driver is the one the user will actually see as opposed to the
'console on the graphics device' hack that people at Wakefield got to see.