Re: public interest in cyberscsi-kernel

[is%beverly.rhein.de@localhost (Ignatios Souvatzis)]

Hi Sigurbjorn, Thomas,

> In "port-amiga%netbsd.org@localhost$s", public interest in cyberscsi-kernel$s 
> wrote:
> 
> > In this mailinglist where some requests for a cyberscsi kernel for NetBSD.
> > Somebody startet to write a driver for this kind of scsi controller. There

It sometimes sounds as if people forget that this is a volunteer effort.
This might well be a misunderstanding, but:

If you paid me full time, and gave me a contract over a lot of years (my
current one is unlimited), I could work on demand. Note that you also should
account for expenses, eg., purchashing a sample of every Amiga model,
and (at least) every Phase5 accellerator board and scsi option.

As this is not the case, (and note that I dont really demand this) we
have to get by with what remains after (e.g.) my job, my PhD research
project, dishwashing, shopping, my girlfriend are satisfied.

Same applies to any other people who contribute to NetBSD-Amiga. 

This said:

> > even exist alpha (beta?) version, but when someone aks for sources or the
> > current development status I couldn't read any answers.
> > 
> > I think there is a public interest and I'd like to know more about a
> > cyberscsi kernel (sources?, further under development?, ...)
> 
> I'd certainly be interested in seeing a working cyberscsi mk2 kernel at 
> least. 
> I can't use my old (and yes,I know it sucks) gvpscsi controller since it
> doesn't function with the mk2, and I had planned to upgrade to a real hard
> disk if I could find a mk2 cyberscsi kernel.  At the moment I'm stuck with no
> scsi and using the internal ide, which is not ideal.  I'd appreciate any sort
> of response.

a) all Phase5 boards are using a variant of the FAS 216 & friends, or
NCR53C9x, (or now Symbios 53C9x ?) chip family, and a custom DMA engine.

The original driver for this was in a file calles "sfas.c", with dma
stuff for fastlane in flsc.c and for blizzard II in blsc.c.

The original driver worked in the authors machine, with the authors mixture
of disk(s) (and tape(s)), but not so well anywhere else. He hasnt any
more time to work on this, and what he contributed in the past, fixed
his problems, but not the ones some other people had.

There are two problems with this driver setup:

a) the Fastlane Z3 DMA engine is difficult to handle. The original
driver seems to fail to do this correctly.

b) the sfas.c, which handles the scsi chip core, doesn't handle all
situations correctly.

You might be aware that I blind-programmed (well, I had docs about the
DMA, but no machine to test myself) dma frontends for cbyerscsi 1, blizzard
2060 scsi and blizzard 1260 scsi, which don't work very well... These are
the "experimental kernels" mentioned. Seeing that they don't work very
well, I hesitate to blindcoding another dma frontend, which surely will
see the same problems, as b) applies to all of them.

I don't have the time and the hardware to really fix this. However, the
problem is most probably the sfas.c, which can be replaced now by
something else:

The NetBSD/Alpha - NetBSD/Sparc driver for this chip family, which was
originally a monolithic file, conditionally compiled for the one or
another, has been split into a machine independent chip driver core and
the respective DMA and bus attachment stuff.

We can now start to write frontends for this. I am aware (and thats what
I mentioned, and what is vaguely cited above) that a certain
NetBSD/Amiga developer who knows how to do such things, has started to
write a Fastlane frontend for this. This is why I don't blindcode any of
the others... I'm awaiting his progress report. And knowing how
disturbing and useless frequent questions about the status of volunteer
projects is, I dont ask him daily or even weekly about the progress, and
I dont tell you who he is so that you wont ask him daily, or even weekly
or monthly _each_. 

If you want to help:

If you're running NetBSD-current on your IDE, and think you know how
to write a frontend to the sys/dev/ic/esp.c driver, I could make the DMA
docs available to you. But you would be on your very own; there is no
example driver which interfaces esp.c to the Amiga busses yet. Well,
other than me, you would at least have a machine to test it on.

If you don't think you can do that, you might consider to offer your
board for a couple of weeks or a few months to somebody who would like
to try it. Note that (judging from my situation) transport risks would be on 
your side, and costs well... at least partially. Most people cant make such
an offer, because they use their board daily with AmigaOS.

If you can't help this way, I'm afraid you have to wait, and in the
meantime (at least the A1200/4000 people out there) run NetBSD on your
IDE disk.

Regards,
        -is