Path: utzoo!attcan!uunet!aplcen!uakari.primate.wisc.edu!zaphod.mps.ohio-state.edu!wuarchive!psuvax1!rutgers!cbmvax!daveh
From: daveh@cbmvax.commodore.com (Dave Haynie)
Newsgroups: comp.sys.amiga.hardware
Subject: Re: DiskFailure
Message-ID: <15315@cbmvax.commodore.com>
Date: 22 Oct 90 16:53:28 GMT
References: <1990Oct15.114329.18125@cbnewsi.att.com> <1990Oct15.160329.28974@cbnewsj.att.com>
Reply-To: daveh@cbmvax.commodore.com (Dave Haynie)
Organization: Commodore, West Chester, PA
Lines: 48

In article <1990Oct15.160329.28974@cbnewsj.att.com> paulm@cbnewsj.att.com (paul.j.maioriello) writes:

>I never had any problems with my 2090 in over a year of use in an A2000.  ...
>I have not been able to get this to work reliably in an A3000.

>My theory is that there is nothing wrong with the controller hardware,
>but rather with the device driver itself.  

That's right on, though probably not for the reasons you're thinking of.  The
A2090, like all Zorro II devices, is a 24 bit addressed device.  That's no 
big deal in the 3000 if it's simply a slave device, but of course the 2090 is
a bus master as well.  The A3000 has memory in various places throughout
the full 32 bit address space of the 68030.  The A2090 as bus master can only
reach a subset of that memory, that in the 24 bit space used by the 68000.
The only memory like that in the A3000 is Chip memory and any Zorro II fast
memory you add.  

The problem with this and many similar devices is that it was never considered
in the device driver that non-24 bit addressable Fast memory would be available
at boot time (remember, the A2090 development dates back to 1985 or so, when
68000s were a pretty exciting thing, 68020s just on the horizon, and 68030s
not even dreamed of yet).  So the A2090, at Binddrivers time, attempts to get a
chunk of memory, and since it figures Fast is better than Chip, it doesn't
specific the kind of memory.  Part of this memory will serve as a DMA buffer
for DMAs involving memory the disks it manages can't reach (as specificed in
the Mask entry for those disks).  Unfortunately, on an A3000, it'll get memory
it can't reach for this private buffer, and that's guaranteed to cause the
thing to fail.  What you need to do is turn off all Fast memory before
starting up the A2090 driver, and then turn it back on later.  If the drives
have a mountlist with "Mask = 0x00fffffe" or something similar, the full 32
bit memory should cause no further problems.  If you have an A2090a drive, 
you'll have to get a disk-loaded A2090 driver and remove the autoboot ROMs,
at least for the moment.

>Now I'm running 2.0.  Given all the work that has been done in AmigaDos 
>(including the exorcism of the BCPL stuff), it seems as if an update to 
>hddisk might be in order.  

I don't personally know of any real software problems beyond that stated
above.  In any case, it shouldn't be any trouble under 1.3, with the Chip
and Fast conflicts resolved, on your A3000.

>Paul M.

-- 
Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests"
   {uunet|pyramid|rutgers}!cbmvax!daveh      PLINK: hazy     BIX: hazy
	Standing on the shoulders of giants leaves me cold	-REM