Subject: Re: StrongARM replacement, 21 names on my list
To: Kjetil B. Thomassen <kjetil@thomassen.priv.no>
From: David Brownlee <abs@anim.dreamworks.com>
List: port-arm32
Date: 09/30/1998 14:21:51
	Continuing:
	http://mail-index.netbsd.org/mlist/port-arm32/1998/04/29/0000.html

	There are a number of possible workarounds involving adding
	a compiler option to:

	a) Avoid using LDM/STM, and load/save registers explicitly.

	b) Write all LDM/STMs as NOP then LDM/STM, and then the linker
	   can swap them as necessary to avoid an LDM/STM on a page
	   boundary.

	The first is easier, but more of a performance hit.

	My inclination would be to make option 'b)' the default, and give
	people without the bugged strongarms an easy way to turn it off,
	but the low number of problem systems may make it better to have
	it off by default and then have an additional 'workaround
	snapshot'.

	Either way - if someone is able to add the compiler switch I'm
	sure people would build usable distributions for those affected.
	
		David/absolute

           -=-  "If there is a hell... I'll see you there"  -=-

On Wed, 30 Sep 1998, Kjetil B. Thomassen wrote:

> On Wed 30 Sep, Laurent Domisse wrote:
> > Kjetil B. Thomassen wrote:
> > > 
> > > I have done a new count of the names on my list of people who may
> > > be interested in getting their SA-110 J/K replaced with an S/T.
> > > I now have 20 names on my list.
> > > 
> > 
> > could you explain in the mailing list the difference between the
> > J/K and S/T release ?
> 
> OK, I'm sorry about that, but before I go into that, it seems as though
> this problem is only hit by users of Acorn Risc PCs.
> 
> The difference is that the J and K versions of the SA-110 has a bug
> that produces all kinds of weird problems. One of them is that cc
> gives an internal error for no apparent reason.
> 
> I don't remember the bug in detail, but that was described in this mailing
> list some time ago. You can probably find the articles in the mailing
> list archive, or by using Olly's search engine at:
> http://noxious.muscat.co.uk/riscbsd/
> 
> The point is that this bug has been fixed in the S and T versions of
> the SA-110.
> 
> You can figure out if you have a buggy version by doing one of the
> following:
> 1) Check the physical chip and see which letter is written after
>    SA-110.
> 2) Boot a recent kernel and see if it claims that you have a buggy
>    SA-110 in the boot message.
> 
> At present there is no workaround in NetBSD for this problem, and
> none is planned for the forseeable future. The only option therefore
> is to replace the chip or use the old ARM610 or ARM710 boards in
> the Risc PC.
> 
> The 233 MHz SA-110s from Acorn seems to be at least version S.
> 
> I hope this clears up the confusion.
> 
> Kjetil B.
> mailto:kjetil@thomassen.priv.no
> http://home.eunet.no/~kjetilbt/
>