>At 9:42 PM -0800 11/10/99, Donald Lee wrote:
>>It's not so much the driver as it is a combination of the
>>way the Adaptec hardware expects mem access to work and
>>the lack of the guard bit.  The pre-G3 mac ROMs didn't envision the
>>future aggressive optimizations in the G3, so they couldn't "guard"
>>against them properly. (sorry ;-> )
>>
>>The driver is "correct" for accesses that have the guard bit set correctly.
>>(The appropriate eieio() instructions are present)
>>There is actually not much that can be done just with SW that is not pretty
>>brutal if the guard bit is not set properly on the HW ranges.
>
>So if I wanted to put my Adaptec 2940UW in the machine I would have a
>problem.  Is there any standard Apple hardware which is a problem?  Is
>there a list somewhere of known problems?

I'm not sure I understand the question.

If you're talking about putting it in a NetBSD OS, then the kernel
should be able to set the appropriate guard bits, and you should be OK.

If you're talking about the Mac OS, then the guard bits are set
by the Mac OS ROM, which does not know about them.   This is
also something that is part of the virtual memory setup and happens
early in the boot, so software in the OS (patches and the like) don't
quite do the job. (It happens with VM off too, so that won't help)

The various manufacturers took different approaches to this problem
(Newer tech, Power Logix, etc.)  Newer tech sort of "patches" the
page tables on the fly, and does not suffer from this problem.  The
other manufacturers use different approaches that appear to work less
well.  I don't know that much about the details.

Dantz (the retrospect backup folks) used to have a bunch of information
on their web site detailing all the things that don't work right with
G3 upgrade cards of various manufacturers.  I don't know if it's still there,
but it was interesting.  The Adaptec cards seemed to be most affected, but
other things broke too.  The Adaptec hardware suffers partly because it
has more functionality in the register accesses that depend on
side-effects, and are therefore more sensitive to out-of-order
memory accesses.

-dgl-