On Wed, 06 Nov 1996 16:56:18 +1000,
Paul Ripke rearranged the electrons to say:
>Tony wrote:
>> 
>> I'm trying to get a Daystar 040 accelerator card to work in a IIci.
>
>I have a Daystar 040 that I've been trying to get working in my IIsi.
>I must admit I haven't tried too much, I've gone as far as building
>a few modified kernels and trying to work out the 040 MMU page table.
>The problem on the IIsi (it may be the same on the IIci - anyone?),
>is that the physical memory is not contiguous, and must be mapped
>with the MMU. NetBSD on 68030 machines reads the page table set up
>by MacOS when it starts, and uses that as a base for its page table.

Well, on boot mrg returns "Mac IIci/Quadra 700 ROM", I'm not sure about
the non-contiguous memory.  Does the Q700 handle memory in the same
way as the IIci?  There are only 4 RAM SIMM slots in a Q700, and
I'm unsure as to how it implements VRAM.

>However, there is no routine to do this on the 68040 class of
>machines (at least there wasn't pre 1.2 - has anything changed?),
>and NetBSD assumes it's contiguous. When the MMU is disabled, it
>crashes.

What kind of symptoms do you get?  We're able to boot single-user,
and issue several commands before the kernel locks up (rarely does it
actually drop in to the debugger!)  Commands like 'w' and 'ps'
give "illegal instruction error", thus the question about FPSP.
Multi-user boots always bomb, but get into the execution of the rc
scripts.

What kind of changes have you been experimenting with?  I'm no
expert when it comes to modifying the kernel code, but I'm willing to
help.  Perhaps we can work together?

Sean

__________________________________________________________________________
Sean Sweda                        :            Owner, Motor City Marauders
sweda@us.itd.umich.edu            :   1994 Internet Baseball League Champs
http://www.umich.edu/~sweda       :                     http://www.ibl.org
__________________________________________________________________________