Subject: MMUs mostly
To: None <port-mac68k@netbsd.org>
From: Andy Ball <ball@cyberspace.org>
List: port-mac68k
Date: 11/21/2001 08:25:58
Hello Todd!

  TV> NetBSD requires booting from MacOS so as to initialize
    > things like the NuBus getting all machines to
    > initialize properly (...whereas having a 5-8MB MacOS
    > boot partition with a System file is guaranteed to get
    > it Right).

I can see how that would work, and thankfully the Mac OS is
a lot smaller than some others.  What should I use to
partition the drive?

  TV> The 68030 has a 68851-compatible (well, almost, it's
    > actually more like a somewhat stripped down 68851)
    > PMMU built onto the chip...

Interesting.  In what way is the internal PMMU stripped down
when compared with the 68851?  How well suited would you say
the 680x0 MMUs were to running BSD?

I don't know if you're familiar with the SUN 3 range of
machines.  They were mostly 68020 based, but some later
models used a 68030, but used what I think was a proprietary
MMU rather than the one in the 030.  I daresay this was for
compatability with their existing system software (BSD-based
at the time I think). I'm wandering off-topic perhaps, but I
am intrigued as to what makes a good MMU for a BSD box.

  TV> The *hardware* ignores the on-board 68851 if one is
    > present -- because there's one internal to the 68030.

The hardware in the microprocessor would ignore the external
one?  I think Eric suggested in the case of the Mac II I was
discussing, that the 68851 couldn't physically be installed
at the same time as the 68030 upgrade, but it's an
interesting (to me) academic question.  Perhaps it's time I
swotted up on how the 68851 interfaces to the 68020.

  TV> Moto bundled PMMU logic inside the CPU with the 68030
    > and up (...although there are some MMUless '040s, aka
    > 68EC040, that have pissed off Amiga users for ages :)

...whereas Mac users were being pissed off by the LC
variants? :-)

Regards,
  - Andy Ball.