Efika OF problems

[Matt Sealey]

The best way I can think of is to instantiate RTAS then copy the device
tree then get the hell out of Open Firmware. You don't have to trash it,
just stop using it after you've set up the MMU.

It's not ideal; you lose a few things in the process, but they are pretty
much nothing anyone is going to use anyway (certainly most of what can
and should be done should be done just before/after you get RTAS and
copy the device tree).

-- 
Matt Sealey <matt at genesi-usa.com>
Genesi, Manager, Developer Relations

Nathan Whitehorn wrote:
> Matt Sealey wrote:
>> There are two problems with OF calls in FreeBSD; one is that FreeBSD 
>> has no
>> support for a "real mode" Open Firmware - Apple's runs in "virtual mode"
>> which means it has all kinds of MMU stuff set up.
>>
>> Solaris had the same problem and Sun fixed it by implementing a slim 
>> "VOF"
>> on top of our real OF, which implemented the virtual mode stuff.
>>
>> There is very, very little point in addressing this issue in firmware and
>> implementing virtual addressing inside firmware, as OF client calls are
>> stupidly slow. What you need to do is get it up and running and get 
>> all the
>> OF stuff *well* out of the way *well* before any MMU stuff is set up by
>> the OS itself.
>>
>> NetBSD manages it somehow, I don't know how. You could have a chat with
>> Frank Wille.. (see the port-ofppc at netbsd.org list)
> We badly need a solution to this problem. I think the VOF route might 
> not be a bad one for machines with hard-to-use firmwares -- we support 
> runtime-switchable Open Firmware interfaces already, and we need support 
> for Linux FDTs as well.
> -Nathan