On Tue, 11 Jan 2000, David A. Gatwood wrote:

> On Tue, 11 Jan 2000, Bill Studenmund wrote:
> 
> > The problem with "emulating" the OF ROMs is that NetBSD eventually takes
> > over all the RAM, and so it'd be tricky to keep this pseudo OF around.
> 
> Not really.  I'd imagine you already make a copy of OF for internal use
> based on the tree, which is located at a static location.  The only issue
> would be passing in a pointer to the location of the custom tree instead
> of using the "real" one that isn't there.  Just put the temporary data
> right after the kernel along with whatever boot parameters you pass in.

Right now we don't copy the OF tree, we talk to OF as needed. Like getting
nodes by name, getting peers of nodes, descendants, etc. All this could be
done from a static tree, but we're not set up for it.

Also, we use OF's standard out and standard in until our drivers come
along take over the console.

> > > On reflection the MacOS psuedo-OpenFirmware booter looks the most elegant
> > > solution, but adding support the 'hard' way to mklinux looks like the
> > > quickest solution. 
> > 
> > Well, we like to keep talking to OF once we've booted, more so than Linux
> > does I think. That might be a problem.
> 
> Define "talking to".  Do you mean reading from the tree or actually doing
> OF calls like for OF-based drivers?

Yes, see above. though I think we only do it with stdout and stdin. But we
were just discussing today the idea of using RTAS for pci config
operations and for nvram and beeping. That would mean we'd need OF much
longer.

Take care,

Bill