[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Janos on Moab on Linux - Node OSKit



> Hello, all.  I have been away on TDY for a while, but
> I'm back now, and breaking Janos with my usual alacrity.
> 
> I am currently working on the JanosVM on top of the
> JNodeOS on top of Moab on top of Linux. That is, I
> want a Moab node OS directly on Linux, without any
> OSKit. At least I THINK that is what I read was possible.

This configuration "should" work, although we usually just run JanosVM on
top of unix or oskit/moab/janosvm on a bare machine...

> The purpose of this is to get a Moab-based active 
> router going on top of Linux, rather than on a standalone
> PC. I wish to explore the use of LKMs to provide
> the level of resource control one would like. I also
> want to explore separating out the packet forwarding
> functions from routing updates and other infrequent
> operations. This is all with an eye to implementing this
> on an NP with a separate control processor.

sounds spiffy

> Now, I have compiled Moab and the J NodeOS with no
> insurmountable problems. I have also compiled Kaffe
> on Linux. But, when I try to compile the JanosVM
> it bletches. It complains that no proper OSKit 
> compiler can be found, which is no surprise since
> I am trying to avoid using the OSKit.

We haven't done this before, but i'll try and wing it here:

you'll need to do a regular configure, not the nodeos-configure with the
following options:

  --with-threads=nodeos-threads --host=i386-nodeos --without-x
  --with-staticlib

Basically, you need to arrange for JanosVM to use nodeos threads, set the
correct host environment which will cause the library in
libraries/extensions/nodeos/clib to be compiled, get rid of x
dependencies, and link the whole thing statically.  And, you'll still need
to set the JNODEOS_CLASSES_ZIP environment variable.

However, for now, i would advise you to do the lkm work and just use
mini test EEs directly on top of moab.  Getting that stuff working will be
the major stumbling block.  The janosvm is just pretty decoration
otherwise...

> Pete Lutz (RIT)

tim stack




[ Janos ] [ OSKit ] [ Network Testbed ] [ Flick ] [ Fluke ]
Flux Research Group / Department of Computer Science / University of Utah