Re: Iguana and the quest for L4::BSD

On Mon May 22, 2006 at 11:04:52 -0500, R. Tyler Ballance wrote:
>Howdy, I've gotten the Kenge-hello world built and running inside of  
>QEMU on top of FreeBSD, which, after digging around through source  
>code was pretty exciting. Now, i'm at the next step. I've been  
>looking into using Iguana somewhat along the lines of how NICTA uses  
>it for the Wombat project, since I think starting with running  
>FreeBSD as a paravirtualized host on top of Iguana is a good place to  
>start
>
>(by the way, before I forget, if you don't know what this L4::BSD  
>nonsense i'm talking about is: https://lists.ira.uni-karlsruhe.de/ 
>pipermail/l4ka/2006-May/001606.html)
>
>I'm somewhat feeling around in the dark right now, I've been fiddling  
>with the big iguana-project--devel-* (53MB!) that's listed here  
>http://www.ertos.nicta.com.au/software/kenge/iguana-project/latest/   
>but I'm not sure if this is the appropriate place to start.

The 53MB mostly contains Linux.

>It seems to me that the best place to start in moving FreeBSD to run  
>on top of Iguana (on top of L4) is to get it the entire thing running  
>as a userland service (a la host OSes on Xen), i'm just not certain  
>as to where a good place to start with this is so I can start putting  
>an intermediary layer to translate some calls so the FreeBSD  
>virtualized OS "thinks" it's got direct access to hardware like it's  
>used to.
>
>Suggestions?

So what to do next really depends on what you want to achieve from this
project and what design you want to take.

Wombat is designed so that we treat l4 as a new architecture and have
basically ported Linux to this architecture. (Although probably, more
accurately we have ported to Iguana, rather than L4).

This compares to for example the afterburning work at Karlsruhe, where
you try to make minimal changes to the code base and actually end up
with a port which can be run both on raw hardware and in a virtualised
environment.

I don't really have a good idea of how to port FreeBSD to a different
achitecture, and would possibly suggest that NetBSD might be a more
suitable target for this work. (In fact a long time ago, summer 2001/02
I ported a lot of the lower levels of NetBSD to run on top L4/Mungi so
that we could get access to the SCSI drivers.)

I'm in #kenge on freenode if you want to discuss this more.

Cheers,

Benno