Re: L4::BSD, Subclass your OS

As I could understand, you want to do something similar to L4Linux but using
BSD kernel instead of the Linux kernel and programming the drivers are as
plugins on top of the L4 microkernel, doesn't you?

A lot of things are not enough clear to me yet, by example, how the shared
libraries are handled? they are loaded by every module? is there some
"library manager"? or should be a library manager implemented as module? if
it is so, the routine calls are done via IPC too?




Ernesto






On 11/05/06, R. Tyler Ballance <[email protected]> wrote:
>
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> - -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> - - -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Well, let me first introduce myself. I'm R. Tyler Ballance, I'm a
> Computer Engineering student at Texas A&M University. I've been using
> BSD (Free/Net/Open/Mir/ekko) since I bought a G3 iBook 4 or 5 years
> ago, and was introduced to open source. I participated in the Google
> Summer of Code last year, and started work on porting Apple's launchd
> utility to FreeBSD, according to Google, I succeeded in my project,
> but it's terribly unfinished, and it's nothing i'm proud of. I now
> run my own software and consulting company to pay the bills, but my
> interest in getting a very Darwin-esque operating system running on
> top of the L4 microkernel is still quite strong.
>
> Thusly, I think starting  project, tentatively named L4::BSD, would
> be a magnificent way to really introduce myself into the inner
> workings of FreeBSD. A project like this has a few wonderful
> possibilities. The most enticing, in my opinion, of which, could be
> platform agnostic device drivers. Since the operating system would
> run in user-space on top of the L4 microkernel, as well as device
> drivers (hopefully I'm not misunderstanding the concept), I believe
> it would be possible to merely define an interface for interacting
> with the driver, which would then run on top of the L4 kernel (in
> user space) regardless of what the main operating system (L4Linux,
> L4:BSD) would be. The basic idea would be that a developer could
> write a driver for his specific network card, let's call it an
> UnrealTech card, which adheres to some L4::Network standard for
> network devices, then in the host operating system, if a the host OS
> can "speak" this L4::Network standard, it could easily use this card,
> as well as any other operating system running on top of L4, that
> spoke the L4::Network standard. (it's a pretty lofty idea, but in my
> opinion, would be an excellent use of the microkernel abstraction)
> Developing something like this would be along the lines of Apple's
> IOKit subsystem, which provides a wonderful interface for dealing
> with hardware through the Mach microkernel.
>
> jkoshy@ of the FreeBSD project also pointed out to me on the freebsd-
> hackers@ list that there would also be the possibility of "migrating
> VMs between machines
> 'on-the-fly', as described in the following paper: http://
> www.cl.cam.ac.uk/Research/SRG/netos/papers/2005-migration-nsdi-
> pre.pdf" Which would be a fantastic use of a microkernel. If i
> remember correctly, Xen uses a microkernel approach to running host
> operating systems.
>
> Presumably we'd use L4::Pistachio, which I believe is a wonderful
> microkernel to start with. Since L4 is largely an academic
> microkernel (as is Mach, and that didn't stop Apple ;)), a lot of
> documentation, white papers, test code, and proof-of-concepts are out
> there. TU Dresden (which i've considered attending) has plenty of
> great resources as well as some of their Fiasco microkernel work all
> posted here: http://os.inf.tu-dresden.de/
>
>
> I'm still researching, I took my last final exam yesterday, so I'm
> looking into and gathering information on the best way to start
> working on this project. I've also registered the channel #l4bsd on
> the Freenode network (irc.freenode.net), since i'm an avid user of
> IRC during the work day :)  If you guys would like, I can setup a
> berlios or sourceforge project for this? Let me know.
>
> I'm going to go ahead and forward this message to the l4ka@ mailing
> list, so I'm certain they'll point out any complete flaws in my
> logic, or otherwise.
>
> So, suggestions, ideas, flames, comments, etc, are all welcome :)
>
>
> Cheers,
>
> - - - -R. Tyler Ballance
> Prospective kernel hacker at large ;)
>
>
> - - -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.3 (Darwin)
>
> iD8DBQFEYxw1qO6nEJfroRsRAmZ/AJ48ENBgF1fD10tjazshJGfO/IB1twCdF9Tl
> Wt7lSdsUMgV0XTpgZvq3KWw=
> =v4FR
> - - -----END PGP SIGNATURE-----
> - -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.3 (Darwin)
>
> iD8DBQFEYxw9qO6nEJfroRsRApmQAKCH6RXmcqYjF+Iqs18KOC1cd7C/vQCcDRO9
> ibh+tHt/B0VJLHUEHIjmFlk=
> =xWaF
> - -----END PGP SIGNATURE-----
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.3 (Darwin)
>
> iD8DBQFEYxxJqO6nEJfroRsRAnqeAJ97NgEhoKrrUCNDBtq9+7jpyCk0cACeKaUa
> eqgW27WXPnjkwR9a5ClLo9g=
> =bR3X
> -----END PGP SIGNATURE-----
>