RFC - Time & space partitioning with RTEMS

[Metge Jean-Jacques]

Hello Pavel,

Thank you very much for these interesting informations.

1- On a short term : 
	- I need now to analyse in depth Xtratum capabilities, in order to check if it could effectively be a good 	baseline.I had never heard about Xtratum before today.
	- I would be interested by a porting of RTEMS on top of Xtratum, in case Xtratum could be the good baseline. Which 	level of effort do you think necessary to ensure this porting ?
2- on medium/ong term, I could be interested by a participation in the EC project about the definition of a base hypervisor for applications requiring IEC61508 safety certifications. Could you, please, provide more informations about this project (eg technical requirements, consortium agreement) or the names of adequate points of contact and/or project leaders ?

Thank you by advance

Jean-Jacques

-----Message d'origine-----
De : Pavel Pisa [mailto:ppisa4lists at pikron.com] 
Envoyé : mardi 29 avril 2008 10:13
À : rtems-users at rtems.org
Cc : Metge Jean-Jacques; Sven.BIEBAUT at be.thalesgroup.com
Objet : Re: RFC - Time & space partitioning with RTEMS


Hello All,

I think, that combination of RTEMS as RT and safety control partition OS and Linux as a generic OS would be extremely interesting for many applications. The ability to run multiple safely separated instances of RTEMS could be interesting as well.

The option to use RTEMS as hypervisor and as guest OS seems interesting to me, but it would require much work and from safety and certification point of view it could be disadvantage.

May it be, that Xtratum nanokernel developed by group around University of Valencia and OpenTech (Nicholas Mc Guire) could be reasonable choice. It is fully open-source with public development. It has been started as a replacement of RTLinux GPL solution to overcome its deficiencies - mainly unmaintenable namespace collision with 2.6 kernels and FSM Labs threats of patent dispute (for IRQ redirections, which has been three from DOS days anyway). The Xtratum mainline is available for i386 now, but there exists port to PowerPC architecture at least now and other ports would be much welcome, but require more people to contribute.

There are aims to evaluate possibilities to use Xtratum as base hypervisor for applications requiring IEC 61508 safety certifications. There is Europen project proposal in preparation for building consortium of universities and companies to prove and extend this concept. The consortium is looking for one or two more companies willing to participate in this project still. Sven, you may know that already from different channels.

The Xtratum can already run Linux 2.6 and PaRTiKle (RTLinux replacement) in parallel. The adaptation of RTEMS to be yet another possible guest OS would be great. I personally like RTEMS API and well defined layered design much and I believe, that RTEMS in parallel with Linux could be interesting choice and if it could fall into scope of project and our university could get funding for such work, I would convice my colleagues and bosses to focus on such work. I hope, that I could contribute by some advises and code as time and my knowledge allows.

More informations about Xtratum can be found there

http://www.xtratum.org/

Best wishes

                Pavel Pisa
        e-mail: pisa at cmp.felk.cvut.cz
        www:    http://cmp.felk.cvut.cz/~pisa


On Tuesday 29 April 2008 09:22, Metge Jean-Jacques wrote:
> Hi Sven,
>
> To be frank, I initially had also the feeling that it would be better
> to port an existing light hypervisor (open source or not) on a LEON, 
> than trying to adapt the LEON version of RTEMS to this kind of 
> feature. In this case, and to go into your direction, RTEMS would have 
> been an OS "personality" that could run over this hypervisor, and any 
> other OS "personality" than RTEMS could also be acceptable on top of 
> this hypervisor (providing some adaptations). The problem is that it 
> really seems impossible, today, to find this kind of product either in 
> open source or in commercial domains, and that I do not have 
> sufficient time & budget to develop (or make develop) an hypervisor 
> from scratch for the LEON or to finance a port on a LEON of an 
> existing commercial hypervisor for embedded RT applications (example : 
> TRANGO, kernel layer of PikeOS, module OS of the VxWorks653 product, 
> LynxSecure, ...) . That's why I came back to RTEMS, because :
> 1- it already exists for the LEON
> 2- taking into account its open source characteristics, I thought it 
> would be maybe easier to reach a honorable result, a providing 
> reasonnable effort
>
> However, I am quite ready to change again my view, if many of us feel
> the same needs around an hypervisor for the LEON and are ready share 
> the same interest around a new type of open source product... I am 
> quite open to any constuctive proposal in this direction.
>
> Regards
>
> Jean-Jacques
>