Re: [Xenomai-help] Some questions about Xenomai

[Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org>]

Charles Clerdan wrote:
> And a last one (for now ...), I read here  :-
> http://www.xenomai.org/documentation/branches/v2.3.x/pdf/Native-API-Tour-rev-C.pdf
> that :
> /"When a realtime
> task executes into the Linux domain on a given processor, the
> Linux kernel as a whole inherits the realtime
> priority of such task, and thus
> competes for the CPU resource by priority with other realtime
> tasks regardless of
> the domain (i.e. Linux or Xenomai) they happen to belong to. In other
> words, the
> realtime
> priority scheme enforced by the native API is consistent across domains. "/
> 
> I was thinking that Linux were a process of Xenomai with the lowest
> effective priority, and a task in secondary mode were only in
> competition with the others Linux threads (or others Xenomai thread in
> 2ndary mode). But if I'm wrong, how can the 2 schedulers share
> ressources in order to keep a hard real-time behaviour (at less for
> Xenomai tasks in 1st mode).

I am not sure I understand your question.

The two schedulers do not share ressources. What is meant by this
sentence is that the Linux kernel is a task, from Xenomai scheduler's
point of view, and this tasks inherits the priority of Xenomai tasks
running in secondary mode. The Linux scheduler is not involved.

And giving to a task running in secondary mode a priority higher than a
task running in primary mode looks like a design error anyway.

If you do not like this behaviour, you can disable it, but, on the whole
it makes things a little more natural.

-- 
					    Gilles.