Montavista realtime scheduler patch

Hi. Is somebody there using this scheduler or has tried it? I couldn't 
find much information o it other than MontaVista's promotional material 
which does not help much.

After a quick glance it seems rather straightforward and not much 
intrusive and besides it's architecture independent. Though the latest 
version available in sourceforge is for kernel 2.4.16 I have been able 
to apply it even to kernel 2.4.25 though I still haven't compiled and 
tried it. Still I am wondering why it hasn't proceeded as its proposal 
looks interesting. Could that be due to O(1) scheduler and 
preemptibility to have make into 2.6 series?

Any insights about its performance and/or gotchas are very welcome. My 
realtime needs are not too stringent (a couple hundreds of microsseconds 
for interrupt latency and perhaps one milissecond for thread scheduling 
after an event) so I am after an adequate solution.

TIA.

Elder.

>Still I am wondering why it hasn't proceeded as its proposal 
>looks interesting. Could that be due to O(1) scheduler and 
>preemptibility to have make into 2.6 series?

I don't know much about all this but it seems that as time goes on, making an
embedded Linux out of the main Linux will become more difficult. Mainstream
Linux has its eyes looking upwards towards bigger systems and will just mean
more for embedded folk to cut out. Actually, you can substitute just about any
mainstream OS name in there.

>Could that be due to O(1) scheduler and
> preemptibility to have make into 2.6 series?
> 

AFAIK, besides other improvements regarding soft real-time performance,
the ideas first introduced by the Montavista realtime scheduler patch
that were continued by community members later are included in Kernel
2.6.

-Michael

"Gary Kato" <garykato@aol.com> wrote in message
news:20040313104028.06694.00001164@mb-m05.aol.com...
> >Still I am wondering why it hasn't proceeded as its proposal
> >looks interesting. Could that be due to O(1) scheduler and
> >preemptibility to have make into 2.6 series?
>
> I don't know much about all this but it seems that as time goes on, making
an
> embedded Linux out of the main Linux will become more difficult.
Mainstream
> Linux has its eyes looking upwards towards bigger systems and will just
mean
> more for embedded folk to cut out. Actually, you can substitute just about
any
> mainstream OS name in there.
>

I haven't done any real work with embedded linux, but I'm planning to (when
I get the time :-).  From what I have read over the past few years, getting
embedded Linux systems running is getting easier.  While the Linux kernel
has support for more and more hardware, it is also becoming increasingly
modular, and 2.6 has had many enhancements aimed at small systems as well as
at big systems.  In particular, it has absorded the ucLinux patches, so that
you can build a main tree linux kernel for non-MMU cpus.  Additionally, the
O(1) schedular and pre-emptive kernel have removed the need for special
schedulers and RT enhancements for many embedded linux uses (though not for
all uses - 2.6 has much faster and more consistent responses to events, but
it is still far from hard real-time).

I'm sure others with more real experiance, rather than just second-hand
knowledge, can comment.

>From what I have read over the past few years, getting
>embedded Linux systems running is getting easier.  While the Linux kernel
>has support for more and more hardware, it is also becoming increasingly
>modular, and 2.6 has had many enhancements aimed at small systems as well as
>at big systems.  In particular, it has absorded the ucLinux patches, so that
>you can build a main tree linux kernel for non-MMU cpus.  Additionally, the
>O(1) schedular and pre-emptive kernel have removed the need for special
>schedulers and RT enhancements for many embedded linux uses (though not for
>all uses - 2.6 has much faster and more consistent responses to events, but
>it is still far from hard real-time).

Sounds promising. Thanks for the information. I'm still learning about Linux. I
have the 2.6 source on my PC here but I haven't looked at it yet. I used to be
a sys admin on Unix and SunOS systems a long time ago and things seemed to have
changed since then, not to mention some of those memories have faded. I think
my first big Linux task is getting Linux 2.6 running on my old 486.
 

On 13 Mar 2004 15:40:28 GMT, garykato@aol.com (Gary Kato) wrote:

>I don't know much about all this but it seems that as time goes on, making an
>embedded Linux out of the main Linux will become more difficult. Mainstream
>Linux has its eyes looking upwards towards bigger systems and will just mean
>more for embedded folk to cut out. Actually, you can substitute just about any
>mainstream OS name in there.

Oh, I don't think so at all.  If you read Kernel Traffic, you'll see
that there's definitely embedded people that do get their concerns
heard and subsystems integrated, and Linus does listen to them
closely.

There are entire systems in there now that are very cool for embedded
use -- MTD among them.  Time to make a battery-backed RAM driver for
my specific hardware that I can put arbitrary filesystems on?  15
minutes and 4 functions to implement.

Great stuff.
-- 
Alex Pavloff - remove BLAH to email
Software Engineer, ESA Technology