[Xenomai-help] Ipipe support for 'longterm' kernel versions

[Xenomai-help] Ipipe support for 'longterm' kernel versions

[Henri Roosen]

Hi all,

I was pleased to read on the Linux Kernel Mailing List Greg's message
about Linux stable kernel release procedure changes, that the longterm
supported stable kernels will be made more explicit by getting the
'longterm' name.

https://lkml.org/lkml/2010/12/2/388

We are embedding Linux in products and therefore those longterm
kernels are very valuable for us. We chose 2.6.32 kernel because of
that a while ago.

However, unfortunately there is no Ipipe patch that applies cleanly on
the latest 2.6.32.26 stable, nor is there an ARM version.

It would be great if Xenomai would also recognize this need and would
do some kind of 'long term' support for the Ipipe for the longterm
kernels. I understand maintaining the Ipipe is a lot of work.
Personally I think it would be an option to only support the longterm
kernels...

What do the Xenomai/Ipipe developers think?

Thanks,
Henri.

Re: [Xenomai-help] Ipipe support for 'longterm' kernel versions

[Jan Kiszka]

Am 07.12.2010 15:47, Henri Roosen wrote:
> Hi all,
> 
> I was pleased to read on the Linux Kernel Mailing List Greg's message
> about Linux stable kernel release procedure changes, that the longterm
> supported stable kernels will be made more explicit by getting the
> 'longterm' name.
> 
> https://lkml.org/lkml/2010/12/2/388
> 
> We are embedding Linux in products and therefore those longterm
> kernels are very valuable for us. We chose 2.6.32 kernel because of
> that a while ago.
> 
> However, unfortunately there is no Ipipe patch that applies cleanly on
> the latest 2.6.32.26 stable, nor is there an ARM version.
> 
> It would be great if Xenomai would also recognize this need and would
> do some kind of 'long term' support for the Ipipe for the longterm
> kernels. I understand maintaining the Ipipe is a lot of work.
> Personally I think it would be an option to only support the longterm
> kernels...
> 
> What do the Xenomai/Ipipe developers think?

Of course, supporting kernel releases over a longer period is a good
thing. However, it takes resources to do this job, to backport fixes or
even features from latest I-pipe versions and to resolve patch conflicts
over older kernels. If you are willing to support our community, you are
always welcome!

How this new (actually it is not that new) longterm strategy of the
Linux kernel will influence I-pipe tip development is not yet well
predictable. Right now I would say the need for certain features only
available in newer kernels will continue to drive updates of I-pipe, not
that some kernel is declared longterm stable. But this also depends on
when the longterm tags will be announced, ahead of releases or much later.

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux

Re: [Xenomai-help] Ipipe support for 'longterm' kernel versions

[Gilles Chanteperdrix]

Henri Roosen wrote:
> Hi all,
> 
> I was pleased to read on the Linux Kernel Mailing List Greg's message
> about Linux stable kernel release procedure changes, that the longterm
> supported stable kernels will be made more explicit by getting the
> 'longterm' name.
> 
> https://lkml.org/lkml/2010/12/2/388
> 
> We are embedding Linux in products and therefore those longterm
> kernels are very valuable for us. We chose 2.6.32 kernel because of
> that a while ago.
> 
> However, unfortunately there is no Ipipe patch that applies cleanly on
> the latest 2.6.32.26 stable, nor is there an ARM version.
> 
> It would be great if Xenomai would also recognize this need and would
> do some kind of 'long term' support for the Ipipe for the longterm
> kernels. I understand maintaining the Ipipe is a lot of work.
> Personally I think it would be an option to only support the longterm
> kernels...
> 
> What do the Xenomai/Ipipe developers think?

Well, 2.6.35 is also a long term version. Maintaining all the long term
versions is a lot of work (it takes time to simply validate a patch for
every new release). So, we probably can not maintain support for all the
long term versions, we would probably have to pick one.

Since the I-pipe patch for ARM skips even versions (in order to reduce
the time spent issuing I-pipe patches), and since it is more recent, I
would rather go for 2.6.35.

-- 
                                                                Gilles.

Re: [Xenomai-help] Ipipe support for 'longterm' kernel versions

[Philippe Gerum]

On Tue, 2010-12-07 at 15:47 +0100, Henri Roosen wrote:
> Hi all,
> 
> I was pleased to read on the Linux Kernel Mailing List Greg's message
> about Linux stable kernel release procedure changes, that the longterm
> supported stable kernels will be made more explicit by getting the
> 'longterm' name.
> 
> https://lkml.org/lkml/2010/12/2/388
> 
> We are embedding Linux in products and therefore those longterm
> kernels are very valuable for us. We chose 2.6.32 kernel because of
> that a while ago.
> 
> However, unfortunately there is no Ipipe patch that applies cleanly on
> the latest 2.6.32.26 stable, nor is there an ARM version.
> 
> It would be great if Xenomai would also recognize this need and would
> do some kind of 'long term' support for the Ipipe for the longterm
> kernels.
>  I understand maintaining the Ipipe is a lot of work.
> Personally I think it would be an option to only support the longterm
> kernels...
> 
> What do the Xenomai/Ipipe developers think?

Only supporting the latest stable release would likely mean stalling the
support for new boards merged over time in regular releases, until a new
stable kernel is picked up -- our arm support would have a real problem
with that, and powerpc as well. Since Xenomai is primarily about
embedded configurations, this would not be acceptable.

Besides, what is deemed stable for the mainline kernel may not always be
the best option when considered from any given architecture. E.g,
picking 2.6.32 for running ARM mpcore is planning for headaches, albeit
it may be fine for others.

We understand the need for stability, and as a matter of fact, we
already have a stable branch for Xenomai releases exactly for this
purpose. But the task of tracking stable kernels over a long period of
time should really be tackled downstream. We could help for sure, but
this is not something we would consider as being on our plate.

> 
> Thanks,
> Henri.
> 
> _______________________________________________
> Xenomai-help mailing list
> Xenomai-help@domain.hid
> https://mail.gna.org/listinfo/xenomai-help

-- 
Philippe.