Forward port of latest RT patch (2.6.21.5-rt20) to 2.6.22 available

Hi Ingo, Thomas, and the greater linux-rt community,
  
	I just wanted to let you guys know that our team has a port of the
21.5-rt20 patch for the 2.6.22 kernel available.  It in no way should be
construed as a substitute for the forthcoming official release from the
RT community.  We are submitting it now in hopes it will be useful to
someone in the interim, even if only for reference for the official
porting effort (since we have resolved numerous issues that were
introduced in the new vanilla code since 21 already).  We hope to rebase
to Ingo's offical patch once it is available.

	As a caveat, I'm sure there are numerous errors which were introduced
by a combination of unnoticed patch fuzz, human error, etc.  But we do
have it building and running pretty reliably on a number of 2-way,
4-way, and 8-way x86_64 SMP systems here in our lab with various
configurations.  It yields very impressive cyclic-test results with our
reference PREEMPT_RT config (see attached) so we know that generally the
port went ok.

	There is some fallout that occurred related to some slob changes
between rc1-rc2 that we could probably use some feedback on.  I have
attached a file (problem-patches.txt) that lists the names/commit-ids of
some of the patches that we know to be problematic from bisecting.
There is also a cheesy hack in fs/buffer.c that is probably just plain
wrong, but we are still trying to find a good way to fix an issue with
the IDE driver.  

	Note also that we find that the resulting kernel does not boot on some
T60p thinkpads without "acpi=off".  We are still investigating the root
cause here.

You can find the patch available here:
ftp://ftp.novell.com/dev/patch-2.6.22-rt1

Best Regards,
-Greg

------
Gregory Haskins [off-list ref]
Linux Solutions Group
Novell

On Wed, 2007-07-11 at 20:58 -0400, Gregory Haskins wrote:

Hi Ingo, Thomas, and the greater linux-rt community,
  
	I just wanted to let you guys know that our team has a port of the
21.5-rt20 patch for the 2.6.22 kernel available.  It in no way should be
construed as a substitute for the forthcoming official release from the
RT community.  We are submitting it now in hopes it will be useful to
someone in the interim, even if only for reference for the official
porting effort (since we have resolved numerous issues that were
introduced in the new vanilla code since 21 already).  We hope to rebase
to Ingo's offical patch once it is available.

	As a caveat, I'm sure there are numerous errors which were introduced
by a combination of unnoticed patch fuzz, human error, etc.  But we do
have it building and running pretty reliably on a number of 2-way,
4-way, and 8-way x86_64 SMP systems here in our lab with various
configurations.  It yields very impressive cyclic-test results with our
reference PREEMPT_RT config (see attached) so we know that generally the
port went ok.

	There is some fallout that occurred related to some slob changes
between rc1-rc2 that we could probably use some feedback on.  I have
attached a file (problem-patches.txt) that lists the names/commit-ids of
some of the patches that we know to be problematic from bisecting.
There is also a cheesy hack in fs/buffer.c that is probably just plain
wrong, but we are still trying to find a good way to fix an issue with
the IDE driver.  

	Note also that we find that the resulting kernel does not boot on some
T60p thinkpads without "acpi=off".  We are still investigating the root
cause here.

You can find the patch available here:
ftp://ftp.novell.com/dev/patch-2.6.22-rt1

Thanks for the effort!

Couple of comments tho .. You might want to change the name of the patch
to distinguish it from Ingo's patch ..

Also it would be nice to move away from the "One Big Patch" release
method , for instance if you could break it up into two or more patches
one with the stuff from the patch-2.6.21.5-rt20 which applied, and
another patch with the reject resolution. That way we can see what was
added ..

Daniel

On Wed, 2007-07-11 at 20:58 -0400, Gregory Haskins wrote:

	Note also that we find that the resulting kernel does not boot on some
T60p thinkpads without "acpi=off".  We are still investigating the root
cause here.

Try "nohpet" for now. I'm investigating it on the -hrt side as well.

	tglx

Hi All,

Sorry for my poor english.

On Wed, 2007-07-11 at 20:58 -0400, Gregory Haskins wrote:
[...]

	I just wanted to let you guys know that our team has a port of the
21.5-rt20 patch for the 2.6.22 kernel available.

Great! Thanks a lot for effort!

I'm doing the same thing and i don't have doubt your work is better than
mine. In any case i have a couple of suggestions for your patch (one is
a simple duplication definition and the other for build kernel with
CONFIG_SMP on i386).

Please keep me updated!

Hope these helps.

Ciao,
Alessio

* Gregory Haskins [off-list ref] wrote:

Hi Ingo, Thomas, and the greater linux-rt community,
  
	I just wanted to let you guys know that our team has a port of 
the 21.5-rt20 patch for the 2.6.22 kernel available. [...]

great! We had the upstream -rt port to .22 in the works too, it was just 
held up by the hpet breakage - which Thomas managed to fix earlier 
today. I've released the 2.6.22.1-rt1 patch to the usual place:

    http://redhat.com/~mingo/realtime-preempt/

I took a quick look at the delta between .22.1-rt1 and your rt1 merge, 
and the delta patch is quite large and i saw no obvious fixes to pick 
up. Could you please filter out any fixes that you might have in your 
tree that you think should be applied, and send it to us as a delta 
against .22.1-rt1? Alessio, could you please also do the same and check 
for any diffs that should be applied? Thanks!

(And, as Daniel suggested, in the future, could you _please_ use a 
different naming scheme instead of using up the upstream -rt namespace.)

	Ingo