[Rtai] RTAI and SMP, any constraints?

[Erwin Burgstaller]

Hi, again!

* Wed, Feb 11, 2009 - 16:09:24 +0100, Erwin Burgstaller - rtai.berknapp at spamgourmet.com:
>
> Having rtai (3.6.1) running on a preemtive smp-kernel (2.6.22), is there
> something which will not work, especially with semaphores, spinlocks and
> shared memories?
> 
> I'm just asking, 'cause we have a problem with an application which
> consists of several processes and these of several threads.
> 
> The box sometimes freezes completely, but doesn't, if all rtai tasks
> are bound to just one cpu.

In the meantime we've found out, that it doesn't freeze too, if we do
not only restart our package but rtai too (unloading and loading the
rtai modules).

Any ideas about that? 

/proc/rtai/names is cleared from all the names our application did
create, so I think we always do a clean shutdown, or maybe we forgot
something?

I also wrote a test, which creates 100 threads with rtai tasks. Each
thread is just incrementing numbers but synced at one point with
exchangeable semaphore or spin lock and the main routine does also
allocate 30 MB of shared memory. This process runs about 12 seconds and
will than be launched again by a while-loop in bash. The test ran almost
48 hours, without any problem.

Regards,
Erwin

-- 
Erwin Burgstaller