MPCI layer problems

Roger Dahlkvist ruppe at
Mon May 11 14:58:51 UTC 2009


When I try to create a global object, the rtems_multiprocessing_announce directive fail and system hang.
I'm using a homogeneous system with two Nios II processors and I have verified that the interprocessor communication works properly and that there are enough packet buffers available.

When tracing what happens, this is what I see on the different nodes.

* During initialization all nodes go thru the MPCI routines initialization, get_packet and send_packet.
* The new message is detected ón the respective node and rtems_multiprocessing_announce is called and as part of that the receive_packet and return_packet routines are called aswell.
* I create a global message que on node 1 and this causes get_packet and then send_packet to be called.
* On node 2 the system hang directly when the new message is detected and rtems_multiprocessing_announce is called. !?!?!

I've looked at the documentation and tried everything I can think of, but still no luck. What have I forgotten?

Any ideas on what might be causing this are very much appreciated.

Thanks in advance.

More information about the users mailing list