[asterisk-dev] Rejected frames occuring after substantial number of calls made over PRI, related to bug 11189

[Jeff Peeler]

For some time now I've been working on trying to fix bug 11189. The only problem that I think remains is that when the number of calls go above 40 or so (I've actually been using the maximum 60 for my particular card), sometimes there are some frames that are rejected. I've seen frames get rejected using both Zaptel/DAHDI on 1.4/trunk, but am currently testing on the latest SVN copies of Asterisk 1.4, Zaptel 1.4, and libpri trunk. It has been noted that high CPU usage will cause rejected frames, but I've never noticed CPU usage in Asterisk 1.4 greater than 50%. These tests were performed on a TE420 set to E1 mode with crossover cables connecting the spans.

Starting at the lowest level, I began trying to see if libpri could be at fault. I've not gotten very far in my attempts. However, perhaps this excerpt from the Q.921 spec is relevant:

5.6.1   Transmitting I frames
Information received by the data link layer entity from layer 3 by means of a DL-DATA request
primitive shall be transmitted in an I frame. The control field parameters N(S) and N(R) shall be
assigned the values of V(S) and V(R), respectively. V(S) shall be incremented by 1 at the end of the
transmission of the I frame.