Recently, I find the Q-Host goes to panic sometimes. I trace the process status by issuing "ps -a" command, finding Mqueue occupies many memory resources.
In the folder "/dev/mqueue", there are thousands of PCM.tapc_req_q.xxxx . It seems that all of them are generated by tapc.
some sample logs captured from Q-host syslog:
Tue Apr 16 16:44:56 2013 0002         tapc           1 22485 1:PCM.tapc_req_q.22485 open on 4
Tue Apr 16 16:44:57 2013 0002         tapc           1 22489 1:PCM.tapc_req_q.22489 open on 4
Tue Apr 16 16:45:15 2013 0002         tapc           1 22622 1:PCM.tapc_req_q.22622 open on 4
....................................
....................................
...................................
Thu Apr 18 11:58:59 2013 0002         tapc           1 12430 1:PCM.tapc_req_q.12430 open on 4
I compare those logs with other Q-Host systems (We have three Q-Host systems here), those events generate normally. The main difference is that ControlMatrix version (4_01b vs 5_00) and tapc_req_q.xxxx files kept in /dev/mqueue. The same folder in the rest of Q-Host systems does not contain suck kind of files.
those kinds of logs appear 73xx within 2 two days. I doubt it should be the root cause to increase the size of QNX Mqueue.
Does anyone know how to fix it?

A full TAPC connection with "full stop" to terminate the connection can help to reduce the memory being exhausted.
It still creates many flag files (with zero byte) onto /dev/mqueue in ControlMatrix version 4.01b.
Those files can be deleted by crob job daily.
There is no such issue on ControlMatrix version 5.00.