[ofw] memory 'leak' running ndconn

Sean Hefty sean.hefty at intel.com
Wed Jan 20 17:15:30 PST 2010


I've traced down the reason, but not the cause, of the memory increase.
The client side sometimes does not finish disconnecting, leaving QPs
lying around until the test finishes.  I added counters to
CClientEp:DisconnectSucceeded() to track the differences between the
number of endpoints created, versus the number destroyed.  At times, the
number created increases faster than the number destroyed.

created 2766,  destroyed 1762
created 4934,  destroyed 2927
created 6295,  destroyed 3292
created 7604,  destroyed 3598
created 8903,  destroyed 3902
created 10169, destroyed 4168
created 11460, destroyed 4457

The result is a pathetic connection rate, with a whole lot of memory
being consumed.  This may be a scheduling issue, but I'm guessing that
somehow specific connections are not disconnecting properly.

- Sean




More information about the ofw mailing list