Hi Roland.
Here is the code from the libibverbs that handles the destroy CQ:
pthread_mutex_lock(&cq->mutex);
while (cq->comp_events_completed != resp.comp_events_reported ||
cq->async_events_completed != resp.async_events_reported)
pthread_cond_wait(&cq->cond, &cq->mutex);
pthread_mutex_unlock(&cq->mutex);
This code will cause for a careless programmer to loop forever if he
acked the events (completion or async)
too many times ....
will you accept a patch that will fix this issue?
thanks
Dotan
_______________________________________________
general mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
