> Wait, I see ... cma_remove_id_dev() would return 0 from the > event_handler, ensuring cma_process_remove() does not invoke > rdma_destroy_id(), is that it? yep - the destruction of the id is controlled by the user