Sean Hefty wrote: >>>+ int (*callback)(int status, >>>+ struct ib_multicast >>>+ *multicast), >>>+ void *context); >>>+ >> >>Is this re-introducing module unload races we had with sa all over again? > > > The call returns a structure that must be freed. If the structure is freed by > returning a non-zero call to the callback, then we have the same problem that > ib_cm and rdma_cm have. Not allowing a return value from the callback is an > easy fix for that though.
I looked at this in more detail. While it's easy on the ib_multicast module to not allow returning a value from the callback, it makes it more difficult on the users to handle failures. (The ipoib changes use this feature.) So, I think we want registration with the ib_multicast module similar to what was done with the ib_sa. - Sean _______________________________________________ openib-general mailing list openib-general@openib.org http://openib.org/mailman/listinfo/openib-general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
