Michael> No, I think there's no problem - user callback has
Michael> finished running and that's all we care about.
I don't think so -- because then the unregister call can return and
the client module free the client struct before the wake_up() runs,
which leads to use-after-free.
Michael> But I'd like to point out that using rwsem gets all these
Michael> corners right automatically. Reconsider that approach?
Again, I don't think so -- rwsem just hides the race from you. There
is a race with semaphores (access semaphore after waking up waiter)
which is the original reason for creating struct completion.
- R.
_______________________________________________
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
