On Thu, 2016-01-28 at 15:30 +0300, Dan Carpenter wrote: > It's not the return where we should trigger the warning it's at the > > rcu_assign_pointer(sta->ampdu_mlme.tid_rx[tid], tid_agg_rx); > > line. That's for correctness, but also it should be slightly easier. > Or it should cut down on false positives if we ignored returns and > only looked global scope type assignements.
That's a good idea! But even that will probably get you a lot of false positives. For example, in this structure, the rcu_head is never initialized until we need it for kfree_rcu() or call_rcu(). I'm sure there are other places like it. johannes
