On Wed, 2012-07-04 at 13:12 +0200, Johannes Berg wrote:

Hi Michal,

A few people noticed that these patches cause deadlocks. I think the
reason is that it's not allowed to lock the device here:

quoted

 	case NETDEV_DOWN:
 		dev_hold(dev);
+		cfg80211_lock_rdev(rdev);

This causes a deadlock since the interface is set down inside mac80211
when inside a handler from cfg80211 already, or so.

I just reproduce it -- it's because we come from nl80211_del_interface()

johannes