On Wed, 2018-01-31 at 15:33 -0600, Denis Kenzior wrote: > > + if (!cfg80211_rx_control_port(dev, skb->data, skb->len, > + ehdr->h_source, > + be16_to_cpu(skb->protocol), > + noencrypt)) > + dev_kfree_skb(skb);
Err, you should free it unconditionally, no? You don't do anything with the skb afterwards, since you can't get into the else branch and deliver it to the netdev now, so afaict it'd be leaked if not freed? Which explains why you didn't notice the error in the previous patch, it was making this code correct by freeing it all the time (never hitting an error, presumably, and even if you wouldn't notice the small leaks) johannes
