On Fri, 2019-02-22 at 13:41 +0100, Johannes Berg wrote: > > +++ b/net/wireless/nl80211.c > > @@ -13098,6 +13098,12 @@ static int nl80211_external_auth(struct > > sk_buff *skb, struct genl_info *info) > > if (!info->attrs[NL80211_ATTR_STATUS_CODE]) > > return -EINVAL; > > > > + if ((info->attrs[NL80211_ATTR_PMK] && > > + !info->attrs[NL80211_ATTR_PMKID]) || > > + (info->attrs[NL80211_ATTR_PMKID] && > > + !info->attrs[NL80211_ATTR_PMK])) > > + return -EINVAL; > > This constitutes a netlink API change, so no, can't be right? PMKID > was > perfectly reasonable to pass by itself before. Good point. Andrei, can you fix this? This can easily be changed to accept PMKID alone but still do what you want when both are included. -- Luca.