On Mon, Feb 24, 2020 at 06:18:45AM +0100, Markus Theil wrote: > This is just how I found the current Linux kernel to be build, when I started to integrate its > control port over nl80211. It allows to set a ethertype, which should be received over the controlled port, > which I set to EAPOL. Additionally and uncoditionally, it forwards all PRE-AUTH frames over the very same > control port (cannot be configured). Therefore I had to do it that way, in order to use the current kernels > implementation. Oh.. That's unfortunate. This is not the way the controlled port was supposed to work and I had not realized the mac80211 implementation ended up doing this. The controlled port is used for frames that are coordinating the _current_ association, i.e., EAPOL frames in the most common case. Ethertype registration is there to allow other protocols (e.g., WAPI) to be implemented. RSN preauthenication is completely different story and it should not have used controlled port at all. While, at least in theory, it would likely be mostly fine to use nl80211 control port for RSN preauth RX side (but I do not want it for TX side), I'll see what can be done with the kernel side to avoid having to do this at all for RX either. I'd prefer to keep wpa_supplicant/hostapd implementation clean on this area (TX&RX for EAPOL only) and not have to mix in unnecessary components to workaround the current mac80211 implementation. -- Jouni Malinen PGP id EFC895FA _______________________________________________ Hostap mailing list Hostap@xxxxxxxxxxxxxxxxxxx http://lists.infradead.org/mailman/listinfo/hostap
