[PATCH v2 00/20] mesh support for wpa_supplicant

[Jouni Malinen]

On Mon, Oct 20, 2014 at 09:07:34AM -0400, Bob Copeland wrote:
> Yeah, if it just sent only commit and transitioned to SAE_COMMITED state
> that should work.  Is the STA<->AP exchange variant documented somewhere?
> I only know about 802.11-2012 11.3.

Figure 4-21 in IEEE Std 802.11-2012 shows an example of this.

> Also it seems wrong that AP->STA would send the tr#2 (confirm) frame
> more than once in the normal case.  It should be transitioning out of
> COMMITTED into ACCEPTED state without replying, i.e., I would expect it to
> look like (as I wrote it):
> 
>     sta --- COMMIT  --> ap
>     sta <-- COMMIT  --- ap
>     sta <-- CONFIRM --- ap
>     sta --- CONFIRM --> ap
> 
> ...without another CONFIRM frame from the AP.

There should certainly not be two copies of the same frame. That
sequence of two Authentication frames from AP to STA looks quite strange
in the context of Authentication (even if allowed for the SAE part). In
addition, the following STA->AP Authentication frame and following
Association Request STA->AP would look a bit strange. It looks more
natural IEEE 802.11 authentication exchange when those frames are sent
in similar order to older Authentication frame exchanges.. And anyway,
this sequence shown here does not match Figure 4-21 (or the current
hostapd/wpa_supplicant implementation for that matter).

This should be:

STA->AP Authentication tr#1 (SAE commit)
STA<-AP Authentication tr#1 (SAE commit)
STA->AP Authentication tr#2 (SAE confirm)
STA<-AP Authentication tr#2 (SAE confirm)
STA->AP Association Request
STA<-AP Association Response
 
-- 
Jouni Malinen                                            PGP id EFC895FA