Delay between Association Resp and Eapol-Start

[Konstantinos Georgantas]

Thanks for the clarification Jouni. I have not observed any problems 
with WPS 2.0 devices. It is WPS 1.0 I wanted to speed up and this change 
does it.

BR,
Kostas


I have not observed any issue with WPS 2.0 devices
 >On Wed, Oct 16, 2013 at 02:00:28PM +0200, Konstantinos Georgantas wrote:
 >> I have been troubleshooting the above mentioned issue and I was
 >> wondering why not to use something like this in eapol_supp_sm.c:
 >>
 >> @@ -288,7 +288,7 @@
 >>           */
 >>  #ifdef CONFIG_WPS
 >>          /* Reduce latency on starting WPS negotiation. */
 >> -        sm->startWhen = 1;
 >> +        sm->startWhen = 0;
 >>  #else /* CONFIG_WPS */
 >>          sm->startWhen = 3;
 >>  #endif /* CONFIG_WPS */
 >>
 >> What complications could such a change introduce when running wps?
 >> Is there an explicit reason for waiting 1 sec between Association
 >> Resp and Eapol-Start? I do not see any side-effects when setting
 >> startWhen to 0.

 >The main reason for startWhen being initialized to non-zero value is in
 >avoiding extra frame exchanges since the normal case for initiating IEEE
 >802.1X/EAPOL authentication is by the authenticator sending out
 >EAP-Request/Identity frame. If the supplicant sends out EAPOL-Start
 >message on association that can result in two rounds of EAP Identity
 >frames.
 >
 >WPS 1.0 is somewhat of an odd design in this context since it uses
 >language to mandate the AP to wait for EAPOL-Start before initiating the
 >exchange. This is an unfortunate workaround for the old spec not
 >mandating clear indication in the Association Request frame of intent to
 >go through WPS exchange. This has been fixed in WSC 2.0 to allow the AP
 >to start EAPOL exchange immediately.
 >
 >Changing sm->startWhen = 0 could potentially speed up some cases with
 >old WPS implementations. However, it would result in extra round trips
 >with more modern implementations and would be undesired for almost all
 >other uses of EAPOL (e.g., WPA2-Enterprise). As such, I would not
 >recommend this type of change.

 >If you are seeing issues with sm->startWhen = 1 with any WPS 2.0 device,
 >it would be nice to hear more details about the problems. I'm assuming
 >the only problem you are talking about here is in delaying WPS 1.0
 >connection by one second.

 >--
 >Jouni Malinen                                            PGP id EFC895FA