Very slow roaming or reassociation in WPA2 mode

[Rob Vandermeulen]

I will try to send a timestamped log today or tomorrow. And i did try 
ap_scan=1 multiple times ( i probably tested almost every configuration 
variance there is) but it is much slower then ap_scan=2, and doesn't even 
work most of the time. The initial association always goes fine, but 
roaming/re-connecting is a problem, ap_scan=1 seems to get confused after a 
few roams/reconnects and doesn't find my network anymore, even though it's 
there (i'll try to find a log with that behaviour to make it more clear).

Kind regards
Rob Vandermeulen

PS. I already posted this to the ipw2200 dev list a weeks ago, and their 
conclusion was that it didn't seem to be a driver issue at first sight.

PS2 : My APs+ MX are supposed to support "fast reauthentication & seamless 
roaming" (vendorspeak), but information on that is very sparse (it could be 
preauthentication or keycaching or hopefully not some proprietary solution). 
wpa_supplicant seems to support preauth but very little documentation is 
available about these subjects - if anyone has more information i would 
gladly appreciate it

On Sunday 13 November 2005 19:03, Jouni Malinen wrote:
> On Sat, Nov 12, 2005 at 05:52:42PM +0100, Rob Vandermeulen wrote:
> > I can connect fine with wpa_supplicant (0.4.6 version, ipw-1.0.8,
> > ieee-1.1.6, fw 2.4, kernel 2.6.13 ) using wext, and the card connects
> > fine to the 17:80 AP (although the other one is much closer, a few meters
> > from the test setup). Everything is fine, an then i disable the 17:80
> > radio. wpa_supplicants detects this very fast, and start to reassociate.
> > This works, but takes somewhere between 30s-1minute. It first has to try
> > a few times to reassociate, and when that works it just "hangs" while
> > trying to reassociate somewhere in the process...
>
> Have you tried ap_scan=1 setting? ap_scan=2 should, at least in theory,
> be a bit faster, assuming the driver has a good roaming implementation.
> However, it would be interesting to see what happens with ap_scan=1.
>
> I would also like to see a wpa_supplicant debug log with time stamps (-t
> added on the command line), so that it would be easier to figure out
> where the delay was. I have not yet had a chance to test this kind of
> roaming case with ipw2x00 myself, but I'll try to do this is at some
> point with ipw2915.