#9936 NORM Not Tri: First WPA2 association (TKIP/AES) always fails, retries succeed

[Zarro Boogs per Child]

#9936: First WPA2 association (TKIP/AES) always fails, retries succeed
------------------------------------+---------------------------------------
           Reporter:  cabalde       |       Owner:  martin.langhoff       
               Type:  defect        |      Status:  new                   
           Priority:  normal        |   Milestone:  Not Triaged           
          Component:  not assigned  |     Version:  1.0 Software Build 802
         Resolution:                |    Keywords:                        
        Next_action:  never set     |    Verified:  0                     
Deployment_affected:                |   Blockedby:                        
           Blocking:                |  
------------------------------------+---------------------------------------

Comment(by martin.langhoff):

 dsd - thanks for chiming in. The Ceibal team is reporting that with their
 APs, WPA always succeeds and WPA2 always fails _on the first connection
 attempt_ (and succeeds in subsequent attempts).

 I am hoping to get a listing of model/chipset from them. If this is not
 AP-specific, it may have to do with memory initialization in supplicant,
 the driver or the firmware.

 Timing-related WPA/WPA2 issues probably remain, but those are hard to
 repro. This seems to be reproduceable 100%, and more clearly measurable
 with the patched NM I prepared (with your dbus/dhcp fix in place and fixed
 logging).

 One oddity is that wpa_supplicant is always running in debug mode, even in
 "normal" operation, due to a bug in NM. It just writes the debug log to
 /dev/null.

 If we think this can be caused by timing issues (wpa_supplicant being too
 slow), I can prep a NM that actually disables debugging in wpa_supp in
 normal operation. But Ceibal's report doesn't sound like a timing issue.

 I left this debug-to-devnull undisturbed so far out of respect towards odd
 timing issues. Some bugs -- inc the ones you mention -- indicate that
 enabling debugging increases our success rate...

-- 
Ticket URL: <http://dev.laptop.org/ticket/9936#comment:6>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system