#10232 HIGH 10.1.2: WiFi dies on suspended XO-1, os300

[Zarro Boogs per Child]

#10232: WiFi dies on suspended XO-1, os300
------------------------------------+---------------------------------------
           Reporter:  hal.murray    |       Owner:               
               Type:  defect        |      Status:  new          
           Priority:  high          |   Milestone:  10.1.2       
          Component:  not assigned  |     Version:  not specified
         Resolution:                |    Keywords:               
        Next_action:  never set     |    Verified:  0            
Deployment_affected:                |   Blockedby:               
           Blocking:                |  
------------------------------------+---------------------------------------

Comment(by hal.murray):

 > hal -- what traffic are you sending to the wlan (from another system)
 when
 > this occurs?  i assume you've taken care of locking down the arp cache
 > somehow?  and the traffic is a ping of some sort? it's possible you're
 > seeing a wakeup race -- i.e., the wlan could send up a packet while the
 > system is going to sleep, and if no other packet is received, or the
 > initial packet isn't taken by the kernel, the wlan module might never do
 a
 > wakeup.

 I'm running a fake ntp server and poking it with real ntpd on 2 PCs.
 One of them pokes it every 64 seconds.  The other varies from 64 to 256
 seconds.

 On both of those PCs, the arp slot for the XO was setup manually.

 I've got boatloads of ntp logging enabled on both PCs, but ntpd only
 logs things when a packet arrives not when it sends one.  We can probably
 figure out when the first packet was lost by stepping forward from the
 last one received.


 > we should also start thinking about workarounds.  e.g., if we wake once
 a
 > minute, on a timer, will that recover well enough?  i'm picturing going
 > right back to sleep unless the network traffic filters notice activity
 > right after that wakeup.

 I'd expect that to work in this case.  It recovers when I poke the
 touchpad.

 But this is only half of the problem.  There is also the case where the
 LED
 is off.  That doesn't recover when I poke the touchpad.  It does recover
 if
 I poke the icon for my AP so we can probably debug that path but it will
 take
 some more work.

-- 
Ticket URL: <http://dev.laptop.org/ticket/10232#comment:21>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system