[Trac #914] Wireless interface enters "frozen mode"

[Zarro Boogs per Child]

#914: Wireless interface enters "frozen mode"
----------------------+-----------------------------------------------------
 Reporter:  jcardona  |        Owner:  josepht
     Type:  defect    |       Status:  new    
 Priority:  high      |    Milestone:  BTest-3
Component:  wireless  |   Resolution:         
 Keywords:            |  
----------------------+-----------------------------------------------------
Comment (by luisca):

 In build 368, the firmware is not frozen anymore, and reloading the module
 will solve the problem. The interface continues acting as a normal mesh
 point, forwarding packets and collaborating in route discovery, and normal
 incoming traffic can be seen with tcpdump. Is only the path from the host
 to the node what is frozen. This causes that no packets or commands can be
 sent to the wireless interface.

 Debugging the driver I found out that priv->dnld_sent ==
 DNLD_{COMMAND,DATA}_SENT and never goes back DNLD_RES_RECEIVED.

 If we manually reset the dnld flag, e hit an error on usb_submit_urb,
 whose first
 lines are:

 if (!urb || urb->hcpriv || !urb->complete)
                return -EINVAL;

 urb is obiously nonzero, and urb->complete, which is a link to the
 handler function on completion, is nonzero too (I checked that). So
 looks like we are exiting because urb->hcpriv is nonzero too, while it
 should be zero. I haven't found in Linux Device Drivers what that
 field means, but looks like the problem might be at usb subsystem
 level.

 Will keep researching and will test tomorrow in build 385.

-- 
Ticket URL: <http://dev.laptop.org/ticket/914#comment:7>
One Laptop Per Child <http://laptop.org/>