#12779 NORM Future : XO-4 custom build hangs SD bus for mmc0 on resume

[Zarro Boogs per Child]

#12779: XO-4 custom build hangs SD bus for mmc0 on resume
--------------------------------+-------------------------------------------
           Reporter:  Quozl     |       Owner:  Quozl                            
               Type:  defect    |      Status:  assigned                         
           Priority:  normal    |   Milestone:  Future Release                   
          Component:  kernel    |     Version:  Development build as of this date
         Resolution:            |    Keywords:                                   
        Next_action:  diagnose  |    Verified:  0                                
Deployment_affected:  au        |   Blockedby:                                   
           Blocking:            |  
--------------------------------+-------------------------------------------

Comment(by Quozl):

 Over three hours I have twice reproduced #12779 on a single laptop using a
 combination of mDNS queries, removal of the iptables rule that classifies
 ICMP packets as user activity, and 1400 byte pings at 630ms to 633ms
 intervals, with WPA2.

 I've also failed to reproduce using the same triggering conditions on an
 open access point, probably because the packets are not held for
 decryption by the wireless card.

 I've also failed to reproduce with other ping intervals outside that
 range.

 Up until the problem occurs, the laptop is waking for about a second or so
 to deal with the incoming packets.  Logs show it is in ''wfi'' for a very
 short time before the wake from WLAN brings the system back up.  The
 kernel takes a few seconds to get USB going.  Packets are handled.  The
 system then goes back to sleep.

 The ping on the other host shows 2.15ms to 4ms RTT most of the time, with
 37ms RTT corresponding to suspend resume cycles of the target.
 (Occasional larger RTTs show buffering, which is probably scans on either
 host).  Therefore the resume takes about 37ms from the transmission of the
 packet.

 When the problem occurs, the laptop stays asleep.  That is, the wake on
 WLAN doesn't happen.  The pings are not returned.  When you wake it
 manually, and mwifiex starts handling an interrupt, you get the "''mmc0:
 Timeout waiting for hardware interrupt.''" in response to reading device
 registers over SD.

-- 
Ticket URL: <http://dev.laptop.org/ticket/12779#comment:16>
One Laptop per Child <http://laptop.org/>
One Laptop per Child bug tracking system