#12590 BLOC 4-firmw: Power LED still on after power off

[Zarro Boogs per Child]

#12590: Power LED still on after power off
-------------------------------------------+--------------------------------
           Reporter:  dsd                  |       Owner:  rsmith       
               Type:  defect               |      Status:  new          
           Priority:  blocker              |   Milestone:  4-firmware   
          Component:  embedded controller  |     Version:  not specified
         Resolution:                       |    Keywords:               
        Next_action:  never set            |    Verified:  0            
Deployment_affected:                       |   Blockedby:               
           Blocking:                       |  
-------------------------------------------+--------------------------------

Comment(by rsmith):

 Replying to [comment:2 dsd]:
 > Richard can't reproduce this after many tries.

 I have finally duplicated this with full logs.  Took 91 cycles of a 10
 minute runin on a machine running build 31 and firmware q7b21.  Here's the
 final bit of kernel log:

 {{{
 [  434.728397] Restarting system.
 [  438.695395] olpc-ec-1.75: EC cmd error: timeout in STATE 2
 [  438.700843] power_supply olpc-battery: driver failed to report
 `charge_full_design' property: 4294967186
 [  438.710475] olpc-ec-1.75: transmit FIFO not empty, resetting
 [  442.705396] olpc-ec-1.75: EC cmd error: timeout in STATE 2
 [  443.705385] Reboot failed -- System halted
 }}}

 And the corresponding EC log:

 {{{
 142857599:Batt runin discharge stop
 MPPT Disabled
 Plimit = 01
 142872718:AC In
 142873024:Life start
 142873128:Chg On
 Trickle Off
 142876555:SDI switchwait noack
 142880551:SDI switchwait noack
 }}}

 > We think this might be a more general failure to shutdown the system.
 This might be related to our occasional EC communication problems e.g.
 #12525. If that is the case, this might not be seen with build 33 (idle
 suspend disabled), as we think suspend/resume is the trigger of such
 communication issues.

 So this is 100% correct.  This bug is triggered when the a shutdown tries
 to happen after #12525 has occurred.

 #12525 seem to only occur after a suspend/resume so this will probably
 never be duplicated on build 33/34 with interactive use as suspend/resume
 is disabled in those builds.  It will probably happen to a few machines in
 runin.  I suspect we will see a few of these in ramp once we up the number
 of machines built.

 There is currently no (EC) fix for this.

 Paul and I have discussed a possible workaround for #12525.  We need to
 implement that and see if it helps.

-- 
Ticket URL: <http://dev.laptop.org/ticket/12590#comment:3>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system