#12039 NORM Not Tri: Botched resume +X on XO-1.5

[Zarro Boogs per Child]

#12039: Botched resume +X on XO-1.5
-------------------------------------------+--------------------------------
           Reporter:  dsd                  |       Owner:  wmb at firmworks.com
               Type:  defect               |      Status:  new              
           Priority:  normal               |   Milestone:  Not Triaged      
          Component:  ofw - open firmware  |     Version:  not specified    
         Resolution:                       |    Keywords:                   
        Next_action:  never set            |    Verified:  0                
Deployment_affected:                       |   Blockedby:                   
           Blocking:                       |  
-------------------------------------------+--------------------------------

Comment(by dsd):

 I can only reproduce this issue on one of my 2 test systems: SHC01601310

 wad suggested on IRC that this may be due to a device not being fully shut
 down during suspend. Any access to main memory (which could happen from
 DMA-using devices) will cause memory rot if done while the RAM is in self-
 refresh during suspend.


 I've reproduced it on the 3.3 kernel trimmed down, with all of the
 following excluded: framebuffer, dcon, usb, mmc, wifi, mouse, sound, ...

 And it reproduces in tiny core linux (with rtcwake added).

 Here is an archive that can be extracted onto a USB disk to replicate this
 setup: http://dev.laptop.org/~dsd/20120804/12039.zip

 It boots and then starts a repeated suspend/resume cycle. Remember to
 attach serial as you won't get anything on screen. And unplug the USB disk
 after booting so that its obvious when the system has rebooted coming out
 of suspend.

 Now I'll attempt to reproduce with old kernels and firmware versions.

-- 
Ticket URL: <http://dev.laptop.org/ticket/12039#comment:1>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system