#11260 NORM Not Tri: i8042 data loss over suspend/resume

[Zarro Boogs per Child]

#11260: i8042 data loss over suspend/resume
-------------------------------------------+--------------------------------
           Reporter:  dsd                  |       Owner:  rsmith     
               Type:  defect               |      Status:  new        
           Priority:  normal               |   Milestone:  Not Triaged
          Component:  embedded controller  |     Version:  1.5-C2     
         Resolution:                       |    Keywords:             
        Next_action:  never set            |    Verified:  0          
Deployment_affected:                       |   Blockedby:             
           Blocking:                       |  
-------------------------------------------+--------------------------------

Comment(by dsd):

 One relevant difference between 2.6.31 kernels (from 10.1 releases) and
 2.6.35 kernels (in 11.x releases) is that 2.6.35 kernels disable i8042
 interrupts much earlier going into suspend. This behaviour was changed to
 fix some races in resume in #10650.

 I tested in 10.1.3 (build 860, kernel 2.6.31) on XO-1.5, running firmware
 Q3A62 (EC code 2.0.1). It is not easy to reproduce the issue there,
 because the interrupts are enabled for much longer, and suspend itself is
 slower than 2.6.35. I did reproduce it exactly once. I also ran several
 times into the issues in #10650 where the interrupts are processed in the
 wrong order during resume (kernel bug, which is why we ended up disabling
 them early). This made it difficult to perform many tests as I had to keep
 rebooting.

 I then retested with Q3B19 against that old OS build. Again reproduced
 exactly once, but with difficulty,  and with several occurances of #10650
 getting in my way.

-- 
Ticket URL: <http://dev.laptop.org/ticket/11260#comment:7>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system