#9869 HIGH 1.5-sof: XO-1.5 screen glitches on resume

[Zarro Boogs per Child]

#9869: XO-1.5 screen glitches on resume
---------------------------------------+------------------------------------
           Reporter:  wad              |       Owner:  cjb                              
               Type:  defect           |      Status:  new                              
           Priority:  high             |   Milestone:  1.5-software-later               
          Component:  x window system  |     Version:  Development build as of this date
         Resolution:                   |    Keywords:  DCON XO-1.5                      
        Next_action:  code             |    Verified:  0                                
Deployment_affected:                   |   Blockedby:                                   
           Blocking:                   |  
---------------------------------------+------------------------------------

Comment(by wad):

 Replying to [comment:7 Quozl]:
 > Replying to [comment:6 pgf]:
 > > the moral?  don't stress the freeze/unfreeze mechanism so hard.
 >
 > Why are DCON interrupts missed during the freeze?  How will we prevent
 unfreeze from potentially being requested within this critical time?
 Should the kernel implement a delay to prevent a race?

 The freeze requires that DCONLOAD be deasserted for two VSYNC pulses.
 The first VSYNC after deassertion starts loading the video into the DCON
 frame buffer, and the second one VSYNC terminates the frame buffer loading
 and triggers the switch from host -> DCON control of the video output.

 The DCON specification doesn't say what should happen if the host
 reasserts DCONLOAD before it has finished loading the frame buffer, but
 given other comments in the spec the behavior seen would make sense.
 Effectively, reasserting DCONLOAD before the freeze has finished is
 aborting the freeze, and therefore no interrupt indicating a successful
 freeze is generated.

 We are still left with the hardware bug that pgf uncovered with the
 occasional "premature DisplayLoad interrupt" being generated.

-- 
Ticket URL: <http://dev.laptop.org/ticket/9869#comment:10>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system