#10161 NORM 11.2.0-: invoking ul-warning from X screen trashes display

[Zarro Boogs per Child]

#10161: invoking ul-warning from X screen trashes display
--------------------------------+-------------------------------------------
           Reporter:  pgf       |       Owner:  jnettlet                         
               Type:  defect    |      Status:  new                              
           Priority:  normal    |   Milestone:  11.2.0-M4                        
          Component:  kernel    |     Version:  Development build as of this date
         Resolution:            |    Keywords:                                   
        Next_action:  diagnose  |    Verified:  0                                
Deployment_affected:            |   Blockedby:                                   
           Blocking:            |  
--------------------------------+-------------------------------------------
Changes (by dsd):

  * next_action:  review => diagnose


Comment:

 Stuck :(

 The patch isn't entirely correct - its confusing VT_ACTIVATE which is used
 for choosing which VT to activate (1, 2, 3, ..) with KDSETMODE which is
 used for choosing KD_GRAPHICS, KD_TEXT, etc.

 We also want to avoid VT_WAITACTIVE at all costs (see #7351) - thats why
 we have chvt_in_userspace()

 chvt_in_userspace() is already called in that path just a few lines down,
 so unless I'm missing something, the patch doesn't actually introduce
 significant changes.


 Moving the chvt above the KDSETMODE (the key thing this patch does)
 appears to help most of the time, but not 100%, I still see the failure
 now and then. Also, the on-boot animation does KDSETMODE then chvt in that
 order and this works fine. Even stranger, if you swap the order in the
 boot path to match what this patch does then you get the screen death like
 this during boot.

 In the kernel source there are various comments about disconnect between
 vt and framebuffer layers in this area so perhaps it's not too surprising
 we're seeing strange behaviour. As this isn't an important bug and I've
 already lost a fair amount of time to it I'm going to forget about it for
 now.

-- 
Ticket URL: <http://dev.laptop.org/ticket/10161#comment:10>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system