#4474 NORM Never A: EC explodes if EC (write) command executed too closely to suspension

[Zarro Boogs per Child]

#4474: EC explodes if EC (write) command executed too closely to suspension
---------------------------------+------------------------------------------
 Reporter:  dilinger             |       Owner:  David.Lin     
     Type:  defect               |      Status:  new           
 Priority:  normal               |   Milestone:  Never Assigned
Component:  embedded controller  |     Version:                
 Keywords:                       |    Verified:  0             
---------------------------------+------------------------------------------
 In the kernel's olpc_fixup_sleep, we need to execute olpc_ec_mask_unset()
 to disable 1% SOC SCIs while we've suspended.  However, if it is run
 without any sort of delay at the end, the following things happen:

  - we execute EC_WRITE_SCI_MASK (0x1b) cmd
  - we wait for the ACK via IBF/OBF flags
  - we send a byte to the EC (the SCI mask)
  - we wait for IBF to be clear
 at this point, the EC transaction should be done.  however..
  - we proceed to suspend

 When we awaken, the EC's IBF/OBF flags are hosed; all EC communication
 fails.  If we reboot into OFW, all of its EC communications fail.

 However, if we add a udelay(250) after the EC_WRITE_SCI_MASK command but
 prior to suspending, when the system resumes the EC is fine.  Another
 thing that works is to run another EC command (EC_READ_SCI_MASK) prior to
 suspend.

 So basically, there's some sort of race in the EC.. if we suspend too
 quickly after a write command, it gets confused.

-- 
Ticket URL: <https://dev.laptop.org/ticket/4474>
One Laptop Per Child <https://dev.laptop.org>
OLPC bug tracking system