#4013 BLOC Update.: sd goes unusable after a suspend cycle

[Zarro Boogs per Child]

#4013: sd goes unusable after a suspend cycle
---------------------------+------------------------------------------------
  Reporter:  PierreOssman  |       Owner:  dwmw2                            
      Type:  defect        |      Status:  new                              
  Priority:  blocker       |   Milestone:  Update.1                         
 Component:  kernel        |     Version:  Development build as of this date
Resolution:                |    Keywords:  killjoy?                         
  Verified:  0             |    Blocking:                                   
 Blockedby:                |  
---------------------------+------------------------------------------------

Comment(by tomeu):

 Ok, just tried in an update.1 build and things seem to be working
 surprisingly well.

 After resume, HAL tells the journal that the old device disappeared. The
 journal then does the same we do when a kid yanks an usb stick, request
 the DS to unmount the xapian DB in that mount point. This fails as the
 files are not there anymore, but the journal copes with that and removes
 the icon from the devices bar.

 Then when HAL announces that the new SD card device has appeared, the
 journal mounts it and notifies the DS to open the xapian index in there.
 The journal looks now as before suspending.

 The only issue I see is that if not all changes had been flushed to the
 index when we suspended, those changes will be lost. This same problem
 arises if a kid yanks a usb stick.

 This last issue is mitigated by the fact that we flush changes every 20
 changes or every minute since the last change.

 So perhaps we can live with that possibility of data loss, or we could
 increase the flush frequency for removable devices, or we could flush at
 every change. Suggestions?

-- 
Ticket URL: <http://dev.laptop.org/ticket/4013#comment:15>
One Laptop Per Child <http://dev.laptop.org>
OLPC bug tracking system