#1905 BLOC Trial-3: Field Return: flash corruption - OpenFirmware complaining of 'unknown node type 2006'.

[Zarro Boogs per Child]

#1905: Field Return: flash corruption - OpenFirmware complaining of 'unknown node
type 2006'.
-----------------------+----------------------------------------------------
  Reporter:  dwmw2     |       Owner:  wad     
      Type:  defect    |      Status:  assigned
  Priority:  blocker   |   Milestone:  Trial-3 
 Component:  hardware  |     Version:          
Resolution:            |    Keywords:          
  Verified:  0         |  
-----------------------+----------------------------------------------------
Comment (by gnu):

 Another useful test would be to have the kernel check for the bad 0x80 bit
 in the length field upon read-back.  If it exists, then go back and see if
 the original memory buffer contains the bad bit.  If it does not, you
 probably have a transient hardware problem (e.g. a RAM bit reads back
 wrongly sometimes, or a bus corruption, or a signal that's moving during
 the setup/hold time).

 Has anyone considered the possibility of a bit-spreader bug in the kernel
 (i.e. software somewhere that writes to this byte when it shouldn't,
 through a bad pointer)?  If there is a way to run this system under a
 logic analyzer or hardware monitor, trigger on accesses to this memory
 buffer location with 0x80 bit set.  The first time it triggers, you can
 catch either a read (i.e. transient RAM or bus failure) or a write (bit
 spreader) that way.  Or, change the code very slightly to use a different
 part of the RAM as the copy-buffer, and see if the problem persists.

 Did I see in Bryan Ma's comment that moving the flash chips from the
 failing unit to a good unit caused the good unit to start failing?  If he
 meant they desoldered the chips and physically moved them, that looks like
 a strong piece of evidence that it isn't a RAM or bus or bit-spreader
 problem; perhaps an incompatibility between CAFE and the NAND chips on
 timing or something.

-- 
Ticket URL: <https://dev.laptop.org/ticket/1905#comment:32>
One Laptop Per Child <http://laptop.org/>