#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:  new    
  Priority:  blocker   |   Milestone:  Trial-3
 Component:  hardware  |     Version:         
Resolution:            |    Keywords:         
  Verified:  0         |  
-----------------------+----------------------------------------------------
Comment (by dwmw2):

 The ECC has been well-tested, so I find that unlikely. Remember also that
 Quanta's machine was suffering only in OpenFirmware, where error
 correction wasn't enabled -- but Linux could read the device just fine
 (and reported ECC corrections while doing so). At the time I thought that
 was a different problem -- but if it's a problem with the internal SRAM of
 the CAFÉ, their machine could actually be exhibiting a different symptom
 of the _same_ problem.

 I'll give a full dump including OOB area of an offending page though, and
 we can verify that ECC is working correctly...

 The nandtest program just failed with this output:
 {{{
 215a0000: checking...
 compare failed. seed 982916440
 Byte 0xb72f is f7 should be f6
 }}}

 This means that the _intended_ data for the offending page are obtained
 by:
 {{{
     srand(982916440);
     for (i=0; i<0xb000; i++) rand();
     for (i=0; i<0x800; i++) buf[i] = rand();
  /* ASSERT(crc32(-1, buf, 0x800) == 0x10358381); */
 }}}

 The contents of the ECC area are:
 {{{
 # nanddump -l 2048 -s $((0x215ab000)) -p /dev/mtd0 2>/dev/null  | grep OOB
 | head -1
   OOB Data: 64 dd 1d e2 2e b2 b4 05 2f 59 a6 e3 3a 3f ff ff
 }}}

 This matches what we get when we write the corrupted contents to flash.
 When the intended version is written to flash correctly, we get the
 following ECC data:
 {{{
   OOB Data: d3 f1 4e 90 54 1f 60 e0 66 ec 93 0e eb 1f ff ff
 }}}

-- 
Ticket URL: <http://dev.laptop.org/ticket/1905#comment:48>
One Laptop Per Child <http://laptop.org/>