Re: [Trac #214] CAFÉ NAND ECC poorly understood and needs optimisation.

[Zarro Boogs per Child]

#214: CAFÉ NAND ECC poorly understood and needs optimisation.
----------------------+-----------------------------------------------------
 Reporter:  dwmw2     |        Owner:  mfoster
     Type:  defect    |       Status:  new    
 Priority:  normal    |    Milestone:  BTest-1
Component:  hardware  |   Resolution:         
 Keywords:            |  
----------------------+-----------------------------------------------------
Changes (by dwmw2):

  * owner:  blizzard => mfoster
  * component:  distro => hardware

Comment:

 We've seen a crash in the error correction -- one of the values in the
 'err_pos' lookup table is out of range, and hits a BUG(). It looks like
 we're calling err_pos() with a value of 0x55e, giving an unknown result
 (which comes out as 0xfff and triggers the BUG()).

 http://dev.laptop.org/~cjb/nand-crash

 We desperately need to understand this code.

 As a separate issue, this was being seen because the ECC hardware seemed
 to be generating spurious errors.

-- 
Ticket URL: <http://dev.laptop.org/ticket/214#comment:1>
One Laptop Per Child <http://laptop.org/>