On 01/05/2012 02:59 AM, Holger Brunck wrote:
Hi Scott,
On 01/04/2012 06:44 PM, Scott Wood wrote:
On 01/04/2012 09:32 AM, Holger Brunck wrote:
Initial implementation for unsafe feature for biterror insertion on NAND-Flash devices. The code flips single bits in the data block of the flash to simulate single bit-errors. Tested with Samsung K9F1G08U0D and Micron MT29F1G08ABADAWP on km_kirkwood boards.
Do we still need dedicated code for this now that we have the ability to do raw writes?
to be able to do raw writes was very usefull for me to reproduce a bug in the UBI layer in the past. Bitflips happen very rarely but they happen, and with this code you are able to force this and see how other SW layer handle this. Due to the fact there was an empty implementation in the u-boot code for this I thought it might be usefull for others too.
The biterr stub has been around since before we had raw writes. If we decide we don't need a separate biterr command, we should remove the stub.
Anything the biterr command does should be possible with raw writes, so while having a dedicated command might be slightly more convenient, I don't think it's worth the extra code size, review/fix iterations, and maintenance.
Does anyone else have a strong opinion on the matter?
-Scott