On Friday 09 March 2007 03:34, yi zhang wrote:
> My system was based on PXA270，consists of
> linux2.6.9 + intel patches + cramfs +　yaffs,
> nandflash was samsung K9F1208, 64M. and this chip was mounted
> as 3 yaffs partitions, After a period of usage of this system,
> there may appear a lot of badblocks,we can see the badblock
> messages in the kernel booting messages, and sometime one of
> the 3 partitions could be damaged entirely. the partitons
> which marked as badblocks cannot be erased by the 'mkyaffs'
> tool.when this occurs, the worst way is that the nandflash
> chip could be changed to another nandflash chip.

Yi, see:
 
http://aleph1.co.uk/lurker/list/message/20070214.151530.af180dc0.en.html

Disable bad-block marking when bringing up new designs/code -- 
hack Yaffs to stop it marking blocks bad -- grep for 
'MarkNANDBlockBad'.

There is probably an issue with reading/writing the 
spare/oob/tags meta data. See the numerous posts on this 
subject. Make sure NAND is working properly at the MTD level.

Make sure the problem of interfacing Yaffs with MTD's oob i/o is 
working -- this may require tweaks to Yaffs and/or MTD depending 
on the versions of both.

Charles/Wookey/Vitaly - Can we add some BOLD statements to the 
Yaffs web pages that advise new users as to the issues, pitfalls 
and best practices for bringing up new MTD based platforms. 
Shouldn't we add a load option or config option to disable 
bad-block marking. Perhaps add some test code that can be used 
to verify i/o to the spare/oob area is working as needed. The 
tags issue comes up weekly as the first (negative) experience 
with new users. [yes, I know if I suggest something like this I 
should code it too, but I freelance with Bright Star... blah 
blah, I'm busy, excuse..., I have kids, excuse...]

-imcd