On Sunday 23 September 2007 17:15, Jonathan McDowell wrote:
> This is just a minor addition; the trace message for when
> yaffs marks a block bad doesn't have a newline on the end
> which causes output like:
>
> nand_erase: attempt to erase a bad block at page 0x000223e0

This doesn't look right - yaffs should not be erasing bad-blocks.
If the block just turned bad AND it got marked bad in MTD then
this might happen - but i don't think that's likely, and most MTD
setups don't have dynamic (mtd) bad-block-list update.  This is
probably an indicator of some other issue, perhaps not reading
the true blockState during initial scan.

> **>> Erasure failed 4320
> yaffs: marking block 4319 bad<4>**>> Block 4320 retired

This is a side-effect of earlier attempt to erase bad-block.

> The attached patch fixes that.

Thanks for the newline!

> Also this set of messages is a bit confusing; the off by one
> appears to be due to yaffs_MarkBlockBad taking account of the
> dev->blockOffset field, but why does it do so when the other
> functions don't seem to?

I think this is covered by the ChunkInNAND = chunkInNAND - 
dev->chunkOffset in the read and write functions, equivalent to 
the blockOffset adjustment in the other functions in 
yaffs_nand.c

-imcd