> 
> > > How YAFFS and YAFFS2 behave?
> > 
> > YAFFS is being more cautious than that, which means that in theory 
> > YAFFS will lose blocks faster than they recommend. However from 
> > accelerated lifetime testing I've done, I have not seen
> this to be a
> > practical problem since ECC errors are so rare once the
> problematic blocks have been removed.
> 
> There is another aspect to that - I've seen yaffs1 leak blocks under 
> power cycling conditions. I suspect power failures were producing 
> incomplete write/erase cycles and ecc errors as a consequence. That 
> led to retirement of good blocks that had bad luck to be written to 
> during power down.
> 
> I would be interested to know if anybody else has seen this behavior.

Do you have your write protect pin tied to a power fail flag or such?
The way the wp pin works is to disable the internal programming voltage
generator. The residual charge might be enough to complete a write, but
is not enough to complete an erase.

If wp is not disabled and the system rails can supply sufficient power
to the NAND chip to complete the erase (~2 to 3 msec) then things should
be fine.

Perhaps another thing worth doing is to do some power state check within
the NAND driver to check a power flag before issuing the write/erase
command.

I wonder whether it is worth trying to detect this case in YAFFS and
give it special treatment. The heuristics to do this detection are a bit
boggling.

Charles