>On Monday 24 April 2006 23:12, Ian Oliver wrote:
>> In article <200604211605.54107.manningc2 at actrix.gen.nz>, Charles
Manning 
wrote:
>> > If you must do FAT then it is probably simpler & more efficient to
just
>> > do  a FAT file system on top of a block driver and leave YAFFS out
of the
>> > picture.
>>
>> Yes, but then we'd need to code all the NAND ECC, wear levelling,
etc.
>
>The fastest line between two points would probably be to run YAFFS
Direct and 
>create a "blockfile". Then write a "block driver" that you use to
access this 
>file as your FAT partition.
>
>That would give you the wear levelling as well as provide you with
YAFFS for 
>system files (ie. non FAT ones). At least then if your FAT gets hosed,
the 
>system does not turn into a brick.
>
>This would likely not be as fast (execution wise) as a
custom-engineered block 
>driver for FAT, but it should get you further down the road faster.
>
>-- Charles

If we were just talking about linux (no direct), wouldn't the loopback
driver
provide exactly this capability?

mount -t yaffs /dev/mtdblock/2 /tmp/yaffs -o sync
cp  vfat.img /tmp/yaffs
mount -t vfat /tmp/yaffs/vfat.img /tmp/vfat -o loop,sync

Chung