We used YAFFS2 in a Linux-based product (with a Marvell Sheeva core and
NAND) recently with great success. I had to tweak mkyaffs2 a bit to
re-position the ECC.

Now we have a third party's design with a PowerPC and a Parallel 64MB
Spansion S29GL512P NOR. Again we're getting some bad behaviour from JFFS2
though only some months in.

>From web pages and this mailing list I understand YAFFS2 does work with
NOR but some things are unclear.

Some code code changes are documented on the YAFFS direct page. Does that
mean you only expect it to work with NOR when the direct interface is used
or just happened to document it there ? Or it should it work automatically
in the non-direct case with what the MTD layer presents ?

We pass mtdparts=physmap-flash.0:0x3c00000(root) to the kernel and mount
JFFS2 on /dev/mtdblock0 which works as far as JFFS2 works.

I've built in version aleph1-release of YAFFS2 on top of 2.6.23, erased
the NOR it and mounted mtdblock0 -t yaffs2 -o inband-tags .

(Without inband-tags it complains "MTD device does not have the right page
sizes" which seems reasonable.)

But it isn't happy as follows.

yaffs_tags_marshall_read+0x94/0x280
yaffs2_checkpt_rd+0x2c0/0x3d4
yaffs2_rd_checkpt_validity_marker+0x20/0x90
yaffs2_checkpt_restore+0x11c/0x650
yaffs_guts_initialise+0x6c8/0xa08
yaffs_internal_read_super+0x460/0x9f8
yaffs2_internal_read_super_mtd+0x24/0x44
get_sb_bdev+0x164/0x1a8
yaffs2_read_super+0x1c/0x2c

What should I look at next ?

Thanks,

Jon