On Wed, Sep 13, 2006 at 07:10:10PM +0400, Vitaly Wool wrote:
> So, the bottom line is the following.
> I suggest you to try these tests:
>
> 1. Create a blank YAFFS filesystem under 2.6.18-git w/ my patch and
> verify it works (that's what I did recently again :) ). That is,
> mount, copy, unmount, diff.
>
> I hope the flash chip you've got is _not_ with syndrome-based ECC,
> otherwise ping me for a patch to current nand_base -- it's broken for
> syndrome case.

Fell at the first hurdle. 2.6.18-rc5 kernel, CONFIG_YAFFS_DOES_ECC is
not set:

root@balloon2:~$ flash_eraseall /dev/mtd4
root@balloon2:~$ mount /dev/mtdblock4 /media/mtdblock4
yaffs: Attempting MTD mount on 31.4, "mtdblock4"
block 4320 is bad
block 7547 is bad
block 8136 is bad
block 10184 is bad
yaffs tragedy: existing chunk < 0 in scan
(repeat at least 390945 times)
root@balloon2:~$ df -h
Filesystem                Size      Used Available Use% Mounted on
/dev/ram0                 7.9M      5.9M      1.6M  78% /
/dev/ram0                 7.9M      5.9M      1.6M  78% /dev/.static/dev
tmpfs                    10.0M     12.0k     10.0M   0% /dev
/dev/mtdblock4          191.0M    191.0M         0 100% /media/mtdblock4

The flash chips are 3 * K9F1208U0Ms ie 64Mx8 with 512 byte page size
(which means we're using YAFFS format rather than YAFFS2 I think?). ECC
is being done in software by the normal MTD bits.

> 2. Create a blank YAFFS filesystem under <2.6.17 *w/
> CONFIG_YAFFS_DOES_ECC unset*. Reboot with 2.6.18-git w/ my patch and
> verify it works.

J.

-- 
Web [        noodles is almost a national pastime in japan         ]
site: http:// [                                          ]       Made by
www.earth.li/~noodles/  [                      ]         HuggieTag 0.0.23