Status of NAND Flash JFFS2 Code

[Rob.McConnell at Zarlink.Com]

Hi Folks,

I am having problems running JFFS2 on NAND flash devices with dodgy blocks
(Toshiba TC58V64AFT).

I started with the  MVista HHL2.0 - 2.4.2 kernel. I updated the mtd code to
the newest available from the CVS repository. This code appears to work
correctly. I can read/write and erase the NAND flash device correctly using
mtd char access. I have also created a block device and this seems to allow
me to "dd" it.

I have tried several version of the JFFS2 code but  I always encounter the
same problems. I started with the JFFS2 code from HHL 2.1 (2.4.17 kernel).
I have also obtained the CVS branches JFFS2-2_4, JFFS2-NAND_BRANCH and
JFFS2-NAND-BRANCHPOINT. I have  back-ported each of these versions into my
2.4.2 kernel and compiled the JFFS2 fs and mtd drivers into the kernel.

This is the sequence of events leading to the problems:

a) Erase NAND flash partition 0 using the "eraseall" utility (./eraseall
/dev/mtd0 and verify that the erase has worked (reading back FF's from all
erased blocks).

b) mount the JFFS2 on the mtdblock0 device (mount -t jffs2 /dev/mtdblock0
/nand/p0)

     This identifies that the NAND flash device is not formatted for JFFS2
and proceeds to do its business. It then comes back to the command prompt
and I can check the /proc/mounts file to see whether it has really been
mounted.

c) I then create a simply file by echo "Hello World" > /nand/p0 and then
sync.  I can see this file in the directory and can read it back.

d) I can also copy this small file to another file or even create another
"Hello World2" file in the same directory without any problems.

e) If I copy a large file onto the JFFS2 fs (e.g. eraseall > 400K), I get
problems when I try to read it back or list the dir contents or even if I
try to execute it.

I get the message "Error reading node from 0x????????: ..." which is
printed from the function "jffs2_read_dnode" in read.c

f) I can't unmount the device (it's busy) and if I then reboot and try to
mount again it fails.

If anyone has any suggestions, then they would be most welcome!

TIA,

Rob ;-)