[Yaffs] Umount and checkpointing problem

Hello,

I'm using a yaffs partition on a EABI arm920t platform and having strange
problems while umounting my partition.
In this case I'm using BusyBox v1.17.4 with the uClibC 0.9.31 for my umount
binary.
I'm looking for help, I can't find the possible reasons... I suppose it
could comes from my yaffs partition or busybox umount applet. I have asked
them for help.
Meantime, I'm exploring the possibility my yaffs is corrupted.

I have the following command in my linuxrc, that mounts my yaffs system

/bin/mount -t yaffs /dev/mtdblock6 /usr

which gets me the following system messages :

yaffs: dev is 32505862 name is "mtdblock6"
yaffs: passed flags ""
yaffs: Attempting MTD mount on 31.6, "mtdblock6"
yaffs: auto selecting yaffs2
yaffs: restored from checkpoint
yaffs_read_super: isCheckpointed 1

In my case (a few seconds later) one of my scripts starts the following
command :

/bin/umount /usr || return 1

For the record, in this particular case, I just want to umount my yaffs
filesystem to flash erase it and get something brand new :)
And before I have the time to flash erase it, that returns these errors :

umount: can't umount /usr: Device or resource busy
save exit: isCheckpointed 1
umount: /dev/mtdblock6 busy - remounted read-only
umount: none busy - remounted read-only

"isCheckpointed 1" rings a bell actually. But I can't really find specs that
explain checkpointing. Except here :
http://www.yaffs.net/lurker/message/20070327.193931.32818f24.pl.html But It
doesn't really help me.
Could someone explain to me what the purpose of checkpointing flag ? I could
imagine there is something wrong with my partition. But I don't really
understand what happened here.

The problem is quite embarrassing because I can't change my system here. I
just want to understand what could have happened :/

Last thing I could add is that I have the following system message while
creating my mtd partitions :

mtd: partition "yaffs" extends beyond the end of device "NAND 256MiB 3,3V
8-bit" -- size truncated to 0xeac0000

Could this be related ?

Thanks,

Yoann Sculo