[JFFS2] Garbage collector says "Checked all inodes but still 0x5e3d0 bytes of unchecked space?"

[Fabrice GIRARDOT]

Hi all.

I'm using a JFFS2 root filesystem on a 2.6.12 Linux kernel,
powered by an ARM9 cpu.
My storage hardware is a 128 MBytes NAND device.

Sometimes, after several unclean system resets, JFFS2's garbage
collector outputs this message :
"Checked all inodes but still 0x5e3d0 bytes of unchecked space?
  kernel BUG at fs/jffs2/gc.c:139!"

As far as I understand, the GC has checked the last inode,
but has still some unchecked space. Since this should not occur,
the code in jffs2's file gc.c reaches a "BUG()" statement.

Some search on this problem did not provides me with much
information.
I found a message in 2002 where the problem could be an hardware
problem (same data not read twice identically), and an other
one sent in 2006 where the "unchecked_size" may be badly computed.

I don't know which exact version of jffs2 I'm using ; my gc.c
file is version 1.144 on 21-dec-2004, which may look a bit old.
Unfortunately, I can not switch to a newer kernel version.

Does somebody has more information on this problem ?
Is it a bug in jffs2 and/or its GC ? Is it fixed ?
Could it be an harwdare problem, confusing jffs2 ?

Any information is welcome.


Regards,

-- 
Fabrice GIRARDOT