+++ Jim Rayner [2009-03-14 14:54 +0000]:
>    Kernel 2.6.29-rc7 Yaffs cvs 05/03/2009 Samsung Nand 256MB hw: Balloon3
>    board
> 
>    I have a recurring set of segfaults with 2.6.29-rc7 and yaffs 05/03/09
>    whilst trying to recover from a large failure.
> 
>    I have just moved up from 2.6.25.2 (with checkpointed root partition yaffs
>    cvs 05/03/09 with 7 bad blocks) to 2.6.29-rc7. On the initial mounting
>    several thousand errors were produced (yaffs cvs 03/03/09), I moved yaffs
>    up to the 05/03/09 cvs version to be consistent but without any reduction
>    in messages.
> 
>    Attempting to do rm -rf /mnt/root results in differing segfaults, the
>    initial one logged below. The same applies when returning to 2.6.25.2 with
>    yaffs 05/03/09. The partitions cannot be unmounted (busy) after the
>    segfaults.
> 
>    I am able to do rm -rf and umount using a recovery initrd running 2.6.25.2
>    & yaffs cvs 13/12/07, returning to yaffs 05/03/09, either kernel, and
>    executing rm -rf causes the same segfaults.
> 
>    I have logs I can email but too large for the ML limit. Initial bits of
>    importance pasted below.
> 
>    Any advice gratefully appreciated,

Hmm, not sure what to make of this lot. I did exactly the same thing
with a couple of boards and this didn't happen. It does sound like
something is not quite right. I guess some more YAFFS TRACING might
help clear up exacty what is falling over.

If you email me (or sned links) to your logs that would be helpful,
and I'll try the transition on a few more boards and see if I can
reproduce it.

I've just made 2.6.29 and yaffs-beginning-of-march the default in
ballon SVN so it would be good to try and get to the bottom of this.
Noodles saw something very similar but put it down to building loads
of different kernels/yaffs variants and accidentally doing something
bad, but now we have two instances that suggests a more significant
issue. 

Wookey
-- 
Principal hats:  iEndian - Balloonboard - Toby Churchill - Emdebian
http://wookware.org/