[ale] Ext4 kernel bug

[Ed Cashin]

On Wed, Oct 24, 2012 at 9:57 PM, Sean Kilpatrick <kilpatms at gmail.com> wrote:
> I'm outta my depth here.  How serious is this bug?
>
> http://www.phoronix.com/scan.php?page=news_item&px=MTIxNDQ
>
> Is there anything I  can do to guarantee the FS journal is complete before
> a shutdown/reboot?

I don't know, but I was interested by your question and found out that
when I create a new ext4 filesystem and grep for "Journal" in the
dumpe2fs output, it says Journal sequence 1.

Then I mount it and it changes to two.  Then I watch the output of,

  echo logdump | debugfs /dev/loop0

... while copying and removing many files in another terminal, doing
sync between each, and I see that it takes a while for the journal to
"wrap", but when it does, the journal sequence number changes to
0x970.

So this answer is too incomplete and technical to help, but I bet you
could force the journal to wrap by repeatedly doing something like
this:

for i in `seq 0 999`; do cp -a /usr/share/doc/e2fsprogs-1.41.12
/mnt/loop0/asd && sync && rm -rf /mnt/loop0/asd && sync; done

... until the sequence number shown by the command below changes:

dumpe2fs /dev/loop0 | grep 'Journal seq'

... and maybe this answer is specific enough that you could ask for
verification in the ext mailing list.  I glanced at the sources but
couldn't see anything to confirm my hunch that the journal sequence
number always changes on fs creation, journal wrap, and mount but not
in the absence of those things.

-- 
  Ed Cashin <ecashin at noserose.net>
  http://noserose.net/e/
  http://www.coraid.com/