Ocfs2 devel - Sep 2011 - block64 failure

All,

So the patches added to allow mounting volumes > 16TB has a problem.
The feature check of the jbd2 superblock is being done before the
jbd2 superblock is actually read.

It is being done after the journal_init_inode() which is incorrect.

ocfs2_check_volume()
   ocfs2_journal_init()
        jbd2_journal_init_inode()
   ocfs2_journal_addressable() <=== check
...

Currently the journal sb is first read during load_journal. Too late for us.
And journal_get_superblock() is not exported.

One solution is to call jbd2_journal_update_format() after
jbd2_journal_init_inode().
This function is used by extX to update the journal format from V1 to V2. In our
case, we are always V2 so will be a null op. But will force read the journal sb.

Icky... hence my post.

The other option is exporting journal_get_superblock() to load the journal sb.

Suggestions?

Sunil

On Wed, Sep 07, 2011 at 10:10:43AM -0700, Sunil Mushran
wrote:
> All,
> 
> So the patches added to allow mounting volumes > 16TB has a problem.
> The feature check of the jbd2 superblock is being done before the
> jbd2 superblock is actually read.
> 
> It is being done after the journal_init_inode() which is incorrect.
> 
> ocfs2_check_volume()
>   ocfs2_journal_init()
>        jbd2_journal_init_inode()
>   ocfs2_journal_addressable() <=== check
> ...
> 
> Currently the journal sb is first read during load_journal. Too late for
us.
> And journal_get_superblock() is not exported.
	I'm confused.  jbd2_journal_check_used_features() reads the
journal superblock if needed.  Why does this not satisfy?

Joel

-- 

"The one important thing i have learned over the years is the
 difference between taking one's work seriously and taking one's self
 seriously.  The first is imperative and the second is disastrous."
	-Margot Fonteyn

			http://www.jlbec.org/
			jlbec at evilplan.org