Re: [PATCH 25/30 V2] libxfs: fix root inode handling inconsistencies

[Christoph Hellwig <hch@xxxxxxxxxxxxx>]

On Tue, Nov 05, 2013 at 10:22:42AM +1100, Dave Chinner wrote:
> > xfs_initialize_perag_data does the following:
> > 
> >  - read in AGI/AGF.  This is something we lazily do whenever we need it
> >    anyway, so no one should rely on it.
> >  - update the in-core superblock global counters.  Seems like the old
> >    xfs_check relies on this and still needs an equivalent if we care
> >    enough.  No one else seems to care.
> 
> So you are suggesting that I move that initialisation to the
> xfs_check code rather than just doing it in the mount code? Or
> something else?

That seems to be the sanest option if it works out.  Otherwise we'd need
to keep a flag to let libxfs do it.