[PATCH 07/18] xfs: don't use vfs writeback for pure metadata modifications

[Christoph Hellwig]

On Wed, Sep 15, 2010 at 10:14:23AM +1000, Dave Chinner wrote:
> Oh, I missed that one in the twisty passages of the setattr code.
> I'll do another (more careful) pass across all the callers, and for
> those that are outside a transaction I'll add a separate call to
> xfs_mark_inode_dirty_sync() for them.

I'd prefer to be on the safe side - rename xfs_ichgtime to
xfs_trans_inode_chgtime or similar, and assert that the added
transaction-argument is non-zero and the inode is locked and added
to the transactions.  That way it's easier to spot which timestamp
updates are transactional and which not.