Re: PATCH] xfs: implement .dirty_inode to fix timestamp handling

[Christoph Hellwig <hch@xxxxxxxxxxxxx>]

On Tue, Sep 01, 2009 at 02:08:20PM -0400, Christoph Hellwig wrote:
> I managed to trigger the ASSERT in the reclaim path, so it looks both
> this version and our previous code is buggy.  It's back to the drawing
> board for now until I gifure out what's going on.

Okay, figured it out - didn't actually test the asserts as the were
added last minute.  Looks like we could always (before and after the
patch) get into the reclaim path with the timestamp in the xfs inode not
uptodate, and due to the I_CLEAR check we would never update it.