Re: Review: remount read-only path is as broken as freezing was....

[Christoph Hellwig <hch@xxxxxxxxxxxxx>]

On Mon, Jun 04, 2007 at 03:14:33PM +1000, David Chinner wrote:
> 
> I recently had a remount,ro test fail in a way I had previously
> only seen freezing fail. That is, it failed because we still
> had active transactions after calling xfs_quiesce_fs(). Further
> investigation shows that this path is broken in the same ways
> that the xfs freeze path was broken (and recently fixed).

Actually it became more broken due to the fix changing things
a little.

In general a mount -o ro should be similar to a quience in
most ways, so the closer we can get them the better.  The patch
looks good to me.  One thing is that SYNC_INODE_QUIESCE should
probably get a more descriptive name and a comment explaining
why it's needed for quience (we want inodes updated not only in
the log) and why not for remount r/o (because no one cares for that
in this case)