On Thu, Mar 12, 2020 at 11:49:32AM +1100, Dave Chinner wrote: > On Wed, Mar 11, 2020 at 07:44:12AM +0100, Christoph Hellwig wrote: > > On Wed, Mar 11, 2020 at 05:39:42PM +1100, Dave Chinner wrote: > > > IOWs, the dax_associate_page() related functionality probably needs > > > to be a filesystem callout - part of the aops vector, I think, so > > > that device dax can still use it. That way XFS can go it's own way, > > > while ext4 and device dax can continue to use the existing mechanism > > > mechanisn that is currently implemented.... > > > > s/XFS/XFS with rmap/, as most XFS file systems currently don't have > > that enabled we'll also need to keep the legacy path around. > > Sure, that's trivially easy to handle in the XFS code once the > callouts are in place. > > But, quite frankly, we can enforce rmap to be enabled > enabled because nobody is using a reflink enabled FS w/ DAX right > now. Everyone will have to mkfs their filesystems anyway to enable > reflink+dax, so we simply don't allow reflink+dax to be enabled > unless rmap is also enabled. Simple, easy, trivial. Heh, this reminds me that I need to get that rmap performance analysis report out to the list... it does have a fairly substantial performance impact (in its current not-terribly-optimized state) but otoh enables self-repair. --D > Cheers, > > Dave. > -- > Dave Chinner > david@xxxxxxxxxxxxx
