On Tue, Mar 10, 2020 at 11:30 PM Christoph Hellwig <hch@xxxxxx> wrote: > > On Tue, Mar 10, 2020 at 08:36:14PM -0700, Darrick J. Wong wrote: > > 1) Leave the inode flag (FS_XFLAG_DAX) as it is, and export the S_DAX > > status via statx. Document that changes to FS_XFLAG_DAX do not take > > effect immediately and that one must check statx to find out the real > > mode. If we choose this, I would also deprecate the dax mount option; > > send in my mkfs.xfs patch to make it so that you can set FS_XFLAG_DAX on > > all files at mkfs time; and we can finally lay this whole thing to rest. > > This is the closest to what we have today. > > > > 2) Withdraw FS_XFLAG_DAX entirely, and let the kernel choose based on > > usage patterns, hardware heuristics, or spiteful arbitrariness. > > 3) Only allow changing FS_XFLAG_DAX on directories or files that do > not have blocks allocated to them yet, and side step all the hard > problems. This sounds reasonable to me. As for deprecating the mount option, I think at a minimum it needs to continue be accepted as an option even if it is ignored to not break existing setups. We're currently going through the prolonged flag day of people discovering that if they update xfsprogs they need to specify "-m reflink=0" to mkfs.xfs. That pain seems to have only been a road bump not a showstopper based on the bug reports I've seen. If anything it has added helpful pressure towards getting reflink support bumped up in the priority. Hopefully the xfs position that the dax mount option can be ignored makes it possible to implement the same policy on ext4, and we can just move on... > Which of course still side steps the hard question of what it actually > is supposed to mean.. If we have statx to indicate the effective dax-state that addresses the pain for applications that want to account for dax in their page cache pressure estimates, and lets FS_XFLAG_DAX not need to specify precise semantics.
