On Thu, Apr 02, 2020 at 01:49:30AM -0700, Christoph Hellwig wrote: > On Thu, Apr 02, 2020 at 12:51:08AM -0700, Christoph Hellwig wrote: > > On Wed, Apr 01, 2020 at 09:17:05PM -0700, Darrick J. Wong wrote: > > > From: Darrick J. Wong <darrick.wong@xxxxxxxxxx> > > > > > > Reflink should force the log out to disk if the filesystem was mounted > > > with wsync, the same as most other operations in xfs. > > > > Looks reasonable. That being said I really hate the way we handle > > this - I've been wanting to rework the wsync/dirsync code to just mark > > as transaction as dirsync or wsync and then let xfs_trans_commit handle > > checking if the file system is mounted with the option to clean this > > mess up. Let me see if I could resurrect that quickly. > > Resurrected and under testing now. While forward porting your patch > I noticed it could be much simpler even without the refactor by just > using xfs_trans_set_sync. The downside of that is that the log force > is under the inode locks, but so are the log forces for all other wysnc > induced log forces. So I think you should just submit this in the > simplified version matching the rest of the wsync users as a fix. If > we want to optimize it later on that should be done as a separate patch > and for all wsync/dirsync users. Can you please send in a Reviewed-by so I can get this moving? :) --D
