Last year, I led a discussion at LSF about the inode change cookie value (aka i_version). The original impetus for this work was the fact that all 3 mainstream local filesystems (btrfs, ext4 and xfs) updated the value for different sorts of activity, which can cause wildly different behavior when the filesystem is exported via NFS. My hope last year was that we could fix this with multigrain timestamps [1], but that didn't pan out. At LSF, I'd like to give a brief overview of state of i_version work, and lead a discussion on what work remains. [1]: https://lwn.net/Articles/938240/ Thanks! -- Jeff Layton <jlayton@xxxxxxxxxx>
