On Sat, Dec 2, 2017 at 11:40 PM, Dave Chinner <david@xxxxxxxxxxxxx> wrote: > On Sat, Dec 02, 2017 at 12:43:18PM +0200, Amir Goldstein wrote: >> On Fri, Dec 1, 2017 at 11:43 PM, Dave Chinner <david@xxxxxxxxxxxxx> wrote: [...] > > Implement FS_IOC_SHUTDOWN on overlay, then you can test it via > software controlled shutdown. i.e. make overlay return true for > _require_scratch_shutdown() and implement what is needed to shut it > all down and prevent further writes of dirty inodes. Thens do > data/metadata checks after a shutdown/unmount/mount cycle. Then > what you have on disk after the mount cycle is what was written by > sync/fsync/syncfs... > > We've only been using this shutdown mechanism to test fsync/sync > mechanisms on XFS for ~20 years. > All right. We don't need to implement FS_IOC_SHUTDOWN on overlay, we only need to shutdown upper fs if upper fs supports shutdown. Doing that would be the equivalent of a dm-flakey test for a blokdev fs, because upper fs is the only persistent storage of overlayfs. Thanks for the advice. Chengguang, I propose that you teach _require_scratch_shutdown about OVL_BASE_SCRATCH_MNT and write and overlay/* test that tests syncfs by shutting down upper fs if upper fs supports shutdown. It doesn't matter much that the test won't run on all underlying fs as long as a regression will be caught by someone running overlay tests over xfs/ext4. Cheers, Amir. -- To unsubscribe from this list: send the line "unsubscribe fstests" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html