On Tue, Dec 29, 2020 at 10:49:55AM +0300, Dmitrii Tcvetkov wrote: > >Please pull the following branch containing all the new xfs code for > >5.11. In this release we add the ability to set a 'needsrepair' flag > >indicating that we /know/ the filesystem requires xfs_repair, but other > >than that, it's the usual strengthening of metadata validation and > >miscellaneous cleanups. > >... > >New code for 5.11: > >- Introduce a "needsrepair" "feature" to flag a filesystem as needing a > > pass through xfs_repair. This is key to enabling filesystem upgrades > > (in xfs_db) that require xfs_repair to make minor adjustments to > >metadata. > > Hello. > > Most likely I miss something obvious but according to xfs_repair(8): > BUGS: > The filesystem to be checked and repaired must have been unmounted > cleanly using normal system administration procedures (the > umount(8) command or system shutdown), not as a result of a > crash or system reset. If the filesystem has not been unmounted > cleanly, mount it and unmount it cleanly before running xfs_repair. > > which is there since commit d321ceac "add libxlog directory" > Date: Wed Oct 17 11:00:32 2001 +0000 in xfsprogs-dev[1]. > > So will be there situation of uncleanly unmounted filesystem with > "needsrepair" bit set? No. If we detect metadata corruption we stop writing metadata and take the fs offline immediately. We would not set needsrepair, for exactly the reasons you outline. --D > Will one be able to mount and umount it before running xfs_repair in > that case? > > [1] git://git.kernel.org/pub/scm/fs/xfs/xfsprogs-dev.git
