A forced reboot left filesystem in inconsistent state (possibly consumer grade SSD issue). It would mount r/w but exhibited runtime errors: some recently modified were malformed; attempted writes were rejected: >NILFS (sdb2): vblocknr = 101945318 has abnormal lifetime: start cno (= 534472) > current cno (= 534431) >NILFS error (device sdb2): nilfs_bmap_propagate: broken bmap (inode number=4) >Remounting filesystem read-only I tried creating a snapshot from recent checkpoint to read recent good verison of files, but that was also rejected with the above error message, and the snapshot reverted to checkpoint upon subsequent mount. Turns out there's a short (~5 seconds) window after "chcp ss ..." when the snapshot can be mounted - about 5 seconds, before the fs driver attempts to create a new checkpoint, ending up in error condition. $ umount /dev/sdb2; mount /dev/sdb2 /mnt/test # to clear out error condition $ chcp ss /dev/sdb2 534430; mount /dev/sdb2 -o ro,cp=534430 /mnt/recover # set up snapshot and immediately mount Hope this helps somebody, maybe even me, in the future. Thanks, -- dx
