To reproduce, set up a zfs with snapshots like so: {{{ # zfs list -t all -r tank/.......1 NAME USED AVAIL REFER MOUNTPOINT tank/.......1 19K 769G 19K /.......1 tank/.......1@....3.........4.........5.........6.........7.........8........ 0 - 19K - tank/.......1@....3.........4.........5.........6.........7.........8......... 0 - 19K - tank/.......1@....3.........4.........5.........6.........7.........8.........9.........0 0 - 19K - }}} These names are carefully calculated to hit or exceed the 88 character limit on mountpoint path lengths when the snapshots are automounted Now, doing something that would cause the snapshots to automount leads to the expected 'name too long' errors: {{{ # find /.......1/.zfs/snapshot /.......1/.zfs/snapshot find: /.......1/.zfs/snapshot/....3.........4.........5.........6.........7.........8.........: File name too long /.......1/.zfs/snapshot/....3.........4.........5.........6.........7.........8........ /.......1/.zfs/snapshot/....3.........4.........5.........6.........7.........8......../aaa /.......1/.zfs/snapshot/....3.........4.........5.........6.........7.........8......../aaa/bbb /.......1/.zfs/snapshot/....3.........4.........5.........6.........7.........8......../aaa/bbb/ccc find: /.......1/.zfs/snapshot/....3.........4.........5.........6.........7.........8.........9.........0: File name too long }}} This is all as expected. Problem is that now the tank/.......1 zfs is left in a busy state: {{{ # zfs destroy -r tank/.......1 cannot unmount '/.......1': Device busy }}} Needs a forced unmount to do anything with it: {{{ # zfs umount -f /.......1 }}} I'd expect that other than an error message there wouldn't be any side effects of failing to automount a snapshot due to overflowing the maximum allowed path length.