Comment 9 for bug 1902588

Steps used to verify the fix (both in focal and groovy):

$ fallocate -l 8G /home/ubuntu/disk.img
$ sudo zpool create pool /home/ubuntu/disk.img
$ for i in {1..20}; do sudo zfs create pool/ds$i; done
$ sudo zfs unmount pool/ds20
$ sudo zfs mount -a & sudo zfs mount -a &
[1] 1964
[2] 1965
ubuntu@groovy:~$ filesystem 'pool/ds20' is already mounted
cannot mount 'pool/ds20': mountpoint or dataset is busy

[1]- Done sudo zfs mount -a
[2]+ Exit 1 sudo zfs mount -a

With the fix applied we should be able to see the error "mountpoint or dataset is busy", due to the concurrent "zfs mount -a" running. That means the old behavior has been restored, since it's not a problem anymore for systemd (and the applied workaround could cause the segfault).