It is well-known that any finite $p$-group in which all its abelian subgroups are cyclic is either a cyclic group or a generalized quaternion group. 

What can be said about $p$-groups in which every normal abelian subgroup is cyclic?