On Grosswald's conjecture on primitive roots