Despite the rapid progress in device efficiency of the perovskite solar cells (PVSCs), the origin of commonly observed large variability of device performance from one-step to two-step approaches is still elusive. Herein, the processing windows of perovskite films prepared by one-step and two-step processes have been investigated. We find that both perovskite films prepared by one-step and two-step methods in air are much rougher than that prepared in glovebox. For the one-step method, complete perovskite conversion can be readily achieved both in air and in glovebox, but totally different in surface morphology. For the two-step method, residual PbI2 are remained in perovskite film, particularly with larger amount for the film prepared in air. Both rough surface and incomplete conversion hinder the realization of high quality perovskite film in air. The photovoltaic performances of the PVSCs prepared by one-step and two-step methods in glovebox are systematically compared. The generally observed lower photovoltaic performance of two-step PVSCs is ascribed to the additional trap-assisted recombination process due to larger trap density in the two-step perovskite film. This is confirmed by carrier recombination study using steady state photoluminescence, transient photovoltage and photocurrent measurements. Finally, combining the high work function hole transporting material poly-TPD, and the high-quality one-step perovskite film, a 17.5% high-efficiency PVSCs can be achieved. The results reveal the major impacts on the morphological and electronic properties of perovskite films between one-step and two-step method, and it brings insight into the strategy for extending the processing windows to achieve highly reliable PVSCs.