In Mobile Crowd Sensing (MCS), mobile users usually can be both the contributor of sensing data and the customer of the service provided by the data collector or service provider. Most of existing work does not consider the dual-role of mobile users in MCS. In this work, by jointly considering users' dual-role as both the contributor of sensing data and the customer of service, we design a Lyapunov based online algorithm to achieve the stability of dynamic MCS system while maximizing the platform utility. Meanwhile, we demonstrate the impacts of parameter on the balance of these two roles of mobile users. Simulation results show that the proposed method outperforms some existing methods in terms of achieved platform utility and system stability. In addition, it is proved that, the time-averaged platform utility could converge to the maximum benchmark asymptotically.