Estimation of carrier frequency offset (CFO) is a challenging task in practical systems specifically in the uplink of multiuser systems where multiple CFOs are present in the received signal. Massive MIMO as a multiuser technique has recently attracted a great deal of attention among researchers. However, to the best of our knowledge, there is no study looking into the joint estimation of CFOs and wireless channel in orthogonal frequency division multiplexing (OFDM) based massive MIMO systems. Therefore, in this paper, we propose joint estimation of multiple CFOs and the users' channel responses based on the maximum likelihood (ML) criteria in such systems. We propose to use the zadoff-chu (ZC) training sequences to reduce the implementation complexity. Additionally, utilization of ZC sequences for training simplifies the multidimensional grid search problem of estimating multiple CFOs and converts it into a set of line search problems, i.e., one line search problem per user. Also this sequence has a low peak to average ratio (PAPR). Finally, we show the efficacy of our proposed algorithm through numerical simulations.