Distributed massive multi-input-multi-output (mMIMO) is a promising architecture which has potential to satisfy the strick latency requirement in Internet of Things (IoT). To further meet the low-cost and low-latency demand in IoT, this paper provides a low-complexity scheme to the access phase for mixed analog-to-digital convertors (ADC) distributed mMIMO. which consists of two steps. In the first step, the clustering behavior among users is detected using large scale fading information, which aims to reduce the complexity. In the second step, with the number of clusters as a priori, a weighted minimum mean square error (WMMSE) clustering algorithm that can provide stable and robust results is proposed. The clustering algorithm aims to maximize the achievable sum rate, in which the nonconvex objective function and constraints are modeled using ell1-norm approximation. Numerical results show that the proposed algorithm has strong convergence, and significant gain can be obtained in various scenarios.