In this article, we propose a channel feature projection clustering (CFPC) based joint channel state information and direction of arrival (JCDoA) estimation scheme for integrated sensing and communication (ISAC) massive multiple-input multiple-output (MIMO) orthogonal frequency division multiplexing (OFDM) system. Specifically, the spatial-delay domain coarse channel estimation scheme is introduced first, and then the angular-domain coarse direction of arrival (DoA) estimation is presented. In order to improve the performance of channel estimation, the angular subspace pruning (ASP) based channel state information (CSI) refinement scheme is proposed. Finally, the DoA refinement scheme based on angular-domain CFPC is proposed. The simulation results show that the proposed JCDoA scheme outperforms traditional least square (LS) and linear minimum mean square error (LMMSE) channel estimation schemes, multiple signal classification (MUSIC) and estimation of signal parameters via rotational invariance technique (ESPRIT) DoA estimation methods and their spatial smoothing (SS) enhancement schemes.