Channel estimation is one of the main challenges for reconfigurable intelligent surface (RIS) assisted communication systems with passive reflective elements due to the high number of parameters to estimate. In this paper, we consider channel estimation for a MIMO FD RIS-assisted wireless communication system and use tensor signal modelling techniques to estimate all channel state information (CSI) involving the self-interference, direct-path, and the RIS assisted channel links. We model the received signal as a tensor composed of two CANDECOMP/PARAFAC (CP) decomposition terms for the non-RIS and the RIS assisted links. Based on this model we extend the alternating least squares algorithm to jointly estimate all channels, then derive the corresponding Cramér-Rao Bounds (CRB). Numerical results show that compared to recent previous works which estimate the non-RIS and RIS links during separate training stages, our method provides a more accurate estimate by efficiently using all pilots transmitted throughout the full training duration without turning the RIS off when comparing the same number of total pilots transmitted. For a sufficient number of transmitted pilots, the proposed method's accuracy comes close to the CRB for the RIS channels and attains the CRB for the direct-path and self-interference channels.