Interior wall blockage affects the performance of indoor cellular networks but has not been sufficiently studied. In this paper, by modelling the spatial distribution of interior walls using the random shape theory and incorporating the attenuation caused by wall blockages into the path loss model, we develop a tractable approach to analyse the performance of indoor small cell networks under the influence of interior wall blockages. Then, the proposed analytical approach is applied to study the downlink coverage probability of two user association strategies: the closest line-of-sight (LOS) base station (BS) strategy, and the closest non-line-of-sight (NLOS) BS strategy. The analytical results are validated by Monte Carlo simulations in comparison with those obtained under the existing wall blockage model that assumes impenetrable walls. Our results reveal the importance of considering the wall attenuation for the accurate analysis of indoor cellular networks. The coverage probability is more sensitive to wall attenuation under the closest NLOS BS user association strategy than under the closest LOS BS user association strategy, and that there is an optimal BS density that maximizes the coverage probability under the closest LOS BS user association strategy, but the coverage probability under the closest NLOS BS user association strategy monotonically decreases with the increase of BS density.