We consider a multi-cell massive multiple-input-multiple-output (MIMO) system with correlated Rayleigh fading channels, where pilot reuse is permitted within each cell (to reduce pilot overhead), and each base station (BS) utilizes multi-cell minimum mean square (M-MMSE) precoders and combining. We derive a large-scale approximation of the uplink signal-to-interference-and-noise ratio (SINR) that is asymptotically tight in the large system limit. By leveraging the derived SINR approximation, we (i) develop a low-complexity multi-cell pilot assignment (PA) scheme aimed at minimizing pilot contamination from pilot-sharing users, via effectively exploiting the channel spatial correlation matrices of all users in the network, (ii) design pilot and data power allocation schemes, using both the weighted sum and max-min spectral efficiency (SE) metrics. Simulations demonstrate the superiority of our multi-cell PA scheme, requiring significantly less pilot overhead. The proposed power allocation schemes also achieve substantial sum SE gains with good fairness among users compared to equal power allocation.