In time-division duplexing (TDD) massive multiple-input multiple-output (MIMO) systems, the reciprocity between the downlink (DL) and the uplink (UL) channels can be utilized to infer the DL channel state information (CSI) with the acquired uplink (UL) CSI. However, the end-to-end UL and DL channel reciprocity are broken when there are mismatches in the radio frequency (RF) analog circuits among different antennas at the base station (BS). Without accurate calibration, there will be severe system performance degradation. This paper studies the internal self-calibration scheme to recover all the unknown full calibration coefficients when different BS antennas are interconnected via hardware transmission lines. On the one hand, we derive the calibration performance for an arbitrary interconnection network in closed-form and obtain closed-form Cramer-Rao lower bound (CRLB) expressions when $(M-1)$ transmission lines are consumed and $M$ denotes the total number of BS antennas. On the other hand, we prove that the star interconnection network is optimal for internal full self-calibration due to its lowest CRLB. Numerical simulation are carried out to verify our theoretical analyses and results.