In strong line-of-sight millimeter-wave (mmWave) wireless systems, the rank-deficient channel severely hampers spatial multiplexing. To address this inherent deficiency, distributed reconfigurable intelligent surfaces (RISs) are introduced in this study to customize the wireless channel. Capitalizing on the ability of the RIS to reshape electromagnetic waves, we theoretically show that a favorable channel with an arbitrary tunable rank and a minimized truncated condition number can be established by elaborately designing the placement and reflection matrix of RISs. Different from existing works on distributed RISs, the number of elements needed for each RIS to combat the path loss and the limited phase control is also considered in this research. On the basis of the proposed channel customization, a joint transmitter-RISs-receiver (Tx-RISs-Rx) design under a hybrid mmWave system is investigated to maximize the downlink spectral efficiency. Using the proposed scheme, the optimal singular value decomposition-based hybrid beamforming at the Tx and Rx can be easily obtained without matrix decomposition for the digital and analog beamforming. The bottoms of the sub-channel mode in the water-filling power allocation algorithm, which are conventionally uncontrollable when the noise power is fixed, are proven to be independently adjustable by RISs. Moreover, the transmit power required for realizing multi-stream transmission is derived. Numerical results are presented to verify our theoretical analysis and exhibit substantial gains over systems without RISs.