You first route the requests from SPI to DMA in DMAMUX,. Then you set the respective DMA Streams' registers (peripheral/memory pointer, NDTR, and the control register) and enable them in the control register. This of course assumes you've already set the DMA's clock in RCC.
Provided the SPI is set up properly, it's enabled for Tx/Rx DMA, this is it. As soon as the Tx DMA is enabled, it transfers enough frames to keep the SPI FIFO full. Then, when SCK clocks arrive, the data are transmitted and DMA fills the succeeding bytes until its NDTR gets zero, unless set to circular. Similarly, received frames fill the FIFO, and when the set threshold is reached, they get transferred by the DMA assigned to Rx into its destination as set in there.
You may want to read the DMA, DMAMUX and SPI chapters in the RM.
I don't Cube.
JW
