the newly merged driver for the mrf24j40 is broken on the stm32f4discovery. From quick analyis, the driver hangs during initialization while setting the channel, explicitly it loops in the function mrf24j40_reset_state_machine(). IMHO, this is not an SPI issues (as we expected earlier), as all SPI transfers before that, and also this specific one look just fine on the logic analyzer. I expect rather something to be broken in the initialization code, but won't have time to look into this any deeper myself...