raspberrypi: SPI driver has several flaws

[Jan Sommer]

I started to work with the SPI of the raspberry pi and noticed several problems with the current driver after recording the frames with a logic analyzer:

• Setting the chip select signal using the CSPOL-bits of the control register creates incorrect chip select signals
• After every byte written to the TX FIFO buffer the driver waits for a signal from an interrupt. This creates very long pauses between consecutive bytes.
• The lsb-first mode does not look as if it would work
• 3-wire/bidirectional mode might not work
• Accessing the driver from a different thread than the one which called spi_init will cause a deadlock
• Simultaneous read-write-operation is not supported

I will submit a patchset for this after some testing.