Testing SPI

Well, the way I'd test it (and have tested it many times) is
to hook a scope probe up to each of the following pins:
 - SPISIMOx
 - SPISOMIx
 - SPICLKx
 - SPISTEx .

where x is a letter signifying which SPI port you are using.

The last applies only if you are using the built-in slave

select line to select your SPI slave. If you are using some
other pin, substitute it where I say "slave select."

I would trigger on the slave select line, run my SPI
application, and look for waveforms that resemble those in
the appropriate section of your SPI Reference Guide.

If you don't have a 4-channel scope, I would put my trigger
probe on the slave select line and one on each of the other
3, and test them one at a time.

If your board is made of 100% SMT parts and you didn't place
intentional test points on those lines for testing, I have
had good luck at soldering AWG 28 wire into vias in the
traces of the lines you want to measure, and clipping scope
probes to them. 

If a trace has no vias, you can try soldering to the chip
pins, but that requires a lot more skill and hand/eye
co-ordination.

If you see no output on the SPI lines while application is
running, I would set a breakpoint right before the return
point in whatever function initializes the SPI port, and
when the IDE hits it, go in and examine all the registers
changed by the function, verifying they match what I just
wrote to them.

This has worked for me, at least.