FSL SPI driver question

Norbert van Bolhuis nvbolhuis at aimvalley.nl
Thu Jan 26 00:33:37 EST 2012

On 01/25/12 01:11, Bruce_Leonard at selinc.com wrote:
> Good afternoon,
> I'm using the 3.0.3 kernel running on an MPC8308 and am trying to
> interface to a Cypress CY14B256Q2A non-volatile RAM via SPI.  I've got the
> SPI infrastructure, the Freescale SPI driver (drivers/spi/spi_fsl_spi.c),
> and spidev built into the kernel and everything on the user space/kernel
> side appears to be working correctly (at least when I try to read the
> NvRAM's config register all the right places in the kernel get hit and I
> see the SPI signals active with an o-scope).
> I think what I'm hitting is a lack of understanding/documentation on the
> SPI controller in the 8308.  To read data from the NvRAM, the Master (the
> 8308 in this case) needs to clock out a byte long "read" command, two
> bytes of address, and then clock in the data from the NvRAM.  However, I
> never get any data back.  I think the problem is that (direct quote from
> the 8308 reference manual) "SPICLK is a gated clock, active only during
> data transfers".  So once the read command and address are sent, the 8308
> considers the data transfer complete and gates off SPICLK.  Without
> SPICLK, the NvRAM has no way to clock out it's data.  I think it's ready
> to, it just can't.
> So the question is, how do I use spidev (or any other means) to get the
> 8308 SPI controller to keep SPICLK active so that the output data from the
> NvRAM gets clocked out to the 8308?
> Thanks.
> Bruce

Did you see Documentation/spi/spidev_fdx.c:do_msg ?
it perform a full-duplex (actually half-duplex) 1 byte transfer.

In your case you need a transfer that outputs 3 bytes (read cmd + address)
and inputs 1? byte.

If you do it this way I would expect the SPICLK to be active
during the 2nd part of the transfer (whenever the CPU "reads" the
data from SPI client).



More information about the Linuxppc-dev mailing list