On Tue, Oct 04, 2016 at 11:56:05PM +0200, Hauke Mehrtens wrote:
> On 10/04/2016 05:13 AM, Mark Brown wrote:
> >> +static void hw_finish_message(const struct intel_ssc_spi *spi)
> >> +{
> >> + /* Disable interrupts */
> >> + intel_ssc_spi_writel(spi, 0, SPI_IRNEN);
> >> +
> >> + /* Disable transmitter and receiver */
> >> + intel_ssc_spi_maskl(spi, 0, SPI_CON_TXOFF | SPI_CON_RXOFF, SPI_CON);
> >> +}
> > This needs to be done per message or just when the hardware idles?
> I can not find anything suggesting that this is needed after every
> message, just when the hardware is idle we should do it.
Make it an unprepare_transfer_hardware() then and let the core figure
out when to call it.
> Thanks for the hint, I will set bits_per_word_mask. Then I do not have
> to manually check if the mask is supported any more, is that correct?
Yes.
> >> +static int intel_ssc_spi_transfer_one_message(struct spi_master *master,
> >> + struct spi_message *msg)
> > Why not implement transfer_one()? This seems to just duplicate the
> > standard implementation.
> Yes I will look into this. The chip documentation says that we should
> not change the clock while a chip select is active becasue this could be
> interpreed as data by the device. How should I handle the speed_hz
> member in the transfer?
Return an error (you need to do that regardless of open coding or not).
Attachment:
signature.asc
Description: PGP signature
