On Wed, Nov 26, 2014 at 12:09:09PM +0200, Laurentiu Palcu wrote:

> +static int dln2_spi_transfer_one_message(struct spi_master *master,
> +					 struct spi_message *msg)
> +{
> +	struct dln2_spi *dln2 = spi_master_get_devdata(master);
> +	struct spi_device *spi = msg->spi;
> +	struct spi_transfer *xfer;
> +	int status;
> +
> +	list_for_each_entry(xfer, &msg->transfers, transfer_list) {
> +		u8 attr = 0;
> +
> +		status = dln2_spi_transfer_setup(dln2, xfer->speed_hz,
> +						 xfer->bits_per_word,
> +						 spi->mode);
> +		if (status < 0) {
> +			dev_err(&dln2->pdev->dev, "Cannot setup transfer\n");
> +			break;
> +		}
> +
> +		if (!xfer->cs_change &&
> +		    !list_is_last(&xfer->transfer_list, &msg->transfers))
> +			attr = xfer->cs_change ? 0 : DLN2_SPI_ATTR_LEAVE_SS_LOW;
> +
> +		status = dln2_spi_rdwr(dln2, xfer->tx_buf, xfer->rx_buf,
> +				       xfer->len, attr);

We recently added a spi_transfer_is_last() operation which should allow
you to move this to using transfer_one() which is a bit nicer.

> +	/* enable SPI module, we're good to go */
> +	ret = dln2_spi_enable(dln2, true);
> +	if (ret < 0) {
> +		dev_err(&pdev->dev, "Failed to enable SPI module\n");
> +		goto exit_free_master;
> +	}

It would be nice to have runtime PM operations which keep the device
disabled when idle, presumably that'd save a bit of power and it's
generally better practice.