Re: [PATCH v3] iio: add driver for Microchip MCP413X/414X/415X/416X/423X/424X/425X/426X

[Slawomir Stepien]

On Mar 20, 2016 19:15, Joachim Eastwood wrote:
> Hi Jonathan,
> 
> On 20 March 2016 at 18:25, Jonathan Cameron <jic23@xxxxxxxxxx> wrote:
> > On 20/03/16 16:12, Joachim Eastwood wrote:
> >>> +static int mcp4131_exec(struct mcp4131_data *data,
> >>> +               u8 addr, u8 cmd,
> >>> +               u16 val)
> >>> +{
> >>> +       int err;
> >>> +       struct spi_device *spi = data->spi;
> >>> +
> >>> +       data->xfer.tx_buf = data->buf;
> >>> +       data->xfer.rx_buf = data->buf;
> >>> +
> >>> +       switch (cmd) {
> >>> +       case MCP4131_READ:
> >>> +               data->xfer.len = 2; /* Two bytes transfer for this command */
> >>> +               data->buf[0] = (addr << MCP4131_WIPER_SHIFT) | MCP4131_READ;
> >>> +               data->buf[1] = 0;
> >>> +               break;
> >>> +
> >>> +       case MCP4131_WRITE:
> >>> +               data->xfer.len = 2;
> >>> +               data->buf[0] = (addr << MCP4131_WIPER_SHIFT) |
> >>> +                       MCP4131_WRITE | (val >> 8);
> >>> +               data->buf[1] = val & 0xFF; /* 8 bits here */
> >>> +               break;
> >>> +
> >>> +       default:
> >>> +               return -EINVAL;
> >>> +       }
> >>> +
> >>> +       dev_dbg(&spi->dev, "mcp4131_exec: tx0: 0x%x tx1: 0x%x\n",
> >>> +                       data->buf[0], data->buf[1]);
> >>> +
> >>> +       spi_message_init(&data->msg);
> >>> +       spi_message_add_tail(&data->xfer, &data->msg);
> >>> +
> >>> +       err = spi_sync(spi, &data->msg);
> >>> +       if (err) {
> >>> +               dev_err(&spi->dev, "spi_sync(): %d\n", err);
> >>> +               return err;
> >>> +       }
> >>
> >> Isn't this init, add, sync sequence basically open coding of what
> >> spi_write/spi_read does?
> >> If you could use those you could also get rid transfer/message structs
> >> in priv data.
> > I initially wrote the same comment, then realised it's more nuanced than
> > that.  Whilst this initially looks like an w8r8 type cycle it's actually
> > something like w4r12 in the read case anyway.  The write case could indeed
> > be done with spi_write.
> 
> Indeed. I didn't notice that for the read case.
> 
> The read case could almost be copy of spi_read, though. One would only
> need to add ".tx_buf = buf" when setting up the transfer struct, I
> think. Having it in its a own function with a comment would make it
> easier to spot the difference.

Just to see if I get it.

For write case I should use the spi_write as it is:

case MCP4131_WRITE:
	spi_write(...);

For read case I should create new function (e.g. mcp4131_read) that will look
like spi_read but with additional tx_buf content so I can read the data on miso?

case MCP4131_READ:
	mcp4131_read(...)

Keep the needed buffers (transfer/message) local.

-- 
Slawomir Stepien