Re: [PATCH v3 1/3] iio: adc: add support for mcp3911

[Jonathan Cameron <jic23@kernel.org>]

On Thu, 2 Aug 2018 22:52:00 +0300
Andy Shevchenko <andy.shevchenko@gmail.com> wrote:

> On Thu, Aug 2, 2018 at 10:15 PM, Marcus Folkesson
> <marcus.folkesson@gmail.com> wrote:
> > MCP3911 is a dual channel Analog Front End (AFE) containing two
> > synchronous sampling delta-sigma Analog-to-Digital Converters (ADC).
> >
> > Signed-off-by: Marcus Folkesson <marcus.folkesson@gmail.com>  
> 
> > Signed-off-by: Kent Gustavsson <kent@minoris.se>  
> 
> What is this? Why it's here (presense and location in the message)?
To clarify... If Kent wrote the patch and you are simply acting
as gatekeeper / upstreamer you should set the mail to be from Kent
and put your own Signed-off after his to basically act as a submaintainer
certifying you believe his sign off and all that entails.

If it is a bit of a joint effort then that's fine but for copyright
purposes there should be some indication of the split.


> 
> > + * Copyright (C) 2018 Kent Gustavsson <kent@minoris.se>  
> 
> > + *  
> 
> Redundant.
> 
> > +static int mcp3911_read(struct mcp3911 *adc, u8 reg, u32 *val, u8 len)
> > +{
> > +       int ret;
> > +
> > +       reg = MCP3911_REG_READ(reg, adc->dev_addr);
> > +       ret = spi_write_then_read(adc->spi, &reg, 1, val, len);
> > +       if (ret < 0)
> > +               return ret;
> > +
> > +       be32_to_cpus(val);
> > +       *val >>= ((4 - len) * 8);
> > +       dev_dbg(&adc->spi->dev, "reading 0x%x from register 0x%x\n", *val,
> > +                       reg>>1);
> > +       return ret;
> > +}
> > +
> > +static int mcp3911_write(struct mcp3911 *adc, u8 reg, u32 val, u8 len)
> > +{
> > +       dev_dbg(&adc->spi->dev, "writing 0x%x to register 0x%x\n", val, reg);
> > +
> > +       val <<= (3 - len) * 8;
> > +       cpu_to_be32s(&val);
> > +       val |= MCP3911_REG_WRITE(reg, adc->dev_addr);
> > +
> > +       return spi_write(adc->spi, &val, len + 1);
> > +}  
> 
> Can't you use REGMAP_SPI ?
My instinct is not really. The magic device-addr doesn't help.
You could work around it but it's not that nice and you'd have
to create the regmap mapping on the fly or carry static ones
for each value of htat.



> 
> > +       of_property_read_u32(of_node, "device-addr", &adc->dev_addr);
> > +       if (adc->dev_addr > 3) {
> > +               dev_err(&adc->spi->dev,
> > +                               "invalid device address (%i). Must be in range 0-3.\n",
> > +                               adc->dev_addr);
> > +               return -EINVAL;
> > +       }
> > +       dev_dbg(&adc->spi->dev, "use device address %i\n", adc->dev_addr);  
> 
> Isn't what we called CS (chip select)?
Nope. We went around this in an earlier revision. It's an address transmitted
in the control byte to allow you to 'share' a chip select line between multiple
chips (crazy but true).

> 
> > +       if (IS_ERR(adc->vref)) {  
> 
> > +  
> 
> Redundant.
> 
> > +       if (adc->clki)  
> 
> Seems to be redundant if clock is optional (it should be NULL and API
> is NULL-aware).
> 
> > +               clk_disable_unprepare(adc->clki);  
> 
> > +       if (adc->clki)
> > +               clk_disable_unprepare(adc->clki);  
> 
> Ditto.
> 
> > +#if defined(CONFIG_OF)  
> 
> This prevents playing with the device on ACPI enabled platforms.
Yeah, that trickery is still little known (and I forget about it from
time to time).

The upshot for those who have missed this is you can use a magic
acpi device to instantiate based on device tree bindings :)

So we need to strip all this 'obvious' protection stuff out of drivers.

> 
> > +               .of_match_table = of_match_ptr(mcp3911_dt_ids),  
> 
> Ditto for macro.
>