Re: [PATCH 2/2] iio: proximity: Add driver support for vcnl3020 proximity sensor

[Andy Shevchenko <andy.shevchenko@gmail.com>]

On Mon, Mar 23, 2020 at 12:41 PM Ivan Mikhaylov <i.mikhaylov@yadro.com> wrote:
>
> Proximity sensor driver based on light/vcnl4000.c code.
> For now supports only the single on-demand measurement.
>
> The VCNL3020 is a fully integrated proximity sensor. Fully
> integrated means that the infrared emitter is included in the
> package. It has 16-bit resolution. It includes a signal
> processing IC and features standard I2C communication
> interface. It features an interrupt function.

Thank you for a patch, my comments below.

> Datasheet available at:
> http://www.vishay.com/docs/84150/vcnl3020.pdf

I'm thinking that we may simple introduce new tag, called Datesheet:
to put such links.

> Signed-off-by: Ivan Mikhaylov <i.mikhaylov@yadro.com>

...

>  obj-$(CONFIG_SRF08)            += srf08.o
>  obj-$(CONFIG_SX9500)           += sx9500.o
>  obj-$(CONFIG_VL53L0X_I2C)      += vl53l0x-i2c.o
> +obj-$(CONFIG_VCNL3020)         += vcnl3020.o

Perhaps keep ordered?

...

> +/*
> + * vcnl3020.c - Support for Vishay VCNL3020 proximity sensor

Using file names in themselves is a bad idea. Whenever you would
rename file (for instance, to support new sensors from the same family
in the future) you will forget (often, I see this in practice!) to
update this line.
Just drop it from here and try to avoid in the future.

> + *
> + * based on vcnl4000.c

This sounds like a continuation of previous sentence. Drop line in
between and use proper English grammar and punctuation.

> + */

...

> +struct vcnl3020_data {
> +       struct i2c_client *client;
> +       u32 rev;

> +       struct mutex vcnl3020_lock; /* for i2c operations */

Simple 'lock' is enough, the rest is dup noise.
Also, consider kernel doc format instead of odd comments.

> +};

...

> +static const struct i2c_device_id vcnl3020_id[] = {
> +       { "vcnl3020", 0 },
> +       {}
> +};
> +MODULE_DEVICE_TABLE(i2c, vcnl3020_id);

Can you group this with OF table below?

...

> +static int32_t vcnl3020_init(struct vcnl3020_data *data)

int32_t...

> +{

> +       s32 rc;

...s32?!

Applies to entire code.

> +       u32 led_current;
> +       struct device *dev = &data->client->dev;

Reversed xmas tree order looks better.

> +       rc = i2c_smbus_read_byte_data(data->client, VCNL_PROD_REV);

Can you use regmap I²C API?

> +       if (rc < 0) {
> +               dev_err(dev, "Error (%d) reading product revision", rc);
> +               goto exit;
> +       }
> +

> +       if (rc == VCNL3020_PROD_ID) {
> +               data->rev = rc & 0xff;

This conjunction looks strange. Also, why type of rev is u32 instead of u8?

> +               mutex_init(&data->vcnl3020_lock);
> +       } else {
> +               dev_err(dev, "Product id (%x) did not match vcnl3020 (%x)", rc,
> +                       VCNL3020_PROD_ID);
> +               rc = -ENODEV;
> +               goto exit;
> +       }
> +
> +       /* set led current */
> +       rc = i2c_smbus_write_byte_data(data->client, VCNL_LED_CURRENT,
> +                                      led_current);
> +       if (rc < 0) {
> +               dev_err(dev, "Error (%d) setting LED current", rc);
> +               goto exit;
> +       }
> +
> +       return 0;

> +exit:
> +       return rc;

Useless. Return directly.

> +};

...

> +       /* wait for data to become ready */
> +       while (tries--) {
> +               rc = i2c_smbus_read_byte_data(data->client, VCNL_COMMAND);
> +               if (rc < 0)
> +                       goto fail;
> +               if (rc & VCNL_PS_RDY)
> +                       break;
> +               msleep(20); /* measurement takes up to 100 ms */
> +       }

Timeout loops look more naturally in do {} while format.

  unsigned int tries = 5;
  ...

  do {
  ...
  } while (--tries);

...

> +       *val = (rc & 0xff) << 8;

> +       *val |= rc & 0xff;

All these conjunctions looks fishy. Why do you need them? Cant you
rely on the returned value by I²C API?

...

> +fail:

Better name is 'err_unlock' or 'out_unlock'. The rule of thumb to
describe in the label what you *about to do* there.

> +       mutex_unlock(&data->vcnl3020_lock);
> +
> +       return rc;
> +}

...

> +                       rc = vcnl3020_measure_proximity(data, val);
> +                       if (rc < 0)

Can rc be positive? Drop all these ' < 0' in cases where it is
guaranteed not to be the case.

> +                               return rc;

...

> +static int32_t vcnl3020_probe(struct i2c_client *client,
> +                             const struct i2c_device_id *id)

Can you switch to ->probe_new() ?

...

> +       dev_info(&client->dev, "Proximity sensor, Rev: %02x\n",
> +                data->rev);

Noise.

...

> +       rc = devm_iio_device_register(&client->dev, indio_dev);
> +       if (rc != 0)

Redundant ' != 0' part.

> +               goto out;
> +
> +       return rc;

> +out:
> +       devm_iio_device_free(&client->dev, indio_dev);
> +       return rc;

Managed resources are exactly for this not to be appeared in the code.

> +}

...

> +static const struct of_device_id vcnl3020_of_match[] = {
> +       {
> +               .compatible = "vishay,vcnl3020",
> +       },

Missed terminator. How did you test this?

> +};

--
With Best Regards,
Andy Shevchenko