Re: [PATCH v6 2/2] iio: (bma400) add driver for the BMA400

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wed, Dec 11, 2019 at 03:21:56PM +0200, Andy Shevchenko wrote:
> On Wed, Dec 11, 2019 at 3:20 AM Dan Robertson <dan@xxxxxxxxxxxxxxx> wrote:
> >
> > Add a IIO driver for the Bosch BMA400 3-axes ultra-low power accelerometer.
> > The driver supports reading from the acceleration and temperature
> > registers. The driver also supports reading and configuring the output data
> > rate, oversampling ratio, and scale.
> 
> > +#define BMA400_LP_OSR_SHIFT         0x05
> > +#define BMA400_NP_OSR_SHIFT         0x04
> > +#define BMA400_SCALE_SHIFT          0x06
> 
> I'm not sure why this is being defined as hex number instead of plain decimal...

Sounds good.

> > +#define BMA400_TWO_BITS_MASK        GENMASK(1, 0)
> > +#define BMA400_LP_OSR_MASK          GENMASK(6, BMA400_LP_OSR_SHIFT)
> > +#define BMA400_NP_OSR_MASK          GENMASK(5, BMA400_NP_OSR_SHIFT)
> > +#define BMA400_ACC_ODR_MASK         GENMASK(3, 0)
> > +#define BMA400_ACC_SCALE_MASK       GENMASK(7, BMA400_SCALE_SHIFT)
> 
> And here simple better to put same numbers. It will help to read.

Do you mean for the shift or for the mask?

> > +const struct regmap_config bma400_regmap_config = {
> > +       .reg_bits = 8,
> > +       .val_bits = 8,
> > +       .max_register = BMA400_CMD_REG,
> > +       .cache_type = REGCACHE_RBTREE,
> > +       .writeable_reg = bma400_is_writable_reg,
> > +       .volatile_reg = bma400_is_volatile_reg,
> > +};
> > +EXPORT_SYMBOL(bma400_regmap_config);
> 
> I'm not sure I got the idea why this one is being exported.

It needs to be exported so that it can be used in the bma400_i2c module and the
future bma400_spi module. In theory, if we _really_ do not want to export this,
then we can define separate regmap configs in each of the bma400_i2c and
(future) bma400_spi modules, but then we would have to export the is_volitile_reg
and is_writable_reg functions. As a result, I do not see any benefits to that
method over exporting the config, but I could be convinced otherwise.

> > +               if (odr < BMA400_ACC_ODR_MIN_RAW ||
> > +                   odr > BMA400_ACC_ODR_MAX_RAW) {
> 
> One line?

It is too long if I simplify to one line.

> > +               if (uhz || hz % BMA400_ACC_ODR_MIN_WHOLE_HZ)
> > +                       return -EINVAL;
> > +
> > +               val = hz / BMA400_ACC_ODR_MIN_WHOLE_HZ;
> > +               idx = __ffs(val);
> > +
> 
> > +               if (val ^ BIT(idx))
> 
> Seems like funny way of checking is_power_of_2(). But it's up to maintainers.
> And your variant may even be better here (in code generation perspective)...
> 
> However, the whole idea here is, IIUC, to have something like
> 
>   hz = 2^idx * BMA400_ACC_ODR_MIN_WHOLE_HZ
> 
> I think you may do it without divisions, i.e. call __ffs() first and then do
>    idx = __ffs(...);
>    val = hz >> idx;
>    if (val != BMA400_ACC_ODR_MIN_WHOLE_HZ)
>     return -EINVAL;
> 
> or something like above.

It would be more obvious what is being done here with is_power_of_two. I'll
revisit this function with your suggestions. If I can make it simpler, I'll
go this route.

> 
> > +                       return -EINVAL;
> 
> ...
> 
> > +       odr = (~BMA400_ACC_ODR_MASK & val) | idx;
> 
> I'm wondering why Yoda style is being used here.

I guess I think like Yoda :) I can update this. I typically do prefer
new_mask | old_mask, but I do not feel too strongly about it.

> > +static void bma400_accel_scale_from_raw(int raw, unsigned int *val)
> > +{
> > +       *val = BMA400_SCALE_MIN * (1 << raw);
> 
> Isn't it the same as
>     *val = BMA400_SCALE_MIN << raw;
> ?

Yes. Good catch. Not sure what I was thinking :)

> 
> > +               return -EINVAL;
> 
> ...
> 
> > +       ret = regmap_read(data->regmap, BMA400_ACC_CONFIG0_REG, &val);
> > +       if (ret < 0)
> 
> I'm wondering if in all of these regmap_read()...
> 
> > +               return ret;
> 
> > +       ret = regmap_write(data->regmap, BMA400_ACC_CONFIG0_REG,
> > +                          mode | (val & ~BMA400_TWO_BITS_MASK));
> > +       if (ret < 0) {
> 
> ...and regmap_write() calls you ever can get a positive returned code.

>From the regmap_read/regmap_write docs:

> * A value of zero will be returned on success, a negative errno will
> * be returned in error cases.

So I assume ret <= 0

Cheers,

 - Dan




[Index of Archives]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Input]     [Linux Kernel]     [Linux SCSI]     [X.org]

  Powered by Linux