Re: [PATCH v4] iio: chemical: Add support for Bosch BME680 sensor

[Himanshu Jha]

> Hi Himanshu,
> 
> This was close to the point where I'd take it and make the few remaining
> fixes myself.  I'm still bothered however by the fact the casts in the
> various calibration functions are still not all justified so please take
> another look at that.  Frankly it looks like the original author
> threw in casts because they didn't want to have to think about which ones
> actually do anything!

Ok. I will remove the ones mentioned below.

> Few other things to fix up for v5 as well.

I will send the fixes in an hour.

> Jonathan
> 
> 
> > +#define   BME680_NB_CONV_MASK			GENMASK(3, 0)
> > +#define	    BME680_RUN_GAS_EN_BIT		BIT(4)
> 
> odd looking spacing above.

I don't know why this is showing like that in the diff output, but I
have checked the code by apllying to my test tree(git am <patch>) and
there was no such spurious spacing!

> > +/*
> > + * Taken from Bosch BME680 API:
> > + * https://github.com/BoschSensortec/BME680_driver/blob/63bb5336/bme680.c#L937
> > + *
> > + * Returns humidity measurement in percent, resolution is 0.001 percent. Output
> > + * value of "43215" represents 43.215 %rH.
> > + */
> > +static u32 bme680_compensate_humid(struct bme680_data *data,
> > +				   u16 adc_humid)
> > +{
> > +	struct bme680_calib *calib = &data->bme680;
> > +	s32 var1, var2, var3, var4, var5, var6, temp_scaled, calc_hum;
> > +
> There is still substantial over the top casting in here.
> 
> data->t_fine is already a 32 bit integer for example.

Will do.

> > +	temp_scaled = (((s32) data->t_fine * 5) + 128) >> 8;
> > +	var1 = (adc_humid - ((s32) ((s32) calib->par_h1 * 16))) -
> 
> the outer s32 looks like it won't do anything to me as the inner bit
> will already be an s32.

OK.

> > +			(((temp_scaled * (s32) calib->par_h3) / 100) >> 1);
> > +	var2 = ((s32) calib->par_h2 * (((temp_scaled * (s32) calib->par_h4) /
> > +			((s32) 100)) + (((temp_scaled * ((temp_scaled *
> > +			(s32) calib->par_h5) / 100)) >> 6) / 100) +
> > +			(s32) (1 << 14))) >> 10;
> 
> Lots more casts of dubious merit in here..

Ok.

> > +	var3 = var1 * var2;
> > +	var4 = (s32) calib->par_h6 << 7;
> > +	var4 = (var4 + ((temp_scaled * (s32) calib->par_h7) / 100)) >> 4;
> > +	var5 = ((var3 >> 14) * (var3 >> 14)) >> 10;
> > +	var6 = (var4 * var5) >> 1;
> > +	calc_hum = (((var3 + var6) >> 10) * 1000) >> 12;
> > +
> > +	if (calc_hum > 100000) /* Cap at 100%rH */
> > +		calc_hum = 100000;
> > +	else if (calc_hum < 0)
> > +		calc_hum = 0;
> > +
> > +	return calc_hum;
> > +}
> > +static u32 bme680_compensate_gas(struct bme680_data *data, u16 gas_res_adc,
> > +				 u8 gas_range)
> > +{
> > +	struct bme680_calib *calib = &data->bme680;
> > +	s64 var1;
> > +	u64 var2;
> > +	s64 var3;
> > +	u32 calc_gas_res;
> > +
> > +	/* Look up table 1 for the possible gas range values */
> > +	u32 lookupTable1[16] = {2147483647u, 2147483647u, 2147483647u,
> > +				2147483647u, 2147483647u, 2126008810u,
> > +				2147483647u, 2130303777u, 2147483647u,
> > +				2147483647u, 2143188679u, 2136746228u,
> > +				2147483647u, 2126008810u, 2147483647u,
> > +				2147483647u};
> > +	/* Look up table 2 for the possible gas range values */
> > +	u32 lookupTable2[16] = {4096000000u, 2048000000u, 1024000000u,
> > +				512000000u, 255744255u, 127110228u, 64000000u,
> > +				32258064u, 16016016u, 8000000u, 4000000u,
> > +				2000000u, 1000000u, 500000u, 250000u, 125000u};
> 
> You can mark these two arrays as const I think.

Sure.

> > +
> > +	var1 = ((1340 + (5 * (s64) calib->range_sw_err)) *
> > +			((s64) lookupTable1[gas_range])) >> 16;
> > +	var2 = (((s64) ((s64) gas_res_adc << 15) - 16777216) + var1);
> 
> Unless something odd is going on the outer cast just casts the bit
> that has already been forced to s64 to s64 so is pointless.

Ok.

> > +	var3 = (((s64) lookupTable2[gas_range] * (s64) var1) >> 9);
> > +	calc_gas_res = (u32) ((var3 + ((s64) var2 >> 1)) / (s64) var2);
> 
> 64 bit division, does that need to be do_div so as to support 32 bit
> platforms? Also, why does var 2 want to be cast to a 64 bit?
> You'll need to go back to 32 bit anyway to use do_div.

I will change to do_div().

> > +
> > +	if ((check & BME680_GAS_STAB_BIT) == 0) {
> > +	/*
> > +	 * occurs if either the gas heating duration was insuffient
> > +	 * to reach the target heater temperature or the target
> > +	 * heater temperature was too high for the heater sink to
> > +	 * reach.
> > +	 */
> 
> Odd comment indenting. I would move it before the if to make it
> more 'natural'.  Still clearly applies to this block without looking 'odd'.

Will do.

> > +static int bme680_read_raw(struct iio_dev *indio_dev,
> > +			   struct iio_chan_spec const *chan,
> > +			   int *val, int *val2, long mask)
> > +{
> > +	struct bme680_data *data = iio_priv(indio_dev);
> > +
> > +	switch (mask) {
> > +	case IIO_CHAN_INFO_PROCESSED:
> > +		switch (chan->type) {
> > +		case IIO_TEMP:
> > +			return bme680_read_temp(data, val, val2);
> > +		case IIO_PRESSURE:
> > +			return bme680_read_press(data, val, val2);
> > +		case IIO_HUMIDITYRELATIVE:
> > +			return bme680_read_humid(data, val, val2);
> > +		case IIO_RESISTANCE:
> > +			return bme680_read_gas(data, val);
> > +		default:
> > +			return -EINVAL;
> > +		}
> > +	case IIO_CHAN_INFO_OVERSAMPLING_RATIO:
> > +		switch (chan->type) {
> > +		case IIO_TEMP:
> > +			*val = 1 << data->oversampling_temp;
> > +			return IIO_VAL_INT;
> > +		case IIO_PRESSURE:
> > +			*val = 1 << data->oversampling_press;
> > +			return IIO_VAL_INT;
> > +		case IIO_HUMIDITYRELATIVE:
> > +			*val = 1 << data->oversampling_humid;
> > +			return IIO_VAL_INT;
> > +		default:
> > +			return -EINVAL;
> > +		}
> > +	default:
> > +		return -EINVAL;
> > +	}
> > +
> 
> Don't think you can get here so this code should not be here.

Ok.

> > +		default:
> > +			return -EINVAL;
> > +		}
> > +	default:
> > +		return -EINVAL;
> > +	}
> > +
> 
> You can't get here so no point in having this return. I'll delete it.

OK.

> > +	ret = devm_add_action(dev, bme680_core_remove, indio_dev);
> > +	if (ret < 0) {
> > +		dev_err(dev, "failed to register remove action\n");
> > +		return ret;
> > +	}
> 
> I think this is logically in the wrong place.  The things it's actually
> doing is unwinding the register below, but at this stage you haven't
> performed that registration.
> 
> If this is all I fine, I 'might' move it and apply anyway.
> 
> I'm assuming that there will shortly be more in there than a simple
> unregister (otherwise you could just have used devm_iio_device_register...
> 
> I'd prefer that for now you just use devm_iio_device_register
> and drop this until you need it.

OK.

> > +
> > +	ret = regmap_write(regmap, BME680_REG_SOFT_RESET_I2C,
> > +			   BME680_CMD_SOFTRESET);
> > +	if (ret < 0)
> 
> It's not something I'm that bothered by, but you are a little random
> in whether a given error outputs a debug message or not...

I missed this.
Will fix.

-- 
Himanshu Jha
Undergraduate Student
Department of Electronics & Communication
Guru Tegh Bahadur Institute of Technology