Re: [PATCH v5 1/4] iio: chemical: scd30: add core driver

[Tomasz Duszynski <tomasz.duszynski@octakon.com>]

On Sat, Jun 20, 2020 at 05:25:02PM +0100, Jonathan Cameron wrote:
> On Sun, 7 Jun 2020 19:58:09 +0200
> Tomasz Duszynski <tomasz.duszynski@octakon.com> wrote:
>
> > Add Sensirion SCD30 carbon dioxide core driver.
> >
> > Signed-off-by: Tomasz Duszynski <tomasz.duszynski@octakon.com>
>
> A few things I'd missed showed up in warnings when I applied this and
> ran a sparse check.  Please fix up and send a v6.
> Also sanity check the rest with sparse. Note if I'd missed this 0-day
> would have sent use these warnings.
>

Ah, forgot to add C=2 to prior building. Thanks for catching this.

> Thanks,
>
> Jonathan
>
> > +
> > +static int scd30_read_meas(struct scd30_state *state)
> > +{
> > +	int i, ret;
> > +
> > +	ret = state->command(state, CMD_READ_MEAS, 0, state->meas, sizeof(state->meas));
> > +	if (ret)
> > +		return ret;
> > +
> > +	be32_to_cpu_array(state->meas, state->meas, ARRAY_SIZE(state->meas));
>
> The type of the input to the above has the wrong endian markings.
>
> CHECK   drivers/iio/chemical/scd30_core.c
> drivers/iio/chemical/scd30_core.c:123:40: warning: incorrect type in argument 2 (different base types)
> drivers/iio/chemical/scd30_core.c:123:40:    expected restricted __be32 const [usertype] *src
> drivers/iio/chemical/scd30_core.c:123:40:    got int *
>
> Whilst you could use a cast, it would be tidier to use an array of __be32.
>

Here's the only place where it's a __be32. All other places assume
cpu endianess which means changing array type generates a few other warnings
here and there. So I'd prefer to be lazy here and use a cast :).

> > +
> > +	for (i = 0; i < ARRAY_SIZE(state->meas); i++)
> > +		state->meas[i] = scd30_float_to_fp(state->meas[i]);
> > +
> > +	/*
> > +	 * co2 is left unprocessed while temperature and humidity are scaled
> > +	 * to milli deg C and milli percent respectively.
> > +	 */
> > +	state->meas[SCD30_TEMP] *= 10;
> > +	state->meas[SCD30_HR] *= 10;
> > +
> > +	return 0;
> > +}
> > +
>
> ...
>
> > +
> > +static irqreturn_t scd30_trigger_handler(int irq, void *p)
> > +{
> > +	struct iio_poll_func *pf = p;
> > +	struct iio_dev *indio_dev = pf->indio_dev;
> > +	struct scd30_state *state = iio_priv(indio_dev);
> > +	struct {
> > +		int data[SCD30_MEAS_COUNT];
> > +		s64 ts __aligned(8);
> > +	} scan = { 0, };
> should be scan = { {0, }, }; or something like that
> as first element happens to be an array.
>
> Actually there is padding in here you need to zero I think.
> So memset is a better bet.
>

Sure.

> > +	int ret;
> > +
> > +	mutex_lock(&state->lock);
> > +	if (!iio_trigger_using_own(indio_dev))
> > +		ret = scd30_read_poll(state);
> > +	else
> > +		ret = scd30_read_meas(state);
> > +	memcpy(scan.data, state->meas, sizeof(state->meas));
> > +	mutex_unlock(&state->lock);
> > +	if (ret)
> > +		goto out;
> > +
> > +	iio_push_to_buffers_with_timestamp(indio_dev, &scan, iio_get_time_ns(indio_dev));
> > +out:
> > +	iio_trigger_notify_done(indio_dev->trig);
> > +	return IRQ_HANDLED;
> > +}
> > +
> ...