Re: [PATCH 2/2 v5] iio: Add SEMTECH SX9310/9311 sensor driver

[Daniel Campello]

On Tue, Mar 24, 2020 at 10:18 AM Andy Shevchenko
<andy.shevchenko@xxxxxxxxx> wrote:
>
> On Mon, Mar 23, 2020 at 8:46 PM Daniel Campello <campello@xxxxxxxxxxxx> wrote:
> >
> > Add SEMTECH SX9310/9311 driver.
> >
> > The device has the following entry points:
> >
> > Usual frequency:
> > - sampling_frequency
> > - sampling_frequency_available
> >
> > Instant reading of current values for different sensors:
> > - in_proximity0_raw
> > - in_proximity1_raw
> > - in_proximity2_raw
> > - in_proximity3_comb_raw
> > and associated events in events/
>
> ...
>
> > +#include <linux/acpi.h>
>
> > +#include <linux/of.h>
>
> No users for these (see ID table handling below), but property.h would
> be needed.
>
>

I removed of.h but kept acpi.h in v6 since it was needed for other
ACPI macros on the probe function
>
> ...
>
> > +#define   SX9310_EVENT_IRQ                             (SX9310_FAR_IRQ | \
> > +                                                        SX9310_CLOSE_IRQ)
>
> Better formatting is
> #define FOO \
>   (BAR | ZOO)
>
> ...
>
> > +struct sx9310_data {
>
> > +       struct i2c_client *client;
>
> Do you really need client? Perhaps struct device will be enough?


There are references to client->irq in some places

>
>
> > +       struct iio_trigger *trig;
> > +       struct regmap *regmap;
> > +       /*
> > +        * Last reading of the proximity status for each channel.
> > +        * We only send an event to user space when this changes.
> > +        */
> > +       bool prox_stat[SX9310_NUM_CHANNELS];
> > +       bool trigger_enabled;
>
> > +       __be16 buffer[SX9310_NUM_CHANNELS +
> > +                     4]; /* 64-bit data + 64-bit timestamp */
>
> Please, fix formatting.
>
> > +};
>
> ...
>
> > +static int sx9310_update_chan_en(struct sx9310_data *data,
> > +                                unsigned int chan_read,
> > +                                unsigned int chan_event)
> > +{
> > +       int ret;
> > +
>
> > +       if ((data->chan_read | data->chan_event) != (chan_read | chan_event)) {
> > +               ret = regmap_update_bits(data->regmap, SX9310_REG_PROX_CTRL0,
> > +                                        SX9310_CHAN_ENABLED_MASK,
> > +                                        chan_read | chan_event);
>
> unsigned int readevent = chan_read | chan_event;
> ...
> if (... != readevent) {
>   ..., readevent);
> }
>
> > +               if (ret)
> > +                       return ret;
> > +       }
> > +       data->chan_read = chan_read;
> > +       data->chan_event = chan_event;
> > +       return 0;
> > +}
>
> ...
>
> > +static int sx9310_read_prox_data(struct sx9310_data *data,
> > +                                const struct iio_chan_spec *chan, __be16 *val)
> > +{
> > +       int ret;
> > +
> > +       ret = regmap_write(data->regmap, SX9310_REG_SENSOR_SEL, chan->channel);
> > +       if (ret < 0)
> > +               return ret;
> > +
>
> > +       return regmap_bulk_read(data->regmap, chan->address, val, 2);
>
> sizeof()?
>
> > +}
>
> ...
>
> > +       if (data->client->irq > 0) {
> > +               ret = wait_for_completion_interruptible(&data->completion);
> > +               reinit_completion(&data->completion);
>
> Logically reinit better to be called before you start measurement.
>
I think this is effectively before measurement, minus error/locking handling.

> > +       } else {
> > +               ret = sx9310_wait_for_sample(data);
> > +       }
>
> ...
>
> > +       *val = sign_extend32(be16_to_cpu(rawval),
> > +                            (chan->address == SX9310_REG_DIFF_MSB ? 11 : 15));
>
> Too many parentheses.
>
> ...
>
> > +       mutex_lock(&data->mutex);
> > +
> > +       ret = regmap_update_bits(data->regmap, SX9310_REG_PROX_CTRL0,
> > +                                SX9310_SCAN_PERIOD_MASK,
> > +                                i << SX9310_SCAN_PERIOD_SHIFT);
> > +
> > +       mutex_unlock(&data->mutex);
>
> Btw, can you use locking provided by regmap?


The lock is protecting critical sections that include multiple regmap
and other operations together in other parts of the code.
>
>
> ...
>
> > +       /*
> > +        * Even if no event is enabled, we need to wake the thread to
>
> > +        * clear the interrupt state by reading SX9310_REG_IRQ_SRC.  It
>
> Move it to the next line.
>
> > +        * is not possible to do that here because regmap_read takes a
> > +        * mutex.
> > +        */
>
> ...
>
> > +       for (chan = 0; chan < SX9310_NUM_CHANNELS; chan++) {
> > +               int dir;
> > +               u64 ev;
> > +               bool new_prox = val & BIT(chan);
> > +
>
> > +               if (!(data->chan_event & BIT(chan)))
> > +                       continue;
>
> for_each_set_bit()


Thanks, more clear now!
>
>
> > +               if (new_prox == data->prox_stat[chan])
> > +                       /* No change on this channel. */
> > +                       continue;
>
> > +       }
>
> ...
>
> > +static struct attribute *sx9310_attributes[] = {
> > +       &iio_dev_attr_sampling_frequency_available.dev_attr.attr,
>
> > +       NULL,
>
> Comma is not needed for terminator.
>
> > +};
>
> ...
>
> > +static int sx9310_buffer_preenable(struct iio_dev *indio_dev)
> > +{
> > +       struct sx9310_data *data = iio_priv(indio_dev);
> > +       unsigned int channels = 0;
> > +       int bit, ret;
> > +
> > +       mutex_lock(&data->mutex);
> > +       for_each_set_bit(bit, indio_dev->active_scan_mask,
> > +                        indio_dev->masklength)
>
> > +               channels |= BIT(indio_dev->channels[bit].channel);
>
> unsigned long channels;
> ...
> __set_bit(...);
>
> > +       ret = sx9310_update_chan_en(data, channels, data->chan_event);
> > +       mutex_unlock(&data->mutex);
> > +       return ret;
> > +}
>
> ...
>
> > +#define SX_INIT(_reg, _def)                    \
> > +       {                                       \
> > +               .reg = SX9310_REG_##_reg,       \
> > +               .def = _def,                    \
> > +       }
>
> Usually it's a good tone to #undef custom macro when they are not
> needed anymore.
>
> ...
>
> > +       for (i = 100; i >= 0; i--) {
> > +               msleep(20);
> > +               ret = regmap_read(data->regmap, SX9310_REG_STAT1, &val);
> > +               if (ret < 0)
> > +                       goto out;
> > +               if (!(val & SX9310_COMPSTAT_MASK))
> > +                       break;
> > +       }
>
> NIH of regmap_real_poll_timeout();


Sorry, somehow I overlooked this on v6, I'll send v7 with it

>
>
> > +
> > +       if (i < 0) {
> > +               dev_err(&data->client->dev,
> > +                       "initial compensation timed out: 0x%02x", val);
> > +               ret = -ETIMEDOUT;
> > +       }
>
> ...
>
> > +               ret = regmap_write(data->regmap, initval->reg, initval->def);
> > +               if (ret < 0)
>
> Do you need all these ' < 0'?


Not sure what do you mean? This one in particular is trying to fail
fast the probe
>
>
> > +                       return ret;
>
> ...
>
> > +       const struct acpi_device_id *acpi_id;
>
> > +       /* id will be NULL when enumerated via ACPI */
> > +       if (id) {
> > +               if (id->driver_data != whoami)
> > +                       dev_err(dev, "WHOAMI does not match i2c_device_id: %s",
> > +                               id->name);
> > +       } else if (ACPI_HANDLE(dev)) {
> > +               acpi_id = acpi_match_device(dev->driver->acpi_match_table, dev);
> > +               if (!acpi_id)
> > +                       return -ENODEV;
> > +               if (acpi_id->driver_data != whoami)
> > +                       dev_err(dev, "WHOAMI does not match acpi_device_id: %s",
> > +                               acpi_id->id);
> > +       } else
> > +               return -ENODEV;
>
> device_get_match_data().
>
THANKS! I was happy to learn about this after you pointed it out!
>
> ...
>
> > +static int sx9310_probe(struct i2c_client *client,
> > +                       const struct i2c_device_id *id)
>
> Can you switch to ->probe_new()?
>
> ...
>
> > +       indio_dev = devm_iio_device_alloc(&client->dev, sizeof(*data));
>
> > +       if (indio_dev == NULL)
>
> if (!indio_dev)
>
> > +               return -ENOMEM;
>
> ...
>
> > +               dev_err(&client->dev, "error in reading WHOAMI register: %d",
> > +                       ret);
>
> If you introduce temporary variable the code will be better to read
>
>   struct device *dev = &client->dev;
>
> > +       ret = sx9310_set_indio_dev_name(&client->dev, indio_dev, id,
> > +                                       data->whoami);
> > +       if (ret < 0)
> > +               return ret;
>
> ...
>
> > +static const struct acpi_device_id sx9310_acpi_match[] = {
> > +       { "STH9310", SX9310_WHOAMI_VALUE },
> > +       { "STH9311", SX9311_WHOAMI_VALUE },
>
> Hmm... May I ask some official proof that these IDs are real and
> issued by vendor?


Not sure how to prove this but they are live in device firmware right
now. One example:
https://chromium.googlesource.com/chromiumos/third_party/coreboot/+/b905beb46935c114ebc416583bb2e5407183af35/src/mainboard/google/zoombini/variants/meowth/devicetree.cb
>
>
> > +       {},
>
> No comma.
>
> > +};
> > +MODULE_DEVICE_TABLE(acpi, sx9310_acpi_match);
>
> > +static const struct of_device_id sx9310_of_match[] = {
> > +       { .compatible = "semtech,sx9310" },
> > +       { .compatible = "semtech,sx9311" },
>
> > +       {},
>
> No comma.
>
> > +};
> > +MODULE_DEVICE_TABLE(of, sx9310_of_match);
> > +
> > +static const struct i2c_device_id sx9310_id[] = {
> > +       { "sx9310", SX9310_WHOAMI_VALUE },
> > +       { "sx9311", SX9311_WHOAMI_VALUE },
>
> > +       {},
>
> No comma.
>
> > +};
> > +MODULE_DEVICE_TABLE(i2c, sx9310_id);
>
> ...
>
> > +               .acpi_match_table = ACPI_PTR(sx9310_acpi_match),
> > +               .of_match_table = of_match_ptr(sx9310_of_match),
>
> Drop these macros. You probably didn't test with !ACPI and/or !OF --
> should be compiler warning.


Why do we want to drop these? I tried compiling with either and both
disabled and did not see a warning on my system.
>
>
> > +               .pm = &sx9310_pm_ops,
> > +       },
>
> --
> With Best Regards,
> Andy Shevchenko

Thanks for the review. Sorry for the delay on this email after sending v6.

Regards,
Daniel