Re: [PATCH v1] iio: adc: aspeed: Support deglitch feature.

From: Billy Tsai
Date: Mon Oct 02 2023 - 04:50:32 EST


On Mon, 25 Sep 2023, at 17:48, Billy Tsai wrote:
> > Create event sysfs for applying the deglitch condition. When
> > in_voltageY_thresh_rising_en/in_voltageY_thresh_falling_en is set to true,
> > the driver will use the in_voltageY_thresh_rising_value and
> > in_voltageY_thresh_falling_value as threshold values. If the ADC value
> > falls outside this threshold, the driver will wait for the ADC sampling
> > period and perform an additional read once to achieve the deglitching
> > purpose.
> >
> > Signed-off-by: Billy Tsai <billy_tsai@xxxxxxxxxxxxxx>
> > ---
> > drivers/iio/adc/aspeed_adc.c | 193 ++++++++++++++++++++++++++++++++++-
> > 1 file changed, 189 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/iio/adc/aspeed_adc.c b/drivers/iio/adc/aspeed_adc.c
> > index 998e8bcc06e1..9e746c81d916 100644
> > --- a/drivers/iio/adc/aspeed_adc.c
> > +++ b/drivers/iio/adc/aspeed_adc.c
> > @@ -95,6 +95,7 @@ struct aspeed_adc_model_data {
> > bool wait_init_sequence;
> > bool need_prescaler;
> > bool bat_sense_sup;
> > + bool require_extra_eoc;

> What is "eoc"? Can we use a better name or add an explanatory comment?

Hi Andrew,
This is the signal name for our ADC controller, it means "End Of Conversion".
The appearance of this signal period indicates that the ADC value is valid and being updated to the register.

> > u8 scaler_bit_width;
> > unsigned int num_channels;
> > const struct aspeed_adc_trim_locate *trim_locate;
> > @@ -120,6 +121,26 @@ struct aspeed_adc_data {
> > int cv;
> > bool battery_sensing;
> > struct adc_gain battery_mode_gain;
> > + unsigned int required_eoc_num;
> > + u16 *upper_bound;
> > + u16 *lower_bound;
> > + bool *upper_en;
> > + bool *lower_en;

> I wonder whether we should instead embed enough memory for these new properties directly into the struct. Take the upper bound on the number of channels across the supported SoCs (`#define ASPEED_ADC_MAX_CHANNELS 16`, from the values defined across the `struct aspeed_adc_model_data` instances down below). From there we could have `u16 upper_bound[ASPEED_ADC_MAX_CHANNELS]` etc instead of the extra allocations in probe(), which get a bit tedious. Also the channel `{upper,lower}_en` values can be bit-masked out of a u16, avoiding the dynamic allocations for those as well.

OK, I will modify it to an array with a size of ASPEED_ADC_MAX_CHANNELS (16) and use bit-mask for *_en.

> > +};
> > +
> > +static const struct iio_event_spec aspeed_adc_events[] = {
> > + {
> > + .type = IIO_EV_TYPE_THRESH,
> > + .dir = IIO_EV_DIR_RISING,
> > + .mask_separate =
> > + BIT(IIO_EV_INFO_VALUE) | BIT(IIO_EV_INFO_ENABLE),
> > + },
> > + {
> > + .type = IIO_EV_TYPE_THRESH,
> > + .dir = IIO_EV_DIR_FALLING,
> > + .mask_separate =
> > + BIT(IIO_EV_INFO_VALUE) | BIT(IIO_EV_INFO_ENABLE),
> > + },
> > };
> >
> > #define ASPEED_CHAN(_idx, _data_reg_addr) { \
> > @@ -131,6 +152,8 @@ struct aspeed_adc_data {
> > .info_mask_shared_by_type = BIT(IIO_CHAN_INFO_SCALE) | \
> > BIT(IIO_CHAN_INFO_SAMP_FREQ) | \
> > BIT(IIO_CHAN_INFO_OFFSET), \
> > + .event_spec = aspeed_adc_events, \
> > + .num_event_specs = ARRAY_SIZE(aspeed_adc_events), \
> > }
> >
> > static const struct iio_chan_spec aspeed_adc_iio_channels[] = {
> > @@ -277,6 +300,35 @@ static int aspeed_adc_set_sampling_rate(struct
> > iio_dev *indio_dev, u32 rate)
> > return 0;
> > }
> >
> > +static int aspeed_adc_get_voltage_raw(struct aspeed_adc_data *data,
> > + struct iio_chan_spec const *chan)
> > +{
> > + int val;
> > +
> > + val = readw(data->base + chan->address);
> > + dev_dbg(data->dev,
> > + "%d upper_bound: %d %x, lower_bound: %d %x, delay: %d * %d ns",
> > + chan->channel, data->upper_en[chan->channel],
> > + data->upper_bound[chan->channel], data->lower_en[chan->channel],
> > + data->lower_bound[chan->channel], data->sample_period_ns,
> > + data->required_eoc_num);
> > + if (data->upper_en[chan->channel]) {
> > + if (val >= data->upper_bound[chan->channel]) {
> > + ndelay(data->sample_period_ns *
> > + data->required_eoc_num);
> > + val = readw(data->base + chan->address);
> > + }
> > + }
> > + if (data->lower_en[chan->channel]) {
> > + if (val <= data->lower_bound[chan->channel]) {
> > + ndelay(data->sample_period_ns *
> > + data->required_eoc_num);
> > + val = readw(data->base + chan->address);
> > + }
> > + }

> Is the potential for a double delay if `data->lower_bound[chan->channel] >= data->upper_bound[chan->channel]` desirable?

I will utilize the "else" condition to prevent the double delay.

> > + return val;
> > +}
> > +
> > static int aspeed_adc_read_raw(struct iio_dev *indio_dev,
> > struct iio_chan_spec const *chan,
> > int *val, int *val2, long mask)
> > @@ -299,14 +351,15 @@ static int aspeed_adc_read_raw(struct iio_dev *indio_dev,
> > * Experiment result is 1ms.
> > */
> > mdelay(1);
> > - *val = readw(data->base + chan->address);
> > + *val = aspeed_adc_get_voltage_raw(data, chan);
> > *val = (*val * data->battery_mode_gain.mult) /
> > data->battery_mode_gain.div;
> > /* Restore control register value */
> > writel(adc_engine_control_reg_val,
> > data->base + ASPEED_REG_ENGINE_CONTROL);
> > - } else
> > - *val = readw(data->base + chan->address);
> > + } else {
> > + *val = aspeed_adc_get_voltage_raw(data, chan);
> > + }
> > return IIO_VAL_INT;
> >
> > case IIO_CHAN_INFO_OFFSET:
> > @@ -369,9 +422,106 @@ static int aspeed_adc_reg_access(struct iio_dev
> > *indio_dev,
> > return 0;
> > }
> >
> > +static int aspeed_adc_read_event_config(struct iio_dev *indio_dev,
> > + const struct iio_chan_spec *chan,
> > + enum iio_event_type type,
> > + enum iio_event_direction dir)
> > +{
> > + struct aspeed_adc_data *data = iio_priv(indio_dev);
> > +
> > + switch (dir) {
> > + case IIO_EV_DIR_RISING:
> > + return data->upper_en[chan->channel];
> > + case IIO_EV_DIR_FALLING:
> > + return data->lower_en[chan->channel];
> > + default:
> > + return -EINVAL;
> > + }
> > +}
> > +
> > +static int aspeed_adc_write_event_config(struct iio_dev *indio_dev,
> > + const struct iio_chan_spec *chan,
> > + enum iio_event_type type,
> > + enum iio_event_direction dir,
> > + int state)
> > +{
> > + struct aspeed_adc_data *data = iio_priv(indio_dev);
> > +
> > + switch (dir) {
> > + case IIO_EV_DIR_RISING:
> > + data->upper_en[chan->channel] = state ? 1 : 0;
> > + break;
> > + case IIO_EV_DIR_FALLING:
> > + data->lower_en[chan->channel] = state ? 1 : 0;
> > + break;
> > + default:
> > + return -EINVAL;
> > + }
> > +
> > + return 0;
> > +}
> > +
> > +static int aspeed_adc_write_event_value(struct iio_dev *indio_dev,
> > + const struct iio_chan_spec *chan,
> > + enum iio_event_type type,
> > + enum iio_event_direction dir,
> > + enum iio_event_info info, int val,
> > + int val2)
> > +{
> > + struct aspeed_adc_data *data = iio_priv(indio_dev);
> > +
> > + if (info != IIO_EV_INFO_VALUE)
> > + return -EINVAL;
> > +
> > + switch (dir) {
> > + case IIO_EV_DIR_RISING:
> > + if (val >= BIT(ASPEED_RESOLUTION_BITS))
> > + return -EINVAL;
> > + data->upper_bound[chan->channel] = val;
> > + break;
> > + case IIO_EV_DIR_FALLING:
> > + data->lower_bound[chan->channel] = val;

> Shouldn't we require the same test against BIT(ASPEED_RESOLUTION_BITS) here? Just because it should be low it doesn't mean that someone won't write a high value. If it is required then you could hoist the test in the IIO_EV_DIR_RISING case above the switch statement to cover both cases.

Thanks, I will move the test above the switch statement.

> > + break;
> > + default:
> > + return -EINVAL;
> > + }
> > +
> > + return 0;
> > +}
> > +
> > +static int aspeed_adc_read_event_value(struct iio_dev *indio_dev,
> > + const struct iio_chan_spec *chan,
> > + enum iio_event_type type,
> > + enum iio_event_direction dir,
> > + enum iio_event_info info, int *val,
> > + int *val2)
> > +{
> > + struct aspeed_adc_data *data = iio_priv(indio_dev);
> > +
> > + if (info != IIO_EV_INFO_VALUE)
> > + return -EINVAL;
> > +
> > + switch (dir) {
> > + case IIO_EV_DIR_RISING:
> > + *val = data->upper_bound[chan->channel];
> > + break;
> > + case IIO_EV_DIR_FALLING:
> > + *val = data->lower_bound[chan->channel];
> > + break;
> > + default:
> > + return -EINVAL;
> > + }
> > +
> > + return IIO_VAL_INT;
> > +}
> > +
> > static const struct iio_info aspeed_adc_iio_info = {
> > .read_raw = aspeed_adc_read_raw,
> > .write_raw = aspeed_adc_write_raw,
> > + .read_event_config = &aspeed_adc_read_event_config,
> > + .write_event_config = &aspeed_adc_write_event_config,
> > + .read_event_value = &aspeed_adc_read_event_value,
> > + .write_event_value = &aspeed_adc_write_event_value,
> > .debugfs_reg_access = aspeed_adc_reg_access,
> > };
> >
> > @@ -502,6 +652,30 @@ static int aspeed_adc_probe(struct platform_device *pdev)
> > if (IS_ERR(data->base))
> > return PTR_ERR(data->base);
> >
> > + data->upper_bound = devm_kzalloc(&pdev->dev,
> > + sizeof(data->upper_bound) *
> > + data->model_data->num_channels,
> > + GFP_KERNEL);
> > + if (!data->upper_bound)
> > + return -ENOMEM;
> > + data->upper_en = devm_kzalloc(&pdev->dev,
> > + sizeof(data->upper_en) *
> > + data->model_data->num_channels,
> > + GFP_KERNEL);
> > + if (!data->upper_en)
> > + return -ENOMEM;
> > + data->lower_bound = devm_kzalloc(&pdev->dev,
> > + sizeof(data->lower_bound) *
> > + data->model_data->num_channels,
> > + GFP_KERNEL);
> > + if (!data->lower_bound)
> > + return -ENOMEM;
> > + data->lower_en = devm_kzalloc(&pdev->dev,
> > + sizeof(data->lower_en) *
> > + data->model_data->num_channels,
> > + GFP_KERNEL);
> > + if (!data->lower_en)
> > + return -ENOMEM;

> See the commentary on the struct definition about potentially avoiding these extra dynamic allocations.

> > /* Register ADC clock prescaler with source specified by device tree.
> > */
> > spin_lock_init(&data->clk_lock);
> > snprintf(clk_parent_name, ARRAY_SIZE(clk_parent_name), "%s",
> > @@ -632,7 +806,14 @@ static int aspeed_adc_probe(struct platform_device
> > *pdev)
> > adc_engine_control_reg_val |= ASPEED_ADC_CTRL_CHANNEL;
> > writel(adc_engine_control_reg_val,
> > data->base + ASPEED_REG_ENGINE_CONTROL);
> > -
> > + adc_engine_control_reg_val =
> > + FIELD_GET(ASPEED_ADC_CTRL_CHANNEL,
> > + readl(data->base + ASPEED_REG_ENGINE_CONTROL));
> > + data->required_eoc_num = hweight_long(adc_engine_control_reg_val);
> > + if (data->model_data->require_extra_eoc &&
> > + (adc_engine_control_reg_val &
> > + BIT(data->model_data->num_channels - 1)))
> > + data->required_eoc_num += 12;

> Why 12? Why add a value to the number of engines enabled? Have I misunderstood?

This behavior is specified by the hardware. In our ADC, it requires 12 dummy sampling
periods to switch the sampling channel from CH7 to CH0. Hence, this condition checks
the enable status of channel 7 to determine the necessary delay period for obtaining the
updated ADC values for each channel.

Best Regards,
Billy Tsai