Re: [PATCH 2/2] iio: adc: Add the TI ads124s08 ADC code

From: Jonathan Cameron
Date: Sat Dec 08 2018 - 06:56:18 EST


On Tue, 4 Dec 2018 11:59:55 -0600
Dan Murphy <dmurphy@xxxxxx> wrote:

> Introduce the TI ADS124S08 and the ADS124S06 ADC
> devices from TI. The ADS124S08 is the 12 channel ADC
> and the ADS124S06 is the 6 channel ADC device
>
> These devices share a common datasheet:
> http://www.ti.com/lit/gpn/ads124s08
>
> Signed-off-by: Dan Murphy <dmurphy@xxxxxx>
Various minor things inline.

Thanks,

Jonathan

> ---
> drivers/iio/adc/Kconfig | 10 +
> drivers/iio/adc/Makefile | 1 +
> drivers/iio/adc/ti-ads124s08.c | 437 +++++++++++++++++++++++++++++++++
> 3 files changed, 448 insertions(+)
> create mode 100644 drivers/iio/adc/ti-ads124s08.c
>
> diff --git a/drivers/iio/adc/Kconfig b/drivers/iio/adc/Kconfig
> index a52fea8749a9..e8c5686e6716 100644
> --- a/drivers/iio/adc/Kconfig
> +++ b/drivers/iio/adc/Kconfig
> @@ -887,6 +887,16 @@ config TI_ADS8688
> This driver can also be built as a module. If so, the module will be
> called ti-ads8688.
>
> +config TI_ADS124S08
> + tristate "Texas Instruments ADS124S08"
> + depends on SPI && OF
> + help
> + If you say yes here you get support for Texas Instruments ADS124S08
> + and ADS124S06 ADC chips
> +
> + This driver can also be built as a module. If so, the module will be
> + called ti-ads124s08.
> +
> config TI_AM335X_ADC
> tristate "TI's AM335X ADC driver"
> depends on MFD_TI_AM335X_TSCADC && HAS_DMA
> diff --git a/drivers/iio/adc/Makefile b/drivers/iio/adc/Makefile
> index a6e6a0b659e2..d70293abfdba 100644
> --- a/drivers/iio/adc/Makefile
> +++ b/drivers/iio/adc/Makefile
> @@ -79,6 +79,7 @@ obj-$(CONFIG_TI_ADC161S626) += ti-adc161s626.o
> obj-$(CONFIG_TI_ADS1015) += ti-ads1015.o
> obj-$(CONFIG_TI_ADS7950) += ti-ads7950.o
> obj-$(CONFIG_TI_ADS8688) += ti-ads8688.o
> +obj-$(CONFIG_TI_ADS124S08) += ti-ads124s08.o
> obj-$(CONFIG_TI_AM335X_ADC) += ti_am335x_adc.o
> obj-$(CONFIG_TI_TLC4541) += ti-tlc4541.o
> obj-$(CONFIG_TWL4030_MADC) += twl4030-madc.o
> diff --git a/drivers/iio/adc/ti-ads124s08.c b/drivers/iio/adc/ti-ads124s08.c
> new file mode 100644
> index 000000000000..b6fc93f37355
> --- /dev/null
> +++ b/drivers/iio/adc/ti-ads124s08.c
> @@ -0,0 +1,437 @@
> +// SPDX-License-Identifier: GPL-2.0
> +// TI ADS124S0X chip family driver
> +// Copyright (C) 2018 Texas Instruments Incorporated - http://www.ti.com/
An oddity of current kernel style is that typically only the spdx
line is //

The rest are a normal kernel multiline comment.

> +
> +#include <linux/err.h>
> +#include <linux/delay.h>
> +#include <linux/device.h>
> +#include <linux/kernel.h>
> +#include <linux/module.h>
> +#include <linux/of.h>
> +#include <linux/of_gpio.h>
> +#include <linux/slab.h>
> +#include <linux/sysfs.h>
> +
> +#include <linux/gpio/consumer.h>
> +#include <linux/regulator/consumer.h>
> +#include <linux/spi/spi.h>
> +
> +#include <linux/iio/iio.h>
> +#include <linux/iio/buffer.h>
> +#include <linux/iio/trigger_consumer.h>
> +#include <linux/iio/triggered_buffer.h>
> +#include <linux/iio/sysfs.h>
> +
> +#define ADS124S08_ID 0x00
> +#define ADS124S06_ID 0x01

Use an enum for these as the actual values don't matter, only that
they are unique.

> +
> +/* Commands */
> +#define ADS124S_CMD_NOP 0x00
Why this shorter prefix? Are all ADS124S parts
compatible with this list?

> +#define ADS124S_CMD_WAKEUP 0x02
> +#define ADS124S_CMD_PWRDWN 0x04
> +#define ADS124S_CMD_RESET 0x06
> +#define ADS124S_CMD_START 0x08
> +#define ADS124S_CMD_STOP 0x0a
> +#define ADS124S_CMD_SYOCAL 0x16
> +#define ADS124S_CMD_SYGCAL 0x17
> +#define ADS124S_CMD_SFOCAL 0x19
> +#define ADS124S_CMD_RDATA 0x12
> +#define ADS124S_CMD_RREG 0x20
> +#define ADS124S_CMD_WREG 0x40
> +
> +/* Registers */
> +#define ADS124S0X_ID 0x00
Avoid wild cards in naming. It always goes wrong when
along comes another part that is similar but not quite the
same yet fits in your naming scheme. Just pick a part
and name after that.

> +#define ADS124S0X_STATUS 0x01
> +#define ADS124S0X_INPUT_MUX 0x02
> +#define ADS124S0X_PGA 0x03
> +#define ADS124S0X_DATA_RATE 0x04
> +#define ADS124S0X_REF 0x05
> +#define ADS124S0X_IDACMAG 0x06
> +#define ADS124S0X_IDACMUX 0x07
> +#define ADS124S0X_VBIAS 0x08
> +#define ADS124S0X_SYS 0x09
> +#define ADS124S0X_OFCAL0 0x0a
> +#define ADS124S0X_OFCAL1 0x0b
> +#define ADS124S0X_OFCAL2 0x0c
> +#define ADS124S0X_FSCAL0 0x0d
> +#define ADS124S0X_FSCAL1 0x0e
> +#define ADS124S0X_FSCAL2 0x0f
> +#define ADS124S0X_GPIODAT 0x10
> +#define ADS124S0X_GPIOCON 0x11
> +
> +/* ADS124S0x common channels */
> +#define ADS124S0X_AIN0 0x00
> +#define ADS124S0X_AIN1 0x01
> +#define ADS124S0X_AIN2 0x02
> +#define ADS124S0X_AIN3 0x03
> +#define ADS124S0X_AIN4 0x04
> +#define ADS124S0X_AIN5 0x05
> +#define ADS124S08_AINCOM 0x0c
> +/* ADS124S08 only channels */
> +#define ADS124S08_AIN6 0x06
> +#define ADS124S08_AIN7 0x07
> +#define ADS124S08_AIN8 0x08
> +#define ADS124S08_AIN9 0x09
> +#define ADS124S08_AIN10 0x0a
> +#define ADS124S08_AIN11 0x0b
> +#define ADS124S0X_MAX_CHANNELS 12
> +
> +#define ADS124S0X_POS_MUX_SHIFT 0x04
> +#define ADS124S_INT_REF 0x09
> +
> +#define ADS124S_START_REG_MASK 0x1f
> +#define ADS124S_NUM_BYTES_MASK 0x1f
> +
> +#define ADS124S_START_CONV 0x01
> +#define ADS124S_STOP_CONV 0x00
> +
> +struct ads124s_chip_info {
> + const struct iio_chan_spec *channels;
> + unsigned int num_channels;
> +};
> +
> +struct ads124s_private {
> + const struct ads124s_chip_info *chip_info;
> + struct gpio_desc *reset_gpio;
> + struct spi_device *spi;
> + struct regulator *reg;
> + struct mutex lock;
> + unsigned int vref_mv;
> + u8 data[ADS124S0X_MAX_CHANNELS];
This will break in horrible ways on spi controllers using DMA.
The data buffer needs to be in it's own cacheline to avoid
possibly overwriting data in the rest of this structure.

___cache_line_aligned is your friend. Wolfram did a very
good talk on this issue at elce.
https://events.linuxfoundation.org/wp-content/uploads/2017/12/20181023-Wolfram-Sang-ELCE18-safe_dma_buffers.pdf

There's a video on youtube as well.

He was addressing i2c, but the arguments still hold and unlike
i2c, spi has always had dma controllers who assume they have
dma safe buffers.

> +};
> +
> +#define ADS124S_CHAN(index) \
> +{ \
> + .type = IIO_VOLTAGE, \
> + .indexed = 1, \
> + .channel = index, \
> + .info_mask_separate = BIT(IIO_CHAN_INFO_RAW), \

It looks like you started adding scale but didn't finish doing so.
It's definitely a good thing to have so please see if that can be
added for V2.

> + .scan_index = index, \
> + .scan_type = { \
> + .sign = 'u', \
> + .realbits = 24, \
> + .storagebits = 24, \
> + .endianness = IIO_BE, \
> + }, \
> +}
> +
> +static const struct iio_chan_spec ads124s06_channels[] = {
> + ADS124S_CHAN(0),
> + ADS124S_CHAN(1),
> + ADS124S_CHAN(2),
> + ADS124S_CHAN(3),
> + ADS124S_CHAN(4),
> + ADS124S_CHAN(5),
> +};
> +
> +static const struct iio_chan_spec ads124s08_channels[] = {
> + ADS124S_CHAN(0),
> + ADS124S_CHAN(1),
> + ADS124S_CHAN(2),
> + ADS124S_CHAN(3),
> + ADS124S_CHAN(4),
> + ADS124S_CHAN(5),
> + ADS124S_CHAN(6),
> + ADS124S_CHAN(7),
> + ADS124S_CHAN(8),
> + ADS124S_CHAN(9),
> + ADS124S_CHAN(10),
> + ADS124S_CHAN(11),
> +};
> +
> +static const struct ads124s_chip_info ads124s_chip_info_tbl[] = {
> + [ADS124S08_ID] = {
> + .channels = ads124s08_channels,
> + .num_channels = ARRAY_SIZE(ads124s08_channels),
> + },
> + [ADS124S06_ID] = {
> + .channels = ads124s06_channels,
> + .num_channels = ARRAY_SIZE(ads124s06_channels),
> + },
> +};
> +
> +static void ads124s_fill_nop(u8 *data, int start, int count)
> +{
> + int i;
> +
> + for (i = start; i <= count; i++)
> + data[i] = ADS124S_CMD_NOP;

memset

> +
> +};
> +
> +static int ads124s_write_cmd(struct iio_dev *indio_dev, u8 command)
> +{
> + struct ads124s_private *priv = iio_priv(indio_dev);
> + struct spi_transfer t[] = {
> + {
> + .tx_buf = &priv->data[0],
> + .len = 1,
> + }
> + };
> +
> + priv->data[0] = command;
> +
> + return spi_sync_transfer(priv->spi, t, ARRAY_SIZE(t));

spi_write?

> +}
> +
> +static int ads124s_write_reg(struct iio_dev *indio_dev, u8 reg, u8 data)
> +{
> + struct ads124s_private *priv = iio_priv(indio_dev);
> + struct spi_transfer t[] = {
> + {
> + .tx_buf = &priv->data[0],
> + .len = 3,
> + }
> + };
> +
> + priv->data[0] = ADS124S_CMD_WREG | reg;
> + priv->data[1] = 0x0;
> + priv->data[2] = data;
> +
> + return spi_sync_transfer(priv->spi, t, ARRAY_SIZE(t));

spi_write?

> +}
> +
> +static int ads124s_reset(struct iio_dev *indio_dev)
> +{
> + struct ads124s_private *priv = iio_priv(indio_dev);
> +
> + if (priv->reset_gpio) {
> + gpiod_direction_output(priv->reset_gpio, 0);
I'd expect that gpio to always be in the output
direction, so just being controlled here.

So gpiod_set_value

> + udelay(200);
> + gpiod_direction_output(priv->reset_gpio, 1);
> + } else {
> + return ads124s_write_cmd(indio_dev, ADS124S_CMD_RESET);
> + }
> +
> + return 0;
> +};
> +
> +static int ads124s_read(struct iio_dev *indio_dev, unsigned int chan)
> +{
> + struct ads124s_private *priv = iio_priv(indio_dev);
> + int ret;
> + u32 tmp;
> + struct spi_transfer t[] = {
> + {
> + .tx_buf = &priv->data[0],
> + .len = 4,
> + .cs_change = 1,
> + }, {
> + .tx_buf = &priv->data[1],
> + .rx_buf = &priv->data[1],
> + .len = 4,
> + },
> + };

Pity we don't have a spi_write_the_read_with_cs or
something like that. I wonder how common this structure is?

> +
> + priv->data[0] = ADS124S_CMD_RDATA;
> + ads124s_fill_nop(priv->data, 1, 3);
> +
> + ret = spi_sync_transfer(priv->spi, t, ARRAY_SIZE(t));
> + if (ret < 0)
> + return ret;
> +
> + tmp = priv->data[2] << 16 | priv->data[3] << 8 | priv->data[4];
> +
> + return tmp;
> +}
> +
> +static int ads124s_read_raw(struct iio_dev *indio_dev,
> + struct iio_chan_spec const *chan,
> + int *val, int *val2, long m)
> +{
> + struct ads124s_private *priv = iio_priv(indio_dev);
> + int ret;
> +
> + mutex_lock(&priv->lock);
> + switch (m) {
> + case IIO_CHAN_INFO_RAW:
> + ret = ads124s_write_reg(indio_dev, ADS124S0X_INPUT_MUX,
> + chan->channel);
> + if (ret) {
> + dev_err(&priv->spi->dev, "Set ADC CH failed\n");

You are eating a bus error in favour of -EINVAL. Please don't!

> + goto out;
> + }
> +
> + ret = ads124s_write_cmd(indio_dev, ADS124S_START_CONV);
> + if (ret) {
> + dev_err(&priv->spi->dev, "Start ADC converions failed\n");
> + goto out;
> + }
> +
> + ret = ads124s_read(indio_dev, chan->channel);
> + if (ret < 0) {
> + dev_err(&priv->spi->dev, "Read ADC failed\n");
> + goto out;
> + } else
> + *val = ret;

This is the same code block as found in the buffered version below. Factor
it out to a utility function and use it both paths.

> +
> + ret = ads124s_write_cmd(indio_dev, ADS124S_STOP_CONV);
> + if (ret) {
> + dev_err(&priv->spi->dev, "Stop ADC converions failed\n");
> + goto out;
> + }
> +
> + mutex_unlock(&priv->lock);
> + if (ret < 0)
> + return ret;
Given you need to unlock in both this path and the error path, cleaner
to just set ret here and break. Then return ret below having set
it appropriately in the error paths.
> +
> + return IIO_VAL_INT;
> + default:
> + break;
> + }
> +out:
> + mutex_unlock(&priv->lock);
> + return -EINVAL;
> +}
> +
> +static const struct iio_info ads124s_info = {
> + .read_raw = &ads124s_read_raw,
> +};
> +
> +static irqreturn_t ads124s_trigger_handler(int irq, void *p)
> +{
> + struct iio_poll_func *pf = p;
> + struct iio_dev *indio_dev = pf->indio_dev;
> + struct ads124s_private *priv = iio_priv(indio_dev);
> + u16 buffer[ADS124S0X_MAX_CHANNELS];

There is an unfortunate oddity in how push_to_buffers_with_timestamp
works in that it builds the data for the kfifo inline in the buffer
provided. Thus that buffer has to have room for the channels and
the 64 bit timestamp.

> + int i, j = 0;
> + int ret;
> +
> + for (i = 0; i < indio_dev->masklength; i++) {
> + if (!test_bit(i, indio_dev->active_scan_mask))

for_each_bit_set

> + continue;
> +
> + ret = ads124s_write_reg(indio_dev, ADS124S0X_INPUT_MUX, i);

Does this need to be rewritten if the channel is already correct?

> + if (ret)
> + dev_err(&priv->spi->dev, "Set ADC CH failed\n");
> +
> + ret = ads124s_write_cmd(indio_dev, ADS124S_START_CONV);
> + if (ret)
> + dev_err(&priv->spi->dev, "Start ADC converions failed\n");
> +
> + buffer[j] = ads124s_read(indio_dev, i);
> + ret = ads124s_write_cmd(indio_dev, ADS124S_STOP_CONV);
> + if (ret)
> + dev_err(&priv->spi->dev, "Stop ADC converions failed\n");
> +
> + j++;
> + }
> +
> + iio_push_to_buffers_with_timestamp(indio_dev, buffer,
> + pf->timestamp);
> +
> + iio_trigger_notify_done(indio_dev->trig);
> +
> + return IRQ_HANDLED;
> +}
> +
> +static int ads124s_probe(struct spi_device *spi)
> +{
> + struct ads124s_private *ads124s_priv;
> + struct iio_dev *indio_dev;
> + int ret;
> +
> + indio_dev = devm_iio_device_alloc(&spi->dev, sizeof(*ads124s_priv));
> + if (indio_dev == NULL)
> + return -ENOMEM;
> +
> + ads124s_priv = iio_priv(indio_dev);
> +
> + ads124s_priv->reg = devm_regulator_get_optional(&spi->dev, "vref");
> + if (!IS_ERR(ads124s_priv->reg)) {
> + ret = regulator_enable(ads124s_priv->reg);
> + if (ret)
> + return ret;

As mentioned below, use devm_add_action_or_reset to provide
automatic disabling of this regulator letting you used managed
cleanup throughout.

> +
> + ads124s_priv->vref_mv = regulator_get_voltage(ads124s_priv->reg);

This doesn't actually seem to be used... As a general rule the only time you
want to read this is to provide a scale value. As that isn't a fast path
it's better to read it where needed in case the value is changing (not unusual
during boot up if the voltage is being used for several things).

> + if (ads124s_priv->vref_mv <= 0)
> + goto err_regulator_disable;
> + } else {
> + /* Use internal reference */
> + ads124s_priv->vref_mv = 0;
> + }
> +
> + ads124s_priv->reset_gpio = devm_gpiod_get_optional(&spi->dev,
> + "reset", GPIOD_OUT_LOW);
> + if (IS_ERR(ads124s_priv->reset_gpio))
> + dev_info(&spi->dev, "Reset GPIO not defined\n");
> +
> + ads124s_priv->chip_info = &ads124s_chip_info_tbl[spi_get_device_id(spi)->driver_data];
> +
> + spi_set_drvdata(spi, indio_dev);
> +
> + ads124s_priv->spi = spi;
> +
> + indio_dev->name = spi_get_device_id(spi)->name;
> + indio_dev->dev.parent = &spi->dev;
> + indio_dev->dev.of_node = spi->dev.of_node;
> + indio_dev->modes = INDIO_DIRECT_MODE;
> + indio_dev->channels = ads124s_priv->chip_info->channels;
> + indio_dev->num_channels = ads124s_priv->chip_info->num_channels;
> + indio_dev->info = &ads124s_info;
> +
> + mutex_init(&ads124s_priv->lock);
> +
> + ret = iio_triggered_buffer_setup(indio_dev, NULL,
> + ads124s_trigger_handler, NULL);
> + if (ret < 0) {
> + dev_err(&spi->dev, "iio triggered buffer setup failed\n");
> + goto err_regulator_disable;
> + }
> +
> + ads124s_reset(indio_dev);
> +
> + ret = iio_device_register(indio_dev);
> + if (ret)
> + goto err_buffer_cleanup;
> +
> + return 0;
> +
> +err_buffer_cleanup:
> + iio_triggered_buffer_cleanup(indio_dev);
> +
> +err_regulator_disable:
> + if (!IS_ERR(ads124s_priv->reg))
> + regulator_disable(ads124s_priv->reg);
> +
> + return ret;
> +}
> +
> +static int ads124s_remove(struct spi_device *spi)
> +{
> + struct iio_dev *indio_dev = spi_get_drvdata(spi);
> +
> + iio_device_unregister(indio_dev);
> + iio_triggered_buffer_cleanup(indio_dev);
Both of these functions can be automatically cleaned
up if you use the devm_ variants of the registration functions
which makes me suspicious.. Ah. The regulator disable
is missing.

You can move to fully managed if you use devm_add_action_or_reset
to turn the regulator off.

> +
> + return 0;
> +}
> +
> +static const struct spi_device_id ads124s_id[] = {
> + { "ads124s08", ADS124S08_ID },
> + { "ads124s06", ADS124S06_ID },
> + { }
> +};
> +MODULE_DEVICE_TABLE(spi, ads124s_id);
> +
> +static const struct of_device_id ads124s_of_table[] = {
> + { .compatible = "ti,ads124s08" },
> + { .compatible = "ti,ads124s06" },

It's trivial but numerical order preferred as it makes it
obvious where to add new devices later.

> + { },
> +};
> +MODULE_DEVICE_TABLE(of, ads124s_of_table);
> +
> +static struct spi_driver ads124s_driver = {
> + .driver = {
> + .name = "ads124s",
> + .of_match_table = ads124s_of_table,
> + },
> + .probe = ads124s_probe,
> + .remove = ads124s_remove,
> + .id_table = ads124s_id,
> +};
> +module_spi_driver(ads124s_driver);
> +
> +MODULE_AUTHOR("Dan Murphy <dmuprhy@xxxxxx>");
> +MODULE_DESCRIPTION("TI TI_ADS12S0X ADC");
> +MODULE_LICENSE("GPL v2");