[2/2] iio: accel: Add driver for Murata SCA3300 accelerometer

Commit Message

Tomas Melin April 12, 2021, 7:50 a.m. UTC

Add initial support for Murata SCA3300 3-axis industrial
accelerometer with digital SPI interface. This device also
provides a temperature measurement.

Device product page including datasheet can be found at:
https://www.murata.com/en-global/products/sensor/accel/sca3300

Signed-off-by: Tomas Melin <[e-mail]>
---
 drivers/iio/accel/Kconfig   |  13 ++
 drivers/iio/accel/Makefile  |   1 +
 drivers/iio/accel/sca3300.c | 434 ++++++++++++++++++++++++++++++++++++
 3 files changed, 448 insertions(+)
 create mode 100644 drivers/iio/accel/sca3300.c

Comments

Jonathan Cameron April 12, 2021, 11:08 a.m. UTC | #1

On Mon, 12 Apr 2021 10:50:56 +0300
Tomas Melin <[e-mail]> wrote:

> Add initial support for Murata SCA3300 3-axis industrial
> accelerometer with digital SPI interface. This device also
> provides a temperature measurement.
> 
> Device product page including datasheet can be found at:
> https://www.murata.com/en-global/products/sensor/accel/sca3300
> 
> Signed-off-by: Tomas Melin <[e-mail]>

Hi Tomas,

Comments below.

Thanks,

Jonathan

> ---
>  drivers/iio/accel/Kconfig   |  13 ++
>  drivers/iio/accel/Makefile  |   1 +
>  drivers/iio/accel/sca3300.c | 434 ++++++++++++++++++++++++++++++++++++
>  3 files changed, 448 insertions(+)
>  create mode 100644 drivers/iio/accel/sca3300.c
> 
> diff --git a/drivers/iio/accel/Kconfig b/drivers/iio/accel/Kconfig
> index cceda3cecbcf..0dbf7b648e8a 100644
> --- a/drivers/iio/accel/Kconfig
> +++ b/drivers/iio/accel/Kconfig
> @@ -450,6 +450,19 @@ config SCA3000
>  	  To compile this driver as a module, say M here: the module will be
>  	  called sca3000.
>  
> +config SCA3300
> +	tristate "Murata SCA3300 3-Axis Accelerometer Driver"
> +	depends on SPI
> +	select CRC8
> +	select IIO_BUFFER
> +	select IIO_TRIGGERED_BUFFER
> +	help
> +	  Say yes here to build support for Murata SCA3300 3-Axis
> +	  accelerometer.
> +
> +	  To compile this driver as a module, choose M here: the module will be
> +	  called sca3300.
> +
>  config STK8312
>  	tristate "Sensortek STK8312 3-Axis Accelerometer Driver"
>  	depends on I2C
> diff --git a/drivers/iio/accel/Makefile b/drivers/iio/accel/Makefile
> index 32cd1342a31a..4b56527a2b97 100644
> --- a/drivers/iio/accel/Makefile
> +++ b/drivers/iio/accel/Makefile
> @@ -50,6 +50,7 @@ obj-$(CONFIG_MXC4005)		+= mxc4005.o
>  obj-$(CONFIG_MXC6255)		+= mxc6255.o
>  
>  obj-$(CONFIG_SCA3000)		+= sca3000.o
> +obj-$(CONFIG_SCA3300)		+= sca3300.o
>  
>  obj-$(CONFIG_STK8312)		+= stk8312.o
>  obj-$(CONFIG_STK8BA50)		+= stk8ba50.o
> diff --git a/drivers/iio/accel/sca3300.c b/drivers/iio/accel/sca3300.c
> new file mode 100644
> index 000000000000..112fb88ecd3a
> --- /dev/null
> +++ b/drivers/iio/accel/sca3300.c
> @@ -0,0 +1,434 @@
> +// SPDX-License-Identifier: GPL-2.0-only
> +/*
> + * Copyright (c) Vaisala Oyj. All rights reserved.

Give a year for the copyright notice if you can.

> + */
> +#include <linux/crc8.h>
> +#include <linux/delay.h>
> +#include <linux/iio/buffer.h>
> +#include <linux/iio/iio.h>
> +#include <linux/iio/sysfs.h>
> +#include <linux/iio/trigger_consumer.h>
> +#include <linux/iio/triggered_buffer.h>
> +#include <linux/kernel.h>
> +#include <linux/module.h>
> +#include <linux/spi/spi.h>
> +
> +#define SCA3300_ALIAS "sca3300"
> +
> +#define SCA3300_REG_STATUS 0x6
> +#define SCA3300_REG_MODE 0xd
> +#define SCA3300_REG_WHOAMI 0x10
> +#define SCA3300_VALUE_SW_RESET 0x20
> +#define SCA3300_CRC8_POLYNOMIAL 0x1d
> +#define SCA3300_X_READ 0

I wouldn't bother defining this.

> +#define SCA3300_X_WRITE BIT(7)

Even this one is something I'd just put inline with a comment.

> +#define SCA3300_DEVICE_ID 0x51
> +#define SCA3300_RS_ERROR 0x3
> +
> +enum sca3300_scan_indexes {
> +	SCA3300_ACC_X = 0,
> +	SCA3300_ACC_Y,
> +	SCA3300_ACC_Z,
> +	SCA3300_TEMP,
> +	SCA3300_TIMESTAMP,
> +};
> +
> +#define SCA3300_ACCEL_CHANNEL(index, reg, axis) {			\
> +		.type = IIO_ACCEL,					\
> +		.address = reg,						\
> +		.modified = 1,						\
> +		.channel2 = IIO_MOD_##axis,				\
> +		.info_mask_separate = BIT(IIO_CHAN_INFO_RAW) |		\
> +				      BIT(IIO_CHAN_INFO_PROCESSED),	\

As mentioned below, don't provide PROCESSED. Userspace is better at handling the
conversion so leave it to them.

> +		.info_mask_shared_by_type = BIT(IIO_CHAN_INFO_SCALE),	\
> +		.scan_index = index,					\
> +		.scan_type = {						\
> +			.sign = 's',					\
> +			.realbits = 16,					\
> +			.storagebits = 16,				\
> +			.shift = 0,					\
> +			.endianness = IIO_CPU,				\
> +		},							\
> +	}
> +
> +static const struct iio_chan_spec sca3300_channels[] = {
> +	SCA3300_ACCEL_CHANNEL(SCA3300_ACC_X, 0x1, X),
> +	SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Y, 0x2, Y),
> +	SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Z, 0x3, Z),
> +	{
> +		.type = IIO_TEMP,
> +		.address = 0x5,
> +		.scan_index = SCA3300_TEMP,
> +		.info_mask_separate = BIT(IIO_CHAN_INFO_RAW),
> +		.scan_type = {
> +			.sign = 's',
> +			.realbits = 16,
> +			.storagebits = 16,
> +			.shift = 0,
> +			.endianness = IIO_CPU,
> +		},
> +	},
> +	IIO_CHAN_SOFT_TIMESTAMP(4),
> +};
> +
> +static const int sca3300_accel_scale[] = {2700, 1350, 5400, 5400};
> +
> +static const unsigned long sca3300_scan_masks[] = {
> +	BIT(SCA3300_ACC_X) | BIT(SCA3300_ACC_Y) | BIT(SCA3300_ACC_Z) |
> +	BIT(SCA3300_TEMP),
> +	0};
> +
> +/**
> + * SCA3300 device data

run scripts/kernel-doc over the file and fix the warnings + errors.

> + *
> + * @spi SPI device structure
> + * @opmode Device operation mode
> + * @lock Data buffer lock
> + * @txbuf Transmit buffer
> + * @rxbuf Receive buffer
> + * @scan Triggered buffer. Four channel 16-bit data + 64-bit timestamp
> + */
> +struct sca3300_data {
> +	struct spi_device *spi;
> +	u32 opmode;
> +	struct mutex lock;
> +	u8 txbuf[4];
> +	u8 rxbuf[4];
> +	struct {
> +		s16 channels[4];
> +		s64 ts __aligned(sizeof(s64));
> +	} scan;
> +};
> +
> +DECLARE_CRC8_TABLE(sca3300_crc_table);
> +
> +static int sca3300_transfer(struct sca3300_data *sca_data, int *val)
> +{
> +	struct spi_delay delay = {.value = 10, .unit = SPI_DELAY_UNIT_USECS};
> +	int32_t ret;
> +	int rs;
> +	u8 crc;
> +	struct spi_transfer xfers[2] = {
> +		{
> +			.tx_buf = sca_data->txbuf,
> +			.rx_buf = NULL,
> +			.len = ARRAY_SIZE(sca_data->txbuf),
> +			.delay = delay,
> +			.cs_change = 1,
> +		},
> +		{
> +			.tx_buf = NULL,
> +			.rx_buf = sca_data->rxbuf,
> +			.len = ARRAY_SIZE(sca_data->rxbuf),
> +			.delay = delay,
> +			.cs_change = 0,
> +		}
> +	};
> +
> +	/* inverted crc value as described in device data sheet */
> +	crc = ~crc8(sca3300_crc_table, &sca_data->txbuf[0], 3, CRC8_INIT_VALUE);
> +	sca_data->txbuf[3] = crc;
> +
> +	ret = spi_sync_transfer(sca_data->spi, xfers, 2);

Use ARRAY_SIZE(xfers) instead of 2.

> +	if (ret < 0) {
> +		dev_err(&sca_data->spi->dev,
> +			"transfer error, error: %d\n", ret);
> +		return -EIO;
> +	}
> +
> +	crc = ~crc8(sca3300_crc_table, &sca_data->rxbuf[0], 3, CRC8_INIT_VALUE);
> +	if (sca_data->rxbuf[3] != crc) {
> +		dev_err(&sca_data->spi->dev, "CRC checksum mismatch");
> +		return -EIO;
> +	}
> +
> +	/* get return status */
> +	rs = sca_data->rxbuf[0] & 0x03;
> +	if (rs == SCA3300_RS_ERROR)
> +		return rs;
> +
> +	*val = (s16)(sca_data->rxbuf[2] | (sca_data->rxbuf[1] << 8));

Preference for an unaligned endian conversion here and explicit sign extend e.g
something like 

sign_extend32(get_unaligned_le16(&sca_data->rxbuf[1]), 15)

> +
> +	return 0;
> +}
> +
> +static int sca3300_read_reg(struct sca3300_data *sca_data, u8 reg, int *val)
> +{
> +	int ret;
> +
> +	mutex_lock(&sca_data->lock);
> +	sca_data->txbuf[0] = SCA3300_X_READ | (reg << 2);
> +	ret = sca3300_transfer(sca_data, val);
> +	if (ret > 0) {
> +		sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS << 2);
> +		ret = sca3300_transfer(sca_data, val);
> +		/* status 0 = startup, 0x2 = mode change */
> +		if (ret > 0 && *val != 0 && *val != 0x2) {
> +			dev_err_ratelimited(&sca_data->spi->dev,
> +					    "device status: %x\n",
> +					    (u16)*val);
> +			mutex_unlock(&sca_data->lock);
> +			return -EIO;
> +		}
> +		if (ret > 0)
> +			ret = 0;
> +	}
> +	mutex_unlock(&sca_data->lock);
> +
> +	return ret;
> +}
> +
> +static int sca3300_write_reg(struct sca3300_data *sca_data, u8 reg, int val)
> +{
> +	int reg_val = 0;
> +	int ret;
> +
> +	mutex_lock(&sca_data->lock);
> +	sca_data->txbuf[0] = SCA3300_X_WRITE | (reg << 2);
> +	sca_data->txbuf[1] = val >> 8;
> +	sca_data->txbuf[2] = val & 0xFF;

Prefer (Slightly) an unaligned put.

> +	ret = sca3300_transfer(sca_data, &reg_val);
> +	if (ret > 0) {

Factor this error handling out to another function, plus trigger it from an
appropriate standard error code rather than a positive return value.

> +		sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS << 2);
> +		ret = sca3300_transfer(sca_data, &reg_val);
> +		/* status 0 = startup, 0x2 = mode change */
> +		if (ret > 0 && reg_val != 0 && reg_val != 0x2) {
> +			dev_err_ratelimited(&sca_data->spi->dev,
> +					    "device status: %x\n",
> +					    (u16)reg_val);
> +			mutex_unlock(&sca_data->lock);
> +			return -EIO;
> +		}
> +		if (ret > 0)
> +			ret = 0;

Whenever you get this sort of juggling it rather implies your design is
less than ideal.

> +	}
> +	mutex_unlock(&sca_data->lock);
> +
> +	return ret;
> +}
> +
> +static int sca3300_write_raw(struct iio_dev *indio_dev,
> +			     struct iio_chan_spec const *chan,
> +			     int val, int val2, long mask)
> +{
> +	struct sca3300_data *data = iio_priv(indio_dev);
> +
> +	switch (mask) {
> +	case IIO_CHAN_INFO_SCALE:
> +		if (val < 0 || val > 3)
> +			return -EINVAL;

No.  Scale values need to be the value of 1LSB not a random integer
that requires you to look up the datasheet.

> +		return sca3300_write_reg(data, SCA3300_REG_MODE, val);
> +	default:
> +		return -EINVAL;
> +	}
> +}
> +
> +static int sca3300_read_raw(struct iio_dev *indio_dev,
> +			    struct iio_chan_spec const *chan,
> +			    int *val, int *val2, long mask)
> +{
> +	struct sca3300_data *data = iio_priv(indio_dev);
> +	int ret;
> +	int reg_val;
> +
> +	switch (mask) {
> +	case IIO_CHAN_INFO_RAW:
> +		ret = sca3300_read_reg(data, chan->address, val);
> +		if (ret < 0)
> +			return ret;
> +		return IIO_VAL_INT;
> +	case IIO_CHAN_INFO_SCALE:
> +		ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
> +		if (ret < 0)
> +			return ret;
> +		*val = sca3300_accel_scale[reg_val];

Scale is very very rarely an integer value which makes me wonder what
you are returning here...  It should be the multiplier needed to
take the raw value to a reading in m/sec^2  Given you use it as fractional
below, I'm guessing this is 1/scale_value?

> +		return IIO_VAL_INT;
> +	case IIO_CHAN_INFO_PROCESSED:
Don't provide processed here.   Userspace needs to do the conversion itself
(and given it has floating point easily available will probably do a better
 job than we can).   I'm assuming the mode can't autonomously change?

> +		ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
> +		if (ret < 0)
> +			return ret;
> +		*val2 = sca3300_accel_scale[reg_val];
> +		ret = sca3300_read_reg(data, chan->address, val);
> +		if (ret < 0)
> +			return ret;
> +		return IIO_VAL_FRACTIONAL;
> +	default:
> +		return -EINVAL;
> +	}
> +}
> +
> +static irqreturn_t sca3300_trigger_handler(int irq, void *p)
> +{
> +	struct iio_poll_func *pf = p;
> +	struct iio_dev *indio_dev = pf->indio_dev;
> +	struct sca3300_data *data = iio_priv(indio_dev);
> +	s64 time_ns = iio_get_time_ns(indio_dev);

Is the timestamp at the start more accurate that that at the end
of reading the channels?  If not, just put this inline in the
iio_push_to_buffers_with_timestamp() call.

> +	int bit, ret, val, i = 0;
> +
> +	for_each_set_bit(bit, indio_dev->active_scan_mask,
> +			 indio_dev->masklength) {
> +		ret = sca3300_read_reg(data, sca3300_channels[bit].address,
> +				       &val);
> +		if (ret < 0)

Given we can't return an error code from interrupt handlers, it is usually
a good idea to print something to the log. Otherwise we get missing data
with no idea of why..

> +			goto out;
> +		if (ARRAY_SIZE(data->scan.channels) > i)

How could this not be true?  If it's always true, don't bother
checking i.

> +			((s16 *)data->scan.channels)[i++] = val;
> +	}
> +
> +	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan, time_ns);
> +out:
> +	iio_trigger_notify_done(indio_dev->trig);
> +
> +	return IRQ_HANDLED;
> +}
> +
> +static int sca3300_init(struct sca3300_data *sca_data,
> +			struct iio_dev *indio_dev)
> +{
> +	int ret;
> +	int value = 0;
> +
> +	if (sca_data->opmode < 1 || sca_data->opmode > 4)
> +		return -EINVAL;

Whilst this is going anyway, better to check that next to where
it is read so we drop out immediately rather than buried in this
function.

> +
> +	ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
> +				SCA3300_VALUE_SW_RESET);
> +	if (ret != 0)
> +		return ret;

For sleeps, good to reference which section in data sheet gives the
timing.

> +	usleep_range(2e3, 10e3);
> +
> +	ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
> +				sca_data->opmode - 1);
> +	if (ret != 0)
if (ret) see below.

> +		return ret;
> +	msleep(100);

Again, document where the time comes from. Saves reviewers time if
they want to check it. (I'm too lazy :)

> +	ret = sca3300_read_reg(sca_data, SCA3300_REG_WHOAMI, &value);
> +	if (ret != 0)

if (ret) is more idiomatic in kernel code.

> +		return ret;
> +
> +	if (value != SCA3300_DEVICE_ID) {
> +		dev_err(&sca_data->spi->dev, "device id not expected value\n");

Perhaps useful to print what was expected and what was seen.  Maybe it's
a new variant and this might help the user to identify that and add it to the
driver.

> +		return -EIO;

-EINVAL probably more appropriate return.

> +	}
> +	return 0;
> +}
> +
> +static int sca3300_debugfs_reg_access(struct iio_dev *indio_dev,
> +				      unsigned int reg, unsigned int writeval,
> +				      unsigned int *readval)
> +{
> +	struct sca3300_data *data = iio_priv(indio_dev);
> +	int value;
> +	int ret;
> +
> +	if (reg > 0x1f)

Use a define that names that register.

> +		return -EINVAL;
> +
> +	if (!readval)
> +		return sca3300_write_reg(data, reg, writeval);
> +
> +	ret = sca3300_read_reg(data, reg, &value);
> +	if (ret < 0)
> +		return ret;
> +
> +	*readval = (unsigned int)value;
> +
> +	return 0;
> +}
> +
> +static const struct iio_info sca3300_info = {
> +	.read_raw = sca3300_read_raw,
> +	.write_raw = sca3300_write_raw,
> +	.debugfs_reg_access = &sca3300_debugfs_reg_access,
> +};
> +
> +static int sca3300_probe(struct spi_device *spi)
> +{
> +	struct sca3300_data *sca_data;
> +	struct iio_dev *indio_dev;
> +	int ret;
> +
> +	indio_dev = devm_iio_device_alloc(&spi->dev, sizeof(*sca_data));
> +	if (!indio_dev) {
> +		dev_err(&spi->dev,
> +			"failed to allocate memory for iio device\n");
> +		return -ENOMEM;
> +	}
> +
> +	sca_data = iio_priv(indio_dev);
> +	mutex_init(&sca_data->lock);
> +	sca_data->spi = spi;
> +	spi_set_drvdata(spi, indio_dev);

Won't be used once using devm as below and remove() is dropped, so drop this.

> +
> +	crc8_populate_msb(sca3300_crc_table, SCA3300_CRC8_POLYNOMIAL);
> +
> +	indio_dev->dev.parent = &spi->dev;
> +	indio_dev->info = &sca3300_info;
> +	indio_dev->name = SCA3300_ALIAS;
> +	indio_dev->modes = INDIO_DIRECT_MODE | INDIO_BUFFER_TRIGGERED;

BUFFER_TRIGGERED bit now set by iio_triggered_buffer_setup() so shouldn't
be ehre as well.

> +	indio_dev->channels = sca3300_channels;
> +	indio_dev->num_channels = ARRAY_SIZE(sca3300_channels);
> +	indio_dev->available_scan_masks = sca3300_scan_masks;
> +
> +	if (spi->dev.of_node) {
> +		ret = of_property_read_u32(spi->dev.of_node, "murata,opmode",

Please use generic firmware access calls rather than the of specific ones.
That lets us get ACPI support for free :)

As per the binding review however, this one needs to go in favour of userspace
control of scale + filter frequencies / sampling freq.

> +					   &sca_data->opmode);
> +		if (ret < 0)
> +			return ret;
> +	}
> +
> +	ret = sca3300_init(sca_data, indio_dev);
> +	if (ret < 0) {
> +		dev_err(&spi->dev, "failed to init device, error: %d\n", ret);
> +		return ret;
> +	}
> +
> +	ret = iio_triggered_buffer_setup(indio_dev, iio_pollfunc_store_time,
> +					 sca3300_trigger_handler, NULL);

devm_iio_triggered_buffer_setup() and you can drop the manual cleanup in remove()
and hence drop remove() entirely as nothing else left.

> +	if (ret < 0) {
> +		dev_err(&spi->dev,
> +			"iio triggered buffer setup failed, error: %d\n", ret);
> +		return ret;
> +	}
> +
> +	ret = devm_iio_device_register(&spi->dev, indio_dev);
> +	if (ret < 0) {
> +		dev_err(&spi->dev, "iio device register failed, error: %d\n",
> +			ret);
> +		iio_triggered_buffer_cleanup(indio_dev);
With devm above, no need to manually clean this up.

> +		return ret;
> +	}
> +
> +	return 0;
> +}
> +
> +static int sca3300_remove(struct spi_device *spi)
> +{
> +	struct iio_dev *indio_dev = spi_get_drvdata(spi);
> +
> +	iio_triggered_buffer_cleanup(indio_dev);
> +	return 0;
> +}
> +
> +static const struct of_device_id sca3300_dt_ids[] = {
> +	{ .compatible = "murata,sca3300"},
> +	{},
> +};
> +MODULE_DEVICE_TABLE(of, sca3300_dt_ids);
> +
> +static struct spi_driver sca3300_driver = {
> +	.driver = {
> +		.name		= SCA3300_ALIAS,
> +		.owner		= THIS_MODULE,
> +		.of_match_table = of_match_ptr(sca3300_dt_ids),
> +	},
> +
> +	.probe	= sca3300_probe,
> +	.remove	= sca3300_remove,
> +};
> +
> +module_spi_driver(sca3300_driver);
> +
> +MODULE_AUTHOR("Tomas Melin <[e-mail]>");
> +MODULE_DESCRIPTION("Murata SCA3300 SPI Accelerometer");
> +MODULE_LICENSE("GPL v2");

Tomas Melin April 15, 2021, 8:41 a.m. UTC | #2

On 4/12/21 2:08 PM, Jonathan Cameron wrote:

> On Mon, 12 Apr 2021 10:50:56 +0300
> Tomas Melin <[e-mail]> wrote:
>
>> Add initial support for Murata SCA3300 3-axis industrial
>> accelerometer with digital SPI interface. This device also
>> provides a temperature measurement.
>>
>> Device product page including datasheet can be found at:
>> https://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.murata.com%2Fen-global%2Fproducts%2Fsensor%2Faccel%2Fsca3300&amp;data=04%7C01%7Ctomas.melin%40vaisala.com%7Ca66811c95e6c404aa5a208d8fda3743c%7C6d7393e041f54c2e9b124c2be5da5c57%7C0%7C0%7C637538225758011604%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&amp;sdata=7bvCd52SFuYdFVlhYrknp0769%2B6hBpxdP6ln0OgKTrE%3D&amp;reserved=0
>>
>> Signed-off-by: Tomas Melin <[e-mail]>
> Hi Tomas,
>
> Comments below.
>
> Thanks,
>
> Jonathan

Hi Jonathan,

Thanks for your comments! I'll get back with a v2.

While working on updates I did notice something new which I cannot

reproduce on older (5.10.17 kernel) version. If compiling this as a 
module, getting error while

unloading module:

[   40.200084] Unable to handle kernel NULL pointer dereference at 
virtual address 00000104
...

[   40.510054] Backtrace:
[   40.512502] [<c06c5e0c>] (iio_device_ioctl_handler_unregister) from 
[<c06ca8c4>] (iio_buffers_free_sysfs_and_mask+0x2c/0x6c)
[   40.523735] [<c06ca898>] (iio_buffers_free_sysfs_and_mask) from 
[<c06c4a18>] (iio_device_unregister+0xa8/0xac)
[   40.533746]  r5:c1811228 r4:c1811000
[   40.537318] [<c06c4970>] (iio_device_unregister) from [<c06c4a38>] 
(devm_iio_device_unreg+0x1c/0x20)
[   40.546461]  r5:c2415000 r4:c25bab80
[   40.550025] [<c06c4a1c>] (devm_iio_device_unreg) from [<c0556820>] 
(release_nodes+0x1c0/0x1f0)
[   40.558654] [<c0556660>] (release_nodes) from [<c0556bb0>] 
(devres_release_all+0x40/0x60)
[   40.566847]  r10:00000081 r9:c2350000 r8:c0100264 r7:00000081 
r6:bf00c010 r5:c19be000
[   40.574669]  r4:c1a91c00
[   40.577194] [<c0556b70>] (devres_release_all) from [<c055216c>] 
(device_release_driver_internal+0x120/0x1cc)
[   40.587031]  r5:c19be000 r4:c1a91c00
[   40.590596] [<c055204c>] (device_release_driver_internal) from 
[<c05522b4>] (driver_detach+0x54/0x90)
[   40.599828]  r7:00000081 r6:00000000 r5:bf00c010 r4:c1a91c00
[   40.605482] [<c0552260>] (driver_detach) from [<c0550ba4>] 
(bus_remove_driver+0x5c/0xb0)
[   40.613583]  r5:00000800 r4:bf00c010
[   40.617148] [<c0550b48>] (bus_remove_driver) from [<c0552c38>] 
(driver_unregister+0x38/0x5c)
[   40.625596]  r5:00000800 r4:bf00c010
[   40.629161] [<c0552c00>] (driver_unregister) from [<bf00a760>] 
(sca3300_driver_exit+0x14/0x8b4 [sca3300])
[   40.638747]  r5:00000800 r4:bf00c080
[   40.642311] [<bf00a74c>] (sca3300_driver_exit [sca3300]) from 
[<c01c06f0>] (sys_delete_module+0x16c/0x238)
[   40.651990] [<c01c0584>] (sys_delete_module) from [<c0100244>] 
(__sys_trace_return+0x0/0x1c)
[   40.660435] Exception stack(0xc2351fa8 to 0xc2351ff0)
[   40.665484] 1fa0:                   0050e5a8 00000000 0050e5e4 
00000800 081d4b00 bec18af4
[   40.673661] 1fc0: 0050e5a8 00000000 bec18b50 00000081 bec18e51 
0050e190 00000001 bec18d3c
[   40.681834] 1fe0: 0050cf70 bec18afc 004f1ec8 b6ecb27c
[   40.686887]  r6:bec18b50 r5:00000000 r4:0050e5a8
[   40.691507] Code: e8bd4000 e1c020d0 e3a0cc01 e3001122 (e5823004)
[   40.707675] ---[ end trace 189882b050077333 ]---

This happens when building against linux-next 5.12.0-rc6-next-20210409. 
I'm failing to see what is wrong. Any ideas?

Thanks,

Tomas


>
>> ---
>>   drivers/iio/accel/Kconfig   |  13 ++
>>   drivers/iio/accel/Makefile  |   1 +
>>   drivers/iio/accel/sca3300.c | 434 ++++++++++++++++++++++++++++++++++++
>>   3 files changed, 448 insertions(+)
>>   create mode 100644 drivers/iio/accel/sca3300.c
>>
>> diff --git a/drivers/iio/accel/Kconfig b/drivers/iio/accel/Kconfig
>> index cceda3cecbcf..0dbf7b648e8a 100644
>> --- a/drivers/iio/accel/Kconfig
>> +++ b/drivers/iio/accel/Kconfig
>> @@ -450,6 +450,19 @@ config SCA3000
>>   	  To compile this driver as a module, say M here: the module will be
>>   	  called sca3000.
>>   
>> +config SCA3300
>> +	tristate "Murata SCA3300 3-Axis Accelerometer Driver"
>> +	depends on SPI
>> +	select CRC8
>> +	select IIO_BUFFER
>> +	select IIO_TRIGGERED_BUFFER
>> +	help
>> +	  Say yes here to build support for Murata SCA3300 3-Axis
>> +	  accelerometer.
>> +
>> +	  To compile this driver as a module, choose M here: the module will be
>> +	  called sca3300.
>> +
>>   config STK8312
>>   	tristate "Sensortek STK8312 3-Axis Accelerometer Driver"
>>   	depends on I2C
>> diff --git a/drivers/iio/accel/Makefile b/drivers/iio/accel/Makefile
>> index 32cd1342a31a..4b56527a2b97 100644
>> --- a/drivers/iio/accel/Makefile
>> +++ b/drivers/iio/accel/Makefile
>> @@ -50,6 +50,7 @@ obj-$(CONFIG_MXC4005)		+= mxc4005.o
>>   obj-$(CONFIG_MXC6255)		+= mxc6255.o
>>   
>>   obj-$(CONFIG_SCA3000)		+= sca3000.o
>> +obj-$(CONFIG_SCA3300)		+= sca3300.o
>>   
>>   obj-$(CONFIG_STK8312)		+= stk8312.o
>>   obj-$(CONFIG_STK8BA50)		+= stk8ba50.o
>> diff --git a/drivers/iio/accel/sca3300.c b/drivers/iio/accel/sca3300.c
>> new file mode 100644
>> index 000000000000..112fb88ecd3a
>> --- /dev/null
>> +++ b/drivers/iio/accel/sca3300.c
>> @@ -0,0 +1,434 @@
>> +// SPDX-License-Identifier: GPL-2.0-only
>> +/*
>> + * Copyright (c) Vaisala Oyj. All rights reserved.
> Give a year for the copyright notice if you can.
>
>> + */
>> +#include <linux/crc8.h>
>> +#include <linux/delay.h>
>> +#include <linux/iio/buffer.h>
>> +#include <linux/iio/iio.h>
>> +#include <linux/iio/sysfs.h>
>> +#include <linux/iio/trigger_consumer.h>
>> +#include <linux/iio/triggered_buffer.h>
>> +#include <linux/kernel.h>
>> +#include <linux/module.h>
>> +#include <linux/spi/spi.h>
>> +
>> +#define SCA3300_ALIAS "sca3300"
>> +
>> +#define SCA3300_REG_STATUS 0x6
>> +#define SCA3300_REG_MODE 0xd
>> +#define SCA3300_REG_WHOAMI 0x10
>> +#define SCA3300_VALUE_SW_RESET 0x20
>> +#define SCA3300_CRC8_POLYNOMIAL 0x1d
>> +#define SCA3300_X_READ 0
> I wouldn't bother defining this.
>
>> +#define SCA3300_X_WRITE BIT(7)
> Even this one is something I'd just put inline with a comment.
>
>> +#define SCA3300_DEVICE_ID 0x51
>> +#define SCA3300_RS_ERROR 0x3
>> +
>> +enum sca3300_scan_indexes {
>> +	SCA3300_ACC_X = 0,
>> +	SCA3300_ACC_Y,
>> +	SCA3300_ACC_Z,
>> +	SCA3300_TEMP,
>> +	SCA3300_TIMESTAMP,
>> +};
>> +
>> +#define SCA3300_ACCEL_CHANNEL(index, reg, axis) {			\
>> +		.type = IIO_ACCEL,					\
>> +		.address = reg,						\
>> +		.modified = 1,						\
>> +		.channel2 = IIO_MOD_##axis,				\
>> +		.info_mask_separate = BIT(IIO_CHAN_INFO_RAW) |		\
>> +				      BIT(IIO_CHAN_INFO_PROCESSED),	\
> As mentioned below, don't provide PROCESSED. Userspace is better at handling the
> conversion so leave it to them.
>
>> +		.info_mask_shared_by_type = BIT(IIO_CHAN_INFO_SCALE),	\
>> +		.scan_index = index,					\
>> +		.scan_type = {						\
>> +			.sign = 's',					\
>> +			.realbits = 16,					\
>> +			.storagebits = 16,				\
>> +			.shift = 0,					\
>> +			.endianness = IIO_CPU,				\
>> +		},							\
>> +	}
>> +
>> +static const struct iio_chan_spec sca3300_channels[] = {
>> +	SCA3300_ACCEL_CHANNEL(SCA3300_ACC_X, 0x1, X),
>> +	SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Y, 0x2, Y),
>> +	SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Z, 0x3, Z),
>> +	{
>> +		.type = IIO_TEMP,
>> +		.address = 0x5,
>> +		.scan_index = SCA3300_TEMP,
>> +		.info_mask_separate = BIT(IIO_CHAN_INFO_RAW),
>> +		.scan_type = {
>> +			.sign = 's',
>> +			.realbits = 16,
>> +			.storagebits = 16,
>> +			.shift = 0,
>> +			.endianness = IIO_CPU,
>> +		},
>> +	},
>> +	IIO_CHAN_SOFT_TIMESTAMP(4),
>> +};
>> +
>> +static const int sca3300_accel_scale[] = {2700, 1350, 5400, 5400};
>> +
>> +static const unsigned long sca3300_scan_masks[] = {
>> +	BIT(SCA3300_ACC_X) | BIT(SCA3300_ACC_Y) | BIT(SCA3300_ACC_Z) |
>> +	BIT(SCA3300_TEMP),
>> +	0};
>> +
>> +/**
>> + * SCA3300 device data
> run scripts/kernel-doc over the file and fix the warnings + errors.
>
>> + *
>> + * @spi SPI device structure
>> + * @opmode Device operation mode
>> + * @lock Data buffer lock
>> + * @txbuf Transmit buffer
>> + * @rxbuf Receive buffer
>> + * @scan Triggered buffer. Four channel 16-bit data + 64-bit timestamp
>> + */
>> +struct sca3300_data {
>> +	struct spi_device *spi;
>> +	u32 opmode;
>> +	struct mutex lock;
>> +	u8 txbuf[4];
>> +	u8 rxbuf[4];
>> +	struct {
>> +		s16 channels[4];
>> +		s64 ts __aligned(sizeof(s64));
>> +	} scan;
>> +};
>> +
>> +DECLARE_CRC8_TABLE(sca3300_crc_table);
>> +
>> +static int sca3300_transfer(struct sca3300_data *sca_data, int *val)
>> +{
>> +	struct spi_delay delay = {.value = 10, .unit = SPI_DELAY_UNIT_USECS};
>> +	int32_t ret;
>> +	int rs;
>> +	u8 crc;
>> +	struct spi_transfer xfers[2] = {
>> +		{
>> +			.tx_buf = sca_data->txbuf,
>> +			.rx_buf = NULL,
>> +			.len = ARRAY_SIZE(sca_data->txbuf),
>> +			.delay = delay,
>> +			.cs_change = 1,
>> +		},
>> +		{
>> +			.tx_buf = NULL,
>> +			.rx_buf = sca_data->rxbuf,
>> +			.len = ARRAY_SIZE(sca_data->rxbuf),
>> +			.delay = delay,
>> +			.cs_change = 0,
>> +		}
>> +	};
>> +
>> +	/* inverted crc value as described in device data sheet */
>> +	crc = ~crc8(sca3300_crc_table, &sca_data->txbuf[0], 3, CRC8_INIT_VALUE);
>> +	sca_data->txbuf[3] = crc;
>> +
>> +	ret = spi_sync_transfer(sca_data->spi, xfers, 2);
> Use ARRAY_SIZE(xfers) instead of 2.
>
>> +	if (ret < 0) {
>> +		dev_err(&sca_data->spi->dev,
>> +			"transfer error, error: %d\n", ret);
>> +		return -EIO;
>> +	}
>> +
>> +	crc = ~crc8(sca3300_crc_table, &sca_data->rxbuf[0], 3, CRC8_INIT_VALUE);
>> +	if (sca_data->rxbuf[3] != crc) {
>> +		dev_err(&sca_data->spi->dev, "CRC checksum mismatch");
>> +		return -EIO;
>> +	}
>> +
>> +	/* get return status */
>> +	rs = sca_data->rxbuf[0] & 0x03;
>> +	if (rs == SCA3300_RS_ERROR)
>> +		return rs;
>> +
>> +	*val = (s16)(sca_data->rxbuf[2] | (sca_data->rxbuf[1] << 8));
> Preference for an unaligned endian conversion here and explicit sign extend e.g
> something like
>
> sign_extend32(get_unaligned_le16(&sca_data->rxbuf[1]), 15)
>
>> +
>> +	return 0;
>> +}
>> +
>> +static int sca3300_read_reg(struct sca3300_data *sca_data, u8 reg, int *val)
>> +{
>> +	int ret;
>> +
>> +	mutex_lock(&sca_data->lock);
>> +	sca_data->txbuf[0] = SCA3300_X_READ | (reg << 2);
>> +	ret = sca3300_transfer(sca_data, val);
>> +	if (ret > 0) {
>> +		sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS << 2);
>> +		ret = sca3300_transfer(sca_data, val);
>> +		/* status 0 = startup, 0x2 = mode change */
>> +		if (ret > 0 && *val != 0 && *val != 0x2) {
>> +			dev_err_ratelimited(&sca_data->spi->dev,
>> +					    "device status: %x\n",
>> +					    (u16)*val);
>> +			mutex_unlock(&sca_data->lock);
>> +			return -EIO;
>> +		}
>> +		if (ret > 0)
>> +			ret = 0;
>> +	}
>> +	mutex_unlock(&sca_data->lock);
>> +
>> +	return ret;
>> +}
>> +
>> +static int sca3300_write_reg(struct sca3300_data *sca_data, u8 reg, int val)
>> +{
>> +	int reg_val = 0;
>> +	int ret;
>> +
>> +	mutex_lock(&sca_data->lock);
>> +	sca_data->txbuf[0] = SCA3300_X_WRITE | (reg << 2);
>> +	sca_data->txbuf[1] = val >> 8;
>> +	sca_data->txbuf[2] = val & 0xFF;
> Prefer (Slightly) an unaligned put.
>
>> +	ret = sca3300_transfer(sca_data, &reg_val);
>> +	if (ret > 0) {
> Factor this error handling out to another function, plus trigger it from an
> appropriate standard error code rather than a positive return value.
>
>> +		sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS << 2);
>> +		ret = sca3300_transfer(sca_data, &reg_val);
>> +		/* status 0 = startup, 0x2 = mode change */
>> +		if (ret > 0 && reg_val != 0 && reg_val != 0x2) {
>> +			dev_err_ratelimited(&sca_data->spi->dev,
>> +					    "device status: %x\n",
>> +					    (u16)reg_val);
>> +			mutex_unlock(&sca_data->lock);
>> +			return -EIO;
>> +		}
>> +		if (ret > 0)
>> +			ret = 0;
> Whenever you get this sort of juggling it rather implies your design is
> less than ideal.
>
>> +	}
>> +	mutex_unlock(&sca_data->lock);
>> +
>> +	return ret;
>> +}
>> +
>> +static int sca3300_write_raw(struct iio_dev *indio_dev,
>> +			     struct iio_chan_spec const *chan,
>> +			     int val, int val2, long mask)
>> +{
>> +	struct sca3300_data *data = iio_priv(indio_dev);
>> +
>> +	switch (mask) {
>> +	case IIO_CHAN_INFO_SCALE:
>> +		if (val < 0 || val > 3)
>> +			return -EINVAL;
> No.  Scale values need to be the value of 1LSB not a random integer
> that requires you to look up the datasheet.
>
>> +		return sca3300_write_reg(data, SCA3300_REG_MODE, val);
>> +	default:
>> +		return -EINVAL;
>> +	}
>> +}
>> +
>> +static int sca3300_read_raw(struct iio_dev *indio_dev,
>> +			    struct iio_chan_spec const *chan,
>> +			    int *val, int *val2, long mask)
>> +{
>> +	struct sca3300_data *data = iio_priv(indio_dev);
>> +	int ret;
>> +	int reg_val;
>> +
>> +	switch (mask) {
>> +	case IIO_CHAN_INFO_RAW:
>> +		ret = sca3300_read_reg(data, chan->address, val);
>> +		if (ret < 0)
>> +			return ret;
>> +		return IIO_VAL_INT;
>> +	case IIO_CHAN_INFO_SCALE:
>> +		ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
>> +		if (ret < 0)
>> +			return ret;
>> +		*val = sca3300_accel_scale[reg_val];
> Scale is very very rarely an integer value which makes me wonder what
> you are returning here...  It should be the multiplier needed to
> take the raw value to a reading in m/sec^2  Given you use it as fractional
> below, I'm guessing this is 1/scale_value?
>
>> +		return IIO_VAL_INT;
>> +	case IIO_CHAN_INFO_PROCESSED:
> Don't provide processed here.   Userspace needs to do the conversion itself
> (and given it has floating point easily available will probably do a better
>   job than we can).   I'm assuming the mode can't autonomously change?
>
>> +		ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
>> +		if (ret < 0)
>> +			return ret;
>> +		*val2 = sca3300_accel_scale[reg_val];
>> +		ret = sca3300_read_reg(data, chan->address, val);
>> +		if (ret < 0)
>> +			return ret;
>> +		return IIO_VAL_FRACTIONAL;
>> +	default:
>> +		return -EINVAL;
>> +	}
>> +}
>> +
>> +static irqreturn_t sca3300_trigger_handler(int irq, void *p)
>> +{
>> +	struct iio_poll_func *pf = p;
>> +	struct iio_dev *indio_dev = pf->indio_dev;
>> +	struct sca3300_data *data = iio_priv(indio_dev);
>> +	s64 time_ns = iio_get_time_ns(indio_dev);
> Is the timestamp at the start more accurate that that at the end
> of reading the channels?  If not, just put this inline in the
> iio_push_to_buffers_with_timestamp() call.
>
>> +	int bit, ret, val, i = 0;
>> +
>> +	for_each_set_bit(bit, indio_dev->active_scan_mask,
>> +			 indio_dev->masklength) {
>> +		ret = sca3300_read_reg(data, sca3300_channels[bit].address,
>> +				       &val);
>> +		if (ret < 0)
> Given we can't return an error code from interrupt handlers, it is usually
> a good idea to print something to the log. Otherwise we get missing data
> with no idea of why..
>
>> +			goto out;
>> +		if (ARRAY_SIZE(data->scan.channels) > i)
> How could this not be true?  If it's always true, don't bother
> checking i.
>
>> +			((s16 *)data->scan.channels)[i++] = val;
>> +	}
>> +
>> +	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan, time_ns);
>> +out:
>> +	iio_trigger_notify_done(indio_dev->trig);
>> +
>> +	return IRQ_HANDLED;
>> +}
>> +
>> +static int sca3300_init(struct sca3300_data *sca_data,
>> +			struct iio_dev *indio_dev)
>> +{
>> +	int ret;
>> +	int value = 0;
>> +
>> +	if (sca_data->opmode < 1 || sca_data->opmode > 4)
>> +		return -EINVAL;
> Whilst this is going anyway, better to check that next to where
> it is read so we drop out immediately rather than buried in this
> function.
>
>> +
>> +	ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
>> +				SCA3300_VALUE_SW_RESET);
>> +	if (ret != 0)
>> +		return ret;
> For sleeps, good to reference which section in data sheet gives the
> timing.
>
>> +	usleep_range(2e3, 10e3);
>> +
>> +	ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
>> +				sca_data->opmode - 1);
>> +	if (ret != 0)
> if (ret) see below.
>
>> +		return ret;
>> +	msleep(100);
> Again, document where the time comes from. Saves reviewers time if
> they want to check it. (I'm too lazy :)
>
>> +	ret = sca3300_read_reg(sca_data, SCA3300_REG_WHOAMI, &value);
>> +	if (ret != 0)
> if (ret) is more idiomatic in kernel code.
>
>> +		return ret;
>> +
>> +	if (value != SCA3300_DEVICE_ID) {
>> +		dev_err(&sca_data->spi->dev, "device id not expected value\n");
> Perhaps useful to print what was expected and what was seen.  Maybe it's
> a new variant and this might help the user to identify that and add it to the
> driver.
>
>> +		return -EIO;
> -EINVAL probably more appropriate return.
>
>> +	}
>> +	return 0;
>> +}
>> +
>> +static int sca3300_debugfs_reg_access(struct iio_dev *indio_dev,
>> +				      unsigned int reg, unsigned int writeval,
>> +				      unsigned int *readval)
>> +{
>> +	struct sca3300_data *data = iio_priv(indio_dev);
>> +	int value;
>> +	int ret;
>> +
>> +	if (reg > 0x1f)
> Use a define that names that register.
>
>> +		return -EINVAL;
>> +
>> +	if (!readval)
>> +		return sca3300_write_reg(data, reg, writeval);
>> +
>> +	ret = sca3300_read_reg(data, reg, &value);
>> +	if (ret < 0)
>> +		return ret;
>> +
>> +	*readval = (unsigned int)value;
>> +
>> +	return 0;
>> +}
>> +
>> +static const struct iio_info sca3300_info = {
>> +	.read_raw = sca3300_read_raw,
>> +	.write_raw = sca3300_write_raw,
>> +	.debugfs_reg_access = &sca3300_debugfs_reg_access,
>> +};
>> +
>> +static int sca3300_probe(struct spi_device *spi)
>> +{
>> +	struct sca3300_data *sca_data;
>> +	struct iio_dev *indio_dev;
>> +	int ret;
>> +
>> +	indio_dev = devm_iio_device_alloc(&spi->dev, sizeof(*sca_data));
>> +	if (!indio_dev) {
>> +		dev_err(&spi->dev,
>> +			"failed to allocate memory for iio device\n");
>> +		return -ENOMEM;
>> +	}
>> +
>> +	sca_data = iio_priv(indio_dev);
>> +	mutex_init(&sca_data->lock);
>> +	sca_data->spi = spi;
>> +	spi_set_drvdata(spi, indio_dev);
> Won't be used once using devm as below and remove() is dropped, so drop this.
>
>> +
>> +	crc8_populate_msb(sca3300_crc_table, SCA3300_CRC8_POLYNOMIAL);
>> +
>> +	indio_dev->dev.parent = &spi->dev;
>> +	indio_dev->info = &sca3300_info;
>> +	indio_dev->name = SCA3300_ALIAS;
>> +	indio_dev->modes = INDIO_DIRECT_MODE | INDIO_BUFFER_TRIGGERED;
> BUFFER_TRIGGERED bit now set by iio_triggered_buffer_setup() so shouldn't
> be ehre as well.
>
>> +	indio_dev->channels = sca3300_channels;
>> +	indio_dev->num_channels = ARRAY_SIZE(sca3300_channels);
>> +	indio_dev->available_scan_masks = sca3300_scan_masks;
>> +
>> +	if (spi->dev.of_node) {
>> +		ret = of_property_read_u32(spi->dev.of_node, "murata,opmode",
> Please use generic firmware access calls rather than the of specific ones.
> That lets us get ACPI support for free :)
>
> As per the binding review however, this one needs to go in favour of userspace
> control of scale + filter frequencies / sampling freq.
>
>> +					   &sca_data->opmode);
>> +		if (ret < 0)
>> +			return ret;
>> +	}
>> +
>> +	ret = sca3300_init(sca_data, indio_dev);
>> +	if (ret < 0) {
>> +		dev_err(&spi->dev, "failed to init device, error: %d\n", ret);
>> +		return ret;
>> +	}
>> +
>> +	ret = iio_triggered_buffer_setup(indio_dev, iio_pollfunc_store_time,
>> +					 sca3300_trigger_handler, NULL);
> devm_iio_triggered_buffer_setup() and you can drop the manual cleanup in remove()
> and hence drop remove() entirely as nothing else left.
>
>> +	if (ret < 0) {
>> +		dev_err(&spi->dev,
>> +			"iio triggered buffer setup failed, error: %d\n", ret);
>> +		return ret;
>> +	}
>> +
>> +	ret = devm_iio_device_register(&spi->dev, indio_dev);
>> +	if (ret < 0) {
>> +		dev_err(&spi->dev, "iio device register failed, error: %d\n",
>> +			ret);
>> +		iio_triggered_buffer_cleanup(indio_dev);
> With devm above, no need to manually clean this up.
>
>> +		return ret;
>> +	}
>> +
>> +	return 0;
>> +}
>> +
>> +static int sca3300_remove(struct spi_device *spi)
>> +{
>> +	struct iio_dev *indio_dev = spi_get_drvdata(spi);
>> +
>> +	iio_triggered_buffer_cleanup(indio_dev);
>> +	return 0;
>> +}
>> +
>> +static const struct of_device_id sca3300_dt_ids[] = {
>> +	{ .compatible = "murata,sca3300"},
>> +	{},
>> +};
>> +MODULE_DEVICE_TABLE(of, sca3300_dt_ids);
>> +
>> +static struct spi_driver sca3300_driver = {
>> +	.driver = {
>> +		.name		= SCA3300_ALIAS,
>> +		.owner		= THIS_MODULE,
>> +		.of_match_table = of_match_ptr(sca3300_dt_ids),
>> +	},
>> +
>> +	.probe	= sca3300_probe,
>> +	.remove	= sca3300_remove,
>> +};
>> +
>> +module_spi_driver(sca3300_driver);
>> +
>> +MODULE_AUTHOR("Tomas Melin <[e-mail]>");
>> +MODULE_DESCRIPTION("Murata SCA3300 SPI Accelerometer");
>> +MODULE_LICENSE("GPL v2");

Tomas Melin April 16, 2021, 12:17 p.m. UTC | #3

On 4/15/21 11:41 AM, Tomas Melin wrote:

>
> While working on updates I did notice something new which I cannot
>
> reproduce on older (5.10.17 kernel) version. If compiling this as a 
> module, getting error while
>
> unloading module:
>
> [   40.200084] Unable to handle kernel NULL pointer dereference at 
> virtual address 00000104
> ...
>
> [   40.510054] Backtrace:
> [   40.512502] [<c06c5e0c>] (iio_device_ioctl_handler_unregister) from 
> [<c06ca8c4>] (iio_buffers_free_sysfs_and_mask+0x2c/0x6c)
> [   40.523735] [<c06ca898>] (iio_buffers_free_sysfs_and_mask) from 
> [<c06c4a18>] (iio_device_unregister+0xa8/0xac)
> [   40.533746]  r5:c1811228 r4:c1811000
> [   40.537318] [<c06c4970>] (iio_device_unregister) from [<c06c4a38>] 
> (devm_iio_device_unreg+0x1c/0x20)
> [   40.546461]  r5:c2415000 r4:c25bab80
> [   40.550025] [<c06c4a1c>] (devm_iio_device_unreg) from [<c0556820>] 
> (release_nodes+0x1c0/0x1f0)
> [   40.558654] [<c0556660>] (release_nodes) from [<c0556bb0>] 
> (devres_release_all+0x40/0x60)
> [   40.566847]  r10:00000081 r9:c2350000 r8:c0100264 r7:00000081 
> r6:bf00c010 r5:c19be000
> [   40.574669]  r4:c1a91c00
> [   40.577194] [<c0556b70>] (devres_release_all) from [<c055216c>] 
> (device_release_driver_internal+0x120/0x1cc)
> [   40.587031]  r5:c19be000 r4:c1a91c00
> [   40.590596] [<c055204c>] (device_release_driver_internal) from 
> [<c05522b4>] (driver_detach+0x54/0x90)
> [   40.599828]  r7:00000081 r6:00000000 r5:bf00c010 r4:c1a91c00
> [   40.605482] [<c0552260>] (driver_detach) from [<c0550ba4>] 
> (bus_remove_driver+0x5c/0xb0)
> [   40.613583]  r5:00000800 r4:bf00c010
> [   40.617148] [<c0550b48>] (bus_remove_driver) from [<c0552c38>] 
> (driver_unregister+0x38/0x5c)
> [   40.625596]  r5:00000800 r4:bf00c010
> [   40.629161] [<c0552c00>] (driver_unregister) from [<bf00a760>] 
> (sca3300_driver_exit+0x14/0x8b4 [sca3300])
> [   40.638747]  r5:00000800 r4:bf00c080
> [   40.642311] [<bf00a74c>] (sca3300_driver_exit [sca3300]) from 
> [<c01c06f0>] (sys_delete_module+0x16c/0x238)
> [   40.651990] [<c01c0584>] (sys_delete_module) from [<c0100244>] 
> (__sys_trace_return+0x0/0x1c)
> [   40.660435] Exception stack(0xc2351fa8 to 0xc2351ff0)
> [   40.665484] 1fa0:                   0050e5a8 00000000 0050e5e4 
> 00000800 081d4b00 bec18af4
> [   40.673661] 1fc0: 0050e5a8 00000000 bec18b50 00000081 bec18e51 
> 0050e190 00000001 bec18d3c
> [   40.681834] 1fe0: 0050cf70 bec18afc 004f1ec8 b6ecb27c
> [   40.686887]  r6:bec18b50 r5:00000000 r4:0050e5a8
> [   40.691507] Code: e8bd4000 e1c020d0 e3a0cc01 e3001122 (e5823004)
> [   40.707675] ---[ end trace 189882b050077333 ]---
>
> This happens when building against linux-next 
> 5.12.0-rc6-next-20210409. I'm failing to see what is wrong. Any ideas?
>
> Thanks,
>
> Tomas


Tested further that for this driver, loading and unloading as module 
works fine until commit:

commit f73f7f4da581875f9b1f2fb8ebd1ab15ed634488
Author: Alexandru Ardelean <[e-mail]>
Date:   Mon Feb 15 12:40:39 2021 +0200

     iio: buffer: add ioctl() to support opening extra buffers for IIO 
device


Any thoughts what causes this issue?

Thanks,

Tomas



>
>
>>
>>> ---
>>>   drivers/iio/accel/Kconfig   |  13 ++
>>>   drivers/iio/accel/Makefile  |   1 +
>>>   drivers/iio/accel/sca3300.c | 434 
>>> ++++++++++++++++++++++++++++++++++++
>>>   3 files changed, 448 insertions(+)
>>>   create mode 100644 drivers/iio/accel/sca3300.c
>>>
>>> diff --git a/drivers/iio/accel/Kconfig b/drivers/iio/accel/Kconfig
>>> index cceda3cecbcf..0dbf7b648e8a 100644
>>> --- a/drivers/iio/accel/Kconfig
>>> +++ b/drivers/iio/accel/Kconfig
>>> @@ -450,6 +450,19 @@ config SCA3000
>>>         To compile this driver as a module, say M here: the module 
>>> will be
>>>         called sca3000.
>>>   +config SCA3300
>>> +    tristate "Murata SCA3300 3-Axis Accelerometer Driver"
>>> +    depends on SPI
>>> +    select CRC8
>>> +    select IIO_BUFFER
>>> +    select IIO_TRIGGERED_BUFFER
>>> +    help
>>> +      Say yes here to build support for Murata SCA3300 3-Axis
>>> +      accelerometer.
>>> +
>>> +      To compile this driver as a module, choose M here: the module 
>>> will be
>>> +      called sca3300.
>>> +
>>>   config STK8312
>>>       tristate "Sensortek STK8312 3-Axis Accelerometer Driver"
>>>       depends on I2C
>>> diff --git a/drivers/iio/accel/Makefile b/drivers/iio/accel/Makefile
>>> index 32cd1342a31a..4b56527a2b97 100644
>>> --- a/drivers/iio/accel/Makefile
>>> +++ b/drivers/iio/accel/Makefile
>>> @@ -50,6 +50,7 @@ obj-$(CONFIG_MXC4005)        += mxc4005.o
>>>   obj-$(CONFIG_MXC6255)        += mxc6255.o
>>>     obj-$(CONFIG_SCA3000)        += sca3000.o
>>> +obj-$(CONFIG_SCA3300)        += sca3300.o
>>>     obj-$(CONFIG_STK8312)        += stk8312.o
>>>   obj-$(CONFIG_STK8BA50)        += stk8ba50.o
>>> diff --git a/drivers/iio/accel/sca3300.c b/drivers/iio/accel/sca3300.c
>>> new file mode 100644
>>> index 000000000000..112fb88ecd3a
>>> --- /dev/null
>>> +++ b/drivers/iio/accel/sca3300.c
>>> @@ -0,0 +1,434 @@
>>> +// SPDX-License-Identifier: GPL-2.0-only
>>> +/*
>>> + * Copyright (c) Vaisala Oyj. All rights reserved.
>> Give a year for the copyright notice if you can.
>>
>>> + */
>>> +#include <linux/crc8.h>
>>> +#include <linux/delay.h>
>>> +#include <linux/iio/buffer.h>
>>> +#include <linux/iio/iio.h>
>>> +#include <linux/iio/sysfs.h>
>>> +#include <linux/iio/trigger_consumer.h>
>>> +#include <linux/iio/triggered_buffer.h>
>>> +#include <linux/kernel.h>
>>> +#include <linux/module.h>
>>> +#include <linux/spi/spi.h>
>>> +
>>> +#define SCA3300_ALIAS "sca3300"
>>> +
>>> +#define SCA3300_REG_STATUS 0x6
>>> +#define SCA3300_REG_MODE 0xd
>>> +#define SCA3300_REG_WHOAMI 0x10
>>> +#define SCA3300_VALUE_SW_RESET 0x20
>>> +#define SCA3300_CRC8_POLYNOMIAL 0x1d
>>> +#define SCA3300_X_READ 0
>> I wouldn't bother defining this.
>>
>>> +#define SCA3300_X_WRITE BIT(7)
>> Even this one is something I'd just put inline with a comment.
>>
>>> +#define SCA3300_DEVICE_ID 0x51
>>> +#define SCA3300_RS_ERROR 0x3
>>> +
>>> +enum sca3300_scan_indexes {
>>> +    SCA3300_ACC_X = 0,
>>> +    SCA3300_ACC_Y,
>>> +    SCA3300_ACC_Z,
>>> +    SCA3300_TEMP,
>>> +    SCA3300_TIMESTAMP,
>>> +};
>>> +
>>> +#define SCA3300_ACCEL_CHANNEL(index, reg, axis) { \
>>> +        .type = IIO_ACCEL,                    \
>>> +        .address = reg,                        \
>>> +        .modified = 1,                        \
>>> +        .channel2 = IIO_MOD_##axis,                \
>>> +        .info_mask_separate = BIT(IIO_CHAN_INFO_RAW) | \
>>> +                      BIT(IIO_CHAN_INFO_PROCESSED),    \
>> As mentioned below, don't provide PROCESSED. Userspace is better at 
>> handling the
>> conversion so leave it to them.
>>
>>> +        .info_mask_shared_by_type = BIT(IIO_CHAN_INFO_SCALE),    \
>>> +        .scan_index = index,                    \
>>> +        .scan_type = {                        \
>>> +            .sign = 's',                    \
>>> +            .realbits = 16,                    \
>>> +            .storagebits = 16,                \
>>> +            .shift = 0,                    \
>>> +            .endianness = IIO_CPU,                \
>>> +        },                            \
>>> +    }
>>> +
>>> +static const struct iio_chan_spec sca3300_channels[] = {
>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_X, 0x1, X),
>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Y, 0x2, Y),
>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Z, 0x3, Z),
>>> +    {
>>> +        .type = IIO_TEMP,
>>> +        .address = 0x5,
>>> +        .scan_index = SCA3300_TEMP,
>>> +        .info_mask_separate = BIT(IIO_CHAN_INFO_RAW),
>>> +        .scan_type = {
>>> +            .sign = 's',
>>> +            .realbits = 16,
>>> +            .storagebits = 16,
>>> +            .shift = 0,
>>> +            .endianness = IIO_CPU,
>>> +        },
>>> +    },
>>> +    IIO_CHAN_SOFT_TIMESTAMP(4),
>>> +};
>>> +
>>> +static const int sca3300_accel_scale[] = {2700, 1350, 5400, 5400};
>>> +
>>> +static const unsigned long sca3300_scan_masks[] = {
>>> +    BIT(SCA3300_ACC_X) | BIT(SCA3300_ACC_Y) | BIT(SCA3300_ACC_Z) |
>>> +    BIT(SCA3300_TEMP),
>>> +    0};
>>> +
>>> +/**
>>> + * SCA3300 device data
>> run scripts/kernel-doc over the file and fix the warnings + errors.
>>
>>> + *
>>> + * @spi SPI device structure
>>> + * @opmode Device operation mode
>>> + * @lock Data buffer lock
>>> + * @txbuf Transmit buffer
>>> + * @rxbuf Receive buffer
>>> + * @scan Triggered buffer. Four channel 16-bit data + 64-bit timestamp
>>> + */
>>> +struct sca3300_data {
>>> +    struct spi_device *spi;
>>> +    u32 opmode;
>>> +    struct mutex lock;
>>> +    u8 txbuf[4];
>>> +    u8 rxbuf[4];
>>> +    struct {
>>> +        s16 channels[4];
>>> +        s64 ts __aligned(sizeof(s64));
>>> +    } scan;
>>> +};
>>> +
>>> +DECLARE_CRC8_TABLE(sca3300_crc_table);
>>> +
>>> +static int sca3300_transfer(struct sca3300_data *sca_data, int *val)
>>> +{
>>> +    struct spi_delay delay = {.value = 10, .unit = 
>>> SPI_DELAY_UNIT_USECS};
>>> +    int32_t ret;
>>> +    int rs;
>>> +    u8 crc;
>>> +    struct spi_transfer xfers[2] = {
>>> +        {
>>> +            .tx_buf = sca_data->txbuf,
>>> +            .rx_buf = NULL,
>>> +            .len = ARRAY_SIZE(sca_data->txbuf),
>>> +            .delay = delay,
>>> +            .cs_change = 1,
>>> +        },
>>> +        {
>>> +            .tx_buf = NULL,
>>> +            .rx_buf = sca_data->rxbuf,
>>> +            .len = ARRAY_SIZE(sca_data->rxbuf),
>>> +            .delay = delay,
>>> +            .cs_change = 0,
>>> +        }
>>> +    };
>>> +
>>> +    /* inverted crc value as described in device data sheet */
>>> +    crc = ~crc8(sca3300_crc_table, &sca_data->txbuf[0], 3, 
>>> CRC8_INIT_VALUE);
>>> +    sca_data->txbuf[3] = crc;
>>> +
>>> +    ret = spi_sync_transfer(sca_data->spi, xfers, 2);
>> Use ARRAY_SIZE(xfers) instead of 2.
>>
>>> +    if (ret < 0) {
>>> +        dev_err(&sca_data->spi->dev,
>>> +            "transfer error, error: %d\n", ret);
>>> +        return -EIO;
>>> +    }
>>> +
>>> +    crc = ~crc8(sca3300_crc_table, &sca_data->rxbuf[0], 3, 
>>> CRC8_INIT_VALUE);
>>> +    if (sca_data->rxbuf[3] != crc) {
>>> +        dev_err(&sca_data->spi->dev, "CRC checksum mismatch");
>>> +        return -EIO;
>>> +    }
>>> +
>>> +    /* get return status */
>>> +    rs = sca_data->rxbuf[0] & 0x03;
>>> +    if (rs == SCA3300_RS_ERROR)
>>> +        return rs;
>>> +
>>> +    *val = (s16)(sca_data->rxbuf[2] | (sca_data->rxbuf[1] << 8));
>> Preference for an unaligned endian conversion here and explicit sign 
>> extend e.g
>> something like
>>
>> sign_extend32(get_unaligned_le16(&sca_data->rxbuf[1]), 15)
>>
>>> +
>>> +    return 0;
>>> +}
>>> +
>>> +static int sca3300_read_reg(struct sca3300_data *sca_data, u8 reg, 
>>> int *val)
>>> +{
>>> +    int ret;
>>> +
>>> +    mutex_lock(&sca_data->lock);
>>> +    sca_data->txbuf[0] = SCA3300_X_READ | (reg << 2);
>>> +    ret = sca3300_transfer(sca_data, val);
>>> +    if (ret > 0) {
>>> +        sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS 
>>> << 2);
>>> +        ret = sca3300_transfer(sca_data, val);
>>> +        /* status 0 = startup, 0x2 = mode change */
>>> +        if (ret > 0 && *val != 0 && *val != 0x2) {
>>> + dev_err_ratelimited(&sca_data->spi->dev,
>>> +                        "device status: %x\n",
>>> +                        (u16)*val);
>>> +            mutex_unlock(&sca_data->lock);
>>> +            return -EIO;
>>> +        }
>>> +        if (ret > 0)
>>> +            ret = 0;
>>> +    }
>>> +    mutex_unlock(&sca_data->lock);
>>> +
>>> +    return ret;
>>> +}
>>> +
>>> +static int sca3300_write_reg(struct sca3300_data *sca_data, u8 reg, 
>>> int val)
>>> +{
>>> +    int reg_val = 0;
>>> +    int ret;
>>> +
>>> +    mutex_lock(&sca_data->lock);
>>> +    sca_data->txbuf[0] = SCA3300_X_WRITE | (reg << 2);
>>> +    sca_data->txbuf[1] = val >> 8;
>>> +    sca_data->txbuf[2] = val & 0xFF;
>> Prefer (Slightly) an unaligned put.
>>
>>> +    ret = sca3300_transfer(sca_data, &reg_val);
>>> +    if (ret > 0) {
>> Factor this error handling out to another function, plus trigger it 
>> from an
>> appropriate standard error code rather than a positive return value.
>>
>>> +        sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS 
>>> << 2);
>>> +        ret = sca3300_transfer(sca_data, &reg_val);
>>> +        /* status 0 = startup, 0x2 = mode change */
>>> +        if (ret > 0 && reg_val != 0 && reg_val != 0x2) {
>>> + dev_err_ratelimited(&sca_data->spi->dev,
>>> +                        "device status: %x\n",
>>> +                        (u16)reg_val);
>>> +            mutex_unlock(&sca_data->lock);
>>> +            return -EIO;
>>> +        }
>>> +        if (ret > 0)
>>> +            ret = 0;
>> Whenever you get this sort of juggling it rather implies your design is
>> less than ideal.
>>
>>> +    }
>>> +    mutex_unlock(&sca_data->lock);
>>> +
>>> +    return ret;
>>> +}
>>> +
>>> +static int sca3300_write_raw(struct iio_dev *indio_dev,
>>> +                 struct iio_chan_spec const *chan,
>>> +                 int val, int val2, long mask)
>>> +{
>>> +    struct sca3300_data *data = iio_priv(indio_dev);
>>> +
>>> +    switch (mask) {
>>> +    case IIO_CHAN_INFO_SCALE:
>>> +        if (val < 0 || val > 3)
>>> +            return -EINVAL;
>> No.  Scale values need to be the value of 1LSB not a random integer
>> that requires you to look up the datasheet.
>>
>>> +        return sca3300_write_reg(data, SCA3300_REG_MODE, val);
>>> +    default:
>>> +        return -EINVAL;
>>> +    }
>>> +}
>>> +
>>> +static int sca3300_read_raw(struct iio_dev *indio_dev,
>>> +                struct iio_chan_spec const *chan,
>>> +                int *val, int *val2, long mask)
>>> +{
>>> +    struct sca3300_data *data = iio_priv(indio_dev);
>>> +    int ret;
>>> +    int reg_val;
>>> +
>>> +    switch (mask) {
>>> +    case IIO_CHAN_INFO_RAW:
>>> +        ret = sca3300_read_reg(data, chan->address, val);
>>> +        if (ret < 0)
>>> +            return ret;
>>> +        return IIO_VAL_INT;
>>> +    case IIO_CHAN_INFO_SCALE:
>>> +        ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
>>> +        if (ret < 0)
>>> +            return ret;
>>> +        *val = sca3300_accel_scale[reg_val];
>> Scale is very very rarely an integer value which makes me wonder what
>> you are returning here...  It should be the multiplier needed to
>> take the raw value to a reading in m/sec^2  Given you use it as 
>> fractional
>> below, I'm guessing this is 1/scale_value?
>>
>>> +        return IIO_VAL_INT;
>>> +    case IIO_CHAN_INFO_PROCESSED:
>> Don't provide processed here.   Userspace needs to do the conversion 
>> itself
>> (and given it has floating point easily available will probably do a 
>> better
>>   job than we can).   I'm assuming the mode can't autonomously change?
>>
>>> +        ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
>>> +        if (ret < 0)
>>> +            return ret;
>>> +        *val2 = sca3300_accel_scale[reg_val];
>>> +        ret = sca3300_read_reg(data, chan->address, val);
>>> +        if (ret < 0)
>>> +            return ret;
>>> +        return IIO_VAL_FRACTIONAL;
>>> +    default:
>>> +        return -EINVAL;
>>> +    }
>>> +}
>>> +
>>> +static irqreturn_t sca3300_trigger_handler(int irq, void *p)
>>> +{
>>> +    struct iio_poll_func *pf = p;
>>> +    struct iio_dev *indio_dev = pf->indio_dev;
>>> +    struct sca3300_data *data = iio_priv(indio_dev);
>>> +    s64 time_ns = iio_get_time_ns(indio_dev);
>> Is the timestamp at the start more accurate that that at the end
>> of reading the channels?  If not, just put this inline in the
>> iio_push_to_buffers_with_timestamp() call.
>>
>>> +    int bit, ret, val, i = 0;
>>> +
>>> +    for_each_set_bit(bit, indio_dev->active_scan_mask,
>>> +             indio_dev->masklength) {
>>> +        ret = sca3300_read_reg(data, sca3300_channels[bit].address,
>>> +                       &val);
>>> +        if (ret < 0)
>> Given we can't return an error code from interrupt handlers, it is 
>> usually
>> a good idea to print something to the log. Otherwise we get missing data
>> with no idea of why..
>>
>>> +            goto out;
>>> +        if (ARRAY_SIZE(data->scan.channels) > i)
>> How could this not be true?  If it's always true, don't bother
>> checking i.
>>
>>> +            ((s16 *)data->scan.channels)[i++] = val;
>>> +    }
>>> +
>>> +    iio_push_to_buffers_with_timestamp(indio_dev, &data->scan, 
>>> time_ns);
>>> +out:
>>> +    iio_trigger_notify_done(indio_dev->trig);
>>> +
>>> +    return IRQ_HANDLED;
>>> +}
>>> +
>>> +static int sca3300_init(struct sca3300_data *sca_data,
>>> +            struct iio_dev *indio_dev)
>>> +{
>>> +    int ret;
>>> +    int value = 0;
>>> +
>>> +    if (sca_data->opmode < 1 || sca_data->opmode > 4)
>>> +        return -EINVAL;
>> Whilst this is going anyway, better to check that next to where
>> it is read so we drop out immediately rather than buried in this
>> function.
>>
>>> +
>>> +    ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
>>> +                SCA3300_VALUE_SW_RESET);
>>> +    if (ret != 0)
>>> +        return ret;
>> For sleeps, good to reference which section in data sheet gives the
>> timing.
>>
>>> +    usleep_range(2e3, 10e3);
>>> +
>>> +    ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
>>> +                sca_data->opmode - 1);
>>> +    if (ret != 0)
>> if (ret) see below.
>>
>>> +        return ret;
>>> +    msleep(100);
>> Again, document where the time comes from. Saves reviewers time if
>> they want to check it. (I'm too lazy :)
>>
>>> +    ret = sca3300_read_reg(sca_data, SCA3300_REG_WHOAMI, &value);
>>> +    if (ret != 0)
>> if (ret) is more idiomatic in kernel code.
>>
>>> +        return ret;
>>> +
>>> +    if (value != SCA3300_DEVICE_ID) {
>>> +        dev_err(&sca_data->spi->dev, "device id not expected 
>>> value\n");
>> Perhaps useful to print what was expected and what was seen. Maybe it's
>> a new variant and this might help the user to identify that and add 
>> it to the
>> driver.
>>
>>> +        return -EIO;
>> -EINVAL probably more appropriate return.
>>
>>> +    }
>>> +    return 0;
>>> +}
>>> +
>>> +static int sca3300_debugfs_reg_access(struct iio_dev *indio_dev,
>>> +                      unsigned int reg, unsigned int writeval,
>>> +                      unsigned int *readval)
>>> +{
>>> +    struct sca3300_data *data = iio_priv(indio_dev);
>>> +    int value;
>>> +    int ret;
>>> +
>>> +    if (reg > 0x1f)
>> Use a define that names that register.
>>
>>> +        return -EINVAL;
>>> +
>>> +    if (!readval)
>>> +        return sca3300_write_reg(data, reg, writeval);
>>> +
>>> +    ret = sca3300_read_reg(data, reg, &value);
>>> +    if (ret < 0)
>>> +        return ret;
>>> +
>>> +    *readval = (unsigned int)value;
>>> +
>>> +    return 0;
>>> +}
>>> +
>>> +static const struct iio_info sca3300_info = {
>>> +    .read_raw = sca3300_read_raw,
>>> +    .write_raw = sca3300_write_raw,
>>> +    .debugfs_reg_access = &sca3300_debugfs_reg_access,
>>> +};
>>> +
>>> +static int sca3300_probe(struct spi_device *spi)
>>> +{
>>> +    struct sca3300_data *sca_data;
>>> +    struct iio_dev *indio_dev;
>>> +    int ret;
>>> +
>>> +    indio_dev = devm_iio_device_alloc(&spi->dev, sizeof(*sca_data));
>>> +    if (!indio_dev) {
>>> +        dev_err(&spi->dev,
>>> +            "failed to allocate memory for iio device\n");
>>> +        return -ENOMEM;
>>> +    }
>>> +
>>> +    sca_data = iio_priv(indio_dev);
>>> +    mutex_init(&sca_data->lock);
>>> +    sca_data->spi = spi;
>>> +    spi_set_drvdata(spi, indio_dev);
>> Won't be used once using devm as below and remove() is dropped, so 
>> drop this.
>>
>>> +
>>> +    crc8_populate_msb(sca3300_crc_table, SCA3300_CRC8_POLYNOMIAL);
>>> +
>>> +    indio_dev->dev.parent = &spi->dev;
>>> +    indio_dev->info = &sca3300_info;
>>> +    indio_dev->name = SCA3300_ALIAS;
>>> +    indio_dev->modes = INDIO_DIRECT_MODE | INDIO_BUFFER_TRIGGERED;
>> BUFFER_TRIGGERED bit now set by iio_triggered_buffer_setup() so 
>> shouldn't
>> be ehre as well.
>>
>>> +    indio_dev->channels = sca3300_channels;
>>> +    indio_dev->num_channels = ARRAY_SIZE(sca3300_channels);
>>> +    indio_dev->available_scan_masks = sca3300_scan_masks;
>>> +
>>> +    if (spi->dev.of_node) {
>>> +        ret = of_property_read_u32(spi->dev.of_node, "murata,opmode",
>> Please use generic firmware access calls rather than the of specific 
>> ones.
>> That lets us get ACPI support for free :)
>>
>> As per the binding review however, this one needs to go in favour of 
>> userspace
>> control of scale + filter frequencies / sampling freq.
>>
>>> + &sca_data->opmode);
>>> +        if (ret < 0)
>>> +            return ret;
>>> +    }
>>> +
>>> +    ret = sca3300_init(sca_data, indio_dev);
>>> +    if (ret < 0) {
>>> +        dev_err(&spi->dev, "failed to init device, error: %d\n", ret);
>>> +        return ret;
>>> +    }
>>> +
>>> +    ret = iio_triggered_buffer_setup(indio_dev, 
>>> iio_pollfunc_store_time,
>>> +                     sca3300_trigger_handler, NULL);
>> devm_iio_triggered_buffer_setup() and you can drop the manual cleanup 
>> in remove()
>> and hence drop remove() entirely as nothing else left.
>>
>>> +    if (ret < 0) {
>>> +        dev_err(&spi->dev,
>>> +            "iio triggered buffer setup failed, error: %d\n", ret);
>>> +        return ret;
>>> +    }
>>> +
>>> +    ret = devm_iio_device_register(&spi->dev, indio_dev);
>>> +    if (ret < 0) {
>>> +        dev_err(&spi->dev, "iio device register failed, error: %d\n",
>>> +            ret);
>>> +        iio_triggered_buffer_cleanup(indio_dev);
>> With devm above, no need to manually clean this up.
>>
>>> +        return ret;
>>> +    }
>>> +
>>> +    return 0;
>>> +}
>>> +
>>> +static int sca3300_remove(struct spi_device *spi)
>>> +{
>>> +    struct iio_dev *indio_dev = spi_get_drvdata(spi);
>>> +
>>> +    iio_triggered_buffer_cleanup(indio_dev);
>>> +    return 0;
>>> +}
>>> +
>>> +static const struct of_device_id sca3300_dt_ids[] = {
>>> +    { .compatible = "murata,sca3300"},
>>> +    {},
>>> +};
>>> +MODULE_DEVICE_TABLE(of, sca3300_dt_ids);
>>> +
>>> +static struct spi_driver sca3300_driver = {
>>> +    .driver = {
>>> +        .name        = SCA3300_ALIAS,
>>> +        .owner        = THIS_MODULE,
>>> +        .of_match_table = of_match_ptr(sca3300_dt_ids),
>>> +    },
>>> +
>>> +    .probe    = sca3300_probe,
>>> +    .remove    = sca3300_remove,
>>> +};
>>> +
>>> +module_spi_driver(sca3300_driver);
>>> +
>>> +MODULE_AUTHOR("Tomas Melin <[e-mail]>");
>>> +MODULE_DESCRIPTION("Murata SCA3300 SPI Accelerometer");
>>> +MODULE_LICENSE("GPL v2");

Tomas Melin April 16, 2021, 12:21 p.m. UTC | #4

Updated email-address for Alexandru.


On 4/16/21 3:17 PM, Tomas Melin wrote:
> On 4/15/21 11:41 AM, Tomas Melin wrote:
>
>>
>> While working on updates I did notice something new which I cannot
>>
>> reproduce on older (5.10.17 kernel) version. If compiling this as a 
>> module, getting error while
>>
>> unloading module:
>>
>> [   40.200084] Unable to handle kernel NULL pointer dereference at 
>> virtual address 00000104
>> ...
>>
>> [   40.510054] Backtrace:
>> [   40.512502] [<c06c5e0c>] (iio_device_ioctl_handler_unregister) 
>> from [<c06ca8c4>] (iio_buffers_free_sysfs_and_mask+0x2c/0x6c)
>> [   40.523735] [<c06ca898>] (iio_buffers_free_sysfs_and_mask) from 
>> [<c06c4a18>] (iio_device_unregister+0xa8/0xac)
>> [   40.533746]  r5:c1811228 r4:c1811000
>> [   40.537318] [<c06c4970>] (iio_device_unregister) from [<c06c4a38>] 
>> (devm_iio_device_unreg+0x1c/0x20)
>> [   40.546461]  r5:c2415000 r4:c25bab80
>> [   40.550025] [<c06c4a1c>] (devm_iio_device_unreg) from [<c0556820>] 
>> (release_nodes+0x1c0/0x1f0)
>> [   40.558654] [<c0556660>] (release_nodes) from [<c0556bb0>] 
>> (devres_release_all+0x40/0x60)
>> [   40.566847]  r10:00000081 r9:c2350000 r8:c0100264 r7:00000081 
>> r6:bf00c010 r5:c19be000
>> [   40.574669]  r4:c1a91c00
>> [   40.577194] [<c0556b70>] (devres_release_all) from [<c055216c>] 
>> (device_release_driver_internal+0x120/0x1cc)
>> [   40.587031]  r5:c19be000 r4:c1a91c00
>> [   40.590596] [<c055204c>] (device_release_driver_internal) from 
>> [<c05522b4>] (driver_detach+0x54/0x90)
>> [   40.599828]  r7:00000081 r6:00000000 r5:bf00c010 r4:c1a91c00
>> [   40.605482] [<c0552260>] (driver_detach) from [<c0550ba4>] 
>> (bus_remove_driver+0x5c/0xb0)
>> [   40.613583]  r5:00000800 r4:bf00c010
>> [   40.617148] [<c0550b48>] (bus_remove_driver) from [<c0552c38>] 
>> (driver_unregister+0x38/0x5c)
>> [   40.625596]  r5:00000800 r4:bf00c010
>> [   40.629161] [<c0552c00>] (driver_unregister) from [<bf00a760>] 
>> (sca3300_driver_exit+0x14/0x8b4 [sca3300])
>> [   40.638747]  r5:00000800 r4:bf00c080
>> [   40.642311] [<bf00a74c>] (sca3300_driver_exit [sca3300]) from 
>> [<c01c06f0>] (sys_delete_module+0x16c/0x238)
>> [   40.651990] [<c01c0584>] (sys_delete_module) from [<c0100244>] 
>> (__sys_trace_return+0x0/0x1c)
>> [   40.660435] Exception stack(0xc2351fa8 to 0xc2351ff0)
>> [   40.665484] 1fa0:                   0050e5a8 00000000 0050e5e4 
>> 00000800 081d4b00 bec18af4
>> [   40.673661] 1fc0: 0050e5a8 00000000 bec18b50 00000081 bec18e51 
>> 0050e190 00000001 bec18d3c
>> [   40.681834] 1fe0: 0050cf70 bec18afc 004f1ec8 b6ecb27c
>> [   40.686887]  r6:bec18b50 r5:00000000 r4:0050e5a8
>> [   40.691507] Code: e8bd4000 e1c020d0 e3a0cc01 e3001122 (e5823004)
>> [   40.707675] ---[ end trace 189882b050077333 ]---
>>
>> This happens when building against linux-next 
>> 5.12.0-rc6-next-20210409. I'm failing to see what is wrong. Any ideas?
>>
>> Thanks,
>>
>> Tomas
>
>
> Tested further that for this driver, loading and unloading as module 
> works fine until commit:
>
> commit f73f7f4da581875f9b1f2fb8ebd1ab15ed634488
> Author: Alexandru Ardelean <[e-mail]>
> Date:   Mon Feb 15 12:40:39 2021 +0200
>
>     iio: buffer: add ioctl() to support opening extra buffers for IIO 
> device
>
>
> Any thoughts what causes this issue?
>
> Thanks,
>
> Tomas
>
>
>
>>
>>
>>>
>>>> ---
>>>>   drivers/iio/accel/Kconfig   |  13 ++
>>>>   drivers/iio/accel/Makefile  |   1 +
>>>>   drivers/iio/accel/sca3300.c | 434 
>>>> ++++++++++++++++++++++++++++++++++++
>>>>   3 files changed, 448 insertions(+)
>>>>   create mode 100644 drivers/iio/accel/sca3300.c
>>>>
>>>> diff --git a/drivers/iio/accel/Kconfig b/drivers/iio/accel/Kconfig
>>>> index cceda3cecbcf..0dbf7b648e8a 100644
>>>> --- a/drivers/iio/accel/Kconfig
>>>> +++ b/drivers/iio/accel/Kconfig
>>>> @@ -450,6 +450,19 @@ config SCA3000
>>>>         To compile this driver as a module, say M here: the module 
>>>> will be
>>>>         called sca3000.
>>>>   +config SCA3300
>>>> +    tristate "Murata SCA3300 3-Axis Accelerometer Driver"
>>>> +    depends on SPI
>>>> +    select CRC8
>>>> +    select IIO_BUFFER
>>>> +    select IIO_TRIGGERED_BUFFER
>>>> +    help
>>>> +      Say yes here to build support for Murata SCA3300 3-Axis
>>>> +      accelerometer.
>>>> +
>>>> +      To compile this driver as a module, choose M here: the 
>>>> module will be
>>>> +      called sca3300.
>>>> +
>>>>   config STK8312
>>>>       tristate "Sensortek STK8312 3-Axis Accelerometer Driver"
>>>>       depends on I2C
>>>> diff --git a/drivers/iio/accel/Makefile b/drivers/iio/accel/Makefile
>>>> index 32cd1342a31a..4b56527a2b97 100644
>>>> --- a/drivers/iio/accel/Makefile
>>>> +++ b/drivers/iio/accel/Makefile
>>>> @@ -50,6 +50,7 @@ obj-$(CONFIG_MXC4005)        += mxc4005.o
>>>>   obj-$(CONFIG_MXC6255)        += mxc6255.o
>>>>     obj-$(CONFIG_SCA3000)        += sca3000.o
>>>> +obj-$(CONFIG_SCA3300)        += sca3300.o
>>>>     obj-$(CONFIG_STK8312)        += stk8312.o
>>>>   obj-$(CONFIG_STK8BA50)        += stk8ba50.o
>>>> diff --git a/drivers/iio/accel/sca3300.c b/drivers/iio/accel/sca3300.c
>>>> new file mode 100644
>>>> index 000000000000..112fb88ecd3a
>>>> --- /dev/null
>>>> +++ b/drivers/iio/accel/sca3300.c
>>>> @@ -0,0 +1,434 @@
>>>> +// SPDX-License-Identifier: GPL-2.0-only
>>>> +/*
>>>> + * Copyright (c) Vaisala Oyj. All rights reserved.
>>> Give a year for the copyright notice if you can.
>>>
>>>> + */
>>>> +#include <linux/crc8.h>
>>>> +#include <linux/delay.h>
>>>> +#include <linux/iio/buffer.h>
>>>> +#include <linux/iio/iio.h>
>>>> +#include <linux/iio/sysfs.h>
>>>> +#include <linux/iio/trigger_consumer.h>
>>>> +#include <linux/iio/triggered_buffer.h>
>>>> +#include <linux/kernel.h>
>>>> +#include <linux/module.h>
>>>> +#include <linux/spi/spi.h>
>>>> +
>>>> +#define SCA3300_ALIAS "sca3300"
>>>> +
>>>> +#define SCA3300_REG_STATUS 0x6
>>>> +#define SCA3300_REG_MODE 0xd
>>>> +#define SCA3300_REG_WHOAMI 0x10
>>>> +#define SCA3300_VALUE_SW_RESET 0x20
>>>> +#define SCA3300_CRC8_POLYNOMIAL 0x1d
>>>> +#define SCA3300_X_READ 0
>>> I wouldn't bother defining this.
>>>
>>>> +#define SCA3300_X_WRITE BIT(7)
>>> Even this one is something I'd just put inline with a comment.
>>>
>>>> +#define SCA3300_DEVICE_ID 0x51
>>>> +#define SCA3300_RS_ERROR 0x3
>>>> +
>>>> +enum sca3300_scan_indexes {
>>>> +    SCA3300_ACC_X = 0,
>>>> +    SCA3300_ACC_Y,
>>>> +    SCA3300_ACC_Z,
>>>> +    SCA3300_TEMP,
>>>> +    SCA3300_TIMESTAMP,
>>>> +};
>>>> +
>>>> +#define SCA3300_ACCEL_CHANNEL(index, reg, axis) { \
>>>> +        .type = IIO_ACCEL,                    \
>>>> +        .address = reg,                        \
>>>> +        .modified = 1,                        \
>>>> +        .channel2 = IIO_MOD_##axis,                \
>>>> +        .info_mask_separate = BIT(IIO_CHAN_INFO_RAW) | \
>>>> +                      BIT(IIO_CHAN_INFO_PROCESSED),    \
>>> As mentioned below, don't provide PROCESSED. Userspace is better at 
>>> handling the
>>> conversion so leave it to them.
>>>
>>>> +        .info_mask_shared_by_type = BIT(IIO_CHAN_INFO_SCALE),    \
>>>> +        .scan_index = index,                    \
>>>> +        .scan_type = {                        \
>>>> +            .sign = 's',                    \
>>>> +            .realbits = 16,                    \
>>>> +            .storagebits = 16,                \
>>>> +            .shift = 0,                    \
>>>> +            .endianness = IIO_CPU,                \
>>>> +        },                            \
>>>> +    }
>>>> +
>>>> +static const struct iio_chan_spec sca3300_channels[] = {
>>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_X, 0x1, X),
>>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Y, 0x2, Y),
>>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Z, 0x3, Z),
>>>> +    {
>>>> +        .type = IIO_TEMP,
>>>> +        .address = 0x5,
>>>> +        .scan_index = SCA3300_TEMP,
>>>> +        .info_mask_separate = BIT(IIO_CHAN_INFO_RAW),
>>>> +        .scan_type = {
>>>> +            .sign = 's',
>>>> +            .realbits = 16,
>>>> +            .storagebits = 16,
>>>> +            .shift = 0,
>>>> +            .endianness = IIO_CPU,
>>>> +        },
>>>> +    },
>>>> +    IIO_CHAN_SOFT_TIMESTAMP(4),
>>>> +};
>>>> +
>>>> +static const int sca3300_accel_scale[] = {2700, 1350, 5400, 5400};
>>>> +
>>>> +static const unsigned long sca3300_scan_masks[] = {
>>>> +    BIT(SCA3300_ACC_X) | BIT(SCA3300_ACC_Y) | BIT(SCA3300_ACC_Z) |
>>>> +    BIT(SCA3300_TEMP),
>>>> +    0};
>>>> +
>>>> +/**
>>>> + * SCA3300 device data
>>> run scripts/kernel-doc over the file and fix the warnings + errors.
>>>
>>>> + *
>>>> + * @spi SPI device structure
>>>> + * @opmode Device operation mode
>>>> + * @lock Data buffer lock
>>>> + * @txbuf Transmit buffer
>>>> + * @rxbuf Receive buffer
>>>> + * @scan Triggered buffer. Four channel 16-bit data + 64-bit 
>>>> timestamp
>>>> + */
>>>> +struct sca3300_data {
>>>> +    struct spi_device *spi;
>>>> +    u32 opmode;
>>>> +    struct mutex lock;
>>>> +    u8 txbuf[4];
>>>> +    u8 rxbuf[4];
>>>> +    struct {
>>>> +        s16 channels[4];
>>>> +        s64 ts __aligned(sizeof(s64));
>>>> +    } scan;
>>>> +};
>>>> +
>>>> +DECLARE_CRC8_TABLE(sca3300_crc_table);
>>>> +
>>>> +static int sca3300_transfer(struct sca3300_data *sca_data, int *val)
>>>> +{
>>>> +    struct spi_delay delay = {.value = 10, .unit = 
>>>> SPI_DELAY_UNIT_USECS};
>>>> +    int32_t ret;
>>>> +    int rs;
>>>> +    u8 crc;
>>>> +    struct spi_transfer xfers[2] = {
>>>> +        {
>>>> +            .tx_buf = sca_data->txbuf,
>>>> +            .rx_buf = NULL,
>>>> +            .len = ARRAY_SIZE(sca_data->txbuf),
>>>> +            .delay = delay,
>>>> +            .cs_change = 1,
>>>> +        },
>>>> +        {
>>>> +            .tx_buf = NULL,
>>>> +            .rx_buf = sca_data->rxbuf,
>>>> +            .len = ARRAY_SIZE(sca_data->rxbuf),
>>>> +            .delay = delay,
>>>> +            .cs_change = 0,
>>>> +        }
>>>> +    };
>>>> +
>>>> +    /* inverted crc value as described in device data sheet */
>>>> +    crc = ~crc8(sca3300_crc_table, &sca_data->txbuf[0], 3, 
>>>> CRC8_INIT_VALUE);
>>>> +    sca_data->txbuf[3] = crc;
>>>> +
>>>> +    ret = spi_sync_transfer(sca_data->spi, xfers, 2);
>>> Use ARRAY_SIZE(xfers) instead of 2.
>>>
>>>> +    if (ret < 0) {
>>>> +        dev_err(&sca_data->spi->dev,
>>>> +            "transfer error, error: %d\n", ret);
>>>> +        return -EIO;
>>>> +    }
>>>> +
>>>> +    crc = ~crc8(sca3300_crc_table, &sca_data->rxbuf[0], 3, 
>>>> CRC8_INIT_VALUE);
>>>> +    if (sca_data->rxbuf[3] != crc) {
>>>> +        dev_err(&sca_data->spi->dev, "CRC checksum mismatch");
>>>> +        return -EIO;
>>>> +    }
>>>> +
>>>> +    /* get return status */
>>>> +    rs = sca_data->rxbuf[0] & 0x03;
>>>> +    if (rs == SCA3300_RS_ERROR)
>>>> +        return rs;
>>>> +
>>>> +    *val = (s16)(sca_data->rxbuf[2] | (sca_data->rxbuf[1] << 8));
>>> Preference for an unaligned endian conversion here and explicit sign 
>>> extend e.g
>>> something like
>>>
>>> sign_extend32(get_unaligned_le16(&sca_data->rxbuf[1]), 15)
>>>
>>>> +
>>>> +    return 0;
>>>> +}
>>>> +
>>>> +static int sca3300_read_reg(struct sca3300_data *sca_data, u8 reg, 
>>>> int *val)
>>>> +{
>>>> +    int ret;
>>>> +
>>>> +    mutex_lock(&sca_data->lock);
>>>> +    sca_data->txbuf[0] = SCA3300_X_READ | (reg << 2);
>>>> +    ret = sca3300_transfer(sca_data, val);
>>>> +    if (ret > 0) {
>>>> +        sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS 
>>>> << 2);
>>>> +        ret = sca3300_transfer(sca_data, val);
>>>> +        /* status 0 = startup, 0x2 = mode change */
>>>> +        if (ret > 0 && *val != 0 && *val != 0x2) {
>>>> + dev_err_ratelimited(&sca_data->spi->dev,
>>>> +                        "device status: %x\n",
>>>> +                        (u16)*val);
>>>> +            mutex_unlock(&sca_data->lock);
>>>> +            return -EIO;
>>>> +        }
>>>> +        if (ret > 0)
>>>> +            ret = 0;
>>>> +    }
>>>> +    mutex_unlock(&sca_data->lock);
>>>> +
>>>> +    return ret;
>>>> +}
>>>> +
>>>> +static int sca3300_write_reg(struct sca3300_data *sca_data, u8 
>>>> reg, int val)
>>>> +{
>>>> +    int reg_val = 0;
>>>> +    int ret;
>>>> +
>>>> +    mutex_lock(&sca_data->lock);
>>>> +    sca_data->txbuf[0] = SCA3300_X_WRITE | (reg << 2);
>>>> +    sca_data->txbuf[1] = val >> 8;
>>>> +    sca_data->txbuf[2] = val & 0xFF;
>>> Prefer (Slightly) an unaligned put.
>>>
>>>> +    ret = sca3300_transfer(sca_data, &reg_val);
>>>> +    if (ret > 0) {
>>> Factor this error handling out to another function, plus trigger it 
>>> from an
>>> appropriate standard error code rather than a positive return value.
>>>
>>>> +        sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS 
>>>> << 2);
>>>> +        ret = sca3300_transfer(sca_data, &reg_val);
>>>> +        /* status 0 = startup, 0x2 = mode change */
>>>> +        if (ret > 0 && reg_val != 0 && reg_val != 0x2) {
>>>> + dev_err_ratelimited(&sca_data->spi->dev,
>>>> +                        "device status: %x\n",
>>>> +                        (u16)reg_val);
>>>> +            mutex_unlock(&sca_data->lock);
>>>> +            return -EIO;
>>>> +        }
>>>> +        if (ret > 0)
>>>> +            ret = 0;
>>> Whenever you get this sort of juggling it rather implies your design is
>>> less than ideal.
>>>
>>>> +    }
>>>> +    mutex_unlock(&sca_data->lock);
>>>> +
>>>> +    return ret;
>>>> +}
>>>> +
>>>> +static int sca3300_write_raw(struct iio_dev *indio_dev,
>>>> +                 struct iio_chan_spec const *chan,
>>>> +                 int val, int val2, long mask)
>>>> +{
>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
>>>> +
>>>> +    switch (mask) {
>>>> +    case IIO_CHAN_INFO_SCALE:
>>>> +        if (val < 0 || val > 3)
>>>> +            return -EINVAL;
>>> No.  Scale values need to be the value of 1LSB not a random integer
>>> that requires you to look up the datasheet.
>>>
>>>> +        return sca3300_write_reg(data, SCA3300_REG_MODE, val);
>>>> +    default:
>>>> +        return -EINVAL;
>>>> +    }
>>>> +}
>>>> +
>>>> +static int sca3300_read_raw(struct iio_dev *indio_dev,
>>>> +                struct iio_chan_spec const *chan,
>>>> +                int *val, int *val2, long mask)
>>>> +{
>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
>>>> +    int ret;
>>>> +    int reg_val;
>>>> +
>>>> +    switch (mask) {
>>>> +    case IIO_CHAN_INFO_RAW:
>>>> +        ret = sca3300_read_reg(data, chan->address, val);
>>>> +        if (ret < 0)
>>>> +            return ret;
>>>> +        return IIO_VAL_INT;
>>>> +    case IIO_CHAN_INFO_SCALE:
>>>> +        ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
>>>> +        if (ret < 0)
>>>> +            return ret;
>>>> +        *val = sca3300_accel_scale[reg_val];
>>> Scale is very very rarely an integer value which makes me wonder what
>>> you are returning here...  It should be the multiplier needed to
>>> take the raw value to a reading in m/sec^2  Given you use it as 
>>> fractional
>>> below, I'm guessing this is 1/scale_value?
>>>
>>>> +        return IIO_VAL_INT;
>>>> +    case IIO_CHAN_INFO_PROCESSED:
>>> Don't provide processed here.   Userspace needs to do the conversion 
>>> itself
>>> (and given it has floating point easily available will probably do a 
>>> better
>>>   job than we can).   I'm assuming the mode can't autonomously change?
>>>
>>>> +        ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
>>>> +        if (ret < 0)
>>>> +            return ret;
>>>> +        *val2 = sca3300_accel_scale[reg_val];
>>>> +        ret = sca3300_read_reg(data, chan->address, val);
>>>> +        if (ret < 0)
>>>> +            return ret;
>>>> +        return IIO_VAL_FRACTIONAL;
>>>> +    default:
>>>> +        return -EINVAL;
>>>> +    }
>>>> +}
>>>> +
>>>> +static irqreturn_t sca3300_trigger_handler(int irq, void *p)
>>>> +{
>>>> +    struct iio_poll_func *pf = p;
>>>> +    struct iio_dev *indio_dev = pf->indio_dev;
>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
>>>> +    s64 time_ns = iio_get_time_ns(indio_dev);
>>> Is the timestamp at the start more accurate that that at the end
>>> of reading the channels?  If not, just put this inline in the
>>> iio_push_to_buffers_with_timestamp() call.
>>>
>>>> +    int bit, ret, val, i = 0;
>>>> +
>>>> +    for_each_set_bit(bit, indio_dev->active_scan_mask,
>>>> +             indio_dev->masklength) {
>>>> +        ret = sca3300_read_reg(data, sca3300_channels[bit].address,
>>>> +                       &val);
>>>> +        if (ret < 0)
>>> Given we can't return an error code from interrupt handlers, it is 
>>> usually
>>> a good idea to print something to the log. Otherwise we get missing 
>>> data
>>> with no idea of why..
>>>
>>>> +            goto out;
>>>> +        if (ARRAY_SIZE(data->scan.channels) > i)
>>> How could this not be true?  If it's always true, don't bother
>>> checking i.
>>>
>>>> +            ((s16 *)data->scan.channels)[i++] = val;
>>>> +    }
>>>> +
>>>> +    iio_push_to_buffers_with_timestamp(indio_dev, &data->scan, 
>>>> time_ns);
>>>> +out:
>>>> +    iio_trigger_notify_done(indio_dev->trig);
>>>> +
>>>> +    return IRQ_HANDLED;
>>>> +}
>>>> +
>>>> +static int sca3300_init(struct sca3300_data *sca_data,
>>>> +            struct iio_dev *indio_dev)
>>>> +{
>>>> +    int ret;
>>>> +    int value = 0;
>>>> +
>>>> +    if (sca_data->opmode < 1 || sca_data->opmode > 4)
>>>> +        return -EINVAL;
>>> Whilst this is going anyway, better to check that next to where
>>> it is read so we drop out immediately rather than buried in this
>>> function.
>>>
>>>> +
>>>> +    ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
>>>> +                SCA3300_VALUE_SW_RESET);
>>>> +    if (ret != 0)
>>>> +        return ret;
>>> For sleeps, good to reference which section in data sheet gives the
>>> timing.
>>>
>>>> +    usleep_range(2e3, 10e3);
>>>> +
>>>> +    ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
>>>> +                sca_data->opmode - 1);
>>>> +    if (ret != 0)
>>> if (ret) see below.
>>>
>>>> +        return ret;
>>>> +    msleep(100);
>>> Again, document where the time comes from. Saves reviewers time if
>>> they want to check it. (I'm too lazy :)
>>>
>>>> +    ret = sca3300_read_reg(sca_data, SCA3300_REG_WHOAMI, &value);
>>>> +    if (ret != 0)
>>> if (ret) is more idiomatic in kernel code.
>>>
>>>> +        return ret;
>>>> +
>>>> +    if (value != SCA3300_DEVICE_ID) {
>>>> +        dev_err(&sca_data->spi->dev, "device id not expected 
>>>> value\n");
>>> Perhaps useful to print what was expected and what was seen. Maybe it's
>>> a new variant and this might help the user to identify that and add 
>>> it to the
>>> driver.
>>>
>>>> +        return -EIO;
>>> -EINVAL probably more appropriate return.
>>>
>>>> +    }
>>>> +    return 0;
>>>> +}
>>>> +
>>>> +static int sca3300_debugfs_reg_access(struct iio_dev *indio_dev,
>>>> +                      unsigned int reg, unsigned int writeval,
>>>> +                      unsigned int *readval)
>>>> +{
>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
>>>> +    int value;
>>>> +    int ret;
>>>> +
>>>> +    if (reg > 0x1f)
>>> Use a define that names that register.
>>>
>>>> +        return -EINVAL;
>>>> +
>>>> +    if (!readval)
>>>> +        return sca3300_write_reg(data, reg, writeval);
>>>> +
>>>> +    ret = sca3300_read_reg(data, reg, &value);
>>>> +    if (ret < 0)
>>>> +        return ret;
>>>> +
>>>> +    *readval = (unsigned int)value;
>>>> +
>>>> +    return 0;
>>>> +}
>>>> +
>>>> +static const struct iio_info sca3300_info = {
>>>> +    .read_raw = sca3300_read_raw,
>>>> +    .write_raw = sca3300_write_raw,
>>>> +    .debugfs_reg_access = &sca3300_debugfs_reg_access,
>>>> +};
>>>> +
>>>> +static int sca3300_probe(struct spi_device *spi)
>>>> +{
>>>> +    struct sca3300_data *sca_data;
>>>> +    struct iio_dev *indio_dev;
>>>> +    int ret;
>>>> +
>>>> +    indio_dev = devm_iio_device_alloc(&spi->dev, sizeof(*sca_data));
>>>> +    if (!indio_dev) {
>>>> +        dev_err(&spi->dev,
>>>> +            "failed to allocate memory for iio device\n");
>>>> +        return -ENOMEM;
>>>> +    }
>>>> +
>>>> +    sca_data = iio_priv(indio_dev);
>>>> +    mutex_init(&sca_data->lock);
>>>> +    sca_data->spi = spi;
>>>> +    spi_set_drvdata(spi, indio_dev);
>>> Won't be used once using devm as below and remove() is dropped, so 
>>> drop this.
>>>
>>>> +
>>>> +    crc8_populate_msb(sca3300_crc_table, SCA3300_CRC8_POLYNOMIAL);
>>>> +
>>>> +    indio_dev->dev.parent = &spi->dev;
>>>> +    indio_dev->info = &sca3300_info;
>>>> +    indio_dev->name = SCA3300_ALIAS;
>>>> +    indio_dev->modes = INDIO_DIRECT_MODE | INDIO_BUFFER_TRIGGERED;
>>> BUFFER_TRIGGERED bit now set by iio_triggered_buffer_setup() so 
>>> shouldn't
>>> be ehre as well.
>>>
>>>> +    indio_dev->channels = sca3300_channels;
>>>> +    indio_dev->num_channels = ARRAY_SIZE(sca3300_channels);
>>>> +    indio_dev->available_scan_masks = sca3300_scan_masks;
>>>> +
>>>> +    if (spi->dev.of_node) {
>>>> +        ret = of_property_read_u32(spi->dev.of_node, "murata,opmode",
>>> Please use generic firmware access calls rather than the of specific 
>>> ones.
>>> That lets us get ACPI support for free :)
>>>
>>> As per the binding review however, this one needs to go in favour of 
>>> userspace
>>> control of scale + filter frequencies / sampling freq.
>>>
>>>> + &sca_data->opmode);
>>>> +        if (ret < 0)
>>>> +            return ret;
>>>> +    }
>>>> +
>>>> +    ret = sca3300_init(sca_data, indio_dev);
>>>> +    if (ret < 0) {
>>>> +        dev_err(&spi->dev, "failed to init device, error: %d\n", 
>>>> ret);
>>>> +        return ret;
>>>> +    }
>>>> +
>>>> +    ret = iio_triggered_buffer_setup(indio_dev, 
>>>> iio_pollfunc_store_time,
>>>> +                     sca3300_trigger_handler, NULL);
>>> devm_iio_triggered_buffer_setup() and you can drop the manual 
>>> cleanup in remove()
>>> and hence drop remove() entirely as nothing else left.
>>>
>>>> +    if (ret < 0) {
>>>> +        dev_err(&spi->dev,
>>>> +            "iio triggered buffer setup failed, error: %d\n", ret);
>>>> +        return ret;
>>>> +    }
>>>> +
>>>> +    ret = devm_iio_device_register(&spi->dev, indio_dev);
>>>> +    if (ret < 0) {
>>>> +        dev_err(&spi->dev, "iio device register failed, error: %d\n",
>>>> +            ret);
>>>> +        iio_triggered_buffer_cleanup(indio_dev);
>>> With devm above, no need to manually clean this up.
>>>
>>>> +        return ret;
>>>> +    }
>>>> +
>>>> +    return 0;
>>>> +}
>>>> +
>>>> +static int sca3300_remove(struct spi_device *spi)
>>>> +{
>>>> +    struct iio_dev *indio_dev = spi_get_drvdata(spi);
>>>> +
>>>> +    iio_triggered_buffer_cleanup(indio_dev);
>>>> +    return 0;
>>>> +}
>>>> +
>>>> +static const struct of_device_id sca3300_dt_ids[] = {
>>>> +    { .compatible = "murata,sca3300"},
>>>> +    {},
>>>> +};
>>>> +MODULE_DEVICE_TABLE(of, sca3300_dt_ids);
>>>> +
>>>> +static struct spi_driver sca3300_driver = {
>>>> +    .driver = {
>>>> +        .name        = SCA3300_ALIAS,
>>>> +        .owner        = THIS_MODULE,
>>>> +        .of_match_table = of_match_ptr(sca3300_dt_ids),
>>>> +    },
>>>> +
>>>> +    .probe    = sca3300_probe,
>>>> +    .remove    = sca3300_remove,
>>>> +};
>>>> +
>>>> +module_spi_driver(sca3300_driver);
>>>> +
>>>> +MODULE_AUTHOR("Tomas Melin <[e-mail]>");
>>>> +MODULE_DESCRIPTION("Murata SCA3300 SPI Accelerometer");
>>>> +MODULE_LICENSE("GPL v2");

Jonathan Cameron April 18, 2021, 10:03 a.m. UTC | #5

On Fri, 16 Apr 2021 15:21:14 +0300
Tomas Melin <[e-mail]> wrote:

> Updated email-address for Alexandru.
> 
> 
> On 4/16/21 3:17 PM, Tomas Melin wrote:
> > On 4/15/21 11:41 AM, Tomas Melin wrote:
> >  
> >>
> >> While working on updates I did notice something new which I cannot
> >>
> >> reproduce on older (5.10.17 kernel) version. If compiling this as a 
> >> module, getting error while
> >>
> >> unloading module:
> >>
> >> [   40.200084] Unable to handle kernel NULL pointer dereference at 
> >> virtual address 00000104
> >> ...
> >>
> >> [   40.510054] Backtrace:
> >> [   40.512502] [<c06c5e0c>] (iio_device_ioctl_handler_unregister) 
> >> from [<c06ca8c4>] (iio_buffers_free_sysfs_and_mask+0x2c/0x6c)
> >> [   40.523735] [<c06ca898>] (iio_buffers_free_sysfs_and_mask) from 
> >> [<c06c4a18>] (iio_device_unregister+0xa8/0xac)
> >> [   40.533746]  r5:c1811228 r4:c1811000
> >> [   40.537318] [<c06c4970>] (iio_device_unregister) from [<c06c4a38>] 
> >> (devm_iio_device_unreg+0x1c/0x20)
> >> [   40.546461]  r5:c2415000 r4:c25bab80
> >> [   40.550025] [<c06c4a1c>] (devm_iio_device_unreg) from [<c0556820>] 
> >> (release_nodes+0x1c0/0x1f0)
> >> [   40.558654] [<c0556660>] (release_nodes) from [<c0556bb0>] 
> >> (devres_release_all+0x40/0x60)
> >> [   40.566847]  r10:00000081 r9:c2350000 r8:c0100264 r7:00000081 
> >> r6:bf00c010 r5:c19be000
> >> [   40.574669]  r4:c1a91c00
> >> [   40.577194] [<c0556b70>] (devres_release_all) from [<c055216c>] 
> >> (device_release_driver_internal+0x120/0x1cc)
> >> [   40.587031]  r5:c19be000 r4:c1a91c00
> >> [   40.590596] [<c055204c>] (device_release_driver_internal) from 
> >> [<c05522b4>] (driver_detach+0x54/0x90)
> >> [   40.599828]  r7:00000081 r6:00000000 r5:bf00c010 r4:c1a91c00
> >> [   40.605482] [<c0552260>] (driver_detach) from [<c0550ba4>] 
> >> (bus_remove_driver+0x5c/0xb0)
> >> [   40.613583]  r5:00000800 r4:bf00c010
> >> [   40.617148] [<c0550b48>] (bus_remove_driver) from [<c0552c38>] 
> >> (driver_unregister+0x38/0x5c)
> >> [   40.625596]  r5:00000800 r4:bf00c010
> >> [   40.629161] [<c0552c00>] (driver_unregister) from [<bf00a760>] 
> >> (sca3300_driver_exit+0x14/0x8b4 [sca3300])
> >> [   40.638747]  r5:00000800 r4:bf00c080
> >> [   40.642311] [<bf00a74c>] (sca3300_driver_exit [sca3300]) from 
> >> [<c01c06f0>] (sys_delete_module+0x16c/0x238)
> >> [   40.651990] [<c01c0584>] (sys_delete_module) from [<c0100244>] 
> >> (__sys_trace_return+0x0/0x1c)
> >> [   40.660435] Exception stack(0xc2351fa8 to 0xc2351ff0)
> >> [   40.665484] 1fa0:                   0050e5a8 00000000 0050e5e4 
> >> 00000800 081d4b00 bec18af4
> >> [   40.673661] 1fc0: 0050e5a8 00000000 bec18b50 00000081 bec18e51 
> >> 0050e190 00000001 bec18d3c
> >> [   40.681834] 1fe0: 0050cf70 bec18afc 004f1ec8 b6ecb27c
> >> [   40.686887]  r6:bec18b50 r5:00000000 r4:0050e5a8
> >> [   40.691507] Code: e8bd4000 e1c020d0 e3a0cc01 e3001122 (e5823004)
> >> [   40.707675] ---[ end trace 189882b050077333 ]---
> >>
> >> This happens when building against linux-next 
> >> 5.12.0-rc6-next-20210409. I'm failing to see what is wrong. Any ideas?
> >>
> >> Thanks,
> >>
> >> Tomas  
> >
> >
> > Tested further that for this driver, loading and unloading as module 
> > works fine until commit:
> >
> > commit f73f7f4da581875f9b1f2fb8ebd1ab15ed634488
> > Author: Alexandru Ardelean <[e-mail]>
> > Date:   Mon Feb 15 12:40:39 2021 +0200
> >
> >     iio: buffer: add ioctl() to support opening extra buffers for IIO 
> > device
> >
> >
> > Any thoughts what causes this issue?

Is this still happening after fixing the ordering in probe / remove?
(devm_iio_triggered_buffer_setup() being easiest way)

As driver currently stands it's calling iio_triggered_buffer_cleanup
before the managed cleanup of the iio_device_register() call.  That
should never happen so would be where I'd look for problems here.

It is possible Alex's work is relying on that ordering being correct
a little more than we previously were. I could be wrong though
and could be something else going on!

Jonathan

+CC Alex's other email address.

> >
> > Thanks,
> >
> > Tomas
> >
> >
> >  
> >>
> >>  
> >>>  
> >>>> ---
> >>>>   drivers/iio/accel/Kconfig   |  13 ++
> >>>>   drivers/iio/accel/Makefile  |   1 +
> >>>>   drivers/iio/accel/sca3300.c | 434 
> >>>> ++++++++++++++++++++++++++++++++++++
> >>>>   3 files changed, 448 insertions(+)
> >>>>   create mode 100644 drivers/iio/accel/sca3300.c
> >>>>
> >>>> diff --git a/drivers/iio/accel/Kconfig b/drivers/iio/accel/Kconfig
> >>>> index cceda3cecbcf..0dbf7b648e8a 100644
> >>>> --- a/drivers/iio/accel/Kconfig
> >>>> +++ b/drivers/iio/accel/Kconfig
> >>>> @@ -450,6 +450,19 @@ config SCA3000
> >>>>         To compile this driver as a module, say M here: the module 
> >>>> will be
> >>>>         called sca3000.
> >>>>   +config SCA3300
> >>>> +    tristate "Murata SCA3300 3-Axis Accelerometer Driver"
> >>>> +    depends on SPI
> >>>> +    select CRC8
> >>>> +    select IIO_BUFFER
> >>>> +    select IIO_TRIGGERED_BUFFER
> >>>> +    help
> >>>> +      Say yes here to build support for Murata SCA3300 3-Axis
> >>>> +      accelerometer.
> >>>> +
> >>>> +      To compile this driver as a module, choose M here: the 
> >>>> module will be
> >>>> +      called sca3300.
> >>>> +
> >>>>   config STK8312
> >>>>       tristate "Sensortek STK8312 3-Axis Accelerometer Driver"
> >>>>       depends on I2C
> >>>> diff --git a/drivers/iio/accel/Makefile b/drivers/iio/accel/Makefile
> >>>> index 32cd1342a31a..4b56527a2b97 100644
> >>>> --- a/drivers/iio/accel/Makefile
> >>>> +++ b/drivers/iio/accel/Makefile
> >>>> @@ -50,6 +50,7 @@ obj-$(CONFIG_MXC4005)        += mxc4005.o
> >>>>   obj-$(CONFIG_MXC6255)        += mxc6255.o
> >>>>     obj-$(CONFIG_SCA3000)        += sca3000.o
> >>>> +obj-$(CONFIG_SCA3300)        += sca3300.o
> >>>>     obj-$(CONFIG_STK8312)        += stk8312.o
> >>>>   obj-$(CONFIG_STK8BA50)        += stk8ba50.o
> >>>> diff --git a/drivers/iio/accel/sca3300.c b/drivers/iio/accel/sca3300.c
> >>>> new file mode 100644
> >>>> index 000000000000..112fb88ecd3a
> >>>> --- /dev/null
> >>>> +++ b/drivers/iio/accel/sca3300.c
> >>>> @@ -0,0 +1,434 @@
> >>>> +// SPDX-License-Identifier: GPL-2.0-only
> >>>> +/*
> >>>> + * Copyright (c) Vaisala Oyj. All rights reserved.  
> >>> Give a year for the copyright notice if you can.
> >>>  
> >>>> + */
> >>>> +#include <linux/crc8.h>
> >>>> +#include <linux/delay.h>
> >>>> +#include <linux/iio/buffer.h>
> >>>> +#include <linux/iio/iio.h>
> >>>> +#include <linux/iio/sysfs.h>
> >>>> +#include <linux/iio/trigger_consumer.h>
> >>>> +#include <linux/iio/triggered_buffer.h>
> >>>> +#include <linux/kernel.h>
> >>>> +#include <linux/module.h>
> >>>> +#include <linux/spi/spi.h>
> >>>> +
> >>>> +#define SCA3300_ALIAS "sca3300"
> >>>> +
> >>>> +#define SCA3300_REG_STATUS 0x6
> >>>> +#define SCA3300_REG_MODE 0xd
> >>>> +#define SCA3300_REG_WHOAMI 0x10
> >>>> +#define SCA3300_VALUE_SW_RESET 0x20
> >>>> +#define SCA3300_CRC8_POLYNOMIAL 0x1d
> >>>> +#define SCA3300_X_READ 0  
> >>> I wouldn't bother defining this.
> >>>  
> >>>> +#define SCA3300_X_WRITE BIT(7)  
> >>> Even this one is something I'd just put inline with a comment.
> >>>  
> >>>> +#define SCA3300_DEVICE_ID 0x51
> >>>> +#define SCA3300_RS_ERROR 0x3
> >>>> +
> >>>> +enum sca3300_scan_indexes {
> >>>> +    SCA3300_ACC_X = 0,
> >>>> +    SCA3300_ACC_Y,
> >>>> +    SCA3300_ACC_Z,
> >>>> +    SCA3300_TEMP,
> >>>> +    SCA3300_TIMESTAMP,
> >>>> +};
> >>>> +
> >>>> +#define SCA3300_ACCEL_CHANNEL(index, reg, axis) { \
> >>>> +        .type = IIO_ACCEL,                    \
> >>>> +        .address = reg,                        \
> >>>> +        .modified = 1,                        \
> >>>> +        .channel2 = IIO_MOD_##axis,                \
> >>>> +        .info_mask_separate = BIT(IIO_CHAN_INFO_RAW) | \
> >>>> +                      BIT(IIO_CHAN_INFO_PROCESSED),    \  
> >>> As mentioned below, don't provide PROCESSED. Userspace is better at 
> >>> handling the
> >>> conversion so leave it to them.
> >>>  
> >>>> +        .info_mask_shared_by_type = BIT(IIO_CHAN_INFO_SCALE),    \
> >>>> +        .scan_index = index,                    \
> >>>> +        .scan_type = {                        \
> >>>> +            .sign = 's',                    \
> >>>> +            .realbits = 16,                    \
> >>>> +            .storagebits = 16,                \
> >>>> +            .shift = 0,                    \
> >>>> +            .endianness = IIO_CPU,                \
> >>>> +        },                            \
> >>>> +    }
> >>>> +
> >>>> +static const struct iio_chan_spec sca3300_channels[] = {
> >>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_X, 0x1, X),
> >>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Y, 0x2, Y),
> >>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Z, 0x3, Z),
> >>>> +    {
> >>>> +        .type = IIO_TEMP,
> >>>> +        .address = 0x5,
> >>>> +        .scan_index = SCA3300_TEMP,
> >>>> +        .info_mask_separate = BIT(IIO_CHAN_INFO_RAW),
> >>>> +        .scan_type = {
> >>>> +            .sign = 's',
> >>>> +            .realbits = 16,
> >>>> +            .storagebits = 16,
> >>>> +            .shift = 0,
> >>>> +            .endianness = IIO_CPU,
> >>>> +        },
> >>>> +    },
> >>>> +    IIO_CHAN_SOFT_TIMESTAMP(4),
> >>>> +};
> >>>> +
> >>>> +static const int sca3300_accel_scale[] = {2700, 1350, 5400, 5400};
> >>>> +
> >>>> +static const unsigned long sca3300_scan_masks[] = {
> >>>> +    BIT(SCA3300_ACC_X) | BIT(SCA3300_ACC_Y) | BIT(SCA3300_ACC_Z) |
> >>>> +    BIT(SCA3300_TEMP),
> >>>> +    0};
> >>>> +
> >>>> +/**
> >>>> + * SCA3300 device data  
> >>> run scripts/kernel-doc over the file and fix the warnings + errors.
> >>>  
> >>>> + *
> >>>> + * @spi SPI device structure
> >>>> + * @opmode Device operation mode
> >>>> + * @lock Data buffer lock
> >>>> + * @txbuf Transmit buffer
> >>>> + * @rxbuf Receive buffer
> >>>> + * @scan Triggered buffer. Four channel 16-bit data + 64-bit 
> >>>> timestamp
> >>>> + */
> >>>> +struct sca3300_data {
> >>>> +    struct spi_device *spi;
> >>>> +    u32 opmode;
> >>>> +    struct mutex lock;
> >>>> +    u8 txbuf[4];
> >>>> +    u8 rxbuf[4];
> >>>> +    struct {
> >>>> +        s16 channels[4];
> >>>> +        s64 ts __aligned(sizeof(s64));
> >>>> +    } scan;
> >>>> +};
> >>>> +
> >>>> +DECLARE_CRC8_TABLE(sca3300_crc_table);
> >>>> +
> >>>> +static int sca3300_transfer(struct sca3300_data *sca_data, int *val)
> >>>> +{
> >>>> +    struct spi_delay delay = {.value = 10, .unit = 
> >>>> SPI_DELAY_UNIT_USECS};
> >>>> +    int32_t ret;
> >>>> +    int rs;
> >>>> +    u8 crc;
> >>>> +    struct spi_transfer xfers[2] = {
> >>>> +        {
> >>>> +            .tx_buf = sca_data->txbuf,
> >>>> +            .rx_buf = NULL,
> >>>> +            .len = ARRAY_SIZE(sca_data->txbuf),
> >>>> +            .delay = delay,
> >>>> +            .cs_change = 1,
> >>>> +        },
> >>>> +        {
> >>>> +            .tx_buf = NULL,
> >>>> +            .rx_buf = sca_data->rxbuf,
> >>>> +            .len = ARRAY_SIZE(sca_data->rxbuf),
> >>>> +            .delay = delay,
> >>>> +            .cs_change = 0,
> >>>> +        }
> >>>> +    };
> >>>> +
> >>>> +    /* inverted crc value as described in device data sheet */
> >>>> +    crc = ~crc8(sca3300_crc_table, &sca_data->txbuf[0], 3, 
> >>>> CRC8_INIT_VALUE);
> >>>> +    sca_data->txbuf[3] = crc;
> >>>> +
> >>>> +    ret = spi_sync_transfer(sca_data->spi, xfers, 2);  
> >>> Use ARRAY_SIZE(xfers) instead of 2.
> >>>  
> >>>> +    if (ret < 0) {
> >>>> +        dev_err(&sca_data->spi->dev,
> >>>> +            "transfer error, error: %d\n", ret);
> >>>> +        return -EIO;
> >>>> +    }
> >>>> +
> >>>> +    crc = ~crc8(sca3300_crc_table, &sca_data->rxbuf[0], 3, 
> >>>> CRC8_INIT_VALUE);
> >>>> +    if (sca_data->rxbuf[3] != crc) {
> >>>> +        dev_err(&sca_data->spi->dev, "CRC checksum mismatch");
> >>>> +        return -EIO;
> >>>> +    }
> >>>> +
> >>>> +    /* get return status */
> >>>> +    rs = sca_data->rxbuf[0] & 0x03;
> >>>> +    if (rs == SCA3300_RS_ERROR)
> >>>> +        return rs;
> >>>> +
> >>>> +    *val = (s16)(sca_data->rxbuf[2] | (sca_data->rxbuf[1] << 8));  
> >>> Preference for an unaligned endian conversion here and explicit sign 
> >>> extend e.g
> >>> something like
> >>>
> >>> sign_extend32(get_unaligned_le16(&sca_data->rxbuf[1]), 15)
> >>>  
> >>>> +
> >>>> +    return 0;
> >>>> +}
> >>>> +
> >>>> +static int sca3300_read_reg(struct sca3300_data *sca_data, u8 reg, 
> >>>> int *val)
> >>>> +{
> >>>> +    int ret;
> >>>> +
> >>>> +    mutex_lock(&sca_data->lock);
> >>>> +    sca_data->txbuf[0] = SCA3300_X_READ | (reg << 2);
> >>>> +    ret = sca3300_transfer(sca_data, val);
> >>>> +    if (ret > 0) {
> >>>> +        sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS 
> >>>> << 2);
> >>>> +        ret = sca3300_transfer(sca_data, val);
> >>>> +        /* status 0 = startup, 0x2 = mode change */
> >>>> +        if (ret > 0 && *val != 0 && *val != 0x2) {
> >>>> + dev_err_ratelimited(&sca_data->spi->dev,
> >>>> +                        "device status: %x\n",
> >>>> +                        (u16)*val);
> >>>> +            mutex_unlock(&sca_data->lock);
> >>>> +            return -EIO;
> >>>> +        }
> >>>> +        if (ret > 0)
> >>>> +            ret = 0;
> >>>> +    }
> >>>> +    mutex_unlock(&sca_data->lock);
> >>>> +
> >>>> +    return ret;
> >>>> +}
> >>>> +
> >>>> +static int sca3300_write_reg(struct sca3300_data *sca_data, u8 
> >>>> reg, int val)
> >>>> +{
> >>>> +    int reg_val = 0;
> >>>> +    int ret;
> >>>> +
> >>>> +    mutex_lock(&sca_data->lock);
> >>>> +    sca_data->txbuf[0] = SCA3300_X_WRITE | (reg << 2);
> >>>> +    sca_data->txbuf[1] = val >> 8;
> >>>> +    sca_data->txbuf[2] = val & 0xFF;  
> >>> Prefer (Slightly) an unaligned put.
> >>>  
> >>>> +    ret = sca3300_transfer(sca_data, &reg_val);
> >>>> +    if (ret > 0) {  
> >>> Factor this error handling out to another function, plus trigger it 
> >>> from an
> >>> appropriate standard error code rather than a positive return value.
> >>>  
> >>>> +        sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS 
> >>>> << 2);
> >>>> +        ret = sca3300_transfer(sca_data, &reg_val);
> >>>> +        /* status 0 = startup, 0x2 = mode change */
> >>>> +        if (ret > 0 && reg_val != 0 && reg_val != 0x2) {
> >>>> + dev_err_ratelimited(&sca_data->spi->dev,
> >>>> +                        "device status: %x\n",
> >>>> +                        (u16)reg_val);
> >>>> +            mutex_unlock(&sca_data->lock);
> >>>> +            return -EIO;
> >>>> +        }
> >>>> +        if (ret > 0)
> >>>> +            ret = 0;  
> >>> Whenever you get this sort of juggling it rather implies your design is
> >>> less than ideal.
> >>>  
> >>>> +    }
> >>>> +    mutex_unlock(&sca_data->lock);
> >>>> +
> >>>> +    return ret;
> >>>> +}
> >>>> +
> >>>> +static int sca3300_write_raw(struct iio_dev *indio_dev,
> >>>> +                 struct iio_chan_spec const *chan,
> >>>> +                 int val, int val2, long mask)
> >>>> +{
> >>>> +    struct sca3300_data *data = iio_priv(indio_dev);
> >>>> +
> >>>> +    switch (mask) {
> >>>> +    case IIO_CHAN_INFO_SCALE:
> >>>> +        if (val < 0 || val > 3)
> >>>> +            return -EINVAL;  
> >>> No.  Scale values need to be the value of 1LSB not a random integer
> >>> that requires you to look up the datasheet.
> >>>  
> >>>> +        return sca3300_write_reg(data, SCA3300_REG_MODE, val);
> >>>> +    default:
> >>>> +        return -EINVAL;
> >>>> +    }
> >>>> +}
> >>>> +
> >>>> +static int sca3300_read_raw(struct iio_dev *indio_dev,
> >>>> +                struct iio_chan_spec const *chan,
> >>>> +                int *val, int *val2, long mask)
> >>>> +{
> >>>> +    struct sca3300_data *data = iio_priv(indio_dev);
> >>>> +    int ret;
> >>>> +    int reg_val;
> >>>> +
> >>>> +    switch (mask) {
> >>>> +    case IIO_CHAN_INFO_RAW:
> >>>> +        ret = sca3300_read_reg(data, chan->address, val);
> >>>> +        if (ret < 0)
> >>>> +            return ret;
> >>>> +        return IIO_VAL_INT;
> >>>> +    case IIO_CHAN_INFO_SCALE:
> >>>> +        ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
> >>>> +        if (ret < 0)
> >>>> +            return ret;
> >>>> +        *val = sca3300_accel_scale[reg_val];  
> >>> Scale is very very rarely an integer value which makes me wonder what
> >>> you are returning here...  It should be the multiplier needed to
> >>> take the raw value to a reading in m/sec^2  Given you use it as 
> >>> fractional
> >>> below, I'm guessing this is 1/scale_value?
> >>>  
> >>>> +        return IIO_VAL_INT;
> >>>> +    case IIO_CHAN_INFO_PROCESSED:  
> >>> Don't provide processed here.   Userspace needs to do the conversion 
> >>> itself
> >>> (and given it has floating point easily available will probably do a 
> >>> better
> >>>   job than we can).   I'm assuming the mode can't autonomously change?
> >>>  
> >>>> +        ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
> >>>> +        if (ret < 0)
> >>>> +            return ret;
> >>>> +        *val2 = sca3300_accel_scale[reg_val];
> >>>> +        ret = sca3300_read_reg(data, chan->address, val);
> >>>> +        if (ret < 0)
> >>>> +            return ret;
> >>>> +        return IIO_VAL_FRACTIONAL;
> >>>> +    default:
> >>>> +        return -EINVAL;
> >>>> +    }
> >>>> +}
> >>>> +
> >>>> +static irqreturn_t sca3300_trigger_handler(int irq, void *p)
> >>>> +{
> >>>> +    struct iio_poll_func *pf = p;
> >>>> +    struct iio_dev *indio_dev = pf->indio_dev;
> >>>> +    struct sca3300_data *data = iio_priv(indio_dev);
> >>>> +    s64 time_ns = iio_get_time_ns(indio_dev);  
> >>> Is the timestamp at the start more accurate that that at the end
> >>> of reading the channels?  If not, just put this inline in the
> >>> iio_push_to_buffers_with_timestamp() call.
> >>>  
> >>>> +    int bit, ret, val, i = 0;
> >>>> +
> >>>> +    for_each_set_bit(bit, indio_dev->active_scan_mask,
> >>>> +             indio_dev->masklength) {
> >>>> +        ret = sca3300_read_reg(data, sca3300_channels[bit].address,
> >>>> +                       &val);
> >>>> +        if (ret < 0)  
> >>> Given we can't return an error code from interrupt handlers, it is 
> >>> usually
> >>> a good idea to print something to the log. Otherwise we get missing 
> >>> data
> >>> with no idea of why..
> >>>  
> >>>> +            goto out;
> >>>> +        if (ARRAY_SIZE(data->scan.channels) > i)  
> >>> How could this not be true?  If it's always true, don't bother
> >>> checking i.
> >>>  
> >>>> +            ((s16 *)data->scan.channels)[i++] = val;
> >>>> +    }
> >>>> +
> >>>> +    iio_push_to_buffers_with_timestamp(indio_dev, &data->scan, 
> >>>> time_ns);
> >>>> +out:
> >>>> +    iio_trigger_notify_done(indio_dev->trig);
> >>>> +
> >>>> +    return IRQ_HANDLED;
> >>>> +}
> >>>> +
> >>>> +static int sca3300_init(struct sca3300_data *sca_data,
> >>>> +            struct iio_dev *indio_dev)
> >>>> +{
> >>>> +    int ret;
> >>>> +    int value = 0;
> >>>> +
> >>>> +    if (sca_data->opmode < 1 || sca_data->opmode > 4)
> >>>> +        return -EINVAL;  
> >>> Whilst this is going anyway, better to check that next to where
> >>> it is read so we drop out immediately rather than buried in this
> >>> function.
> >>>  
> >>>> +
> >>>> +    ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
> >>>> +                SCA3300_VALUE_SW_RESET);
> >>>> +    if (ret != 0)
> >>>> +        return ret;  
> >>> For sleeps, good to reference which section in data sheet gives the
> >>> timing.
> >>>  
> >>>> +    usleep_range(2e3, 10e3);
> >>>> +
> >>>> +    ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
> >>>> +                sca_data->opmode - 1);
> >>>> +    if (ret != 0)  
> >>> if (ret) see below.
> >>>  
> >>>> +        return ret;
> >>>> +    msleep(100);  
> >>> Again, document where the time comes from. Saves reviewers time if
> >>> they want to check it. (I'm too lazy :)
> >>>  
> >>>> +    ret = sca3300_read_reg(sca_data, SCA3300_REG_WHOAMI, &value);
> >>>> +    if (ret != 0)  
> >>> if (ret) is more idiomatic in kernel code.
> >>>  
> >>>> +        return ret;
> >>>> +
> >>>> +    if (value != SCA3300_DEVICE_ID) {
> >>>> +        dev_err(&sca_data->spi->dev, "device id not expected 
> >>>> value\n");  
> >>> Perhaps useful to print what was expected and what was seen. Maybe it's
> >>> a new variant and this might help the user to identify that and add 
> >>> it to the
> >>> driver.
> >>>  
> >>>> +        return -EIO;  
> >>> -EINVAL probably more appropriate return.
> >>>  
> >>>> +    }
> >>>> +    return 0;
> >>>> +}
> >>>> +
> >>>> +static int sca3300_debugfs_reg_access(struct iio_dev *indio_dev,
> >>>> +                      unsigned int reg, unsigned int writeval,
> >>>> +                      unsigned int *readval)
> >>>> +{
> >>>> +    struct sca3300_data *data = iio_priv(indio_dev);
> >>>> +    int value;
> >>>> +    int ret;
> >>>> +
> >>>> +    if (reg > 0x1f)  
> >>> Use a define that names that register.
> >>>  
> >>>> +        return -EINVAL;
> >>>> +
> >>>> +    if (!readval)
> >>>> +        return sca3300_write_reg(data, reg, writeval);
> >>>> +
> >>>> +    ret = sca3300_read_reg(data, reg, &value);
> >>>> +    if (ret < 0)
> >>>> +        return ret;
> >>>> +
> >>>> +    *readval = (unsigned int)value;
> >>>> +
> >>>> +    return 0;
> >>>> +}
> >>>> +
> >>>> +static const struct iio_info sca3300_info = {
> >>>> +    .read_raw = sca3300_read_raw,
> >>>> +    .write_raw = sca3300_write_raw,
> >>>> +    .debugfs_reg_access = &sca3300_debugfs_reg_access,
> >>>> +};
> >>>> +
> >>>> +static int sca3300_probe(struct spi_device *spi)
> >>>> +{
> >>>> +    struct sca3300_data *sca_data;
> >>>> +    struct iio_dev *indio_dev;
> >>>> +    int ret;
> >>>> +
> >>>> +    indio_dev = devm_iio_device_alloc(&spi->dev, sizeof(*sca_data));
> >>>> +    if (!indio_dev) {
> >>>> +        dev_err(&spi->dev,
> >>>> +            "failed to allocate memory for iio device\n");
> >>>> +        return -ENOMEM;
> >>>> +    }
> >>>> +
> >>>> +    sca_data = iio_priv(indio_dev);
> >>>> +    mutex_init(&sca_data->lock);
> >>>> +    sca_data->spi = spi;
> >>>> +    spi_set_drvdata(spi, indio_dev);  
> >>> Won't be used once using devm as below and remove() is dropped, so 
> >>> drop this.
> >>>  
> >>>> +
> >>>> +    crc8_populate_msb(sca3300_crc_table, SCA3300_CRC8_POLYNOMIAL);
> >>>> +
> >>>> +    indio_dev->dev.parent = &spi->dev;
> >>>> +    indio_dev->info = &sca3300_info;
> >>>> +    indio_dev->name = SCA3300_ALIAS;
> >>>> +    indio_dev->modes = INDIO_DIRECT_MODE | INDIO_BUFFER_TRIGGERED;  
> >>> BUFFER_TRIGGERED bit now set by iio_triggered_buffer_setup() so 
> >>> shouldn't
> >>> be ehre as well.
> >>>  
> >>>> +    indio_dev->channels = sca3300_channels;
> >>>> +    indio_dev->num_channels = ARRAY_SIZE(sca3300_channels);
> >>>> +    indio_dev->available_scan_masks = sca3300_scan_masks;
> >>>> +
> >>>> +    if (spi->dev.of_node) {
> >>>> +        ret = of_property_read_u32(spi->dev.of_node, "murata,opmode",  
> >>> Please use generic firmware access calls rather than the of specific 
> >>> ones.
> >>> That lets us get ACPI support for free :)
> >>>
> >>> As per the binding review however, this one needs to go in favour of 
> >>> userspace
> >>> control of scale + filter frequencies / sampling freq.
> >>>  
> >>>> + &sca_data->opmode);
> >>>> +        if (ret < 0)
> >>>> +            return ret;
> >>>> +    }
> >>>> +
> >>>> +    ret = sca3300_init(sca_data, indio_dev);
> >>>> +    if (ret < 0) {
> >>>> +        dev_err(&spi->dev, "failed to init device, error: %d\n", 
> >>>> ret);
> >>>> +        return ret;
> >>>> +    }
> >>>> +
> >>>> +    ret = iio_triggered_buffer_setup(indio_dev, 
> >>>> iio_pollfunc_store_time,
> >>>> +                     sca3300_trigger_handler, NULL);  
> >>> devm_iio_triggered_buffer_setup() and you can drop the manual 
> >>> cleanup in remove()
> >>> and hence drop remove() entirely as nothing else left.
> >>>  
> >>>> +    if (ret < 0) {
> >>>> +        dev_err(&spi->dev,
> >>>> +            "iio triggered buffer setup failed, error: %d\n", ret);
> >>>> +        return ret;
> >>>> +    }
> >>>> +
> >>>> +    ret = devm_iio_device_register(&spi->dev, indio_dev);
> >>>> +    if (ret < 0) {
> >>>> +        dev_err(&spi->dev, "iio device register failed, error: %d\n",
> >>>> +            ret);
> >>>> +        iio_triggered_buffer_cleanup(indio_dev);  
> >>> With devm above, no need to manually clean this up.
> >>>  
> >>>> +        return ret;
> >>>> +    }
> >>>> +
> >>>> +    return 0;
> >>>> +}
> >>>> +
> >>>> +static int sca3300_remove(struct spi_device *spi)
> >>>> +{
> >>>> +    struct iio_dev *indio_dev = spi_get_drvdata(spi);
> >>>> +
> >>>> +    iio_triggered_buffer_cleanup(indio_dev);
> >>>> +    return 0;
> >>>> +}
> >>>> +
> >>>> +static const struct of_device_id sca3300_dt_ids[] = {
> >>>> +    { .compatible = "murata,sca3300"},
> >>>> +    {},
> >>>> +};
> >>>> +MODULE_DEVICE_TABLE(of, sca3300_dt_ids);
> >>>> +
> >>>> +static struct spi_driver sca3300_driver = {
> >>>> +    .driver = {
> >>>> +        .name        = SCA3300_ALIAS,
> >>>> +        .owner        = THIS_MODULE,
> >>>> +        .of_match_table = of_match_ptr(sca3300_dt_ids),
> >>>> +    },
> >>>> +
> >>>> +    .probe    = sca3300_probe,
> >>>> +    .remove    = sca3300_remove,
> >>>> +};
> >>>> +
> >>>> +module_spi_driver(sca3300_driver);
> >>>> +
> >>>> +MODULE_AUTHOR("Tomas Melin <[e-mail]>");
> >>>> +MODULE_DESCRIPTION("Murata SCA3300 SPI Accelerometer");
> >>>> +MODULE_LICENSE("GPL v2");

Tomas Melin April 19, 2021, 1:34 p.m. UTC | #6

Hi,

On 4/18/21 1:03 PM, Jonathan Cameron wrote:
> On Fri, 16 Apr 2021 15:21:14 +0300
> Tomas Melin <[e-mail]> wrote:
>
>> Updated email-address for Alexandru.
>>
>>
>> On 4/16/21 3:17 PM, Tomas Melin wrote:
>>> On 4/15/21 11:41 AM, Tomas Melin wrote:
>>>   
>>>> While working on updates I did notice something new which I cannot
>>>>
>>>> reproduce on older (5.10.17 kernel) version. If compiling this as a
>>>> module, getting error while
>>>>
>>>> unloading module:
>>>>
>>>> [   40.200084] Unable to handle kernel NULL pointer dereference at
>>>> virtual address 00000104
>>>> ...
>>>>
>>>> [   40.510054] Backtrace:
>>>> [   40.512502] [<c06c5e0c>] (iio_device_ioctl_handler_unregister)
>>>> from [<c06ca8c4>] (iio_buffers_free_sysfs_and_mask+0x2c/0x6c)
>>>> [   40.523735] [<c06ca898>] (iio_buffers_free_sysfs_and_mask) from
>>>> [<c06c4a18>] (iio_device_unregister+0xa8/0xac)
>>>> [   40.533746]  r5:c1811228 r4:c1811000
>>>> [   40.537318] [<c06c4970>] (iio_device_unregister) from [<c06c4a38>]
>>>> (devm_iio_device_unreg+0x1c/0x20)
>>>> [   40.546461]  r5:c2415000 r4:c25bab80
>>>> [   40.550025] [<c06c4a1c>] (devm_iio_device_unreg) from [<c0556820>]
>>>> (release_nodes+0x1c0/0x1f0)
>>>> [   40.558654] [<c0556660>] (release_nodes) from [<c0556bb0>]
>>>> (devres_release_all+0x40/0x60)
>>>> [   40.566847]  r10:00000081 r9:c2350000 r8:c0100264 r7:00000081
>>>> r6:bf00c010 r5:c19be000
>>>> [   40.574669]  r4:c1a91c00
>>>> [   40.577194] [<c0556b70>] (devres_release_all) from [<c055216c>]
>>>> (device_release_driver_internal+0x120/0x1cc)
>>>> [   40.587031]  r5:c19be000 r4:c1a91c00
>>>> [   40.590596] [<c055204c>] (device_release_driver_internal) from
>>>> [<c05522b4>] (driver_detach+0x54/0x90)
>>>> [   40.599828]  r7:00000081 r6:00000000 r5:bf00c010 r4:c1a91c00
>>>> [   40.605482] [<c0552260>] (driver_detach) from [<c0550ba4>]
>>>> (bus_remove_driver+0x5c/0xb0)
>>>> [   40.613583]  r5:00000800 r4:bf00c010
>>>> [   40.617148] [<c0550b48>] (bus_remove_driver) from [<c0552c38>]
>>>> (driver_unregister+0x38/0x5c)
>>>> [   40.625596]  r5:00000800 r4:bf00c010
>>>> [   40.629161] [<c0552c00>] (driver_unregister) from [<bf00a760>]
>>>> (sca3300_driver_exit+0x14/0x8b4 [sca3300])
>>>> [   40.638747]  r5:00000800 r4:bf00c080
>>>> [   40.642311] [<bf00a74c>] (sca3300_driver_exit [sca3300]) from
>>>> [<c01c06f0>] (sys_delete_module+0x16c/0x238)
>>>> [   40.651990] [<c01c0584>] (sys_delete_module) from [<c0100244>]
>>>> (__sys_trace_return+0x0/0x1c)
>>>> [   40.660435] Exception stack(0xc2351fa8 to 0xc2351ff0)
>>>> [   40.665484] 1fa0:                   0050e5a8 00000000 0050e5e4
>>>> 00000800 081d4b00 bec18af4
>>>> [   40.673661] 1fc0: 0050e5a8 00000000 bec18b50 00000081 bec18e51
>>>> 0050e190 00000001 bec18d3c
>>>> [   40.681834] 1fe0: 0050cf70 bec18afc 004f1ec8 b6ecb27c
>>>> [   40.686887]  r6:bec18b50 r5:00000000 r4:0050e5a8
>>>> [   40.691507] Code: e8bd4000 e1c020d0 e3a0cc01 e3001122 (e5823004)
>>>> [   40.707675] ---[ end trace 189882b050077333 ]---
>>>>
>>>> This happens when building against linux-next
>>>> 5.12.0-rc6-next-20210409. I'm failing to see what is wrong. Any ideas?
>>>>
>>>> Thanks,
>>>>
>>>> Tomas
>>>
>>> Tested further that for this driver, loading and unloading as module
>>> works fine until commit:
>>>
>>> commit f73f7f4da581875f9b1f2fb8ebd1ab15ed634488
>>> Author: Alexandru Ardelean <[e-mail]>
>>> Date:   Mon Feb 15 12:40:39 2021 +0200
>>>
>>>      iio: buffer: add ioctl() to support opening extra buffers for IIO
>>> device
>>>
>>>
>>> Any thoughts what causes this issue?
> Is this still happening after fixing the ordering in probe / remove?
> (devm_iio_triggered_buffer_setup() being easiest way)
>
> As driver currently stands it's calling iio_triggered_buffer_cleanup
> before the managed cleanup of the iio_device_register() call.  That
> should never happen so would be where I'd look for problems here.
>
> It is possible Alex's work is relying on that ordering being correct
> a little more than we previously were. I could be wrong though
> and could be something else going on!
>
> Jonathan
>
> +CC Alex's other email address.

This is still happening with v3 of driver patchset, which uses only devm_*

versions in probing. v5.12-rc7 works ok, I'm currently only seeing this 
using linux-next.

Thanks,

Tomas


>>> Thanks,
>>>
>>> Tomas
>>>
>>>
>>>   
>>>>   
>>>>>   
>>>>>> ---
>>>>>>    drivers/iio/accel/Kconfig   |  13 ++
>>>>>>    drivers/iio/accel/Makefile  |   1 +
>>>>>>    drivers/iio/accel/sca3300.c | 434
>>>>>> ++++++++++++++++++++++++++++++++++++
>>>>>>    3 files changed, 448 insertions(+)
>>>>>>    create mode 100644 drivers/iio/accel/sca3300.c
>>>>>>
>>>>>> diff --git a/drivers/iio/accel/Kconfig b/drivers/iio/accel/Kconfig
>>>>>> index cceda3cecbcf..0dbf7b648e8a 100644
>>>>>> --- a/drivers/iio/accel/Kconfig
>>>>>> +++ b/drivers/iio/accel/Kconfig
>>>>>> @@ -450,6 +450,19 @@ config SCA3000
>>>>>>          To compile this driver as a module, say M here: the module
>>>>>> will be
>>>>>>          called sca3000.
>>>>>>    +config SCA3300
>>>>>> +    tristate "Murata SCA3300 3-Axis Accelerometer Driver"
>>>>>> +    depends on SPI
>>>>>> +    select CRC8
>>>>>> +    select IIO_BUFFER
>>>>>> +    select IIO_TRIGGERED_BUFFER
>>>>>> +    help
>>>>>> +      Say yes here to build support for Murata SCA3300 3-Axis
>>>>>> +      accelerometer.
>>>>>> +
>>>>>> +      To compile this driver as a module, choose M here: the
>>>>>> module will be
>>>>>> +      called sca3300.
>>>>>> +
>>>>>>    config STK8312
>>>>>>        tristate "Sensortek STK8312 3-Axis Accelerometer Driver"
>>>>>>        depends on I2C
>>>>>> diff --git a/drivers/iio/accel/Makefile b/drivers/iio/accel/Makefile
>>>>>> index 32cd1342a31a..4b56527a2b97 100644
>>>>>> --- a/drivers/iio/accel/Makefile
>>>>>> +++ b/drivers/iio/accel/Makefile
>>>>>> @@ -50,6 +50,7 @@ obj-$(CONFIG_MXC4005)        += mxc4005.o
>>>>>>    obj-$(CONFIG_MXC6255)        += mxc6255.o
>>>>>>      obj-$(CONFIG_SCA3000)        += sca3000.o
>>>>>> +obj-$(CONFIG_SCA3300)        += sca3300.o
>>>>>>      obj-$(CONFIG_STK8312)        += stk8312.o
>>>>>>    obj-$(CONFIG_STK8BA50)        += stk8ba50.o
>>>>>> diff --git a/drivers/iio/accel/sca3300.c b/drivers/iio/accel/sca3300.c
>>>>>> new file mode 100644
>>>>>> index 000000000000..112fb88ecd3a
>>>>>> --- /dev/null
>>>>>> +++ b/drivers/iio/accel/sca3300.c
>>>>>> @@ -0,0 +1,434 @@
>>>>>> +// SPDX-License-Identifier: GPL-2.0-only
>>>>>> +/*
>>>>>> + * Copyright (c) Vaisala Oyj. All rights reserved.
>>>>> Give a year for the copyright notice if you can.
>>>>>   
>>>>>> + */
>>>>>> +#include <linux/crc8.h>
>>>>>> +#include <linux/delay.h>
>>>>>> +#include <linux/iio/buffer.h>
>>>>>> +#include <linux/iio/iio.h>
>>>>>> +#include <linux/iio/sysfs.h>
>>>>>> +#include <linux/iio/trigger_consumer.h>
>>>>>> +#include <linux/iio/triggered_buffer.h>
>>>>>> +#include <linux/kernel.h>
>>>>>> +#include <linux/module.h>
>>>>>> +#include <linux/spi/spi.h>
>>>>>> +
>>>>>> +#define SCA3300_ALIAS "sca3300"
>>>>>> +
>>>>>> +#define SCA3300_REG_STATUS 0x6
>>>>>> +#define SCA3300_REG_MODE 0xd
>>>>>> +#define SCA3300_REG_WHOAMI 0x10
>>>>>> +#define SCA3300_VALUE_SW_RESET 0x20
>>>>>> +#define SCA3300_CRC8_POLYNOMIAL 0x1d
>>>>>> +#define SCA3300_X_READ 0
>>>>> I wouldn't bother defining this.
>>>>>   
>>>>>> +#define SCA3300_X_WRITE BIT(7)
>>>>> Even this one is something I'd just put inline with a comment.
>>>>>   
>>>>>> +#define SCA3300_DEVICE_ID 0x51
>>>>>> +#define SCA3300_RS_ERROR 0x3
>>>>>> +
>>>>>> +enum sca3300_scan_indexes {
>>>>>> +    SCA3300_ACC_X = 0,
>>>>>> +    SCA3300_ACC_Y,
>>>>>> +    SCA3300_ACC_Z,
>>>>>> +    SCA3300_TEMP,
>>>>>> +    SCA3300_TIMESTAMP,
>>>>>> +};
>>>>>> +
>>>>>> +#define SCA3300_ACCEL_CHANNEL(index, reg, axis) { \
>>>>>> +        .type = IIO_ACCEL,                    \
>>>>>> +        .address = reg,                        \
>>>>>> +        .modified = 1,                        \
>>>>>> +        .channel2 = IIO_MOD_##axis,                \
>>>>>> +        .info_mask_separate = BIT(IIO_CHAN_INFO_RAW) | \
>>>>>> +                      BIT(IIO_CHAN_INFO_PROCESSED),    \
>>>>> As mentioned below, don't provide PROCESSED. Userspace is better at
>>>>> handling the
>>>>> conversion so leave it to them.
>>>>>   
>>>>>> +        .info_mask_shared_by_type = BIT(IIO_CHAN_INFO_SCALE),    \
>>>>>> +        .scan_index = index,                    \
>>>>>> +        .scan_type = {                        \
>>>>>> +            .sign = 's',                    \
>>>>>> +            .realbits = 16,                    \
>>>>>> +            .storagebits = 16,                \
>>>>>> +            .shift = 0,                    \
>>>>>> +            .endianness = IIO_CPU,                \
>>>>>> +        },                            \
>>>>>> +    }
>>>>>> +
>>>>>> +static const struct iio_chan_spec sca3300_channels[] = {
>>>>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_X, 0x1, X),
>>>>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Y, 0x2, Y),
>>>>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Z, 0x3, Z),
>>>>>> +    {
>>>>>> +        .type = IIO_TEMP,
>>>>>> +        .address = 0x5,
>>>>>> +        .scan_index = SCA3300_TEMP,
>>>>>> +        .info_mask_separate = BIT(IIO_CHAN_INFO_RAW),
>>>>>> +        .scan_type = {
>>>>>> +            .sign = 's',
>>>>>> +            .realbits = 16,
>>>>>> +            .storagebits = 16,
>>>>>> +            .shift = 0,
>>>>>> +            .endianness = IIO_CPU,
>>>>>> +        },
>>>>>> +    },
>>>>>> +    IIO_CHAN_SOFT_TIMESTAMP(4),
>>>>>> +};
>>>>>> +
>>>>>> +static const int sca3300_accel_scale[] = {2700, 1350, 5400, 5400};
>>>>>> +
>>>>>> +static const unsigned long sca3300_scan_masks[] = {
>>>>>> +    BIT(SCA3300_ACC_X) | BIT(SCA3300_ACC_Y) | BIT(SCA3300_ACC_Z) |
>>>>>> +    BIT(SCA3300_TEMP),
>>>>>> +    0};
>>>>>> +
>>>>>> +/**
>>>>>> + * SCA3300 device data
>>>>> run scripts/kernel-doc over the file and fix the warnings + errors.
>>>>>   
>>>>>> + *
>>>>>> + * @spi SPI device structure
>>>>>> + * @opmode Device operation mode
>>>>>> + * @lock Data buffer lock
>>>>>> + * @txbuf Transmit buffer
>>>>>> + * @rxbuf Receive buffer
>>>>>> + * @scan Triggered buffer. Four channel 16-bit data + 64-bit
>>>>>> timestamp
>>>>>> + */
>>>>>> +struct sca3300_data {
>>>>>> +    struct spi_device *spi;
>>>>>> +    u32 opmode;
>>>>>> +    struct mutex lock;
>>>>>> +    u8 txbuf[4];
>>>>>> +    u8 rxbuf[4];
>>>>>> +    struct {
>>>>>> +        s16 channels[4];
>>>>>> +        s64 ts __aligned(sizeof(s64));
>>>>>> +    } scan;
>>>>>> +};
>>>>>> +
>>>>>> +DECLARE_CRC8_TABLE(sca3300_crc_table);
>>>>>> +
>>>>>> +static int sca3300_transfer(struct sca3300_data *sca_data, int *val)
>>>>>> +{
>>>>>> +    struct spi_delay delay = {.value = 10, .unit =
>>>>>> SPI_DELAY_UNIT_USECS};
>>>>>> +    int32_t ret;
>>>>>> +    int rs;
>>>>>> +    u8 crc;
>>>>>> +    struct spi_transfer xfers[2] = {
>>>>>> +        {
>>>>>> +            .tx_buf = sca_data->txbuf,
>>>>>> +            .rx_buf = NULL,
>>>>>> +            .len = ARRAY_SIZE(sca_data->txbuf),
>>>>>> +            .delay = delay,
>>>>>> +            .cs_change = 1,
>>>>>> +        },
>>>>>> +        {
>>>>>> +            .tx_buf = NULL,
>>>>>> +            .rx_buf = sca_data->rxbuf,
>>>>>> +            .len = ARRAY_SIZE(sca_data->rxbuf),
>>>>>> +            .delay = delay,
>>>>>> +            .cs_change = 0,
>>>>>> +        }
>>>>>> +    };
>>>>>> +
>>>>>> +    /* inverted crc value as described in device data sheet */
>>>>>> +    crc = ~crc8(sca3300_crc_table, &sca_data->txbuf[0], 3,
>>>>>> CRC8_INIT_VALUE);
>>>>>> +    sca_data->txbuf[3] = crc;
>>>>>> +
>>>>>> +    ret = spi_sync_transfer(sca_data->spi, xfers, 2);
>>>>> Use ARRAY_SIZE(xfers) instead of 2.
>>>>>   
>>>>>> +    if (ret < 0) {
>>>>>> +        dev_err(&sca_data->spi->dev,
>>>>>> +            "transfer error, error: %d\n", ret);
>>>>>> +        return -EIO;
>>>>>> +    }
>>>>>> +
>>>>>> +    crc = ~crc8(sca3300_crc_table, &sca_data->rxbuf[0], 3,
>>>>>> CRC8_INIT_VALUE);
>>>>>> +    if (sca_data->rxbuf[3] != crc) {
>>>>>> +        dev_err(&sca_data->spi->dev, "CRC checksum mismatch");
>>>>>> +        return -EIO;
>>>>>> +    }
>>>>>> +
>>>>>> +    /* get return status */
>>>>>> +    rs = sca_data->rxbuf[0] & 0x03;
>>>>>> +    if (rs == SCA3300_RS_ERROR)
>>>>>> +        return rs;
>>>>>> +
>>>>>> +    *val = (s16)(sca_data->rxbuf[2] | (sca_data->rxbuf[1] << 8));
>>>>> Preference for an unaligned endian conversion here and explicit sign
>>>>> extend e.g
>>>>> something like
>>>>>
>>>>> sign_extend32(get_unaligned_le16(&sca_data->rxbuf[1]), 15)
>>>>>   
>>>>>> +
>>>>>> +    return 0;
>>>>>> +}
>>>>>> +
>>>>>> +static int sca3300_read_reg(struct sca3300_data *sca_data, u8 reg,
>>>>>> int *val)
>>>>>> +{
>>>>>> +    int ret;
>>>>>> +
>>>>>> +    mutex_lock(&sca_data->lock);
>>>>>> +    sca_data->txbuf[0] = SCA3300_X_READ | (reg << 2);
>>>>>> +    ret = sca3300_transfer(sca_data, val);
>>>>>> +    if (ret > 0) {
>>>>>> +        sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS
>>>>>> << 2);
>>>>>> +        ret = sca3300_transfer(sca_data, val);
>>>>>> +        /* status 0 = startup, 0x2 = mode change */
>>>>>> +        if (ret > 0 && *val != 0 && *val != 0x2) {
>>>>>> + dev_err_ratelimited(&sca_data->spi->dev,
>>>>>> +                        "device status: %x\n",
>>>>>> +                        (u16)*val);
>>>>>> +            mutex_unlock(&sca_data->lock);
>>>>>> +            return -EIO;
>>>>>> +        }
>>>>>> +        if (ret > 0)
>>>>>> +            ret = 0;
>>>>>> +    }
>>>>>> +    mutex_unlock(&sca_data->lock);
>>>>>> +
>>>>>> +    return ret;
>>>>>> +}
>>>>>> +
>>>>>> +static int sca3300_write_reg(struct sca3300_data *sca_data, u8
>>>>>> reg, int val)
>>>>>> +{
>>>>>> +    int reg_val = 0;
>>>>>> +    int ret;
>>>>>> +
>>>>>> +    mutex_lock(&sca_data->lock);
>>>>>> +    sca_data->txbuf[0] = SCA3300_X_WRITE | (reg << 2);
>>>>>> +    sca_data->txbuf[1] = val >> 8;
>>>>>> +    sca_data->txbuf[2] = val & 0xFF;
>>>>> Prefer (Slightly) an unaligned put.
>>>>>   
>>>>>> +    ret = sca3300_transfer(sca_data, &reg_val);
>>>>>> +    if (ret > 0) {
>>>>> Factor this error handling out to another function, plus trigger it
>>>>> from an
>>>>> appropriate standard error code rather than a positive return value.
>>>>>   
>>>>>> +        sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS
>>>>>> << 2);
>>>>>> +        ret = sca3300_transfer(sca_data, &reg_val);
>>>>>> +        /* status 0 = startup, 0x2 = mode change */
>>>>>> +        if (ret > 0 && reg_val != 0 && reg_val != 0x2) {
>>>>>> + dev_err_ratelimited(&sca_data->spi->dev,
>>>>>> +                        "device status: %x\n",
>>>>>> +                        (u16)reg_val);
>>>>>> +            mutex_unlock(&sca_data->lock);
>>>>>> +            return -EIO;
>>>>>> +        }
>>>>>> +        if (ret > 0)
>>>>>> +            ret = 0;
>>>>> Whenever you get this sort of juggling it rather implies your design is
>>>>> less than ideal.
>>>>>   
>>>>>> +    }
>>>>>> +    mutex_unlock(&sca_data->lock);
>>>>>> +
>>>>>> +    return ret;
>>>>>> +}
>>>>>> +
>>>>>> +static int sca3300_write_raw(struct iio_dev *indio_dev,
>>>>>> +                 struct iio_chan_spec const *chan,
>>>>>> +                 int val, int val2, long mask)
>>>>>> +{
>>>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
>>>>>> +
>>>>>> +    switch (mask) {
>>>>>> +    case IIO_CHAN_INFO_SCALE:
>>>>>> +        if (val < 0 || val > 3)
>>>>>> +            return -EINVAL;
>>>>> No.  Scale values need to be the value of 1LSB not a random integer
>>>>> that requires you to look up the datasheet.
>>>>>   
>>>>>> +        return sca3300_write_reg(data, SCA3300_REG_MODE, val);
>>>>>> +    default:
>>>>>> +        return -EINVAL;
>>>>>> +    }
>>>>>> +}
>>>>>> +
>>>>>> +static int sca3300_read_raw(struct iio_dev *indio_dev,
>>>>>> +                struct iio_chan_spec const *chan,
>>>>>> +                int *val, int *val2, long mask)
>>>>>> +{
>>>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
>>>>>> +    int ret;
>>>>>> +    int reg_val;
>>>>>> +
>>>>>> +    switch (mask) {
>>>>>> +    case IIO_CHAN_INFO_RAW:
>>>>>> +        ret = sca3300_read_reg(data, chan->address, val);
>>>>>> +        if (ret < 0)
>>>>>> +            return ret;
>>>>>> +        return IIO_VAL_INT;
>>>>>> +    case IIO_CHAN_INFO_SCALE:
>>>>>> +        ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
>>>>>> +        if (ret < 0)
>>>>>> +            return ret;
>>>>>> +        *val = sca3300_accel_scale[reg_val];
>>>>> Scale is very very rarely an integer value which makes me wonder what
>>>>> you are returning here...  It should be the multiplier needed to
>>>>> take the raw value to a reading in m/sec^2  Given you use it as
>>>>> fractional
>>>>> below, I'm guessing this is 1/scale_value?
>>>>>   
>>>>>> +        return IIO_VAL_INT;
>>>>>> +    case IIO_CHAN_INFO_PROCESSED:
>>>>> Don't provide processed here.   Userspace needs to do the conversion
>>>>> itself
>>>>> (and given it has floating point easily available will probably do a
>>>>> better
>>>>>    job than we can).   I'm assuming the mode can't autonomously change?
>>>>>   
>>>>>> +        ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
>>>>>> +        if (ret < 0)
>>>>>> +            return ret;
>>>>>> +        *val2 = sca3300_accel_scale[reg_val];
>>>>>> +        ret = sca3300_read_reg(data, chan->address, val);
>>>>>> +        if (ret < 0)
>>>>>> +            return ret;
>>>>>> +        return IIO_VAL_FRACTIONAL;
>>>>>> +    default:
>>>>>> +        return -EINVAL;
>>>>>> +    }
>>>>>> +}
>>>>>> +
>>>>>> +static irqreturn_t sca3300_trigger_handler(int irq, void *p)
>>>>>> +{
>>>>>> +    struct iio_poll_func *pf = p;
>>>>>> +    struct iio_dev *indio_dev = pf->indio_dev;
>>>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
>>>>>> +    s64 time_ns = iio_get_time_ns(indio_dev);
>>>>> Is the timestamp at the start more accurate that that at the end
>>>>> of reading the channels?  If not, just put this inline in the
>>>>> iio_push_to_buffers_with_timestamp() call.
>>>>>   
>>>>>> +    int bit, ret, val, i = 0;
>>>>>> +
>>>>>> +    for_each_set_bit(bit, indio_dev->active_scan_mask,
>>>>>> +             indio_dev->masklength) {
>>>>>> +        ret = sca3300_read_reg(data, sca3300_channels[bit].address,
>>>>>> +                       &val);
>>>>>> +        if (ret < 0)
>>>>> Given we can't return an error code from interrupt handlers, it is
>>>>> usually
>>>>> a good idea to print something to the log. Otherwise we get missing
>>>>> data
>>>>> with no idea of why..
>>>>>   
>>>>>> +            goto out;
>>>>>> +        if (ARRAY_SIZE(data->scan.channels) > i)
>>>>> How could this not be true?  If it's always true, don't bother
>>>>> checking i.
>>>>>   
>>>>>> +            ((s16 *)data->scan.channels)[i++] = val;
>>>>>> +    }
>>>>>> +
>>>>>> +    iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
>>>>>> time_ns);
>>>>>> +out:
>>>>>> +    iio_trigger_notify_done(indio_dev->trig);
>>>>>> +
>>>>>> +    return IRQ_HANDLED;
>>>>>> +}
>>>>>> +
>>>>>> +static int sca3300_init(struct sca3300_data *sca_data,
>>>>>> +            struct iio_dev *indio_dev)
>>>>>> +{
>>>>>> +    int ret;
>>>>>> +    int value = 0;
>>>>>> +
>>>>>> +    if (sca_data->opmode < 1 || sca_data->opmode > 4)
>>>>>> +        return -EINVAL;
>>>>> Whilst this is going anyway, better to check that next to where
>>>>> it is read so we drop out immediately rather than buried in this
>>>>> function.
>>>>>   
>>>>>> +
>>>>>> +    ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
>>>>>> +                SCA3300_VALUE_SW_RESET);
>>>>>> +    if (ret != 0)
>>>>>> +        return ret;
>>>>> For sleeps, good to reference which section in data sheet gives the
>>>>> timing.
>>>>>   
>>>>>> +    usleep_range(2e3, 10e3);
>>>>>> +
>>>>>> +    ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
>>>>>> +                sca_data->opmode - 1);
>>>>>> +    if (ret != 0)
>>>>> if (ret) see below.
>>>>>   
>>>>>> +        return ret;
>>>>>> +    msleep(100);
>>>>> Again, document where the time comes from. Saves reviewers time if
>>>>> they want to check it. (I'm too lazy :)
>>>>>   
>>>>>> +    ret = sca3300_read_reg(sca_data, SCA3300_REG_WHOAMI, &value);
>>>>>> +    if (ret != 0)
>>>>> if (ret) is more idiomatic in kernel code.
>>>>>   
>>>>>> +        return ret;
>>>>>> +
>>>>>> +    if (value != SCA3300_DEVICE_ID) {
>>>>>> +        dev_err(&sca_data->spi->dev, "device id not expected
>>>>>> value\n");
>>>>> Perhaps useful to print what was expected and what was seen. Maybe it's
>>>>> a new variant and this might help the user to identify that and add
>>>>> it to the
>>>>> driver.
>>>>>   
>>>>>> +        return -EIO;
>>>>> -EINVAL probably more appropriate return.
>>>>>   
>>>>>> +    }
>>>>>> +    return 0;
>>>>>> +}
>>>>>> +
>>>>>> +static int sca3300_debugfs_reg_access(struct iio_dev *indio_dev,
>>>>>> +                      unsigned int reg, unsigned int writeval,
>>>>>> +                      unsigned int *readval)
>>>>>> +{
>>>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
>>>>>> +    int value;
>>>>>> +    int ret;
>>>>>> +
>>>>>> +    if (reg > 0x1f)
>>>>> Use a define that names that register.
>>>>>   
>>>>>> +        return -EINVAL;
>>>>>> +
>>>>>> +    if (!readval)
>>>>>> +        return sca3300_write_reg(data, reg, writeval);
>>>>>> +
>>>>>> +    ret = sca3300_read_reg(data, reg, &value);
>>>>>> +    if (ret < 0)
>>>>>> +        return ret;
>>>>>> +
>>>>>> +    *readval = (unsigned int)value;
>>>>>> +
>>>>>> +    return 0;
>>>>>> +}
>>>>>> +
>>>>>> +static const struct iio_info sca3300_info = {
>>>>>> +    .read_raw = sca3300_read_raw,
>>>>>> +    .write_raw = sca3300_write_raw,
>>>>>> +    .debugfs_reg_access = &sca3300_debugfs_reg_access,
>>>>>> +};
>>>>>> +
>>>>>> +static int sca3300_probe(struct spi_device *spi)
>>>>>> +{
>>>>>> +    struct sca3300_data *sca_data;
>>>>>> +    struct iio_dev *indio_dev;
>>>>>> +    int ret;
>>>>>> +
>>>>>> +    indio_dev = devm_iio_device_alloc(&spi->dev, sizeof(*sca_data));
>>>>>> +    if (!indio_dev) {
>>>>>> +        dev_err(&spi->dev,
>>>>>> +            "failed to allocate memory for iio device\n");
>>>>>> +        return -ENOMEM;
>>>>>> +    }
>>>>>> +
>>>>>> +    sca_data = iio_priv(indio_dev);
>>>>>> +    mutex_init(&sca_data->lock);
>>>>>> +    sca_data->spi = spi;
>>>>>> +    spi_set_drvdata(spi, indio_dev);
>>>>> Won't be used once using devm as below and remove() is dropped, so
>>>>> drop this.
>>>>>   
>>>>>> +
>>>>>> +    crc8_populate_msb(sca3300_crc_table, SCA3300_CRC8_POLYNOMIAL);
>>>>>> +
>>>>>> +    indio_dev->dev.parent = &spi->dev;
>>>>>> +    indio_dev->info = &sca3300_info;
>>>>>> +    indio_dev->name = SCA3300_ALIAS;
>>>>>> +    indio_dev->modes = INDIO_DIRECT_MODE | INDIO_BUFFER_TRIGGERED;
>>>>> BUFFER_TRIGGERED bit now set by iio_triggered_buffer_setup() so
>>>>> shouldn't
>>>>> be ehre as well.
>>>>>   
>>>>>> +    indio_dev->channels = sca3300_channels;
>>>>>> +    indio_dev->num_channels = ARRAY_SIZE(sca3300_channels);
>>>>>> +    indio_dev->available_scan_masks = sca3300_scan_masks;
>>>>>> +
>>>>>> +    if (spi->dev.of_node) {
>>>>>> +        ret = of_property_read_u32(spi->dev.of_node, "murata,opmode",
>>>>> Please use generic firmware access calls rather than the of specific
>>>>> ones.
>>>>> That lets us get ACPI support for free :)
>>>>>
>>>>> As per the binding review however, this one needs to go in favour of
>>>>> userspace
>>>>> control of scale + filter frequencies / sampling freq.
>>>>>   
>>>>>> + &sca_data->opmode);
>>>>>> +        if (ret < 0)
>>>>>> +            return ret;
>>>>>> +    }
>>>>>> +
>>>>>> +    ret = sca3300_init(sca_data, indio_dev);
>>>>>> +    if (ret < 0) {
>>>>>> +        dev_err(&spi->dev, "failed to init device, error: %d\n",
>>>>>> ret);
>>>>>> +        return ret;
>>>>>> +    }
>>>>>> +
>>>>>> +    ret = iio_triggered_buffer_setup(indio_dev,
>>>>>> iio_pollfunc_store_time,
>>>>>> +                     sca3300_trigger_handler, NULL);
>>>>> devm_iio_triggered_buffer_setup() and you can drop the manual
>>>>> cleanup in remove()
>>>>> and hence drop remove() entirely as nothing else left.
>>>>>   
>>>>>> +    if (ret < 0) {
>>>>>> +        dev_err(&spi->dev,
>>>>>> +            "iio triggered buffer setup failed, error: %d\n", ret);
>>>>>> +        return ret;
>>>>>> +    }
>>>>>> +
>>>>>> +    ret = devm_iio_device_register(&spi->dev, indio_dev);
>>>>>> +    if (ret < 0) {
>>>>>> +        dev_err(&spi->dev, "iio device register failed, error: %d\n",
>>>>>> +            ret);
>>>>>> +        iio_triggered_buffer_cleanup(indio_dev);
>>>>> With devm above, no need to manually clean this up.
>>>>>   
>>>>>> +        return ret;
>>>>>> +    }
>>>>>> +
>>>>>> +    return 0;
>>>>>> +}
>>>>>> +
>>>>>> +static int sca3300_remove(struct spi_device *spi)
>>>>>> +{
>>>>>> +    struct iio_dev *indio_dev = spi_get_drvdata(spi);
>>>>>> +
>>>>>> +    iio_triggered_buffer_cleanup(indio_dev);
>>>>>> +    return 0;
>>>>>> +}
>>>>>> +
>>>>>> +static const struct of_device_id sca3300_dt_ids[] = {
>>>>>> +    { .compatible = "murata,sca3300"},
>>>>>> +    {},
>>>>>> +};
>>>>>> +MODULE_DEVICE_TABLE(of, sca3300_dt_ids);
>>>>>> +
>>>>>> +static struct spi_driver sca3300_driver = {
>>>>>> +    .driver = {
>>>>>> +        .name        = SCA3300_ALIAS,
>>>>>> +        .owner        = THIS_MODULE,
>>>>>> +        .of_match_table = of_match_ptr(sca3300_dt_ids),
>>>>>> +    },
>>>>>> +
>>>>>> +    .probe    = sca3300_probe,
>>>>>> +    .remove    = sca3300_remove,
>>>>>> +};
>>>>>> +
>>>>>> +module_spi_driver(sca3300_driver);
>>>>>> +
>>>>>> +MODULE_AUTHOR("Tomas Melin <[e-mail]>");
>>>>>> +MODULE_DESCRIPTION("Murata SCA3300 SPI Accelerometer");
>>>>>> +MODULE_LICENSE("GPL v2");

Jonathan Cameron April 19, 2021, 7:45 p.m. UTC | #7

On Mon, 19 Apr 2021 16:34:35 +0300
Tomas Melin <[e-mail]> wrote:

> Hi,
> 
> On 4/18/21 1:03 PM, Jonathan Cameron wrote:
> > On Fri, 16 Apr 2021 15:21:14 +0300
> > Tomas Melin <[e-mail]> wrote:
> >  
> >> Updated email-address for Alexandru.
> >>
> >>
> >> On 4/16/21 3:17 PM, Tomas Melin wrote:  
> >>> On 4/15/21 11:41 AM, Tomas Melin wrote:
> >>>     
> >>>> While working on updates I did notice something new which I cannot
> >>>>
> >>>> reproduce on older (5.10.17 kernel) version. If compiling this as a
> >>>> module, getting error while
> >>>>
> >>>> unloading module:
> >>>>
> >>>> [   40.200084] Unable to handle kernel NULL pointer dereference at
> >>>> virtual address 00000104
> >>>> ...
> >>>>
> >>>> [   40.510054] Backtrace:
> >>>> [   40.512502] [<c06c5e0c>] (iio_device_ioctl_handler_unregister)
> >>>> from [<c06ca8c4>] (iio_buffers_free_sysfs_and_mask+0x2c/0x6c)
> >>>> [   40.523735] [<c06ca898>] (iio_buffers_free_sysfs_and_mask) from
> >>>> [<c06c4a18>] (iio_device_unregister+0xa8/0xac)
> >>>> [   40.533746]  r5:c1811228 r4:c1811000
> >>>> [   40.537318] [<c06c4970>] (iio_device_unregister) from [<c06c4a38>]
> >>>> (devm_iio_device_unreg+0x1c/0x20)
> >>>> [   40.546461]  r5:c2415000 r4:c25bab80
> >>>> [   40.550025] [<c06c4a1c>] (devm_iio_device_unreg) from [<c0556820>]
> >>>> (release_nodes+0x1c0/0x1f0)
> >>>> [   40.558654] [<c0556660>] (release_nodes) from [<c0556bb0>]
> >>>> (devres_release_all+0x40/0x60)
> >>>> [   40.566847]  r10:00000081 r9:c2350000 r8:c0100264 r7:00000081
> >>>> r6:bf00c010 r5:c19be000
> >>>> [   40.574669]  r4:c1a91c00
> >>>> [   40.577194] [<c0556b70>] (devres_release_all) from [<c055216c>]
> >>>> (device_release_driver_internal+0x120/0x1cc)
> >>>> [   40.587031]  r5:c19be000 r4:c1a91c00
> >>>> [   40.590596] [<c055204c>] (device_release_driver_internal) from
> >>>> [<c05522b4>] (driver_detach+0x54/0x90)
> >>>> [   40.599828]  r7:00000081 r6:00000000 r5:bf00c010 r4:c1a91c00
> >>>> [   40.605482] [<c0552260>] (driver_detach) from [<c0550ba4>]
> >>>> (bus_remove_driver+0x5c/0xb0)
> >>>> [   40.613583]  r5:00000800 r4:bf00c010
> >>>> [   40.617148] [<c0550b48>] (bus_remove_driver) from [<c0552c38>]
> >>>> (driver_unregister+0x38/0x5c)
> >>>> [   40.625596]  r5:00000800 r4:bf00c010
> >>>> [   40.629161] [<c0552c00>] (driver_unregister) from [<bf00a760>]
> >>>> (sca3300_driver_exit+0x14/0x8b4 [sca3300])
> >>>> [   40.638747]  r5:00000800 r4:bf00c080
> >>>> [   40.642311] [<bf00a74c>] (sca3300_driver_exit [sca3300]) from
> >>>> [<c01c06f0>] (sys_delete_module+0x16c/0x238)
> >>>> [   40.651990] [<c01c0584>] (sys_delete_module) from [<c0100244>]
> >>>> (__sys_trace_return+0x0/0x1c)
> >>>> [   40.660435] Exception stack(0xc2351fa8 to 0xc2351ff0)
> >>>> [   40.665484] 1fa0:                   0050e5a8 00000000 0050e5e4
> >>>> 00000800 081d4b00 bec18af4
> >>>> [   40.673661] 1fc0: 0050e5a8 00000000 bec18b50 00000081 bec18e51
> >>>> 0050e190 00000001 bec18d3c
> >>>> [   40.681834] 1fe0: 0050cf70 bec18afc 004f1ec8 b6ecb27c
> >>>> [   40.686887]  r6:bec18b50 r5:00000000 r4:0050e5a8
> >>>> [   40.691507] Code: e8bd4000 e1c020d0 e3a0cc01 e3001122 (e5823004)
> >>>> [   40.707675] ---[ end trace 189882b050077333 ]---
> >>>>
> >>>> This happens when building against linux-next
> >>>> 5.12.0-rc6-next-20210409. I'm failing to see what is wrong. Any ideas?
> >>>>
> >>>> Thanks,
> >>>>
> >>>> Tomas  
> >>>
> >>> Tested further that for this driver, loading and unloading as module
> >>> works fine until commit:
> >>>
> >>> commit f73f7f4da581875f9b1f2fb8ebd1ab15ed634488
> >>> Author: Alexandru Ardelean <[e-mail]>
> >>> Date:   Mon Feb 15 12:40:39 2021 +0200
> >>>
> >>>      iio: buffer: add ioctl() to support opening extra buffers for IIO
> >>> device
> >>>
> >>>
> >>> Any thoughts what causes this issue?  
> > Is this still happening after fixing the ordering in probe / remove?
> > (devm_iio_triggered_buffer_setup() being easiest way)
> >
> > As driver currently stands it's calling iio_triggered_buffer_cleanup
> > before the managed cleanup of the iio_device_register() call.  That
> > should never happen so would be where I'd look for problems here.
> >
> > It is possible Alex's work is relying on that ordering being correct
> > a little more than we previously were. I could be wrong though
> > and could be something else going on!
> >
> > Jonathan
> >
> > +CC Alex's other email address.  
> 
> This is still happening with v3 of driver patchset, which uses only devm_*
> 
> versions in probing. v5.12-rc7 works ok, I'm currently only seeing this 
> using linux-next.

Thanks for checking.  I've just fired up a VM and confirmed that the same
thing is happening with creating and removing instances of the dummy driver.

Will see if I can figure out what's going wrong.

Jonathan

> 
> Thanks,
> 
> Tomas
> 
> 
> >>> Thanks,
> >>>
> >>> Tomas
> >>>
> >>>
> >>>     
> >>>>     
> >>>>>     
> >>>>>> ---
> >>>>>>    drivers/iio/accel/Kconfig   |  13 ++
> >>>>>>    drivers/iio/accel/Makefile  |   1 +
> >>>>>>    drivers/iio/accel/sca3300.c | 434
> >>>>>> ++++++++++++++++++++++++++++++++++++
> >>>>>>    3 files changed, 448 insertions(+)
> >>>>>>    create mode 100644 drivers/iio/accel/sca3300.c
> >>>>>>
> >>>>>> diff --git a/drivers/iio/accel/Kconfig b/drivers/iio/accel/Kconfig
> >>>>>> index cceda3cecbcf..0dbf7b648e8a 100644
> >>>>>> --- a/drivers/iio/accel/Kconfig
> >>>>>> +++ b/drivers/iio/accel/Kconfig
> >>>>>> @@ -450,6 +450,19 @@ config SCA3000
> >>>>>>          To compile this driver as a module, say M here: the module
> >>>>>> will be
> >>>>>>          called sca3000.
> >>>>>>    +config SCA3300
> >>>>>> +    tristate "Murata SCA3300 3-Axis Accelerometer Driver"
> >>>>>> +    depends on SPI
> >>>>>> +    select CRC8
> >>>>>> +    select IIO_BUFFER
> >>>>>> +    select IIO_TRIGGERED_BUFFER
> >>>>>> +    help
> >>>>>> +      Say yes here to build support for Murata SCA3300 3-Axis
> >>>>>> +      accelerometer.
> >>>>>> +
> >>>>>> +      To compile this driver as a module, choose M here: the
> >>>>>> module will be
> >>>>>> +      called sca3300.
> >>>>>> +
> >>>>>>    config STK8312
> >>>>>>        tristate "Sensortek STK8312 3-Axis Accelerometer Driver"
> >>>>>>        depends on I2C
> >>>>>> diff --git a/drivers/iio/accel/Makefile b/drivers/iio/accel/Makefile
> >>>>>> index 32cd1342a31a..4b56527a2b97 100644
> >>>>>> --- a/drivers/iio/accel/Makefile
> >>>>>> +++ b/drivers/iio/accel/Makefile
> >>>>>> @@ -50,6 +50,7 @@ obj-$(CONFIG_MXC4005)        += mxc4005.o
> >>>>>>    obj-$(CONFIG_MXC6255)        += mxc6255.o
> >>>>>>      obj-$(CONFIG_SCA3000)        += sca3000.o
> >>>>>> +obj-$(CONFIG_SCA3300)        += sca3300.o
> >>>>>>      obj-$(CONFIG_STK8312)        += stk8312.o
> >>>>>>    obj-$(CONFIG_STK8BA50)        += stk8ba50.o
> >>>>>> diff --git a/drivers/iio/accel/sca3300.c b/drivers/iio/accel/sca3300.c
> >>>>>> new file mode 100644
> >>>>>> index 000000000000..112fb88ecd3a
> >>>>>> --- /dev/null
> >>>>>> +++ b/drivers/iio/accel/sca3300.c
> >>>>>> @@ -0,0 +1,434 @@
> >>>>>> +// SPDX-License-Identifier: GPL-2.0-only
> >>>>>> +/*
> >>>>>> + * Copyright (c) Vaisala Oyj. All rights reserved.  
> >>>>> Give a year for the copyright notice if you can.
> >>>>>     
> >>>>>> + */
> >>>>>> +#include <linux/crc8.h>
> >>>>>> +#include <linux/delay.h>
> >>>>>> +#include <linux/iio/buffer.h>
> >>>>>> +#include <linux/iio/iio.h>
> >>>>>> +#include <linux/iio/sysfs.h>
> >>>>>> +#include <linux/iio/trigger_consumer.h>
> >>>>>> +#include <linux/iio/triggered_buffer.h>
> >>>>>> +#include <linux/kernel.h>
> >>>>>> +#include <linux/module.h>
> >>>>>> +#include <linux/spi/spi.h>
> >>>>>> +
> >>>>>> +#define SCA3300_ALIAS "sca3300"
> >>>>>> +
> >>>>>> +#define SCA3300_REG_STATUS 0x6
> >>>>>> +#define SCA3300_REG_MODE 0xd
> >>>>>> +#define SCA3300_REG_WHOAMI 0x10
> >>>>>> +#define SCA3300_VALUE_SW_RESET 0x20
> >>>>>> +#define SCA3300_CRC8_POLYNOMIAL 0x1d
> >>>>>> +#define SCA3300_X_READ 0  
> >>>>> I wouldn't bother defining this.
> >>>>>     
> >>>>>> +#define SCA3300_X_WRITE BIT(7)  
> >>>>> Even this one is something I'd just put inline with a comment.
> >>>>>     
> >>>>>> +#define SCA3300_DEVICE_ID 0x51
> >>>>>> +#define SCA3300_RS_ERROR 0x3
> >>>>>> +
> >>>>>> +enum sca3300_scan_indexes {
> >>>>>> +    SCA3300_ACC_X = 0,
> >>>>>> +    SCA3300_ACC_Y,
> >>>>>> +    SCA3300_ACC_Z,
> >>>>>> +    SCA3300_TEMP,
> >>>>>> +    SCA3300_TIMESTAMP,
> >>>>>> +};
> >>>>>> +
> >>>>>> +#define SCA3300_ACCEL_CHANNEL(index, reg, axis) { \
> >>>>>> +        .type = IIO_ACCEL,                    \
> >>>>>> +        .address = reg,                        \
> >>>>>> +        .modified = 1,                        \
> >>>>>> +        .channel2 = IIO_MOD_##axis,                \
> >>>>>> +        .info_mask_separate = BIT(IIO_CHAN_INFO_RAW) | \
> >>>>>> +                      BIT(IIO_CHAN_INFO_PROCESSED),    \  
> >>>>> As mentioned below, don't provide PROCESSED. Userspace is better at
> >>>>> handling the
> >>>>> conversion so leave it to them.
> >>>>>     
> >>>>>> +        .info_mask_shared_by_type = BIT(IIO_CHAN_INFO_SCALE),    \
> >>>>>> +        .scan_index = index,                    \
> >>>>>> +        .scan_type = {                        \
> >>>>>> +            .sign = 's',                    \
> >>>>>> +            .realbits = 16,                    \
> >>>>>> +            .storagebits = 16,                \
> >>>>>> +            .shift = 0,                    \
> >>>>>> +            .endianness = IIO_CPU,                \
> >>>>>> +        },                            \
> >>>>>> +    }
> >>>>>> +
> >>>>>> +static const struct iio_chan_spec sca3300_channels[] = {
> >>>>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_X, 0x1, X),
> >>>>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Y, 0x2, Y),
> >>>>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Z, 0x3, Z),
> >>>>>> +    {
> >>>>>> +        .type = IIO_TEMP,
> >>>>>> +        .address = 0x5,
> >>>>>> +        .scan_index = SCA3300_TEMP,
> >>>>>> +        .info_mask_separate = BIT(IIO_CHAN_INFO_RAW),
> >>>>>> +        .scan_type = {
> >>>>>> +            .sign = 's',
> >>>>>> +            .realbits = 16,
> >>>>>> +            .storagebits = 16,
> >>>>>> +            .shift = 0,
> >>>>>> +            .endianness = IIO_CPU,
> >>>>>> +        },
> >>>>>> +    },
> >>>>>> +    IIO_CHAN_SOFT_TIMESTAMP(4),
> >>>>>> +};
> >>>>>> +
> >>>>>> +static const int sca3300_accel_scale[] = {2700, 1350, 5400, 5400};
> >>>>>> +
> >>>>>> +static const unsigned long sca3300_scan_masks[] = {
> >>>>>> +    BIT(SCA3300_ACC_X) | BIT(SCA3300_ACC_Y) | BIT(SCA3300_ACC_Z) |
> >>>>>> +    BIT(SCA3300_TEMP),
> >>>>>> +    0};
> >>>>>> +
> >>>>>> +/**
> >>>>>> + * SCA3300 device data  
> >>>>> run scripts/kernel-doc over the file and fix the warnings + errors.
> >>>>>     
> >>>>>> + *
> >>>>>> + * @spi SPI device structure
> >>>>>> + * @opmode Device operation mode
> >>>>>> + * @lock Data buffer lock
> >>>>>> + * @txbuf Transmit buffer
> >>>>>> + * @rxbuf Receive buffer
> >>>>>> + * @scan Triggered buffer. Four channel 16-bit data + 64-bit
> >>>>>> timestamp
> >>>>>> + */
> >>>>>> +struct sca3300_data {
> >>>>>> +    struct spi_device *spi;
> >>>>>> +    u32 opmode;
> >>>>>> +    struct mutex lock;
> >>>>>> +    u8 txbuf[4];
> >>>>>> +    u8 rxbuf[4];
> >>>>>> +    struct {
> >>>>>> +        s16 channels[4];
> >>>>>> +        s64 ts __aligned(sizeof(s64));
> >>>>>> +    } scan;
> >>>>>> +};
> >>>>>> +
> >>>>>> +DECLARE_CRC8_TABLE(sca3300_crc_table);
> >>>>>> +
> >>>>>> +static int sca3300_transfer(struct sca3300_data *sca_data, int *val)
> >>>>>> +{
> >>>>>> +    struct spi_delay delay = {.value = 10, .unit =
> >>>>>> SPI_DELAY_UNIT_USECS};
> >>>>>> +    int32_t ret;
> >>>>>> +    int rs;
> >>>>>> +    u8 crc;
> >>>>>> +    struct spi_transfer xfers[2] = {
> >>>>>> +        {
> >>>>>> +            .tx_buf = sca_data->txbuf,
> >>>>>> +            .rx_buf = NULL,
> >>>>>> +            .len = ARRAY_SIZE(sca_data->txbuf),
> >>>>>> +            .delay = delay,
> >>>>>> +            .cs_change = 1,
> >>>>>> +        },
> >>>>>> +        {
> >>>>>> +            .tx_buf = NULL,
> >>>>>> +            .rx_buf = sca_data->rxbuf,
> >>>>>> +            .len = ARRAY_SIZE(sca_data->rxbuf),
> >>>>>> +            .delay = delay,
> >>>>>> +            .cs_change = 0,
> >>>>>> +        }
> >>>>>> +    };
> >>>>>> +
> >>>>>> +    /* inverted crc value as described in device data sheet */
> >>>>>> +    crc = ~crc8(sca3300_crc_table, &sca_data->txbuf[0], 3,
> >>>>>> CRC8_INIT_VALUE);
> >>>>>> +    sca_data->txbuf[3] = crc;
> >>>>>> +
> >>>>>> +    ret = spi_sync_transfer(sca_data->spi, xfers, 2);  
> >>>>> Use ARRAY_SIZE(xfers) instead of 2.
> >>>>>     
> >>>>>> +    if (ret < 0) {
> >>>>>> +        dev_err(&sca_data->spi->dev,
> >>>>>> +            "transfer error, error: %d\n", ret);
> >>>>>> +        return -EIO;
> >>>>>> +    }
> >>>>>> +
> >>>>>> +    crc = ~crc8(sca3300_crc_table, &sca_data->rxbuf[0], 3,
> >>>>>> CRC8_INIT_VALUE);
> >>>>>> +    if (sca_data->rxbuf[3] != crc) {
> >>>>>> +        dev_err(&sca_data->spi->dev, "CRC checksum mismatch");
> >>>>>> +        return -EIO;
> >>>>>> +    }
> >>>>>> +
> >>>>>> +    /* get return status */
> >>>>>> +    rs = sca_data->rxbuf[0] & 0x03;
> >>>>>> +    if (rs == SCA3300_RS_ERROR)
> >>>>>> +        return rs;
> >>>>>> +
> >>>>>> +    *val = (s16)(sca_data->rxbuf[2] | (sca_data->rxbuf[1] << 8));  
> >>>>> Preference for an unaligned endian conversion here and explicit sign
> >>>>> extend e.g
> >>>>> something like
> >>>>>
> >>>>> sign_extend32(get_unaligned_le16(&sca_data->rxbuf[1]), 15)
> >>>>>     
> >>>>>> +
> >>>>>> +    return 0;
> >>>>>> +}
> >>>>>> +
> >>>>>> +static int sca3300_read_reg(struct sca3300_data *sca_data, u8 reg,
> >>>>>> int *val)
> >>>>>> +{
> >>>>>> +    int ret;
> >>>>>> +
> >>>>>> +    mutex_lock(&sca_data->lock);
> >>>>>> +    sca_data->txbuf[0] = SCA3300_X_READ | (reg << 2);
> >>>>>> +    ret = sca3300_transfer(sca_data, val);
> >>>>>> +    if (ret > 0) {
> >>>>>> +        sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS
> >>>>>> << 2);
> >>>>>> +        ret = sca3300_transfer(sca_data, val);
> >>>>>> +        /* status 0 = startup, 0x2 = mode change */
> >>>>>> +        if (ret > 0 && *val != 0 && *val != 0x2) {
> >>>>>> + dev_err_ratelimited(&sca_data->spi->dev,
> >>>>>> +                        "device status: %x\n",
> >>>>>> +                        (u16)*val);
> >>>>>> +            mutex_unlock(&sca_data->lock);
> >>>>>> +            return -EIO;
> >>>>>> +        }
> >>>>>> +        if (ret > 0)
> >>>>>> +            ret = 0;
> >>>>>> +    }
> >>>>>> +    mutex_unlock(&sca_data->lock);
> >>>>>> +
> >>>>>> +    return ret;
> >>>>>> +}
> >>>>>> +
> >>>>>> +static int sca3300_write_reg(struct sca3300_data *sca_data, u8
> >>>>>> reg, int val)
> >>>>>> +{
> >>>>>> +    int reg_val = 0;
> >>>>>> +    int ret;
> >>>>>> +
> >>>>>> +    mutex_lock(&sca_data->lock);
> >>>>>> +    sca_data->txbuf[0] = SCA3300_X_WRITE | (reg << 2);
> >>>>>> +    sca_data->txbuf[1] = val >> 8;
> >>>>>> +    sca_data->txbuf[2] = val & 0xFF;  
> >>>>> Prefer (Slightly) an unaligned put.
> >>>>>     
> >>>>>> +    ret = sca3300_transfer(sca_data, &reg_val);
> >>>>>> +    if (ret > 0) {  
> >>>>> Factor this error handling out to another function, plus trigger it
> >>>>> from an
> >>>>> appropriate standard error code rather than a positive return value.
> >>>>>     
> >>>>>> +        sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS
> >>>>>> << 2);
> >>>>>> +        ret = sca3300_transfer(sca_data, &reg_val);
> >>>>>> +        /* status 0 = startup, 0x2 = mode change */
> >>>>>> +        if (ret > 0 && reg_val != 0 && reg_val != 0x2) {
> >>>>>> + dev_err_ratelimited(&sca_data->spi->dev,
> >>>>>> +                        "device status: %x\n",
> >>>>>> +                        (u16)reg_val);
> >>>>>> +            mutex_unlock(&sca_data->lock);
> >>>>>> +            return -EIO;
> >>>>>> +        }
> >>>>>> +        if (ret > 0)
> >>>>>> +            ret = 0;  
> >>>>> Whenever you get this sort of juggling it rather implies your design is
> >>>>> less than ideal.
> >>>>>     
> >>>>>> +    }
> >>>>>> +    mutex_unlock(&sca_data->lock);
> >>>>>> +
> >>>>>> +    return ret;
> >>>>>> +}
> >>>>>> +
> >>>>>> +static int sca3300_write_raw(struct iio_dev *indio_dev,
> >>>>>> +                 struct iio_chan_spec const *chan,
> >>>>>> +                 int val, int val2, long mask)
> >>>>>> +{
> >>>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
> >>>>>> +
> >>>>>> +    switch (mask) {
> >>>>>> +    case IIO_CHAN_INFO_SCALE:
> >>>>>> +        if (val < 0 || val > 3)
> >>>>>> +            return -EINVAL;  
> >>>>> No.  Scale values need to be the value of 1LSB not a random integer
> >>>>> that requires you to look up the datasheet.
> >>>>>     
> >>>>>> +        return sca3300_write_reg(data, SCA3300_REG_MODE, val);
> >>>>>> +    default:
> >>>>>> +        return -EINVAL;
> >>>>>> +    }
> >>>>>> +}
> >>>>>> +
> >>>>>> +static int sca3300_read_raw(struct iio_dev *indio_dev,
> >>>>>> +                struct iio_chan_spec const *chan,
> >>>>>> +                int *val, int *val2, long mask)
> >>>>>> +{
> >>>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
> >>>>>> +    int ret;
> >>>>>> +    int reg_val;
> >>>>>> +
> >>>>>> +    switch (mask) {
> >>>>>> +    case IIO_CHAN_INFO_RAW:
> >>>>>> +        ret = sca3300_read_reg(data, chan->address, val);
> >>>>>> +        if (ret < 0)
> >>>>>> +            return ret;
> >>>>>> +        return IIO_VAL_INT;
> >>>>>> +    case IIO_CHAN_INFO_SCALE:
> >>>>>> +        ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
> >>>>>> +        if (ret < 0)
> >>>>>> +            return ret;
> >>>>>> +        *val = sca3300_accel_scale[reg_val];  
> >>>>> Scale is very very rarely an integer value which makes me wonder what
> >>>>> you are returning here...  It should be the multiplier needed to
> >>>>> take the raw value to a reading in m/sec^2  Given you use it as
> >>>>> fractional
> >>>>> below, I'm guessing this is 1/scale_value?
> >>>>>     
> >>>>>> +        return IIO_VAL_INT;
> >>>>>> +    case IIO_CHAN_INFO_PROCESSED:  
> >>>>> Don't provide processed here.   Userspace needs to do the conversion
> >>>>> itself
> >>>>> (and given it has floating point easily available will probably do a
> >>>>> better
> >>>>>    job than we can).   I'm assuming the mode can't autonomously change?
> >>>>>     
> >>>>>> +        ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
> >>>>>> +        if (ret < 0)
> >>>>>> +            return ret;
> >>>>>> +        *val2 = sca3300_accel_scale[reg_val];
> >>>>>> +        ret = sca3300_read_reg(data, chan->address, val);
> >>>>>> +        if (ret < 0)
> >>>>>> +            return ret;
> >>>>>> +        return IIO_VAL_FRACTIONAL;
> >>>>>> +    default:
> >>>>>> +        return -EINVAL;
> >>>>>> +    }
> >>>>>> +}
> >>>>>> +
> >>>>>> +static irqreturn_t sca3300_trigger_handler(int irq, void *p)
> >>>>>> +{
> >>>>>> +    struct iio_poll_func *pf = p;
> >>>>>> +    struct iio_dev *indio_dev = pf->indio_dev;
> >>>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
> >>>>>> +    s64 time_ns = iio_get_time_ns(indio_dev);  
> >>>>> Is the timestamp at the start more accurate that that at the end
> >>>>> of reading the channels?  If not, just put this inline in the
> >>>>> iio_push_to_buffers_with_timestamp() call.
> >>>>>     
> >>>>>> +    int bit, ret, val, i = 0;
> >>>>>> +
> >>>>>> +    for_each_set_bit(bit, indio_dev->active_scan_mask,
> >>>>>> +             indio_dev->masklength) {
> >>>>>> +        ret = sca3300_read_reg(data, sca3300_channels[bit].address,
> >>>>>> +                       &val);
> >>>>>> +        if (ret < 0)  
> >>>>> Given we can't return an error code from interrupt handlers, it is
> >>>>> usually
> >>>>> a good idea to print something to the log. Otherwise we get missing
> >>>>> data
> >>>>> with no idea of why..
> >>>>>     
> >>>>>> +            goto out;
> >>>>>> +        if (ARRAY_SIZE(data->scan.channels) > i)  
> >>>>> How could this not be true?  If it's always true, don't bother
> >>>>> checking i.
> >>>>>     
> >>>>>> +            ((s16 *)data->scan.channels)[i++] = val;
> >>>>>> +    }
> >>>>>> +
> >>>>>> +    iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
> >>>>>> time_ns);
> >>>>>> +out:
> >>>>>> +    iio_trigger_notify_done(indio_dev->trig);
> >>>>>> +
> >>>>>> +    return IRQ_HANDLED;
> >>>>>> +}
> >>>>>> +
> >>>>>> +static int sca3300_init(struct sca3300_data *sca_data,
> >>>>>> +            struct iio_dev *indio_dev)
> >>>>>> +{
> >>>>>> +    int ret;
> >>>>>> +    int value = 0;
> >>>>>> +
> >>>>>> +    if (sca_data->opmode < 1 || sca_data->opmode > 4)
> >>>>>> +        return -EINVAL;  
> >>>>> Whilst this is going anyway, better to check that next to where
> >>>>> it is read so we drop out immediately rather than buried in this
> >>>>> function.
> >>>>>     
> >>>>>> +
> >>>>>> +    ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
> >>>>>> +                SCA3300_VALUE_SW_RESET);
> >>>>>> +    if (ret != 0)
> >>>>>> +        return ret;  
> >>>>> For sleeps, good to reference which section in data sheet gives the
> >>>>> timing.
> >>>>>     
> >>>>>> +    usleep_range(2e3, 10e3);
> >>>>>> +
> >>>>>> +    ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
> >>>>>> +                sca_data->opmode - 1);
> >>>>>> +    if (ret != 0)  
> >>>>> if (ret) see below.
> >>>>>     
> >>>>>> +        return ret;
> >>>>>> +    msleep(100);  
> >>>>> Again, document where the time comes from. Saves reviewers time if
> >>>>> they want to check it. (I'm too lazy :)
> >>>>>     
> >>>>>> +    ret = sca3300_read_reg(sca_data, SCA3300_REG_WHOAMI, &value);
> >>>>>> +    if (ret != 0)  
> >>>>> if (ret) is more idiomatic in kernel code.
> >>>>>     
> >>>>>> +        return ret;
> >>>>>> +
> >>>>>> +    if (value != SCA3300_DEVICE_ID) {
> >>>>>> +        dev_err(&sca_data->spi->dev, "device id not expected
> >>>>>> value\n");  
> >>>>> Perhaps useful to print what was expected and what was seen. Maybe it's
> >>>>> a new variant and this might help the user to identify that and add
> >>>>> it to the
> >>>>> driver.
> >>>>>     
> >>>>>> +        return -EIO;  
> >>>>> -EINVAL probably more appropriate return.
> >>>>>     
> >>>>>> +    }
> >>>>>> +    return 0;
> >>>>>> +}
> >>>>>> +
> >>>>>> +static int sca3300_debugfs_reg_access(struct iio_dev *indio_dev,
> >>>>>> +                      unsigned int reg, unsigned int writeval,
> >>>>>> +                      unsigned int *readval)
> >>>>>> +{
> >>>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
> >>>>>> +    int value;
> >>>>>> +    int ret;
> >>>>>> +
> >>>>>> +    if (reg > 0x1f)  
> >>>>> Use a define that names that register.
> >>>>>     
> >>>>>> +        return -EINVAL;
> >>>>>> +
> >>>>>> +    if (!readval)
> >>>>>> +        return sca3300_write_reg(data, reg, writeval);
> >>>>>> +
> >>>>>> +    ret = sca3300_read_reg(data, reg, &value);
> >>>>>> +    if (ret < 0)
> >>>>>> +        return ret;
> >>>>>> +
> >>>>>> +    *readval = (unsigned int)value;
> >>>>>> +
> >>>>>> +    return 0;
> >>>>>> +}
> >>>>>> +
> >>>>>> +static const struct iio_info sca3300_info = {
> >>>>>> +    .read_raw = sca3300_read_raw,
> >>>>>> +    .write_raw = sca3300_write_raw,
> >>>>>> +    .debugfs_reg_access = &sca3300_debugfs_reg_access,
> >>>>>> +};
> >>>>>> +
> >>>>>> +static int sca3300_probe(struct spi_device *spi)
> >>>>>> +{
> >>>>>> +    struct sca3300_data *sca_data;
> >>>>>> +    struct iio_dev *indio_dev;
> >>>>>> +    int ret;
> >>>>>> +
> >>>>>> +    indio_dev = devm_iio_device_alloc(&spi->dev, sizeof(*sca_data));
> >>>>>> +    if (!indio_dev) {
> >>>>>> +        dev_err(&spi->dev,
> >>>>>> +            "failed to allocate memory for iio device\n");
> >>>>>> +        return -ENOMEM;
> >>>>>> +    }
> >>>>>> +
> >>>>>> +    sca_data = iio_priv(indio_dev);
> >>>>>> +    mutex_init(&sca_data->lock);
> >>>>>> +    sca_data->spi = spi;
> >>>>>> +    spi_set_drvdata(spi, indio_dev);  
> >>>>> Won't be used once using devm as below and remove() is dropped, so
> >>>>> drop this.
> >>>>>     
> >>>>>> +
> >>>>>> +    crc8_populate_msb(sca3300_crc_table, SCA3300_CRC8_POLYNOMIAL);
> >>>>>> +
> >>>>>> +    indio_dev->dev.parent = &spi->dev;
> >>>>>> +    indio_dev->info = &sca3300_info;
> >>>>>> +    indio_dev->name = SCA3300_ALIAS;
> >>>>>> +    indio_dev->modes = INDIO_DIRECT_MODE | INDIO_BUFFER_TRIGGERED;  
> >>>>> BUFFER_TRIGGERED bit now set by iio_triggered_buffer_setup() so
> >>>>> shouldn't
> >>>>> be ehre as well.
> >>>>>     
> >>>>>> +    indio_dev->channels = sca3300_channels;
> >>>>>> +    indio_dev->num_channels = ARRAY_SIZE(sca3300_channels);
> >>>>>> +    indio_dev->available_scan_masks = sca3300_scan_masks;
> >>>>>> +
> >>>>>> +    if (spi->dev.of_node) {
> >>>>>> +        ret = of_property_read_u32(spi->dev.of_node, "murata,opmode",  
> >>>>> Please use generic firmware access calls rather than the of specific
> >>>>> ones.
> >>>>> That lets us get ACPI support for free :)
> >>>>>
> >>>>> As per the binding review however, this one needs to go in favour of
> >>>>> userspace
> >>>>> control of scale + filter frequencies / sampling freq.
> >>>>>     
> >>>>>> + &sca_data->opmode);
> >>>>>> +        if (ret < 0)
> >>>>>> +            return ret;
> >>>>>> +    }
> >>>>>> +
> >>>>>> +    ret = sca3300_init(sca_data, indio_dev);
> >>>>>> +    if (ret < 0) {
> >>>>>> +        dev_err(&spi->dev, "failed to init device, error: %d\n",
> >>>>>> ret);
> >>>>>> +        return ret;
> >>>>>> +    }
> >>>>>> +
> >>>>>> +    ret = iio_triggered_buffer_setup(indio_dev,
> >>>>>> iio_pollfunc_store_time,
> >>>>>> +                     sca3300_trigger_handler, NULL);  
> >>>>> devm_iio_triggered_buffer_setup() and you can drop the manual
> >>>>> cleanup in remove()
> >>>>> and hence drop remove() entirely as nothing else left.
> >>>>>     
> >>>>>> +    if (ret < 0) {
> >>>>>> +        dev_err(&spi->dev,
> >>>>>> +            "iio triggered buffer setup failed, error: %d\n", ret);
> >>>>>> +        return ret;
> >>>>>> +    }
> >>>>>> +
> >>>>>> +    ret = devm_iio_device_register(&spi->dev, indio_dev);
> >>>>>> +    if (ret < 0) {
> >>>>>> +        dev_err(&spi->dev, "iio device register failed, error: %d\n",
> >>>>>> +            ret);
> >>>>>> +        iio_triggered_buffer_cleanup(indio_dev);  
> >>>>> With devm above, no need to manually clean this up.
> >>>>>     
> >>>>>> +        return ret;
> >>>>>> +    }
> >>>>>> +
> >>>>>> +    return 0;
> >>>>>> +}
> >>>>>> +
> >>>>>> +static int sca3300_remove(struct spi_device *spi)
> >>>>>> +{
> >>>>>> +    struct iio_dev *indio_dev = spi_get_drvdata(spi);
> >>>>>> +
> >>>>>> +    iio_triggered_buffer_cleanup(indio_dev);
> >>>>>> +    return 0;
> >>>>>> +}
> >>>>>> +
> >>>>>> +static const struct of_device_id sca3300_dt_ids[] = {
> >>>>>> +    { .compatible = "murata,sca3300"},
> >>>>>> +    {},
> >>>>>> +};
> >>>>>> +MODULE_DEVICE_TABLE(of, sca3300_dt_ids);
> >>>>>> +
> >>>>>> +static struct spi_driver sca3300_driver = {
> >>>>>> +    .driver = {
> >>>>>> +        .name        = SCA3300_ALIAS,
> >>>>>> +        .owner        = THIS_MODULE,
> >>>>>> +        .of_match_table = of_match_ptr(sca3300_dt_ids),
> >>>>>> +    },
> >>>>>> +
> >>>>>> +    .probe    = sca3300_probe,
> >>>>>> +    .remove    = sca3300_remove,
> >>>>>> +};
> >>>>>> +
> >>>>>> +module_spi_driver(sca3300_driver);
> >>>>>> +
> >>>>>> +MODULE_AUTHOR("Tomas Melin <[e-mail]>");
> >>>>>> +MODULE_DESCRIPTION("Murata SCA3300 SPI Accelerometer");
> >>>>>> +MODULE_LICENSE("GPL v2");

Jonathan Cameron April 19, 2021, 8:16 p.m. UTC | #8

On Mon, 19 Apr 2021 20:45:08 +0100
Jonathan Cameron <[e-mail]> wrote:

> On Mon, 19 Apr 2021 16:34:35 +0300
> Tomas Melin <[e-mail]> wrote:
> 
> > Hi,
> > 
> > On 4/18/21 1:03 PM, Jonathan Cameron wrote:  
> > > On Fri, 16 Apr 2021 15:21:14 +0300
> > > Tomas Melin <[e-mail]> wrote:
> > >    
> > >> Updated email-address for Alexandru.
> > >>
> > >>
> > >> On 4/16/21 3:17 PM, Tomas Melin wrote:    
> > >>> On 4/15/21 11:41 AM, Tomas Melin wrote:
> > >>>       
> > >>>> While working on updates I did notice something new which I cannot
> > >>>>
> > >>>> reproduce on older (5.10.17 kernel) version. If compiling this as a
> > >>>> module, getting error while
> > >>>>
> > >>>> unloading module:
> > >>>>
> > >>>> [   40.200084] Unable to handle kernel NULL pointer dereference at
> > >>>> virtual address 00000104
> > >>>> ...
> > >>>>
> > >>>> [   40.510054] Backtrace:
> > >>>> [   40.512502] [<c06c5e0c>] (iio_device_ioctl_handler_unregister)
> > >>>> from [<c06ca8c4>] (iio_buffers_free_sysfs_and_mask+0x2c/0x6c)
> > >>>> [   40.523735] [<c06ca898>] (iio_buffers_free_sysfs_and_mask) from
> > >>>> [<c06c4a18>] (iio_device_unregister+0xa8/0xac)
> > >>>> [   40.533746]  r5:c1811228 r4:c1811000
> > >>>> [   40.537318] [<c06c4970>] (iio_device_unregister) from [<c06c4a38>]
> > >>>> (devm_iio_device_unreg+0x1c/0x20)
> > >>>> [   40.546461]  r5:c2415000 r4:c25bab80
> > >>>> [   40.550025] [<c06c4a1c>] (devm_iio_device_unreg) from [<c0556820>]
> > >>>> (release_nodes+0x1c0/0x1f0)
> > >>>> [   40.558654] [<c0556660>] (release_nodes) from [<c0556bb0>]
> > >>>> (devres_release_all+0x40/0x60)
> > >>>> [   40.566847]  r10:00000081 r9:c2350000 r8:c0100264 r7:00000081
> > >>>> r6:bf00c010 r5:c19be000
> > >>>> [   40.574669]  r4:c1a91c00
> > >>>> [   40.577194] [<c0556b70>] (devres_release_all) from [<c055216c>]
> > >>>> (device_release_driver_internal+0x120/0x1cc)
> > >>>> [   40.587031]  r5:c19be000 r4:c1a91c00
> > >>>> [   40.590596] [<c055204c>] (device_release_driver_internal) from
> > >>>> [<c05522b4>] (driver_detach+0x54/0x90)
> > >>>> [   40.599828]  r7:00000081 r6:00000000 r5:bf00c010 r4:c1a91c00
> > >>>> [   40.605482] [<c0552260>] (driver_detach) from [<c0550ba4>]
> > >>>> (bus_remove_driver+0x5c/0xb0)
> > >>>> [   40.613583]  r5:00000800 r4:bf00c010
> > >>>> [   40.617148] [<c0550b48>] (bus_remove_driver) from [<c0552c38>]
> > >>>> (driver_unregister+0x38/0x5c)
> > >>>> [   40.625596]  r5:00000800 r4:bf00c010
> > >>>> [   40.629161] [<c0552c00>] (driver_unregister) from [<bf00a760>]
> > >>>> (sca3300_driver_exit+0x14/0x8b4 [sca3300])
> > >>>> [   40.638747]  r5:00000800 r4:bf00c080
> > >>>> [   40.642311] [<bf00a74c>] (sca3300_driver_exit [sca3300]) from
> > >>>> [<c01c06f0>] (sys_delete_module+0x16c/0x238)
> > >>>> [   40.651990] [<c01c0584>] (sys_delete_module) from [<c0100244>]
> > >>>> (__sys_trace_return+0x0/0x1c)
> > >>>> [   40.660435] Exception stack(0xc2351fa8 to 0xc2351ff0)
> > >>>> [   40.665484] 1fa0:                   0050e5a8 00000000 0050e5e4
> > >>>> 00000800 081d4b00 bec18af4
> > >>>> [   40.673661] 1fc0: 0050e5a8 00000000 bec18b50 00000081 bec18e51
> > >>>> 0050e190 00000001 bec18d3c
> > >>>> [   40.681834] 1fe0: 0050cf70 bec18afc 004f1ec8 b6ecb27c
> > >>>> [   40.686887]  r6:bec18b50 r5:00000000 r4:0050e5a8
> > >>>> [   40.691507] Code: e8bd4000 e1c020d0 e3a0cc01 e3001122 (e5823004)
> > >>>> [   40.707675] ---[ end trace 189882b050077333 ]---
> > >>>>
> > >>>> This happens when building against linux-next
> > >>>> 5.12.0-rc6-next-20210409. I'm failing to see what is wrong. Any ideas?
> > >>>>
> > >>>> Thanks,
> > >>>>
> > >>>> Tomas    
> > >>>
> > >>> Tested further that for this driver, loading and unloading as module
> > >>> works fine until commit:
> > >>>
> > >>> commit f73f7f4da581875f9b1f2fb8ebd1ab15ed634488
> > >>> Author: Alexandru Ardelean <[e-mail]>
> > >>> Date:   Mon Feb 15 12:40:39 2021 +0200
> > >>>
> > >>>      iio: buffer: add ioctl() to support opening extra buffers for IIO
> > >>> device
> > >>>
> > >>>
> > >>> Any thoughts what causes this issue?    
> > > Is this still happening after fixing the ordering in probe / remove?
> > > (devm_iio_triggered_buffer_setup() being easiest way)
> > >
> > > As driver currently stands it's calling iio_triggered_buffer_cleanup
> > > before the managed cleanup of the iio_device_register() call.  That
> > > should never happen so would be where I'd look for problems here.
> > >
> > > It is possible Alex's work is relying on that ordering being correct
> > > a little more than we previously were. I could be wrong though
> > > and could be something else going on!
> > >
> > > Jonathan
> > >
> > > +CC Alex's other email address.    
> > 
> > This is still happening with v3 of driver patchset, which uses only devm_*
> > 
> > versions in probing. v5.12-rc7 works ok, I'm currently only seeing this 
> > using linux-next.  
> 
> Thanks for checking.  I've just fired up a VM and confirmed that the same
> thing is happening with creating and removing instances of the dummy driver.
> 
> Will see if I can figure out what's going wrong.

So the issue proved reasonably easy to track down.
We remove the buffer ioctl handler twice.  First in iio_device_unregister() in a loop
over the list, and again later in iio_buffer_sysfs_and_mask() (called from
the same function). (note it would happen later in the equivalent event related
version but we have already died on the first instance which is the one Tomas hit).

@Alex.  Which one should we keep?  Given the buffer and event code both
clean up manually I think what we want is:

@@ -1939,9 +1943,6 @@ void iio_device_unregister(struct iio_dev *indio_dev)
 
        indio_dev->info = NULL;
 
-       list_for_each_entry_safe(h, t, &iio_dev_opaque->ioctl_handlers, entry)
-               list_del(&h->entry);
-
        iio_device_wakeup_eventset(indio_dev);
        iio_buffer_wakeup_poll(indio_dev);
 

I'm going to guess something went wrong in a merge or rebased and left that
block when it should have been removed.

Anyhow, if the above is the right fix I can roll it up into a patch tomorrow and
get it into linux-next at least.

thanks,

Jonathan

> 
> Jonathan
> 
> > 
> > Thanks,
> > 
> > Tomas
> > 
> >   
> > >>> Thanks,
> > >>>
> > >>> Tomas
> > >>>
> > >>>
> > >>>       
> > >>>>       
> > >>>>>       
> > >>>>>> ---
> > >>>>>>    drivers/iio/accel/Kconfig   |  13 ++
> > >>>>>>    drivers/iio/accel/Makefile  |   1 +
> > >>>>>>    drivers/iio/accel/sca3300.c | 434
> > >>>>>> ++++++++++++++++++++++++++++++++++++
> > >>>>>>    3 files changed, 448 insertions(+)
> > >>>>>>    create mode 100644 drivers/iio/accel/sca3300.c
> > >>>>>>
> > >>>>>> diff --git a/drivers/iio/accel/Kconfig b/drivers/iio/accel/Kconfig
> > >>>>>> index cceda3cecbcf..0dbf7b648e8a 100644
> > >>>>>> --- a/drivers/iio/accel/Kconfig
> > >>>>>> +++ b/drivers/iio/accel/Kconfig
> > >>>>>> @@ -450,6 +450,19 @@ config SCA3000
> > >>>>>>          To compile this driver as a module, say M here: the module
> > >>>>>> will be
> > >>>>>>          called sca3000.
> > >>>>>>    +config SCA3300
> > >>>>>> +    tristate "Murata SCA3300 3-Axis Accelerometer Driver"
> > >>>>>> +    depends on SPI
> > >>>>>> +    select CRC8
> > >>>>>> +    select IIO_BUFFER
> > >>>>>> +    select IIO_TRIGGERED_BUFFER
> > >>>>>> +    help
> > >>>>>> +      Say yes here to build support for Murata SCA3300 3-Axis
> > >>>>>> +      accelerometer.
> > >>>>>> +
> > >>>>>> +      To compile this driver as a module, choose M here: the
> > >>>>>> module will be
> > >>>>>> +      called sca3300.
> > >>>>>> +
> > >>>>>>    config STK8312
> > >>>>>>        tristate "Sensortek STK8312 3-Axis Accelerometer Driver"
> > >>>>>>        depends on I2C
> > >>>>>> diff --git a/drivers/iio/accel/Makefile b/drivers/iio/accel/Makefile
> > >>>>>> index 32cd1342a31a..4b56527a2b97 100644
> > >>>>>> --- a/drivers/iio/accel/Makefile
> > >>>>>> +++ b/drivers/iio/accel/Makefile
> > >>>>>> @@ -50,6 +50,7 @@ obj-$(CONFIG_MXC4005)        += mxc4005.o
> > >>>>>>    obj-$(CONFIG_MXC6255)        += mxc6255.o
> > >>>>>>      obj-$(CONFIG_SCA3000)        += sca3000.o
> > >>>>>> +obj-$(CONFIG_SCA3300)        += sca3300.o
> > >>>>>>      obj-$(CONFIG_STK8312)        += stk8312.o
> > >>>>>>    obj-$(CONFIG_STK8BA50)        += stk8ba50.o
> > >>>>>> diff --git a/drivers/iio/accel/sca3300.c b/drivers/iio/accel/sca3300.c
> > >>>>>> new file mode 100644
> > >>>>>> index 000000000000..112fb88ecd3a
> > >>>>>> --- /dev/null
> > >>>>>> +++ b/drivers/iio/accel/sca3300.c
> > >>>>>> @@ -0,0 +1,434 @@
> > >>>>>> +// SPDX-License-Identifier: GPL-2.0-only
> > >>>>>> +/*
> > >>>>>> + * Copyright (c) Vaisala Oyj. All rights reserved.    
> > >>>>> Give a year for the copyright notice if you can.
> > >>>>>       
> > >>>>>> + */
> > >>>>>> +#include <linux/crc8.h>
> > >>>>>> +#include <linux/delay.h>
> > >>>>>> +#include <linux/iio/buffer.h>
> > >>>>>> +#include <linux/iio/iio.h>
> > >>>>>> +#include <linux/iio/sysfs.h>
> > >>>>>> +#include <linux/iio/trigger_consumer.h>
> > >>>>>> +#include <linux/iio/triggered_buffer.h>
> > >>>>>> +#include <linux/kernel.h>
> > >>>>>> +#include <linux/module.h>
> > >>>>>> +#include <linux/spi/spi.h>
> > >>>>>> +
> > >>>>>> +#define SCA3300_ALIAS "sca3300"
> > >>>>>> +
> > >>>>>> +#define SCA3300_REG_STATUS 0x6
> > >>>>>> +#define SCA3300_REG_MODE 0xd
> > >>>>>> +#define SCA3300_REG_WHOAMI 0x10
> > >>>>>> +#define SCA3300_VALUE_SW_RESET 0x20
> > >>>>>> +#define SCA3300_CRC8_POLYNOMIAL 0x1d
> > >>>>>> +#define SCA3300_X_READ 0    
> > >>>>> I wouldn't bother defining this.
> > >>>>>       
> > >>>>>> +#define SCA3300_X_WRITE BIT(7)    
> > >>>>> Even this one is something I'd just put inline with a comment.
> > >>>>>       
> > >>>>>> +#define SCA3300_DEVICE_ID 0x51
> > >>>>>> +#define SCA3300_RS_ERROR 0x3
> > >>>>>> +
> > >>>>>> +enum sca3300_scan_indexes {
> > >>>>>> +    SCA3300_ACC_X = 0,
> > >>>>>> +    SCA3300_ACC_Y,
> > >>>>>> +    SCA3300_ACC_Z,
> > >>>>>> +    SCA3300_TEMP,
> > >>>>>> +    SCA3300_TIMESTAMP,
> > >>>>>> +};
> > >>>>>> +
> > >>>>>> +#define SCA3300_ACCEL_CHANNEL(index, reg, axis) { \
> > >>>>>> +        .type = IIO_ACCEL,                    \
> > >>>>>> +        .address = reg,                        \
> > >>>>>> +        .modified = 1,                        \
> > >>>>>> +        .channel2 = IIO_MOD_##axis,                \
> > >>>>>> +        .info_mask_separate = BIT(IIO_CHAN_INFO_RAW) | \
> > >>>>>> +                      BIT(IIO_CHAN_INFO_PROCESSED),    \    
> > >>>>> As mentioned below, don't provide PROCESSED. Userspace is better at
> > >>>>> handling the
> > >>>>> conversion so leave it to them.
> > >>>>>       
> > >>>>>> +        .info_mask_shared_by_type = BIT(IIO_CHAN_INFO_SCALE),    \
> > >>>>>> +        .scan_index = index,                    \
> > >>>>>> +        .scan_type = {                        \
> > >>>>>> +            .sign = 's',                    \
> > >>>>>> +            .realbits = 16,                    \
> > >>>>>> +            .storagebits = 16,                \
> > >>>>>> +            .shift = 0,                    \
> > >>>>>> +            .endianness = IIO_CPU,                \
> > >>>>>> +        },                            \
> > >>>>>> +    }
> > >>>>>> +
> > >>>>>> +static const struct iio_chan_spec sca3300_channels[] = {
> > >>>>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_X, 0x1, X),
> > >>>>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Y, 0x2, Y),
> > >>>>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Z, 0x3, Z),
> > >>>>>> +    {
> > >>>>>> +        .type = IIO_TEMP,
> > >>>>>> +        .address = 0x5,
> > >>>>>> +        .scan_index = SCA3300_TEMP,
> > >>>>>> +        .info_mask_separate = BIT(IIO_CHAN_INFO_RAW),
> > >>>>>> +        .scan_type = {
> > >>>>>> +            .sign = 's',
> > >>>>>> +            .realbits = 16,
> > >>>>>> +            .storagebits = 16,
> > >>>>>> +            .shift = 0,
> > >>>>>> +            .endianness = IIO_CPU,
> > >>>>>> +        },
> > >>>>>> +    },
> > >>>>>> +    IIO_CHAN_SOFT_TIMESTAMP(4),
> > >>>>>> +};
> > >>>>>> +
> > >>>>>> +static const int sca3300_accel_scale[] = {2700, 1350, 5400, 5400};
> > >>>>>> +
> > >>>>>> +static const unsigned long sca3300_scan_masks[] = {
> > >>>>>> +    BIT(SCA3300_ACC_X) | BIT(SCA3300_ACC_Y) | BIT(SCA3300_ACC_Z) |
> > >>>>>> +    BIT(SCA3300_TEMP),
> > >>>>>> +    0};
> > >>>>>> +
> > >>>>>> +/**
> > >>>>>> + * SCA3300 device data    
> > >>>>> run scripts/kernel-doc over the file and fix the warnings + errors.
> > >>>>>       
> > >>>>>> + *
> > >>>>>> + * @spi SPI device structure
> > >>>>>> + * @opmode Device operation mode
> > >>>>>> + * @lock Data buffer lock
> > >>>>>> + * @txbuf Transmit buffer
> > >>>>>> + * @rxbuf Receive buffer
> > >>>>>> + * @scan Triggered buffer. Four channel 16-bit data + 64-bit
> > >>>>>> timestamp
> > >>>>>> + */
> > >>>>>> +struct sca3300_data {
> > >>>>>> +    struct spi_device *spi;
> > >>>>>> +    u32 opmode;
> > >>>>>> +    struct mutex lock;
> > >>>>>> +    u8 txbuf[4];
> > >>>>>> +    u8 rxbuf[4];
> > >>>>>> +    struct {
> > >>>>>> +        s16 channels[4];
> > >>>>>> +        s64 ts __aligned(sizeof(s64));
> > >>>>>> +    } scan;
> > >>>>>> +};
> > >>>>>> +
> > >>>>>> +DECLARE_CRC8_TABLE(sca3300_crc_table);
> > >>>>>> +
> > >>>>>> +static int sca3300_transfer(struct sca3300_data *sca_data, int *val)
> > >>>>>> +{
> > >>>>>> +    struct spi_delay delay = {.value = 10, .unit =
> > >>>>>> SPI_DELAY_UNIT_USECS};
> > >>>>>> +    int32_t ret;
> > >>>>>> +    int rs;
> > >>>>>> +    u8 crc;
> > >>>>>> +    struct spi_transfer xfers[2] = {
> > >>>>>> +        {
> > >>>>>> +            .tx_buf = sca_data->txbuf,
> > >>>>>> +            .rx_buf = NULL,
> > >>>>>> +            .len = ARRAY_SIZE(sca_data->txbuf),
> > >>>>>> +            .delay = delay,
> > >>>>>> +            .cs_change = 1,
> > >>>>>> +        },
> > >>>>>> +        {
> > >>>>>> +            .tx_buf = NULL,
> > >>>>>> +            .rx_buf = sca_data->rxbuf,
> > >>>>>> +            .len = ARRAY_SIZE(sca_data->rxbuf),
> > >>>>>> +            .delay = delay,
> > >>>>>> +            .cs_change = 0,
> > >>>>>> +        }
> > >>>>>> +    };
> > >>>>>> +
> > >>>>>> +    /* inverted crc value as described in device data sheet */
> > >>>>>> +    crc = ~crc8(sca3300_crc_table, &sca_data->txbuf[0], 3,
> > >>>>>> CRC8_INIT_VALUE);
> > >>>>>> +    sca_data->txbuf[3] = crc;
> > >>>>>> +
> > >>>>>> +    ret = spi_sync_transfer(sca_data->spi, xfers, 2);    
> > >>>>> Use ARRAY_SIZE(xfers) instead of 2.
> > >>>>>       
> > >>>>>> +    if (ret < 0) {
> > >>>>>> +        dev_err(&sca_data->spi->dev,
> > >>>>>> +            "transfer error, error: %d\n", ret);
> > >>>>>> +        return -EIO;
> > >>>>>> +    }
> > >>>>>> +
> > >>>>>> +    crc = ~crc8(sca3300_crc_table, &sca_data->rxbuf[0], 3,
> > >>>>>> CRC8_INIT_VALUE);
> > >>>>>> +    if (sca_data->rxbuf[3] != crc) {
> > >>>>>> +        dev_err(&sca_data->spi->dev, "CRC checksum mismatch");
> > >>>>>> +        return -EIO;
> > >>>>>> +    }
> > >>>>>> +
> > >>>>>> +    /* get return status */
> > >>>>>> +    rs = sca_data->rxbuf[0] & 0x03;
> > >>>>>> +    if (rs == SCA3300_RS_ERROR)
> > >>>>>> +        return rs;
> > >>>>>> +
> > >>>>>> +    *val = (s16)(sca_data->rxbuf[2] | (sca_data->rxbuf[1] << 8));    
> > >>>>> Preference for an unaligned endian conversion here and explicit sign
> > >>>>> extend e.g
> > >>>>> something like
> > >>>>>
> > >>>>> sign_extend32(get_unaligned_le16(&sca_data->rxbuf[1]), 15)
> > >>>>>       
> > >>>>>> +
> > >>>>>> +    return 0;
> > >>>>>> +}
> > >>>>>> +
> > >>>>>> +static int sca3300_read_reg(struct sca3300_data *sca_data, u8 reg,
> > >>>>>> int *val)
> > >>>>>> +{
> > >>>>>> +    int ret;
> > >>>>>> +
> > >>>>>> +    mutex_lock(&sca_data->lock);
> > >>>>>> +    sca_data->txbuf[0] = SCA3300_X_READ | (reg << 2);
> > >>>>>> +    ret = sca3300_transfer(sca_data, val);
> > >>>>>> +    if (ret > 0) {
> > >>>>>> +        sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS
> > >>>>>> << 2);
> > >>>>>> +        ret = sca3300_transfer(sca_data, val);
> > >>>>>> +        /* status 0 = startup, 0x2 = mode change */
> > >>>>>> +        if (ret > 0 && *val != 0 && *val != 0x2) {
> > >>>>>> + dev_err_ratelimited(&sca_data->spi->dev,
> > >>>>>> +                        "device status: %x\n",
> > >>>>>> +                        (u16)*val);
> > >>>>>> +            mutex_unlock(&sca_data->lock);
> > >>>>>> +            return -EIO;
> > >>>>>> +        }
> > >>>>>> +        if (ret > 0)
> > >>>>>> +            ret = 0;
> > >>>>>> +    }
> > >>>>>> +    mutex_unlock(&sca_data->lock);
> > >>>>>> +
> > >>>>>> +    return ret;
> > >>>>>> +}
> > >>>>>> +
> > >>>>>> +static int sca3300_write_reg(struct sca3300_data *sca_data, u8
> > >>>>>> reg, int val)
> > >>>>>> +{
> > >>>>>> +    int reg_val = 0;
> > >>>>>> +    int ret;
> > >>>>>> +
> > >>>>>> +    mutex_lock(&sca_data->lock);
> > >>>>>> +    sca_data->txbuf[0] = SCA3300_X_WRITE | (reg << 2);
> > >>>>>> +    sca_data->txbuf[1] = val >> 8;
> > >>>>>> +    sca_data->txbuf[2] = val & 0xFF;    
> > >>>>> Prefer (Slightly) an unaligned put.
> > >>>>>       
> > >>>>>> +    ret = sca3300_transfer(sca_data, &reg_val);
> > >>>>>> +    if (ret > 0) {    
> > >>>>> Factor this error handling out to another function, plus trigger it
> > >>>>> from an
> > >>>>> appropriate standard error code rather than a positive return value.
> > >>>>>       
> > >>>>>> +        sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS
> > >>>>>> << 2);
> > >>>>>> +        ret = sca3300_transfer(sca_data, &reg_val);
> > >>>>>> +        /* status 0 = startup, 0x2 = mode change */
> > >>>>>> +        if (ret > 0 && reg_val != 0 && reg_val != 0x2) {
> > >>>>>> + dev_err_ratelimited(&sca_data->spi->dev,
> > >>>>>> +                        "device status: %x\n",
> > >>>>>> +                        (u16)reg_val);
> > >>>>>> +            mutex_unlock(&sca_data->lock);
> > >>>>>> +            return -EIO;
> > >>>>>> +        }
> > >>>>>> +        if (ret > 0)
> > >>>>>> +            ret = 0;    
> > >>>>> Whenever you get this sort of juggling it rather implies your design is
> > >>>>> less than ideal.
> > >>>>>       
> > >>>>>> +    }
> > >>>>>> +    mutex_unlock(&sca_data->lock);
> > >>>>>> +
> > >>>>>> +    return ret;
> > >>>>>> +}
> > >>>>>> +
> > >>>>>> +static int sca3300_write_raw(struct iio_dev *indio_dev,
> > >>>>>> +                 struct iio_chan_spec const *chan,
> > >>>>>> +                 int val, int val2, long mask)
> > >>>>>> +{
> > >>>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
> > >>>>>> +
> > >>>>>> +    switch (mask) {
> > >>>>>> +    case IIO_CHAN_INFO_SCALE:
> > >>>>>> +        if (val < 0 || val > 3)
> > >>>>>> +            return -EINVAL;    
> > >>>>> No.  Scale values need to be the value of 1LSB not a random integer
> > >>>>> that requires you to look up the datasheet.
> > >>>>>       
> > >>>>>> +        return sca3300_write_reg(data, SCA3300_REG_MODE, val);
> > >>>>>> +    default:
> > >>>>>> +        return -EINVAL;
> > >>>>>> +    }
> > >>>>>> +}
> > >>>>>> +
> > >>>>>> +static int sca3300_read_raw(struct iio_dev *indio_dev,
> > >>>>>> +                struct iio_chan_spec const *chan,
> > >>>>>> +                int *val, int *val2, long mask)
> > >>>>>> +{
> > >>>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
> > >>>>>> +    int ret;
> > >>>>>> +    int reg_val;
> > >>>>>> +
> > >>>>>> +    switch (mask) {
> > >>>>>> +    case IIO_CHAN_INFO_RAW:
> > >>>>>> +        ret = sca3300_read_reg(data, chan->address, val);
> > >>>>>> +        if (ret < 0)
> > >>>>>> +            return ret;
> > >>>>>> +        return IIO_VAL_INT;
> > >>>>>> +    case IIO_CHAN_INFO_SCALE:
> > >>>>>> +        ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
> > >>>>>> +        if (ret < 0)
> > >>>>>> +            return ret;
> > >>>>>> +        *val = sca3300_accel_scale[reg_val];    
> > >>>>> Scale is very very rarely an integer value which makes me wonder what
> > >>>>> you are returning here...  It should be the multiplier needed to
> > >>>>> take the raw value to a reading in m/sec^2  Given you use it as
> > >>>>> fractional
> > >>>>> below, I'm guessing this is 1/scale_value?
> > >>>>>       
> > >>>>>> +        return IIO_VAL_INT;
> > >>>>>> +    case IIO_CHAN_INFO_PROCESSED:    
> > >>>>> Don't provide processed here.   Userspace needs to do the conversion
> > >>>>> itself
> > >>>>> (and given it has floating point easily available will probably do a
> > >>>>> better
> > >>>>>    job than we can).   I'm assuming the mode can't autonomously change?
> > >>>>>       
> > >>>>>> +        ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
> > >>>>>> +        if (ret < 0)
> > >>>>>> +            return ret;
> > >>>>>> +        *val2 = sca3300_accel_scale[reg_val];
> > >>>>>> +        ret = sca3300_read_reg(data, chan->address, val);
> > >>>>>> +        if (ret < 0)
> > >>>>>> +            return ret;
> > >>>>>> +        return IIO_VAL_FRACTIONAL;
> > >>>>>> +    default:
> > >>>>>> +        return -EINVAL;
> > >>>>>> +    }
> > >>>>>> +}
> > >>>>>> +
> > >>>>>> +static irqreturn_t sca3300_trigger_handler(int irq, void *p)
> > >>>>>> +{
> > >>>>>> +    struct iio_poll_func *pf = p;
> > >>>>>> +    struct iio_dev *indio_dev = pf->indio_dev;
> > >>>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
> > >>>>>> +    s64 time_ns = iio_get_time_ns(indio_dev);    
> > >>>>> Is the timestamp at the start more accurate that that at the end
> > >>>>> of reading the channels?  If not, just put this inline in the
> > >>>>> iio_push_to_buffers_with_timestamp() call.
> > >>>>>       
> > >>>>>> +    int bit, ret, val, i = 0;
> > >>>>>> +
> > >>>>>> +    for_each_set_bit(bit, indio_dev->active_scan_mask,
> > >>>>>> +             indio_dev->masklength) {
> > >>>>>> +        ret = sca3300_read_reg(data, sca3300_channels[bit].address,
> > >>>>>> +                       &val);
> > >>>>>> +        if (ret < 0)    
> > >>>>> Given we can't return an error code from interrupt handlers, it is
> > >>>>> usually
> > >>>>> a good idea to print something to the log. Otherwise we get missing
> > >>>>> data
> > >>>>> with no idea of why..
> > >>>>>       
> > >>>>>> +            goto out;
> > >>>>>> +        if (ARRAY_SIZE(data->scan.channels) > i)    
> > >>>>> How could this not be true?  If it's always true, don't bother
> > >>>>> checking i.
> > >>>>>       
> > >>>>>> +            ((s16 *)data->scan.channels)[i++] = val;
> > >>>>>> +    }
> > >>>>>> +
> > >>>>>> +    iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
> > >>>>>> time_ns);
> > >>>>>> +out:
> > >>>>>> +    iio_trigger_notify_done(indio_dev->trig);
> > >>>>>> +
> > >>>>>> +    return IRQ_HANDLED;
> > >>>>>> +}
> > >>>>>> +
> > >>>>>> +static int sca3300_init(struct sca3300_data *sca_data,
> > >>>>>> +            struct iio_dev *indio_dev)
> > >>>>>> +{
> > >>>>>> +    int ret;
> > >>>>>> +    int value = 0;
> > >>>>>> +
> > >>>>>> +    if (sca_data->opmode < 1 || sca_data->opmode > 4)
> > >>>>>> +        return -EINVAL;    
> > >>>>> Whilst this is going anyway, better to check that next to where
> > >>>>> it is read so we drop out immediately rather than buried in this
> > >>>>> function.
> > >>>>>       
> > >>>>>> +
> > >>>>>> +    ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
> > >>>>>> +                SCA3300_VALUE_SW_RESET);
> > >>>>>> +    if (ret != 0)
> > >>>>>> +        return ret;    
> > >>>>> For sleeps, good to reference which section in data sheet gives the
> > >>>>> timing.
> > >>>>>       
> > >>>>>> +    usleep_range(2e3, 10e3);
> > >>>>>> +
> > >>>>>> +    ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
> > >>>>>> +                sca_data->opmode - 1);
> > >>>>>> +    if (ret != 0)    
> > >>>>> if (ret) see below.
> > >>>>>       
> > >>>>>> +        return ret;
> > >>>>>> +    msleep(100);    
> > >>>>> Again, document where the time comes from. Saves reviewers time if
> > >>>>> they want to check it. (I'm too lazy :)
> > >>>>>       
> > >>>>>> +    ret = sca3300_read_reg(sca_data, SCA3300_REG_WHOAMI, &value);
> > >>>>>> +    if (ret != 0)    
> > >>>>> if (ret) is more idiomatic in kernel code.
> > >>>>>       
> > >>>>>> +        return ret;
> > >>>>>> +
> > >>>>>> +    if (value != SCA3300_DEVICE_ID) {
> > >>>>>> +        dev_err(&sca_data->spi->dev, "device id not expected
> > >>>>>> value\n");    
> > >>>>> Perhaps useful to print what was expected and what was seen. Maybe it's
> > >>>>> a new variant and this might help the user to identify that and add
> > >>>>> it to the
> > >>>>> driver.
> > >>>>>       
> > >>>>>> +        return -EIO;    
> > >>>>> -EINVAL probably more appropriate return.
> > >>>>>       
> > >>>>>> +    }
> > >>>>>> +    return 0;
> > >>>>>> +}
> > >>>>>> +
> > >>>>>> +static int sca3300_debugfs_reg_access(struct iio_dev *indio_dev,
> > >>>>>> +                      unsigned int reg, unsigned int writeval,
> > >>>>>> +                      unsigned int *readval)
> > >>>>>> +{
> > >>>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
> > >>>>>> +    int value;
> > >>>>>> +    int ret;
> > >>>>>> +
> > >>>>>> +    if (reg > 0x1f)    
> > >>>>> Use a define that names that register.
> > >>>>>       
> > >>>>>> +        return -EINVAL;
> > >>>>>> +
> > >>>>>> +    if (!readval)
> > >>>>>> +        return sca3300_write_reg(data, reg, writeval);
> > >>>>>> +
> > >>>>>> +    ret = sca3300_read_reg(data, reg, &value);
> > >>>>>> +    if (ret < 0)
> > >>>>>> +        return ret;
> > >>>>>> +
> > >>>>>> +    *readval = (unsigned int)value;
> > >>>>>> +
> > >>>>>> +    return 0;
> > >>>>>> +}
> > >>>>>> +
> > >>>>>> +static const struct iio_info sca3300_info = {
> > >>>>>> +    .read_raw = sca3300_read_raw,
> > >>>>>> +    .write_raw = sca3300_write_raw,
> > >>>>>> +    .debugfs_reg_access = &sca3300_debugfs_reg_access,
> > >>>>>> +};
> > >>>>>> +
> > >>>>>> +static int sca3300_probe(struct spi_device *spi)
> > >>>>>> +{
> > >>>>>> +    struct sca3300_data *sca_data;
> > >>>>>> +    struct iio_dev *indio_dev;
> > >>>>>> +    int ret;
> > >>>>>> +
> > >>>>>> +    indio_dev = devm_iio_device_alloc(&spi->dev, sizeof(*sca_data));
> > >>>>>> +    if (!indio_dev) {
> > >>>>>> +        dev_err(&spi->dev,
> > >>>>>> +            "failed to allocate memory for iio device\n");
> > >>>>>> +        return -ENOMEM;
> > >>>>>> +    }
> > >>>>>> +
> > >>>>>> +    sca_data = iio_priv(indio_dev);
> > >>>>>> +    mutex_init(&sca_data->lock);
> > >>>>>> +    sca_data->spi = spi;
> > >>>>>> +    spi_set_drvdata(spi, indio_dev);    
> > >>>>> Won't be used once using devm as below and remove() is dropped, so
> > >>>>> drop this.
> > >>>>>       
> > >>>>>> +
> > >>>>>> +    crc8_populate_msb(sca3300_crc_table, SCA3300_CRC8_POLYNOMIAL);
> > >>>>>> +
> > >>>>>> +    indio_dev->dev.parent = &spi->dev;
> > >>>>>> +    indio_dev->info = &sca3300_info;
> > >>>>>> +    indio_dev->name = SCA3300_ALIAS;
> > >>>>>> +    indio_dev->modes = INDIO_DIRECT_MODE | INDIO_BUFFER_TRIGGERED;    
> > >>>>> BUFFER_TRIGGERED bit now set by iio_triggered_buffer_setup() so
> > >>>>> shouldn't
> > >>>>> be ehre as well.
> > >>>>>       
> > >>>>>> +    indio_dev->channels = sca3300_channels;
> > >>>>>> +    indio_dev->num_channels = ARRAY_SIZE(sca3300_channels);
> > >>>>>> +    indio_dev->available_scan_masks = sca3300_scan_masks;
> > >>>>>> +
> > >>>>>> +    if (spi->dev.of_node) {
> > >>>>>> +        ret = of_property_read_u32(spi->dev.of_node, "murata,opmode",    
> > >>>>> Please use generic firmware access calls rather than the of specific
> > >>>>> ones.
> > >>>>> That lets us get ACPI support for free :)
> > >>>>>
> > >>>>> As per the binding review however, this one needs to go in favour of
> > >>>>> userspace
> > >>>>> control of scale + filter frequencies / sampling freq.
> > >>>>>       
> > >>>>>> + &sca_data->opmode);
> > >>>>>> +        if (ret < 0)
> > >>>>>> +            return ret;
> > >>>>>> +    }
> > >>>>>> +
> > >>>>>> +    ret = sca3300_init(sca_data, indio_dev);
> > >>>>>> +    if (ret < 0) {
> > >>>>>> +        dev_err(&spi->dev, "failed to init device, error: %d\n",
> > >>>>>> ret);
> > >>>>>> +        return ret;
> > >>>>>> +    }
> > >>>>>> +
> > >>>>>> +    ret = iio_triggered_buffer_setup(indio_dev,
> > >>>>>> iio_pollfunc_store_time,
> > >>>>>> +                     sca3300_trigger_handler, NULL);    
> > >>>>> devm_iio_triggered_buffer_setup() and you can drop the manual
> > >>>>> cleanup in remove()
> > >>>>> and hence drop remove() entirely as nothing else left.
> > >>>>>       
> > >>>>>> +    if (ret < 0) {
> > >>>>>> +        dev_err(&spi->dev,
> > >>>>>> +            "iio triggered buffer setup failed, error: %d\n", ret);
> > >>>>>> +        return ret;
> > >>>>>> +    }
> > >>>>>> +
> > >>>>>> +    ret = devm_iio_device_register(&spi->dev, indio_dev);
> > >>>>>> +    if (ret < 0) {
> > >>>>>> +        dev_err(&spi->dev, "iio device register failed, error: %d\n",
> > >>>>>> +            ret);
> > >>>>>> +        iio_triggered_buffer_cleanup(indio_dev);    
> > >>>>> With devm above, no need to manually clean this up.
> > >>>>>       
> > >>>>>> +        return ret;
> > >>>>>> +    }
> > >>>>>> +
> > >>>>>> +    return 0;
> > >>>>>> +}
> > >>>>>> +
> > >>>>>> +static int sca3300_remove(struct spi_device *spi)
> > >>>>>> +{
> > >>>>>> +    struct iio_dev *indio_dev = spi_get_drvdata(spi);
> > >>>>>> +
> > >>>>>> +    iio_triggered_buffer_cleanup(indio_dev);
> > >>>>>> +    return 0;
> > >>>>>> +}
> > >>>>>> +
> > >>>>>> +static const struct of_device_id sca3300_dt_ids[] = {
> > >>>>>> +    { .compatible = "murata,sca3300"},
> > >>>>>> +    {},
> > >>>>>> +};
> > >>>>>> +MODULE_DEVICE_TABLE(of, sca3300_dt_ids);
> > >>>>>> +
> > >>>>>> +static struct spi_driver sca3300_driver = {
> > >>>>>> +    .driver = {
> > >>>>>> +        .name        = SCA3300_ALIAS,
> > >>>>>> +        .owner        = THIS_MODULE,
> > >>>>>> +        .of_match_table = of_match_ptr(sca3300_dt_ids),
> > >>>>>> +    },
> > >>>>>> +
> > >>>>>> +    .probe    = sca3300_probe,
> > >>>>>> +    .remove    = sca3300_remove,
> > >>>>>> +};
> > >>>>>> +
> > >>>>>> +module_spi_driver(sca3300_driver);
> > >>>>>> +
> > >>>>>> +MODULE_AUTHOR("Tomas Melin <[e-mail]>");
> > >>>>>> +MODULE_DESCRIPTION("Murata SCA3300 SPI Accelerometer");
> > >>>>>> +MODULE_LICENSE("GPL v2");    
>

Alexandru Ardelean April 20, 2021, 8:12 a.m. UTC | #9

(

On Mon, Apr 19, 2021 at 11:15 PM Jonathan Cameron <[e-mail]> wrote:
>
> On Mon, 19 Apr 2021 20:45:08 +0100
> Jonathan Cameron <[e-mail]> wrote:
>
> > On Mon, 19 Apr 2021 16:34:35 +0300
> > Tomas Melin <[e-mail]> wrote:
> >
> > > Hi,
> > >
> > > On 4/18/21 1:03 PM, Jonathan Cameron wrote:
> > > > On Fri, 16 Apr 2021 15:21:14 +0300
> > > > Tomas Melin <[e-mail]> wrote:
> > > >
> > > >> Updated email-address for Alexandru.
> > > >>
> > > >>
> > > >> On 4/16/21 3:17 PM, Tomas Melin wrote:
> > > >>> On 4/15/21 11:41 AM, Tomas Melin wrote:
> > > >>>
> > > >>>> While working on updates I did notice something new which I cannot
> > > >>>>
> > > >>>> reproduce on older (5.10.17 kernel) version. If compiling this as a
> > > >>>> module, getting error while
> > > >>>>
> > > >>>> unloading module:
> > > >>>>
> > > >>>> [   40.200084] Unable to handle kernel NULL pointer dereference at
> > > >>>> virtual address 00000104
> > > >>>> ...
> > > >>>>
> > > >>>> [   40.510054] Backtrace:
> > > >>>> [   40.512502] [<c06c5e0c>] (iio_device_ioctl_handler_unregister)
> > > >>>> from [<c06ca8c4>] (iio_buffers_free_sysfs_and_mask+0x2c/0x6c)
> > > >>>> [   40.523735] [<c06ca898>] (iio_buffers_free_sysfs_and_mask) from
> > > >>>> [<c06c4a18>] (iio_device_unregister+0xa8/0xac)
> > > >>>> [   40.533746]  r5:c1811228 r4:c1811000
> > > >>>> [   40.537318] [<c06c4970>] (iio_device_unregister) from [<c06c4a38>]
> > > >>>> (devm_iio_device_unreg+0x1c/0x20)
> > > >>>> [   40.546461]  r5:c2415000 r4:c25bab80
> > > >>>> [   40.550025] [<c06c4a1c>] (devm_iio_device_unreg) from [<c0556820>]
> > > >>>> (release_nodes+0x1c0/0x1f0)
> > > >>>> [   40.558654] [<c0556660>] (release_nodes) from [<c0556bb0>]
> > > >>>> (devres_release_all+0x40/0x60)
> > > >>>> [   40.566847]  r10:00000081 r9:c2350000 r8:c0100264 r7:00000081
> > > >>>> r6:bf00c010 r5:c19be000
> > > >>>> [   40.574669]  r4:c1a91c00
> > > >>>> [   40.577194] [<c0556b70>] (devres_release_all) from [<c055216c>]
> > > >>>> (device_release_driver_internal+0x120/0x1cc)
> > > >>>> [   40.587031]  r5:c19be000 r4:c1a91c00
> > > >>>> [   40.590596] [<c055204c>] (device_release_driver_internal) from
> > > >>>> [<c05522b4>] (driver_detach+0x54/0x90)
> > > >>>> [   40.599828]  r7:00000081 r6:00000000 r5:bf00c010 r4:c1a91c00
> > > >>>> [   40.605482] [<c0552260>] (driver_detach) from [<c0550ba4>]
> > > >>>> (bus_remove_driver+0x5c/0xb0)
> > > >>>> [   40.613583]  r5:00000800 r4:bf00c010
> > > >>>> [   40.617148] [<c0550b48>] (bus_remove_driver) from [<c0552c38>]
> > > >>>> (driver_unregister+0x38/0x5c)
> > > >>>> [   40.625596]  r5:00000800 r4:bf00c010
> > > >>>> [   40.629161] [<c0552c00>] (driver_unregister) from [<bf00a760>]
> > > >>>> (sca3300_driver_exit+0x14/0x8b4 [sca3300])
> > > >>>> [   40.638747]  r5:00000800 r4:bf00c080
> > > >>>> [   40.642311] [<bf00a74c>] (sca3300_driver_exit [sca3300]) from
> > > >>>> [<c01c06f0>] (sys_delete_module+0x16c/0x238)
> > > >>>> [   40.651990] [<c01c0584>] (sys_delete_module) from [<c0100244>]
> > > >>>> (__sys_trace_return+0x0/0x1c)
> > > >>>> [   40.660435] Exception stack(0xc2351fa8 to 0xc2351ff0)
> > > >>>> [   40.665484] 1fa0:                   0050e5a8 00000000 0050e5e4
> > > >>>> 00000800 081d4b00 bec18af4
> > > >>>> [   40.673661] 1fc0: 0050e5a8 00000000 bec18b50 00000081 bec18e51
> > > >>>> 0050e190 00000001 bec18d3c
> > > >>>> [   40.681834] 1fe0: 0050cf70 bec18afc 004f1ec8 b6ecb27c
> > > >>>> [   40.686887]  r6:bec18b50 r5:00000000 r4:0050e5a8
> > > >>>> [   40.691507] Code: e8bd4000 e1c020d0 e3a0cc01 e3001122 (e5823004)
> > > >>>> [   40.707675] ---[ end trace 189882b050077333 ]---
> > > >>>>
> > > >>>> This happens when building against linux-next
> > > >>>> 5.12.0-rc6-next-20210409. I'm failing to see what is wrong. Any ideas?
> > > >>>>
> > > >>>> Thanks,
> > > >>>>
> > > >>>> Tomas
> > > >>>
> > > >>> Tested further that for this driver, loading and unloading as module
> > > >>> works fine until commit:
> > > >>>
> > > >>> commit f73f7f4da581875f9b1f2fb8ebd1ab15ed634488
> > > >>> Author: Alexandru Ardelean <[e-mail]>
> > > >>> Date:   Mon Feb 15 12:40:39 2021 +0200
> > > >>>
> > > >>>      iio: buffer: add ioctl() to support opening extra buffers for IIO
> > > >>> device
> > > >>>
> > > >>>
> > > >>> Any thoughts what causes this issue?
> > > > Is this still happening after fixing the ordering in probe / remove?
> > > > (devm_iio_triggered_buffer_setup() being easiest way)
> > > >
> > > > As driver currently stands it's calling iio_triggered_buffer_cleanup
> > > > before the managed cleanup of the iio_device_register() call.  That
> > > > should never happen so would be where I'd look for problems here.
> > > >
> > > > It is possible Alex's work is relying on that ordering being correct
> > > > a little more than we previously were. I could be wrong though
> > > > and could be something else going on!
> > > >
> > > > Jonathan
> > > >
> > > > +CC Alex's other email address.
> > >
> > > This is still happening with v3 of driver patchset, which uses only devm_*
> > >
> > > versions in probing. v5.12-rc7 works ok, I'm currently only seeing this
> > > using linux-next.
> >
> > Thanks for checking.  I've just fired up a VM and confirmed that the same
> > thing is happening with creating and removing instances of the dummy driver.
> >
> > Will see if I can figure out what's going wrong.
>
> So the issue proved reasonably easy to track down.
> We remove the buffer ioctl handler twice.  First in iio_device_unregister() in a loop
> over the list, and again later in iio_buffer_sysfs_and_mask() (called from
> the same function). (note it would happen later in the equivalent event related
> version but we have already died on the first instance which is the one Tomas hit).
>
> @Alex.  Which one should we keep?  Given the buffer and event code both
> clean up manually I think what we want is:
>
> @@ -1939,9 +1943,6 @@ void iio_device_unregister(struct iio_dev *indio_dev)
>
>         indio_dev->info = NULL;
>
> -       list_for_each_entry_safe(h, t, &iio_dev_opaque->ioctl_handlers, entry)
> -               list_del(&h->entry);
> -
>         iio_device_wakeup_eventset(indio_dev);
>         iio_buffer_wakeup_poll(indio_dev);
>
>
> I'm going to guess something went wrong in a merge or rebased and left that
> block when it should have been removed.

No, that's not a merge/rebase gone wrong.
It's me being stupid.

Removing that loop is the good fix.
Each module doing a iio_device_ioctl_handler_register() should also do
a iio_device_ioctl_handler_unregister().

>
> Anyhow, if the above is the right fix I can roll it up into a patch tomorrow and
> get it into linux-next at least.

Sure :)

Apologies for the breakage.

>
> thanks,
>
> Jonathan
>
> >
> > Jonathan
> >
> > >
> > > Thanks,
> > >
> > > Tomas
> > >
> > >
> > > >>> Thanks,
> > > >>>
> > > >>> Tomas
> > > >>>
> > > >>>
> > > >>>
> > > >>>>
> > > >>>>>
> > > >>>>>> ---
> > > >>>>>>    drivers/iio/accel/Kconfig   |  13 ++
> > > >>>>>>    drivers/iio/accel/Makefile  |   1 +
> > > >>>>>>    drivers/iio/accel/sca3300.c | 434
> > > >>>>>> ++++++++++++++++++++++++++++++++++++
> > > >>>>>>    3 files changed, 448 insertions(+)
> > > >>>>>>    create mode 100644 drivers/iio/accel/sca3300.c
> > > >>>>>>
> > > >>>>>> diff --git a/drivers/iio/accel/Kconfig b/drivers/iio/accel/Kconfig
> > > >>>>>> index cceda3cecbcf..0dbf7b648e8a 100644
> > > >>>>>> --- a/drivers/iio/accel/Kconfig
> > > >>>>>> +++ b/drivers/iio/accel/Kconfig
> > > >>>>>> @@ -450,6 +450,19 @@ config SCA3000
> > > >>>>>>          To compile this driver as a module, say M here: the module
> > > >>>>>> will be
> > > >>>>>>          called sca3000.
> > > >>>>>>    +config SCA3300
> > > >>>>>> +    tristate "Murata SCA3300 3-Axis Accelerometer Driver"
> > > >>>>>> +    depends on SPI
> > > >>>>>> +    select CRC8
> > > >>>>>> +    select IIO_BUFFER
> > > >>>>>> +    select IIO_TRIGGERED_BUFFER
> > > >>>>>> +    help
> > > >>>>>> +      Say yes here to build support for Murata SCA3300 3-Axis
> > > >>>>>> +      accelerometer.
> > > >>>>>> +
> > > >>>>>> +      To compile this driver as a module, choose M here: the
> > > >>>>>> module will be
> > > >>>>>> +      called sca3300.
> > > >>>>>> +
> > > >>>>>>    config STK8312
> > > >>>>>>        tristate "Sensortek STK8312 3-Axis Accelerometer Driver"
> > > >>>>>>        depends on I2C
> > > >>>>>> diff --git a/drivers/iio/accel/Makefile b/drivers/iio/accel/Makefile
> > > >>>>>> index 32cd1342a31a..4b56527a2b97 100644
> > > >>>>>> --- a/drivers/iio/accel/Makefile
> > > >>>>>> +++ b/drivers/iio/accel/Makefile
> > > >>>>>> @@ -50,6 +50,7 @@ obj-$(CONFIG_MXC4005)        += mxc4005.o
> > > >>>>>>    obj-$(CONFIG_MXC6255)        += mxc6255.o
> > > >>>>>>      obj-$(CONFIG_SCA3000)        += sca3000.o
> > > >>>>>> +obj-$(CONFIG_SCA3300)        += sca3300.o
> > > >>>>>>      obj-$(CONFIG_STK8312)        += stk8312.o
> > > >>>>>>    obj-$(CONFIG_STK8BA50)        += stk8ba50.o
> > > >>>>>> diff --git a/drivers/iio/accel/sca3300.c b/drivers/iio/accel/sca3300.c
> > > >>>>>> new file mode 100644
> > > >>>>>> index 000000000000..112fb88ecd3a
> > > >>>>>> --- /dev/null
> > > >>>>>> +++ b/drivers/iio/accel/sca3300.c
> > > >>>>>> @@ -0,0 +1,434 @@
> > > >>>>>> +// SPDX-License-Identifier: GPL-2.0-only
> > > >>>>>> +/*
> > > >>>>>> + * Copyright (c) Vaisala Oyj. All rights reserved.
> > > >>>>> Give a year for the copyright notice if you can.
> > > >>>>>
> > > >>>>>> + */
> > > >>>>>> +#include <linux/crc8.h>
> > > >>>>>> +#include <linux/delay.h>
> > > >>>>>> +#include <linux/iio/buffer.h>
> > > >>>>>> +#include <linux/iio/iio.h>
> > > >>>>>> +#include <linux/iio/sysfs.h>
> > > >>>>>> +#include <linux/iio/trigger_consumer.h>
> > > >>>>>> +#include <linux/iio/triggered_buffer.h>
> > > >>>>>> +#include <linux/kernel.h>
> > > >>>>>> +#include <linux/module.h>
> > > >>>>>> +#include <linux/spi/spi.h>
> > > >>>>>> +
> > > >>>>>> +#define SCA3300_ALIAS "sca3300"
> > > >>>>>> +
> > > >>>>>> +#define SCA3300_REG_STATUS 0x6
> > > >>>>>> +#define SCA3300_REG_MODE 0xd
> > > >>>>>> +#define SCA3300_REG_WHOAMI 0x10
> > > >>>>>> +#define SCA3300_VALUE_SW_RESET 0x20
> > > >>>>>> +#define SCA3300_CRC8_POLYNOMIAL 0x1d
> > > >>>>>> +#define SCA3300_X_READ 0
> > > >>>>> I wouldn't bother defining this.
> > > >>>>>
> > > >>>>>> +#define SCA3300_X_WRITE BIT(7)
> > > >>>>> Even this one is something I'd just put inline with a comment.
> > > >>>>>
> > > >>>>>> +#define SCA3300_DEVICE_ID 0x51
> > > >>>>>> +#define SCA3300_RS_ERROR 0x3
> > > >>>>>> +
> > > >>>>>> +enum sca3300_scan_indexes {
> > > >>>>>> +    SCA3300_ACC_X = 0,
> > > >>>>>> +    SCA3300_ACC_Y,
> > > >>>>>> +    SCA3300_ACC_Z,
> > > >>>>>> +    SCA3300_TEMP,
> > > >>>>>> +    SCA3300_TIMESTAMP,
> > > >>>>>> +};
> > > >>>>>> +
> > > >>>>>> +#define SCA3300_ACCEL_CHANNEL(index, reg, axis) { \
> > > >>>>>> +        .type = IIO_ACCEL,                    \
> > > >>>>>> +        .address = reg,                        \
> > > >>>>>> +        .modified = 1,                        \
> > > >>>>>> +        .channel2 = IIO_MOD_##axis,                \
> > > >>>>>> +        .info_mask_separate = BIT(IIO_CHAN_INFO_RAW) | \
> > > >>>>>> +                      BIT(IIO_CHAN_INFO_PROCESSED),    \
> > > >>>>> As mentioned below, don't provide PROCESSED. Userspace is better at
> > > >>>>> handling the
> > > >>>>> conversion so leave it to them.
> > > >>>>>
> > > >>>>>> +        .info_mask_shared_by_type = BIT(IIO_CHAN_INFO_SCALE),    \
> > > >>>>>> +        .scan_index = index,                    \
> > > >>>>>> +        .scan_type = {                        \
> > > >>>>>> +            .sign = 's',                    \
> > > >>>>>> +            .realbits = 16,                    \
> > > >>>>>> +            .storagebits = 16,                \
> > > >>>>>> +            .shift = 0,                    \
> > > >>>>>> +            .endianness = IIO_CPU,                \
> > > >>>>>> +        },                            \
> > > >>>>>> +    }
> > > >>>>>> +
> > > >>>>>> +static const struct iio_chan_spec sca3300_channels[] = {
> > > >>>>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_X, 0x1, X),
> > > >>>>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Y, 0x2, Y),
> > > >>>>>> +    SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Z, 0x3, Z),
> > > >>>>>> +    {
> > > >>>>>> +        .type = IIO_TEMP,
> > > >>>>>> +        .address = 0x5,
> > > >>>>>> +        .scan_index = SCA3300_TEMP,
> > > >>>>>> +        .info_mask_separate = BIT(IIO_CHAN_INFO_RAW),
> > > >>>>>> +        .scan_type = {
> > > >>>>>> +            .sign = 's',
> > > >>>>>> +            .realbits = 16,
> > > >>>>>> +            .storagebits = 16,
> > > >>>>>> +            .shift = 0,
> > > >>>>>> +            .endianness = IIO_CPU,
> > > >>>>>> +        },
> > > >>>>>> +    },
> > > >>>>>> +    IIO_CHAN_SOFT_TIMESTAMP(4),
> > > >>>>>> +};
> > > >>>>>> +
> > > >>>>>> +static const int sca3300_accel_scale[] = {2700, 1350, 5400, 5400};
> > > >>>>>> +
> > > >>>>>> +static const unsigned long sca3300_scan_masks[] = {
> > > >>>>>> +    BIT(SCA3300_ACC_X) | BIT(SCA3300_ACC_Y) | BIT(SCA3300_ACC_Z) |
> > > >>>>>> +    BIT(SCA3300_TEMP),
> > > >>>>>> +    0};
> > > >>>>>> +
> > > >>>>>> +/**
> > > >>>>>> + * SCA3300 device data
> > > >>>>> run scripts/kernel-doc over the file and fix the warnings + errors.
> > > >>>>>
> > > >>>>>> + *
> > > >>>>>> + * @spi SPI device structure
> > > >>>>>> + * @opmode Device operation mode
> > > >>>>>> + * @lock Data buffer lock
> > > >>>>>> + * @txbuf Transmit buffer
> > > >>>>>> + * @rxbuf Receive buffer
> > > >>>>>> + * @scan Triggered buffer. Four channel 16-bit data + 64-bit
> > > >>>>>> timestamp
> > > >>>>>> + */
> > > >>>>>> +struct sca3300_data {
> > > >>>>>> +    struct spi_device *spi;
> > > >>>>>> +    u32 opmode;
> > > >>>>>> +    struct mutex lock;
> > > >>>>>> +    u8 txbuf[4];
> > > >>>>>> +    u8 rxbuf[4];
> > > >>>>>> +    struct {
> > > >>>>>> +        s16 channels[4];
> > > >>>>>> +        s64 ts __aligned(sizeof(s64));
> > > >>>>>> +    } scan;
> > > >>>>>> +};
> > > >>>>>> +
> > > >>>>>> +DECLARE_CRC8_TABLE(sca3300_crc_table);
> > > >>>>>> +
> > > >>>>>> +static int sca3300_transfer(struct sca3300_data *sca_data, int *val)
> > > >>>>>> +{
> > > >>>>>> +    struct spi_delay delay = {.value = 10, .unit =
> > > >>>>>> SPI_DELAY_UNIT_USECS};
> > > >>>>>> +    int32_t ret;
> > > >>>>>> +    int rs;
> > > >>>>>> +    u8 crc;
> > > >>>>>> +    struct spi_transfer xfers[2] = {
> > > >>>>>> +        {
> > > >>>>>> +            .tx_buf = sca_data->txbuf,
> > > >>>>>> +            .rx_buf = NULL,
> > > >>>>>> +            .len = ARRAY_SIZE(sca_data->txbuf),
> > > >>>>>> +            .delay = delay,
> > > >>>>>> +            .cs_change = 1,
> > > >>>>>> +        },
> > > >>>>>> +        {
> > > >>>>>> +            .tx_buf = NULL,
> > > >>>>>> +            .rx_buf = sca_data->rxbuf,
> > > >>>>>> +            .len = ARRAY_SIZE(sca_data->rxbuf),
> > > >>>>>> +            .delay = delay,
> > > >>>>>> +            .cs_change = 0,
> > > >>>>>> +        }
> > > >>>>>> +    };
> > > >>>>>> +
> > > >>>>>> +    /* inverted crc value as described in device data sheet */
> > > >>>>>> +    crc = ~crc8(sca3300_crc_table, &sca_data->txbuf[0], 3,
> > > >>>>>> CRC8_INIT_VALUE);
> > > >>>>>> +    sca_data->txbuf[3] = crc;
> > > >>>>>> +
> > > >>>>>> +    ret = spi_sync_transfer(sca_data->spi, xfers, 2);
> > > >>>>> Use ARRAY_SIZE(xfers) instead of 2.
> > > >>>>>
> > > >>>>>> +    if (ret < 0) {
> > > >>>>>> +        dev_err(&sca_data->spi->dev,
> > > >>>>>> +            "transfer error, error: %d\n", ret);
> > > >>>>>> +        return -EIO;
> > > >>>>>> +    }
> > > >>>>>> +
> > > >>>>>> +    crc = ~crc8(sca3300_crc_table, &sca_data->rxbuf[0], 3,
> > > >>>>>> CRC8_INIT_VALUE);
> > > >>>>>> +    if (sca_data->rxbuf[3] != crc) {
> > > >>>>>> +        dev_err(&sca_data->spi->dev, "CRC checksum mismatch");
> > > >>>>>> +        return -EIO;
> > > >>>>>> +    }
> > > >>>>>> +
> > > >>>>>> +    /* get return status */
> > > >>>>>> +    rs = sca_data->rxbuf[0] & 0x03;
> > > >>>>>> +    if (rs == SCA3300_RS_ERROR)
> > > >>>>>> +        return rs;
> > > >>>>>> +
> > > >>>>>> +    *val = (s16)(sca_data->rxbuf[2] | (sca_data->rxbuf[1] << 8));
> > > >>>>> Preference for an unaligned endian conversion here and explicit sign
> > > >>>>> extend e.g
> > > >>>>> something like
> > > >>>>>
> > > >>>>> sign_extend32(get_unaligned_le16(&sca_data->rxbuf[1]), 15)
> > > >>>>>
> > > >>>>>> +
> > > >>>>>> +    return 0;
> > > >>>>>> +}
> > > >>>>>> +
> > > >>>>>> +static int sca3300_read_reg(struct sca3300_data *sca_data, u8 reg,
> > > >>>>>> int *val)
> > > >>>>>> +{
> > > >>>>>> +    int ret;
> > > >>>>>> +
> > > >>>>>> +    mutex_lock(&sca_data->lock);
> > > >>>>>> +    sca_data->txbuf[0] = SCA3300_X_READ | (reg << 2);
> > > >>>>>> +    ret = sca3300_transfer(sca_data, val);
> > > >>>>>> +    if (ret > 0) {
> > > >>>>>> +        sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS
> > > >>>>>> << 2);
> > > >>>>>> +        ret = sca3300_transfer(sca_data, val);
> > > >>>>>> +        /* status 0 = startup, 0x2 = mode change */
> > > >>>>>> +        if (ret > 0 && *val != 0 && *val != 0x2) {
> > > >>>>>> + dev_err_ratelimited(&sca_data->spi->dev,
> > > >>>>>> +                        "device status: %x\n",
> > > >>>>>> +                        (u16)*val);
> > > >>>>>> +            mutex_unlock(&sca_data->lock);
> > > >>>>>> +            return -EIO;
> > > >>>>>> +        }
> > > >>>>>> +        if (ret > 0)
> > > >>>>>> +            ret = 0;
> > > >>>>>> +    }
> > > >>>>>> +    mutex_unlock(&sca_data->lock);
> > > >>>>>> +
> > > >>>>>> +    return ret;
> > > >>>>>> +}
> > > >>>>>> +
> > > >>>>>> +static int sca3300_write_reg(struct sca3300_data *sca_data, u8
> > > >>>>>> reg, int val)
> > > >>>>>> +{
> > > >>>>>> +    int reg_val = 0;
> > > >>>>>> +    int ret;
> > > >>>>>> +
> > > >>>>>> +    mutex_lock(&sca_data->lock);
> > > >>>>>> +    sca_data->txbuf[0] = SCA3300_X_WRITE | (reg << 2);
> > > >>>>>> +    sca_data->txbuf[1] = val >> 8;
> > > >>>>>> +    sca_data->txbuf[2] = val & 0xFF;
> > > >>>>> Prefer (Slightly) an unaligned put.
> > > >>>>>
> > > >>>>>> +    ret = sca3300_transfer(sca_data, &reg_val);
> > > >>>>>> +    if (ret > 0) {
> > > >>>>> Factor this error handling out to another function, plus trigger it
> > > >>>>> from an
> > > >>>>> appropriate standard error code rather than a positive return value.
> > > >>>>>
> > > >>>>>> +        sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS
> > > >>>>>> << 2);
> > > >>>>>> +        ret = sca3300_transfer(sca_data, &reg_val);
> > > >>>>>> +        /* status 0 = startup, 0x2 = mode change */
> > > >>>>>> +        if (ret > 0 && reg_val != 0 && reg_val != 0x2) {
> > > >>>>>> + dev_err_ratelimited(&sca_data->spi->dev,
> > > >>>>>> +                        "device status: %x\n",
> > > >>>>>> +                        (u16)reg_val);
> > > >>>>>> +            mutex_unlock(&sca_data->lock);
> > > >>>>>> +            return -EIO;
> > > >>>>>> +        }
> > > >>>>>> +        if (ret > 0)
> > > >>>>>> +            ret = 0;
> > > >>>>> Whenever you get this sort of juggling it rather implies your design is
> > > >>>>> less than ideal.
> > > >>>>>
> > > >>>>>> +    }
> > > >>>>>> +    mutex_unlock(&sca_data->lock);
> > > >>>>>> +
> > > >>>>>> +    return ret;
> > > >>>>>> +}
> > > >>>>>> +
> > > >>>>>> +static int sca3300_write_raw(struct iio_dev *indio_dev,
> > > >>>>>> +                 struct iio_chan_spec const *chan,
> > > >>>>>> +                 int val, int val2, long mask)
> > > >>>>>> +{
> > > >>>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
> > > >>>>>> +
> > > >>>>>> +    switch (mask) {
> > > >>>>>> +    case IIO_CHAN_INFO_SCALE:
> > > >>>>>> +        if (val < 0 || val > 3)
> > > >>>>>> +            return -EINVAL;
> > > >>>>> No.  Scale values need to be the value of 1LSB not a random integer
> > > >>>>> that requires you to look up the datasheet.
> > > >>>>>
> > > >>>>>> +        return sca3300_write_reg(data, SCA3300_REG_MODE, val);
> > > >>>>>> +    default:
> > > >>>>>> +        return -EINVAL;
> > > >>>>>> +    }
> > > >>>>>> +}
> > > >>>>>> +
> > > >>>>>> +static int sca3300_read_raw(struct iio_dev *indio_dev,
> > > >>>>>> +                struct iio_chan_spec const *chan,
> > > >>>>>> +                int *val, int *val2, long mask)
> > > >>>>>> +{
> > > >>>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
> > > >>>>>> +    int ret;
> > > >>>>>> +    int reg_val;
> > > >>>>>> +
> > > >>>>>> +    switch (mask) {
> > > >>>>>> +    case IIO_CHAN_INFO_RAW:
> > > >>>>>> +        ret = sca3300_read_reg(data, chan->address, val);
> > > >>>>>> +        if (ret < 0)
> > > >>>>>> +            return ret;
> > > >>>>>> +        return IIO_VAL_INT;
> > > >>>>>> +    case IIO_CHAN_INFO_SCALE:
> > > >>>>>> +        ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
> > > >>>>>> +        if (ret < 0)
> > > >>>>>> +            return ret;
> > > >>>>>> +        *val = sca3300_accel_scale[reg_val];
> > > >>>>> Scale is very very rarely an integer value which makes me wonder what
> > > >>>>> you are returning here...  It should be the multiplier needed to
> > > >>>>> take the raw value to a reading in m/sec^2  Given you use it as
> > > >>>>> fractional
> > > >>>>> below, I'm guessing this is 1/scale_value?
> > > >>>>>
> > > >>>>>> +        return IIO_VAL_INT;
> > > >>>>>> +    case IIO_CHAN_INFO_PROCESSED:
> > > >>>>> Don't provide processed here.   Userspace needs to do the conversion
> > > >>>>> itself
> > > >>>>> (and given it has floating point easily available will probably do a
> > > >>>>> better
> > > >>>>>    job than we can).   I'm assuming the mode can't autonomously change?
> > > >>>>>
> > > >>>>>> +        ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
> > > >>>>>> +        if (ret < 0)
> > > >>>>>> +            return ret;
> > > >>>>>> +        *val2 = sca3300_accel_scale[reg_val];
> > > >>>>>> +        ret = sca3300_read_reg(data, chan->address, val);
> > > >>>>>> +        if (ret < 0)
> > > >>>>>> +            return ret;
> > > >>>>>> +        return IIO_VAL_FRACTIONAL;
> > > >>>>>> +    default:
> > > >>>>>> +        return -EINVAL;
> > > >>>>>> +    }
> > > >>>>>> +}
> > > >>>>>> +
> > > >>>>>> +static irqreturn_t sca3300_trigger_handler(int irq, void *p)
> > > >>>>>> +{
> > > >>>>>> +    struct iio_poll_func *pf = p;
> > > >>>>>> +    struct iio_dev *indio_dev = pf->indio_dev;
> > > >>>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
> > > >>>>>> +    s64 time_ns = iio_get_time_ns(indio_dev);
> > > >>>>> Is the timestamp at the start more accurate that that at the end
> > > >>>>> of reading the channels?  If not, just put this inline in the
> > > >>>>> iio_push_to_buffers_with_timestamp() call.
> > > >>>>>
> > > >>>>>> +    int bit, ret, val, i = 0;
> > > >>>>>> +
> > > >>>>>> +    for_each_set_bit(bit, indio_dev->active_scan_mask,
> > > >>>>>> +             indio_dev->masklength) {
> > > >>>>>> +        ret = sca3300_read_reg(data, sca3300_channels[bit].address,
> > > >>>>>> +                       &val);
> > > >>>>>> +        if (ret < 0)
> > > >>>>> Given we can't return an error code from interrupt handlers, it is
> > > >>>>> usually
> > > >>>>> a good idea to print something to the log. Otherwise we get missing
> > > >>>>> data
> > > >>>>> with no idea of why..
> > > >>>>>
> > > >>>>>> +            goto out;
> > > >>>>>> +        if (ARRAY_SIZE(data->scan.channels) > i)
> > > >>>>> How could this not be true?  If it's always true, don't bother
> > > >>>>> checking i.
> > > >>>>>
> > > >>>>>> +            ((s16 *)data->scan.channels)[i++] = val;
> > > >>>>>> +    }
> > > >>>>>> +
> > > >>>>>> +    iio_push_to_buffers_with_timestamp(indio_dev, &data->scan,
> > > >>>>>> time_ns);
> > > >>>>>> +out:
> > > >>>>>> +    iio_trigger_notify_done(indio_dev->trig);
> > > >>>>>> +
> > > >>>>>> +    return IRQ_HANDLED;
> > > >>>>>> +}
> > > >>>>>> +
> > > >>>>>> +static int sca3300_init(struct sca3300_data *sca_data,
> > > >>>>>> +            struct iio_dev *indio_dev)
> > > >>>>>> +{
> > > >>>>>> +    int ret;
> > > >>>>>> +    int value = 0;
> > > >>>>>> +
> > > >>>>>> +    if (sca_data->opmode < 1 || sca_data->opmode > 4)
> > > >>>>>> +        return -EINVAL;
> > > >>>>> Whilst this is going anyway, better to check that next to where
> > > >>>>> it is read so we drop out immediately rather than buried in this
> > > >>>>> function.
> > > >>>>>
> > > >>>>>> +
> > > >>>>>> +    ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
> > > >>>>>> +                SCA3300_VALUE_SW_RESET);
> > > >>>>>> +    if (ret != 0)
> > > >>>>>> +        return ret;
> > > >>>>> For sleeps, good to reference which section in data sheet gives the
> > > >>>>> timing.
> > > >>>>>
> > > >>>>>> +    usleep_range(2e3, 10e3);
> > > >>>>>> +
> > > >>>>>> +    ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
> > > >>>>>> +                sca_data->opmode - 1);
> > > >>>>>> +    if (ret != 0)
> > > >>>>> if (ret) see below.
> > > >>>>>
> > > >>>>>> +        return ret;
> > > >>>>>> +    msleep(100);
> > > >>>>> Again, document where the time comes from. Saves reviewers time if
> > > >>>>> they want to check it. (I'm too lazy :)
> > > >>>>>
> > > >>>>>> +    ret = sca3300_read_reg(sca_data, SCA3300_REG_WHOAMI, &value);
> > > >>>>>> +    if (ret != 0)
> > > >>>>> if (ret) is more idiomatic in kernel code.
> > > >>>>>
> > > >>>>>> +        return ret;
> > > >>>>>> +
> > > >>>>>> +    if (value != SCA3300_DEVICE_ID) {
> > > >>>>>> +        dev_err(&sca_data->spi->dev, "device id not expected
> > > >>>>>> value\n");
> > > >>>>> Perhaps useful to print what was expected and what was seen. Maybe it's
> > > >>>>> a new variant and this might help the user to identify that and add
> > > >>>>> it to the
> > > >>>>> driver.
> > > >>>>>
> > > >>>>>> +        return -EIO;
> > > >>>>> -EINVAL probably more appropriate return.
> > > >>>>>
> > > >>>>>> +    }
> > > >>>>>> +    return 0;
> > > >>>>>> +}
> > > >>>>>> +
> > > >>>>>> +static int sca3300_debugfs_reg_access(struct iio_dev *indio_dev,
> > > >>>>>> +                      unsigned int reg, unsigned int writeval,
> > > >>>>>> +                      unsigned int *readval)
> > > >>>>>> +{
> > > >>>>>> +    struct sca3300_data *data = iio_priv(indio_dev);
> > > >>>>>> +    int value;
> > > >>>>>> +    int ret;
> > > >>>>>> +
> > > >>>>>> +    if (reg > 0x1f)
> > > >>>>> Use a define that names that register.
> > > >>>>>
> > > >>>>>> +        return -EINVAL;
> > > >>>>>> +
> > > >>>>>> +    if (!readval)
> > > >>>>>> +        return sca3300_write_reg(data, reg, writeval);
> > > >>>>>> +
> > > >>>>>> +    ret = sca3300_read_reg(data, reg, &value);
> > > >>>>>> +    if (ret < 0)
> > > >>>>>> +        return ret;
> > > >>>>>> +
> > > >>>>>> +    *readval = (unsigned int)value;
> > > >>>>>> +
> > > >>>>>> +    return 0;
> > > >>>>>> +}
> > > >>>>>> +
> > > >>>>>> +static const struct iio_info sca3300_info = {
> > > >>>>>> +    .read_raw = sca3300_read_raw,
> > > >>>>>> +    .write_raw = sca3300_write_raw,
> > > >>>>>> +    .debugfs_reg_access = &sca3300_debugfs_reg_access,
> > > >>>>>> +};
> > > >>>>>> +
> > > >>>>>> +static int sca3300_probe(struct spi_device *spi)
> > > >>>>>> +{
> > > >>>>>> +    struct sca3300_data *sca_data;
> > > >>>>>> +    struct iio_dev *indio_dev;
> > > >>>>>> +    int ret;
> > > >>>>>> +
> > > >>>>>> +    indio_dev = devm_iio_device_alloc(&spi->dev, sizeof(*sca_data));
> > > >>>>>> +    if (!indio_dev) {
> > > >>>>>> +        dev_err(&spi->dev,
> > > >>>>>> +            "failed to allocate memory for iio device\n");
> > > >>>>>> +        return -ENOMEM;
> > > >>>>>> +    }
> > > >>>>>> +
> > > >>>>>> +    sca_data = iio_priv(indio_dev);
> > > >>>>>> +    mutex_init(&sca_data->lock);
> > > >>>>>> +    sca_data->spi = spi;
> > > >>>>>> +    spi_set_drvdata(spi, indio_dev);
> > > >>>>> Won't be used once using devm as below and remove() is dropped, so
> > > >>>>> drop this.
> > > >>>>>
> > > >>>>>> +
> > > >>>>>> +    crc8_populate_msb(sca3300_crc_table, SCA3300_CRC8_POLYNOMIAL);
> > > >>>>>> +
> > > >>>>>> +    indio_dev->dev.parent = &spi->dev;
> > > >>>>>> +    indio_dev->info = &sca3300_info;
> > > >>>>>> +    indio_dev->name = SCA3300_ALIAS;
> > > >>>>>> +    indio_dev->modes = INDIO_DIRECT_MODE | INDIO_BUFFER_TRIGGERED;
> > > >>>>> BUFFER_TRIGGERED bit now set by iio_triggered_buffer_setup() so
> > > >>>>> shouldn't
> > > >>>>> be ehre as well.
> > > >>>>>
> > > >>>>>> +    indio_dev->channels = sca3300_channels;
> > > >>>>>> +    indio_dev->num_channels = ARRAY_SIZE(sca3300_channels);
> > > >>>>>> +    indio_dev->available_scan_masks = sca3300_scan_masks;
> > > >>>>>> +
> > > >>>>>> +    if (spi->dev.of_node) {
> > > >>>>>> +        ret = of_property_read_u32(spi->dev.of_node, "murata,opmode",
> > > >>>>> Please use generic firmware access calls rather than the of specific
> > > >>>>> ones.
> > > >>>>> That lets us get ACPI support for free :)
> > > >>>>>
> > > >>>>> As per the binding review however, this one needs to go in favour of
> > > >>>>> userspace
> > > >>>>> control of scale + filter frequencies / sampling freq.
> > > >>>>>
> > > >>>>>> + &sca_data->opmode);
> > > >>>>>> +        if (ret < 0)
> > > >>>>>> +            return ret;
> > > >>>>>> +    }
> > > >>>>>> +
> > > >>>>>> +    ret = sca3300_init(sca_data, indio_dev);
> > > >>>>>> +    if (ret < 0) {
> > > >>>>>> +        dev_err(&spi->dev, "failed to init device, error: %d\n",
> > > >>>>>> ret);
> > > >>>>>> +        return ret;
> > > >>>>>> +    }
> > > >>>>>> +
> > > >>>>>> +    ret = iio_triggered_buffer_setup(indio_dev,
> > > >>>>>> iio_pollfunc_store_time,
> > > >>>>>> +                     sca3300_trigger_handler, NULL);
> > > >>>>> devm_iio_triggered_buffer_setup() and you can drop the manual
> > > >>>>> cleanup in remove()
> > > >>>>> and hence drop remove() entirely as nothing else left.
> > > >>>>>
> > > >>>>>> +    if (ret < 0) {
> > > >>>>>> +        dev_err(&spi->dev,
> > > >>>>>> +            "iio triggered buffer setup failed, error: %d\n", ret);
> > > >>>>>> +        return ret;
> > > >>>>>> +    }
> > > >>>>>> +
> > > >>>>>> +    ret = devm_iio_device_register(&spi->dev, indio_dev);
> > > >>>>>> +    if (ret < 0) {
> > > >>>>>> +        dev_err(&spi->dev, "iio device register failed, error: %d\n",
> > > >>>>>> +            ret);
> > > >>>>>> +        iio_triggered_buffer_cleanup(indio_dev);
> > > >>>>> With devm above, no need to manually clean this up.
> > > >>>>>
> > > >>>>>> +        return ret;
> > > >>>>>> +    }
> > > >>>>>> +
> > > >>>>>> +    return 0;
> > > >>>>>> +}
> > > >>>>>> +
> > > >>>>>> +static int sca3300_remove(struct spi_device *spi)
> > > >>>>>> +{
> > > >>>>>> +    struct iio_dev *indio_dev = spi_get_drvdata(spi);
> > > >>>>>> +
> > > >>>>>> +    iio_triggered_buffer_cleanup(indio_dev);
> > > >>>>>> +    return 0;
> > > >>>>>> +}
> > > >>>>>> +
> > > >>>>>> +static const struct of_device_id sca3300_dt_ids[] = {
> > > >>>>>> +    { .compatible = "murata,sca3300"},
> > > >>>>>> +    {},
> > > >>>>>> +};
> > > >>>>>> +MODULE_DEVICE_TABLE(of, sca3300_dt_ids);
> > > >>>>>> +
> > > >>>>>> +static struct spi_driver sca3300_driver = {
> > > >>>>>> +    .driver = {
> > > >>>>>> +        .name        = SCA3300_ALIAS,
> > > >>>>>> +        .owner        = THIS_MODULE,
> > > >>>>>> +        .of_match_table = of_match_ptr(sca3300_dt_ids),
> > > >>>>>> +    },
> > > >>>>>> +
> > > >>>>>> +    .probe    = sca3300_probe,
> > > >>>>>> +    .remove    = sca3300_remove,
> > > >>>>>> +};
> > > >>>>>> +
> > > >>>>>> +module_spi_driver(sca3300_driver);
> > > >>>>>> +
> > > >>>>>> +MODULE_AUTHOR("Tomas Melin <[e-mail]>");
> > > >>>>>> +MODULE_DESCRIPTION("Murata SCA3300 SPI Accelerometer");
> > > >>>>>> +MODULE_LICENSE("GPL v2");
> >
>

Patch

diff --git a/drivers/iio/accel/Kconfig b/drivers/iio/accel/Kconfig
index cceda3cecbcf..0dbf7b648e8a 100644
--- a/drivers/iio/accel/Kconfig
+++ b/drivers/iio/accel/Kconfig
@@ -450,6 +450,19 @@  config SCA3000
 	  To compile this driver as a module, say M here: the module will be
 	  called sca3000.
 
+config SCA3300
+	tristate "Murata SCA3300 3-Axis Accelerometer Driver"
+	depends on SPI
+	select CRC8
+	select IIO_BUFFER
+	select IIO_TRIGGERED_BUFFER
+	help
+	  Say yes here to build support for Murata SCA3300 3-Axis
+	  accelerometer.
+
+	  To compile this driver as a module, choose M here: the module will be
+	  called sca3300.
+
 config STK8312
 	tristate "Sensortek STK8312 3-Axis Accelerometer Driver"
 	depends on I2C
diff --git a/drivers/iio/accel/Makefile b/drivers/iio/accel/Makefile
index 32cd1342a31a..4b56527a2b97 100644
--- a/drivers/iio/accel/Makefile
+++ b/drivers/iio/accel/Makefile
@@ -50,6 +50,7 @@  obj-$(CONFIG_MXC4005)		+= mxc4005.o
 obj-$(CONFIG_MXC6255)		+= mxc6255.o
 
 obj-$(CONFIG_SCA3000)		+= sca3000.o
+obj-$(CONFIG_SCA3300)		+= sca3300.o
 
 obj-$(CONFIG_STK8312)		+= stk8312.o
 obj-$(CONFIG_STK8BA50)		+= stk8ba50.o
diff --git a/drivers/iio/accel/sca3300.c b/drivers/iio/accel/sca3300.c
new file mode 100644
index 000000000000..112fb88ecd3a
--- /dev/null
+++ b/drivers/iio/accel/sca3300.c
@@ -0,0 +1,434 @@ 
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (c) Vaisala Oyj. All rights reserved.
+ */
+#include <linux/crc8.h>
+#include <linux/delay.h>
+#include <linux/iio/buffer.h>
+#include <linux/iio/iio.h>
+#include <linux/iio/sysfs.h>
+#include <linux/iio/trigger_consumer.h>
+#include <linux/iio/triggered_buffer.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/spi/spi.h>
+
+#define SCA3300_ALIAS "sca3300"
+
+#define SCA3300_REG_STATUS 0x6
+#define SCA3300_REG_MODE 0xd
+#define SCA3300_REG_WHOAMI 0x10
+#define SCA3300_VALUE_SW_RESET 0x20
+#define SCA3300_CRC8_POLYNOMIAL 0x1d
+#define SCA3300_X_READ 0
+#define SCA3300_X_WRITE BIT(7)
+#define SCA3300_DEVICE_ID 0x51
+#define SCA3300_RS_ERROR 0x3
+
+enum sca3300_scan_indexes {
+	SCA3300_ACC_X = 0,
+	SCA3300_ACC_Y,
+	SCA3300_ACC_Z,
+	SCA3300_TEMP,
+	SCA3300_TIMESTAMP,
+};
+
+#define SCA3300_ACCEL_CHANNEL(index, reg, axis) {			\
+		.type = IIO_ACCEL,					\
+		.address = reg,						\
+		.modified = 1,						\
+		.channel2 = IIO_MOD_##axis,				\
+		.info_mask_separate = BIT(IIO_CHAN_INFO_RAW) |		\
+				      BIT(IIO_CHAN_INFO_PROCESSED),	\
+		.info_mask_shared_by_type = BIT(IIO_CHAN_INFO_SCALE),	\
+		.scan_index = index,					\
+		.scan_type = {						\
+			.sign = 's',					\
+			.realbits = 16,					\
+			.storagebits = 16,				\
+			.shift = 0,					\
+			.endianness = IIO_CPU,				\
+		},							\
+	}
+
+static const struct iio_chan_spec sca3300_channels[] = {
+	SCA3300_ACCEL_CHANNEL(SCA3300_ACC_X, 0x1, X),
+	SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Y, 0x2, Y),
+	SCA3300_ACCEL_CHANNEL(SCA3300_ACC_Z, 0x3, Z),
+	{
+		.type = IIO_TEMP,
+		.address = 0x5,
+		.scan_index = SCA3300_TEMP,
+		.info_mask_separate = BIT(IIO_CHAN_INFO_RAW),
+		.scan_type = {
+			.sign = 's',
+			.realbits = 16,
+			.storagebits = 16,
+			.shift = 0,
+			.endianness = IIO_CPU,
+		},
+	},
+	IIO_CHAN_SOFT_TIMESTAMP(4),
+};
+
+static const int sca3300_accel_scale[] = {2700, 1350, 5400, 5400};
+
+static const unsigned long sca3300_scan_masks[] = {
+	BIT(SCA3300_ACC_X) | BIT(SCA3300_ACC_Y) | BIT(SCA3300_ACC_Z) |
+	BIT(SCA3300_TEMP),
+	0};
+
+/**
+ * SCA3300 device data
+ *
+ * @spi SPI device structure
+ * @opmode Device operation mode
+ * @lock Data buffer lock
+ * @txbuf Transmit buffer
+ * @rxbuf Receive buffer
+ * @scan Triggered buffer. Four channel 16-bit data + 64-bit timestamp
+ */
+struct sca3300_data {
+	struct spi_device *spi;
+	u32 opmode;
+	struct mutex lock;
+	u8 txbuf[4];
+	u8 rxbuf[4];
+	struct {
+		s16 channels[4];
+		s64 ts __aligned(sizeof(s64));
+	} scan;
+};
+
+DECLARE_CRC8_TABLE(sca3300_crc_table);
+
+static int sca3300_transfer(struct sca3300_data *sca_data, int *val)
+{
+	struct spi_delay delay = {.value = 10, .unit = SPI_DELAY_UNIT_USECS};
+	int32_t ret;
+	int rs;
+	u8 crc;
+	struct spi_transfer xfers[2] = {
+		{
+			.tx_buf = sca_data->txbuf,
+			.rx_buf = NULL,
+			.len = ARRAY_SIZE(sca_data->txbuf),
+			.delay = delay,
+			.cs_change = 1,
+		},
+		{
+			.tx_buf = NULL,
+			.rx_buf = sca_data->rxbuf,
+			.len = ARRAY_SIZE(sca_data->rxbuf),
+			.delay = delay,
+			.cs_change = 0,
+		}
+	};
+
+	/* inverted crc value as described in device data sheet */
+	crc = ~crc8(sca3300_crc_table, &sca_data->txbuf[0], 3, CRC8_INIT_VALUE);
+	sca_data->txbuf[3] = crc;
+
+	ret = spi_sync_transfer(sca_data->spi, xfers, 2);
+	if (ret < 0) {
+		dev_err(&sca_data->spi->dev,
+			"transfer error, error: %d\n", ret);
+		return -EIO;
+	}
+
+	crc = ~crc8(sca3300_crc_table, &sca_data->rxbuf[0], 3, CRC8_INIT_VALUE);
+	if (sca_data->rxbuf[3] != crc) {
+		dev_err(&sca_data->spi->dev, "CRC checksum mismatch");
+		return -EIO;
+	}
+
+	/* get return status */
+	rs = sca_data->rxbuf[0] & 0x03;
+	if (rs == SCA3300_RS_ERROR)
+		return rs;
+
+	*val = (s16)(sca_data->rxbuf[2] | (sca_data->rxbuf[1] << 8));
+
+	return 0;
+}
+
+static int sca3300_read_reg(struct sca3300_data *sca_data, u8 reg, int *val)
+{
+	int ret;
+
+	mutex_lock(&sca_data->lock);
+	sca_data->txbuf[0] = SCA3300_X_READ | (reg << 2);
+	ret = sca3300_transfer(sca_data, val);
+	if (ret > 0) {
+		sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS << 2);
+		ret = sca3300_transfer(sca_data, val);
+		/* status 0 = startup, 0x2 = mode change */
+		if (ret > 0 && *val != 0 && *val != 0x2) {
+			dev_err_ratelimited(&sca_data->spi->dev,
+					    "device status: %x\n",
+					    (u16)*val);
+			mutex_unlock(&sca_data->lock);
+			return -EIO;
+		}
+		if (ret > 0)
+			ret = 0;
+	}
+	mutex_unlock(&sca_data->lock);
+
+	return ret;
+}
+
+static int sca3300_write_reg(struct sca3300_data *sca_data, u8 reg, int val)
+{
+	int reg_val = 0;
+	int ret;
+
+	mutex_lock(&sca_data->lock);
+	sca_data->txbuf[0] = SCA3300_X_WRITE | (reg << 2);
+	sca_data->txbuf[1] = val >> 8;
+	sca_data->txbuf[2] = val & 0xFF;
+	ret = sca3300_transfer(sca_data, &reg_val);
+	if (ret > 0) {
+		sca_data->txbuf[0] = SCA3300_X_READ | (SCA3300_REG_STATUS << 2);
+		ret = sca3300_transfer(sca_data, &reg_val);
+		/* status 0 = startup, 0x2 = mode change */
+		if (ret > 0 && reg_val != 0 && reg_val != 0x2) {
+			dev_err_ratelimited(&sca_data->spi->dev,
+					    "device status: %x\n",
+					    (u16)reg_val);
+			mutex_unlock(&sca_data->lock);
+			return -EIO;
+		}
+		if (ret > 0)
+			ret = 0;
+	}
+	mutex_unlock(&sca_data->lock);
+
+	return ret;
+}
+
+static int sca3300_write_raw(struct iio_dev *indio_dev,
+			     struct iio_chan_spec const *chan,
+			     int val, int val2, long mask)
+{
+	struct sca3300_data *data = iio_priv(indio_dev);
+
+	switch (mask) {
+	case IIO_CHAN_INFO_SCALE:
+		if (val < 0 || val > 3)
+			return -EINVAL;
+		return sca3300_write_reg(data, SCA3300_REG_MODE, val);
+	default:
+		return -EINVAL;
+	}
+}
+
+static int sca3300_read_raw(struct iio_dev *indio_dev,
+			    struct iio_chan_spec const *chan,
+			    int *val, int *val2, long mask)
+{
+	struct sca3300_data *data = iio_priv(indio_dev);
+	int ret;
+	int reg_val;
+
+	switch (mask) {
+	case IIO_CHAN_INFO_RAW:
+		ret = sca3300_read_reg(data, chan->address, val);
+		if (ret < 0)
+			return ret;
+		return IIO_VAL_INT;
+	case IIO_CHAN_INFO_SCALE:
+		ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
+		if (ret < 0)
+			return ret;
+		*val = sca3300_accel_scale[reg_val];
+		return IIO_VAL_INT;
+	case IIO_CHAN_INFO_PROCESSED:
+		ret = sca3300_read_reg(data, SCA3300_REG_MODE, &reg_val);
+		if (ret < 0)
+			return ret;
+		*val2 = sca3300_accel_scale[reg_val];
+		ret = sca3300_read_reg(data, chan->address, val);
+		if (ret < 0)
+			return ret;
+		return IIO_VAL_FRACTIONAL;
+	default:
+		return -EINVAL;
+	}
+}
+
+static irqreturn_t sca3300_trigger_handler(int irq, void *p)
+{
+	struct iio_poll_func *pf = p;
+	struct iio_dev *indio_dev = pf->indio_dev;
+	struct sca3300_data *data = iio_priv(indio_dev);
+	s64 time_ns = iio_get_time_ns(indio_dev);
+	int bit, ret, val, i = 0;
+
+	for_each_set_bit(bit, indio_dev->active_scan_mask,
+			 indio_dev->masklength) {
+		ret = sca3300_read_reg(data, sca3300_channels[bit].address,
+				       &val);
+		if (ret < 0)
+			goto out;
+		if (ARRAY_SIZE(data->scan.channels) > i)
+			((s16 *)data->scan.channels)[i++] = val;
+	}
+
+	iio_push_to_buffers_with_timestamp(indio_dev, &data->scan, time_ns);
+out:
+	iio_trigger_notify_done(indio_dev->trig);
+
+	return IRQ_HANDLED;
+}
+
+static int sca3300_init(struct sca3300_data *sca_data,
+			struct iio_dev *indio_dev)
+{
+	int ret;
+	int value = 0;
+
+	if (sca_data->opmode < 1 || sca_data->opmode > 4)
+		return -EINVAL;
+
+	ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
+				SCA3300_VALUE_SW_RESET);
+	if (ret != 0)
+		return ret;
+	usleep_range(2e3, 10e3);
+
+	ret = sca3300_write_reg(sca_data, SCA3300_REG_MODE,
+				sca_data->opmode - 1);
+	if (ret != 0)
+		return ret;
+	msleep(100);
+	ret = sca3300_read_reg(sca_data, SCA3300_REG_WHOAMI, &value);
+	if (ret != 0)
+		return ret;
+
+	if (value != SCA3300_DEVICE_ID) {
+		dev_err(&sca_data->spi->dev, "device id not expected value\n");
+		return -EIO;
+	}
+	return 0;
+}
+
+static int sca3300_debugfs_reg_access(struct iio_dev *indio_dev,
+				      unsigned int reg, unsigned int writeval,
+				      unsigned int *readval)
+{
+	struct sca3300_data *data = iio_priv(indio_dev);
+	int value;
+	int ret;
+
+	if (reg > 0x1f)
+		return -EINVAL;
+
+	if (!readval)
+		return sca3300_write_reg(data, reg, writeval);
+
+	ret = sca3300_read_reg(data, reg, &value);
+	if (ret < 0)
+		return ret;
+
+	*readval = (unsigned int)value;
+
+	return 0;
+}
+
+static const struct iio_info sca3300_info = {
+	.read_raw = sca3300_read_raw,
+	.write_raw = sca3300_write_raw,
+	.debugfs_reg_access = &sca3300_debugfs_reg_access,
+};
+
+static int sca3300_probe(struct spi_device *spi)
+{
+	struct sca3300_data *sca_data;
+	struct iio_dev *indio_dev;
+	int ret;
+
+	indio_dev = devm_iio_device_alloc(&spi->dev, sizeof(*sca_data));
+	if (!indio_dev) {
+		dev_err(&spi->dev,
+			"failed to allocate memory for iio device\n");
+		return -ENOMEM;
+	}
+
+	sca_data = iio_priv(indio_dev);
+	mutex_init(&sca_data->lock);
+	sca_data->spi = spi;
+	spi_set_drvdata(spi, indio_dev);
+
+	crc8_populate_msb(sca3300_crc_table, SCA3300_CRC8_POLYNOMIAL);
+
+	indio_dev->dev.parent = &spi->dev;
+	indio_dev->info = &sca3300_info;
+	indio_dev->name = SCA3300_ALIAS;
+	indio_dev->modes = INDIO_DIRECT_MODE | INDIO_BUFFER_TRIGGERED;
+	indio_dev->channels = sca3300_channels;
+	indio_dev->num_channels = ARRAY_SIZE(sca3300_channels);
+	indio_dev->available_scan_masks = sca3300_scan_masks;
+
+	if (spi->dev.of_node) {
+		ret = of_property_read_u32(spi->dev.of_node, "murata,opmode",
+					   &sca_data->opmode);
+		if (ret < 0)
+			return ret;
+	}
+
+	ret = sca3300_init(sca_data, indio_dev);
+	if (ret < 0) {
+		dev_err(&spi->dev, "failed to init device, error: %d\n", ret);
+		return ret;
+	}
+
+	ret = iio_triggered_buffer_setup(indio_dev, iio_pollfunc_store_time,
+					 sca3300_trigger_handler, NULL);
+	if (ret < 0) {
+		dev_err(&spi->dev,
+			"iio triggered buffer setup failed, error: %d\n", ret);
+		return ret;
+	}
+
+	ret = devm_iio_device_register(&spi->dev, indio_dev);
+	if (ret < 0) {
+		dev_err(&spi->dev, "iio device register failed, error: %d\n",
+			ret);
+		iio_triggered_buffer_cleanup(indio_dev);
+		return ret;
+	}
+
+	return 0;
+}
+
+static int sca3300_remove(struct spi_device *spi)
+{
+	struct iio_dev *indio_dev = spi_get_drvdata(spi);
+
+	iio_triggered_buffer_cleanup(indio_dev);
+	return 0;
+}
+
+static const struct of_device_id sca3300_dt_ids[] = {
+	{ .compatible = "murata,sca3300"},
+	{},
+};
+MODULE_DEVICE_TABLE(of, sca3300_dt_ids);
+
+static struct spi_driver sca3300_driver = {
+	.driver = {
+		.name		= SCA3300_ALIAS,
+		.owner		= THIS_MODULE,
+		.of_match_table = of_match_ptr(sca3300_dt_ids),
+	},
+
+	.probe	= sca3300_probe,
+	.remove	= sca3300_remove,
+};
+
+module_spi_driver(sca3300_driver);
+
+MODULE_AUTHOR("Tomas Melin <[e-mail]>");
+MODULE_DESCRIPTION("Murata SCA3300 SPI Accelerometer");
+MODULE_LICENSE("GPL v2");