Message-ID: <20191223160255.742eb6e4@archlinux>
Date:   Mon, 23 Dec 2019 16:02:55 +0000
From:   Jonathan Cameron <jic23@...nel.org>
To:     YuDong Zhang <mtwget@...il.com>
Cc:     knaack.h@....de, lars@...afoo.de, pmeerw@...erw.net,
        tduszyns@...il.com, mike.looijmans@...ic.nl,
        matt.ranostay@...sulko.com, gregkh@...uxfoundation.org,
        tglx@...utronix.de, arnd@...db.de, linux-iio@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 1/2] iio: chemical: add support for Dynament Premier
 series single gas sensor

On Tue, 17 Dec 2019 18:24:33 +0800
YuDong Zhang <mtwget@...il.com> wrote:

> Add support for Dynament Premier series single gas sensor.
> 
> Signed-off-by: YuDong Zhang <mtwget@...il.com>
Hi YuDong Zhang,

A few additional comments in here.  

Thanks,

Jonathan

> ---
>  MAINTAINERS                    |   5 +
>  drivers/iio/chemical/Kconfig   |  10 +
>  drivers/iio/chemical/Makefile  |   1 +
>  drivers/iio/chemical/premier.c | 366 +++++++++++++++++++++++++++++++++
>  4 files changed, 382 insertions(+)
>  create mode 100644 drivers/iio/chemical/premier.c
> 
> diff --git a/MAINTAINERS b/MAINTAINERS
> index a049abccaa26..ae228ac7adc9 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -5792,6 +5792,11 @@ S:	Maintained
>  F:	drivers/media/usb/dvb-usb-v2/dvb_usb*
>  F:	drivers/media/usb/dvb-usb-v2/usb_urb.c
>  
> +DYNAMENT PREMIER SERIES SINGLE GAS SENSOR DRIVER
> +M:	YuDong Zhang <mtwget@...il.com>
> +S:	Maintained
> +F:	drivers/iio/chemical/premier.c
> +
>  DYNAMIC DEBUG
>  M:	Jason Baron <jbaron@...mai.com>
>  S:	Maintained
> diff --git a/drivers/iio/chemical/Kconfig b/drivers/iio/chemical/Kconfig
> index fa4586037bb8..93c0c108245b 100644
> --- a/drivers/iio/chemical/Kconfig
> +++ b/drivers/iio/chemical/Kconfig
> @@ -62,6 +62,16 @@ config IAQCORE
>  	  iAQ-Core Continuous/Pulsed VOC (Volatile Organic Compounds)
>  	  sensors
>  
> +config PREMIER
> +	tristate "Dynament Premier series sensor"
> +	depends on SERIAL_DEV_BUS
> +	help
> +	  Say Y here to build support for the Dynament Premier
> +	  series sensor.
> +
> +	  To compile this driver as a module, choose M here: the module will
> +	  be called premier.
> +
>  config PMS7003
>  	tristate "Plantower PMS7003 particulate matter sensor"
>  	depends on SERIAL_DEV_BUS
> diff --git a/drivers/iio/chemical/Makefile b/drivers/iio/chemical/Makefile
> index f97270bc4034..c8e779d7cf4a 100644
> --- a/drivers/iio/chemical/Makefile
> +++ b/drivers/iio/chemical/Makefile
> @@ -10,6 +10,7 @@ obj-$(CONFIG_BME680_I2C) += bme680_i2c.o
>  obj-$(CONFIG_BME680_SPI) += bme680_spi.o
>  obj-$(CONFIG_CCS811)		+= ccs811.o
>  obj-$(CONFIG_IAQCORE)		+= ams-iaq-core.o
> +obj-$(CONFIG_PREMIER)		+= premier.o
>  obj-$(CONFIG_PMS7003) += pms7003.o
>  obj-$(CONFIG_SENSIRION_SGP30)	+= sgp30.o
>  obj-$(CONFIG_SPS30) += sps30.o
> diff --git a/drivers/iio/chemical/premier.c b/drivers/iio/chemical/premier.c
> new file mode 100644
> index 000000000000..a226dd9d78cb
> --- /dev/null
> +++ b/drivers/iio/chemical/premier.c
> @@ -0,0 +1,366 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Dynament Premier series single gas sensor driver
> + *
> + * Copyright (c) YuDong Zhang <mtwget@...il.com>
> + */
> +
> +#include <asm/unaligned.h>
> +#include <linux/completion.h>
> +#include <linux/device.h>
> +#include <linux/errno.h>
> +#include <linux/iio/iio.h>
> +#include <linux/jiffies.h>
> +#include <linux/kernel.h>
> +#include <linux/mod_devicetable.h>
> +#include <linux/module.h>
> +#include <linux/mutex.h>
> +#include <linux/of.h>
> +#include <linux/regulator/consumer.h>
> +#include <linux/serdev.h>
> +
> +#define PREMIER_DRIVER_NAME "dynament-premier"
> +
> +#define PREMIER_DLE (0x10)
> +#define PREMIER_CMD_RD (0x13)
> +#define PREMIER_CMD_NAK (0x19)
> +#define PREMIER_CMD_DAT (0x1a)
> +#define PREMIER_EOF (0x1f)
> +
> +#define PREMIER_TIMEOUT msecs_to_jiffies(6000)
> +
> +/*
> + * commands have following format:
> + *
> + * +-----+-----+---------+-----+-----+-----------+-----------+
> + * | DLE | CMD | PAYLOAD | DLE | EOF | CKSUM MSB | CKSUM LSB |
> + * +-----+-----+---------+-----+-----+-----------+-----------+
> + */
> +static const u8 premier_cmd_read_live_data_simple[] = { 0x10, 0x13, 0x06, 0x10,
> +							0x1F, 0x00, 0x58 };
> +
> +struct premier_frame {
> +	u8 state;
> +	u8 is_dat;
> +	u8 is_nak;
> +	u8 data_len;
> +	u8 vi, si, gi, gj;
> +	u8 gas[4];
> +	u8 byte_stuffing;
> +	u8 checksum_received[2];
> +	u16 checksum_calculated;
> +};
> +
> +struct premier_data {
> +	struct serdev_device *serdev;
> +	struct premier_frame frame;
> +	struct completion frame_ready;
> +	struct mutex lock; /* must be held whenever state gets touched */
> +	struct regulator *vcc;
> +};
> +
> +static int premier_do_cmd_read_live_data(struct premier_data *state)
> +{
> +	int ret;
> +
> +	ret = serdev_device_write(state->serdev,
> +				  premier_cmd_read_live_data_simple,
> +				  sizeof(premier_cmd_read_live_data_simple),
> +				  PREMIER_TIMEOUT);
> +	if (ret < sizeof(premier_cmd_read_live_data_simple))
> +		return ret < 0 ? ret : -EIO;
> +
> +	ret = wait_for_completion_interruptible_timeout(&state->frame_ready,
> +							PREMIER_TIMEOUT);
> +
> +	if (!ret)
> +		ret = -ETIMEDOUT;
> +
> +	return ret < 0 ? ret : 0;
> +}
> +
> +static s32 premier_float_to_int_clamped(const u8 *fp)
> +{
> +	int val = get_unaligned_le32(fp);
> +	int mantissa = val & GENMASK(22, 0);
> +	/* this is fine since passed float is always non-negative */
> +	int exp = val >> 23;
> +	int fraction, shift;
> +
> +	/* special case 0 */
> +	if (!exp && !mantissa)
> +		return 0;
> +
> +	exp -= 127;
> +	if (exp < 0) {
> +		/* return values ranging from 1 to 99 */
> +		return ((((1 << 23) + mantissa) * 100) >> 23) >> (-exp);
> +	}
> +
> +	/* return values ranging from 100 to int_max */
> +	shift = 23 - exp;
> +	val = (1 << exp) + (mantissa >> shift);
> +
> +	fraction = mantissa & GENMASK(shift - 1, 0);
> +
> +	return val * 100 + ((fraction * 100) >> shift);
> +}
> +
> +static int premier_read_raw(struct iio_dev *indio_dev,
> +			    struct iio_chan_spec const *chan, int *val,
> +			    int *val2, long mask)
> +{
> +	struct premier_data *state = iio_priv(indio_dev);
> +	struct premier_frame *frame = &state->frame;
> +	int ret;
> +	s32 val_tmp;
> +
> +	switch (mask) {
> +	case IIO_CHAN_INFO_PROCESSED:
> +
> +		mutex_lock(&state->lock);
> +		ret = premier_do_cmd_read_live_data(state);
> +		if (ret) {
> +			mutex_unlock(&state->lock);
> +			return ret;
> +		}
> +		val_tmp = premier_float_to_int_clamped(frame->gas);
> +		mutex_unlock(&state->lock);
> +
> +		*val = val_tmp / 100;
> +		*val2 = (val_tmp % 100) * 10000;
> +		return IIO_VAL_INT_PLUS_MICRO;
> +	default:
> +		return -EINVAL;
> +	}
> +
> +	return -EINVAL;
> +}
> +
> +static const struct iio_info premier_info = {
> +	.read_raw = premier_read_raw,
> +};
> +
> +static const struct iio_chan_spec premier_channels[] = {
> +	{
> +		.type = IIO_MASSCONCENTRATION,
> +		.channel = 1,
> +		.channel2 = IIO_MOD_CO2,
> +		.scan_index = -1,
> +		.info_mask_separate = BIT(IIO_CHAN_INFO_PROCESSED),
> +		.modified = 1,
> +	},
> +	IIO_CHAN_SOFT_TIMESTAMP(0),
This is odd.  The scan_index = -1 above tells the device not to expose
the channel via the buffered interface.   That means we just have
a timestamp channel exposed via that interface.  What use is that?

You don't register that interface at all. Hence no point in either
providing a timestamp here, or in providing '.scan_index = -1' for
the massconcentration channel.  It doesn't add anything.

> +};
> +
> +static int premier_receive_buf(struct serdev_device *serdev,
> +			       const unsigned char *buf, size_t size)
> +{
> +	struct iio_dev *indio_dev = serdev_device_get_drvdata(serdev);
> +	struct premier_data *state = iio_priv(indio_dev);
> +	struct premier_frame *frame = &state->frame;
> +	int i;
> +
> +	for (i = 0; i < size; i++) {
> +		if (frame->state > 0 && frame->state <= 7)
> +			frame->checksum_calculated += buf[i];
> +
> +		switch (frame->state) {
> +		case 0:
> +			if (buf[i] == PREMIER_DLE) {
> +				frame->is_dat = 0;
> +				frame->is_nak = 0;
> +				frame->checksum_calculated = buf[i];
> +				/* We don't initialize checksum_calculated in
> +				 * the last state in case we didn't go
> +				 * there because of noise
> +				 */
> +				frame->state++;
> +			}
> +			break;
> +		case 1:
> +			/*
> +			 * If noise corrupts a byte in the FSM sequence,
> +			 * we loop between state 0 and 1,
> +			 * until we have a valid sequence of DLE&DAT or DLE&NAK
> +			 */
> +			if (buf[i] == PREMIER_CMD_DAT) {
> +				frame->is_dat = 1;
> +				frame->state++;
> +			} else if (buf[i] == PREMIER_CMD_NAK) {
> +				frame->is_nak = 1;
> +				frame->state++;
> +			} else
> +				frame->state = 0;
> +			break;
> +		case 2:
> +			if (frame->is_nak)
> +				frame->state = 0;
> +			else if (frame->is_dat) {
> +				frame->data_len = buf[i] - 4;
> +				/* remove version and status bytes from count */
> +				if (frame->data_len < 4)
> +					frame->state = 0;
> +				/* we check for the upper limit in state 5 */
> +				else
> +					frame->state++;
> +			} else
> +				frame->state = 0;
> +			break;
> +		case 3:
> +			/* Just do nothing for 2 rounds to bypass
> +			 * the 2 version bytes
> +			 */
> +			if (frame->vi < 2 - 1)
> +				frame->vi++;
> +			else {
> +				frame->vi = 0;
> +				frame->state++;
> +			}
> +			break;
> +		case 4:
> +			if (frame->si < 2 - 1)
> +				frame->si++;
> +			else {
> +				frame->si = 0;
> +				frame->state++;
> +			}
> +			break;
> +		case 5:
> +			if (frame->gi < frame->data_len - 1) {
> +				if (buf[i] != PREMIER_DLE ||
> +				    frame->byte_stuffing) {
> +					frame->gas[frame->gj] = buf[i];
> +					frame->byte_stuffing = 0;
> +					frame->gj++;
> +					if (frame->gj >= 4)
> +						frame->state = 0;
> +					/* Don't violate array limits
> +					 * if data_len corrupt
> +					 */
> +				} else
> +					frame->byte_stuffing = 1;
> +				frame->gi++;
> +			} else {
> +				frame->gas[frame->gj] = buf[i];
> +				frame->byte_stuffing = 0;
> +				frame->gi = 0;
> +				frame->gj = 0;
> +				frame->state++;
> +			}
> +			break;
> +		case 6:
> +			if (buf[i] == PREMIER_DLE)
> +				frame->state++;
> +			else
> +				frame->state = 0;
> +			break;
> +		case 7:
> +			if (buf[i] == PREMIER_EOF)
> +				frame->state++;
> +			else
> +				frame->state = 0;
> +			break;
> +		case 8:
> +			frame->checksum_received[1] = buf[i];
> +
> +			frame->state++;
> +			break;
> +		case 9:
> +			frame->checksum_received[0] = buf[i];
> +
> +			if (frame->checksum_calculated ==
> +			    get_unaligned_le16(frame->checksum_received))
> +				complete(&state->frame_ready);
> +
> +			frame->state = 0;
> +			break;
> +		}
> +	}
> +
> +	return size;
> +}
> +
> +static const struct serdev_device_ops premier_serdev_ops = {
> +	.receive_buf = premier_receive_buf,
> +	.write_wakeup = serdev_device_write_wakeup,
> +};
> +
> +static int premier_probe(struct serdev_device *serdev)
> +{
> +	struct premier_data *state;
> +	struct iio_dev *indio_dev;
> +	int ret;
> +
> +	indio_dev = devm_iio_device_alloc(&serdev->dev, sizeof(*state));
> +	if (!indio_dev)
> +		return -ENOMEM;
> +
> +	state = iio_priv(indio_dev);
> +	serdev_device_set_drvdata(serdev, indio_dev);
> +	state->serdev = serdev;
> +	indio_dev->dev.parent = &serdev->dev;
> +	indio_dev->info = &premier_info;
> +	indio_dev->name = PREMIER_DRIVER_NAME;
> +	indio_dev->channels = premier_channels;
> +	indio_dev->num_channels = ARRAY_SIZE(premier_channels);
> +	indio_dev->modes = INDIO_DIRECT_MODE;
> +
> +	mutex_init(&state->lock);
> +	init_completion(&state->frame_ready);
> +
> +	state->vcc = devm_regulator_get(&serdev->dev, "vcc");
> +	if (IS_ERR(state->vcc)) {
> +		ret = PTR_ERR(state->vcc);
> +		return ret;
> +	}
> +
> +	serdev_device_set_client_ops(serdev, &premier_serdev_ops);
> +	ret = devm_serdev_device_open(&serdev->dev, serdev);
> +	if (ret)
> +		return ret;
> +
> +	serdev_device_set_baudrate(serdev, 9600);

Please check return codes from these and handle them.

> +	serdev_device_set_flow_control(serdev, false);
> +
> +	ret = serdev_device_set_parity(serdev, SERDEV_PARITY_NONE);
> +	if (ret)
> +		return ret;
> +
> +	if (state->vcc) {

I don't think you ever need to check if a regulator is null or not.
If you have regulators built in then you will get either a real
regulator, or a stub. Either of those will be fine.

If you don't have regulators built in you will indeed get state->vcc == NULL.
However, then regulator_enable will also be stubbed out so: 

https://elixir.bootlin.com/linux/latest/source/include/linux/regulator/consumer.h#L398

Result is you can drop checks on validity of state->vcc.

> +		ret = regulator_enable(state->vcc);
> +		if (ret)
> +			return ret;
> +	}
> +
> +	return devm_iio_device_register(&serdev->dev, indio_dev);
This just introduced a race condition.   When removing, any managed function
unwind callbacks (effectively iio_device_unregister here) will be called
after the remove function has finished.  In this particular case that means
you only remove the userspace interface after you have removed the power.
Any reads that occur in that interval will get unpredictable results.

The basic rule on this is you can only use managed functions until you first
have to use a non managed function.  Once that occurs you must stop using them.

Note however, that devm_add_action_or_reset can be used to safe add a callback
to your probe code, making additional functions managed.  Look at other drivers
to see how this works.

> +}
> +
> +static void premier_remove(struct serdev_device *serdev)
> +{
> +	struct iio_dev *indio_dev = serdev_device_get_drvdata(serdev);
> +	struct premier_data *state = iio_priv(indio_dev);
> +
> +	if (state->vcc)
> +		regulator_disable(state->vcc);
> +}
> +
> +static const struct of_device_id premier_of_match[] = {
> +	{ .compatible = "dynament,premier" },
> +	{}
> +};
> +MODULE_DEVICE_TABLE(of, premier_of_match);
> +
> +static struct serdev_device_driver premier_driver = {
> +	.driver = {
> +		.name = PREMIER_DRIVER_NAME,
> +		.of_match_table = premier_of_match,
> +	},
> +	.probe = premier_probe,
> +	.remove = premier_remove,
> +};
> +module_serdev_device_driver(premier_driver);
> +
> +MODULE_AUTHOR("YuDong Zhang <mtwget@...il.com>");
> +MODULE_DESCRIPTION("Dynament Premier series single gas sensor driver");
> +MODULE_LICENSE("GPL v2");

Powered by blists - more mailing lists