Date:   Mon, 8 Jun 2020 18:34:47 +0200
From:   Álvaro Fernández Rojas <noltari@...il.com>
To:     Guenter Roeck <linux@...ck-us.net>
Cc:     jdelvare@...e.com, thierry.reding@...il.com,
        u.kleine-koenig@...gutronix.de, linux-kernel@...r.kernel.org,
        linux-hwmon@...r.kernel.org, linux-pwm@...r.kernel.org
Subject: Re: [PATCH 2/2] hwmon: Add SFR NB6 sensor driver

Hi Guenter,

Thanks for your comments, I will reimplement it as MFD.

Regards,
Álvaro.

> El 8 jun 2020, a las 16:03, Guenter Roeck <linux@...ck-us.net> escribió:
> 
> On 6/7/20 11:25 AM, Álvaro Fernández Rojas wrote:
>> SFR NeufBox 6 is equipped with an I2C connected chip to monitor voltage,
>> temperature and other stats. It also has a PWM LEDs that can be manually
>> controlled or set to specific hardware monitoring modes.
>> 
>> Signed-off-by: Álvaro Fernández Rojas <noltari@...il.com>
>> ---
>> drivers/hwmon/Kconfig     |  11 +
>> drivers/hwmon/Makefile    |   1 +
>> drivers/hwmon/nb6-hwmon.c | 466 ++++++++++++++++++++++++++++++++++++++
> 
> Missing Documentation/hwmon/nb6-hwmon.rst
> 
>> 3 files changed, 478 insertions(+)
>> create mode 100644 drivers/hwmon/nb6-hwmon.c
>> 
>> diff --git a/drivers/hwmon/Kconfig b/drivers/hwmon/Kconfig
>> index 288ae9f63588..5523de75868d 100644
>> --- a/drivers/hwmon/Kconfig
>> +++ b/drivers/hwmon/Kconfig
>> @@ -1092,6 +1092,17 @@ config SENSORS_MENF21BMC_HWMON
>> 	  This driver can also be built as a module. If so the module
>> 	  will be called menf21bmc_hwmon.
>> 
>> +config SENSORS_NB6_HWMON
>> +	tristate "SFR NeufBox 6 (NB6) Hardware Monitoring"
>> +	depends on I2C
>> +	depends on PWM
>> +	help
>> +	  If you say yes here you get support for SFR NeufBox 6 PWM LEDs,
>> +	  temperature and voltage sensors.
>> +
> This driver should not support LEDs.
> 
>> +	  This driver can also be built as a module. If so, the module
>> +	  will be called nb6-hwmon.
>> +
>> config SENSORS_ADCXX
>> 	tristate "National Semiconductor ADCxxxSxxx"
>> 	depends on SPI_MASTER
>> diff --git a/drivers/hwmon/Makefile b/drivers/hwmon/Makefile
>> index 3e32c21f5efe..5a35fde42bd8 100644
>> --- a/drivers/hwmon/Makefile
>> +++ b/drivers/hwmon/Makefile
>> @@ -141,6 +141,7 @@ obj-$(CONFIG_SENSORS_MCP3021)	+= mcp3021.o
>> obj-$(CONFIG_SENSORS_TC654)	+= tc654.o
>> obj-$(CONFIG_SENSORS_MLXREG_FAN) += mlxreg-fan.o
>> obj-$(CONFIG_SENSORS_MENF21BMC_HWMON) += menf21bmc_hwmon.o
>> +obj-$(CONFIG_SENSORS_NB6_HWMON)	+= nb6-hwmon.o
>> obj-$(CONFIG_SENSORS_NCT6683)	+= nct6683.o
>> obj-$(CONFIG_SENSORS_NCT6775)	+= nct6775.o
>> obj-$(CONFIG_SENSORS_NCT7802)	+= nct7802.o
>> diff --git a/drivers/hwmon/nb6-hwmon.c b/drivers/hwmon/nb6-hwmon.c
>> new file mode 100644
>> index 000000000000..aebb6ebbf3a8
>> --- /dev/null
>> +++ b/drivers/hwmon/nb6-hwmon.c
>> @@ -0,0 +1,466 @@
>> +// SPDX-License-Identifier: GPL-2.0-or-later
>> +/*
>> + * Driver for SFR NeufBox 6 Hardware Monitoring
>> + *
>> + * Copyright 2020 Álvaro Fernández Rojas <noltari@...il.com>
>> + */
>> +#include <linux/delay.h>
>> +#include <linux/kernel.h>
>> +#include <linux/init.h>
>> +#include <linux/module.h>
>> +#include <linux/i2c.h>
>> +#include <linux/hwmon.h>
>> +#include <linux/err.h>
>> +#include <linux/mutex.h>
>> +#include <linux/pwm.h>
>> +#include <linux/sysfs.h>
>> +
>> +#define NB6_LEDS_PWM_REG	0x10
>> +#define NB6_TEMP_REG		0x22
>> +#define NB6_VOLTAGE_REG		0x24
>> +#define NB6_DMESG_CTL_REG	0x26
>> +#define NB6_DMESG_VAL_REG	0x27
>> +#define NB6_LEDS_MODE_REG	0x31
>> +#define NB6_RELEASE_REG		0x90
>> +#define NB6_STATS_BOOT_REG	0xa0
>> +#define NB6_STATS_PANIC_REG	0xa1
>> +#define NB6_STATS_OOPS_REG	0xa2
>> +#define NB6_WDT_REG		0xee
>> +
>> +#define NB6_DELAY		1000
>> +#define NB6_DMESG_SIZE		512
>> +#define NB6_I2C_RETRIES		5
>> +#define NB6_LEDS_MODE_LEN	32
>> +#define NB6_LEDS_PWM_COUNT	9
>> +#define NB6_PWM(x)		(NB6_LEDS_PWM_REG + x)
>> +#define NB6_WDT_LEN		10
>> +
>> +#define ADC_quantum(Vref)	((1000 * (Vref)) / 1024)
>> +#define ADC_mV(Vref,x)		((ADC_quantum(Vref) * (x)) / 1000)
>> +#define ADC_Temperature(t)	(1000 * (100 * ADC_mV(1800, t)) / 349)
>> +#define MR1			82
>> +#define MR2			20
>> +#define ADC_Voltage(v)		((ADC_mV(2400, v) * ((10 * (MR1 + MR2)) / MR2)) / 10)
>> +
>> +struct nb6_data {
>> +	struct pwm_chip pwm;
>> +	struct i2c_client *i2c;
>> +	struct device *dev;
>> +	struct mutex lock;
>> +	u8 leds_mode;
>> +	u8 leds_pwm[NB6_LEDS_PWM_COUNT];
>> +	u8 release;
>> +	u8 stats_boot;
>> +	u8 stats_panic;
>> +	u8 stats_oops;
>> +	u16 temperature;
>> +	u16 voltage;
>> +	u8 watchdog;
>> +};
>> +
>> +enum LEDS_MODE {
>> +	LEDS_MODE_DISABLE,
>> +	LEDS_MODE_BOOT,
>> +	LEDS_MODE_BOOT_MAIN,
>> +	LEDS_MODE_BOOT_TFTP,
>> +	LEDS_MODE_BOOT_RESCUE,
>> +	LEDS_MODE_LOGIN,
>> +	LEDS_MODE_BURNING,
>> +	LEDS_MODE_DOWNLOAD,
>> +	LEDS_MODE_WDT_TEMPERATURE,
>> +	LEDS_MODE_WDT_VOLTAGE,
>> +	LEDS_MODE_PANIC,
>> +	LEDS_MODE_CONTROL,
>> +	LEDS_MODE_NUM
>> +};
>> +
>> +static char const *leds_modes_str[] = {
>> +	[LEDS_MODE_DISABLE] = "disable",
>> +	[LEDS_MODE_BOOT] = "boot",
>> +	[LEDS_MODE_BOOT_MAIN] = "boot-main",
>> +	[LEDS_MODE_BOOT_TFTP] = "boot-tftp",
>> +	[LEDS_MODE_BOOT_RESCUE] = "boot-rescue",
>> +	[LEDS_MODE_LOGIN] = "login",
>> +	[LEDS_MODE_BURNING] = "burning",
>> +	[LEDS_MODE_DOWNLOAD] = "downloading",
>> +	[LEDS_MODE_WDT_TEMPERATURE] = "wdt-temperature",
>> +	[LEDS_MODE_WDT_VOLTAGE] = "wdt-voltage",
>> +	[LEDS_MODE_PANIC] = "panic",
>> +	[LEDS_MODE_CONTROL] = "control",
>> +};
>> +
>> +/* I2C Helpers */
>> +
>> +static u8 nb6_readb(struct nb6_data *data, u8 addr, u8 val)
> 
> This will never return an error.
> 
>> +{
>> +	int status;
>> +	unsigned i;
>> +
>> +	for (i = 0; i < NB6_I2C_RETRIES; i++) {
>> +		status = i2c_smbus_read_byte_data(data->i2c, addr);
>> +		if (status >= 0)
>> +			return status;
>> +		udelay(NB6_DELAY);
>> +	}
>> +
> 
> Is that system really that bad ? That needs an explanation.
> 
>> +	dev_err(data->dev, "read error (%d): addr=0x%02x", status, addr);
>> +
>> +	return val;
> 
> Obviously this error case was never reached.
> 
>> +}
>> +
>> +static u16 nb6_readw(struct nb6_data *data, u8 addr, u16 val)
> 
> This will never return an error.
> 
>> +{
>> +	u8 tmp;
>> +
>> +	tmp = nb6_readb(data, addr, (val >> 8) & 0xff);
> 
> This ignores any received errors and, worse, converts errors
> into pretty much random data.
> 
>> +	val = (val & 0xff) | (tmp << 8);
>> +
>> +	tmp = nb6_readb(data, addr + 1, val & 0xff);
>> +	val = (val & 0xff00) | tmp;
>> +
>> +	return val;
>> +}
>> +
>> +static s32 nb6_writeb(struct nb6_data *data, u8 addr, u8 val)
>> +{
>> +	int status;
>> +	unsigned i;
>> +
>> +	for (i = 0; i < NB6_I2C_RETRIES; i++) {
>> +		status = i2c_smbus_write_byte_data(data->i2c, addr, val);
>> +		if (!status)
>> +			return 0;
>> +		udelay(NB6_DELAY);
>> +	}
>> +
>> +	dev_err(data->dev, "write error (%d): addr=0x%02x val=0x%02x", status, addr, val);
>> +
>> +	return status;
>> +}
>> +
>> +static inline void leds_mode_update(struct nb6_data *data, u8 val)
>> +{
>> +	if ((data->leds_mode != val) &&
>> +	    !nb6_writeb(data, NB6_LEDS_MODE_REG, val))
>> +		data->leds_mode = val;
>> +}
>> +
>> +static inline void leds_pwm_update(struct nb6_data *data, u8 id, u8 val)
>> +{
>> +	if ((data->leds_pwm[id] != val) &&
>> +	    !nb6_writeb(data, NB6_PWM(id), val))
>> +		data->leds_pwm[id] = val;
>> +}
>> +
>> +/* Hardware Monitoring */
>> +
>> +static ssize_t dmesg_show(struct device *dev, struct device_attribute *attr,
>> +			  char *buf)
>> +{
>> +	struct nb6_data *data = dev_get_drvdata(dev);
>> +	unsigned i;
>> +
>> +	mutex_lock(&data->lock);
>> +
>> +	if (nb6_writeb(data, NB6_DMESG_CTL_REG, 0)) {
>> +		mutex_unlock(&data->lock);
>> +		return -EINVAL;
>> +	}
>> +
>> +	for (i = 0; i < NB6_DMESG_SIZE; i++)
>> +		buf[i] = nb6_readb(data, NB6_DMESG_VAL_REG, ~0);
>> +
>> +	mutex_unlock(&data->lock);
>> +
>> +	*buf = '\0';
>> +
>> +	return i + 1;
>> +}
>> +
>> +static DEVICE_ATTR_RO(dmesg);
>> +
> 
> Non-standard attributes need to be explained and their need must be discussed.
> 
>> +static ssize_t leds_mode_show(struct device *dev,
>> +			      struct device_attribute *attr, char *buf)
>> +{
>> +	struct nb6_data *data = dev_get_drvdata(dev);
>> +	loff_t off = 0;
>> +	unsigned i;
>> +
>> +	mutex_lock(&data->lock);
>> +	data->leds_mode = nb6_readb(data, NB6_LEDS_MODE_REG, data->leds_mode);
>> +	mutex_unlock(&data->lock);
>> +
>> +	for (i = 0; i < ARRAY_SIZE(leds_modes_str); i++) {
>> +		off += sprintf(buf + off,
>> +			       (i == data->leds_mode) ? "[%s] " : "%s ",
>> +			       leds_modes_str[i]);
>> +	}
>> +
>> +	off += sprintf(buf + off, "\n");
>> +
>> +	return off;
>> +}
>> +
>> +static ssize_t leds_mode_store(struct device *dev,
>> +			       struct device_attribute *attr,
>> +			       const char *buf, size_t count)
>> +{
>> +	struct nb6_data *data = dev_get_drvdata(dev);
>> +	char _s[NB6_LEDS_MODE_LEN];
>> +	char *s;
>> +	unsigned i;
>> +
>> +	snprintf(_s, sizeof(_s), "%s", buf);
>> +	s = strstrip(_s);
>> +	for (i = 0; i < ARRAY_SIZE(leds_modes_str); i++) {
>> +		if (!strcmp(s, leds_modes_str[i])) {
>> +			leds_mode_update(data, i);
>> +			break;
>> +		}
>> +	}
>> +
>> +	return count;
>> +}
>> +
>> +static DEVICE_ATTR_RW(leds_mode);
>> +
> 
> I won't accept any attributes not associated with hardware monitoring.
> 
>> +static ssize_t release_show(struct device *dev, struct device_attribute *attr,
>> +			    char *buf)
>> +{
>> +	struct nb6_data *data = dev_get_drvdata(dev);
>> +
>> +	mutex_lock(&data->lock);
>> +	data->release = nb6_readb(data, NB6_RELEASE_REG, data->release);
>> +	mutex_unlock(&data->lock);
>> +
>> +	return sprintf(buf, "%u\n", data->release);
>> +}
>> +
>> +static DEVICE_ATTR_RO(release);
>> +
> 
> And I will most definitely not accept unexplained non-standard attributes.
> 
>> +static ssize_t stats_show(struct device *dev, struct device_attribute *attr,
>> +			  char *buf)
>> +{
>> +	struct nb6_data *data = dev_get_drvdata(dev);
>> +
>> +	mutex_lock(&data->lock);
>> +	data->stats_boot = nb6_readb(data, NB6_STATS_BOOT_REG,
>> +				     data->stats_boot);
>> +	data->stats_panic = nb6_readb(data, NB6_STATS_PANIC_REG,
>> +				      data->stats_panic);
>> +	data->stats_oops = nb6_readb(data, NB6_STATS_OOPS_REG,
>> +				     data->stats_oops);
>> +	mutex_unlock(&data->lock);
>> +
>> +	return sprintf(buf, "boot: %u\npanic: %u\noops: %u\n",
>> +		       data->stats_boot, data->stats_panic, data->stats_oops);
>> +}
>> +
>> +static DEVICE_ATTR_RO(stats);
>> +
>> +static ssize_t temperature_show(struct device *dev,
>> +				struct device_attribute *da, char *buf)
>> +{
>> +	struct nb6_data *data = dev_get_drvdata(dev);
>> +
>> +	mutex_lock(&data->lock);
>> +	data->temperature = nb6_readw(data, NB6_TEMP_REG, data->temperature);
>> +	mutex_unlock(&data->lock);
>> +
>> +	return sprintf(buf, "%u\n", ADC_Temperature(data->temperature));
>> +}
>> +
>> +static DEVICE_ATTR_RO(temperature);
>> +
>> +static ssize_t voltage_show(struct device *dev, struct device_attribute *attr,
>> +			    char *buf)
>> +{
>> +	struct nb6_data *data = dev_get_drvdata(dev);
>> +
>> +	mutex_lock(&data->lock);
>> +	data->voltage = nb6_readw(data, NB6_VOLTAGE_REG, data->voltage);
>> +	mutex_unlock(&data->lock);
>> +
>> +	return sprintf(buf, "%u\n", ADC_Voltage(data->voltage));
>> +}
>> +
>> +static DEVICE_ATTR_RO(voltage);
>> +
> 
> This are not standard hardware monitoring attributes.
> 
>> +static ssize_t watchdog_show(struct device *dev,
>> +			     struct device_attribute *attr, char *buf)
>> +{
>> +	struct nb6_data *data = dev_get_drvdata(dev);
>> +
>> +	mutex_lock(&data->lock);
>> +	data->watchdog = nb6_readb(data, NB6_WDT_REG, data->watchdog);
>> +	mutex_unlock(&data->lock);
>> +
>> +	return sprintf(buf, "%u\n", data->watchdog);
>> +}
>> +
>> +static ssize_t watchdog_store(struct device *dev,
>> +			      struct device_attribute *attr, const char *buf,
>> +			      size_t len)
>> +{
>> +	struct nb6_data *data = dev_get_drvdata(dev);
>> +	unsigned long val;
>> +	int ret;
>> +
>> +	ret = kstrtoul(buf, 0, &val);
>> +	if (ret < 0)
>> +		return ret;
>> +
>> +	mutex_lock(&data->lock);
>> +	if (!nb6_writeb(data, NB6_WDT_REG, val))
>> +		data->watchdog = val;
>> +	mutex_unlock(&data->lock);
>> +
>> +	return len;
>> +}
>> +
>> +static DEVICE_ATTR_RW(watchdog);
>> +
> 
> While there are situations where we accept a watchdog driver as part
> of a hardware monitoring driver, it has to be a standard watchdog
> driver, not a random attribute such as "watchdog" and "status".
> 
>> +static struct attribute *nb6_attrs[] = {
>> +	&dev_attr_dmesg.attr,
>> +	&dev_attr_leds_mode.attr,
>> +	&dev_attr_release.attr,
>> +	&dev_attr_stats.attr,
>> +	&dev_attr_temperature.attr,
>> +	&dev_attr_voltage.attr,
>> +	&dev_attr_watchdog.attr,
>> +	NULL,
>> +};
> 
> Not a single standard hardware monitoring attribute.
> 
>> +
>> +ATTRIBUTE_GROUPS(nb6);
>> +
>> +/* PWM */
>> +
>> +static inline struct nb6_data *to_nb6_pwm(struct pwm_chip *chip)
>> +{
>> +	return container_of(chip, struct nb6_data, pwm);
>> +}
>> +
>> +static int nb6_pwm_request(struct pwm_chip *chip, struct pwm_device *pwm)
>> +{
>> +	struct nb6_data *data = to_nb6_pwm(chip);
>> +
>> +	mutex_lock(&data->lock);
>> +	if (!nb6_writeb(data, NB6_PWM(pwm->hwpwm), 0))
>> +		data->leds_pwm[pwm->hwpwm] = 0;
>> +	else
>> +		data->leds_pwm[pwm->hwpwm] = ~0;
>> +	mutex_unlock(&data->lock);
>> +
>> +	return 0;
>> +}
>> +
>> +static void nb6_pwm_free(struct pwm_chip *chip, struct pwm_device *pwm)
>> +{
>> +	struct nb6_data *data = to_nb6_pwm(chip);
>> +	unsigned i;
>> +
>> +	mutex_lock(&data->lock);
>> +	for (i = 0; i < NB6_LEDS_PWM_COUNT; i++) {
>> +		if (data->leds_pwm[i]) {
>> +			leds_mode_update(data, LEDS_MODE_DISABLE);
>> +			break;
>> +		}
>> +	}
>> +	mutex_unlock(&data->lock);
>> +}
>> +
>> +static int nb6_pwm_config(struct pwm_chip *chip, struct pwm_device *pwm,
>> +			  int duty_ns, int period_ns)
>> +{
>> +	struct nb6_data *data = to_nb6_pwm(chip);
>> +
>> +	mutex_lock(&data->lock);
>> +	leds_mode_update(data, LEDS_MODE_CONTROL);
>> +	leds_pwm_update(data, pwm->hwpwm, duty_ns);
>> +	mutex_unlock(&data->lock);
>> +
>> +	return 0;
>> +}
>> +
>> +static int nb6_pwm_enable(struct pwm_chip *chip, struct pwm_device *pwm)
>> +{
>> +	struct nb6_data *data = to_nb6_pwm(chip);
>> +
>> +	mutex_lock(&data->lock);
>> +	leds_mode_update(data, LEDS_MODE_CONTROL);
>> +	mutex_unlock(&data->lock);
>> +
>> +	return 0;
>> +}
>> +
>> +static void nb6_pwm_disable(struct pwm_chip *chip, struct pwm_device *pwm)
>> +{
>> +	struct nb6_data *data = to_nb6_pwm(chip);
>> +
>> +	mutex_lock(&data->lock);
>> +	leds_pwm_update(data, pwm->hwpwm, 0);
>> +	mutex_unlock(&data->lock);
>> +}
>> +
>> +static const struct pwm_ops nb6_pwm_ops = {
>> +	.request = nb6_pwm_request,
>> +	.free = nb6_pwm_free,
>> +	.config = nb6_pwm_config,
>> +	.enable = nb6_pwm_enable,
>> +	.disable = nb6_pwm_disable,
>> +	.owner = THIS_MODULE,
>> +};
>> +
>> +/* Driver */
>> +
>> +static int nb6_hwmon_probe(struct i2c_client *client,
>> +			   const struct i2c_device_id *id)
>> +{
>> +	struct device *dev = &client->dev;
>> +	struct nb6_data *data;
>> +	struct device *hwmon_dev;
>> +	int ret;
>> +
>> +	data = devm_kzalloc(dev, sizeof(struct nb6_data), GFP_KERNEL);
>> +	if (!data)
>> +		return -ENOMEM;
>> +
>> +	data->dev = dev;
>> +	data->i2c = client;
>> +	data->leds_mode = LEDS_MODE_NUM;
>> +	i2c_set_clientdata(client, data);
>> +	mutex_init(&data->lock);
>> +
>> +	hwmon_dev = devm_hwmon_device_register_with_groups(dev, client->name,
>> +							   data,
>> +							   nb6_groups);
> 
> New drivers must use devm_hwmon_device_register_with_info().
> 
>> +	if (IS_ERR(hwmon_dev))
>> +		return PTR_ERR(hwmon_dev);
>> +
>> +	data->pwm.dev = dev;
>> +	data->pwm.ops = &nb6_pwm_ops;
>> +	data->pwm.base = -1;
>> +	data->pwm.npwm = NB6_LEDS_PWM_COUNT;
>> +
>> +	ret = pwmchip_add(&data->pwm);
>> +	if (ret < 0)
>> +		return ret;
>> +
> 
> This very much looks like a multi-function chip (hardware monitoring,
> pwm, LEDs, and possibly watchdog) and should be implemented as such.
> 
>> +	return 0;
>> +};
>> +
>> +static const struct of_device_id nb6_hwmon_of_match[] = {
>> +	{ .compatible = "sfr,nb6_hwmon" },
>> +	{ },
>> +};
>> +MODULE_DEVICE_TABLE(of, nb6_hwmon_of_match);
>> +
>> +static struct i2c_driver nb6_hwmon_driver = {
>> +	.class = I2C_CLASS_HWMON,
>> +	.driver = {
>> +		.name = "nb6-hwmon",
>> +		.of_match_table = of_match_ptr(nb6_hwmon_of_match),
>> +	},
>> +	.probe = nb6_hwmon_probe,
>> +};
>> +module_i2c_driver(nb6_hwmon_driver);

Powered by blists - more mailing lists