Re: [PATCH v3 2/3] hwmon: add ChromeOS EC driver

[Thomas Weißschuh]

On 2024-05-28 06:39:25+0000, Tzung-Bi Shih wrote:
> On Mon, May 27, 2024 at 10:58:32PM +0200, Thomas Weißschuh wrote:
> > diff --git a/drivers/hwmon/cros_ec_hwmon.c b/drivers/hwmon/cros_ec_hwmon.c
> [...]
> > + *  ChromesOS EC driver for hwmon
> 
> s/ChromesOS/ChromeOS/.

Ack. Copy-and-paste...

> > +struct cros_ec_hwmon_priv {
> > +	struct cros_ec_device *cros_ec;
> > +	const char *temp_sensor_names[EC_TEMP_SENSOR_ENTRIES + EC_TEMP_SENSOR_B_ENTRIES];
> > +	u8 usable_fans;
> > +};
> > +
> > +static int cros_ec_hwmon_read_fan_speed(struct cros_ec_device *cros_ec, u8 index, u16 *speed)
> > +{
> > +	u16 data;
> > +	int ret;
> > +
> > +	ret = cros_ec_cmd_readmem(cros_ec, EC_MEMMAP_FAN + index * 2, 2, &data);
> > +	if (ret < 0)
> > +		return ret;
> > +
> > +	data = le16_to_cpu(data);
> > +	*speed = data;
> > +
> > +	if (data == EC_FAN_SPEED_NOT_PRESENT || data == EC_FAN_SPEED_STALLED)
> > +		return -EIO;
> 
> `data` could be eliminated; use `*speed` instead.

Then the le16 value would need to be written directly to the out
parameter. The current usage relies on *speed (sometimes) being set even
if ret != 0.

(See next response block)

> 
> > +static int cros_ec_hwmon_read(struct device *dev, enum hwmon_sensor_types type,
> > +			      u32 attr, int channel, long *val)
> > +{
> [...]
> > +	u16 speed = 0;
> > +	u8 temp = 0;
> 
> They don't need to initialize.

They need to.

The logic

if (ret == -EIO && speed == EC_FAN_SPEED_STALLED)
	ret = 0;

relies on -EIO and a write to speed from cros_ec_hwmon_read_fan_speed().

But if cros_ec_cmd_readmem() already returns -EIO, then speed would be
uninitialized.

I'll see if I can make this clearer somehow.

> 
> > +	if (type == hwmon_fan && attr == hwmon_fan_input) {
> > +		ret = cros_ec_hwmon_read_fan_speed(priv->cros_ec, channel, &speed);
> > +		if (ret == 0)
> > +			*val = speed;
> > +	} else if (type == hwmon_fan && attr == hwmon_fan_fault) {
> > +		ret = cros_ec_hwmon_read_fan_speed(priv->cros_ec, channel, &speed);
> > +		if (ret == -EIO && speed == EC_FAN_SPEED_STALLED)
> > +			ret = 0;
> > +		if (ret == 0)
> > +			*val = speed == EC_FAN_SPEED_STALLED;
> > +	} else if (type == hwmon_temp && attr == hwmon_temp_input) {
> > +		ret = cros_ec_hwmon_read_temp(priv->cros_ec, channel, &temp);
> > +		if (ret == 0)
> > +			*val = kelvin_to_millicelsius((((long)temp) + EC_TEMP_SENSOR_OFFSET));
> > +	} else if (type == hwmon_temp && attr == hwmon_temp_fault) {
> > +		ret = cros_ec_hwmon_read_temp(priv->cros_ec, channel, &temp);
> > +		if (ret == -EIO && speed == EC_TEMP_SENSOR_ERROR)
> > +			ret = 0;
> > +		if (ret == 0)
> > +			*val = temp == EC_TEMP_SENSOR_ERROR;
> > +	}
> 
> Refactoring them by switch .. case .. may improve the readability.

It would introduce another level of indentation, which I tried to avoid.
But some vertical whitespace would be useful indeed.