[lm-sensors] [PATCH 2.6.25.4] f71882.c driver, Added PWM/FAN control.

[Hans de Goede]

Mark van Doesburg wrote:
> Hi,
> 
> I've finally added some missing features to the driver.

Good work, thanks!

> There are still
> some things unimplemented, such as:
> 
> 	1. Changing the fan target reached timeout.
> 	2. Something to set FAN?_RATE_SEL.  
> 	3. FAN?_STOP_DUTY
> 	4. FAN?_JUMP_HIGH_EN and FAN?_JUMP_LOW_EN 
> 
> Mostly because I don't need them, and there is no default name for
> these features.
> 

Yes, I saw all those when reviewing the datasheet too, and I agree its best to 
just not export them.

> I did add one entry to set/reset the interpolation mode of the auto_point
> follower, to make the pwm control act as a P-regulator. Simply because
> I like it, it prevents rapid speedups and slowdowns of the fan.
> 

Ok.

> Since the auto_point follower is able to track only one temperature at
> a time, I gave a preference to the lowest numbered one. Maybe returning
> an error would be better if the user tries to set an impossible value ?
> 

Yes, please just return -EINVAL on an impossible value.

I've taken a quick peek at it, and all in all it looks good, except for the 
stuffing of pwm and point together in one integer that isn't necessary, I 
believe it would be the best to instead use struct sensor_device_attribute_2 
instead of struct sensor_device_attribute for the f71882fg_fan_attr array, as 
that allows specifying both an index and a nr per attr, so instead of:

static struct sensor_device_attribute f71882fg_fan_attr[] =
{
   	SENSOR_ATTR(fan1_input, S_IRUGO, show_fan, NULL, 0),
...
	SENSOR_ATTR(pwm1_auto_point1_pwm, S_IRUGO|S_IWUSR,
		show_pwm_auto_point_pwm, store_pwm_auto_point_pwm,
		AUTO_POINT(0,0)),


You would write:

static struct sensor_device_attribute f71882fg_fan_attr[] =
{
	SENSOR_ATTR_2(fan1_input, S_IRUGO, show_fan, NULL, 0, 0),
...
	SENSOR_ATTR_2(pwm1_auto_point1_pwm, S_IRUGO|S_IWUSR,
		show_pwm_auto_point_pwm, store_pwm_auto_point_pwm,
		0, 0),


And then instead of:

static ssize_t store_pwm_auto_point_temp(struct device *dev,
	struct device_attribute *devattr,
	const char *buf, size_t count)
{
	int point, pwm;
	struct f71882fg_data *data = f71882fg_update_device(dev);
	int nr = to_sensor_dev_attr(devattr)->index;
	int val = simple_strtoul(buf, NULL, 10);

	pwm=AUTO_POINT_PWM(nr);
	point=AUTO_POINT_POINT(nr);
...

You would write:

static ssize_t store_pwm_auto_point_temp(struct device *dev,
	struct device_attribute *devattr,
	const char *buf, size_t count)
{
	struct f71882fg_data *data = f71882fg_update_device(dev);
	int pwm = to_sensor_dev_attr_2(devattr)->index;
	int point = to_sensor_dev_attr_2(devattr)->nr;
	int val = simple_strtoul(buf, NULL, 10);
...


If you could make a new patch with this changed (and returning -EINVAL for 
trying to bind a auto pwm to more then one temp), I'll do a full review and run 
a few tests on my f71882fg test system. I think it would be best if you could 
split this new patch in 2 parts:
1) Convert the current f71882.c from "struct sensor_device_attribute" ->
    "struct sensor_device_attribute_2" without any functional changes
2) Your patch adding pwm support (thanks!!) on top of this

I would like to see it this way for easier review.

Thanks & Regards,

Hans