[PATCH v4] drm/i915/hwmon: expose fan speed

[Raag Jadav]

On Fri, Aug 09, 2024 at 02:48:08PM +0300, Andy Shevchenko wrote:
> On Fri, Aug 09, 2024 at 11:45:25AM +0530, Raag Jadav wrote:
> > Add hwmon support for fan1_input attribute, which will expose fan speed
> > in RPM. With this in place we can monitor fan speed using lm-sensors tool.
> > 
> > $ sensors
> > i915-pci-0300
> > Adapter: PCI adapter
> > in0:         653.00 mV
> > fan1:        3833 RPM
> > power1:           N/A  (max =  43.00 W)
> > energy1:      32.02 kJ
> 
> > v2:
> > - Add mutex protection
> > - Handle overflow
> > - Add ABI documentation
> > - Aesthetic adjustments (Riana)
> > 
> > v3:
> > - Declare rotations as "long" and drop redundant casting
> > - Change date and version in ABI documentation
> > - Add commenter name in changelog (Riana)
> > 
> > v4:
> > - Fix wakeref leak
> > - Drop switch case and simplify hwm_fan_xx() (Andi)
> 
> I do not understand why we pollute Git history with changelogs, but it's
> probably the ugly atavism in DRM workflow.

Yeah I'm still getting used to it.
Also welcome back, hope it's not a bad start ;)

> ...
> 
> > +hwm_fan_is_visible(const struct hwm_drvdata *ddat, u32 attr)
> > +{
> > +	struct i915_hwmon *hwmon = ddat->hwmon;
> > +
> > +	return attr == hwmon_fan_input &&
> > +	       i915_mmio_reg_valid(hwmon->rg.fan_speed) ? 0444 : 0;
> 
> Not sure why ternary here, it's not well readable in my opinion.
> 
> 	if (attr == hwmon_fan_input && i915_mmio_reg_valid(hwmon->rg.fan_speed))
> 		return 0444;
> 
> 	return 0;
> 
> looks better, no?

Andi had a preference for single return statement.
I'm personally fine with both.

> ...
> 
> > +	/*
> > +	 * HW register value is accumulated count of pulses from
> > +	 * PWM fan with the scale of 2 pulses per rotation.
> > +	 */
> > +	rotations = pulses >> 1;
> 
> In accordance with the comment the
> 
> 	rotations = pulses / 2;
> 
> looks better.
> 
> ...
> 
> (1)
> 
> > +	time = time_now - fi->time_prev;
> > +
> 
> I think location of this blank line is better at (1) above.
> 
> > +	if (unlikely(!time)) {
> > +		ret = -EAGAIN;
> > +		goto exit;
> > +	}

Sure.

> ...
> 
> > +	/* Convert to minutes for calculating RPM */
> > +	*val = DIV_ROUND_UP(rotations * (60 * MSEC_PER_SEC), time);
> 
> Have you considered to keep jiffies in the fi and use something from jiffies.h
> here? To me it feels like we multiply and divide when it can be avoided.
> Please, think about it (I haven't checked myself, just an idea to share).

Will explore.

Raag