Re: (EXT) Re: (EXT) Re: [PATCH v2 1/1] hwmon: pwm-fan: dynamically switch regulator

[Alexander Stein <alexander.stein@ew.tq-group.com>]

Am Freitag, 6. Mai 2022, 10:20:01 CEST schrieb Uwe Kleine-König:
> * PGP Signed by an unknown key
> 
> Hello,
> 
> On Fri, May 06, 2022 at 09:15:55AM +0200, Alexander Stein wrote:
> > Am Freitag, 6. Mai 2022, 00:00:37 CEST schrieb Guenter Roeck:
> > > On Wed, May 04, 2022 at 02:45:51PM +0200, Alexander Stein wrote:
> > > > From: Markus Niebel <Markus.Niebel@ew.tq-group.com>
> > > > 
> > > > A pwm value equal to zero is meant to switch off the pwm
> > > > hence also switching off the fan. Currently the optional
> > > > regulator is always on. When using this driver on boards
> > > > with an inverted pwm signal polarity this can cause running
> > > > the fan at maximum speed when setting pwm to zero.
> > > 
> > > The appropriate solution in this case would be to tell the
> > > software that the pwm is inverted. Turning off the regulator
> > > in that situation is a bad idea since setting the pwm value to
> > > 1 would set it to almost full speed. That does not really make
> > > sense.
> > 
> > The pwm-fan driver is already configured for inverted PWM (ommited some
> > properties for shortness):
> > fan0: pwm-fan {
> > 
> > 	compatible = "pwm-fan";
> > 	fan-supply = <&reg_pwm_fan>;
> > 	pwms = <&pwm3 0 40000 PWM_POLARITY_INVERTED>;
> > 	cooling-levels = <0 32 64 128 196 240>;
> > 
> > [...]
> > };
> > 
> > The problem here is that the pwm-fan driver currently enables the
> > regulator
> > unconditionally, but the PWM only when the fan is enabled, refer to
> > __set_pwm(). This results in a fan at full speed when pwm-fan is idle, as
> > pwm state is not enabled.
> 
> Which PWM driver are you using?

It's pwm-imx27 on a imx8mp based board.

> There is an implicit assumption in some PWM consumers that a disabled
> PWM emits the inactive level. However not all PWMs do this. Is this such
> a case?

Oh, I was not aware of that assumption. As far I can tell, this assumption 
might actually be violated in pwm-imx27.
If state->enabled==false then the EN Bit in PWMCR is not set which most 
probably renders the output polarity in POUTC as inactive.

Best regards,
Alexander