Re: [PATCH] arm: dts: fix rk3066a based boards vdd_log voltage initialization

[Heiko Stuebner]

Am Montag, 19. September 2016, 19:13:27 CEST schrieb Boris Brezillon:
> On Mon, 19 Sep 2016 09:38:34 -0700
> 
> Doug Anderson <dianders@xxxxxxxxxxxx> wrote:
> > Hi,
> > 
> > On Mon, Sep 19, 2016 at 9:15 AM, Heiko Stuebner <heiko@xxxxxxxxx> wrote:
> > > Am Montag, 19. September 2016, 08:15:30 CEST schrieb Doug Anderson:
> > >> Hi,
> > >> 
> > >> On Mon, Sep 19, 2016 at 1:44 AM, Andy Yan <andy.yan@xxxxxxxxxxxxxx> 
wrote:
> > >> > The current rk3066a based boards(Rayeager, Bqcurie2, Marsboard) use
> > >> > pwm modulate vdd_logic voltage, but the pwm is default disabled and
> > >> > the pwm pin acts as a gpio before pwm regulator probed, so the pwm
> > >> > regulator driver will get a zero dutycycle at probe time, so change
> > >> > the initial dutycycle to zero to match pwm_regulator_init_state
> > >> > check.
> > >> > 
> > >> > Signed-off-by: Andy Yan <andy.yan@xxxxxxxxxxxxxx>
> > >> > 
> > >> > ---
> > >> > 
> > >> >  arch/arm/boot/dts/rk3066a-bqcurie2.dts  | 2 +-
> > >> >  arch/arm/boot/dts/rk3066a-marsboard.dts | 2 +-
> > >> >  arch/arm/boot/dts/rk3066a-rayeager.dts  | 2 +-
> > >> >  3 files changed, 3 insertions(+), 3 deletions(-)
> > >> > 
> > >> > diff --git a/arch/arm/boot/dts/rk3066a-bqcurie2.dts
> > >> > b/arch/arm/boot/dts/rk3066a-bqcurie2.dts index bc674ee..618450d
> > >> > 100644
> > >> > --- a/arch/arm/boot/dts/rk3066a-bqcurie2.dts
> > >> > +++ b/arch/arm/boot/dts/rk3066a-bqcurie2.dts
> > >> > @@ -61,7 +61,7 @@
> > >> > 
> > >> >                 regulator-min-microvolt = <1200000>;
> > >> >                 regulator-max-microvolt = <1200000>;
> > >> >                 regulator-always-on;
> > >> > 
> > >> > -               voltage-table = <1000000 100>,
> > >> > +               voltage-table = <1000000 0>,
> > >> 
> > >> In my opinion this isn't quite the right answer.  I think that you
> > >> should add a new property describing the voltage in the case that the
> > >> 
> > >> pin is an input and you should fill that property in, like:
> > >>   voltage-when-input = <1000000>;
> > > 
> > > I'd think this would be more of a pwm issue, not something the
> > > pwm-regulator should need to care about.
> > > 
> > > Ideally the pwm driver should be able to return some state information
> > > even if disabled? I.e. deriving a duty-cycle value from its pin state
> > > similar to what Doug described below (it's either 0% or 100%)
> > > 
> > > But right now I have a hard time understanding how the pwm could return
> > > any
> > > duty-cycle information for an input gpio to the pwm-regulator, as I
> > > assume the pwm-driver has to probe (and thus set pinctrl to the pwm
> > > function) before the pwm-regulator is able to get the pwm handle?
> > 
> > Hrm, right.  The PWM ought to own the pinctrl, not the regulator.
> > Hrm.  Then I guess this gets more complicated.
> > 
> > One thing to point out, though, is that an EE I talked to said that
> > the "voltage when input" is actually a well defined property and is
> > unrelated to the min/max voltage.  AKA: it's not guaranteed to be
> > equal to the 50% duty cycle.  ...so adding a property to the PWM
> > regulator that includes this value is something very sane.  The
> > "voltage when input" is defined by the pile of resistors and
> > capacitors that are used to actually make the PWM control the
> > regulator.
> > 
> > The "voltage when input" is super important because this is the
> > voltage that's used at bootup (when all pins are configured as inputs,
> > possible with a pull applied) and that's used during suspend time when
> > the PWM stops.
> 
> Correct me if I'm wrong, but the main problem here is that, when we try
> to detect the initial regulator state, we ran into a "missing entry in
> the duty-cycle <-> voltage table" error, which then triggers an -EINVAL
> error preventing the PWM regulator probe to succeed.

I'm unsure about this ... i.e. according to Andy's description, the pin is 
muxed as gpio and set to input, which in turn results in the voltage-when-
input scenario.

So what happens with that voltage when the pwm-driver (= driver-core after 
probe) switches the pinmux to the pwm output, but that pwm itself is not 
running?

For veyrons it was easy, as the firmware set up the pwm to what it wanted, so 
the state could be read back.