Re: [PATCH v2 4/4] arm64: dts: qcom: sdm845: mtp: Add vadc channels and thermal zones

[Bjorn Andersson]

On Tue 05 Oct 15:21 PDT 2021, Thara Gopinath wrote:

> 
> 
> On 10/5/21 5:52 PM, Bjorn Andersson wrote:
> > On Tue 05 Oct 13:30 PDT 2021, Thara Gopinath wrote:
> > 
> > > 
> > > 
> > > On 10/4/21 4:56 PM, Dmitry Baryshkov wrote:
> > > > On Mon, 4 Oct 2021 at 23:13, Bjorn Andersson <bjorn.andersson@xxxxxxxxxx> wrote:
> > > > > 
> > > > > On Wed 29 Sep 11:40 PDT 2021, Matthias Kaehlcke wrote:
> > > > > 
> > > > > > On Thu, Sep 23, 2021 at 02:23:11PM -0700, Bjorn Andersson wrote:
> > > > > > > Downstream defines four ADC channels related to thermal sensors external
> > > > > > > to the PM8998 and two channels for internal voltage measurements.
> > > > > > > 
> > > > > > > Add these to the upstream SDM845 MTP, describe the thermal monitor
> > > > > > > channels and add thermal_zones for these.
> > > > > > > 
> > > > > > > Signed-off-by: Bjorn Andersson <bjorn.andersson@xxxxxxxxxx>
> > > > > > > ---
> > > > > > > 
> > > > > > > In addition to the iio channels exposed by v1, Daniel wanted thermal_zones...
> > > > > > > 
> > > > > > > Changes since v1:
> > > > > > > - Enable the pm8998_adc_tm and describe the ADC channels
> > > > > > > - Add thermal-zones for the new channels
> > > > > > > 
> > > > > > >    arch/arm64/boot/dts/qcom/sdm845-mtp.dts | 128 ++++++++++++++++++++++++
> > > > > > >    1 file changed, 128 insertions(+)
> > > > > > > 
> > > > > > > diff --git a/arch/arm64/boot/dts/qcom/sdm845-mtp.dts b/arch/arm64/boot/dts/qcom/sdm845-mtp.dts
> > > > > [..]
> > > > > > > +&pm8998_adc {
> > > > > > > +   adc-chan@4c {
> > > > > > > +           reg = <ADC5_XO_THERM_100K_PU>;
> > > > > > > +           label = "xo_therm";
> > > > > > > +   };
> > > > > > > +
> > > > > > > +   adc-chan@4d {
> > > > > > > +           reg = <ADC5_AMUX_THM1_100K_PU>;
> > > > > > > +           label = "msm_therm";
> > > > > > > +   };
> > > > > > > +
> > > > > > > +   adc-chan@4f {
> > > > > > > +           reg = <ADC5_AMUX_THM3_100K_PU>;
> > > > > > > +           label = "pa_therm1";
> > > > > > > +   };
> > > > > > > +
> > > > > > > +   adc-chan@51 {
> > > > > > > +           reg = <ADC5_AMUX_THM5_100K_PU>;
> > > > > > > +           label = "quiet_therm";
> > > > > > > +   };
> > > > > > > +
> > > > > > > +   adc-chan@83 {
> > > > > > > +           reg = <ADC5_VPH_PWR>;
> > > > > > > +           label = "vph_pwr";
> > > > > > > +   };
> > > > > > > +
> > > > > > > +   adc-chan@85 {
> > > > > > > +           reg = <ADC5_VCOIN>;
> > > > > > > +           label = "vcoin";
> > > > > > > +   };
> > > > > > > +};
> > > > > > > +
> > > > > > > +&pm8998_adc_tm {
> > > > > > > +   status = "okay";
> > > > > > > +
> > > > > > > +   xo-thermistor@1 {
> > > > > > > +           reg = <1>;
> > > > > > > +           io-channels = <&pm8998_adc ADC5_XO_THERM_100K_PU>;
> > > > > > > +           qcom,ratiometric;
> > > > > > > +           qcom,hw-settle-time-us = <200>;
> > > > > > > +   };
> > > > > > > +
> > > > > > > +   msm-thermistor@2 {
> > > > > > > +           reg = <2>;
> > > > > > > +           io-channels = <&pm8998_adc ADC5_AMUX_THM1_100K_PU>;
> > > > > > > +           qcom,ratiometric;
> > > > > > > +           qcom,hw-settle-time-us = <200>;
> > > > > > > +   };
> > > > > > > +
> > > > > > > +   pa-thermistor@3 {
> > > > > > > +           reg = <3>;
> > > > > > > +           io-channels = <&pm8998_adc ADC5_AMUX_THM3_100K_PU>;
> > > > > > > +           qcom,ratiometric;
> > > > > > > +           qcom,hw-settle-time-us = <200>;
> > > > > > > +   };
> > > > > > > +
> > > > > > > +   quiet-thermistor@4 {
> > > > > > > +           reg = <4>;
> > > > > > > +           io-channels = <&pm8998_adc ADC5_AMUX_THM5_100K_PU>;
> > > > > > > +           qcom,ratiometric;
> > > > > > > +           qcom,hw-settle-time-us = <200>;
> > > > > > > +   };
> > > > > > > +};
> > > > > > > +
> > > > > > 
> > > > > > The example in the 'qcom,spmi-adc-tm5' binding specifies 'qcom,ratiometric'
> > > > > > and 'qcom,hw-settle-time-us' for both the ADC and the thermal monitor, so do
> > > > > > several board files (e.g. sm8250-mtp.dts and qrb5165-rb5.dts). This apparent
> > > > > > redundancy bothered me earlier, it's not really clear to me whether it's
> > > > > > needed/recommended or not. Do you happen to have any insights on this?
> > > > > 
> > > > > Hmm, you're right and I missed this in defining my channels. I've not
> > > > > looked at this detail, just got reasonable readings from my thermal
> > > > > zones and was happy about that.
> > > > > 
> > > > > Dmitry, do you have any further insights why these properties are
> > > > > supposed to be duplicated between the adc channel and the thermal zones?
> > > > 
> > > > Because both ADC channel and thermal zone registers should be
> > > > programmed accordingly.
> > > > 
> > > > One not-so-perfect approach would be to use io-channels property to
> > > > locate the adc's adc-chan node and to parse it. However this way
> > > > thermal driver would have to know the exact structure of adc's device
> > > > tree nodes.
> > > > Another (even worse) solution would be to introduce qcom-specific API
> > > > to query these properties from the IIO channel.
> > > > 
> > > > Selecting between these two options I decided to follow the downstream
> > > > path and just to duplicate these few properties.
> > > 
> > > I know I am joining late to this party, since v3 is out. But before I send
> > > my acked-by, wanted to check/state a few things.
> > > To me, the not-so-perfect approach that you mention here is actually not so
> > > bad, considering the thermal sensor driver already is aware of internals of
> > > the adc device. The duplicating of node is kind of confusing.
> > > Also did you explore using one of _channel_write apis in iio framework to
> > > move all the writes into the adc device instead of thermal sensor driver ?
> > > 
> > 
> > To me the TM hardware block seems to just duplicate some parts of the
> > ADC block. Without understanding why the two would need to be configured
> > in sync, I don't think we should tie them together further.
> 
> You are right. It is duplicating most of the ADC parts except for
> temperature read for which it calls into the adc device. And I don't see any
> interface from the ADC driver that can be used to do the configuration TM
> needs. Hence the question, have we ever explored this path ?
> 
> 
> > 
> > In particular, I get the same readings with or without configuring the
> > ADC channels, does this perhaps imply that the TM configuration is used
> > to configure the ADC during it's reads (i.e. we don't need to define the
> > ADC channels after all), or is this just a coincidence?
> 
> Maybe because of what we do in adc_tm5_init ?
> 
> If needed, we should be able to duplicate the read in the TM as well and get
> rid of ADC channels , right ? I don't like this half configuration in TM and
> half in ADC device unless there is a reason behind this.
> 

Looking at the registers once more makes me more convinced that the PMIC
is using the ADC hardware behind the back of the ADC driver - for the
purpose of detecting if we pass the trip points.

And then we use the ADC driver to read the temperature, separate of
that. To expose a thermal_zone with both trip points and temperature.

That way it would be in our best interest to configure the TM and the
ADC in the same way, so the trips and temperature readings are matching,
but it really is two different interfaces to the underlying ADC
hardware.

Regards,
Bjorn

> > 
> > There are other things in the PMIC, using the ADC seemingly without you
> > having to explicitly configure the ADC and there's definitely things
> > pointing to the ADC being "timeshared" between various users.
> > 
> > Regards,
> > Bjorn
> > 
> 
> -- 
> Warm Regards
> Thara (She/Her/Hers)