Re: [PATCH 2/3] pinctrl: Device tree bindings for Qualcomm pm8xxx gpio block

[Stephen Boyd]

On 07/10/14 02:53, Linus Walleij wrote:
> On Wed, Jul 9, 2014 at 11:18 PM, Bjorn Andersson <bjorn@xxxxxxx> wrote:
>> On Wed, Jul 9, 2014 at 1:53 AM, Linus Walleij <linus.walleij@xxxxxxxxxx> wrote:
>>> On Tue, Jul 8, 2014 at 3:26 AM, Bjorn Andersson
>>> <bjorn.andersson@xxxxxxxxxxxxxx> wrote:
>>>
>>> +- function:
>>> +       Usage: optional
>>> +       Value type: <string>
>>> +       Definition: Specify the alternative function to be configured for the
>>> +                   specified pins.  Valid values are:
>>> +                       "normal",
>>> +                       "paired",
>>> +                       "func1",
>>> +                       "func2",
>>> +                       "dtest1",
>>> +                       "dtest2",
>>> +                       "dtest3",
>>> +                       "dtest4"
>>> These are a bit ambigous, why doesn't the driver present functions that
>>> are more specific than "func1", "func2"? Or "dtest1"?
>> I agree, unfortunately I have only seen traces of the actual function matrix;
>> for pm8xxx I have no documentation and for pm8x41 they are only listed as
>> func[1-2] and dtest[1-4].
>>
>> Maybe if someone at Qualcomm could release such a list we could provide a
>> proper table instead.
> I guess Stephen Boyd can help us. (?)

Ok. "normal" is pretty much gpio mode, i.e. don't mux anything. "paired"
is where we take the output of the gpio next to it and loop it back into
this gpio (and vice versa). So gpio1 is paired with gpio2, gpio 3 is
paired with gpio 4, etc. This allows us to make level translators by
choosing different supply voltages for the paired gpios. "func1" and
"func2" are used for muxing things internally. "dtest" is used to mux
specific things out for testing purposes, not really used in any
end-products but still useful while debugging. I can provide the
function to pin mapping if necessary. There are lots of pmics.

>
>>>> +- bias-pull-up:
>>>> +       Usage: optional
>>>> +       Value type: <u32> (optional)
>>>> +       Definition: The specified pins should be configued as pull up. An
>>>> +                   optional argument can be used to configure the strength.
>>>> +                   Valid values are; as defined in
>>>> +                   <dt-bindings/pinctrl/qcom,pm8xxx-gpio.h>:
>>>> +                   1: 30uA                     (PM8XXX_GPIO_PULL_UP_30)
>>>> +                   2: 1.5uA                    (PM8XXX_GPIO_PULL_UP_1P5)
>>>> +                   3: 31.5uA                   (PM8XXX_GPIO_PULL_UP_31P5)
>>>> +                   4: 1.5uA + 30uA boost       (PM8XXX_GPIO_PULL_UP_1P5_30)
>>> Hm, I don't know of the internal kernel API or so, but I'm thinking that
>>> for the DT bindings, this definition should be generic in
>>> Documentation/devicetree/bindings/pinctrl/pinctrl-bindings.txt
>>> and put in SI units like uA.
>> Totally agree with you; and this is already specified in pinctrl-binding.txt as
>> being Ohm.
>>
>> So I first did a spin with the strength as a separate property, but as that
>> because the only part that pinconf-generic didn't parse for me I merged it and
>> wanted your comment on it.
> Yeah. And thinking of it.... how can it be uA? It has to be Ohms... it's a pull
> up resistor thing after all. I suspect the uA value is just something like the
> maximum current drawn through the pullup given a certain voltage?
>
>>> So I would prefer:
>>>
>>> bias-pull-up = <30>;
>>>
>> Yeah, but that's the easy one ;)
>>
>> How do you say 1.5 or 31.5 and how do you differ that from 1.5 + 30 boot?
> It needs to be set using Ohms.
>
>>> for 30 uA. Maybe we want nA even? I'm uncertain about the proper granularity
>>> here :-/
>>>
>>> Magic enumerators 1,2,3,4 doesn't seem so good, that seems more like it's
>>> trying to match the magic value that is to be poked into a register or
>>> something like that.
>> The stuff going into the hardware is a value 0-3 for pull up; so these values
>> are "only" an enum with the additional benefit of saying "bias-pull-up;"
>> results in 30uA pull up which is the most commonly used form (hence being
>> optional).
> What is the nominal voltage of these pins? GIven that you can figure
> out the Ohms. And I suspect it to be something very close to N times
> the resistance of a depleted transistor in this technology.

I believe the nominal voltage changes depending on which supply you
choose (power-source in this document). Basically the gpio can be
connected to different regulators on the pmic so you can choose
different voltages, i.e. 1.8V, 3.0V, 3.3V etc. Furthermore, some of the
regulators you can choose have variable voltage, although it may not be
variable enough to have much effect on this. So it would seem that the
pull-up resistance would be directly affected by which power-source is
chosen. Maybe we just shouldn't use the generic properties for this?

BTW, I see that power-source has made a comeback. What are the units? Is
that in mV? If it is I'm slightly concerned that we're not accurately
describing the hardware in cases where the voltage can actually be
different. And I worry about configurations where we may have the same
power source muxed into the gpio twice from different places on the
pmic. Sometimes we do this and actually need to choose the "right" power
source even though they're technically running at the same voltage (one
may be slightly cleaner signal or something).

>>>> +- drive-strength:
>>>> +       Usage: optional
>>>> +       Value type: <u32>
>>>> +       Definition: Selects the drive strength for the specified pins. Value
>>>> +                   drive strengths are:
>>>> +                       0: no   (PM8XXX_GPIO_STRENGTH_NO)
>>>> +                       1: high (PM8XXX_GPIO_STRENGTH_HIGH)
>>>> +                       2: medium       (PM8XXX_GPIO_STRENGTH_MED)
>>>> +                       3: low  (PM8XXX_GPIO_STRENGTH_LOW)
>>> I would really prefer to have these in mA, because the genric pinconf
>>> bindings say they should be! SI units are so much more understandable.
>> This is all the information to be found in the available documentation and
>> code. Maybe someone from Qualcomm can shed some light on it?
> Stephen?

I've emailed the hardware engineers. I'm pretty sure this is the same
story as the pull-up though. It varies depending on the input voltage. I
hope to get more information soon.

-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by The Linux Foundation

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/