Message-ID: <c6ab0452-f976-469c-be85-9b065db46361@roeck-us.net>
Date:   Tue, 22 Aug 2023 09:50:19 -0700
From:   Guenter Roeck <linux@...ck-us.net>
To:     Naresh Solanki <naresh.solanki@...ements.com>
Cc:     Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>,
        Conor Dooley <conor@...nel.org>,
        Jean Delvare <jdelvare@...e.com>,
        krzysztof.kozlowski+dt@...aro.org,
        Rob Herring <robh+dt@...nel.org>,
        Conor Dooley <conor+dt@...nel.org>,
        linux-hwmon@...r.kernel.org,
        Patrick Rudolph <patrick.rudolph@...ements.com>,
        Rob Herring <robh@...nel.org>, devicetree@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3 1/3] dt-bindings: hwmon: Add Infineon TDA38640

On Tue, Aug 22, 2023 at 09:41:48PM +0530, Naresh Solanki wrote:
> Hi,
> 
> On Tue, 22 Aug 2023 at 18:47, Guenter Roeck <linux@...ck-us.net> wrote:
> >
> > On Tue, Aug 22, 2023 at 02:32:31PM +0530, Naresh Solanki wrote:
> > > Hi
> > >
> > > On Fri, 18 Aug 2023 at 14:53, Krzysztof Kozlowski
> > > <krzysztof.kozlowski@...aro.org> wrote:
> > > >
> > > > On 16/08/2023 10:51, Naresh Solanki wrote:
> > > > > Hi Krzysztof,
> > > > >
> > > > > On Tue, 15 Aug 2023 at 01:02, Krzysztof Kozlowski
> > > > > <krzysztof.kozlowski@...aro.org> wrote:
> > > > >>
> > > > >> On 11/08/2023 18:00, Naresh Solanki wrote:
> > > > >>> Hi,
> > > > >>>
> > > > >>> On Tue, 8 Aug 2023 at 19:58, Conor Dooley <conor@...nel.org> wrote:
> > > > >>>>
> > > > >>>> On Tue, Aug 08, 2023 at 07:10:08AM -0700, Guenter Roeck wrote:
> > > > >>>>> On 8/8/23 04:46, Conor Dooley wrote:
> > > > >>>>>> On Wed, Aug 02, 2023 at 09:31:51PM +0200, Naresh Solanki wrote:
> > > > >>>>>>> From: Patrick Rudolph <patrick.rudolph@...ements.com>
> > > > >>>>>>>
> > > > >>>>>>> The TDA38640 chip has different output control mechanisms depending on
> > > > >>>>>>> its mode of operation. When the chip is in SVID mode, only
> > > > >>>>>>> hardware-based output control is supported via ENABLE pin. However, when
> > > > >>>>>>> it operates in PMBus mode, software control works perfectly.
> > > > >>>>>>>
> > > > >>>>>>> To enable software control as a workaround in SVID mode, add the DT
> > > > >>>>>>> property 'infineon,en-svid-control'. This property will enable the
> > > > >>>>>>> workaround, which utilizes ENABLE pin polarity flipping for output when
> > > > >>>>>>> the chip is in SVID mode.
> > > > >>>>>>
> > > > >>>>>> Why do you need a custom property for this? How come it is not possible
> > > > >>>>>> to determine what bus you are on?
> > > > >>>>>>
> > > > >>>>>
> > > > >>>>> That is not the point. Yes, it can be detected if the control method is
> > > > >>>>> PMBus or SVID. However, in SVID mode, SVID is supposed to control the
> > > > >>>>> output, not PMBUs. This is bypassed by controlling the polarity of the
> > > > >>>>> (physical) output enable signal. We do _not_ want this enabled automatically
> > > > >>>>> in SVID mode. Its side effects on random boards using this chip are unknown.
> > > > >>>>> Thus, this needs a property which specifically enables this functionality
> > > > >>>>> for users who _really_ need to use it and (hopefully) know what they are
> > > > >>>>> doing.
> > > > >>>>
> > > > >>>> Hmm, reading this it makes a lot more sense why this is a property - I
> > > > >>>> guess I just struggled to understand the commit message here,
> > > > >>>> particularly what the benefit of using the workaround is. I'm still
> > > > >>>> having difficulty parsing the commit & property text though - its
> > > > >>>> unclear to me when you would need to use it - so I will stay out
> > > > >>>> of the way & let Rob or Krzysztof handle things.
> > > > >>>
> > > > >>> To provide context, my system employs a unique power sequence
> > > > >>> strategy utilizing a BMC (Baseboard Management Controller),
> > > > >>> rendering the reliance on the ENABLE pin unnecessary.
> > > > >>> In this configuration, the ENABLE pin is grounded in the hardware.
> > > > >>> While most regulators facilitate PMBus Operation for output control,
> > > > >>> the TDA38640 chip, when in SVID mode, is constrained by the
> > > > >>> ENABLE pin to align with Intel specifications.
> > > > >>> My communication with Infineon confirmed that the recommended
> > > > >>> approach is to invert the Enable Pin for my use case.
> > > > >>>
> > > > >>> Since this is not typically the use case for most setup & hence DT property
> > > > >>> is must for enabling the special case.
> > > > >>>
> > > > >>> For further insight into my setup's power sequence strategy, you can
> > > > >>> refer to the following link: https://github.com/9elements/pwrseqd
> > > > >>>
> > > > >>
> > > > >> This justifies to me the property, but still you described desired
> > > > >> driver behavior, not the hardware characteristic. Don't describe what
> > > > >> you want to control, but describe the entire system.
> > > > > I guess by entire system you mean how the regulators(including
> > > > > TDA38640) connected & operated in our setup ?
> > > >
> > > > I mean, property name and description should say what is the
> > > > characteristic of the hardware/firmware/entire system.
> > > Based on your feedback, will update to below:
> > > infineon,fixed-level-en-pin:
> > >     description: |
> > >       Indicate the ENABLE pin is set at fixed level or left
> > >       unconnected(has internal pull-up).
> > >     type: boolean
> >
> > Messy, because while it reflects physical connectivity, it doesn't reflect
> > its use in the system at all. The pin may be at fixed level or left
> > unconnected, but the system vendor doesn't want to give users the
> > means to control output power (which would be the normal situation).
> Maybe this would be better ?
> infineon,svid-mode-fixed-en-pin
>   description: |
>     Indicate the ENABLE pin is set at fixed level or left
>     unconnected(has internal pull-up) which chip in
>     SVID mode.

which chip ? Do you mean "with the chip" ?

I don't think that makes a difference. It still doesn't describe
your use case (which is something like "the chip is in SVID mode,
its enable pin is set to fixed level, and we need to manipulate its
interpretation by the chip so we can control the enable status from
software). I have no idea how to express that in a way that would be
acceptable as devicetree property.

It doesn't seem to me that we are making much progress here.
I know it isn't supposed to be done, and I don't really like it,
but could you use a module parameter in your system ? I'd be
open to accept that to make progress.

Thanks,
Guenter

Powered by blists - more mailing lists