Re: [PATCH 9/9] arm64: dts: renesas: rzg3s-smarc-som: Enable I3C

[Claudiu Beznea <claudiu.beznea@kernel.org>]

Hi, Wolfram,

On 5/26/26 19:55, Wolfram Sang wrote:
> Hi Claudiu,
> 
>> Could you please let me know what do you consider we should do here? Do you
>> think we could drop these pin controller setting and do some particular I3C
>> controller settings instead?
> 
> My original thought was: If it was a boolean state which is active when
> suspending and disabled when resuming, then we wouldn't need a customer
> specific binding for it and just do this in the suspend/resume functions
> of the pin-controller...
> 
> ... BUT ...
> 
> reading more about this in the manual, just raises more questions for
> me.
> 
>> The output is fixed at Hi-Z and no data is transmitted to the inside even if
>> data is input from outside. “Standby mode” is available when using I2C mode
>> only. (Not available when using I3C mode).
>>
>> The current driver proposal don't take into account the IP mode when setting
>> STBN though
>> pinctrl_pm_select_sleep_state()/pinctrl_pm_select_default_state() to keep
>> the code simpler, relying on the "Not available when using I3C mode" part of
>> the note, and considering setting it when the IP is in I3C mode is harmless.
> 
> This is one question I also had: What does "not available" actually
> mean? Did you confirm with HW guys that it is really harmles?

I asked this yesterday, but I don't have an answer to it.

In my testing I haven't noticed any issues with it.

> 
> I also wonder about the intended use-case of this mode. "no data is
> transmitted to the inside even if data is input from outside" doesn't
> really sound like a mode intended when the whole SoC goes to sleep.

Indeed, that's why I chose in the driver to touch it though the runtime PM APIs, 
so it can be configured after the I3C IP is no longer used.

> Why
> or how would input be even transmitted to the inside if everything is in
> a deep-sleep state? I could also imagine that this mode is rather used
> to hide from the bus for a while for some corner-case reason.
> 
> And finally: does this really save energy? 

I don't expect. I'm not sure I can measure it. I also asked to the HW team.

> Could you measure a benefit?
> Maybe there is nothing driven at all in the sleep state? Then, nothing
> is gained? Not clear from the datasheet.
> 
> Because the datasheet is so sparse with information and because it
> doesn't say how STBN is intended to be used, I would argue we should
> skip it until we know what it is for and how it is used. If we know this
> somewhen, we can still add this in a second step.

OK for me.

> 
> But for now, enabling I3C realiably is the first step, and for that we
> surley need the POC bit to select the voltage. This is easily
> understandable and straightforward to do. So, my suggestion is to pick
> this low-hanging fruit now and reach for the other one once we have more
> information about it.

OK, I'll re-spin this dropping the STBN feature until further clarification.

-- 
Thank you,
Claudiu