Re: [PATCH 1/2] dt-bindings: net: bluetooth: nxp: add support for supply and reset

[Krzysztof Kozlowski <krzk@kernel.org>]

On 21/10/2024 12:25, Marco Felsch wrote:
> On 24-10-21, Krzysztof Kozlowski wrote:
>> On 21/10/2024 08:41, Marco Felsch wrote:
>>> On 24-10-07, Krzysztof Kozlowski wrote:
>>>> On 07/10/2024 14:58, POPESCU Catalin wrote:
>>>>>>>>
>>>>>>>> +  vcc-supply:
>>>>>>>> +    description:
>>>>>>>> +      phandle of the regulator that provides the supply voltage.
>>>>>>>> +
>>>>>>>> +  reset-gpios:
>>>>>>>> +    description:
>>>>>>>> +      Chip powerdown/reset signal (PDn).
>>>>>>>> +
>>>>>>> Hi Catalin,
>>>>>>>
>>>>>>> For NXP WIFI/BT chip, WIFI and BT share the one PDn pin, which means that both wifi and BT controller will be powered on and off at the same time.
>>>>>>> Taking the M.2 NXP WIFI/BT module as an example, pin56(W_DISABLE1) is connected to the WIFI/BT chip PDn pin, we has already controlled this pin in the corresponding PCIe/SDIO controller dts nodes.
>>>>>>> It is not clear to me what exactly pins for vcc-supply and reset-gpios you describing here. Can you help understand the corresponding pins on M.2 interface as an example? Thanks.
>>>>>
>>>>> Hi Sherry,
>>>>>
>>>>> Regulators and reset controls being refcounted, we can then implement 
>>>>> powerup sequence in both bluetooth/wlan drivers and have the drivers 
>>>>> operate independently. This way bluetooth driver would has no dependance 
>>>>> on the wlan driver for :
>>>>>
>>>>> - its power supply
>>>>>
>>>>> - its reset pin (PDn)
>>>>>
>>>>> - its firmware (being downloaded as part of the combo firmware)
>>>>>
>>>>> For the wlan driver we use mmc power sequence to drive the chip reset 
>>>>> pin and there's another patchset that adds support for reset control 
>>>>> into the mmc pwrseq simple driver.
>>>>>
>>>>>> Please wrap your replies.
>>>>>>
>>>>>> It seems you need power sequencing just like Bartosz did for Qualcomm WCN.
>>>>>
>>>>> Hi Krzysztof,
>>>>>
>>>>> I'm not familiar with power sequencing, but looks like way more 
>>>>> complicated than reset controls. So, why power sequencing is recommended 
>>>>> here ? Is it b/c a supply is involved ?
>>>>
>>>> Based on earlier message:
>>>>
>>>> "For NXP WIFI/BT chip, WIFI and BT share the one PDn pin, which means
>>>> that both wifi and BT controller will be powered on and off at the same
>>>> time."
>>>>
>>>> but maybe that's not needed. No clue, I don't know the hardware. But be
>>>> carefully what you write in the bindings, because then it will be ABI.
>>>
>>> We noticed the new power-sequencing infrastructure which is part of 6.11
>>> too but I don't think that this patch is wrong. The DT ABI won't break
>>> if we switch to the power-sequencing later on since the "reset-gpios"
>>> are not marked as required. So it is up to the driver to handle it
>>> either via a separate power-sequence driver or via "power-supply" and
>>> "reset-gpios" directly.
>>
>> That's not the point. We expect correct hardware description. If you say
>> now it has "reset-gpios" but later say "actually no, because it has
>> PMU", I respond: no. Describe the hardware, not current Linux.
> 
> I know that DT abstracts the HW. That said I don't see the problem with
> this patch. The HW is abstracted just fine:
> 
> shared PDn          -> reset-gpios
> shared power-supply -> vcc-supply
> 
> Right now the DT ABI for the BT part is incomplete since it assume a
> running WLAN part or some hog-gpios to pull the device out-of-reset
> which is addressed by this patchset.
> 
> Making use of the new power-sequencing fw is a Linux detail and I don't
> see why the DT can't be extended later on. We always extend the DT if
> something is missing or if we found a better way to handle devices.

Sure, although I am not really confident that you understand the
implications - you will not be able to switch to proper power-sequencing
with above bindings, because it will not be just possible without
breaking the ABI or changing hardware description (which you say it is
"fine", so complete/done). I am fine with it, just mind the implications.

Best regards,
Krzysztof