Re: [PATCH v1 5/6] dt-bindings: phy: tegra: Document Nvidia Tegra XMM6260 PHY

[Svyatoslav Ryhel]

пн, 18 трав. 2026 р. о 15:14 Krzysztof Kozlowski <krzk@xxxxxxxxxx> пише:
>
> On Fri, May 15, 2026 at 11:37:34AM +0300, Svyatoslav Ryhel wrote:
> > пт, 15 трав. 2026 р. о 11:20 Krzysztof Kozlowski <krzk@xxxxxxxxxx> пише:
> > >
> > > On Mon, May 11, 2026 at 04:57:00PM +0300, Svyatoslav Ryhel wrote:
> > > > Document the XMM6260 PHY used by various devices based on the Nvidia Tegra
> > > > SoC, describing its usage
> > > >
> > > > Signed-off-by: Svyatoslav Ryhel <clamor95@xxxxxxxxx>
> > > > ---
> > > >  .../bindings/phy/nvidia,tegra-xmm6260.yaml    | 58 +++++++++++++++++++
> > > >  1 file changed, 58 insertions(+)
> > > >  create mode 100644 Documentation/devicetree/bindings/phy/nvidia,tegra-xmm6260.yaml
> > > >
> > > > diff --git a/Documentation/devicetree/bindings/phy/nvidia,tegra-xmm6260.yaml b/Documentation/devicetree/bindings/phy/nvidia,tegra-xmm6260.yaml
> > > > new file mode 100644
> > > > index 000000000000..0346433c9772
> > > > --- /dev/null
> > > > +++ b/Documentation/devicetree/bindings/phy/nvidia,tegra-xmm6260.yaml
> > > > @@ -0,0 +1,58 @@
> > > > +# SPDX-License-Identifier: (GPL-2.0 OR BSD-2-Clause)
> > > > +%YAML 1.2
> > > > +---
> > > > +$id: http://devicetree.org/schemas/phy/nvidia,tegra-xmm6260.yaml#
> > > > +$schema: http://devicetree.org/meta-schemas/core.yaml#
> > > > +
> > > > +title: Nvidia Tegra PHY for XMM6260 modem
> > >
> > > XMM6260 is Infineon modem, so any combination with nvidia,tegra is very
> > > confusing.
> > >
> >
> > May you please suggest how to adjust the name then? Thank you.
>
> Depending what is that. Start describing hardware, not driver behavior
> to help in that.
>
> >
> > > > +
> > > > +description:
> > > > +  A hardware configuration used in Tegra SoCs to provide proper interaction
> > > > +  between the application processor and the modem, as well as control over
> > > > +  one of the SoC's USB lines for the modem.
> > > > +
> > > > +maintainers:
> > > > +  - Svyatoslav Ryhel <clamor95@xxxxxxxxx>
> > > > +
> > > > +properties:
> > > > +  compatible:
> > > > +    const: nvidia,tegra-xmm6260
> > >
> > > Also here.
> > >
> > > What sort of phy is this? So far looks more like a software construct.
> > >
> >
> > Infineon XMM6260 does not work as an ordinary USB modem, it is a
> > standalone CPU which just exposes itself to AP via USB. In order to do
> > so, it has to have control over a USB bus of AP which is dedicated to
> > it. In case of Tegra - XMM6260 interaction it looks like this: second
> > Tegra USB controller is set into HSIC mode and is dedicated solely to
> > the modem, modem controls this USB bus. Then the main XMM6260 driver
> > performs power and init sequence and once it is ready it calls phy to
> > register controller. Phy has its own supply, controls USB controller
> > de/register and using enable GPIO sends signal to modem to proceed.
> > Additionally, since some XMM626 versions have a few steps to setup
> > exposing different USB devices, phy handles controller reinit for each
> > step. If treat XMM6260 as an simple USB modem it will never init.
> >
> > One more benefit of having PHY is that modem driver itself is generic
> > and PHY handles SoC specific configurations required by the modem.
> > Since this modem was used on a variety of different SoC's (Exynos and
> > OMAP for example) they can reuse modem's driver and provide only PHY
> > which handles modem interactions with the USB bus.
>
> Without any registers here, this is not a PHY but a power sequencing,
> just like we do for other USB or PCI devices.
>
> Optionally, it could be part of existing USB phy, when configuring it in
> HSIC mode, but it seems you add here supplies for the modem, not actual
> phy as the phy is undefined.
>
> The problem is that in the patch and explanation you mix driver model
> and driver behavior, so I really don't know what is this hardware. And
> it is not my job to guess, btw. A partial argument/proof why this is not
> a PHY, is that you reference the USB in the node, so phy-provider
> references the phy-consumer. That's reverse. If this is PHY, it's USB's
> HSIC phy, thus this needs to be referenced by USB.
>
> If this is power sequencing, then it can be represented as USB device,
> just like we do for all USB devices, but then it is not PHY and
> phy-cells are not appropriate.

I was looking into power sequencing and using it instead of phy is
perfectly fine for me. How I see XMM modem configuration:
- main device which handles most of generic power sequences,
init/deinit and detection for USB device (both, target USB device and
intermediate stages for fw loading)
- phy or power seq, does not matter, both fit, that performs
manipulations with SoC specific USB controller in order to ensure
correct modem boot

If I use power seq then modem will require link to power seq and power
seq will work with USB line, this should not create confusion of phy
acquiring USB that is phy consumer. Will this approach be acceptable?

>
>
> Best regards,
> Krzysztof
>