Re: [PATCH v4 1/2] dt-bindings: qcom,snps-dwc3: Add property indicating presence of eUSB2 phy

From: Thinh Nguyen

Date: Wed Jul 15 2026 - 20:10:22 EST


On Tue, Jul 14, 2026, Peter Chen wrote:
> On 26-07-14 23:23:01, Thinh Nguyen wrote:
> > On Mon, Jul 13, 2026, Peter Chen wrote:
> > >
> > > Yes, eUSB2 is based on UTMI Parallel mode, but it has dedicated
> > > Physical Layer Supplement [1], eg at CH2.4, it lists eUSB2 PHY
> > > Features.
> > >
> > >
> > > > To go this route properly, we'd need to introduce a new phy type
> > > > attribute in the phy framework, which will be a bigger change that may
> > > > impact more than this driver.
> > > >
> > >
> > > It is not at generic PHY framework, it is just for USB PHY dedicated.
> > >
> > > I do not see big changes, it only needs to change above files and related
> > > dt-binding files, and other users may leverage it if the controller
> > > has special sequence or settings for eUSB2.
> > >
> >
> > The usbphy_modes describes the interface between the controller and the
> > phy. eusb2 still uses utmi, so adding eusb2 there is sematically
> > incorrect. If we introduce the eusb2 mode there, we'd have to audit dwc3
> > and every other driver that uses usbphy_modes to handle the new mode
> > correctly.
> >
> > That said, I have no objection to adding a new phy type for eusb2. I
> > just want to note it is a bigger change relative to adding a boolean
> > property to the DT binding.
> >
>
> Hi Thinh,
>
> Only the new user for adding phy_type="eusb2" at dts needs to check the
> controller logic, and this is what the user needs to do that check the
> new added device tree property logic at the driver.

To clarify, are you suggesting to introduce a new phy_type property to
indicate phy type? If so, that sounds similar to qcom,has-eusb2-phy but
more generic. Also, where do you plan to define it (dwc3 or phy node)?

>
> So, it will not affect current user for phy_type, and I checked the dwc3
> core, it does not need to change code for new phy type.
>
> Currently, it is only benefit for qcom platform, but eusb2 (uesb2v2 later)
> is the common USB2 mode, other users may leverage for this new phy type.
>

If you mean defining a new PHY_TYPE_EUSB2 constant in
include/dt-bindings/phy/phy.h, the glue driver needs to parse the
phandle args, and the eusb2 bindings would also need to change the
number of #phy-cells. That's a bigger change.

Thanks,
Thinh