Re: [PATCH v6 1/3] dt-bindings: phy: Add CIX Sky1 USB3 PHY

From: Hongliang Yang

Date: Tue Nov 10 2026 - 22:11:11 EST


> So previously you had two phys, now you have only one. Confusing. Where
> did the second phy go?

The second PHY did not go anywhere. It is the second lane of the same
block, and it has no resources of its own.

v5 described one child node per lane:

usb-phy@9210000 {
...
usb-port@0 {
reg = <0>;
#phy-cells = <0>;
};

usb-port@1 {
reg = <1>;
#phy-cells = <0>;
};
};

Those child nodes had no register range, clocks or resets of their own;
everything they described belongs to the parent. Both lanes are
programmed through the same register block, with the same clocks and
resets, and by the same initialization sequence, so a consumer has
nothing to configure per lane. The two children only mirrored the two
controllers.

v6 makes the parent the provider, with #phy-cells = <0>, and both
controllers reference the same phandle:

usb3_phy4: usb-phy@9210000 {
compatible = "cix,sky1-usb3-phy";
reg = <0x9210000 0x40000>;
clocks = <&scmi_clk CLK_TREE_USB3A_PHY3_GATE>,
<&scmi_clk CLK_TREE_USB3A_PHY_x2_REF>;
clock-names = "apb", "ref";
resets = <&s5_syscon SKY1_USBPHY_SS_RST_N>,
<&s5_syscon SKY1_USBPHY_SS_PST_N>;
reset-names = "core", "apb";
#phy-cells = <0>;
};

&usb4 {
phys = <&usb3_phy4>;
phy-names = "cdns3,usb3-phy";
};

&usb5 {
phys = <&usb3_phy4>;
phy-names = "cdns3,usb3-phy";
};

The driver reference counts phy_init() and phy_exit(), so the block is
initialized once, when the first consumer comes up, and stays powered
while any of the two is active.

The v6 changelog only said that the per-port child nodes were dropped and
did not say why, which is what made the change look like a PHY had gone
missing. The binding description and the commit message of v7 state the
reason, and I will resend the series with that.

Also, the example in the USBSSP binding still referenced two phandles in
the v5 form, which did not help; that is fixed in the USB series.

Best regards,
Hongliang Yang