Re: [PATCH v4 1/6] dt-bindings: phy: qcom: add Nord QMP PCIe PHY binding
From: Krzysztof Kozlowski
Date: Fri Sep 25 2026 - 02:29:13 EST
On 25/09/2026 06:08, Krishna Chaitanya Chundru wrote:
>
>
> On 9/24/2026 6:52 PM, Krzysztof Kozlowski wrote:
>> On Mon, Sep 21, 2026 at 11:07:09AM +0530, Krishna Chaitanya Chundru wrote:
>>> +
>>> + vdda-pll-d-supply: true
>>> +
>>> + qcom,link-mode:
>>> + description:
>>> + Reference to the TCSR syscon register containing the PCIe link mode
>>> + selected by firmware. The property consists of the TCSR syscon
>>> + phandle followed by the offset of the TCSR_PCIE_LINK_CONFIG_MODE
>>> + register.
>>> + The register value selects one of these topologies.
>>> + 0 (QMP_PCIE_NORD_MODE_X16) — all ports combined as one x16 PHY
>>> + 1 (QMP_PCIE_NORD_MODE_X8_X8) — port A as x8 and ports B+C+D as x8
>>> + 2 (QMP_PCIE_NORD_MODE_X8_X4_X4) — port A as x8, port B as x4, and ports C+D as x4
>>> + 3 (QMP_PCIE_NORD_MODE_X8_X4_X2_X2) — port A as x8, port B as x4, and ports C and D as x2
>> Same question as last time. phy-cells tells which mode you are here.
>> This is duplicating it and the register value CANNOT choose the
>> topology. If it does, then phy-cells should be 0 as topology is defined
>> by hardware already, no?
> phy-cells does not describe the topology. The PHY argument is only used by the
> PCIe controller to identify the PHY instance associated with that controller
> (for example, phys = <&pcie_phy 2> selects PHY instance 2).
>
> The qcom,link-mode property serves a different purpose. It provides access to
> the firmware-programmed TCSR_PCIE_LINK_CONFIG_MODE register, which describes
> the global lane partitioning (x16, x8+x8, x8+x4+x4, x8+x4+x2+x2, etc.).
>
> - phy-cells identifies which PHY instance a controller is using.
> - qcom,link-mode reports the overall topology selected by firmware.
OK, I assumed the cell argument defines the instance in that topology,
but it seems not. I think that was the misunderstanding.
Best regards,
Krzysztof