On 2023/3/22 16:00, Roger Quadros wrote:
> Hi Minda,
>
> On 21/03/2023 14:35, Minda Chen wrote:
>>
>>
>> On 2023/3/20 23:34, Rob Herring wrote:
>>> On Wed, Mar 15, 2023 at 06:44:11PM +0800, Minda Chen wrote:
>>>> USB Glue layer and Cadence USB subnode configuration,
>>>> also includes USB and PCIe phy dts configuration.
>>>>
>>>> Signed-off-by: Minda Chen <minda.chen@xxxxxxxxxxxxxxxx>
>>>> ---
>>>> .../jh7110-starfive-visionfive-2.dtsi | 7 +++
>>>> arch/riscv/boot/dts/starfive/jh7110.dtsi | 54 +++++++++++++++++++
>>>> 2 files changed, 61 insertions(+)
>>>>
>>>> diff --git a/arch/riscv/boot/dts/starfive/jh7110-starfive-visionfive-2.dtsi b/arch/riscv/boot/dts/starfive/jh7110-starfive-visionfive-2.dtsi
>>>> index a132debb9b53..c64476aebc1a 100644
>>>> --- a/arch/riscv/boot/dts/starfive/jh7110-starfive-visionfive-2.dtsi
>>>> +++ b/arch/riscv/boot/dts/starfive/jh7110-starfive-visionfive-2.dtsi
>>>> @@ -236,3 +236,10 @@
>>>> pinctrl-0 = <&uart0_pins>;
>>>> status = "okay";
>>>> };
>>>> +
>>>> +&usb0 {
>>>> + status = "okay";
>>>> + usbdrd_cdns3: usb@0 {
>>>> + dr_mode = "peripheral";
>>>> + };
>>>> +};
>>>> diff --git a/arch/riscv/boot/dts/starfive/jh7110.dtsi b/arch/riscv/boot/dts/starfive/jh7110.dtsi
>>>> index f70a4ed47eb4..17722fd1be62 100644
>>>> --- a/arch/riscv/boot/dts/starfive/jh7110.dtsi
>>>> +++ b/arch/riscv/boot/dts/starfive/jh7110.dtsi
>>>> @@ -362,6 +362,60 @@
>>>> status = "disabled";
>>>> };
>>>>
>>>> + usb0: usb@10100000 {
>>>> + compatible = "starfive,jh7110-usb";
>>>> + clocks = <&stgcrg JH7110_STGCLK_USB0_LPM>,
>>>> + <&stgcrg JH7110_STGCLK_USB0_STB>,
>>>> + <&stgcrg JH7110_STGCLK_USB0_APB>,
>>>> + <&stgcrg JH7110_STGCLK_USB0_AXI>,
>>>> + <&stgcrg JH7110_STGCLK_USB0_UTMI_APB>;
>>>> + clock-names = "lpm", "stb", "apb", "axi", "utmi_apb";
>>>> + resets = <&stgcrg JH7110_STGRST_USB0_PWRUP>,
>>>> + <&stgcrg JH7110_STGRST_USB0_APB>,
>>>> + <&stgcrg JH7110_STGRST_USB0_AXI>,
>>>> + <&stgcrg JH7110_STGRST_USB0_UTMI_APB>;
>>>> + starfive,stg-syscon = <&stg_syscon 0x4 0xc4 0x148 0x1f4>;
>>>> + starfive,sys-syscon = <&sys_syscon 0x18>;
>>>> + status = "disabled";
>>>> + #address-cells = <1>;
>>>> + #size-cells = <1>;
>>>> + ranges = <0x0 0x0 0x10100000 0x100000>;
>>>> +
>>>> + usbdrd_cdns3: usb@0 {
>>>> + compatible = "cdns,usb3";
>>>
>>> This pattern of USB wrapper and then a "generic" IP node is discouraged
>>> if it is just clocks, resets, power-domains, etc. IOW, unless there's an
>>> actual wrapper h/w block with its own registers, then don't do this
>>> split. Merge it all into a single node.
>>>
>> I am afraid it is difficult to merge in one single node.
>>
>> 1.If cadence3 usb device is still the sub device. All the dts setting are in
>> StarFive node. This can not work.
>> StarFive driver code Using platform_device_add generate cadenc3 usb platform device.
>> Even IO memory space setting can be passed to cadence3 USB, PHY setting can not be passed.
>> For the PHY driver using dts now. But in this case, Cadence3 USB no dts configure.
>>
>> 2. Just one USB Cadence platform device.
>> Maybe this can work. But Cadence USB driver code cdns3-plat.c required to changed.
>>
>> Hi Peter Pawel and Roger
>> There is a "platform_suspend" function pointer in "struct cdns3_platform_data"，
>> Add "platform_init" and "platform_exit" for our JH7110 platform. Maybe it can work.
>> Is it OK?
>
> Once you move all the syscon register modifications from your wrapper driver
> to your PHY driver, only clock and reset control are left in your wrapper driver.
> This is generic enough to be done in the cdns3,usb driver itself so you don't need a
> wrapper node.
>
> Pawel, do you agree?
>
move all the syscon codes to PHY driver is ok. I found dwc3/dwc3-of-simple.c is public codes
and contain just clock and reset control codes. I can change the residual codes to public codes.
But I found rockchip 3399 usb dts which is one of dwc3 simple platform still contain two nodes.
See Documentation/devicetree/bindings/usb/rockchip,rk3399-dwc3.yaml
>>>> + reg = <0x0 0x10000>,
>>>> + <0x10000 0x10000>,
>>>> + <0x20000 0x10000>;
>>>> + reg-names = "otg", "xhci", "dev";
>>>> + interrupts = <100>, <108>, <110>;
>>>> + interrupt-names = "host", "peripheral", "otg";
>>>> + phys = <&usbphy0>;
>>>> + phy-names = "cdns3,usb2-phy";
>>>
>>> No need for *-names when there is only 1 entry. Names are local to the
>>> device and only to distinguish entries, so 'usb2' would be sufficient
>>> here.
>>>
>> The PHY name 'cdns3,usb2-phy' is defined in cadence3 usb driver code.
>> Cadence USB3 driver code using this name to get PHY instance.
>> And all the PHY ops used in Cadence3 USB sub device.
>>>> + maximum-speed = "super-speed";
>>>> + };
>>>> + };
>>>> +
>>>> + usbphy0: phy@10200000 {
>>>> + compatible = "starfive,jh7110-usb-phy";
>>>> + reg = <0x0 0x10200000 0x0 0x10000>;
>>>> + clocks = <&syscrg JH7110_SYSCLK_USB_125M>,
>>>> + <&stgcrg JH7110_STGCLK_USB0_APP_125>;
>>>> + clock-names = "125m", "app_125";
>>>> + #phy-cells = <0>;
>>>> + };
>>>> +
>>>> + pciephy0: phy@10210000 {
>>>> + compatible = "starfive,jh7110-pcie-phy";
>>>> + reg = <0x0 0x10210000 0x0 0x10000>;
>>>> + #phy-cells = <0>;
>>>> + };
>>>> +
>>>> + pciephy1: phy@10220000 {
>>>> + compatible = "starfive,jh7110-pcie-phy";
>>>> + reg = <0x0 0x10220000 0x0 0x10000>;
>>>> + #phy-cells = <0>;
>>>> + };
>>>> +
>>>> stgcrg: clock-controller@10230000 {
>>>> compatible = "starfive,jh7110-stgcrg";
>>>> reg = <0x0 0x10230000 0x0 0x10000>;
>>>> --
>>>> 2.17.1
>>>>
>
> cheers,
> -roger
