Re: [PATCH v5 1/7] dt-bindings: connector: Add mikrobus-connector

[Ayush Singh <ayush@beagleboard.org>]

On 9/4/24 23:19, Rob Herring wrote:

> On Wed, Sep 4, 2024 at 12:09 PM Ayush Singh <ayush@beagleboard.org> wrote:
>>>> gpio-map is what you are looking for. It's documented in the DT spec.
>>>> It was created exactly for this purpose of remapping GPIO lines on a
>>>> connector.
>>>>
>>>> Rob
>>
>> Hi. I found docs on nexus nodes [1] and tried using it for mikroBUS, but
>> it does not seem to be working. Here is my connector:
>>
>> ```
>>
>>       mikrobus_gpio0: mikrobus-gpio0 {
>>           #gpio-cells = <2>;
>>           gpio-map =
>>           <0 0 &main_gpio1 11 0>, <1 0 &main_gpio1 9 0>,
>>           <2 0 &main_gpio1 24 0>, <3 0 &main_gpio1 25 0>,
>>           <4 0 &main_gpio1 22 0>, <5 0 &main_gpio1 23 0>,
>>           <6 0 &main_gpio1 7 0>, <7 0 &main_gpio1 8 0>,
>>           <8 0 &main_gpio1 14 0>, <9 0 &main_gpio1 13 0>,
>>           <10 0 &main_gpio1 12 0>, <11 0 &main_gpio1 10 0>;
>>           gpio-map-mask = <0xf 0x0>;
>>           gpio-map-pass-thru = <0x0 0x1>;
>>       };
>>
>> ...
>>
>> &main_uart5 {
>>       status = "okay";
>>       pinctrl-names = "default";
>>       pinctrl-0 = <&mikrobus_uart_pins_default>;
>>
>>       gnss {
>>           compatible = "u-blox,neo-8";
>>           reset-gpios = <&mikrobus_gpio0 10 GPIO_ACTIVE_LOW>;
>>       };
>> };
>>
>> ```
>>
>>
>> After some fdtdump, I can see that at least the dtc compiler does not
>> seem to do the forwarding at dt compile time. Here is the dump:
> dtc knows nothing about it.
>
>> ```
>>
>> mikrobus-gpio0 {
>>           #gpio-cells = <0x00000002>;
>>           gpio-map = <0x00000000 0x00000000 0x00000025 0x0000000b
>> 0x00000000 0x00000001 0x00000000 0x00000025 0x00000009 0x00000000
>> 0x00000002 0x00000000 0x00000025 0x00000018 0x00000000 0x00000003
>> 0x00000000 0x00000025 0x00000019 0x00000000 0x00000004 0x00000000
>> 0x00000025 0x00000016 0x00000000 0x00000005 0x00000000 0x00000025
>> 0x00000017 0x00000000 0x00000006 0x00000000 0x00000025 0x00000007
>> 0x00000000 0x00000007 0x00000000 0x00000025 0x00000008 0x00000000
>> 0x00000008 0x00000000 0x00000025 0x0000000e 0x00000000 0x00000009
>> 0x00000000 0x00000025 0x0000000d 0x00000000 0x0000000a 0x00000000
>> 0x00000025 0x0000000c 0x00000000 0x0000000b 0x00000000 0x00000025
>> 0x0000000a 0x00000000>;
>>           gpio-map-mask = <0x0000000f 0x00000000>;
>>           gpio-map-pass-thru = <0x00000000 0x00000001>;
>>           phandle = <0x0000000e>;
>>       };
> You might need "gpio-controller" here. Though if you do, I think
> that's a mistake in the kernel. It should work like interrupt-map and
> generally you have either interrupt-controller or interrupt-map, but
> not both (though that is allowed now too).
>
>> ...
>>
>> serial@2850000 {
>>               compatible = "ti,am64-uart", "ti,am654-uart";
>>               reg = <0x00000000 0x02850000 0x00000000 0x00000100>;
>>               interrupts = <0x00000000 0x000000b7 0x00000004>;
>>               power-domains = <0x00000003 0x0000009c 0x00000001>;
>>               clocks = <0x00000002 0x0000009c 0x00000000>;
>>               clock-names = "fclk";
>>               status = "okay";
>>               pinctrl-names = "default";
>>               pinctrl-0 = <0x0000000d>;
>>               phandle = <0x00000081>;
>>               gnss {
>>                   compatible = "u-blox,neo-8";
>>                   reset-gpios = <0x0000000e 0x0000000a 0x00000001>;
>>               };
>>    };
>>
>> ```
>>
>>
>> So I am a bit unsure. Is the dtc parser in the kernel supposed to do the
> No such thing as "dtc parser in the kernel".
>
>> mapping, or is it supposed to be done by `dtc` at compile time?
> No.
>
>> Maybe we
>> do not have support for it in upstream kernel yet?
> Yes, there is upstream support. Grep for of_parse_phandle_with_args_map.
>
> Rob


So, after a bit of troubleshooting, it seems that a nexus node should 
not be present at root level (unless it also has an actual driver). If 
the nexus node is a root node without an actual driver, anything 
referring to the node is deferred.


I am still a bit unsure if I should make the `mikrobus-connector` itself 
a nexus node or if I should have a subnode named `mikrobus_gpio`, but at 
least things seem to be working now. So, thanks for your help.


Ayush Singh