Re: [PATCH v3 2/9] clk: ralink: add clock and reset driver for MTMIPS SoCs

[Krzysztof Kozlowski]

On 17/06/2023 15:31, Sergio Paracuellos wrote:
> Hi Krzysztof,
> 
> On Sat, Jun 17, 2023 at 3:07 PM Krzysztof Kozlowski
> <krzysztof.kozlowski@xxxxxxxxxx> wrote:
>>
>> On 17/06/2023 14:54, Shiji Yang wrote:
>>>> void __init plat_time_init(void)
>>>> {
>>>> +    struct of_phandle_args clkspec;
>>>>      struct clk *clk;
>>>> +    int cpu_clk_idx;
>>>>
>>>>      ralink_of_remap();
>>>>
>>>> -    ralink_clk_init();
>>>> -    clk = clk_get_sys("cpu", NULL);
>>>> +    cpu_clk_idx = clk_cpu_index();
>>>> +    if (cpu_clk_idx == -1)
>>>> +            panic("unable to get CPU clock index");
>>>> +
>>>> +    of_clk_init(NULL);
>>>> +    clkspec.np = of_find_node_by_name(NULL, "sysc");
>>>
>>> The node name should be "syscon" as the example node name in the
>>> dt-bindings document is "syscon".
>>
>> NAK for both.
>>
>> Node names must not be an ABI, unless you talk about child of some
>> device node. I don't think this is the case here. Look by phandle (for a
>> device context) or by compatible (looks the case here).
> 
> We need to get the cpu clock to set the initial cpu clock here. Search
> by 'sysc' is the only  shared in all the dtsi files since it is the
> clock provider node. Why is this not correct?

Because device node name can change anytime and your entire Linux driver
gets broken. Node name is not an ABI.

> I don't understand what
> you mean with look by phandle for a device context. 

Your device node should contain phandle to the other node.

> The case of
> searching for compatible is a mess since as you can see in the
> bindings there are tons of compatibles to search for, then (this code
> is common to all ralink platforms).

Compatible is one of the ways using ABI.

Best regards,
Krzysztof