Re: [PATCH v4 11/21] riscv: Add Canaan Kendryte K210 clock driver

[Damien Le Moal <Damien.LeMoal@xxxxxxx>]

On 2020/12/07 19:06, Geert Uytterhoeven wrote:
> Hi Damien,
> 
> On Mon, Dec 7, 2020 at 10:55 AM Damien Le Moal <Damien.LeMoal@xxxxxxx> wrote:
>> On 2020/12/07 17:44, Geert Uytterhoeven wrote:
>>> On Mon, Dec 7, 2020 at 4:52 AM Damien Le Moal <Damien.LeMoal@xxxxxxx> wrote:
>>>> I prepared a v5 series addressing your comments (and other comments).
>>>> I will post that later today after some more tests.
>>>
>>> Thanks, already looking at k210-sysctl-v18...
>>>
>>>> On Fri, 2020-12-04 at 22:19 -0800, Stephen Boyd wrote:
>>>>>> --- /dev/null
>>>>>> +++ b/drivers/clk/clk-k210.c
>>>
>>>>>> +       in0_clk = of_clk_get(np, 0);
>>>>>> +       if (IS_ERR(in0_clk)) {
>>>>>> +               pr_warn("%pOFP: in0 oscillator not found\n", np);
>>>>>> +               hws[K210_CLK_IN0] =
>>>>>> +                       clk_hw_register_fixed_rate(NULL, "in0", NULL,
>>>>>> +                                                  0, K210_IN0_RATE);
>>>>>> +       } else {
>>>>>> +               hws[K210_CLK_IN0] = __clk_get_hw(in0_clk);
>>>>>> +       }
>>>>>> +       if (IS_ERR(hws[K210_CLK_IN0])) {
>>>>>> +               pr_err("%pOFP: failed to get base oscillator\n", np);
>>>>>> +               goto err;
>>>>>> +       }
>>>>>> +
>>>>>> +       in0 = clk_hw_get_name(hws[K210_CLK_IN0]);
>>>>>> +       aclk_parents[0] = in0;
>>>>>> +       pll_parents[0] = in0;
>>>>>> +       mux_parents[0] = in0;
>>>>>
>>>>> Can we use the new way of specifying clk parents so that we don't have
>>>>> to use __clk_get_hw(), of_clk_get(), and clk_hw_get_name()? Hopefully
>>>>> the core can handl that all instead of this driver.
>>>>
>>>> I removed all this by adding:
>>>>
>>>> clock-output-names = "in0";
>>>>
>>>> to the DT fixed-rate oscillator clock node (and documented that too). Doing so,
>>>> clk_hw_get_name(), __clk_get_hw() and of_clk_get() are not needed anymore and
>>>> the parents clock names arrays do not need run-time update.
>>>
>>> "clock-output-names" is deprecated for clocks with a single output:
>>> the clock name will be taken from the node name.
>>
>> Arg. I missed that.
>>
>>> However, relying on a clock name like this is fragile.
>>> Instead, your driver should use the phandle from the clocks property,
>>> using of_clk_get_by_name() or of_clk_get().
>>
>> That is what all versions before V5 used. But Stephen mentioned that the driver
>> should not, if possible, use of_clk_get()/__clk_get_name(). Hence the change.
>> Easy to revert back.
>>
>>> Stephen: I'm a bit puzzled, as you suggest _not_ using of_clk_get()?
>>
>> Another solution to this would be to simply remove the fixed-rate clock node
>> from the DT and have the k210 clock driver unconditionally create that clock
>> (that is one line !). That actually may be even more simple than the previous
>> version, albeit at the cost of having the DT not being a perfect description of
>> the hardware. I am fine with that though.
>>
>> Thoughts ?
> 
> If there's an external crystal, DT should describe it.
> Does the K210 support different crystal frequencies?

I am not 100% sure if this oscillator is part of the SoC or if it is an external
input to it. Probably not. Hard to tell by just looking at the boards. I have
the boards drawings though, so  I will check. The frequency seems to be fixed by
hardware: frequencies of the PLLs can be changed to change the CPU frequency,
but the Kendryte SDK does not point to any way allowing changing the base
frequency of the oscillator.

> Anyway, I'm very interested in what the (new) proper way of handling this
> is...
> 
> Gr{oetje,eeting}s,
> 
>                         Geert
> 
> --
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
> 
> In personal conversations with technical people, I call myself a hacker. But
> when I'm talking to journalists I just say "programmer" or something like that.
>                                 -- Linus Torvalds
> 


-- 
Damien Le Moal
Western Digital Research