Re: [PATCH v5 3/3] clk: imx: imx8: add audio clock mux driver

[Stephen Boyd <sboyd@kernel.org>]

Quoting Shengjiu Wang (2023-07-20 00:31:05)
> On Thu, Jul 20, 2023 at 4:17 AM Stephen Boyd <sboyd@kernel.org> wrote:
>     Quoting Shengjiu Wang (2023-07-14 01:10:59)
>     > +/**
>     > + * struct imx8_acm_soc_data: soc specific data
>     > + * @sels: pointer to struct clk_imx8_acm_sel
>     > + * @num_sels: numbers of items
>     > + */
>     > +struct imx8_acm_soc_data {
>     > +       struct clk_imx8_acm_sel *sels;
>     > +       unsigned int num_sels;
>     > +};
>     > +
>     > +/**
>     > + * struct imx8_acm_priv: private structure
> 
>     Compile with W=1 and see that this isn't kerneldoc. Please fix.
> 
> 
> I have used the W=1,  but there is not error or warning from kerneldoc
> 
> My command is:
> make W=1 ARCH=arm64 CROSS_COMPILE=/opt/toolchain/
> gcc-linaro-14.0.0-2023.06-x86_64_aarch64-linux-gnu/bin/aarch64-linux-gnu- -j4
> 
> Anything I missed?

I thought W=1 ran kerneldoc check, but maybe not by default. Look at
the docs:

https://docs.kernel.org/doc-guide/kernel-doc.html#structure-union-and-enumeration-documentation

> 
> 
> 
>     > + * @dev_pm: multi power domain
>     > + * @soc_data: pointer to soc data
>     > + * @reg: base address of registers
>     > + * @regs: save registers for suspend
>     > + */
>     > +struct imx8_acm_priv {
>     > +       struct clk_imx_acm_pm_domains dev_pm;
>     > +       const struct imx8_acm_soc_data *soc_data;
>     > +       void __iomem *reg;
>     > +       u32 regs[IMX_ADMA_ACM_CLK_END];
>     > +};
>     > +
>     > +static const struct clk_parent_data imx8qm_aud_clk_sels[] = {
>     > +       {.fw_name = "aud_rec_clk0_lpcg_clk", .name =
>     "aud_rec_clk0_lpcg_clk" },
> 
>     There should only be fw_name here, or use an index. Presumably this
>     isn't migrating old code or bindings. Also, please add space after { and
>     before }.
> 
> 
> The clock  "aud_rec_clk0_lpcg_clk" is not defined in this provider. if
> remove .name There is an issue for finding parent.

The .fw_name should be part of the DT binding. If it isn't part of the
binding as an element of clock-names then it shouldn't be in the driver
as a .fw_name. If it isn't part of the provider, then it should be
referenced by DT index or clock-names. If you use the index, you avoid a
possibly time consuming set of string comparisons. So if you can, use
the index as much as possible. If you're converting a pre-existing
binding you will have to use .name as well to fallback to the globally
unique string name for a clk.

> 
> Should I remove .fw_name , only keep .name?

You should remove .name it sounds like.