Re: [PATCH v5 3/4] rust: clk: add devres-managed clks

On Mon, 06 Jul 2026 11:37:14 -0300=0D
Daniel Almeida <[email protected]> wrote:=0D
=0D
> The clk API allows fine-grained control, but some drivers might be=0D
> more interested in a "set and forget" API.=0D
> =0D
> Expand the current API to support this. The clock will automatically be=0D
> disabled, unprepared and freed when the device is unbound from the bus=0D
> without further intervention by the driver.=0D
> =0D
> Signed-off-by: Daniel Almeida <[email protected]>=0D
> ---=0D
>  rust/kernel/clk.rs | 46 ++++++++++++++++++++++++++++++++++++++++++++++=0D
>  1 file changed, 46 insertions(+)=0D
> =0D
> diff --git a/rust/kernel/clk.rs b/rust/kernel/clk.rs=0D
> index dd5fd656271e..b9a44f83843a 100644=0D
> --- a/rust/kernel/clk.rs=0D
> +++ b/rust/kernel/clk.rs=0D
> @@ -96,6 +96,52 @@ impl Sealed for super::Prepared {}=0D
>          impl Sealed for super::Enabled {}=0D
>      }=0D
>  =0D
> +    /// Obtains and enables a [`devres`]-managed [`Clk`] for a bound dev=
ice.=0D
> +    ///=0D
> +    /// [`devres`]: crate::devres::Devres=0D
> +    #[inline]=0D
> +    pub fn devm_enable(dev: &Device<Bound>, name: Option<&CStr>) -> Resu=
lt {=0D
> +        let name =3D name.map_or(ptr::null(), |n| n.as_char_ptr());=0D
> +=0D
> +        // SAFETY: It is safe to call [`devm_clk_get_enabled`] with a va=
lid=0D
=0D
Intra-doc links won't work in regular comments.=0D
=0D
> +        // device pointer.=0D
> +        from_err_ptr(unsafe { bindings::devm_clk_get_enabled(dev.as_raw(=
), name) })?;=0D
> +        Ok(())=0D
> +    }=0D
> +=0D
> +    /// Obtains and enables a [`devres`]-managed [`Clk`] for a bound dev=
ice.=0D
> +    ///=0D
> +    /// This does not print any error messages if the clock is not found=
.=0D
> +    ///=0D
> +    /// [`devres`]: crate::devres::Devres=0D
> +    #[inline]=0D
> +    pub fn devm_enable_optional(dev: &Device<Bound>, name: Option<&CStr>=
) -> Result {=0D
> +        let name =3D name.map_or(ptr::null(), |n| n.as_char_ptr());=0D
> +=0D
> +        // SAFETY: It is safe to call [`devm_clk_get_optional_enabled`] =
with a=0D
> +        // valid device pointer.=0D
> +        from_err_ptr(unsafe { bindings::devm_clk_get_optional_enabled(de=
v.as_raw(), name) })?;=0D
> +        Ok(())=0D
> +    }=0D
> +=0D
> +    /// Same as [`devm_enable_optional`], but also sets the rate.=0D
> +    #[inline]=0D
> +    pub fn devm_enable_optional_with_rate(=0D
> +        dev: &Device<Bound>,=0D
> +        name: Option<&CStr>,=0D
> +        rate: Hertz,=0D
> +    ) -> Result {=0D
> +        let name =3D name.map_or(ptr::null(), |n| n.as_char_ptr());=0D
> +=0D
> +        // SAFETY: It is safe to call=0D
> +        // [`devm_clk_get_optional_enabled_with_rate`] with a valid devi=
ce=0D
> +        // pointer.=0D
> +        from_err_ptr(unsafe {=0D
> +            bindings::devm_clk_get_optional_enabled_with_rate(dev.as_raw=
(), name, rate.as_hz())=0D
> +        })?;=0D
=0D
Just fyi, there's currently a bug in the C implementation of=0D
devm_clk_get_optional_enabled_with_rate() which I caught while reviewing yo=
ur=0D
series. The fix patch is already on the list [1]. Other than that, this LGT=
M.=0D
=0D
[1]: https://lore.kernel.org/all/[email protected]=
=0D
=0D
Regards,=0D
Onur=0D
=0D
> +        Ok(())=0D
> +    }=0D
> +=0D
>      /// A trait representing the different states that a [`Clk`] can be =
in.=0D
>      pub trait ClkState: private::Sealed {=0D
>          /// Whether the clock is enabled in this state.=0D
> =0D
> -- =0D
> 2.54.0=0D
> =0D