Re: [PATCH RESEND v2] rust: regulator: add a bare minimum regulator abstraction

[Daniel Almeida <daniel.almeida@collabora.com>]

Hi Mark,

> On 27 Mar 2025, at 08:32, Mark Brown <broonie@kernel.org> wrote:
> 
> On Wed, Mar 26, 2025 at 04:49:26PM -0300, Daniel Almeida wrote:
>>> On 26 Mar 2025, at 15:56, Mark Brown <broonie@kernel.org> wrote:
> 
>>>> +    /// Disables the regulator.
>>>> +    pub fn disable(self) -> Result<Regulator> {
>>>> +        // Keep the count on `regulator_get()`.
>>>> +        let regulator = ManuallyDrop::new(self);
> 
>>> This looks like user code could manually call it which feels like asking
>>> for trouble?
> 
>> Yes, user code can call this. My understanding is that drivers may want to
>> disable the regulator at runtime, possibly to save power when the device is
>> idle?
> 
>> What trouble are you referring to?
> 
> My understanding was that the enable was done by transforming a
> Regulator into an EnabledRegulator but if you can explicitly call
> disable() on an EnabledRegulator without destroying it then you've got
> an EnabledRegulator which isn't actually enabled.  Perhaps it's not
> clear to me how the API should work?

No, you misunderstood a bit, but that’s on me, I should have included examples.

> +impl EnabledRegulator {

> +    /// Disables the regulator.
> +    pub fn disable(self) -> Result<Regulator> 

disable() consumes EnabledRegulator to return Regulator.

Any function that takes 'self' by value (i.e.: “self" instead of “&self” )
effectively kills it. So, in that sense, disable() performs a conversion
between the two types after calling regulator_disable().

— Daniel