Re: [PATCH v3] rust: regulator: add a bare minimum regulator abstraction

["Benno Lossin" <lossin@kernel.org>]

On Wed May 14, 2025 at 6:19 PM CEST, Daniel Almeida wrote:
>> On 14 May 2025, at 13:05, Benno Lossin <lossin@kernel.org> wrote:
>> On Wed May 14, 2025 at 5:50 PM CEST, Mark Brown wrote:
>>> On Wed, May 14, 2025 at 05:38:40PM +0200, Benno Lossin wrote:
>>>> On Wed May 14, 2025 at 4:40 PM CEST, Daniel Almeida wrote:
>>>>> By the way, IIUC, regulator_disable() does not disable a regulator necessarily.
>>>>> It just tells the system that you don't care about it being enabled anymore. It can
>>>>> still remain on if there are other users.
>>> 
>>>> Hmm, so a `struct regulator` might already be enabled and calling
>>>> `regulator_enable` doesn't do anything?
>>> 
>>> It takes a reference to the regulator.  This may or may not result in a
>>> change in an underlying physical regulator.
>> 
>> Gotcha. So calling `regulator_enable` twice on the same regulator is
>> fine?
>> 
>> If that is the case -- and after re-reading the functions exposed on
>> both types `EnabledRegulator` and `Regulator` -- I am confused why we
>> even need two different type states? Both expose the same functions
>> (except `enable` and `disable`) and I don't otherwise see the purpose of
>> having two types.
>> 
>> ---
>> Cheers,
>> Benno
>> 
>
>
> As Mark said:
>
>> IIUC the point is to allow Rust's type system to keep track of the
>> reference on the regulator, otherwise the user code has to keep track of
>> the number of enables it's done like it currently does in C code.
>
> So this all started because keeping track of the enables was rather clunky. See
> v1 [0].
>
> [0] https://lore.kernel.org/rust-for-linux/20250219162517.278362-1-daniel.almeida@collabora.com/

Ahh thanks for the pointer this makes much more sense now. Yeah that's a
reason to have two types.

Please document this properly. A good reference is
`rust/kernel/fs/file.rs`. It also deals with different refcounts and
ownership, so that might give you some idea for how to write the
comments around changes in ownership. Also have a general comment about
the two different refcounts in a `struct regulator` on the two regulator
wrappers.

---
Cheers,
Benno