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

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

On Mon May 19, 2025 at 1:46 PM CEST, Mark Brown wrote:
> On Mon, May 19, 2025 at 01:25:56PM +0200, Benno Lossin wrote:
>> On Mon May 19, 2025 at 11:56 AM CEST, Mark Brown wrote:
>
>> > No.  You should not leak any refcount, the per consumer refcount
>> > duplicates what's being done for the regulator as a whole, one should
>> > never be incremented or decremented without the other (but there may be
>> > multiple consumers to choose from).
>
>> What stops the last `regulator_put` to also call `regulator_disable` a
>> correct number of times?
>
> We obviously could but the regulator API defaults to not doing anything
> unless explicitly told to since getting things wrong can physically
> damage the system.  We've no idea if just disabling the regulator would
> be safe at this point so we just don't touch anything and complain about
> it.

Gotcha.

>> What are the kinds of problems that one could encounter when not calling
>> `regulator_disable` before `regulator_put` or if `regulator_enable` was
>> never called to begin with?
>
> If you don't disable the regulator you've just leaked a reference which
> is obviously a problem.

For sure. But I'm trying to figure out if this is a safety-related issue
or not. Safety in Rust has a rather specific meaning that can be
summarized with "no UB". So since the C side does nothing if the user
screwed up the refcounts, it lets me to believe that we don't have any
safety related issues when forgetting to call `regulator_disable`.

Of course we still should strive for an API that makes that impossible
or at least very hard, but we don't need to make the API `unsafe` or
have to take special care. (At least if I understood correctly)

---
Cheers,
Benno