Re: [PATCH] rust: i2c: avoid locking when calling I2cAdapter::inc_ref

On Sun Aug 9, 2026 at 10:17 AM -03, Igor Korotin wrote:
> Hello Nicolás
>
> Sorry for the delay.
>
> Yes, worth fixing regardless of real users -- inc_ref taking a lock it
> doesn't need is a real bug, not speculative work.
>
> Instead of adding a separate Rust-only wrapper, please consider splitting
> i2c_get_adapter itself: pull the increment into a helper, export it, and
> call it directly from inc_ref -- no lock, no lookup:
>
> +bool __i2c_adapter_get(struct i2c_adapter *adapter)
> +{
> +	if (try_module_get(adapter->owner)) {
> +		get_device(&adapter->dev);
> +		return true;
> +	}
> +	return false;
> +}
> +EXPORT_SYMBOL(__i2c_adapter_get);
> +
>  struct i2c_adapter *i2c_get_adapter(int nr)
>  {
>  	struct i2c_adapter *adapter;
>
>  	mutex_lock(&core_lock);
>  	adapter = idr_find(&i2c_adapter_idr, nr);
> -	if (!adapter)
> -		goto exit;
> -
> -	if (try_module_get(adapter->owner))
> -		get_device(&adapter->dev);
> -	else
> +	if (adapter && !__i2c_adapter_get(adapter))
>  		adapter = NULL;
> -
> - exit:
>  	mutex_unlock(&core_lock);
>  	return adapter;
>  }
>  EXPORT_SYMBOL(i2c_get_adapter);

Excellent! Thank you for this suggestion and the explanation!

>
> Heads up for v2: Trevor Chan has a patch changing AlwaysRefCounted::inc_ref
> to an associated function (fn inc_ref(obj: &Self)). Not merged yet --
> rebase onto it if it lands first.

I think it is better to wait until Trevor Chan's patch gets merged since
it is in v7 [1] and already has been reviewed by some mantainers. If I send 
this fix and gets merged first he will probably have conflicts and have
to send a v8.

I'll keep an eye on the progress of that patch and once it is merged I
am going to send this fix.

Thank you,
Nicolás

[1] https://lore.kernel.org/rust-for-linux/[email protected]/T/#u