Re: [PATCH net-next v7 2/7] net: add framework to support Ethernet PSE and PDs devices

[Jakub Kicinski <kuba@kernel.org>]

On Mon, 26 Sep 2022 13:24:55 +0200 Oleksij Rempel wrote:
> +static inline int pse_controller_register(struct pse_controller_dev *pcdev)
> +{
> +	return -ENOTSUPP;
> +}
> +
> +static inline void pse_controller_unregister(struct pse_controller_dev *pcdev)
> +{
> +}
> +
> +static inline int devm_pse_controller_register(struct device *dev,
> +						 struct pse_controller_dev *pcdev)
> +{
> +	return -ENOTSUPP;
> +}

Presumably only PSE controller drivers would try to register themselves.
Such drivers should depend on the right config, and therefore we don't
need static inline stubs for the register/unregister API.

> +static inline struct pse_control *pse_control_get(struct device *dev)
> +{
> +	return ERR_PTR(-ENOTSUPP);
> +}
> +
> +static inline struct pse_control *devm_pse_control_get( struct device *dev)

nit: extra space after (

> +{
> +	return ERR_PTR(-ENOTSUPP);
> +}

These two I don't see any calls to outside drivers/net/pse-pd/pse_core.c
so they should go from the API until we get an in-tree caller.

> +static inline struct pse_control *of_pse_control_get(struct device_node *node)
> +{
> +	return ERR_PTR(-ENOTSUPP);
> +}

This one should prolly return -ENOENT as noted on patch 4.

If you could sed -i 's/ENOTSUPP/EOPNOTSUPP/' on the patches that'd be
great, I don't think those errno can leak to user space but why risk
it...