Re: [net-next RFC PATCH 04/14] net: phy: add initial support for PHY package in DT

[Simon Horman <horms@kernel.org>]

On Mon, Nov 20, 2023 at 02:50:31PM +0100, Christian Marangi wrote:
> Add initial support for PHY package in DT.
> 
> Make it easier to define PHY package and describe the global PHY
> directly in DT by refereincing them by phandles instead of custom
> functions in each PHY driver.
> 
> Each PHY in a package needs to be defined in a dedicated node in the
> mdio node. This dedicated node needs to have the compatible set to
> "ethernet-phy-package" and define "global-phys" and "#global-phy-cells"
> respectively to a list of phandle to the global phy to define for the
> PHY package and 0 for cells as the phandle won't take any args.
> 
> With this defined, the generic PHY probe will join each PHY in this
> dedicated node to the package.
> 
> PHY driver MUST set the required global PHY count in
> .phy_package_global_phy_num to correctly verify that DT define the
> correct number of phandle to the required global PHY.
> 
> mdio_bus.c and of_mdio.c is updated to now support and parse also
> PHY package subnote that have the compatible "phy-package".
> 
> Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>

Hi Christian,

I was a little hasty in hitting send on my previous message.
Please find some more minor feedback from my side below.

...

> diff --git a/drivers/net/mdio/of_mdio.c b/drivers/net/mdio/of_mdio.c
> index 64ebcb6d235c..bb910651118f 100644
> --- a/drivers/net/mdio/of_mdio.c
> +++ b/drivers/net/mdio/of_mdio.c
> @@ -139,6 +139,44 @@ bool of_mdiobus_child_is_phy(struct device_node *child)
>  }
>  EXPORT_SYMBOL(of_mdiobus_child_is_phy);
>  
> +static int __of_mdiobus_parse_phys(struct mii_bus *mdio, struct device_node *np,
> +				   bool *scanphys)
> +{
> +	struct device_node *child;
> +	int addr, rc;
> +
> +	/* Loop over the child nodes and register a phy_device for each phy */
> +	for_each_available_child_of_node(np, child) {
> +		if (of_device_is_compatible(child, "ethernet-phy-package")) {
> +			rc = __of_mdiobus_parse_phys(mdio, child, scanphys);
> +			if (rc && rc != -ENODEV)
> +				return rc;

for_each_available_child_of_node() makes calls to of_node_get() and
of_node_put(), so when jumping out of a loop it is necessary to call
of_node_put(), in this case of_node_put(child).

As flagged by Coccinelle.

Also flagged in of_mdiobus_find_phy() both before and after this patch.

> +
> +			continue;
> +		}
> +
> +		addr = of_mdio_parse_addr(&mdio->dev, child);
> +		if (addr < 0) {
> +			*scanphys = true;
> +			continue;
> +		}
> +
> +		if (of_mdiobus_child_is_phy(child))
> +			rc = of_mdiobus_register_phy(mdio, child, addr);
> +		else
> +			rc = of_mdiobus_register_device(mdio, child, addr);
> +
> +		if (rc == -ENODEV)
> +			dev_err(&mdio->dev,
> +				"MDIO device at address %d is missing.\n",
> +				addr);
> +		else if (rc)
> +			return rc;
> +	}
> +
> +	return 0;
> +}
> +
>  /**
>   * __of_mdiobus_register - Register mii_bus and create PHYs from the device tree
>   * @mdio: pointer to mii_bus structure
> @@ -180,25 +218,9 @@ int __of_mdiobus_register(struct mii_bus *mdio, struct device_node *np,
>  		return rc;
>  
>  	/* Loop over the child nodes and register a phy_device for each phy */
> -	for_each_available_child_of_node(np, child) {
> -		addr = of_mdio_parse_addr(&mdio->dev, child);
> -		if (addr < 0) {
> -			scanphys = true;
> -			continue;
> -		}
> -
> -		if (of_mdiobus_child_is_phy(child))
> -			rc = of_mdiobus_register_phy(mdio, child, addr);
> -		else
> -			rc = of_mdiobus_register_device(mdio, child, addr);
> -
> -		if (rc == -ENODEV)
> -			dev_err(&mdio->dev,
> -				"MDIO device at address %d is missing.\n",
> -				addr);
> -		else if (rc)
> -			goto unregister;
> -	}
> +	rc = __of_mdiobus_parse_phys(mdio, np, &scanphys);
> +	if (rc)
> +		goto unregister;

Jumping to unregister will call of_node_put(child),
however child appears to be uninitialised here.

Flagged by clang-16 W=1 build, and Smatch.

>  
>  	if (!scanphys)
>  		return 0;

...