Re: [net-next PATCH v5 3/9] net: phy: add devm/of_phy_package_join helper

[Christian Marangi]

On Thu, Feb 01, 2024 at 05:40:28PM +0100, Antoine Tenart wrote:
> Quoting Christian Marangi (2024-02-01 16:17:29)
> > +/**
> > + * of_phy_package_join - join a common PHY group in PHY package
> > + * @phydev: target phy_device struct
> > + * @priv_size: if non-zero allocate this amount of bytes for private data
> > + *
> > + * This is a variant of phy_package_join for PHY package defined in DT.
> > + *
> > + * The parent node of the @phydev is checked as a valid PHY package node
> > + * structure (by matching the node name "ethernet-phy-package") and the
> > + * base_addr for the PHY package is passed to phy_package_join.
> > + *
> > + * With this configuration the shared struct will also have the np value
> > + * filled to use additional DT defined properties in PHY specific
> > + * probe_once and config_init_once PHY package OPs.
> > + *
> > + * Returns < 1 on error, 0 on success. Esp. calling phy_package_join()
> 
> So, < 0 on error ?
> 
> > +int of_phy_package_join(struct phy_device *phydev, size_t priv_size)
> > +{
> > +       struct device_node *node = phydev->mdio.dev.of_node;
> > +       struct device_node *package_node;
> > +       u32 base_addr;
> > +       int ret;
> > +
> > +       if (!node)
> > +               return -EINVAL;
> > +
> > +       package_node = of_get_parent(node);
> 
> Is the node put on package leave?
>

Didn't read the get_parent was incrementing the refcount... Will update
the leave accordingly!

> > +       if (!package_node)
> > +               return -EINVAL;
> > +
> > +       if (!of_node_name_eq(package_node, "ethernet-phy-package")
> 
> of_put_node? + below.
> 
> > +               return -EINVAL;
> > +
> > +       if (of_property_read_u32(package_node, "reg", &base_addr))
> > +               return -EINVAL;
> > +
> > +       ret = phy_package_join(phydev, base_addr, priv_size);
> > +       if (ret)
> > +               return ret;
> > +
> > +       phydev->shared->np = package_node;
> 
> Just looked quickly, looks like ->np is uninitialized in the !of join
> case.
> 

Correct, I will update the non of variant to inizialize ->np to NULL.
(in theory we alloc every struct as zero so it should not be a problem
but you are right with being safe on this.)

Thanks!

> > +
> > +       return 0;
> > +}
> > +EXPORT_SYMBOL_GPL(of_phy_package_join);

-- 
	Ansuel