Re: [net-next PATCH v1] net: dpaa2-mac: Add ACPI support for DPAA2 MAC driver

[Calvin Johnson <calvin.johnson@oss.nxp.com>]

On Thu, Jun 25, 2020 at 09:42:11PM +0200, Andrew Lunn wrote:
> On Thu, Jun 25, 2020 at 10:05:38AM +0530, Calvin Johnson wrote:
> 
> > +static struct phy_device *dpaa2_find_phy_device(struct fwnode_handle *fwnode)
> > +{
> > +	struct fwnode_reference_args args;
> > +	struct platform_device *pdev;
> > +	struct mii_bus *mdio;
> > +	struct device *dev;
> > +	acpi_status status;
> > +	int addr;
> > +	int err;
> > +
> > +	status = acpi_node_get_property_reference(fwnode, "mdio-handle",
> > +						  0, &args);
> > +
> > +	if (ACPI_FAILURE(status) || !is_acpi_device_node(args.fwnode))
> > +		return NULL;
> > +
> > +	dev = bus_find_device_by_fwnode(&platform_bus_type, args.fwnode);
> > +	if (IS_ERR_OR_NULL(dev))
> > +		return NULL;
> > +	pdev =  to_platform_device(dev);
> > +	mdio = platform_get_drvdata(pdev);
> > +
> > +	err = fwnode_property_read_u32(fwnode, "phy-channel", &addr);
> > +	if (err < 0 || addr < 0 || addr >= PHY_MAX_ADDR)
> > +		return NULL;
> > +
> > +	return mdiobus_get_phy(mdio, addr);
> > +}
> 
> Hi Calvin
> 
> I think this needs putting somewhere global, since you are effectively
> defines how all ACPI MACs will find their PHY. This becomes the
> defacto standard until the ACPI standards committee comes along and
> tells you, you are doing it wrong, it should be like....
> 
> Does Linux have a location to document all its defacto standard ACPI
> stuff? It would be good to document this somewhere.

Hi Andrew

As you know, making this code generic would bring us back to waiting for
ACPI team's approval which is very difficult to get as the ACPI doesn't
have any opinion on MDIO bus.

Like other ACPI ethernet drivers, can't we keep it local to this driver to
avoid the above issue?

I can add more documentation to this function itself.
If we plan to make this approach generic, then it may have to be put in:
Documentation/firmware-guide/acpi/

Please advice.

Thanks
Calvin