pon., 20 cze 2022 o 19:41 Andy Shevchenko
[off-list ref] napisał(a):

On Mon, Jun 20, 2022 at 05:02:16PM +0200, Marcin Wojtas wrote:

quoted

In order to support both ACPI and DT, modify the generic
DSA code to use device_/fwnode_ equivalent routines.
No functional change is introduced by this patch.

...

quoted

      struct device_node      *dn;

What prevents us from removing this?

I left it to satisfy possible issues with backward compatibility - I
migrated mv88e6xxx, other DSA drivers still rely on of_* and may use
this field.

quoted

+     struct fwnode_handle    *fwnode;

...

quoted

-             dn = of_get_child_by_name(ds->dev->of_node, "mdio");
+             fwnode = fwnode_get_named_child_node(ds->dev->fwnode, "mdio");

The rule of thumb is avoid dereferencing fwnode from struct device. So
dev_fwnode(), but here it would be achieved by device_get_named_child_node().

Ok, thanks - will do for all occurences.

...

quoted

-static int dsa_switch_parse_of(struct dsa_switch *ds, struct device_node *dn)
+static int dsa_switch_parse_of(struct dsa_switch *ds, struct fwnode_handle *fwnode)

Shouldn't _of suffix be replaced by, let's say, _fw?

I thought about it and can perform such naming update in next iteration.

...

quoted

-     return dsa_switch_parse_ports_of(ds, dn);
+     return dsa_switch_parse_ports_of(ds, fwnode);

Ditto.

...

quoted

+     fwnode = ds->dev->fwnode;

dev_fwnode() or corresponding device_property_ API.

OK.

...

quoted

      slave_dev->dev.of_node = port->dn;
+     slave_dev->dev.fwnode = port->fwnode;

device_set_node()

OK.

Thanks,
Marcin