On Thu, Oct 07, 2021 at 08:41:55PM +0530, Prasanna Vengateshan wrote:

+static int lan937x_mdio_register(struct dsa_switch *ds)
+{
+	struct device_node *mdio_np;
+	int ret;
+
+	mdio_np = of_get_child_by_name(ds->dev->of_node, "mdio");
+	if (!mdio_np) {
+		dev_err(ds->dev, "no MDIO bus node\n");
+		return -ENODEV;
+	}
+
+	ds->slave_mii_bus = devm_mdiobus_alloc(ds->dev);
+	if (!ds->slave_mii_bus)
+		return -ENOMEM;
+
+	ds->slave_mii_bus->priv = ds->priv;
+	ds->slave_mii_bus->read = lan937x_sw_mdio_read;
+	ds->slave_mii_bus->write = lan937x_sw_mdio_write;
+	ds->slave_mii_bus->name = "lan937x slave smi";
+	snprintf(ds->slave_mii_bus->id, MII_BUS_ID_SIZE, "SMI-%d", ds->index);
+	ds->slave_mii_bus->parent = ds->dev;
+	ds->slave_mii_bus->phy_mask = ~ds->phys_mii_mask;
+
+	ret = of_mdiobus_register(ds->slave_mii_bus, mdio_np);

Please use devm_of_mdiobus_register if you're going to use
devm_mdiobus_alloc, or no devres at all.
https://patchwork.kernel.org/project/netdevbpf/patch/20210920214209.1733768-3-vladimir.oltean@nxp.com/

+	if (ret) {
+		dev_err(ds->dev, "unable to register MDIO bus %s\n",
+			ds->slave_mii_bus->id);
+	}
+
+	of_node_put(mdio_np);
+
+	return ret;
+}
+
+static phy_interface_t lan937x_get_interface(struct ksz_device *dev, int port)
+{
+	phy_interface_t interface;
+	u8 data8;
+	int ret;
+
+	if (lan937x_is_internal_phy_port(dev, port))
+		return PHY_INTERFACE_MODE_NA;

Typically we use PHY_INTERFACE_MODE_INTERNAL.

+
+	/* read interface from REG_PORT_XMII_CTRL_1 register */
+	ret = lan937x_pread8(dev, port, REG_PORT_XMII_CTRL_1, &data8);
+	if (ret < 0)
+		return PHY_INTERFACE_MODE_NA;
+
+	switch (data8 & PORT_MII_SEL_M) {
+	case PORT_RMII_SEL:
+		interface = PHY_INTERFACE_MODE_RMII;
+		break;
+	case PORT_RGMII_SEL:
+		interface = PHY_INTERFACE_MODE_RGMII;
+		if (data8 & PORT_RGMII_ID_EG_ENABLE)
+			interface = PHY_INTERFACE_MODE_RGMII_TXID;
+		if (data8 & PORT_RGMII_ID_IG_ENABLE) {
+			interface = PHY_INTERFACE_MODE_RGMII_RXID;
+			if (data8 & PORT_RGMII_ID_EG_ENABLE)
+				interface = PHY_INTERFACE_MODE_RGMII_ID;
+		}
+		break;
+	case PORT_MII_SEL:
+	default:
+		/* Interface is MII */
+		interface = PHY_INTERFACE_MODE_MII;
+		break;
+	}
+
+	return interface;
+}
+
+static void lan937x_config_cpu_port(struct dsa_switch *ds)
+{
+	struct ksz_device *dev = ds->priv;
+	struct ksz_port *p;
+	int i;
+
+	ds->num_ports = dev->port_cnt;
+
+	for (i = 0; i < dev->port_cnt; i++) {
+		if (dsa_is_cpu_port(ds, i) && (dev->cpu_ports & (1 << i))) {
+			phy_interface_t interface;
+
+			dev->cpu_port = i;
+			dev->host_mask = (1 << dev->cpu_port);
+			dev->port_mask |= dev->host_mask;
+			p = &dev->ports[i];
+
+			/* Check if the device tree have specific interface
+			 * setting otherwise read & assign from XMII register
+			 * for host port interface
+			 */
+			interface = lan937x_get_interface(dev, i);

What does the CPU port have so special that you override it here?
Again some compatibility with out-of-tree DT bindings?

+			if (!p->interface)
+				p->interface = interface;
+
+			dev_info(dev->dev,
+				 "Port%d: using phy mode %s\n",
+				 i,
+				 phy_modes(p->interface));
+
+			/* enable cpu port */
+			lan937x_port_setup(dev, i, true);
+			p->vid_member = dev->port_mask;
+		}
+	}
+
+static u8 lan937x_rgmii_dly_reg_val(int port, u32 val)
+{
+	u8 reg_val;
+
+	/* force minimum delay if delay is less than min delay */
+	if (val && val < 2170)
+		val = 2170;
+
+	/* maximum delay is 4ns */
+	if (val > 4000)
+		val = 4000;

These bindings are new. Given that you also document their min and max
values, why don't you just error out on out-of-range values instead of
silently doing what you think is going to be fine?