On 19/08/2021 17:17:05+0000, Vladimir Oltean wrote:

On Thu, Aug 19, 2021 at 08:04:16PM +0300, Vladimir Oltean wrote:

quoted

The ocelot driver was converted to phylink, and that expects a valid
phy_interface_t. Without a phy-mode, of_get_phy_mode returns
PHY_INTERFACE_MODE_NA, which is not ideal because phylink rejects that.

The ocelot driver was patched to treat PHY_INTERFACE_MODE_NA as
PHY_INTERFACE_MODE_INTERNAL to work with the broken DT blobs, but we
should fix the device trees and specify the phy-mode too.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
---

Please note that the pre-phylink driver has this check:

		switch (ocelot_port->phy_mode) {
		case PHY_INTERFACE_MODE_NA:
			(...)
		case PHY_INTERFACE_MODE_SGMII:
			(...)
		case PHY_INTERFACE_MODE_QSGMII:
			(...)
		default:
			dev_err(ocelot->dev,
				"invalid phy mode for port%d, (Q)SGMII only\n",
				port);
			of_node_put(portnp);
			err = -EINVAL;
			goto out_teardown;
		}

So it does not actually expect PHY_INTERFACE_MODE_INTERNAL and will
error out.

Are we okay with the new device tree blobs breaking the old kernel?

From my point of view, newer device trees are not required to work on
older kernel, this would impose an unreasonable limitation and the use
case is very limited.

-- 
Alexandre Belloni, co-owner and COO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com