On Mon, Oct 3, 2022 at 3:35 PM Fabio Estevam festevam@denx.de wrote:
On 03/10/2022 16:09, Tim Harvey wrote:
Right - so it looks like the 'net: fec: add support for DM_MDIO' patch is not doing its job to probe the mdio bus and ports.
Is CONFIG_DM_MDIO enabled and is dm_fec_bind_mdio returning successfully?
Yes, for both.
Is CONFIG_DM_ETH_PHY enabled? I don't use that and if I recall that code never seemed right to me.
CONFIG_DM_ETH_PHY is not enabled.
Fabio,
I'm not sure why your configuration is behaving differently and think you'll have to instrument code to help understand it.
Here is the chain of events I see on my imx6q fec <-> mv88e6085 configuration: fecmxc_probe() calls dm_fec_bind_mdio() which finds the mdio node in dt and binds the 'fec_mdio' driver to it then probes that driver. When it binds via device_bind_driver_to_node() it should end up calling dsa_post_bind() to bind to the dsa_port driver. Can you verify that you have DM_DSA enabled and that dsa_post_bind() gets called?
You can enable debug for dsa-uclass.c and perhaps that will explain more. In dsa_post_bind() there is a loop over pdata->num_ports which calls device_bind_driver_to_node() to bind the node to DSA_PORT_CHILD_DRV_NAME - I imagine that isn't happening due to one of the checks prior to this failing silently.
Tim