(CC'ing Johannes) On Sun, Mar 17, 2024 at 10:55:16PM +0100, Andrew Lunn wrote: > On Sun, Mar 17, 2024 at 08:57:22PM +0200, Laurent Pinchart wrote: > > On Sun, Mar 17, 2024 at 06:17:17PM +0100, Andrew Lunn wrote: > > > > +&eqos { > > > > + pinctrl-names = "default"; > > > > + pinctrl-0 = <&pinctrl_eqos>; > > > > + phy-mode = "rgmii-id"; > > > > + phy-handle = <ðphy0>; > > > > + > > > > + mdio { > > > > + compatible = "snps,dwmac-mdio"; > > > > + #address-cells = <1>; > > > > + #size-cells = <0>; > > > > + > > > > + /* Atheros AR8033 on v1.0, Realtek RTL8211E on v1.1 */ > > > > + ethphy0: ethernet-phy@0 { > > > > + compatible = "ethernet-phy-ieee802.3-c22"; > > > > + reg = <0>; > > > > + eee-broken-1000t; > > > > + }; > > > > > > Hi Laurent > > > > > > Do you happen to know what is broken with respect to EEE? It seems > > > like a lot of IMX boards have this, so i suspect it is the MAC. Maybe > > > we should be keying off the MAC compatible and disabling this in the > > > ethernet driver rather than have every .dts file needing it? > > > > I wonder if this could be cargo-cult. To be honest, I've copied it from > > the BSP and haven't investigated it. I've tried dropping that and > > haven't noticed any difference, but I'm not sure how I should test it > > properly. > > Maybe a better approach is to find the errata. It could be some older > version of the eqos was broken, and it got fixed along the way? If > that is so, moving it into the driver would be better, assuming there > is some sort of hardware version register in the eqos. I don't know if there are public errata about this issue. It is beyong my areas of expertise. I've found a relatively recent e-mail on the netdev mailing list that seems related ([1]), but there was no reply. [1] https://lore.kernel.org/netdev/9c1c9408-88ac-4ade-b8ec-2ae5d8922cac@xxxxxxxxxxxxxx/ -- Regards, Laurent Pinchart