The PHY address bit 2 is configured by the LED pin. Attaching a LED
to this pin is not sufficient to guarantee this configuration pin is
correctly read. This leads to some platforms having their PHY at
address 0 and others at address 4.
If there is no phy-handle specified, the FEC driver will scan the PHY
bus for a PHY and use that. Consequently, adding the DT configuration
of the PHY and the phy properties to the FEC driver broke some boards.
Fix this by removing the phy-handle property, and listing two PHY
entries for both possible PHY addresses, so that the DT configuration
for the PHY can be found by the PHY driver.
Fixes: 86b08bd5b994 ("ARM: dts: imx6-sr-som: add ethernet PHY configuration")
Reported-by: Christoph Mattheis <christoph.mattheis@arcor.de>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Shawn Guo <shawnguo@kernel.org>
 &fec {
        pinctrl-names = "default";
        pinctrl-0 = <&pinctrl_microsom_enet_ar8035>;
-       phy-handle = <&phy>;
        phy-mode = "rgmii-id";
        phy-reset-duration = <2>;
        phy-reset-gpios = <&gpio4 15 GPIO_ACTIVE_LOW>;
                #address-cells = <1>;
                #size-cells = <0>;
 
-               phy: ethernet-phy@0 {
+               /*
+                * The PHY can appear at either address 0 or 4 due to the
+                * configuration (LED) pin not being pulled sufficiently.
+                */
+               ethernet-phy@0 {
                        reg = <0>;
                        qca,clk-out-frequency = <125000000>;
                };
+
+               ethernet-phy@4 {
+                       reg = <4>;
+                       qca,clk-out-frequency = <125000000>;
+               };
        };
 };