I have a beaglebone black running debian, which sometimes boots with a dead NIC.

When it is not working, the lights of the NIC stay on or off permanenty, indepent of actual connection state. No packets are received or any communication is possible. The NIC drivers are loaded and dmesg shows no abnormalities. I can compare all of that output with a working board.

I noticed that when the NIC is not working though, the output of dmesg and ethtool are showing a different 'physical address' (not the MAC).

Working:
dmesg: [ 1.473566] davinci_mdio 4a101000.mdio: phy[0]: device 4a101000.mdio:00, driver SMSC LAN8710/LAN8720
ethtool: Settings for eth0:
...
PHYAD: 0

Not working:
dmesg: [ 1.473676] davinci_mdio 4a101000.mdio: phy[2]: device 4a101000.mdio:02, driver SMSC LAN8710/LAN8720
ethtool: Settings for eth0:
...
PHYAD: 2

My question is, can I set/force this phyad to 0, to solve this problem?
I tried it using ethtool, but it refuses:
root@beaglebone:# ethtool -s eth0 phyad 0
Cannot set new settings: Invalid argument
not setting phy_address

root@beaglebone:# ethtool -s eth0 phyad 2
<this command completes>

Can somebody at least tell me, what that sub address is? Is it in index in a list of found devices, an address on a bus, or something else?

This subaddress, as you call it? 4a101000.mdio:00, that's the phy. It's enumerated somewhere in /sys. Here's the datasheet for the device: http://www.sunnyqi.com/upLoad/produc...06/LAN8710.pdf