LAN Auto-Negotiation fails (LED blinks orange) on Gigabit Switch

[acidicX]

Hi there,
when connected to a gigabit router (in this case a Netgear WNDR 3700 v2), the auto-negotiation fails and the LAN port only blinks orange (indicating wrong speed). Auto-negotiation works fine with the same router and a RBPi (also 10/100 Mbit card).

I can get a connection when I use a crossover cable and connect it to my Linux PC with a static 10 or 100 Mbit/s speed (set via mii-tool on the PC). Auto-negotiation however also fails with the crossover cable (PC has a gigabit ethernet card).

I have not tried it on a 100 Mbit/s switch/router yet.
// edit:
It did not work with the stock image, and still does not work after running apt-get upgrade.

[acidicX]

I've just tried it with the upgraded beta Debian7 NAND image, did an apt-get upgrade, but still the same behaviour.

[ZubairLK]

Strange. It should not do that. I don't have a gigabit switch handy to check myself. But you could force the ethernet speed to 100 perhaps. Or disable autonegotiation and see if that works.

If you have wifi, please download ethtool using

apt-get install ethtool

Then use ethtool to disable autonegotiation. You could also force it to 10/100Mb perhaps.

root@ci20-1277:# ethtool eth0
Settings for eth0:
Supported ports: [ TP MII ]
Supported link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full
Supported pause frame use: No
Supports auto-negotiation: Yes
Advertised link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full
Advertised pause frame use: No
Advertised auto-negotiation: Yes
Link partner advertised link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full
Link partner advertised pause frame use: Symmetric
Link partner advertised auto-negotiation: Yes
Speed: 100Mb/s
Duplex: Full
Port: MII
PHYAD: 0
Transceiver: internal
Auto-negotiation: on
Supports Wake-on: d
Wake-on: d
Current message level: 0x00000004 (4)
link
Link detected: yes
root@ci20-1277:# ethtool -s eth0 autoneg off
root@ci20-1277:# ethtool eth0
Settings for eth0:
Supported ports: [ TP MII ]
Supported link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full
Supported pause frame use: No
Supports auto-negotiation: Yes
Advertised link modes: Not reported
Advertised pause frame use: No
Advertised auto-negotiation: No
Speed: 100Mb/s
Duplex: Full
Port: MII
PHYAD: 0
Transceiver: internal
Auto-negotiation: off
Supports Wake-on: d
Wake-on: d
Current message level: 0x00000004 (4)
link
Link detected: yes
root@ci20-1277:#

Source
http://www.shellhacks.com/en/HowTo-Change-Speed-and-Duplex-of-Ethernet-card-in-Linux

Hope this helps.
ZubairLK

[acidicX]

Hi ZubairLK,

thanks for your answer.
I already tried setting it with mii-tool, which is (or should be) equal to ethtool. So just to make sure I've tried setting it with ethtool, and - voilà - it works on 10 Mbit/s now... at least ;-)

Force to 10 Mbits FD & HD:
ethtool reports link OK, green LED indicating link traffic
mii-tools reports link OK, but no LEDs, no working connection

Force to 100 Mbits FD & HD:
ethtool & mii-tool report no link, LED blinking orange, no connection possible

Autoneg still does not work, obviously.
My project requires the LAN port as we need the WiFi to create an AP, so I really need the LAN port - thanks a bunch for that workaround. It still is a major bug though.

Cheers,
acidicX

[ZubairLK]

Hi AcidicX,

Would it be possible for you to try a different cable/gigabit router.

I just tried a gigabit switch with cat 5,5e and 6 cables. They all worked. Autonegotiated and everything..

ZubairLK

[acidicX]

Right, cables are not the issue (tried that already) - but with a GbE switch, it worked, autonegotiated without problem, even with other 1000M-cards on the switch. Problem seems to be router + ci20. The funny part is - the switch is now connected to my router and autoneg still works.

However, the router works flawlessly with a direct connection to a RBPi Model B, autoneg to 100M. I'll try a B+ soon enough, but I guess this will also work. Therefore I'm pretty sure it's not the router that has the issue...

[ZubairLK]

tracing such corner issues can be a really wild goose chase.. :)

Glag you have a working setup now.

I'll close the issue..

[acidicX]

well... I wouldn't call it 'working setup'. I have to put a (otherwise useless) switch between the ci20 and the router... I would appreciate a follow up. It's also probably not the only GbE router that will have the issue with the ci20.

[quantumaardvark]

I don't know if you still have this problem. I had a severe case of it a couple of years ago in an embedded development. 10Mb Ethernet is essentially a 'differential digital' connection, i.e Manchester encoded 1's and 0's. 100Mb and upwards are 'more analog' in their nature, and depend on a PLL to use the 4B/5B encoding. Static or similar disruption to the PHY interface can destroy the analog interface but leave the digital (i.e. 10Mb) interface working. Auto-negotiation takes place at 10Mb which is more-or-less always guaranteed to work. As far as the PHY is concerned, it believes it has the capability to run at 100Mb. So during auto-negotiation it will advertise it has that capability to do so. At the end of auto-negotiation it will switch to 100Mb and then become 'deaf'. Forcing 10Mb ensures the working interface i.e. the 10Mb one is the only one that is used.The only fix for this is replacing the Ethernet chip or the C120! We debugged this (i.e. we worked out what was happening) by using a debugger in the Ethernet driver to read back the PHY registers before, during and after auto-negotiation. As you seem to have 100Mb working but not 1Gb, it would seem likely a similar problem is in play. 1G uses all four pairs, 100Mb only uses two pairs. It seems probable you have either 'lost' the PHY receiver on one of the other two pairs, or a transformer winding or miid-point bias capacitor have been lost, resulting in wildly incorrect common-mode voltages.

[acidicX]

@quantumaardvark I don't think this is the case. If it would be a faulty connection, it would not work on any GbE switches, right?
I only have the problem with some switches (e.g. the one on my router).

[quantumaardvark]

It would only be a faulty connection it it failed to work on ALL GbE switches. If the other switches you have work with auto-negotiation but are only 10/100Mb, then the negotiation will work with those. If on the other hand you have it working with other routers at 1Gb then it is more likely to be a marginal tolerance issue due to PHY damage.

Your comments on the failure mode still indicate it is PHY damage on the CI20. The switch must have a better tolerance and/or be better balanced than the router. Seeing as all of these board-level products are susceptible to static damage though handling, I'd still favour the explanation of the PHY being 'walking wounded' in this respect. Your only solution is to fix the data rate advertised in the negotiation as suggested by ZubairLK. FWIW, I haven't had any problems at 1GbE... yet!

[acidicX]

OK... well, as I have no other ci20 board available and therefore can't check it by method of elimination, I'll mail to support and see what they make of it.
Thanks for your help!