Duplicate adress (mac adress) detected !

[pichel]

Hello,
I've configured a Linux box with two network cards. All statically configured, since its a network supervisor machine. The second NIC, eth1, gives me an unsolved problem : pinging any adress using this interface takes often more than 500 ms !
I noticed that in dmesg, I had the title message when activating eth1 : duplicate adress detected

It's not an IP problem (as far as I could diagnostic)
Well if it's MAC related, I tested the following :
Changing the end of the mac adress to random values, which I did sufficent times to be conviced that I should have found at least one which is not used over my network (less than 80 machines including bridges and infrastructure).
I also tried a weird thing : giving a mac adress starting with 99:xxxxxxxx which does not correspond to any vendor. Whe I've done that, I don't have the offending message in dmesg anymore  , but I then cannot use the interface at all : No route to any host 

I tried different NICs with the same result. The only special thing about this comuter is that this is a Compaq desktop computer, wich handles the to PCI expansion slots on a daughterboard itslef slotted on the motherboard...

Well, I stucked with the silly problem without knowing where to look now to solve it. I've discovered this forum where I hope to see someone clever enough on this issue.

Thanks in advance

Michel Roche
(France)

[Ricky]

I guess, PCI expansion is I think treating multiple NIC as single NIC on OS .. so if there are two NICs then MAC address conflicting.
Can you verify if it is conflicting with MAC of first card itself ?.. note down MAC of first card when there is only one nic.. then remove first card and insert second and see what MAC is now .. are they appearing same to OS ?

Seems to be problem with expansion card ..

[pichel]

Hello,
I'm just realizing that my former answer hasn't been published, surely I missed a step or have been disturbed 

Ricky : I understand your back thoughts I think. But ifconfig gives different MAC adresses, the arp table doesn't contain any duplicated, as far as I go in the subnet. And as I mentionned former, I tried to force different arbitrary adresses without succes, except for the one beginning with 99 which gave me half a success.

Another thing that may be worthy. I'm usually connecting remotely to this machine which runs nagios and other things like that, making it my eyes on the subnet. I'm supposed to be connecting via eth0 wich is in a 172.18.x.x subnet, and eth1 is my probe on the 172.19 (and also others).x.x side of the gateway. The 172.19/16 network is connected to the Internet via the gateway which leads in the 172.18/16 network which has the connected modem in it.
From time to time, when playing with eth1 to seek for this problem, I've been disconnected from my ssh session when putting eth1 down, which supposes that paquets where escaping from it, where I expected them going through eth0, as the route table pretended :

Code: [Select]

Destination     Passerelle      Genmask         Indic Metric Ref    Use Iface
172.19.254.0    *               255.255.255.0   U     0      0        0 eth1
172.20.254.0    172.19.254.252  255.255.255.0   UG    0      0        0 eth1
172.18.254.0    *               255.255.255.0   U     0      0        0 eth0
172.17.0.0      172.19.254.252  255.255.0.0     UG    0      0        0 eth1
default         172.18.254.254  0.0.0.0         UG    0      0        0 eth0

I'm now wondering if there's not an IP mess that leads this machine to see its NIC by two different ways...
I should add that everything is connected on the same layer2 switch, with adequate (I hope at last  ) traffic isolation, so it's seen as two switches for the two networks.

Thanks for your support,
Michel

PS: BTW, is there a way to inquire for where the ssh traffic come ? When I trace a route from this pc to the internet, I see a correct one, via 172.18 at once....