Hello,
I have installed SL6.0 on a system, which has two NICs connected to independent switches.
Host network setup:
|-> bond0: eth0 and eth1 in active-backup bond0 (mode=1)
|-> bond0.100: tagged vlan attached to bridge br100
|-> br100: bridge with static IP
This setup produces messages in logs like this: bond0.100: received packet with own address as source address.
Running tcpdump (on br100 and bond0.100) shows that this is related to ARP requests, which i see twice - untagged and vlan tagged:
06:06:16.376513 00:22:bb:66:b1:23 > ff:ff:ff:ff:ff:ff, ethertype ARP (0x0806), length 42: Ethernet (len 6), IPv4 (len 4), Request who-has 172.16.0.11 tell 172.16.0.19, length 28
06:06:16.376566 00:22:bb:66:b1:23 > ff:ff:ff:ff:ff:ff, ethertype 802.1Q (0x8100), length 64: vlan 100, p 0, ethertype ARP, Ethernet (len 6), IPv4 (len 4), Request who-has 172.16.0.11 tell 172.16.0.19, length 46
Then I start a KVM virtual machine with an interface attached to bridge br100.
The KVM guest cannot be randomly pinged (i.e. 16% packet loss).
brctl showmacs br100 sometimes shows the mac address of the kvm guest in the same port 1 as the bond0.100 iface.
However, when I put one interface of the bond down (ifdown eth1), the ping gets replies like it should and the messages disappear from logs. Also the vlan tagged ARPs are no longer seen on br100/bond0.100.
We also use this network configuration on CentoOS 5.5 systems, where it's working as expected (although there are other problems).
There is a bug report describing quite similar problem, but it seems to be ignored by Red Hat:
https://partner-bugzilla.redhat.com/show_bug.cgi?id=623199
This problem was already discussed on other lists, but I could not spot a satisfying solution there:
https://lists.linux-foundation.org/pipermail/bridge/2009-December/006863.html
http://www.spinics.net/lists/linux-net/msg17574.html
I would appreciate any suggestions on how to resolve this problem.
Thanks,
dano
|