[ovs-discuss] Linux kernel panics when attaching eth cable

Paolo Cravero pcravero at as2594.net
Fri Jan 29 14:55:21 UTC 2010


Good morning/evening,
I tried both with 0.90.x and 0.99.1: as soon as I attach the network cable to 
a physical interface managed by openvswitch module, the system goes into 
kernel panic.

Configuration in /etc/ovs-vswitchd.conf is:

bridge.bridge0.port=peth0
bridge.bridge0.port=eth1
bridge.bridge0.port=oveth0
bridge.bridge0.port=oveth1
iface.oveth0.internal=true
iface.oveth1.internal=true

OS is CentOS 5.3 (yum updated today):
Linux 2.6.18-164.11.1.el5xen #1 SMP Wed Jan 20 08:06:04 EST 2010 x86_64 x86_64 
x86_64 GNU/Linux

Note: kernel panics occur with the non-xen kernel too.

Physical interface peth0 is a Marvell Technology Group Ltd. 88E8053 PCI-E 
Gigabit Ethernet Controller (rev 20)

I remove the bridge module. Load OVS:

/sbin/insmod /root/setup/a/datapath/linux-2.6/openvswitch_mod.ko
ovs-vswitchd /etc/ovs-vswitchd.conf

As long as the cable is unplugged I can do all sort of configurations as of 
documentation. When I attach the network cable ... bump!

Upstream switch is a Cisco device with parameters I cannot control.

Am I doing something wrong? (quite possible since I'm new to linux kernel 
switches) Is it a hardware problem? Else?

Thank you in advance for any support you'll be able to give.
Paolo Cravero






More information about the discuss mailing list