Xen users - Feb 2010 - Re: pv_ops kernel and network problems (checksumoffloading?)

Hi,

I found this mail about network problems in the archive.
I think I have the same network problems with a routing (with or without
NAT) domU connected by two network bridges.
The dom0 logs thousands of following messages:
--
Attempting to checksum a non-TCP/UDP packet, dropping a protocol 1 packet
--
and all outgoing TCP packets have wrong TCP checksums (traced on another
machine on the external network)

My setup:

Private Network ->
dom0: peth0 -> eth0(bridge) -> vif1.0 ->
domU: eth0 -> routing (with or without NAT) -> eth1 ->
dom0: vif1.1 -> eth1(bridge) -> peth1 ->
Public Network

Packets coming in from peth0 are either dropped (ICMP) by dom0 after routing
through domU (see error above) or for TCP, they have wrong checksums.
The checksums traced with tcpdump on dom0 eth1 are correct, but the
checksums of the same packets traced on another machine on the public
network are wrong. The packets are dropped therefore by the receiver.

dom0 and domU are Fedora 12.
Kernels used (all pv_ops)
dom0: kernel-PAE-2.6.31.9-1.2.82.xendom0.fc12.i686.rpm (from
http://fedorapeople.org/~myoung/dom0/i386/)
domU: kernel-PAE-2.6.31.12-174.2.3.fc12.i686.rpm
xen: xen-hypervisor-3.4.2-1.fc12.i686.rpm

Hardware:
Pentium 4 without any virtualization support
peth0: on board PCI-E Marvell Yukon 88E8053 (driver sky2)
peth1: PCI Intel Pro/1000GT (driver e1000)

I can reproduce this 100%. Turning off tx chksumming (eththool -K {dev} tx
off) on all devices in dom0 and domU does not help.

I would be happy if someone can give me a hint, if there is any fix
available, either by configuration settings or replacing one of the kernels?

Regards,
Bernhard


_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users