Slow network outbound throughput with tso enabled in veth pair with newer kernels

[Oliver Böttcher]

Hi all,

first of, this isn't really a bug report but a request for a workaround.

With (obviously) kernels below 3.8 and OpenVSwitch < 2.0.0 the tcp segmentaion offload was off by default:

$ ethtool -k qvb056d020b-d6
Offload parameters for qvb056d020b-d6:
rx-checksumming: off
tx-checksumming: off
scatter-gather: off
tcp-segmentation-offload: off
udp-fragmentation-offload: off
generic-segmentation-offload: off
generic-receive-offload: on
large-receive-offload: off
rx-vlan-offload: off
tx-vlan-offload: off
ntuple-filters: off
receive-hashing: off

With newer kernels (3.16.3-200.fc20.x86_64 in our case) tso is on by default which leads to slow outbound speed:
Features for qvb5979dad2-b3:
rx-checksumming: on
tx-checksumming: on
 tx-checksum-ipv4: off [fixed]
 tx-checksum-ip-generic: on
 tx-checksum-ipv6: off [fixed]
 tx-checksum-fcoe-crc: off [fixed]
 tx-checksum-sctp: off [fixed]
scatter-gather: on
 tx-scatter-gather: on
 tx-scatter-gather-fraglist: on
tcp-segmentation-offload: on
 tx-tcp-segmentation: on
 tx-tcp-ecn-segmentation: on
 tx-tcp6-segmentation: on
udp-fragmentation-offload: on
generic-segmentation-offload: on
generic-receive-offload: on
large-receive-offload: off [fixed]
rx-vlan-offload: on
tx-vlan-offload: on
ntuple-filters: off [fixed]
receive-hashing: off [fixed]
highdma: on
rx-vlan-filter: off [fixed]
vlan-challenged: off [fixed]
tx-lockless: on [fixed]
netns-local: off [fixed]
tx-gso-robust: off [fixed]
tx-fcoe-segmentation: off [fixed]
tx-gre-segmentation: on
tx-ipip-segmentation: on
tx-sit-segmentation: on
tx-udp_tnl-segmentation: on
tx-mpls-segmentation: off [fixed]
fcoe-mtu: off [fixed]
tx-nocache-copy: off
loopback: off [fixed]
rx-fcs: off [fixed]
rx-all: off [fixed]
tx-vlan-stag-hw-insert: on
rx-vlan-stag-hw-parse: on
rx-vlan-stag-filter: off [fixed]
l2-fwd-offload: off [fixed]
busy-poll: off [fixed]

Another workaround would be to disable tso in every VM which is using the virtio net driver which is often impossible since VMs are managed by customers.

After creation of the veth pair a ethtool -K qvbXXX tso off should be triggered.

Relevant used versions:
kernel-3.16.6-200.fc20.x86_64
openvswitch-2.3.0-1.fc20.x86_64
qemu-system-x86-1.6.2-8.fc20.x86_64

[Kyle Mestery (mestery)]

Is this still a relevant bug? It hasn't seen any updates since November, 2014.

[Kyle Mestery (mestery)]

Ramping down priority given this hasn't seen updates for almost a year.

[Ihar Hrachyshka (ihar-hrachyshka)]

Similar issue was raised for kernel at: http://comments.gmane.org/gmane.linux.network/358321

But the thread does not suggest that there is consensus that there is anything to fix and it's not a local configuration issue.

[Launchpad Janitor (janitor)]

[Expired for neutron because there has been no activity for 60 days.]