bpf: A veth perf test
A test that test the veth performance. To avoid the ip routing
overhead, bpf_redirect_peer (phy-eth0 => veth) and
bpf_redirect_neigh (veth => phy-eth0) is used. In particular,
veth backlog queue performance is the interest here.
eth0 (phy-nic) < -- > veth0 (host) < -- > veth_eth0 (netns)
host:
VETH_ETH0_ADDR6="ip6-addr-at-veth_eth0(netns)" VETH0_ADDR="ip6-addr-at-veth0(host)" ./test_progs -v -t veth_perf
... ...
Press any key to cleanup:
netns:
ip netns exec ns_test bash # enter netns
./tcp_rr -6 -P 12345 -Q 64 -R 128 -l 30 -T 36 -U -F 1024 #run test server
client:
./tcp_rr -c -H 2803:6080:c124:5ab::3 -P 12345 -Q 64 -R 128 -l 30 -T 36 -U -F 256
Signed-off-by: Martin KaFai Lau <martin.lau@kernel.org>
2 files changed