IPv6 Flow Label Corruption

[mmisamor]

Hi,

I'm using Cisco Trex to generate IPv6 traffic for a different set of flow labels to test my lab setup. There are no issues if I send normal IPv6 packets, i.e. without setting any flow label value on the script. The problem comes when I set a value for this flow label field in the IPv6 header which, after a long research and debugging, we discover that Cisco Trex induces a corruption in the flow label value [take a look at the packet with Wireshark].

At this point, it can be mentioned that the lab setup is working properly because by using iperf3 we manage to test this flow label feature but we would like to generate traffic with Cisco Trex and DPDK.

This is the script that I'm currently using for generating the IPv6 packets for a defined flow label value. The weird thing here is that, if I use a low range number for the flow label, for example, flow label = 10or flow label = 20 everything works but once I set a high range number like flow label = 88000 it is corrupted somehow.

from trex_stl_lib.api import *
import argparse

class STLS1(object):

    def create_stream (self):
        pkt_size = 1500
        base_pkt = Ether()/Dot1Q(vlan=50)/IPv6(src="fc01:16::1",dst="fc01:48::1",fl=88000)/UDP(dport=12,sport=1025)
        payload_size = pkt_size-len(base_pkt)
        my_pkt = base_pkt/('5'*(payload_size))
        return STLStream(packet = STLPktBuilder(pkt = my_pkt), mode = STLTXCont())

    def get_streams (self, tunables, **kwargs):
        parser = argparse.ArgumentParser(description='Argparser for {}'.format(os.path.basename(__file__)), formatter_class=argparse.ArgumentDefaultsHelpFormatter)
        args = parser.parse_args(tunables)
        return [ self.create_stream() ]

def register():
    return STLS1()

This is the Cisco Trex IPv4 address port configuration:

root@srv:/opt/trex/v2.92# cat /etc/trex_cfg.yaml
### Config file generated by dpdk_setup_ports.py ###

- port_limit: 2
  version: 2
  interfaces: ['02:00.0', '05:00.1']
  stack: linux\_based
  port_info:
      - ip: 4.4.4.1
        default_gw: 4.4.4.2
        vlan: 50
      - ip: 3.3.3.1
        default_gw: 3.3.3.2
        vlan: 60

  platform:
      master_thread_id: 0
      latency_thread_id: 8
      dual_if:
        - socket: 0
          threads: [1,2,3,4,5,6,7,16,17,18,19,20,21,22,23]

Now, I launch Cisco Trex server:

root@srv:/opt/trex/v2.92\# ./t-rex-64 -i --stl

And Cisco Trex client:

root@srv:/opt/trex/v2.92\# ./trex-console

After that, I need to configure the IPv6 address of Cisco Trex ports under service mode on the client side:

trex> start -f stl/my_ipv6_test.py -m 1mbps -p 0
trex> service
trex(service)> ipv6 -p 0 -s fc01:50::5
trex(service)> ipv6 -p 1 -s fc01:60::5
trex(service)> vlan -p 0 --vlan 50
trex(service)> vlan -p 1 --vlan 60

And, at this point, I ping to the switch in order to verify that I have connection and the destination is reachable:

trex(service)> ping -d fc01:50::3 -p 0
Pinging fc01:50::3 from port 0 over VLAN '50' with 64 bytes of data:
Reply from fc01:50::3: bytes=68, time=0.87ms, hlim=255
Reply from fc01:50::3: bytes=68, time=0.87ms, hlim=255
Reply from fc01:50::3: bytes=68, time=0.87ms, hlim=255

trex(service)> ping -d fc01:60::3 -p 1
Pinging fc01:60::3 from port 1 over VLAN '60' with 64 bytes of data:
Reply from fc01:60::3: bytes=68, time=1.26ms, hlim=255
Reply from fc01:60::3: bytes=68, time=1.26ms, hlim=255
Reply from fc01:60::3: bytes=68, time=1.26ms, hlim=255

trex(service)> service --off

Does anyone have any idea about what is going on?

Thanks for your time, Best regards, Carmen Misa