Ipv6; BGP peering

[uablrek]

BGP peering in a ipv6-only cluster

This PR takes ipv6 support as far as bgp peering for nodes inside the cluster;

# gobgp neighbor
Peer                AS  Up/Down State       |#Received  Accepted
1000::1:c0a8:101 64512 00:00:37 Establ      |        0         0
1000::1:c0a8:102 64512 00:00:37 Establ      |        0         0
1000::1:c0a8:103 64512 00:00:40 Establ      |        0         0

NOTE; with this PR ipv6 code is actually used if the nodeIP is recognised as an ipv6 address. Before ipv4 operation was hard-coded.

The CIDR handling does not work yet so the veth devices does not get any (ipv6) addresses. The kube-bridge device is howerver assigned an ipv6 address;

# ip -6 addr show dev kube-bridge
11: kube-bridge: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 state UP qlen 1000
    inet6 1000::2:b00:301/120 scope global 
       valid_lft forever preferred_lft forever
    inet6 fe80::6409:c0ff:fe65:9419/64 scope link 
       valid_lft forever preferred_lft forever

Design notes

The iptablesCmdHandler is now created with protocol;

iptables.NewWithProtocol(iptables.ProtocolIPv6)

A function for creating the right iptablesCmdHandler is provided to avoid code duplication. The PodEgressRules funtions are now methods in NetworkRoutingController to be able to access ipv6 stuff.

The ipset 31-caracter name limitation

This is a problem. The Refresh function added "-temp" to the name which together with the "inet6:" prefix made the temp names too long. As a temporary solution just a "-" is added.
IMO some hash string like "tmp_d2d362cdc6579390f1c0617d7" shall be used for temporary ipset's to make the temp name independent from the original set name.

Please comment on this.

[uablrek]

Took a closer look and the CIDR allocation does work.

# jq . < /etc/cni/net.d/10-kuberouter.conf  | cat
{
  "bridge": "kube-bridge",
  "ipam": {
    "subnet": "1000::2:b00:100/120",
    "type": "host-local"
  },
  "isDefaultGateway": true,
  "isGateway": true,
  "name": "ekvm",
  "type": "bridge"
}

It is routing to the node cids that fails. Calls to a service that happens to be local works.

[murali-reddy]

LGTM.

Verified with https://github.com/Nordix/xcluster/blob/master/ovl/kube-router-ipv6/README.md

thanks @uablrek for the patch and helping out with testing