collisions of duplicate ipvlan and macvlan network IDs during createNetwork

[eyz]

PR #2055 resolves the issue shown below, where an existing copy of a network has not been removed on Docker Engine shutdown, and still persists when the config-only network is created once again on a subsequent createNetwork call

existing network config = (*ipvlan.configuration)(0xc420322e80)({\n
  ID: (string) (len=25) \"hwqzu5i4kmrd2uzi34zpnpy1m\",\n
  Mtu: (int) 0,\n
  dbIndex: (uint64) 6,\n
  dbExists: (bool) true,\n
  Internal: (bool) false,\n
  Parent: (string) (len=4) \"eth0\",\n
  IpvlanMode: (string) (len=2) \"l2\",\n
  CreatedSlaveLink: (bool) false,\n
  Ipv4Subnets: ([]*ipvlan.ipv4Subnet) (len=1 cap=4) {\n
    (*ipvlan.ipv4Subnet)(0xc4206413a0)({\n
      SubnetIP: (string) (len=13) \"10.100.0.0/16\",\n
      GwIP: (string) (len=13) \"10.100.0.1/16\"\n
    })\n
  },\n
  Ipv6Subnets: ([]*ipvlan.ipv6Subnet) <nil>\n})\n

new network config = (*ipvlan.configuration)(0xc4203c5b80)({\n
  ID: (string) (len=25) \"hwqzu5i4kmrd2uzi34zpnpy1m\",\n
  Mtu: (int) 0,\n
  dbIndex: (uint64) 0,\n
  dbExists: (bool) false,\n
  Internal: (bool) false,\n
  Parent: (string) (len=4) \"eth0\",\n
  IpvlanMode: (string) (len=2) \"l2\",\n
  CreatedSlaveLink: (bool) false,\n
  Ipv4Subnets: ([]*ipvlan.ipv4Subnet) (len=1 cap=1) {\n
    (*ipvlan.ipv4Subnet)(0xc42149af00)({\n
      SubnetIP: (string) (len=13) \"10.100.0.0/16\",\n
      GwIP: (string) (len=13) \"10.100.0.1/16\"\n
    })\n
  },\n
  Ipv6Subnets: ([]*ipvlan.ipv6Subnet) <nil>\n})\n

The symptom of an existing and now-conflicting network is usually as follows, including showing up as a rejected task's error -

docker.err.log:time="2018-01-15T23:32:21.993891743Z" level=error msg="fatal task error" error="network di-hwqzu5i4kmrd is already using parent interface eth0

[eyz]

Mentioned this to @mavenugo quite a while ago. Not sure if cleanup can be done on the shutdown side, but I think this may be safer, to cleanup if needed on creation instead.

[kensel]

Bumping as it seems to resolve my issue and the issue mentioned moby/moby/issues/35101

[eyz]

I am no longer actively working on PRs for Docker Engine as I've moved my greenfield development away from Docker Swarm to Kubernetes, but I appreciate your support, @kensel. If anyone else finds more information, please do not hesitate to add more detail. Thanks again!
@fcrisciani and @mavenugo, is anyone available to take this over?