Address already in use for IPv6 Local-Link

[nerdalert]

Issue Description

• When assigning an IPv6 Link-Local address to a container it uncovered an issue. When specifying a Link-Local --ipv6=, the allocation from the default IPAM driver works fine the first time, but when you exit the endpoint the ipv6 release for the endpoint does not occur. When you go to start the same container config again it will throw an error with docker: Error response from daemon: Address already in use.
• While assigning a Link-Local is questionable from a practical standpoint, I think it makes sense to either A) block it from creation or B) Don't perform v6 categorical type checks outside of is it 64-bit etc.
• On the error returned it might be worth considering adding more information to address in use errors with the ability to set an IP addr --ip=x.x.x.x as of v1.10. In the err msg of Address already in use. specifying the address that is already in use could make it a bit easier to track down when it comes to dual stack v4/v6 and multi-subnet v4 and/or v6 e.g. Address %s already in use.
• The error does not occur with the -d bridge driver since it uses a bit different of v6 support to support backwards compatable bridge use cases.

Build Details

• Ver: Docker version 1.11.0-dev, build 921c744-unsupported, experimental
• Branch: PR Added macvlan and ipvlan drivers #964

Steps to reproduce - Create a Dual Stack v4/v6 Network (Also the same results with v6 only etc.)

• Create a network using the `-d ipvlan' or '-d macvlan' drivers

docker network create -d macvlan  \
  --subnet=192.168.123.0/24 \
  --gateway=192.168.123.1 \
  --subnet=fe96::/64 \
  --gateway=fe96::1 \
  -o host_iface=eth0 macnet

• Daemon debug logs

DEBU[0705] form data: {"CheckDuplicate":true,"Driver":"macvlan","IPAM":{"Config":[{"Gateway":"192.168.123.1","Subnet":"192.168.123.0/24"},{"Gateway":"fe96::1","Subnet":"fe96::/64"}],"Driver":"default","Options":{}},"Internal":false,"Name":"macnet","Options":{"host_iface":"eth0"}}
DEBU[0705] Allocating IPv4 pools for network macnet (fcc6b19e01ee0b2e1a24c1da76d8b2ee7c97ed83e01ec9903b2d48d6090d67ec)
DEBU[0705] RequestPool(LocalDefault, 192.168.123.0/24, , map[], false)
DEBU[0705] RequestAddress(LocalDefault/192.168.123.0/24, 192.168.123.1, map[RequestAddressType:com.docker.network.gateway])
DEBU[0705] Allocating IPv6 pools for network macnet (fcc6b19e01ee0b2e1a24c1da76d8b2ee7c97ed83e01ec9903b2d48d6090d67ec)
DEBU[0705] RequestPool(LocalDefault, fe96::/64, , map[], true)
DEBU[0705] RequestAddress(LocalDefault/fe96::/64, fe96::1, map[RequestAddressType:com.docker.network.gateway])

Start container # 1 and then exit

• Specify an explicit --ip6 address:

docker run -it --net=macnet --ip6=fe96::10 --rm alpine /bin/sh

• Daemon debug logs

DEBU[0803] RequestAddress(LocalDefault/192.168.123.0/24, <nil>, map[])
DEBU[0803] RequestAddress(LocalDefault/fe96::/64, fe96::10, map[])
DEBU[0803] Assigning addresses for endpoint clever_morse's interface on network macnet
DEBU[0803] Macvlan Endpoint Joined with IPv4_Addr: 192.168.123.2, Gateway: 192.168.123.1, MacVlan_Mode: bridge, Host_Iface: eth0
DEBU[0803] Macvlan Endpoint Joined with IPv6_Addr: fe96::10 Gateway: fe96::1 MacVlan_Mode: bridge, Host_Iface: eth0

• Logs from exiting the container with v4 only release, v6 doesn't release:

DEBU[0827] ReleaseAddress(LocalDefault/192.168.123.0/24, 192.168.123.2)

Recreate container # 1

• Container output:

docker run -it --net=macnet --ip6=fe96::10 --rm alpine /bin/sh
docker: Error response from daemon: Address already in use.

• Daemon Debug logs:

DEBU[0888] Assigning addresses for endpoint boring_hypatia's interface on network macnet
DEBU[0888] RequestAddress(LocalDefault/192.168.123.0/24, <nil>, map[])
DEBU[0888] RequestAddress(LocalDefault/fe96::/64, fe96::10, map[])
WARN[0888] failed to cleanup ipc mounts:
...
(trunc)
...
ERRO[0888] Handler for POST /v1.22/containers/51e020f96de4ab2ee7b9a995577f0550a81fbda36eb659306a6d0c0f166e9b45/start returned error: Address already in use

[GordonTheTurtle]

@nerdalert It has been detected that this issue has not received any activity in over 6 months. Can you please let us know if it is still relevant:

• For a bug: do you still experience the issue with the latest version?
• For a feature request: was your request appropriately answered in a later version?

Thank you!
This issue will be automatically closed in 1 week unless it is commented on.
For more information please refer to #1926