autogenerated LinkLocalIPv6Address can create two containers with the same ipv6 address

[gissehel]

Description of problem:

When I create containers without specifying any ip or mac address, ipv4, mac address and ipv6 are autogenerated (at least since version 1.5 for ipv6).

It looks like ip v4 and mac address can't conflicts, but you can generate two containers with the same ip v6:

docker version:

Client version: 1.6.1
Client API version: 1.18
Go version (client): go1.4.2
Git commit (client): 97cd073
OS/Arch (client): linux/amd64
Server version: 1.6.1
Server API version: 1.18
Go version (server): go1.4.2
Git commit (server): 97cd073
OS/Arch (server): linux/amd64

docker info:

Containers: 11
Images: 185
Storage Driver: aufs
Root Dir: /var/lib/docker/aufs
Backing Filesystem: extfs
Dirs: 207
Dirperm1 Supported: false
Execution Driver: native-0.2
Kernel Version: 3.13.0-39-generic
Operating System: Ubuntu 14.04.1 LTS
CPUs: 4
Total Memory: 15.59 GiB
Name: the server's name
ID: KU57:P757:EDQZ:P4Q2:6SF2:MLXX:WY2K:TW62:CX7C:EVHS:DORS:K6ZH
WARNING: No swap limit support

uname -a:

Linux the server's name 3.13.0-39-generic #66-Ubuntu SMP Tue Oct 28 13:30:27 UTC 2014 x86_64 x86_64 x86_64 GNU/Linux

Environment details (AWS, VirtualBox, physical, etc.):

physical server

How reproducible:

Always

Steps to Reproduce:

• start docker daemon (because it will restart the auto assignation of ips to 172.17.0.1)
• create 17 containers
• create a container "test1"
• create 238 other useless containers
• create a container "test2"

Expected results based on how ipv6 affectation seems to work in spirit:

name	IPv4	mac	IPv6
test1	172.17.0.18	02:42:ac:11:00:12	fe80::42:acff:fe11:12
test2	172.17.1.2	02:42:ac:11:01:02	fe80::42:acff:fe11:102

Actual results:

name	IPv4	mac	IPv6
test1	172.17.0.18	02:42:ac:11:00:12	fe80::42:acff:fe11:12
test2	172.17.1.2	02:42:ac:11:01:02	fe80::42:acff:fe11:12

Additional info:

It seems to me that the problem lies in #8947 in the file daemon/networkdriver/bridge/driver.go with the line:

return fmt.Sprintf("fe80::%x%x:%xff:fe%x:%x%x/64", hw[0], hw[1], hw[2], hw[3], hw[4], hw[5]), nil

which should look like (in spirit, I don't do go):

return fmt.Sprintf("fe80::%x:%x:%x:%x/64", hw[0]*0x100+hw[1], hw[2]*0x100+0xff, 0xfe*0x100+hw[3], hw[4]*0x100+hw[5]), nil

As I never wrote a line of go, I won't create a PR: I prefer to have someone who know what he's doing to do it.

[thaJeztah]

Thanks for reporting this, and researching! As you might know, networking will get a major overhaul in Docker 1.7 with the integration of libnetwork. However, we're not there yet, so I'll include people from both "old" and "new" (libnetwork) here.

ping @MalteJ @mrjana @LK4D4

[MalteJ]

@gissehel thank you very much for your great work!
With your analysis I could write a fix really fast. I have opened a Pull Request.

BTW: Your code was quite good - it just has raised a type mismatch error. The hw array holds bytes (8 Bit) that cannot be multiplied with 0x100 (> 8 Bits), so I have had to cast the hw entities to int. Apart from that your code proposal is working. You really should write some Go code and contribute to Docker 😉

Thanks!

[aboch]

As explained by @mrjana in #13226 (comment), link local IPv6 generation is now left to linux (libnetwork will just construct the MAC from the IPv4) and this issue no longer exists in current docker (at least since 1.7).

See o/p from latest docker:

$ ./docker run --rm --net nw2 --ip 172.18.1.2 busybox ifconfig
eth0      Link encap:Ethernet  HWaddr 02:42:AC:12:01:02  
          inet addr:172.18.1.2  Bcast:0.0.0.0  Mask:255.255.0.0
          inet6 addr: fe80::42:acff:fe12:102/64 Scope:Link    <===
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:2 errors:0 dropped:0 overruns:0 frame:0
          TX packets:2 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:0 
          RX bytes:180 (180.0 B)  TX bytes:168 (168.0 B)

Therefore I am suggesting to close this issue.
ping @thaJeztah

[thaJeztah]

Thanks @aboch, I'll close this issue, but @gissehel feel free to comment after I've closed 👍