Fixes #13211 linkLocalIPv6FromMac

[MalteJ]

Kudos to @gissehel for his great work on investigation and the fabulous bug report!

fixes #13211

[thaJeztah]

thanks @MalteJ!

ping @mrjana to check if Libnetwork doesn't have the same bug.

[mrjana]

In libnetwork we are deprecating this as inspect value. The eth0 interface inside the container will have this linklocal ipv6 address assigned to the interface itself. So if you do ip -6 addr show dev eth0 you should see the link local ipv6 address.

[jessfraz]

so I think this needs to be opened on libnetwork now? then pulled in here am i correct

[thaJeztah]

@jfrazelle according to @mrjana's #13226 (comment) this shouldn't be a problem in Libnetwork. But I'm interested to hear if there's a regression test in place

[jessfraz]

cool @MalteJ can you rebase please :)

[MalteJ]

@jfrazelle well, it looks like the files I have edited disappeared from master branch 😄

[thaJeztah]

@MalteJ is the test case still useful for libnetwork?

[mavenugo]

@mrjana can you suggest the course of action for this PR ?

[mrjana]

@MalteJ @thaJeztah @jfrazelle @mavenugo This test is moot because in the libnetwork bridge driver we don't generate the linklocal address from MAC at all. We leave it to the kernel to do it for us because of the way we plumb the interfaces. So this(the generation part) doesn't need to be tested. What may be tested is add an integration-cli to ensure that this is the case when container is brought up and the the container eth0 address has the linklocalFromMac setup correctly.

But, we should probably remove the linklocalFromMAC field from NetworkSettings because it's not populated any more.

[icecrime]

Sorry @MalteJ, another one that we seem to have dropped with the big move to libnetwork =/ I'm closing this as my understanding from the above comments is that it's not relevant anymore.

[MalteJ]

no problem. I am happy we have the new network stack!

[icecrime]

❤️