ICMPv6: Recognise ND option 14 (Nonce)

[bonktree]

This option has been observed to be included in neighbour solicitations sent by Linux kernel when conf/*/enhanced_dad is enabled on the relevant network interface. It looks like it appeared in the linked kernel commit.

In accordance with RFC 7527, a nonce can help distinguish valid DAD NS messages in the presence of Ethernet loopbacks: after a DAD probe is sent on the network, if an NS is received with the same nonce it is considered looped back and ignored.

We implement the printer for this option and add a simple test with 3 different verbosity levels.

[infrastation]

Thank you for waiting. I have only a cosmetic remark, in that

nonce option (14), length 8 (1): 6069604c0aaa

should be formatted as

nonce option (14), length 8 (1): 60:69:60:4c:0a:aa

to minimize confusion when the value is something such as 112233445566. There is a number of occurrences in the existing code that do exactly that, perhaps it is time to deduplicate that code and to make it a function you could use instead of implementing your own.

[vt-alt]

Perhaps 60:69:60:4c:0a:aa could be confused with a mac address, but this is just a nonce number which does not need to be extremely readable.

[bonktree]

I have only a cosmetic remark, in that

nonce option (14), length 8 (1): 6069604c0aaa

should be formatted as

nonce option (14), length 8 (1): 60:69:60:4c:0a:aa

to minimize confusion when the value is something such as 112233445566. There is a number of occurrences in the existing code that do exactly that, perhaps it is time to deduplicate that code and to make it a function you could use instead of implementing your own.

I like this notation with colons and did consider it for a while, but, as @vt-alt has noted above, the nonce is extremely likely to resemble an Ethernet hardware address, since in practice it will almost definitely be 6 bytes in length (the smallest length possible). So there is a slight chance an inexperienced reader examining the dump might wonder if there is a device with that address.

I'd looked for other established formats for variable-length nonces, and the only other suitable example was in the IKE printer, e. g. in tests/ikev2fourv.out. Other supported protocols use a fixed-size nonce, e. g. a single octet or a 32-bit word, like mptcp and babel; those are printed like 0xaabbccdd.

[bonktree]

There might be other options, though; like this:

nonce option (14), length 8 (1): [60:69:60:4c:0a:aa]

or this:

nonce option (14), length 8 (1): nonce(60:69:60:4c:0a:aa)

[infrastation]

Either that, or the value could use the proper 0x prefix. Bare hex digits is a classic space for misinterpretation.

[vt-alt]

I vote for 0x6069604c0aaa 😄

[fxlb]

My doubt about 0x6069604c0aaa was unfounded.

[bonktree]

or the value could use the proper 0x prefix.

Yes. I'm in favor of 0x6069604c0aaa.

OK, so be it.

[bonktree]

Rebased on latest master with no further changes.