fix(iroh): Accurately set RecvMeta::dst_ip, remove normalized_local_addr

[matheus23]

Description

Now that we have multipath and don't rely on quinn only seeing "EndpointMappedAddr"s, but instead have it see the actual addresses (minus the one MixedMappedAddr), we can actually address #2632 very easily. We just pass it through from how we get it from the socket/OS.

We don't have to worry about MixedMappedAddrs, because we only use them to send, they never end up at dst_ip in our transmits, because they get translated to actual IP adresses before hitting the network, and then get received as actual dst_ips on the other side.

Notes & open questions

Closes #2632

I think we can also close #2491 then? AFAICT, we removed the is_unicast_link_local IP check with QNT, so we actually send over link-local addresses now. This could mean that we fully support link-local IPs with this PR now 🎉

These changes make sense to me, but I didn't test them super thoroughly. FWIW, all tests that fail in this PR already fail before the changes for me locally.

Change checklist

• Self-review.

[github-actions]

Documentation for this PR has been generated and is available at: https://n0-computer.github.io/iroh/pr/3770/docs/iroh/

Last updated: 2025-12-12T17:48:58Z

[flub]

🎉

[flub]

I think we can also close #2491 then? AFAICT, we removed the is_unicast_link_local IP check with QNT, so we actually send over link-local addresses now. This could mean that we fully support link-local IPs with this PR now 🎉

Do we not filter out the link-local addresses from the local interfaces? And we also filter out localhost somewhere unless that's the only one found I think. But maybe that's fine because the OS will do hairpinning on the LAN interface address anyway.

[dignifiedquire]

I love deleting code

[matheus23]

Do we not filter out the link-local addresses from the local interfaces? And we also filter out localhost somewhere unless that's the only one found I think. But maybe that's fine because the OS will do hairpinning on the LAN interface address anyway.

I can find the code that used to do that, but I can't find it in the current codebase. We directly use all addresses from our local interface.

[dignifiedquire]

@matheus23 see n0-computer/net-tools#50 for the important code