libp2p / libp2p/go-libp2p

Unstable NAT traversal from a natmanger port collision

Open
#2,059 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

kind/bug need/analysis
Dominant language
Go
Stars
6.9k
Forks
1.3k
Avg merge
13d 21h
Merged PRs (30d)
1

Description

The natmanger does not check to see if a port is already in use by another peer on the network before trying to clobber it with a new route:
https://github.com/libp2p/go-libp2p/blob/master/p2p/host/basic/natmgr.go#L129-L134

The result is that if a number of go-libp2p clients all try to use QUIC UDP in a large NAT like a collage dorm room, then no one will be able to do so in a stable way. This will cause problems for issue #1785.

Worse yet - you can't know how a given router is going to respond to aggressive multi-pronged TCP/UDP hole punching for the same port. Dozens of client is an extreme, but even two or more clients may case problems. Even one attempt at an existing route could disrupt network traffic for a buggy or cheap router, the client needs to do its best to avoid existing routes.

If the network interface is reporting a non-public IP address (RFC1918) then we can assume we are behind a NAT. We should see if there is an existing route before trying to setup any new routes. Which is not trivial which is why its not being currently done. It is better to randomly select a UDP port from a large range as to reduce the likelihood of collision as much as possible. If there is a UDP or TCP holepunch failure, the client should try selecting another random port above 1024 and try again, do so on three random ports before giving up.

Now if two libp2p nodes are on the same local network - then they might have found each-other using mDNS. If mDNS is reporting the use of a specific port, natmanger should avoid making routes using this external port. (Which would only work if the clients follow the pattern of mapping the same external port the internal - which afaik go-libp2p natmanger is doing already.)

Both avoiding already used ports reported by mDNS and choosing a random port each time will greatly reduce the likelihood of a NAT table collision.

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with p2p/host/basic/natmgr.go at lines 129-134 and review how routes and ports are selected. Check the NAT traversal and mDNS-related entry points described in the issue, then define tests for avoiding reported ports, retrying with random ports, and stopping after three failures.

Written by the indexing model from the issue text.

Assessment

Tech stack
go
Domain
networking
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.