IPv6 support: discovery and listeners bind IPv4 only

未关闭
#853 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 6 天内回复

还没有人认领这个 Issue。

评估

难度
5/5
预计耗时
一周以上
新手友好度
18/100
Issue 类型
功能
描述清晰度
基本清楚
活跃度
活跃
技术栈
rust
领域
networking

调研方向

Start in bin/ethlambda/src/main.rs: run_node builds p2p_socket on 0.0.0.0, and build_swarm listens on /ip4/0.0.0.0 for QUIC/TCP; spawn_discovery hands one UDP socket to ethrex. Read read_ip and Options::validate_ports for ENR and flag handling, plus #825 for the bootnode refusal. Done means IPv6 listeners/discovery (or a documented ethrex dependency), ip6/udp6/tcp6/quic6 read and written, CLI flags validated, and tests covering dual-stack and IPv6-only peers.

由索引模型根据 Issue 内容生成。

描述

Problem

The node binds IPv4 sockets only, so IPv6 works only for outbound connections:

  • Discovery (discv5) is IPv4-only. run_node builds p2p_socket on 0.0.0.0 (bin/ethlambda/src/main.rs) and passes its IP as DiscoverySpawnConfig::bind_ip. spawn_discovery opens one UdpSocket there and hands it to ethrex's DiscoveryServer. An IPv4 socket cannot send to an IPv6 address or receive from one, so discv5 never reaches an IPv6-only peer, whether it was found by discovery or given as a bootnode.
  • The swarm listens on IPv4 only. build_swarm listens on /ip4/0.0.0.0 for QUIC and TCP, so no peer can connect in over IPv6. Outbound dials to IPv6 do work: libp2p-quic opens a [::]:0 dialer endpoint when no IPv6 listener exists, and TCP opens a socket of the target's family.
  • --discovery.advertise-ip accepts an IPv6 address. It writes ip6 into the ENR, but the sockets behind the record are IPv4, so the published record is one no peer can reach.
  • ENRs from peers. read_ip prefers ip over ip6, and the IPv6-specific ports (udp6, tcp6, quic6) are never read or written.

Today this matters only for IPv6-only peers. A dual-stack record carries ip, which wins.

What support would take

Lighthouse runs dual-stack: --listen-address can be given once per family, the IPv6 ports have their own flags (--port6, --discovery-port6, --quic-port6), and discv5 gets one socket per family. Covering that here means:

  1. Discovery on IPv6 too. ethrex's DiscoveryServer takes a single socket, so this needs either a dual-stack socket with IPv4-mapped addresses handled throughout, or ethrex accepting a second socket.
  2. QUIC and TCP listeners on :: as well as 0.0.0.0.
  3. The ENR's ip6/udp6/tcp6/quic6 entries, written for our record and read from peers'.
  4. CLI flags for the IPv6 listen address and ports, and Options::validate_ports rules for them.
  5. Refusing an IPv6 --discovery.advertise-ip until the above exists, or warning that the record is unreachable.

Related

#825 (open) makes beacon --bootnodes refuse an IPv6 entry that is only a discv5 contact (a bare /ip6/.../udp/<port> multiaddr, or an ENR whose only address is ip6 and whose only port is udp) when discovery is bound to IPv4. The caller passes in the bind address, so the refusal lifts on its own once discovery binds IPv6.

主要语言
Rust
星标
86
派生
34
平均合并
9 天 20 小时
30 天内合并 PR
77

环境准备

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

lambdaclass/ethlambda 的其他 Issue

查看 lambdaclass/ethlambda 的全部 Issue

相似的 Issue

更多 Rust Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。