查看“︁IPv6 术语表 Glossary”︁的源代码

本页收录配置和排障 IPv6 时最常见的术语，按主题分类，中英双语对照。配置实战请参考 [[Photoincat 2 IPv6 PPPoE]] 与 [[Photonicat IPv6 Native, NAT6模式]]。

This page collects the most common IPv6 jargon for configuration and troubleshooting, sorted by topic, bilingual (Chinese / English). For hands-on setup see [[Photoincat 2 IPv6 PPPoE]] and [[Photonicat IPv6 Native, NAT6模式]].

__TOC__

== 一、基础概念 Basics ==
{| class="wikitable" style="width:100%"
! style="width:16%" | 术语 Term !! style="width:44%" | 中文说明 !! English
|-
| '''IPv6''' || 第 6 版互联网协议，128 位地址（约 3.4×10³⁸ 个），解决 IPv4 地址枯竭问题，原生支持自动配置与端到端直连。 || Internet Protocol version 6, with 128-bit addresses (~3.4×10³⁸), solving IPv4 exhaustion; natively supports autoconfiguration and end-to-end connectivity.
|-
| '''IPv4''' || 第 4 版互联网协议，32 位地址（约 43 亿个），已枯竭，普遍依赖 NAT 共享公网地址。 || Internet Protocol version 4, 32-bit addresses (~4.3 billion), exhausted; commonly shared behind NAT.
|-
| '''双栈 Dual Stack''' || 设备/网络同时运行 IPv4 和 IPv6，两套协议并行、互不依赖，是目前最主流的过渡方式。 || Running IPv4 and IPv6 simultaneously and independently on the same device/network; the most common transition strategy today.
|-
| '''地址格式 Address Format''' || 8 组 16 位十六进制数，冒号分隔，如 <code>240e:38c:8449:2200::1</code>。前导零可省略；连续全零组可用 <code>::</code> 压缩一次；<code>%eth0</code> 表示链路本地地址的出口接口（Zone ID）。 || Eight groups of 16-bit hex separated by colons, e.g. <code>240e:38c:8449:2200::1</code>. Leading zeros may be dropped; one run of zero groups can be compressed to <code>::</code>; <code>%eth0</code> is a zone ID selecting the interface for link-local addresses.
|-
| '''前缀 Prefix''' || 地址的高位网络部分，写作 <code>地址/长度</code>，如 <code>240e:38c:8449:2200::/56</code>。类似 IPv4 的 CIDR。 || The high-order network part of an address, written <code>address/length</code>, e.g. <code>240e:38c:8449:2200::/56</code>; analogous to IPv4 CIDR.
|-
| '''/64''' || IPv6 局域网子网的标准大小。SLAAC 只能工作在 /64 上；运营商委托的 /56、/60 会被路由器切分成若干 /64 分给各内网接口。 || The standard IPv6 LAN subnet size. SLAAC only works on /64; an ISP-delegated /56 or /60 is split by the router into multiple /64s for LAN interfaces.
|-
| '''接口标识符 IID (Interface Identifier)''' || 地址的低 64 位，标识子网内的具体接口，可由 EUI-64、随机数或 DHCPv6 生成。 || The low 64 bits of an address identifying an interface within the subnet; generated via EUI-64, randomization, or DHCPv6.
|-
| '''EUI-64''' || 由网卡 48 位 MAC 地址扩展生成 64 位 IID 的方法（中间插入 <code>ff:fe</code> 并翻转 U/L 位），如 MAC <code>bc:24:11:75:97:aa</code> → <code>be24:11ff:fe75:97aa</code>。会暴露硬件信息，故有隐私地址扩展。 || Deriving a 64-bit IID from the 48-bit MAC (insert <code>ff:fe</code>, flip the U/L bit), e.g. MAC <code>bc:24:11:75:97:aa</code> → <code>be24:11ff:fe75:97aa</code>. Leaks hardware identity, hence privacy extensions.
|-
| '''作用域 Scope''' || 地址的有效范围：链路本地（link-local）、唯一本地（ULA）、全球（global）。不同作用域地址不可互相路由。 || The reach of an address: link-local, unique-local (ULA), or global. Addresses of different scopes are not interchangeable for routing.
|-
| '''Hop Limit''' || IPv6 头部字段，等价于 IPv4 的 TTL，每经过一跳减 1，为 0 时丢弃。 || IPv6 header field equivalent to IPv4 TTL; decremented per hop, packet dropped at 0.
|-
| '''MTU''' || 链路最大传输单元。IPv6 最小要求 1280 字节；PPPoE 线路常见 1492。IPv6 路由器不做分片，全靠 PMTUD。 || Maximum Transmission Unit. IPv6 requires ≥1280 bytes; PPPoE links are typically 1492. IPv6 routers never fragment — path MTU discovery is mandatory.
|}

== 二、地址类型 Address Types ==
{| class="wikitable" style="width:100%"
! style="width:16%" | 术语 Term !! style="width:44%" | 中文说明 !! English
|-
| '''GUA (Global Unicast Address)''' || 全球单播地址（<code>2000::/3</code>，如 240e:、2408:、2409: 开头），公网可路由，即"全球直连"地址。 || Globally routable unicast address (<code>2000::/3</code>, e.g. prefixes 240e:, 2408:, 2409:) — the "directly reachable from the internet" address.
|-
| '''ULA (Unique Local Address)''' || 唯一本地地址（<code>fc00::/7</code>，实际使用 <code>fd00::/8</code>），类似 IPv4 私网地址，仅内网有效，不被公网路由。OpenWrt 默认自动生成一个 fd 前缀。 || Unique Local Address (<code>fc00::/7</code>, in practice <code>fd00::/8</code>) — like RFC1918 private space; LAN-only, never routed on the internet. OpenWrt auto-generates one fd prefix by default.
|-
| '''LLA (Link-Local Address)''' || 链路本地地址（<code>fe80::/10</code>），每个 IPv6 接口必有，仅本链路有效，用于邻居发现、路由协商；IPv6 默认网关通常就是路由器的 fe80 地址。 || Link-local address (<code>fe80::/10</code>); mandatory on every IPv6 interface, valid only on-link, used for NDP and routing. The IPv6 default gateway is usually the router's fe80 address.
|-
| '''组播 Multicast''' || <code>ff00::/8</code>，一对多通信。IPv6 没有广播，全部用组播替代，如 <code>ff02::1</code>（本链路所有节点）、<code>ff02::2</code>（所有路由器）。 || <code>ff00::/8</code>, one-to-many. IPv6 has no broadcast — multicast replaces it, e.g. <code>ff02::1</code> (all nodes on link), <code>ff02::2</code> (all routers).
|-
| '''任播 Anycast''' || 同一地址配置在多个节点上，路由把流量送到"最近"的一个，常用于 DNS、CDN。 || The same address on multiple nodes; routing delivers to the "nearest" one. Common for DNS and CDNs.
|-
| '''环回地址 Loopback (::1)''' || 等价于 IPv4 的 127.0.0.1，仅本机有效。 || Equivalent of 127.0.0.1; host-internal only.
|-
| '''未指定地址 Unspecified (::)''' || 全零地址，表示"尚无地址"，如 DAD 探测时的源地址；也用于表示默认路由 <code>::/0</code>。 || The all-zeros address meaning "no address yet" (e.g. source during DAD); <code>::/0</code> also denotes the default route.
|-
| '''被请求节点组播 Solicited-Node Multicast''' || <code>ff02::1:ffXX:XXXX</code>，由单播地址末 24 位生成，邻居发现（解析 MAC）时用它替代广播，效率更高。 || <code>ff02::1:ffXX:XXXX</code>, derived from the last 24 bits of a unicast address; used by NDP for address resolution instead of broadcast.
|-
| '''隐私地址 Privacy / Temporary Address (RFC 4941)''' || 随机生成、定期更换的 SLAAC 地址，用作出站连接的源地址，避免通过固定 IID 被长期跟踪。 || Randomized, periodically rotated SLAAC addresses used as source for outbound connections, preventing tracking via a stable IID.
|-
| '''稳定隐私地址 Stable Privacy (RFC 7217)''' || 不暴露 MAC 但在同一网络内保持稳定的 IID 生成方式（按前缀+密钥哈希），兼顾隐私与可管理性，现代系统默认采用。 || Opaque, per-network-stable IIDs (hash of prefix+secret) that hide the MAC yet stay constant on a given network; the modern OS default.
|-
| '''弃用地址 Deprecated Address''' || preferred lifetime 已过期但 valid lifetime 未到的地址：现有连接可继续用，新连接不再选它作源地址。前缀更换时常见。 || An address past its preferred lifetime but still valid: existing sessions continue, new connections stop selecting it as source. Common during prefix changes.
|-
| '''生存期 Lifetimes (valid / preferred)''' || RA/DHCPv6 为每个地址/前缀附带两个倒计时：preferred（优先使用期）与 valid（有效期）。动态前缀的地址会显示剩余秒数而非 forever。 || Every RA/DHCPv6 address or prefix carries two timers: preferred and valid lifetime. Dynamic-prefix addresses show a countdown instead of "forever".
|}

== 三、地址获取与分配 Address Acquisition ==
{| class="wikitable" style="width:100%"
! style="width:16%" | 术语 Term !! style="width:44%" | 中文说明 !! English
|-
| '''SLAAC (Stateless Address Autoconfiguration)''' || 无状态地址自动配置：主机收到路由器 RA 中带 A 标志的 /64 前缀后，自己拼出完整地址（前缀+IID），无需服务器记录状态。最常用的 IPv6 分配方式，Android 只支持 SLAAC。 || Stateless autoconfiguration: a host combines an RA-advertised /64 prefix (with the A flag) with its own IID to form an address — no server keeps state. The most common IPv6 method; Android supports only SLAAC.
|-
| '''DHCPv6 (Stateful)''' || 有状态地址分配：服务器（如 odhcpd）记录并下发具体 /128 地址与 DNS 等信息，便于集中管理和固定地址。 || Stateful assignment: a server (e.g. odhcpd) leases specific /128 addresses plus options like DNS, enabling central management and static leases.
|-
| '''无状态 DHCPv6 (Stateless DHCPv6)''' || 地址由 SLAAC 生成，仅通过 DHCPv6 获取 DNS 等附加信息（对应 RA 的 O 标志）。 || Addresses come from SLAAC; DHCPv6 only supplies extra options such as DNS (signaled by the RA O flag).
|-
| '''PD (Prefix Delegation, 前缀委托)''' || 运营商通过 DHCPv6-PD 把一整段前缀（常见 /56、/60）委托给你的路由器，路由器再切成 /64 分给各内网接口。这是内网获得公网 IPv6 的关键机制。 || The ISP delegates a whole prefix (typically /56 or /60) to your router via DHCPv6-PD; the router splits it into /64s for LAN interfaces. This is how the LAN gets global IPv6.
|-
| '''RA (Router Advertisement)''' || 路由器周期性（或应 RS 请求）组播的 ICMPv6 通告，携带前缀、默认网关生存期、M/O/A 标志、DNS 等，是 SLAAC 的载体。 || ICMPv6 announcements a router multicasts periodically (or in reply to an RS), carrying prefixes, router lifetime, M/O/A flags, DNS, etc.; the vehicle of SLAAC.
|-
| '''RS (Router Solicitation)''' || 主机上线时主动发出的"请路由器立刻发 RA"的请求（发往 ff02::2）。 || A host's "please send an RA now" request (to ff02::2) when it comes up.
|-
| '''M 标志 (Managed flag)''' || RA 中的标志位：建议客户端用有状态 DHCPv6 获取地址。 || RA flag telling clients to obtain addresses via stateful DHCPv6.
|-
| '''O 标志 (Other-config flag)''' || RA 标志位：地址用 SLAAC，但 DNS 等其他配置去问 DHCPv6。 || RA flag: addresses via SLAAC, but fetch other options (DNS etc.) from DHCPv6.
|-
| '''A 标志 (Autonomous flag)''' || RA 前缀选项里的标志位：允许主机用该前缀做 SLAAC。'''没有 A 标志就没有 SLAAC'''。 || Per-prefix RA flag permitting SLAAC with that prefix. '''No A flag, no SLAAC.'''
|-
| '''L 标志 (On-link flag)''' || RA 前缀选项标志位：该前缀在本链路直达，同前缀主机间直接通信、不经路由器。 || Per-prefix RA flag declaring the prefix on-link, so same-prefix hosts talk directly without the router.
|-
| '''Router Lifetime''' || RA 中宣告"我可当默认网关"的有效期；设为 0 表示"我发前缀/DNS 但别把我当默认网关"。多路由环境排障关键参数。 || How long the sender may serve as default gateway; 0 means "use my prefixes/DNS but not me as gateway". Key knob in multi-router LANs.
|-
| '''路由器优先级 Router Preference (RFC 4191)''' || RA 中的 high/medium/low 优先级，让客户端在多个网关间选优。OpenWrt 对应 <code>ra_preference</code>。 || The high/medium/low preference in RAs guiding clients among multiple gateways; <code>ra_preference</code> in OpenWrt.
|-
| '''RDNSS / DNSSL''' || RA 直接携带的 DNS 服务器地址（RDNSS）与 DNS 搜索域（DNSSL），使纯 SLAAC 环境无需 DHCPv6 也能拿到 DNS。 || DNS server addresses (RDNSS) and search domains (DNSSL) carried directly in RAs, so pure-SLAAC hosts get DNS without DHCPv6.
|-
| '''DUID (DHCP Unique Identifier)''' || DHCPv6 客户端/服务器的唯一标识（替代 IPv4 时代的 MAC），做 IPv6 静态租约时要绑定 DUID 而不是 MAC。 || The unique identifier of DHCPv6 endpoints (replacing MAC from the IPv4 era); IPv6 static leases bind to DUID, not MAC.
|-
| '''IA_NA / IA_PD''' || DHCPv6 请求中的两类"身份关联"：IA_NA 要一个普通地址（/128），IA_PD 要一段委托前缀。路由器 WAN 口两者都会请求。 || The two DHCPv6 identity associations: IA_NA requests a normal /128 address, IA_PD requests a delegated prefix. A router WAN requests both.
|-
| '''源地址选择 Source Address Selection (RFC 6724)''' || 主机有多个地址时选哪个作源地址的规则（作用域匹配、标签匹配、最长前缀匹配等）。多运营商前缀共存时，选错源会被上游按源过滤丢包。 || The rule set (scope match, label match, longest prefix match…) picking a source among multiple addresses. With multiple ISP prefixes, a wrong source gets dropped by upstream source filtering.
|-
| '''地址标签 Address Label''' || RFC 6724 策略表中的标签，可人为调整（如 <code>ip addrlabel</code> / networkd <code>[IPv6AddressLabel]</code>）来控制源地址选择倾向。 || Labels in the RFC 6724 policy table, tunable (via <code>ip addrlabel</code> or networkd <code>[IPv6AddressLabel]</code>) to steer source selection.
|}

== 四、邻居发现与协议机制 NDP & Protocol Mechanics ==
{| class="wikitable" style="width:100%"
! style="width:16%" | 术语 Term !! style="width:44%" | 中文说明 !! English
|-
| '''ICMPv6''' || IPv6 的控制协议，承载差错报告、Ping、邻居发现、MLD 等。'''防火墙不能一刀切禁 ICMPv6'''，否则 NDP/PMTUD 全部失效。 || IPv6's control protocol: errors, ping, NDP, MLD. '''Never blanket-block ICMPv6''' — NDP and PMTUD break.
|-
| '''NDP (Neighbor Discovery Protocol)''' || 邻居发现协议，基于 ICMPv6，替代 IPv4 的 ARP + DHCP 发现 + ICMP 重定向：地址解析、路由器发现、重复检测、不可达检测都靠它。 || The ICMPv6-based protocol replacing ARP, DHCP discovery and redirects: address resolution, router discovery, DAD, and reachability detection.
|-
| '''NS / NA (Neighbor Solicitation / Advertisement)''' || 邻居请求/通告，等价于 ARP 的请求/应答，把 IPv6 地址解析为 MAC。 || The IPv6 equivalents of ARP request/reply, resolving IPv6 addresses to MACs.
|-
| '''DAD (Duplicate Address Detection)''' || 重复地址检测：地址启用前先发 NS 探测有没有人已占用，防止冲突。 || Probing (via NS) that a tentative address is unused before enabling it.
|-
| '''NUD (Neighbor Unreachability Detection)''' || 邻居不可达检测：持续验证邻居（含默认网关）是否仍然可达，不可达时切换路由。<code>ip -6 neigh</code> 里的 REACHABLE/STALE 状态即由它维护。 || Continuously verifying neighbors (incl. gateways) are still reachable, switching routes when not. Maintains the REACHABLE/STALE states seen in <code>ip -6 neigh</code>.
|-
| '''MLD (Multicast Listener Discovery)''' || 组播侦听发现（ICMPv6 版 IGMP），交换机/路由器据此裁剪组播转发。NDP 依赖组播，故桥接设备错误过滤 MLD 会导致 IPv6 玄学故障。 || The ICMPv6 counterpart of IGMP for pruning multicast. NDP relies on multicast, so bridges mis-filtering MLD cause mysterious IPv6 failures.
|-
| '''NDP 代理 NDP Proxy (ndppd / ndp relay)''' || 路由器代替下游设备应答 NS，让上游以为这些地址在本链路，用于旁路由/中继/没有 PD 的场景（OpenWrt <code>ndp: relay</code>）。 || The router answers NS on behalf of downstream hosts so upstream believes they're on-link; used for bridged/relay setups without PD (OpenWrt <code>ndp: relay</code>).
|-
| '''PMTUD (Path MTU Discovery)''' || 路径 MTU 发现：IPv6 路由器不分片，靠 ICMPv6 "Packet Too Big" 让源端减小包长。被防火墙误杀时表现为小包通、大包（网页/下载）挂。 || IPv6 routers don't fragment; ICMPv6 "Packet Too Big" tells the sender to shrink packets. When firewalled away: pings work, big transfers hang.
|-
| '''ECMP (Equal-Cost Multi-Path)''' || 等价多路径：多个同优先级默认路由按流哈希分担。多路由器 LAN 中若某网关无 IPv6 上行，会出现"部分连接黑洞"式间歇故障。 || Flow-hashed load sharing across equal default routes. In multi-router LANs a gateway without IPv6 upstream causes intermittent per-flow blackholes.
|}

== 五、过渡与转换技术 Transition & Translation ==
{| class="wikitable" style="width:100%"
! style="width:16%" | 术语 Term !! style="width:44%" | 中文说明 !! English
|-
| '''NAT6 / NAT66''' || 对 IPv6 做源地址转换：内网用 ULA，出口伪装成路由器的 GUA。牺牲了端到端直连，仅建议在运营商不给 PD 前缀时使用。参见 [[Photonicat IPv6 Native, NAT6模式]]。 || Source-NAT for IPv6: LAN uses ULA, egress masqueraded to the router's GUA. Sacrifices end-to-end reachability; only recommended when the ISP delegates no prefix. See [[Photonicat IPv6 Native, NAT6模式]].
|-
| '''NPTv6 (Network Prefix Translation)''' || 无状态 1:1 前缀替换（RFC 6296），内外前缀一一映射，比 NAT66 干净，常用于多线/前缀变化场景。 || Stateless 1:1 prefix rewriting (RFC 6296); cleaner than NAT66, used for multihoming or unstable prefixes.
|-
| '''NAT64''' || 让纯 IPv6 网络访问 IPv4 互联网的协议转换器，把 <code>64:ff9b::/96</code> 内嵌的 IPv4 地址翻译成真实 IPv4 流量。 || Translates IPv6-only clients to the IPv4 internet by mapping IPv4 addresses embedded in <code>64:ff9b::/96</code>.
|-
| '''DNS64''' || 与 NAT64 配套的 DNS：目标只有 A 记录时合成假 AAAA（指向 NAT64 前缀），引导纯 IPv6 客户端走 NAT64。 || Companion to NAT64: synthesizes AAAA records (pointing into the NAT64 prefix) for IPv4-only destinations.
|-
| '''464XLAT''' || 移动网络主流方案：终端侧 CLAT + 运营商侧 NAT64/PLAT，让纯 IPv6 蜂窝网上的 IPv4-only 应用也能工作。 || The mobile-network staple: client-side CLAT plus provider-side NAT64 (PLAT), letting IPv4-only apps work over IPv6-only cellular.
|-
| '''DS-Lite (Dual-Stack Lite)''' || 运营商只给 IPv6，IPv4 流量封装进 IPv6 隧道送到运营商侧 CGN 统一做 NAT。 || ISP provides only IPv6; IPv4 is tunneled over IPv6 to a carrier-grade NAT.
|-
| '''6in4 / SIT 隧道''' || 把 IPv6 包封装在 IPv4 协议 41 里的静态隧道，如 HE tunnelbroker，常用于运营商无 IPv6 时体验 IPv6。 || Static IPv6-in-IPv4 (protocol 41) tunnels, e.g. HE tunnelbroker; a way to get IPv6 when the ISP offers none.
|-
| '''6to4 / 6rd''' || 由 IPv4 地址自动派生 IPv6 前缀的隧道技术：6to4（2002::/16，已淘汰）和运营商可控的改进版 6rd。 || Tunnels deriving IPv6 prefixes from IPv4 addresses: legacy 6to4 (2002::/16, deprecated) and its ISP-managed successor 6rd.
|-
| '''Teredo''' || 穿越 NAT 的 IPv6-over-UDP 隧道（2001::/32），Windows 曾内置，现基本淘汰。 || IPv6-over-UDP tunneling through NAT (2001::/32); once built into Windows, now essentially retired.
|-
| '''MAP-E / MAP-T''' || 无状态的运营商级 IPv4-over-IPv6 方案，按规则把 IPv4 地址+端口段映射进 IPv6 前缀（封装 E / 翻译 T）。 || Stateless carrier-grade IPv4-over-IPv6: rule-based mapping of IPv4 address+port ranges into IPv6 prefixes (Encapsulation / Translation).
|-
| '''CGN / CGNAT''' || 运营商级 NAT，多个用户共享一个公网 IPv4。IPv6 的普及正是为了摆脱它。 || Carrier-Grade NAT sharing one public IPv4 among many subscribers — exactly what IPv6 adoption avoids.
|}

== 六、OpenWrt / photonicat 相关 OpenWrt & photonicat Specifics ==
{| class="wikitable" style="width:100%"
! style="width:16%" | 术语 Term !! style="width:44%" | 中文说明 !! English
|-
| '''odhcpd''' || OpenWrt 的 RA / DHCPv6 / NDP 代理服务端，向 LAN 下发前缀和地址；LuCI 里 LAN 的 "DHCPv6/RA 服务" 就是它。 || OpenWrt's RA/DHCPv6/NDP-proxy server distributing prefixes and addresses to the LAN; the "DHCPv6/RA service" settings in LuCI drive it.
|-
| '''odhcp6c''' || OpenWrt 的 WAN 侧 DHCPv6 客户端，负责向运营商请求 IA_NA 地址和 IA_PD 前缀。 || OpenWrt's WAN-side DHCPv6 client requesting IA_NA addresses and IA_PD prefixes from the ISP.
|-
| '''wan_6 (动态别名接口)''' || WAN 设为 <code>ipv6='auto'</code> 时，netifd 在 PPPoE/DHCP 逻辑口上自动创建的 IPv6 虚拟接口（注意下划线，区别于手工建的 wan6），负责 DHCPv6/PD。26.04 固件无需手工创建 wan6。 || The dynamic IPv6 alias interface netifd spawns on the PPPoE/DHCP link when WAN has <code>ipv6='auto'</code> (underscore — distinct from a manual wan6). Handles DHCPv6/PD; no manual wan6 needed on 26.04 firmware.
|-
| '''ip6assign''' || LAN 接口选项：从委托前缀中给该接口分配多长的子网（通常 60 或 64）。'''不设则 LAN 拿不到公网前缀'''。 || LAN option: the subnet length assigned to this interface from the delegated pool (usually 60 or 64). '''Unset = LAN never gets a global prefix.'''
|-
| '''ip6class (IPv6 前缀过滤器)''' || 限制接口只接受某类前缀（如 local=ULA、wan_6）。'''最常见故障源：被设成 local 后内网只有 fd 地址、无公网 IPv6'''。一般应留空。 || Restricts which prefix classes an interface accepts (e.g. local=ULA, wan_6). '''Top failure cause: set to "local" the LAN only gets fd addresses and no global IPv6.''' Normally leave empty.
|-
| '''ip6hint''' || 指定接口从委托前缀中取哪个 /64 子网的十六进制序号，便于固定各 VLAN 的子网编号。 || Hex hint choosing which /64 out of the delegation an interface takes; handy for stable per-VLAN subnet numbering.
|-
| '''ula_prefix (network.globals)''' || OpenWrt 全局自动生成的 ULA 前缀（fd 开头），保证公网前缀变化/断线时内网 IPv6 互访依然稳定。 || The auto-generated global ULA prefix (fd…) keeping LAN-internal IPv6 stable even when the ISP prefix changes or the WAN drops.
|-
| '''混合模式 Hybrid mode''' || LuCI 中 RA/DHCPv6/NDP 的 "hybrid"：有委托前缀时当服务器，配置了中继上游时当中继，自动切换，日常建议选它。 || LuCI's "hybrid" for RA/DHCPv6/NDP: acts as server when holding a delegated prefix, as relay when an upstream master is set. The sensible default.
|-
| '''fw4 / nftables''' || OpenWrt 22.03 起的防火墙（基于 nftables），IPv4/IPv6 统一规则。默认放行 lan→wan 转发与必要 ICMPv6，'''IPv6 上网无需加规则'''。 || OpenWrt's firewall since 22.03 (nftables-based) with unified v4/v6 rules. Allows lan→wan forwarding and essential ICMPv6 by default — '''no extra rules needed for outbound IPv6'''.
|-
| '''Docker ip6tables 坑 (Docker ip6tables pitfall)''' || Docker 27+ 默认接管 ip6tables，创建 policy drop 的 <code>ip6 filter FORWARD</code> 链，'''丢掉所有转发的 IPv6'''（fw4 放行也没用）。修复见 [[Photoincat 2 IPv6 PPPoE]] 第 4 节。 || Docker 27+ manages ip6tables by default, creating an <code>ip6 filter FORWARD</code> chain with policy drop that '''kills all forwarded IPv6''' regardless of fw4. Fix in [[Photoincat 2 IPv6 PPPoE]] §4.
|-
| '''源限定默认路由 Source-restricted default route''' || OpenWrt 安装的 IPv6 默认路由形如 <code>default from 240e:xxxx::/56 via …</code>，只转发源地址属于本运营商前缀的流量，防止源地址伪造/多线源错配。 || OpenWrt installs IPv6 defaults like <code>default from 240e:xxxx::/56 via …</code>, forwarding only traffic sourced from that ISP's prefix — anti-spoofing, and a gotcha in multihomed LANs.
|-
| '''sourcefilter''' || OpenWrt DHCPv6 接口选项，设 0 可关闭上述按源过滤（放宽多线场景限制）。 || OpenWrt DHCPv6 interface option; 0 disables the source filtering above (loosens multihoming).
|}

== 七、DNS 与诊断 DNS & Diagnostics ==
{| class="wikitable" style="width:100%"
! style="width:16%" | 术语 Term !! style="width:44%" | 中文说明 !! English
|-
| '''AAAA 记录''' || 域名到 IPv6 地址的 DNS 记录（A 记录的 IPv6 版）。<code>nslookup -type=AAAA photonicat.com</code> 可查询。 || The DNS record mapping a name to an IPv6 address (IPv6 counterpart of A). Query with <code>nslookup -type=AAAA photonicat.com</code>.
|-
| '''Happy Eyeballs (RFC 8305)''' || 双栈主机同时尝试 IPv6 和 IPv4 连接、择快而用的算法，避免一边故障导致网页卡顿。也因此 IPv6 半瘫时症状常是"偶尔慢"而非"全断"。 || Dual-stack hosts race IPv6 vs IPv4 and use the faster; broken IPv6 thus shows up as "sometimes slow", not "fully down".
|-
| '''ping -6 / ping6''' || IPv6 连通性测试，如 <code>ping -6 photonicat.com</code>；链路本地地址要带接口：<code>ping fe80::1%br-lan</code>。 || IPv6 reachability test, e.g. <code>ping -6 photonicat.com</code>; link-local targets need a zone: <code>ping fe80::1%br-lan</code>.
|-
| '''traceroute6 / mtr''' || 逐跳追踪 IPv6 路径，定位断在内网、运营商还是远端。 || Hop-by-hop IPv6 path tracing to locate whether failures are local, ISP-side, or remote.
|-
| '''ip -6 addr / ip -6 route / ip -6 neigh''' || Linux 查看 IPv6 地址、路由表（注意 <code>default via fe80::…</code>）和邻居表（等价 ARP 表）的三板斧。 || The Linux trio for inspecting addresses, routes (note <code>default via fe80::…</code>) and the neighbor (ARP-equivalent) table.
|-
| '''ifstatus''' || OpenWrt 查询接口运行状态的命令：<code>ifstatus wan_6</code> 看运营商给的地址/前缀，<code>ifstatus lan</code> 看前缀是否分到内网。 || OpenWrt's interface state tool: <code>ifstatus wan_6</code> shows ISP-issued address/prefix, <code>ifstatus lan</code> shows whether it reached the LAN.
|-
| '''RA 抓包 (tcpdump)''' || <code>tcpdump -i br-lan -vv 'icmp6 and ip6[40]==134'</code> 直接看 RA 的前缀和标志位，是排查 SLAAC 的终极手段。 || <code>tcpdump -i br-lan -vv 'icmp6 and ip6[40]==134'</code> shows RA prefixes and flags on the wire — the definitive SLAAC debugging tool.
|-
| '''test-ipv6.com''' || 浏览器端一键检测 IPv6 连通性和评分的网站（国内亦可用 testipv6.cn）。 || Browser-based one-click IPv6 connectivity scoring (testipv6.cn inside China).
|}

[[Photoincat 2 IPv6 PPPoE|← 返回 IPv6 PPPoE 配置指南 / Back to the IPv6 PPPoE guide]]