fix(dns): adapt resolver chain to dns policy branch

# Conflicts:
#	Cargo.lock
#	easytier/build/main.rs

.github/workflows/core.yml

@@ -157,6 +157,9 @@ jobs:
 
       - uses: mlugg/setup-zig@v2
         if: ${{ contains(matrix.OS, 'ubuntu') }}
+        with:
+          version: 0.16.0
+          use-cache: true
 
       - uses: taiki-e/install-action@v2
         if: ${{ contains(matrix.OS, 'ubuntu') }}
@@ -227,7 +230,7 @@ jobs:
               *)       UPX_ARCH="amd64" ;;
             esac
 
-            UPX_VERSION=5.1.1
+            UPX_VERSION=4.2.4
             UPX_PKG="upx-${UPX_VERSION}-${UPX_ARCH}_linux"
             curl -L "https://github.com/upx/upx/releases/download/v${UPX_VERSION}/${UPX_PKG}.tar.xz" -s | tar xJvf -
             cp "${UPX_PKG}/upx" .

.github/workflows/docker.yml

@@ -11,7 +11,7 @@ on:
       image_tag:
         description: 'Tag for this image build'
         type: string
-        default: 'v2.6.3'
+        default: 'v2.6.4'
         required: true
       mark_latest:
         description: 'Mark this image as latest'

.github/workflows/release.yml

@@ -18,7 +18,7 @@ on:
       version:
         description: 'Version for this release'
         type: string
-        default: 'v2.6.3'
+        default: 'v2.6.4'
         required: true
       make_latest:
         description: 'Mark this release as latest'

Cargo.lock

@@ -915,7 +915,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "bfcfdc083699101d5a7965e49925975f2f55060f94f9a05e7187be95d530ca59"
 dependencies = [
  "once_cell",
- "proc-macro-crate 3.2.0",
+ "proc-macro-crate 3.5.0",
  "proc-macro2",
  "quote",
  "syn 2.0.117",
@@ -1981,37 +1981,6 @@ dependencies = [
  "syn 2.0.117",
 ]
 
-[[package]]
-name = "derive_builder"
-version = "0.20.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "507dfb09ea8b7fa618fcf76e953f4f5e192547945816d5358edffe39f6f94947"
-dependencies = [
- "derive_builder_macro",
-]
-
-[[package]]
-name = "derive_builder_core"
-version = "0.20.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2d5bcf7b024d6835cfb3d473887cd966994907effbe9227e8c8219824d06c4e8"
-dependencies = [
- "darling 0.20.10",
- "proc-macro2",
- "quote",
- "syn 2.0.117",
-]
-
-[[package]]
-name = "derive_builder_macro"
-version = "0.20.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ab63b0e2bf4d5928aff72e83a7dace85d7bba5fe12dcc3c5a572d78caffd3f3c"
-dependencies = [
- "derive_builder_core",
- "syn 2.0.117",
-]
-
 [[package]]
 name = "derive_more"
 version = "0.99.20"
@@ -2229,7 +2198,7 @@ checksum = "d0881ea181b1df73ff77ffaaf9c7544ecc11e82fba9b5f27b262a3c73a332555"
 
 [[package]]
 name = "easytier"
-version = "2.6.3"
+version = "2.6.4"
 dependencies = [
  "aes-gcm",
  "anyhow",
@@ -2262,35 +2231,36 @@ dependencies = [
  "defguard_wireguard_rs",
  "delegate",
  "derivative",
- "derive_builder",
  "derive_more 2.1.1",
- "easytier-rpc-build",
  "encoding",
  "flume 0.12.0",
  "forwarded-header-value",
  "futures",
  "futures-util",
- "gethostname 0.5.0",
+ "getset",
  "git-version",
  "globwalk",
  "guarden",
- "hickory-client",
+ "hickory-net",
  "hickory-proto",
  "hickory-resolver",
  "hickory-server",
  "hmac",
+ "hostname",
  "http",
  "http_req",
  "humansize",
  "humantime-serde",
  "idna 1.0.3",
  "igd-next",
+ "indexmap 2.14.0",
  "indoc",
  "itertools 0.14.0",
  "kcp-sys",
  "machine-uid",
  "maplit",
  "mimalloc",
+ "moka",
  "multimap",
  "natpmp",
  "netlink-packet-core",
@@ -2301,6 +2271,7 @@ dependencies = [
  "nix 0.29.0",
  "once_cell",
  "openssl",
+ "optionize",
  "ordered_hash_map",
  "parking_lot",
  "paste",
@@ -2308,7 +2279,8 @@ dependencies = [
  "petgraph 0.8.1",
  "pin-project-lite",
  "pnet",
- "prefix-trie",
+ "prefix-trie 0.7.0",
+ "proc-macro2",
  "prost",
  "prost-build",
  "prost-reflect",
@@ -2318,6 +2290,7 @@ dependencies = [
  "prost-wkt-types",
  "quinn",
  "quinn-plaintext",
+ "quote",
  "rand 0.8.5",
  "rcgen",
  "regex",
@@ -2330,6 +2303,7 @@ dependencies = [
  "rustls",
  "serde",
  "serde_json",
+ "serde_with",
  "serial_test",
  "service-manager",
  "sha2",
@@ -2357,7 +2331,6 @@ dependencies = [
  "tokio-util",
  "tokio-websockets",
  "toml 0.8.19",
- "tonic-build",
  "tracing",
  "tracing-subscriber",
  "tun-easytier",
@@ -2384,7 +2357,7 @@ version = "0.1.0"
 dependencies = [
  "android_logger",
  "easytier",
- "jni",
+ "jni 0.21.1",
  "log",
  "once_cell",
  "serde",
@@ -2405,7 +2378,7 @@ dependencies = [
 
 [[package]]
 name = "easytier-gui"
-version = "2.6.3"
+version = "2.6.4"
 dependencies = [
  "anyhow",
  "async-trait",
@@ -2413,7 +2386,7 @@ dependencies = [
  "dashmap",
  "dunce",
  "easytier",
- "gethostname 1.1.0",
+ "gethostname",
  "libc",
  "once_cell",
  "security-framework-sys",
@@ -2436,14 +2409,6 @@ dependencies = [
  "windows 0.52.0",
 ]
 
-[[package]]
-name = "easytier-rpc-build"
-version = "0.1.0"
-dependencies = [
- "heck 0.5.0",
- "prost-build",
-]
-
 [[package]]
 name = "easytier-uptime"
 version = "0.1.0"
@@ -2486,7 +2451,7 @@ dependencies = [
 
 [[package]]
 name = "easytier-web"
-version = "2.6.3"
+version = "2.6.4"
 dependencies = [
  "anyhow",
  "async-trait",
@@ -2708,24 +2673,6 @@ version = "1.1.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "66b7e2430c6dff6a955451e2cfc438f09cea1965a9d6f87f7e3b90decc014099"
 
-[[package]]
-name = "endian-type"
-version = "0.1.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c34f04666d835ff5d62e058c3995147c06f42fe86ff053337632bca83e42702d"
-
-[[package]]
-name = "enum-as-inner"
-version = "0.6.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a1e6a265c649f3f5979b601d26f1d05ada116434c87741c9493cb56218f76cbc"
-dependencies = [
- "heck 0.5.0",
- "proc-macro2",
- "quote",
- "syn 2.0.117",
-]
-
 [[package]]
 name = "enumflags2"
 version = "0.7.12"
@@ -3307,16 +3254,6 @@ dependencies = [
  "zeroize",
 ]
 
-[[package]]
-name = "gethostname"
-version = "0.5.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "dc3655aa6818d65bc620d6911f05aa7b6aeb596291e1e9f79e52df85583d1e30"
-dependencies = [
- "rustix 0.38.34",
- "windows-targets 0.52.6",
-]
-
 [[package]]
 name = "gethostname"
 version = "1.1.0"
@@ -3379,6 +3316,18 @@ dependencies = [
  "wasip3",
 ]
 
+[[package]]
+name = "getset"
+version = "0.1.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9cf0fc11e47561d47397154977bc219f4cf809b2974facc3ccb3b89e2436f912"
+dependencies = [
+ "proc-macro-error2",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
 [[package]]
 name = "ghash"
 version = "0.5.1"
@@ -3742,88 +3691,103 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "7f24254aa9a54b5c858eaee2f5bccdb46aaf0e486a595ed5fd8f86ba55232a70"
 
 [[package]]
-name = "hickory-client"
-version = "0.25.2"
+name = "hickory-net"
+version = "0.26.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c466cd63a4217d5b2b8e32f23f58312741ce96e3c84bf7438677d2baff0fc555"
-dependencies = [
- "cfg-if",
- "data-encoding",
- "futures-channel",
- "futures-util",
- "hickory-proto",
- "once_cell",
- "radix_trie",
- "rand 0.9.1",
- "thiserror 2.0.11",
- "tokio",
- "tracing",
-]
-
-[[package]]
-name = "hickory-proto"
-version = "0.25.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f8a6fe56c0038198998a6f217ca4e7ef3a5e51f46163bd6dd60b5c71ca6c6502"
+checksum = "e2295ed2f9c31e471e1428a8f88a3f0e1f4b27c15049592138d1eebe9c35b183"
 dependencies = [
  "async-trait",
+ "bytes",
  "cfg-if",
  "data-encoding",
- "enum-as-inner",
  "futures-channel",
  "futures-io",
  "futures-util",
+ "h2",
+ "hickory-proto",
+ "http",
  "idna 1.0.3",
  "ipnet",
- "once_cell",
- "rand 0.9.1",
- "ring",
+ "jni 0.22.4",
+ "rand 0.10.1",
+ "rustls",
  "serde",
  "thiserror 2.0.11",
  "tinyvec",
  "tokio",
+ "tokio-rustls",
+ "tracing",
+ "url",
+ "webpki-roots 1.0.6",
+]
+
+[[package]]
+name = "hickory-proto"
+version = "0.26.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0bab31817bfb44672a252e97fe81cd0c18d1b2cf892108922f6818820df8c643"
+dependencies = [
+ "data-encoding",
+ "idna 1.0.3",
+ "ipnet",
+ "jni 0.22.4",
+ "once_cell",
+ "prefix-trie 0.8.4",
+ "rand 0.10.1",
+ "ring",
+ "serde",
+ "thiserror 2.0.11",
+ "tinyvec",
  "tracing",
  "url",
 ]
 
 [[package]]
 name = "hickory-resolver"
-version = "0.25.2"
+version = "0.26.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "dc62a9a99b0bfb44d2ab95a7208ac952d31060efc16241c87eaf36406fecf87a"
+checksum = "f0d58d28879ceecde6607729660c2667a081ccdc082e082675042793960f178c"
 dependencies = [
  "cfg-if",
  "futures-util",
+ "hickory-net",
  "hickory-proto",
  "ipconfig",
+ "ipnet",
+ "jni 0.22.4",
  "moka",
+ "ndk-context",
  "once_cell",
  "parking_lot",
- "rand 0.9.1",
+ "rand 0.10.1",
  "resolv-conf",
+ "rustls",
  "serde",
  "smallvec",
+ "system-configuration 0.7.0",
  "thiserror 2.0.11",
  "tokio",
+ "tokio-rustls",
  "tracing",
+ "webpki-roots 1.0.6",
 ]
 
 [[package]]
 name = "hickory-server"
-version = "0.25.2"
+version = "0.26.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d53e5fe811b941c74ee46b8818228bfd2bc2688ba276a0eaeb0f2c95ea3b2585"
+checksum = "130236ba6abba90da6a7acf7a87b27d862b592c3145dc74bc47bf86d8ff198ec"
 dependencies = [
  "async-trait",
  "bytes",
  "cfg-if",
  "data-encoding",
- "enum-as-inner",
  "futures-util",
+ "hickory-net",
  "hickory-proto",
  "hickory-resolver",
  "ipnet",
- "prefix-trie",
+ "prefix-trie 0.8.4",
  "serde",
  "thiserror 2.0.11",
  "time",
@@ -3859,6 +3823,17 @@ dependencies = [
  "windows-sys 0.52.0",
 ]
 
+[[package]]
+name = "hostname"
+version = "0.4.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "617aaa3557aef3810a6369d0a99fac8a080891b68bd9f9812a1eeda0c0730cbd"
+dependencies = [
+ "cfg-if",
+ "libc",
+ "windows-link 0.2.1",
+]
+
 [[package]]
 name = "html5ever"
 version = "0.29.1"
@@ -4044,7 +4019,7 @@ dependencies = [
  "libc",
  "percent-encoding",
  "pin-project-lite",
- "socket2 0.5.10",
+ "socket2 0.6.3",
  "tokio",
  "tower-service",
  "tracing",
@@ -4556,19 +4531,68 @@ dependencies = [
  "cesu8",
  "cfg-if",
  "combine",
- "jni-sys",
+ "jni-sys 0.3.0",
  "log",
  "thiserror 1.0.63",
  "walkdir",
  "windows-sys 0.45.0",
 ]
 
+[[package]]
+name = "jni"
+version = "0.22.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5efd9a482cf3a427f00d6b35f14332adc7902ce91efb778580e180ff90fa3498"
+dependencies = [
+ "cfg-if",
+ "combine",
+ "jni-macros",
+ "jni-sys 0.4.1",
+ "log",
+ "simd_cesu8",
+ "thiserror 2.0.11",
+ "walkdir",
+ "windows-link 0.2.1",
+]
+
+[[package]]
+name = "jni-macros"
+version = "0.22.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a00109accc170f0bdb141fed3e393c565b6f5e072365c3bd58f5b062591560a3"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "rustc_version",
+ "simd_cesu8",
+ "syn 2.0.117",
+]
+
 [[package]]
 name = "jni-sys"
 version = "0.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "8eaf4bc02d17cbdd7ff4c7438cafcdf7fb9a4613313ad11b4f8fefe7d3fa0130"
 
+[[package]]
+name = "jni-sys"
+version = "0.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c6377a88cb3910bee9b0fa88d4f42e1d2da8e79915598f65fb0c7ee14c878af2"
+dependencies = [
+ "jni-sys-macros",
+]
+
+[[package]]
+name = "jni-sys-macros"
+version = "0.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "38c0b942f458fe50cdac086d2f946512305e5631e720728f2a61aabcd47a6264"
+dependencies = [
+ "quote",
+ "syn 2.0.117",
+]
+
 [[package]]
 name = "jobserver"
 version = "0.1.32"
@@ -5103,9 +5127,12 @@ version = "0.12.10"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "a9321642ca94a4282428e6ea4af8cc2ca4eac48ac7a6a4ea8f33f76d0ce70926"
 dependencies = [
+ "async-lock",
  "crossbeam-channel",
  "crossbeam-epoch",
  "crossbeam-utils",
+ "event-listener",
+ "futures-util",
  "loom",
  "parking_lot",
  "portable-atomic",
@@ -5207,7 +5234,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c3f42e7bbe13d351b6bead8286a43aac9534b82bd3cc43e47037f012ebfd62d4"
 dependencies = [
  "bitflags 2.8.0",
- "jni-sys",
+ "jni-sys 0.3.0",
  "log",
  "ndk-sys",
  "num_enum",
@@ -5227,7 +5254,7 @@ version = "0.6.0+11769913"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ee6cda3051665f1fb8d9e08fc35c96d5a244fb1be711a03b71118828afc9a873"
 dependencies = [
- "jni-sys",
+ "jni-sys 0.3.0",
 ]
 
 [[package]]
@@ -5243,7 +5270,7 @@ dependencies = [
  "netlink-packet-route 0.17.1",
  "netlink-sys",
  "once_cell",
- "system-configuration",
+ "system-configuration 0.6.1",
  "windows-sys 0.52.0",
 ]
 
@@ -5354,15 +5381,6 @@ version = "1.0.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "650eef8c711430f1a879fdd01d4745a7deea475becfb90269c06775983bbf086"
 
-[[package]]
-name = "nibble_vec"
-version = "0.1.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "77a5d83df9f36fe23f0c3648c6bbb8b0298bb5f1939c8f2704431371f4b84d43"
-dependencies = [
- "smallvec",
-]
-
 [[package]]
 name = "nix"
 version = "0.25.1"
@@ -5581,7 +5599,7 @@ version = "0.7.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "680998035259dcfcafe653688bf2aa6d3e2dc05e98be6ab46afb089dc84f1df8"
 dependencies = [
- "proc-macro-crate 3.2.0",
+ "proc-macro-crate 2.0.0",
  "proc-macro2",
  "quote",
  "syn 2.0.117",
@@ -5932,6 +5950,30 @@ version = "0.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "04744f49eae99ab78e0d5c0b603ab218f515ea8cfe5a456d7629ad883a3b6e7d"
 
+[[package]]
+name = "optionize"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f15b218321873a7919c8b2a188ad488bf18c067478274be4c69c3d402b2be3ea"
+dependencies = [
+ "delegate",
+ "derive_more 2.1.1",
+ "optionize-macros",
+]
+
+[[package]]
+name = "optionize-macros"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1af5a2aba73a1808cbb39b451aabb5f8dc5c9c73d68ce5e61b79f8892f775a0f"
+dependencies = [
+ "darling 0.23.0",
+ "proc-macro-crate 3.5.0",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
 [[package]]
 name = "ordered-float"
 version = "2.10.1"
@@ -6661,6 +6703,17 @@ dependencies = [
  "num-traits",
 ]
 
+[[package]]
+name = "prefix-trie"
+version = "0.8.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4cf6e3177f0684016a5c209b00882e15f8bdd3f3bb48f0491df10cd102d0c6e7"
+dependencies = [
+ "either",
+ "ipnet",
+ "num-traits",
+]
+
 [[package]]
 name = "prettyplease"
 version = "0.2.20"
@@ -6701,11 +6754,11 @@ dependencies = [
 
 [[package]]
 name = "proc-macro-crate"
-version = "3.2.0"
+version = "3.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8ecf48c7ca261d60b74ab1a7b20da18bede46776b2e55535cb958eb595c5fa7b"
+checksum = "e67ba7e9b2b56446f1d419b1d807906278ffa1a658a8a5d8a39dcb1f5a78614f"
 dependencies = [
- "toml_edit 0.22.20",
+ "toml_edit 0.25.12+spec-1.1.0",
 ]
 
 [[package]]
@@ -7058,16 +7111,6 @@ version = "0.7.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "dc33ff2d4973d518d823d61aa239014831e521c75da58e3df4840d3f47749d09"
 
-[[package]]
-name = "radix_trie"
-version = "0.2.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c069c179fcdc6a2fe24d8d18305cf085fdbd4f922c041943e203685d6a1c58fd"
-dependencies = [
- "endian-type",
- "nibble_vec",
-]
-
 [[package]]
 name = "rand"
 version = "0.7.3"
@@ -7394,7 +7437,7 @@ dependencies = [
  "serde_json",
  "serde_urlencoded",
  "sync_wrapper",
- "system-configuration",
+ "system-configuration 0.6.1",
  "tokio",
  "tokio-native-tls",
  "tokio-rustls",
@@ -7569,7 +7612,7 @@ checksum = "1f168d99749d307be9de54d23fd226628d99768225ef08f6ffb52e0182a27746"
 dependencies = [
  "cfg-if",
  "glob",
- "proc-macro-crate 3.2.0",
+ "proc-macro-crate 3.5.0",
  "proc-macro2",
  "quote",
  "regex",
@@ -7732,6 +7775,7 @@ version = "0.23.27"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "730944ca083c1c233a75c09f199e973ca499344a2b7ba9e755c457e86fb4a321"
 dependencies = [
+ "log",
  "once_cell",
  "ring",
  "rustls-pki-types",
@@ -7778,7 +7822,7 @@ checksum = "19787cda76408ec5404443dc8b31795c87cd8fec49762dc75fa727740d34acc1"
 dependencies = [
  "core-foundation 0.10.0",
  "core-foundation-sys",
- "jni",
+ "jni 0.21.1",
  "log",
  "once_cell",
  "rustls",
@@ -8577,6 +8621,16 @@ version = "0.3.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d66dc143e6b11c1eddc06d5c423cfc97062865baf299914ab64caa38182078fe"
 
+[[package]]
+name = "simd_cesu8"
+version = "1.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "94f90157bb87cddf702797c5dadfa0be7d266cdf49e22da2fcaa32eff75b2c33"
+dependencies = [
+ "rustc_version",
+ "simdutf8",
+]
+
 [[package]]
 name = "simdutf8"
 version = "0.1.5"
@@ -9136,6 +9190,17 @@ dependencies = [
  "system-configuration-sys",
 ]
 
+[[package]]
+name = "system-configuration"
+version = "0.7.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a13f3d0daba03132c0aa9767f98351b3488edc2c100cda2d2ec2b04f3d8d3c8b"
+dependencies = [
+ "bitflags 2.8.0",
+ "core-foundation 0.9.4",
+ "system-configuration-sys",
+]
+
 [[package]]
 name = "system-configuration-sys"
 version = "0.6.0"
@@ -9205,7 +9270,7 @@ dependencies = [
  "gdkwayland-sys",
  "gdkx11-sys",
  "gtk",
- "jni",
+ "jni 0.21.1",
  "libc",
  "log",
  "ndk",
@@ -9267,7 +9332,7 @@ dependencies = [
  "heck 0.5.0",
  "http",
  "image 0.25.10",
- "jni",
+ "jni 0.21.1",
  "libc",
  "log",
  "mime",
@@ -9402,7 +9467,7 @@ version = "2.3.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d8f08346c8deb39e96f86973da0e2d76cbb933d7ac9b750f6dc4daf955a6f997"
 dependencies = [
- "gethostname 1.1.0",
+ "gethostname",
  "log",
  "os_info",
  "serde",
@@ -9495,7 +9560,7 @@ dependencies = [
  "dpi",
  "gtk",
  "http",
- "jni",
+ "jni 0.21.1",
  "objc2",
  "objc2-ui-kit",
  "objc2-web-kit",
@@ -9518,7 +9583,7 @@ checksum = "e11ea2e6f801d275fdd890d6c9603736012742a1c33b96d0db788c9cdebf7f9e"
 dependencies = [
  "gtk",
  "http",
- "jni",
+ "jni 0.21.1",
  "log",
  "objc2",
  "objc2-app-kit",
@@ -9969,6 +10034,15 @@ dependencies = [
  "serde_core",
 ]
 
+[[package]]
+name = "toml_datetime"
+version = "1.1.1+spec-1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3165f65f62e28e0115a00b2ebdd37eb6f3b641855f9d636d3cd4103767159ad7"
+dependencies = [
+ "serde_core",
+]
+
 [[package]]
 name = "toml_edit"
 version = "0.19.15"
@@ -10006,6 +10080,18 @@ dependencies = [
  "winnow 0.6.18",
 ]
 
+[[package]]
+name = "toml_edit"
+version = "0.25.12+spec-1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d2153edc6955a6c354fad8f5efd38b6a8769bdccf9fe50f8e1329f81b0baa5d7"
+dependencies = [
+ "indexmap 2.14.0",
+ "toml_datetime 1.1.1+spec-1.1.0",
+ "toml_parser",
+ "winnow 1.0.1",
+]
+
 [[package]]
 name = "toml_parser"
 version = "1.1.2+spec-1.1.0"
@@ -10051,19 +10137,6 @@ dependencies = [
  "tracing",
 ]
 
-[[package]]
-name = "tonic-build"
-version = "0.12.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "568392c5a2bd0020723e3f387891176aabafe36fd9fcd074ad309dfa0c8eb964"
-dependencies = [
- "prettyplease",
- "proc-macro2",
- "prost-build",
- "quote",
- "syn 2.0.117",
-]
-
 [[package]]
 name = "tower"
 version = "0.4.13"
@@ -11866,6 +11939,9 @@ name = "winnow"
 version = "1.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "09dac053f1cd375980747450bfc7250c264eaae0583872e845c0c7cd578872b5"
+dependencies = [
+ "memchr",
+]
 
 [[package]]
 name = "winreg"
@@ -12067,7 +12143,7 @@ dependencies = [
  "html5ever",
  "http",
  "javascriptcore-rs",
- "jni",
+ "jni 0.21.1",
  "kuchikiki",
  "libc",
  "ndk",
@@ -12130,7 +12206,7 @@ version = "0.13.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9993aa5be5a26815fe2c3eacfc1fde061fc1a1f094bf1ad2a18bf9c495dd7414"
 dependencies = [
- "gethostname 1.1.0",
+ "gethostname",
  "rustix 1.0.7",
  "x11rb-protocol",
 ]
@@ -12247,7 +12323,7 @@ version = "5.14.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "897e79616e84aac4b2c46e9132a4f63b93105d54fe8c0e8f6bffc21fa8d49222"
 dependencies = [
- "proc-macro-crate 3.2.0",
+ "proc-macro-crate 3.5.0",
  "proc-macro2",
  "quote",
  "syn 2.0.117",
@@ -12484,7 +12560,7 @@ version = "5.10.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "5b59b012ebe9c46656f9cc08d8da8b4c726510aef12559da3e5f1bf72780752c"
 dependencies = [
- "proc-macro-crate 3.2.0",
+ "proc-macro-crate 3.5.0",
  "proc-macro2",
  "quote",
  "syn 2.0.117",

Cargo.toml

@@ -3,7 +3,6 @@ resolver = "2"
 members = [
     "easytier",
     "easytier-gui/src-tauri",
-    "easytier-rpc-build",
     "easytier-web",
     "easytier-contrib/easytier-ffi",
     "easytier-contrib/easytier-uptime",

easytier-contrib/easytier-magisk/module.prop

@@ -1,6 +1,6 @@
 id=easytier_magisk
 name=EasyTier_Magisk
-version=v2.6.3
+version=v2.6.4
 versionCode=1
 author=EasyTier
 description=easytier magisk module @EasyTier(https://github.com/EasyTier/EasyTier)

easytier-contrib/easytier-ohrs/Cargo.toml

@@ -7,6 +7,10 @@ edition = "2024"
 crate-type=["cdylib"]
 
 [dependencies]
+async-trait = "0.1"
+base64 = "0.22"
+flate2 = "1.1"
+gethostname = "1.1"
 ohos-hilog-binding = {version = "*", features = ["redirect"]}
 easytier = { path = "../../easytier" }
 napi-derive-ohos = "1.1"
@@ -26,10 +30,16 @@ napi-ohos = { version = "1.1", default-features = false, features = [
     "web_stream",
 ] }
 once_cell = "1.21.3"
+ipnet = "2.10"
+serde = { version = "1.0", features = ["derive"] }
 serde_json = "1.0.125"
+prost-reflect = { version = "0.14.5", default-features = false, features = ["derive"] }
+rusqlite = { version = "0.32", features = ["bundled"] }
 tracing-subscriber = "0.3.19"
 tracing-core = "0.1.33"
 tracing = "0.1.41"
+tokio = { version = "1", features = ["rt-multi-thread", "sync", "time"] }
+url = "2.5"
 uuid = { version = "1.5.0", features = [
     "v4",
     "fast-rng",

easytier-contrib/easytier-ohrs/src/config.rs

@@ -0,0 +1,4 @@
+pub(crate) mod repository;
+pub(crate) mod services;
+pub(crate) mod storage;
+pub(crate) mod types;

easytier-contrib/easytier-ohrs/src/config/repository/mod.rs

@@ -0,0 +1,13 @@
+#[path = "../../config_repo/field_store.rs"]
+mod field_store;
+#[path = "../../config_repo/import_export.rs"]
+mod import_export;
+#[path = "../../config_repo/legacy_migration.rs"]
+mod legacy_migration;
+#[path = "../../config_repo/validation.rs"]
+mod validation;
+
+#[path = "../../config_repo.rs"]
+mod repo;
+
+pub use repo::*;

easytier-contrib/easytier-ohrs/src/config/services/mod.rs

@@ -0,0 +1,2 @@
+pub(crate) mod schema_service;
+pub(crate) mod share_link_service;

easytier-contrib/easytier-ohrs/src/config/services/schema_service.rs

@@ -0,0 +1,414 @@
+use easytier::proto::ALL_DESCRIPTOR_BYTES;
+use napi_derive_ohos::napi;
+use once_cell::sync::Lazy;
+use prost_reflect::{Cardinality, DescriptorPool, FieldDescriptor, Kind, MessageDescriptor};
+use serde::Serialize;
+
+#[derive(Debug, Clone, Serialize)]
+#[napi(object)]
+pub struct FieldOption {
+    pub label: String,
+    pub value: String,
+}
+
+#[derive(Debug, Clone, Serialize)]
+#[napi(object)]
+pub struct ValidationRule {
+    pub rule_type: String,
+    pub arg: String,
+    pub message: String,
+}
+
+#[derive(Debug, Clone, Serialize)]
+#[napi(object)]
+pub struct NetworkConfigSchema {
+    pub node_kind: String,
+    pub name: String,
+    pub field_number: i32,
+    pub type_name: Option<String>,
+    pub semantic_type: Option<String>,
+    pub value_kind: String,
+    pub is_list: bool,
+    pub required: bool,
+    pub default_value_text: Option<String>,
+    pub enum_options: Vec<FieldOption>,
+    pub validations: Vec<ValidationRule>,
+    pub children: Vec<NetworkConfigSchema>,
+    pub definitions: Vec<NetworkConfigSchema>,
+}
+
+#[derive(Debug, Clone, Serialize)]
+#[napi(object)]
+pub struct ConfigFieldMapping {
+    pub field_name: String,
+    pub field_number: i32,
+}
+
+static DESCRIPTOR_POOL: Lazy<DescriptorPool> = Lazy::new(|| {
+    DescriptorPool::decode(ALL_DESCRIPTOR_BYTES)
+        .expect("easytier descriptor pool should decode from embedded protobuf descriptors")
+});
+
+const NETWORK_CONFIG_MESSAGE_NAME: &str = "api.manage.NetworkConfig";
+
+fn descriptor_pool() -> &'static DescriptorPool {
+    &DESCRIPTOR_POOL
+}
+
+fn network_config_descriptor() -> MessageDescriptor {
+    descriptor_pool()
+        .get_message_by_name(NETWORK_CONFIG_MESSAGE_NAME)
+        .expect("api.manage.NetworkConfig descriptor should exist")
+}
+
+fn field_default_value_text(field: &FieldDescriptor) -> Option<String> {
+    if field.is_list() || field.is_map() {
+        return Some("[]".to_string());
+    }
+
+    match field.kind() {
+        Kind::Bool => Some("false".to_string()),
+        Kind::String => Some("\"\"".to_string()),
+        Kind::Bytes => Some("\"\"".to_string()),
+        Kind::Int32
+        | Kind::Sint32
+        | Kind::Sfixed32
+        | Kind::Int64
+        | Kind::Sint64
+        | Kind::Sfixed64
+        | Kind::Uint32
+        | Kind::Fixed32
+        | Kind::Uint64
+        | Kind::Fixed64
+        | Kind::Float
+        | Kind::Double => Some("0".to_string()),
+        Kind::Enum(enum_desc) => enum_desc
+            .get_value(0)
+            .map(|value| value.number().to_string()),
+        Kind::Message(_) => None,
+    }
+}
+
+fn field_type_name(field: &FieldDescriptor) -> Option<String> {
+    match field.kind() {
+        Kind::Enum(enum_desc) => Some(enum_desc.full_name().to_string()),
+        Kind::Message(message_desc) => Some(message_desc.full_name().to_string()),
+        _ => None,
+    }
+}
+
+fn field_semantic_type(field: &FieldDescriptor) -> Option<String> {
+    match field.name() {
+        "virtual_ipv4" => Some("cidr_ip".to_string()),
+        "network_length" => Some("cidr_mask".to_string()),
+        "peer_urls" => Some("peer[]".to_string()),
+        "proxy_cidrs" => Some("cidr[]".to_string()),
+        "listener_urls" => Some("listener[]".to_string()),
+        "routes" => Some("route[]".to_string()),
+        "exit_nodes" => Some("ip[]".to_string()),
+        "relay_network_whitelist" => Some("network_name[]".to_string()),
+        "mapped_listeners" => Some("mapped_listener[]".to_string()),
+        "port_forwards" => Some("port_forward[]".to_string()),
+        _ => None,
+    }
+}
+
+fn enum_options(kind: Kind) -> Vec<FieldOption> {
+    match kind {
+        Kind::Enum(enum_desc) => enum_desc
+            .values()
+            .map(|value| FieldOption {
+                label: value.name().to_string(),
+                value: value.number().to_string(),
+            })
+            .collect(),
+        _ => Vec::new(),
+    }
+}
+
+fn should_expose_field(field: &FieldDescriptor) -> bool {
+    match field.containing_oneof() {
+        Some(_) => field
+            .field_descriptor_proto()
+            .proto3_optional
+            .unwrap_or(false),
+        None => true,
+    }
+}
+
+fn build_validations(field: &FieldDescriptor) -> Vec<ValidationRule> {
+    if field.cardinality() == Cardinality::Required {
+        return vec![ValidationRule {
+            rule_type: "required".to_string(),
+            arg: String::new(),
+            message: format!("{} is required", field.name()),
+        }];
+    }
+
+    Vec::new()
+}
+
+fn kind_to_value_kind(field: &FieldDescriptor) -> String {
+    if field.is_map() {
+        return "object".to_string();
+    }
+
+    match field.kind() {
+        Kind::Bool => "boolean".to_string(),
+        Kind::String | Kind::Bytes => "string".to_string(),
+        Kind::Int32
+        | Kind::Sint32
+        | Kind::Sfixed32
+        | Kind::Int64
+        | Kind::Sint64
+        | Kind::Sfixed64
+        | Kind::Uint32
+        | Kind::Fixed32
+        | Kind::Uint64
+        | Kind::Fixed64
+        | Kind::Float
+        | Kind::Double => "number".to_string(),
+        Kind::Enum(_) => "enum".to_string(),
+        Kind::Message(_) => "object".to_string(),
+    }
+}
+
+fn build_node(
+    node_kind: &str,
+    name: String,
+    field_number: i32,
+    type_name: Option<String>,
+    semantic_type: Option<String>,
+    value_kind: String,
+    is_list: bool,
+    required: bool,
+    default_value_text: Option<String>,
+    enum_options: Vec<FieldOption>,
+    validations: Vec<ValidationRule>,
+    children: Vec<NetworkConfigSchema>,
+    definitions: Vec<NetworkConfigSchema>,
+) -> NetworkConfigSchema {
+    NetworkConfigSchema {
+        node_kind: node_kind.to_string(),
+        name,
+        field_number,
+        type_name,
+        semantic_type,
+        value_kind,
+        is_list,
+        required,
+        default_value_text,
+        enum_options,
+        validations,
+        children,
+        definitions,
+    }
+}
+
+fn build_map_entry_node(message_desc: &MessageDescriptor) -> NetworkConfigSchema {
+    let key_field = message_desc.map_entry_key_field();
+    let value_field = message_desc.map_entry_value_field();
+
+    build_node(
+        "object",
+        message_desc.name().to_string(),
+        0,
+        Some(message_desc.full_name().to_string()),
+        None,
+        "object".to_string(),
+        false,
+        true,
+        None,
+        Vec::new(),
+        Vec::new(),
+        vec![
+            build_schema_field_node(&key_field),
+            build_schema_field_node(&value_field),
+        ],
+        Vec::new(),
+    )
+}
+
+fn field_children(field: &FieldDescriptor) -> Vec<NetworkConfigSchema> {
+    if field.is_map() {
+        if let Kind::Message(message_desc) = field.kind() {
+            return vec![build_map_entry_node(&message_desc)];
+        }
+    }
+
+    match field.kind() {
+        Kind::Message(message_desc) => build_message_children(&message_desc),
+        _ => Vec::new(),
+    }
+}
+
+fn build_message_children(message_desc: &MessageDescriptor) -> Vec<NetworkConfigSchema> {
+    message_desc
+        .fields()
+        .filter(should_expose_field)
+        .map(|field| build_schema_field_node(&field))
+        .collect()
+}
+
+fn build_schema_field_node(field: &FieldDescriptor) -> NetworkConfigSchema {
+    build_node(
+        "field",
+        field.name().to_string(),
+        field.number() as i32,
+        field_type_name(field),
+        field_semantic_type(field),
+        kind_to_value_kind(field),
+        field.is_list() || field.is_map(),
+        field.cardinality() == Cardinality::Required,
+        field_default_value_text(field),
+        enum_options(field.kind()),
+        build_validations(field),
+        field_children(field),
+        Vec::new(),
+    )
+}
+
+fn collect_definitions() -> Vec<NetworkConfigSchema> {
+    let mut definitions = Vec::new();
+
+    for message_desc in descriptor_pool().all_messages() {
+        let full_name = message_desc.full_name();
+        if full_name == NETWORK_CONFIG_MESSAGE_NAME || message_desc.is_map_entry() {
+            continue;
+        }
+
+        definitions.push(build_node(
+            "object",
+            full_name.to_string(),
+            0,
+            Some(full_name.to_string()),
+            None,
+            "object".to_string(),
+            false,
+            true,
+            None,
+            Vec::new(),
+            Vec::new(),
+            build_message_children(&message_desc),
+            Vec::new(),
+        ));
+    }
+
+    for enum_desc in descriptor_pool().all_enums() {
+        definitions.push(build_node(
+            "enum",
+            enum_desc.full_name().to_string(),
+            0,
+            Some(enum_desc.full_name().to_string()),
+            None,
+            "enum".to_string(),
+            false,
+            false,
+            None,
+            enum_options(Kind::Enum(enum_desc.clone())),
+            Vec::new(),
+            Vec::new(),
+            Vec::new(),
+        ));
+    }
+
+    definitions.sort_by(|a, b| a.name.cmp(&b.name));
+    definitions
+}
+
+fn build_network_config_schema() -> NetworkConfigSchema {
+    let network_config = network_config_descriptor();
+    build_node(
+        "schema",
+        network_config.name().to_string(),
+        0,
+        Some(network_config.full_name().to_string()),
+        None,
+        "object".to_string(),
+        false,
+        true,
+        None,
+        Vec::new(),
+        Vec::new(),
+        build_message_children(&network_config),
+        collect_definitions(),
+    )
+}
+
+fn build_network_config_field_mappings() -> Vec<ConfigFieldMapping> {
+    network_config_descriptor()
+        .fields()
+        .filter(should_expose_field)
+        .map(|field| ConfigFieldMapping {
+            field_name: field.name().to_string(),
+            field_number: field.number() as i32,
+        })
+        .collect()
+}
+
+pub fn get_network_config_schema() -> NetworkConfigSchema {
+    build_network_config_schema()
+}
+
+pub fn get_network_config_field_mappings() -> Vec<ConfigFieldMapping> {
+    build_network_config_field_mappings()
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn schema_is_exposed_as_single_tree_type() {
+        let schema = get_network_config_schema();
+        assert_eq!(schema.node_kind, "schema");
+        assert_eq!(schema.name, "NetworkConfig");
+        assert_eq!(
+            schema.type_name.as_deref(),
+            Some("api.manage.NetworkConfig")
+        );
+
+        let virtual_ipv4 = schema
+            .children
+            .iter()
+            .find(|field| field.name == "virtual_ipv4")
+            .expect("virtual_ipv4 field");
+        assert_eq!(virtual_ipv4.semantic_type.as_deref(), Some("cidr_ip"));
+
+        let secure_mode = schema
+            .children
+            .iter()
+            .find(|field| field.name == "secure_mode")
+            .expect("secure_mode field");
+        assert!(
+            secure_mode
+                .children
+                .iter()
+                .any(|field| field.name == "enabled")
+        );
+
+        let secure_mode_definition = schema
+            .definitions
+            .iter()
+            .find(|definition| definition.name == "common.SecureModeConfig")
+            .expect("secure mode definition");
+        assert!(
+            secure_mode_definition
+                .children
+                .iter()
+                .any(|field| field.name == "local_private_key")
+        );
+
+        let networking_method_definition = schema
+            .definitions
+            .iter()
+            .find(|definition| definition.name == "api.manage.NetworkingMethod")
+            .expect("networking method enum definition");
+        assert!(
+            networking_method_definition
+                .enum_options
+                .iter()
+                .any(|option| option.label == "PublicServer")
+        );
+    }
+}

easytier-contrib/easytier-ohrs/src/config/services/share_link_service.rs

@@ -0,0 +1,197 @@
+use crate::config::repository::{get_config_record, save_config_record};
+use crate::config::services::schema_service::get_network_config_field_mappings;
+use crate::config::types::stored_config::SharedConfigLinkPayload;
+use base64::{Engine as _, engine::general_purpose::URL_SAFE_NO_PAD};
+use easytier::proto::api::manage::NetworkConfig;
+use flate2::{Compression, read::ZlibDecoder, write::ZlibEncoder};
+use gethostname::gethostname;
+use std::collections::HashMap;
+use std::io::{Read, Write};
+use url::Url;
+use uuid::Uuid;
+
+const SHARE_LINK_HOST: &str = "easytier.cn";
+const SHARE_LINK_PATH: &str = "/comp_cfg";
+
+fn field_name_to_id_map() -> HashMap<String, String> {
+    get_network_config_field_mappings()
+        .into_iter()
+        .map(|mapping| (mapping.field_name, mapping.field_number.to_string()))
+        .collect()
+}
+
+fn field_id_to_name_map() -> HashMap<String, String> {
+    get_network_config_field_mappings()
+        .into_iter()
+        .map(|mapping| (mapping.field_number.to_string(), mapping.field_name))
+        .collect()
+}
+
+fn prune_empty(value: &serde_json::Value) -> Option<serde_json::Value> {
+    match value {
+        serde_json::Value::Null => None,
+        serde_json::Value::Array(values) if values.is_empty() => None,
+        _ => Some(value.clone()),
+    }
+}
+
+fn map_config_json(config: &NetworkConfig) -> Result<String, String> {
+    let field_name_to_id = field_name_to_id_map();
+    let raw = serde_json::to_value(config).map_err(|err| err.to_string())?;
+    let mut mapped = serde_json::Map::new();
+
+    for (key, value) in raw.as_object().cloned().unwrap_or_default() {
+        let Some(value) = prune_empty(&value) else {
+            continue;
+        };
+        let mapped_key = field_name_to_id.get(&key).cloned().unwrap_or(key);
+        mapped.insert(mapped_key, value);
+    }
+
+    serde_json::to_string(&mapped).map_err(|err| err.to_string())
+}
+
+fn unmap_config_json(raw: &str) -> Result<NetworkConfig, String> {
+    let field_id_to_name = field_id_to_name_map();
+    let value = serde_json::from_str::<serde_json::Value>(raw).map_err(|err| err.to_string())?;
+    let mut mapped = serde_json::Map::new();
+    for (key, value) in value.as_object().cloned().unwrap_or_default() {
+        let field_name = field_id_to_name.get(&key).cloned().unwrap_or(key);
+        mapped.insert(field_name, value);
+    }
+    serde_json::from_value(serde_json::Value::Object(mapped)).map_err(|err| err.to_string())
+}
+
+fn compress_to_base64url(raw: &str) -> Result<String, String> {
+    let mut encoder = ZlibEncoder::new(Vec::new(), Compression::best());
+    encoder
+        .write_all(raw.as_bytes())
+        .map_err(|err| err.to_string())?;
+    let compressed = encoder.finish().map_err(|err| err.to_string())?;
+    Ok(URL_SAFE_NO_PAD.encode(compressed))
+}
+
+fn decompress_from_base64url(raw: &str) -> Result<String, String> {
+    let compressed = URL_SAFE_NO_PAD.decode(raw).map_err(|err| err.to_string())?;
+    let mut decoder = ZlibDecoder::new(compressed.as_slice());
+    let mut out = String::new();
+    decoder
+        .read_to_string(&mut out)
+        .map_err(|err| err.to_string())?;
+    Ok(out)
+}
+
+pub fn build_config_share_link(
+    config_id: &str,
+    display_name: Option<String>,
+    only_start: bool,
+) -> Option<String> {
+    let record = get_config_record(config_id)?;
+    let config = serde_json::from_str::<NetworkConfig>(&record.config_json).ok()?;
+    let mapped_json = map_config_json(&config).ok()?;
+    let compressed = compress_to_base64url(&mapped_json).ok()?;
+    let final_name = display_name
+        .or(Some(record.meta.display_name))
+        .filter(|name| !name.is_empty());
+
+    let mut url = Url::parse(&format!("https://{SHARE_LINK_HOST}{SHARE_LINK_PATH}")).ok()?;
+    url.query_pairs_mut().append_pair("cfg", &compressed);
+    if let Some(name) = final_name {
+        url.query_pairs_mut().append_pair("name", &name);
+    }
+    if only_start {
+        url.query_pairs_mut().append_pair("only_start", "true");
+    }
+    Some(url.to_string())
+}
+
+pub fn parse_config_share_link(share_link: &str) -> Option<SharedConfigLinkPayload> {
+    let url = Url::parse(share_link).ok()?;
+    if url.host_str()? != SHARE_LINK_HOST || url.path() != SHARE_LINK_PATH {
+        return None;
+    }
+
+    let cfg = url
+        .query_pairs()
+        .find(|(key, _)| key == "cfg")?
+        .1
+        .to_string();
+    let mapped_json = decompress_from_base64url(&cfg).ok()?;
+    let mut config = unmap_config_json(&mapped_json).ok()?;
+    config.instance_id = Some(Uuid::new_v4().to_string());
+    let hostname = gethostname().to_string_lossy().to_string();
+    if !hostname.is_empty() {
+        config.hostname = Some(hostname);
+    }
+
+    let config_json = serde_json::to_string(&config).ok()?;
+    let display_name = url
+        .query_pairs()
+        .find(|(key, _)| key == "name")
+        .map(|(_, value)| value.to_string())
+        .filter(|name| !name.is_empty());
+    let only_start = url
+        .query_pairs()
+        .find(|(key, _)| key == "only_start")
+        .map(|(_, value)| value == "true")
+        .unwrap_or(false);
+
+    Some(SharedConfigLinkPayload {
+        config_json,
+        display_name,
+        only_start,
+    })
+}
+
+pub fn import_config_share_link(
+    share_link: &str,
+    display_name_override: Option<String>,
+) -> Option<String> {
+    let payload = parse_config_share_link(share_link)?;
+    let config = serde_json::from_str::<NetworkConfig>(&payload.config_json).ok()?;
+    let config_id = config.instance_id.clone()?;
+    let display_name = display_name_override
+        .filter(|name| !name.is_empty())
+        .or(payload.display_name)
+        .unwrap_or_else(|| config_id.clone());
+
+    save_config_record(config_id.clone(), display_name, payload.config_json)?;
+    Some(config_id)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::config_repo::{create_config_record, init_config_store};
+    use std::time::{SystemTime, UNIX_EPOCH};
+
+    fn test_root() -> String {
+        let unique = SystemTime::now()
+            .duration_since(UNIX_EPOCH)
+            .unwrap()
+            .as_nanos();
+        std::env::temp_dir()
+            .join(format!("easytier_ohrs_share_test_{unique}"))
+            .to_string_lossy()
+            .into_owned()
+    }
+
+    #[test]
+    fn share_link_roundtrip_works() {
+        assert!(init_config_store(test_root()));
+        create_config_record("cfg-share".to_string(), "share-demo".to_string())
+            .expect("create config");
+
+        let link = build_config_share_link("cfg-share", None, true).expect("share link");
+        let payload = parse_config_share_link(&link).expect("parse link");
+        let config =
+            serde_json::from_str::<NetworkConfig>(&payload.config_json).expect("config json");
+
+        assert!(payload.only_start);
+        assert_eq!(payload.display_name.as_deref(), Some("share-demo"));
+        assert_ne!(config.instance_id.as_deref(), Some("cfg-share"));
+
+        let imported_id = import_config_share_link(&link, None).expect("import link");
+        assert_ne!(imported_id, "cfg-share");
+    }
+}

easytier-contrib/easytier-ohrs/src/config/storage/config_meta.rs

@@ -0,0 +1,333 @@
+use crate::config::types::stored_config::{StoredConfigList, StoredConfigMeta};
+use ohos_hilog_binding::{hilog_debug, hilog_error};
+use rusqlite::{Connection, OptionalExtension, params};
+use std::path::PathBuf;
+use std::sync::Mutex;
+use std::time::{SystemTime, UNIX_EPOCH};
+
+static CONFIG_DB_PATH: Mutex<Option<PathBuf>> = Mutex::new(None);
+const CONFIG_DB_FILE_NAME: &str = "easytier-config-store.db";
+
+#[derive(Debug, Clone)]
+struct StoredConfigMetaRecord {
+    config_id: String,
+    display_name: String,
+    created_at: String,
+    updated_at: String,
+    favorite: bool,
+    temporary: bool,
+}
+
+pub(crate) fn now_ts_string() -> String {
+    SystemTime::now()
+        .duration_since(UNIX_EPOCH)
+        .map(|d| d.as_secs().to_string())
+        .unwrap_or_else(|_| "0".to_string())
+}
+
+fn db_file_path() -> Option<PathBuf> {
+    CONFIG_DB_PATH
+        .lock()
+        .ok()
+        .and_then(|guard| guard.as_ref().cloned())
+}
+
+fn init_schema(conn: &Connection) -> rusqlite::Result<()> {
+    conn.execute_batch(
+        "PRAGMA foreign_keys = ON;
+         CREATE TABLE IF NOT EXISTS stored_configs (
+             config_id TEXT PRIMARY KEY,
+             display_name TEXT NOT NULL,
+             created_at TEXT NOT NULL,
+             updated_at TEXT NOT NULL,
+             favorite INTEGER NOT NULL DEFAULT 0,
+             temporary INTEGER NOT NULL DEFAULT 0
+         );
+         CREATE TABLE IF NOT EXISTS stored_config_fields (
+             config_id TEXT NOT NULL,
+             field_name TEXT NOT NULL,
+             field_json TEXT NOT NULL,
+             updated_at TEXT NOT NULL,
+             PRIMARY KEY (config_id, field_name),
+             FOREIGN KEY (config_id) REFERENCES stored_configs(config_id) ON DELETE CASCADE
+         );
+         CREATE INDEX IF NOT EXISTS idx_stored_config_fields_config_id
+             ON stored_config_fields(config_id);",
+    )
+}
+
+pub(crate) fn open_db() -> Option<Connection> {
+    let path = db_file_path()?;
+    let conn = match Connection::open(&path) {
+        Ok(conn) => conn,
+        Err(e) => {
+            hilog_error!("[Rust] failed to open config db {}: {}", path.display(), e);
+            return None;
+        }
+    };
+
+    if let Err(e) = init_schema(&conn) {
+        hilog_error!(
+            "[Rust] failed to initialize config db {}: {}",
+            path.display(),
+            e
+        );
+        return None;
+    }
+
+    Some(conn)
+}
+
+fn row_to_meta(row: &rusqlite::Row<'_>) -> rusqlite::Result<StoredConfigMetaRecord> {
+    Ok(StoredConfigMetaRecord {
+        config_id: row.get(0)?,
+        display_name: row.get(1)?,
+        created_at: row.get(2)?,
+        updated_at: row.get(3)?,
+        favorite: row.get::<_, i64>(4)? != 0,
+        temporary: row.get::<_, i64>(5)? != 0,
+    })
+}
+
+fn load_meta_record(conn: &Connection, config_id: &str) -> Option<StoredConfigMetaRecord> {
+    conn.query_row(
+        "SELECT config_id, display_name, created_at, updated_at, favorite, temporary
+         FROM stored_configs WHERE config_id = ?1",
+        params![config_id],
+        row_to_meta,
+    )
+    .optional()
+    .ok()
+    .flatten()
+}
+
+fn to_meta(record: StoredConfigMetaRecord) -> StoredConfigMeta {
+    StoredConfigMeta {
+        config_id: record.config_id,
+        display_name: record.display_name,
+        created_at: record.created_at,
+        updated_at: record.updated_at,
+        favorite: record.favorite,
+        temporary: record.temporary,
+    }
+}
+
+pub fn init_config_meta_store(root_dir: String) -> bool {
+    let root = PathBuf::from(root_dir);
+    if let Err(e) = std::fs::create_dir_all(&root) {
+        hilog_error!(
+            "[Rust] failed to create config db dir {}: {}",
+            root.display(),
+            e
+        );
+        return false;
+    }
+
+    let db_path = root.join(CONFIG_DB_FILE_NAME);
+    match CONFIG_DB_PATH.lock() {
+        Ok(mut guard) => {
+            *guard = Some(db_path.clone());
+        }
+        Err(e) => {
+            hilog_error!("[Rust] failed to lock config db path: {}", e);
+            return false;
+        }
+    }
+
+    if open_db().is_none() {
+        return false;
+    }
+
+    hilog_debug!("[Rust] initialized config db at {}", db_path.display());
+    true
+}
+
+pub fn list_config_meta_entries() -> StoredConfigList {
+    let Some(conn) = open_db() else {
+        return StoredConfigList { configs: vec![] };
+    };
+
+    let mut stmt = match conn.prepare(
+        "SELECT config_id, display_name, created_at, updated_at, favorite, temporary
+         FROM stored_configs
+         ORDER BY updated_at DESC, display_name ASC",
+    ) {
+        Ok(stmt) => stmt,
+        Err(e) => {
+            hilog_error!("[Rust] failed to prepare list meta query: {}", e);
+            return StoredConfigList { configs: vec![] };
+        }
+    };
+
+    let rows = match stmt.query_map([], row_to_meta) {
+        Ok(rows) => rows,
+        Err(e) => {
+            hilog_error!("[Rust] failed to list config meta rows: {}", e);
+            return StoredConfigList { configs: vec![] };
+        }
+    };
+
+    let configs = rows.filter_map(Result::ok).map(to_meta).collect();
+    StoredConfigList { configs }
+}
+
+pub fn get_config_display_name(config_id: &str) -> Option<String> {
+    let conn = open_db()?;
+    load_meta_record(&conn, config_id).map(|record| record.display_name)
+}
+
+pub fn get_config_meta(config_id: &str) -> Option<StoredConfigMeta> {
+    let conn = open_db()?;
+    load_meta_record(&conn, config_id).map(to_meta)
+}
+
+pub fn upsert_config_meta(
+    config_id: String,
+    display_name: String,
+    favorite: bool,
+    temporary: bool,
+) -> StoredConfigMeta {
+    let now = now_ts_string();
+    let Some(conn) = open_db() else {
+        return StoredConfigMeta {
+            config_id,
+            display_name,
+            created_at: now.clone(),
+            updated_at: now,
+            favorite,
+            temporary,
+        };
+    };
+
+    let created_at = load_meta_record(&conn, &config_id)
+        .map(|record| record.created_at)
+        .unwrap_or_else(|| now.clone());
+
+    if let Err(e) = conn.execute(
+        "INSERT INTO stored_configs (
+             config_id, display_name, created_at, updated_at, favorite, temporary
+         ) VALUES (?1, ?2, ?3, ?4, ?5, ?6)
+         ON CONFLICT(config_id) DO UPDATE SET
+             display_name = excluded.display_name,
+             updated_at = excluded.updated_at,
+             favorite = excluded.favorite,
+             temporary = excluded.temporary",
+        params![
+            config_id,
+            display_name,
+            created_at,
+            now,
+            if favorite { 1 } else { 0 },
+            if temporary { 1 } else { 0 }
+        ],
+    ) {
+        hilog_error!("[Rust] failed to upsert config meta: {}", e);
+    }
+
+    get_config_meta(&config_id).unwrap_or(StoredConfigMeta {
+        config_id,
+        display_name,
+        created_at,
+        updated_at: now,
+        favorite,
+        temporary,
+    })
+}
+
+pub(crate) fn upsert_config_meta_in_tx(
+    tx: &rusqlite::Transaction<'_>,
+    config_id: String,
+    display_name: String,
+    favorite: bool,
+    temporary: bool,
+) -> Option<StoredConfigMeta> {
+    let now = now_ts_string();
+    let created_at = tx
+        .query_row(
+            "SELECT config_id, display_name, created_at, updated_at, favorite, temporary
+             FROM stored_configs WHERE config_id = ?1",
+            params![config_id],
+            row_to_meta,
+        )
+        .optional()
+        .ok()
+        .flatten()
+        .map(|record| record.created_at)
+        .unwrap_or_else(|| now.clone());
+
+    tx.execute(
+        "INSERT INTO stored_configs (
+             config_id, display_name, created_at, updated_at, favorite, temporary
+         ) VALUES (?1, ?2, ?3, ?4, ?5, ?6)
+         ON CONFLICT(config_id) DO UPDATE SET
+             display_name = excluded.display_name,
+             updated_at = excluded.updated_at,
+             favorite = excluded.favorite,
+             temporary = excluded.temporary",
+        params![
+            config_id,
+            display_name,
+            created_at,
+            now,
+            if favorite { 1 } else { 0 },
+            if temporary { 1 } else { 0 }
+        ],
+    )
+    .ok()?;
+
+    tx.query_row(
+        "SELECT config_id, display_name, created_at, updated_at, favorite, temporary
+         FROM stored_configs WHERE config_id = ?1",
+        params![config_id],
+        row_to_meta,
+    )
+    .optional()
+    .ok()
+    .flatten()
+    .map(to_meta)
+    .or(Some(StoredConfigMeta {
+        config_id,
+        display_name,
+        created_at,
+        updated_at: now,
+        favorite,
+        temporary,
+    }))
+}
+
+pub fn set_config_display_name(
+    config_id: String,
+    display_name: String,
+) -> Option<StoredConfigMeta> {
+    let conn = open_db()?;
+    let mut record = load_meta_record(&conn, &config_id)?;
+    record.display_name = display_name;
+    record.updated_at = now_ts_string();
+
+    conn.execute(
+        "UPDATE stored_configs
+         SET display_name = ?2, updated_at = ?3
+         WHERE config_id = ?1",
+        params![config_id, record.display_name, record.updated_at],
+    )
+    .ok()?;
+
+    Some(to_meta(record))
+}
+
+pub fn delete_config_meta(config_id: &str) -> bool {
+    let Some(conn) = open_db() else {
+        return false;
+    };
+
+    match conn.execute(
+        "DELETE FROM stored_configs WHERE config_id = ?1",
+        params![config_id],
+    ) {
+        Ok(rows) => rows > 0,
+        Err(e) => {
+            hilog_error!("[Rust] failed to delete config meta {}: {}", config_id, e);
+            false
+        }
+    }
+}

easytier-contrib/easytier-ohrs/src/config/storage/mod.rs

@@ -0,0 +1 @@
+pub(crate) mod config_meta;

easytier-contrib/easytier-ohrs/src/config/types/mod.rs

@@ -0,0 +1 @@
+pub(crate) mod stored_config;

easytier-contrib/easytier-ohrs/src/config/types/stored_config.rs

@@ -0,0 +1,68 @@
+use napi_derive_ohos::napi;
+use serde::Serialize;
+
+#[derive(Debug, Clone, Serialize)]
+#[serde(rename_all = "camelCase")]
+#[napi(object)]
+pub struct StoredConfigMeta {
+    pub config_id: String,
+    pub display_name: String,
+    pub created_at: String,
+    pub updated_at: String,
+    pub favorite: bool,
+    pub temporary: bool,
+}
+
+#[derive(Debug, Clone, Serialize)]
+#[serde(rename_all = "camelCase")]
+#[napi(object)]
+pub struct StoredConfigRecord {
+    pub meta: StoredConfigMeta,
+    pub config_json: String,
+}
+
+#[derive(Debug, Clone, Serialize)]
+#[serde(rename_all = "camelCase")]
+#[napi(object)]
+pub struct StoredConfigList {
+    pub configs: Vec<StoredConfigMeta>,
+}
+
+#[derive(Debug, Clone, Serialize)]
+#[serde(rename_all = "camelCase")]
+#[napi(object)]
+pub struct ExportTomlResult {
+    pub toml_text: String,
+}
+
+#[derive(Debug, Clone, Serialize)]
+#[serde(rename_all = "camelCase")]
+#[napi(object)]
+pub struct StoredConfigSummary {
+    pub config_id: String,
+    pub display_name: String,
+}
+
+#[derive(Debug, Clone, Serialize)]
+#[serde(rename_all = "camelCase")]
+#[napi(object)]
+pub struct SharedConfigLinkPayload {
+    pub config_json: String,
+    pub display_name: Option<String>,
+    pub only_start: bool,
+}
+
+#[derive(Debug, Clone, Serialize)]
+#[serde(rename_all = "camelCase")]
+#[napi(object)]
+pub struct LocalSocketSyncMessage {
+    pub message_type: String,
+    pub payload_json: String,
+}
+
+#[derive(Debug, Clone, Serialize)]
+#[napi(object)]
+pub struct KeyValuePair {
+    pub key: String,
+    pub value: String,
+}

easytier-contrib/easytier-ohrs/src/config_repo.rs

@@ -0,0 +1,349 @@
+use super::{field_store, import_export, legacy_migration, validation};
+use crate::config::storage::config_meta::{
+    delete_config_meta, get_config_meta, init_config_meta_store, list_config_meta_entries, open_db,
+    upsert_config_meta_in_tx,
+};
+use crate::config::types::stored_config::{ExportTomlResult, StoredConfigRecord};
+use easytier::common::config::ConfigLoader;
+use easytier::proto::api::manage::NetworkConfig;
+use ohos_hilog_binding::{hilog_debug, hilog_error};
+use rusqlite::params;
+use serde_json::Value;
+use std::path::PathBuf;
+use std::sync::Mutex;
+
+static CONFIG_ROOT_DIR: Mutex<Option<PathBuf>> = Mutex::new(None);
+pub(crate) const CONFIG_DIR_NAME: &str = "easytier-configs";
+pub(crate) const KERNEL_SOCKET_FILE_NAME: &str = "easytier-kernel.sock";
+
+pub(crate) fn config_root_dir() -> Option<PathBuf> {
+    CONFIG_ROOT_DIR
+        .lock()
+        .ok()
+        .and_then(|guard| guard.as_ref().cloned())
+}
+
+pub(crate) fn kernel_socket_path() -> Option<PathBuf> {
+    config_root_dir().map(|root| root.join(KERNEL_SOCKET_FILE_NAME))
+}
+
+pub(crate) fn legacy_config_file_path(config_id: &str) -> Option<PathBuf> {
+    legacy_migration::legacy_config_file_path(&config_root_dir(), CONFIG_DIR_NAME, config_id)
+}
+
+pub fn init_config_store(root_dir: String) -> bool {
+    let root = PathBuf::from(root_dir);
+    let configs_dir = root.join(CONFIG_DIR_NAME);
+    if let Err(e) = std::fs::create_dir_all(&configs_dir) {
+        hilog_error!(
+            "[Rust] failed to create config dir {}: {}",
+            configs_dir.display(),
+            e
+        );
+        return false;
+    }
+
+    match CONFIG_ROOT_DIR.lock() {
+        Ok(mut guard) => {
+            *guard = Some(root.clone());
+        }
+        Err(e) => {
+            hilog_error!("[Rust] failed to lock config root dir: {}", e);
+            return false;
+        }
+    }
+
+    if !init_config_meta_store(root.to_string_lossy().into_owned()) {
+        return false;
+    }
+
+    hilog_debug!(
+        "[Rust] initialized config repo at {}",
+        configs_dir.display()
+    );
+    true
+}
+
+fn migrate_legacy_file_if_needed(config_id: &str) -> Option<()> {
+    legacy_migration::migrate_legacy_file_if_needed(
+        &config_root_dir(),
+        CONFIG_DIR_NAME,
+        config_id,
+        save_config_record,
+    )
+}
+
+pub fn save_config_record(
+    config_id: String,
+    display_name: String,
+    config_json: String,
+) -> Option<StoredConfigRecord> {
+    let config = match validation::validate_config_json(&config_json, config_id.clone()) {
+        Ok(config) => config,
+        Err(e) => {
+            hilog_error!("[Rust] save_config_record failed {}", e);
+            return None;
+        }
+    };
+
+    let normalized_json = match serde_json::to_string(&config) {
+        Ok(raw) => raw,
+        Err(e) => {
+            hilog_error!(
+                "[Rust] failed to serialize normalized config {}: {}",
+                config_id,
+                e
+            );
+            return None;
+        }
+    };
+
+    let fields = match validation::config_to_top_level_map(&config) {
+        Some(fields) => fields,
+        None => return None,
+    };
+
+    let conn = open_db()?;
+    let tx = conn.unchecked_transaction().ok()?;
+    let existing_meta = get_config_meta(&config_id);
+    let favorite = existing_meta
+        .as_ref()
+        .map(|meta| meta.favorite)
+        .unwrap_or(false);
+    let temporary = existing_meta
+        .as_ref()
+        .map(|meta| meta.temporary)
+        .unwrap_or(false);
+    let meta = upsert_config_meta_in_tx(&tx, config_id.clone(), display_name, favorite, temporary)?;
+
+    field_store::replace_config_fields(&tx, &config_id, fields)?;
+
+    tx.commit().ok()?;
+
+    if let Some(legacy_path) = legacy_config_file_path(&config_id) {
+        if legacy_path.exists() {
+            let _ = std::fs::remove_file(legacy_path);
+        }
+    }
+
+    Some(StoredConfigRecord {
+        meta,
+        config_json: normalized_json,
+    })
+}
+
+pub fn load_config_json(config_id: &str) -> Option<String> {
+    migrate_legacy_file_if_needed(config_id)?;
+    let object = field_store::load_config_map_from_db(config_id)?;
+    serde_json::to_string(&Value::Object(object)).ok()
+}
+
+pub fn get_config_record(config_id: &str) -> Option<StoredConfigRecord> {
+    let config_json = load_config_json(config_id)?;
+    let meta = get_config_meta(config_id)?;
+    Some(StoredConfigRecord { meta, config_json })
+}
+
+pub fn get_config_field_value(config_id: &str, field: &str) -> Option<String> {
+    migrate_legacy_file_if_needed(config_id)?;
+    let conn = open_db()?;
+    conn.query_row(
+        "SELECT field_json FROM stored_config_fields
+         WHERE config_id = ?1 AND field_name = ?2",
+        params![config_id, field],
+        |row| row.get::<_, String>(0),
+    )
+    .ok()
+}
+
+pub fn set_config_field_value(config_id: &str, field: &str, json_value: &str) -> bool {
+    if field.contains('.') {
+        return false;
+    }
+
+    let raw = match load_config_json(config_id) {
+        Some(raw) => raw,
+        None => return false,
+    };
+    let mut value = match serde_json::from_str::<Value>(&raw) {
+        Ok(value) => value,
+        Err(_) => return false,
+    };
+    let new_field_value = match serde_json::from_str::<Value>(json_value) {
+        Ok(value) => value,
+        Err(_) => return false,
+    };
+    let object = match value.as_object_mut() {
+        Some(object) => object,
+        None => return false,
+    };
+    object.insert(field.to_string(), new_field_value);
+
+    let normalized = match serde_json::to_string(&value) {
+        Ok(raw) => raw,
+        Err(_) => return false,
+    };
+
+    let display_name = get_config_meta(config_id)
+        .map(|meta| meta.display_name)
+        .unwrap_or_else(|| config_id.to_string());
+
+    save_config_record(config_id.to_string(), display_name, normalized).is_some()
+}
+
+pub fn get_display_name(config_id: &str) -> Option<String> {
+    get_config_meta(config_id).map(|meta| meta.display_name)
+}
+
+pub fn get_default_config_json() -> Option<String> {
+    crate::build_default_network_config_json().ok()
+}
+
+pub fn create_config_record(config_id: String, display_name: String) -> Option<StoredConfigRecord> {
+    let raw = get_default_config_json()?;
+    let mut config = serde_json::from_str::<NetworkConfig>(&raw).ok()?;
+    config.instance_id = Some(config_id.clone());
+    let normalized_json = serde_json::to_string(&config).ok()?;
+    save_config_record(config_id, display_name, normalized_json)
+}
+
+pub fn start_kernel_with_config_id(config_id: &str) -> bool {
+    let raw = match load_config_json(config_id) {
+        Some(raw) => raw,
+        None => return false,
+    };
+    crate::run_network_instance_from_json(&raw)
+}
+
+pub fn list_config_meta_json() -> String {
+    serde_json::to_string(&list_config_meta_entries().configs).unwrap_or_else(|_| "[]".to_string())
+}
+
+pub fn delete_config_record(config_id: &str) -> bool {
+    if let Some(path) = legacy_config_file_path(config_id) {
+        if path.exists() {
+            let _ = std::fs::remove_file(path);
+        }
+    }
+
+    let conn = match open_db() {
+        Some(conn) => conn,
+        None => return false,
+    };
+    if let Err(e) = conn.execute(
+        "DELETE FROM stored_config_fields WHERE config_id = ?1",
+        params![config_id],
+    ) {
+        hilog_error!("[Rust] failed to delete config fields {}: {}", config_id, e);
+        return false;
+    }
+
+    delete_config_meta(config_id)
+}
+
+pub fn export_config_toml(config_id: &str) -> Option<ExportTomlResult> {
+    let record = get_config_record(config_id)?;
+    import_export::export_config_toml_from_record(&record)
+}
+
+pub fn import_toml_config(
+    toml_text: String,
+    display_name: Option<String>,
+) -> Option<StoredConfigRecord> {
+    import_export::import_toml_to_record(toml_text, display_name, save_config_record)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use rusqlite::params;
+    use std::path::PathBuf;
+    use std::time::{SystemTime, UNIX_EPOCH};
+
+    fn test_root() -> String {
+        let unique = SystemTime::now()
+            .duration_since(UNIX_EPOCH)
+            .unwrap()
+            .as_nanos();
+        let dir = std::env::temp_dir().join(format!("easytier_ohrs_test_{}", unique));
+        dir.to_string_lossy().into_owned()
+    }
+
+    #[test]
+    fn save_get_export_delete_roundtrip() {
+        let root = test_root();
+        assert!(init_config_store(root.clone()));
+
+        let config_json = crate::build_default_network_config_json().expect("default config");
+        let saved = save_config_record("cfg-1".to_string(), "test-config".to_string(), config_json)
+            .expect("save config");
+
+        assert_eq!(saved.meta.config_id, "cfg-1");
+        assert_eq!(saved.meta.display_name, "test-config");
+
+        let loaded = get_config_record("cfg-1").expect("load config");
+        assert_eq!(loaded.meta.display_name, "test-config");
+        assert!(loaded.config_json.contains("cfg-1"));
+
+        let legacy_json_path = PathBuf::from(&root)
+            .join(CONFIG_DIR_NAME)
+            .join("cfg-1.json");
+        assert!(
+            !legacy_json_path.exists(),
+            "config should no longer be persisted as a per-config json file"
+        );
+
+        let conn = open_db().expect("db should be open");
+        let field_count: i64 = conn
+            .query_row(
+                "SELECT COUNT(*) FROM stored_config_fields WHERE config_id = ?1",
+                params!["cfg-1"],
+                |row| row.get(0),
+            )
+            .expect("count config fields");
+        assert!(field_count > 0, "config fields should be stored in sqlite");
+
+        let exported = export_config_toml("cfg-1").expect("export toml");
+        assert!(exported.toml_text.contains("instance_id"));
+
+        assert!(delete_config_record("cfg-1"));
+        assert!(get_config_record("cfg-1").is_none());
+    }
+
+    #[test]
+    fn set_config_field_updates_only_requested_top_level_field() {
+        let root = test_root();
+        assert!(init_config_store(root));
+
+        let config_json = crate::build_default_network_config_json().expect("default config");
+        save_config_record(
+            "cfg-field".to_string(),
+            "field-config".to_string(),
+            config_json,
+        )
+        .expect("save config");
+
+        let before_network_name = get_config_field_value("cfg-field", "network_name");
+        let before_instance_id = get_config_field_value("cfg-field", "instance_id")
+            .expect("instance id field should exist");
+
+        assert!(set_config_field_value(
+            "cfg-field",
+            "network_name",
+            "\"changed-network\""
+        ));
+
+        assert_eq!(
+            get_config_field_value("cfg-field", "network_name"),
+            Some("\"changed-network\"".to_string())
+        );
+        assert_eq!(
+            get_config_field_value("cfg-field", "instance_id"),
+            Some(before_instance_id)
+        );
+        assert_ne!(
+            get_config_field_value("cfg-field", "network_name"),
+            before_network_name
+        );
+    }
+}

easytier-contrib/easytier-ohrs/src/config_repo/field_store.rs

@@ -0,0 +1,67 @@
+use crate::config::storage::config_meta::{now_ts_string, open_db};
+use ohos_hilog_binding::hilog_error;
+use rusqlite::{Connection, params};
+use serde_json::{Map, Value};
+
+pub(super) fn load_config_map_from_db(config_id: &str) -> Option<Map<String, Value>> {
+    let conn = open_db()?;
+    let mut stmt = conn
+        .prepare(
+            "SELECT field_name, field_json
+             FROM stored_config_fields
+             WHERE config_id = ?1",
+        )
+        .ok()?;
+    let rows = stmt
+        .query_map(params![config_id], |row| {
+            let field_name: String = row.get(0)?;
+            let field_json: String = row.get(1)?;
+            Ok((field_name, field_json))
+        })
+        .ok()?;
+
+    let mut object = Map::new();
+    for row in rows {
+        let (field_name, field_json) = row.ok()?;
+        let value = serde_json::from_str::<Value>(&field_json).ok()?;
+        object.insert(field_name, value);
+    }
+
+    if object.is_empty() {
+        None
+    } else {
+        Some(object)
+    }
+}
+
+pub(super) fn replace_config_fields(
+    tx: &Connection,
+    config_id: &str,
+    fields: Map<String, Value>,
+) -> Option<()> {
+    if let Err(e) = tx.execute(
+        "DELETE FROM stored_config_fields WHERE config_id = ?1",
+        params![config_id],
+    ) {
+        hilog_error!(
+            "[Rust] failed to clear existing config fields {}: {}",
+            config_id,
+            e
+        );
+        return None;
+    }
+
+    for (field_name, value) in fields {
+        let field_json = serde_json::to_string(&value).ok()?;
+        if let Err(e) = tx.execute(
+            "INSERT INTO stored_config_fields (config_id, field_name, field_json, updated_at)
+             VALUES (?1, ?2, ?3, ?4)",
+            params![config_id, field_name, field_json, now_ts_string()],
+        ) {
+            hilog_error!("[Rust] failed to persist config field {}: {}", config_id, e);
+            return None;
+        }
+    }
+
+    Some(())
+}

easytier-contrib/easytier-ohrs/src/config_repo/import_export.rs

@@ -0,0 +1,48 @@
+use crate::config::types::stored_config::{ExportTomlResult, StoredConfigRecord};
+use easytier::common::config::{ConfigLoader, TomlConfigLoader};
+use easytier::proto::api::manage::NetworkConfig;
+
+pub(super) fn export_config_toml_from_record(
+    record: &StoredConfigRecord,
+) -> Option<ExportTomlResult> {
+    let config = serde_json::from_str::<NetworkConfig>(&record.config_json).ok()?;
+    let toml = config.gen_config().ok()?;
+    Some(ExportTomlResult {
+        toml_text: toml.dump(),
+    })
+}
+
+pub(super) fn import_toml_to_record(
+    toml_text: String,
+    display_name: Option<String>,
+    save_config_record: impl Fn(String, String, String) -> Option<StoredConfigRecord>,
+) -> Option<StoredConfigRecord> {
+    let config =
+        NetworkConfig::new_from_config(TomlConfigLoader::new_from_str(&toml_text).ok()?).ok()?;
+
+    let config_id = config.instance_id.clone()?;
+    let name_from_toml = toml_text
+        .lines()
+        .find_map(|line| {
+            let trimmed = line.trim();
+            if !trimmed.starts_with("instance_name") {
+                return None;
+            }
+            trimmed.split_once('=').map(|(_, value)| {
+                value
+                    .trim()
+                    .trim_matches('"')
+                    .trim_matches('\'')
+                    .to_string()
+            })
+        })
+        .filter(|name| !name.is_empty());
+
+    let final_name = display_name
+        .filter(|name| !name.is_empty())
+        .or(name_from_toml)
+        .unwrap_or_else(|| config_id.clone());
+
+    let config_json = serde_json::to_string(&config).ok()?;
+    save_config_record(config_id, final_name, config_json)
+}

easytier-contrib/easytier-ohrs/src/config_repo/legacy_migration.rs

@@ -0,0 +1,45 @@
+use crate::config::storage::config_meta::get_config_meta;
+use ohos_hilog_binding::hilog_error;
+use std::path::PathBuf;
+
+pub(super) fn legacy_config_file_path(
+    root_dir: &Option<PathBuf>,
+    config_dir_name: &str,
+    config_id: &str,
+) -> Option<PathBuf> {
+    root_dir.as_ref().map(|root| {
+        root.join(config_dir_name)
+            .join(format!("{}.json", config_id))
+    })
+}
+
+pub(super) fn migrate_legacy_file_if_needed(
+    root_dir: &Option<PathBuf>,
+    config_dir_name: &str,
+    config_id: &str,
+    save_config_record: impl Fn(
+        String,
+        String,
+        String,
+    ) -> Option<crate::config::types::stored_config::StoredConfigRecord>,
+) -> Option<()> {
+    let legacy_path = legacy_config_file_path(root_dir, config_dir_name, config_id)?;
+    if !legacy_path.exists() {
+        return Some(());
+    }
+
+    let raw = std::fs::read_to_string(&legacy_path).ok()?;
+    let display_name = get_config_meta(config_id)
+        .map(|meta| meta.display_name)
+        .unwrap_or_else(|| config_id.to_string());
+    save_config_record(config_id.to_string(), display_name, raw)?;
+
+    if let Err(e) = std::fs::remove_file(&legacy_path) {
+        hilog_error!(
+            "[Rust] failed to remove legacy config file {}: {}",
+            legacy_path.display(),
+            e
+        );
+    }
+    Some(())
+}

easytier-contrib/easytier-ohrs/src/config_repo/validation.rs

@@ -0,0 +1,30 @@
+use easytier::proto::api::manage::NetworkConfig;
+use serde_json::{Map, Value};
+
+pub(super) fn normalize_config_id(
+    mut config: NetworkConfig,
+    requested_id: String,
+) -> Result<NetworkConfig, String> {
+    if requested_id.is_empty() {
+        return Err("config_id is required".to_string());
+    }
+    config.instance_id = Some(requested_id);
+    Ok(config)
+}
+
+pub(super) fn validate_config_json(
+    config_json: &str,
+    config_id: String,
+) -> Result<NetworkConfig, String> {
+    let config = serde_json::from_str::<NetworkConfig>(config_json)
+        .map_err(|e| format!("parse config json failed: {}", e))?;
+    let config = normalize_config_id(config, config_id)?;
+    config
+        .gen_config()
+        .map_err(|e| format!("generate toml failed: {}", e))?;
+    Ok(config)
+}
+
+pub(super) fn config_to_top_level_map(config: &NetworkConfig) -> Option<Map<String, Value>> {
+    serde_json::to_value(config).ok()?.as_object().cloned()
+}

easytier-contrib/easytier-ohrs/src/exports.rs

@@ -0,0 +1,2 @@
+pub(crate) mod config_api;
+pub(crate) mod runtime_api;

easytier-contrib/easytier-ohrs/src/exports/config_api.rs

@@ -0,0 +1,46 @@
+use crate::config;
+
+pub(crate) fn init_config_store(root_dir: String) -> bool {
+    config::repository::init_config_store(root_dir)
+}
+
+pub(crate) fn list_configs() -> String {
+    config::repository::list_config_meta_json()
+}
+
+pub(crate) fn save_config(config_id: String, display_name: String, config_json: String) -> bool {
+    config::repository::save_config_record(config_id, display_name, config_json).is_some()
+}
+
+pub(crate) fn create_config(config_id: String, display_name: String) -> bool {
+    config::repository::create_config_record(config_id, display_name).is_some()
+}
+
+pub(crate) fn delete_stored_config_meta(config_id: String) -> bool {
+    config::repository::delete_config_record(&config_id)
+}
+
+pub(crate) fn get_config(config_id: String) -> Option<String> {
+    config::repository::load_config_json(&config_id)
+}
+
+pub(crate) fn get_default_config() -> Option<String> {
+    config::repository::get_default_config_json()
+}
+
+pub(crate) fn get_config_field(config_id: String, field: String) -> Option<String> {
+    config::repository::get_config_field_value(&config_id, &field)
+}
+
+pub(crate) fn set_config_field(config_id: String, field: String, json_value: String) -> bool {
+    config::repository::set_config_field_value(&config_id, &field, &json_value)
+}
+
+pub(crate) fn import_toml(toml_text: String, display_name: Option<String>) -> Option<String> {
+    config::repository::import_toml_config(toml_text, display_name)
+        .map(|record| record.meta.config_id)
+}
+
+pub(crate) fn export_toml(config_id: String) -> Option<String> {
+    config::repository::export_config_toml(&config_id).map(|ret| ret.toml_text)
+}

easytier-contrib/easytier-ohrs/src/exports/runtime_api.rs

@@ -0,0 +1,184 @@
+use crate::config::repository::load_config_json;
+use crate::config::storage::config_meta::get_config_display_name;
+use crate::config::types::stored_config::KeyValuePair;
+use crate::kernel_bridge::{
+    aggregate_requested_tun_routes, start_local_socket_server as start_local_socket_server_inner,
+    stop_local_socket_server as stop_local_socket_server_inner,
+};
+use crate::runtime::state::runtime_state::{
+    RuntimeAggregateState, TunAggregateState, clear_tun_attached, mark_tun_attached,
+    runtime_instance_from_running_info,
+};
+use crate::{ASYNC_RUNTIME, EASYTIER_VERSION, INSTANCE_MANAGER, WEB_CLIENTS};
+use easytier::proto::api::manage::NetworkConfig;
+use ohos_hilog_binding::{hilog_error, hilog_info};
+use std::sync::Arc;
+
+pub(crate) fn start_kernel(
+    config_id: String,
+    start_kernel_with_config_id: impl Fn(&str) -> bool,
+) -> bool {
+    start_kernel_with_config_id(&config_id)
+}
+
+pub(crate) fn stop_kernel(
+    config_id: String,
+    stop_web_client: impl Fn(&str) -> bool,
+    parse_instance_uuid: impl Fn(&str) -> Option<uuid::Uuid>,
+    maybe_stop_local_socket_server: impl Fn(),
+) -> bool {
+    clear_tun_attached(&config_id);
+    if stop_web_client(&config_id) {
+        return true;
+    }
+
+    let Some(instance_id) = parse_instance_uuid(&config_id) else {
+        return false;
+    };
+
+    let ret = INSTANCE_MANAGER
+        .delete_network_instance(vec![instance_id])
+        .map(|_| true)
+        .unwrap_or_else(|err| {
+            hilog_error!("[Rust] stop_kernel failed {}: {}", config_id, err);
+            false
+        });
+    maybe_stop_local_socket_server();
+    ret
+}
+
+pub(crate) fn stop_network_instance(
+    config_ids: Vec<String>,
+    stop_kernel: impl Fn(String) -> bool,
+) -> bool {
+    let mut ok = true;
+    for config_id in config_ids {
+        ok = stop_kernel(config_id) && ok;
+    }
+    ok
+}
+
+pub(crate) fn collect_network_infos() -> Vec<KeyValuePair> {
+    let infos = match INSTANCE_MANAGER.collect_network_infos_sync() {
+        Ok(infos) => infos,
+        Err(err) => {
+            hilog_error!("[Rust] collect network infos failed {}", err);
+            return vec![];
+        }
+    };
+
+    infos
+        .into_iter()
+        .filter_map(|(key, value)| {
+            serde_json::to_string(&value)
+                .ok()
+                .map(|value_json| KeyValuePair {
+                    key: key.to_string(),
+                    value: value_json,
+                })
+        })
+        .collect()
+}
+
+pub(crate) fn set_tun_fd(
+    config_id: String,
+    fd: i32,
+    parse_instance_uuid: impl Fn(&str) -> Option<uuid::Uuid>,
+) -> bool {
+    let Some(instance_id) = parse_instance_uuid(&config_id) else {
+        hilog_error!("[Rust] set_tun_fd invalid instance id: {}", config_id);
+        return false;
+    };
+
+    INSTANCE_MANAGER
+        .set_tun_fd(&instance_id, fd)
+        .map(|_| {
+            mark_tun_attached(&config_id);
+            hilog_info!(
+                "[Rust] set_tun_fd success instance={} fd={} marked_attached=true",
+                config_id,
+                fd
+            );
+            true
+        })
+        .unwrap_or_else(|err| {
+            hilog_error!("[Rust] set_tun_fd failed {}: {}", config_id, err);
+            false
+        })
+}
+
+pub(crate) fn get_runtime_snapshot() -> RuntimeAggregateState {
+    get_runtime_snapshot_inner()
+}
+
+pub(crate) fn get_runtime_snapshot_inner() -> RuntimeAggregateState {
+    let infos = match INSTANCE_MANAGER.collect_network_infos_sync() {
+        Ok(infos) => infos,
+        Err(err) => {
+            hilog_error!("[Rust] collect network infos failed {}", err);
+            return RuntimeAggregateState {
+                instances: vec![],
+                tun: TunAggregateState {
+                    active: false,
+                    attached_instance_ids: vec![],
+                    aggregated_routes: vec![],
+                    dns_servers: vec![],
+                    need_rebuild: false,
+                },
+                running_instance_count: 0,
+            };
+        }
+    };
+
+    let mut instances = Vec::with_capacity(infos.len());
+    for (instance_uuid, info) in infos {
+        let config_id = instance_uuid.to_string();
+        let display_name = get_config_display_name(&config_id).unwrap_or_else(|| config_id.clone());
+        let config_json = load_config_json(&config_id);
+        let stored_config = config_json
+            .as_deref()
+            .and_then(|raw| serde_json::from_str::<NetworkConfig>(raw).ok());
+        let magic_dns_enabled = stored_config
+            .as_ref()
+            .and_then(|cfg| cfg.enable_magic_dns)
+            .unwrap_or(false);
+        let need_exit_node = stored_config
+            .as_ref()
+            .map(|cfg| !cfg.exit_nodes.is_empty())
+            .unwrap_or(false);
+        instances.push(runtime_instance_from_running_info(
+            config_id,
+            display_name,
+            magic_dns_enabled,
+            need_exit_node,
+            info,
+        ));
+    }
+
+    instances.sort_by(|a, b| {
+        a.display_name
+            .cmp(&b.display_name)
+            .then_with(|| a.instance_id.cmp(&b.instance_id))
+    });
+    let attached_instance_ids = instances
+        .iter()
+        .filter(|instance| instance.tun_required)
+        .map(|instance| instance.instance_id.clone())
+        .collect::<Vec<_>>();
+    let aggregated_routes = aggregate_requested_tun_routes(&instances);
+    let running_instance_count =
+        instances.iter().filter(|instance| instance.running).count() as i32;
+    let tun_active = !attached_instance_ids.is_empty();
+
+    RuntimeAggregateState {
+        instances,
+        tun: TunAggregateState {
+            active: tun_active,
+            attached_instance_ids,
+            aggregated_routes,
+            dns_servers: vec![],
+            need_rebuild: false,
+        },
+        running_instance_count,
+    }
+}

easytier-contrib/easytier-ohrs/src/kernel_bridge.rs

@@ -0,0 +1,6 @@
+mod protocol;
+mod routing;
+mod socket_server;
+
+pub(crate) use routing::aggregate_requested_tun_routes;
+pub use socket_server::{start_local_socket_server, stop_local_socket_server};

easytier-contrib/easytier-ohrs/src/kernel_bridge/protocol.rs

@@ -0,0 +1,50 @@
+use crate::config::types::stored_config::LocalSocketSyncMessage;
+use serde::Serialize;
+use std::io::{Error, ErrorKind, Write};
+use std::os::unix::net::UnixStream;
+
+#[derive(Debug, Clone, Serialize)]
+#[serde(rename_all = "camelCase")]
+pub(crate) struct TunRequestPayload {
+    pub config_id: String,
+    pub instance_id: String,
+    pub display_name: String,
+    pub virtual_ipv4: Option<String>,
+    pub virtual_ipv4_cidr: Option<String>,
+    pub aggregated_routes: Vec<String>,
+    pub magic_dns_enabled: bool,
+    pub need_exit_node: bool,
+}
+
+pub(crate) fn send_local_socket_message(
+    stream: &mut UnixStream,
+    message_type: &str,
+    payload_json: String,
+) -> std::io::Result<()> {
+    let message = LocalSocketSyncMessage {
+        message_type: message_type.to_string(),
+        payload_json,
+    };
+    let mut raw = serde_json::to_vec(&message)
+        .map_err(|err| Error::new(ErrorKind::InvalidData, err.to_string()))?;
+    raw.push(b'\n');
+    stream.write_all(&raw)?;
+    Ok(())
+}
+
+pub(crate) fn broadcast_local_socket_message(
+    clients: &mut Vec<UnixStream>,
+    message_type: &str,
+    payload_json: &str,
+) -> bool {
+    let mut active_clients = Vec::with_capacity(clients.len());
+    let mut delivered = false;
+    for mut client in clients.drain(..) {
+        if send_local_socket_message(&mut client, message_type, payload_json.to_string()).is_ok() {
+            delivered = true;
+            active_clients.push(client);
+        }
+    }
+    *clients = active_clients;
+    delivered
+}

easytier-contrib/easytier-ohrs/src/kernel_bridge/routing.rs

@@ -0,0 +1,105 @@
+use crate::config::repository::load_config_json;
+use crate::runtime::state::runtime_state::RuntimeInstanceState;
+use easytier::proto::api::manage::NetworkConfig;
+use ipnet::IpNet;
+use ohos_hilog_binding::hilog_debug;
+use std::collections::HashSet;
+use std::net::IpAddr;
+
+pub(crate) fn load_manual_routes(config_id: &str) -> Vec<String> {
+    load_config_json(config_id)
+        .and_then(|raw| serde_json::from_str::<NetworkConfig>(&raw).ok())
+        .map(|config| config.routes)
+        .unwrap_or_default()
+}
+
+fn normalize_route_cidr(route: &str) -> Option<String> {
+    route
+        .parse::<IpNet>()
+        .ok()
+        .map(|network| match network {
+            IpNet::V4(net) => net.trunc().to_string(),
+            IpNet::V6(net) => net.trunc().to_string(),
+        })
+        .or_else(|| {
+            route.parse::<IpAddr>().ok().map(|addr| match addr {
+                IpAddr::V4(ip) => format!("{}/32", ip),
+                IpAddr::V6(ip) => format!("{}/128", ip),
+            })
+        })
+}
+
+fn simplify_routes(routes: Vec<String>) -> Vec<String> {
+    let mut parsed = routes
+        .into_iter()
+        .filter_map(|route| normalize_route_cidr(&route))
+        .filter_map(|route| route.parse::<IpNet>().ok())
+        .collect::<Vec<_>>();
+    parsed.sort_by(|left, right| {
+        left.prefix_len()
+            .cmp(&right.prefix_len())
+            .then_with(|| left.network().to_string().cmp(&right.network().to_string()))
+    });
+
+    let mut simplified = Vec::<IpNet>::new();
+    'outer: for route in parsed {
+        for existing in &simplified {
+            if existing.contains(&route.network()) && existing.prefix_len() <= route.prefix_len() {
+                continue 'outer;
+            }
+        }
+        simplified.retain(|existing| {
+            !(route.contains(&existing.network()) && route.prefix_len() <= existing.prefix_len())
+        });
+        simplified.push(route);
+    }
+
+    let mut seen = HashSet::new();
+    simplified
+        .into_iter()
+        .map(|route| route.to_string())
+        .filter(|route| seen.insert(route.clone()))
+        .collect()
+}
+
+pub(crate) fn aggregate_tun_routes(instance: &RuntimeInstanceState) -> Vec<String> {
+    let virtual_ipv4_cidr = instance
+        .my_node_info
+        .as_ref()
+        .and_then(|info| info.virtual_ipv4_cidr.clone());
+    let manual_routes = load_manual_routes(&instance.config_id);
+    let proxy_cidrs = instance
+        .routes
+        .iter()
+        .flat_map(|route| route.proxy_cidrs.iter().cloned())
+        .collect::<Vec<_>>();
+    let mut raw_routes = Vec::new();
+
+    if let Some(cidr) = virtual_ipv4_cidr.clone() {
+        raw_routes.push(cidr);
+    }
+
+    raw_routes.extend(manual_routes.iter().cloned());
+    raw_routes.extend(proxy_cidrs.iter().cloned());
+    let aggregated_routes = simplify_routes(raw_routes);
+    hilog_debug!(
+        "[Rust] aggregate_tun_routes instance={} proxy_cidrs={:?} aggregated_routes={:?}",
+        instance.instance_id,
+        proxy_cidrs,
+        aggregated_routes
+    );
+    aggregated_routes
+}
+
+pub(crate) fn aggregate_requested_tun_routes(instances: &[RuntimeInstanceState]) -> Vec<String> {
+    let mut aggregated_routes = Vec::new();
+    let mut seen_routes = HashSet::new();
+    for instance in instances.iter().filter(|instance| instance.tun_required) {
+        for route in aggregate_tun_routes(instance) {
+            if seen_routes.insert(route.clone()) {
+                aggregated_routes.push(route);
+            }
+        }
+    }
+    aggregated_routes
+}

easytier-contrib/easytier-ohrs/src/kernel_bridge/socket_server.rs

@@ -0,0 +1,196 @@
+use super::protocol::{TunRequestPayload, broadcast_local_socket_message};
+use crate::config::repository::kernel_socket_path;
+use crate::get_runtime_snapshot_inner;
+use crate::kernel_bridge::routing::aggregate_tun_routes;
+use ohos_hilog_binding::{hilog_error, hilog_info};
+use once_cell::sync::Lazy;
+use std::collections::{HashMap, HashSet};
+use std::io::ErrorKind;
+use std::os::unix::net::{UnixListener, UnixStream};
+use std::path::PathBuf;
+use std::sync::Mutex;
+use std::sync::atomic::{AtomicBool, Ordering};
+use std::thread::{self, JoinHandle};
+use std::time::Duration;
+
+struct LocalSocketState {
+    stop_flag: std::sync::Arc<AtomicBool>,
+    socket_path: PathBuf,
+    worker: JoinHandle<()>,
+}
+
+static LOCAL_SOCKET_STATE: Lazy<Mutex<Option<LocalSocketState>>> = Lazy::new(|| Mutex::new(None));
+
+pub fn start_local_socket_server() -> bool {
+    let socket_path = match kernel_socket_path() {
+        Some(path) => path,
+        None => {
+            hilog_error!("[Rust] kernel socket path unavailable");
+            return false;
+        }
+    };
+
+    match LOCAL_SOCKET_STATE.lock() {
+        Ok(guard) if guard.is_some() => return true,
+        Ok(_) => {}
+        Err(err) => {
+            hilog_error!("[Rust] lock localsocket state failed: {}", err);
+            return false;
+        }
+    }
+
+    if socket_path.exists() {
+        let _ = std::fs::remove_file(&socket_path);
+    }
+
+    let listener = match UnixListener::bind(&socket_path) {
+        Ok(listener) => listener,
+        Err(err) => {
+            hilog_error!(
+                "[Rust] bind localsocket failed {}: {}",
+                socket_path.display(),
+                err
+            );
+            return false;
+        }
+    };
+    if let Err(err) = listener.set_nonblocking(true) {
+        hilog_error!("[Rust] set localsocket nonblocking failed: {}", err);
+        let _ = std::fs::remove_file(&socket_path);
+        return false;
+    }
+
+    let stop_flag = std::sync::Arc::new(AtomicBool::new(false));
+    let worker_stop_flag = stop_flag.clone();
+    let worker = thread::spawn(move || {
+        let mut last_snapshot_json = String::new();
+        let mut delivered_tun_requests = HashSet::new();
+        let mut last_tun_route_signatures = HashMap::<String, String>::new();
+        let mut clients = Vec::<UnixStream>::new();
+
+        while !worker_stop_flag.load(Ordering::Relaxed) {
+            let mut accepted_client = false;
+            loop {
+                match listener.accept() {
+                    Ok((stream, _addr)) => {
+                        accepted_client = true;
+                        clients.push(stream);
+                    }
+                    Err(err) if err.kind() == ErrorKind::WouldBlock => break,
+                    Err(err) => {
+                        hilog_error!("[Rust] accept localsocket failed: {}", err);
+                        break;
+                    }
+                }
+            }
+
+            let snapshot = get_runtime_snapshot_inner();
+            let snapshot_json = match serde_json::to_string(&snapshot) {
+                Ok(json) => json,
+                Err(err) => {
+                    hilog_error!("[Rust] serialize runtime snapshot failed: {}", err);
+                    thread::sleep(Duration::from_millis(250));
+                    continue;
+                }
+            };
+
+            if accepted_client || snapshot_json != last_snapshot_json {
+                let _ = broadcast_local_socket_message(
+                    &mut clients,
+                    "runtime_snapshot",
+                    &snapshot_json,
+                );
+                last_snapshot_json = snapshot_json;
+            }
+
+            for instance in snapshot.instances.iter() {
+                if instance.running && instance.tun_required {
+                    let virtual_ipv4 = instance
+                        .my_node_info
+                        .as_ref()
+                        .and_then(|info| info.virtual_ipv4.clone());
+                    let virtual_ipv4_cidr = instance
+                        .my_node_info
+                        .as_ref()
+                        .and_then(|info| info.virtual_ipv4_cidr.clone());
+                    if clients.is_empty() {
+                        continue;
+                    }
+                    if virtual_ipv4.is_none() || virtual_ipv4_cidr.is_none() {
+                        continue;
+                    }
+                    let aggregated_routes = aggregate_tun_routes(instance);
+                    let route_signature = serde_json::to_string(&aggregated_routes)
+                        .unwrap_or_else(|_| "[]".to_string());
+                    let should_send = !delivered_tun_requests.contains(&instance.instance_id)
+                        || last_tun_route_signatures
+                            .get(&instance.instance_id)
+                            .map(|value| value != &route_signature)
+                            .unwrap_or(true);
+                    if !should_send {
+                        continue;
+                    }
+                    let payload = TunRequestPayload {
+                        config_id: instance.config_id.clone(),
+                        instance_id: instance.instance_id.clone(),
+                        display_name: instance.display_name.clone(),
+                        virtual_ipv4,
+                        virtual_ipv4_cidr,
+                        aggregated_routes,
+                        magic_dns_enabled: instance.magic_dns_enabled,
+                        need_exit_node: instance.need_exit_node,
+                    };
+                    let payload_json = match serde_json::to_string(&payload) {
+                        Ok(json) => json,
+                        Err(err) => {
+                            hilog_error!("[Rust] serialize tun request failed: {}", err);
+                            continue;
+                        }
+                    };
+                    if broadcast_local_socket_message(&mut clients, "tun_request", &payload_json) {
+                        delivered_tun_requests.insert(instance.instance_id.clone());
+                        last_tun_route_signatures
+                            .insert(instance.instance_id.clone(), route_signature);
+                    }
+                } else {
+                    delivered_tun_requests.remove(&instance.instance_id);
+                    last_tun_route_signatures.remove(&instance.instance_id);
+                }
+            }
+
+            thread::sleep(Duration::from_millis(250));
+        }
+    });
+
+    match LOCAL_SOCKET_STATE.lock() {
+        Ok(mut guard) => {
+            *guard = Some(LocalSocketState {
+                stop_flag,
+                socket_path,
+                worker,
+            });
+            true
+        }
+        Err(err) => {
+            hilog_error!("[Rust] lock localsocket state failed: {}", err);
+            false
+        }
+    }
+}
+
+pub fn stop_local_socket_server() -> bool {
+    let state = match LOCAL_SOCKET_STATE.lock() {
+        Ok(mut guard) => guard.take(),
+        Err(err) => {
+            hilog_error!("[Rust] lock localsocket state failed: {}", err);
+            return false;
+        }
+    };
+
+    if let Some(state) = state {
+        state.stop_flag.store(true, Ordering::Relaxed);
+        let _ = state.worker.join();
+        let _ = std::fs::remove_file(state.socket_path);
+    }
+    true
+}

easytier-contrib/easytier-ohrs/src/lib.rs

@@ -1,185 +1,485 @@
-mod native_log;
+mod config;
+mod exports;
+mod kernel_bridge;
+mod platform;
+mod runtime;
 
-use easytier::common::config::{ConfigFileControl, ConfigLoader, TomlConfigLoader};
+use config::repository::{
+    create_config_record, delete_config_record, export_config_toml, get_config_field_value,
+    get_default_config_json, import_toml_config, init_config_store as init_repo_store,
+    list_config_meta_json, save_config_record, set_config_field_value, start_kernel_with_config_id,
+};
+use config::services::schema_service::{
+    ConfigFieldMapping, NetworkConfigSchema,
+    get_network_config_field_mappings as build_network_config_field_mappings,
+    get_network_config_schema as build_network_config_schema,
+};
+use config::services::share_link_service::{
+    build_config_share_link as build_config_share_link_inner,
+    import_config_share_link as import_config_share_link_inner,
+    parse_config_share_link as parse_config_share_link_inner,
+};
+use config::storage::config_meta::get_config_display_name;
+use config::types::stored_config::{KeyValuePair, SharedConfigLinkPayload};
 use easytier::common::constants::EASYTIER_VERSION;
+use easytier::common::{
+    MachineIdOptions,
+    config::{ConfigFileControl, ConfigLoader, TomlConfigLoader},
+};
 use easytier::instance_manager::NetworkInstanceManager;
 use easytier::proto::api::manage::NetworkConfig;
+use easytier::proto::api::manage::NetworkingMethod;
+use easytier::web_client::{WebClient, WebClientHooks, run_web_client};
+use kernel_bridge::{
+    aggregate_requested_tun_routes, start_local_socket_server as start_local_socket_server_inner,
+    stop_local_socket_server as stop_local_socket_server_inner,
+};
 use napi_derive_ohos::napi;
-use ohos_hilog_binding::{hilog_debug, hilog_error};
+use ohos_hilog_binding::{hilog_error, hilog_info};
+use runtime::state::runtime_state::{
+    RuntimeAggregateState, TunAggregateState, clear_tun_attached, mark_tun_attached,
+    runtime_instance_from_running_info,
+};
+use std::collections::{HashMap, HashSet};
 use std::format;
+use std::sync::{Arc, Mutex};
+use tokio::runtime::{Builder, Runtime};
 use uuid::Uuid;
 
-static INSTANCE_MANAGER: once_cell::sync::Lazy<NetworkInstanceManager> =
-    once_cell::sync::Lazy::new(NetworkInstanceManager::new);
+pub(crate) static INSTANCE_MANAGER: once_cell::sync::Lazy<Arc<NetworkInstanceManager>> =
+    once_cell::sync::Lazy::new(|| Arc::new(NetworkInstanceManager::new()));
+static ASYNC_RUNTIME: once_cell::sync::Lazy<Runtime> = once_cell::sync::Lazy::new(|| {
+    Builder::new_multi_thread()
+        .enable_all()
+        .build()
+        .expect("tokio runtime for easytier-ohrs")
+});
+static WEB_CLIENTS: once_cell::sync::Lazy<Mutex<HashMap<String, ManagedWebClient>>> =
+    once_cell::sync::Lazy::new(|| Mutex::new(HashMap::new()));
 
-#[napi(object)]
-pub struct KeyValuePair {
-    pub key: String,
-    pub value: String,
+#[derive(Default)]
+struct TrackedWebClientHooks {
+    instance_ids: Mutex<HashSet<Uuid>>,
 }
 
-#[napi]
-pub fn easytier_version() -> String {
-    EASYTIER_VERSION.to_string()
+struct ManagedWebClient {
+    _client: WebClient,
+    hooks: Arc<TrackedWebClientHooks>,
 }
 
-#[napi]
-pub fn set_tun_fd(inst_id: String, fd: i32) -> bool {
-    match Uuid::try_parse(&inst_id) {
-        Ok(uuid) => match INSTANCE_MANAGER.set_tun_fd(&uuid, fd) {
-            Ok(_) => {
-                hilog_debug!("[Rust] set tun fd {} to {}.", fd, inst_id);
-                true
-            }
-            Err(e) => {
-                hilog_error!("[Rust] cant set tun fd {} to {}. {}", fd, inst_id, e);
-                false
-            }
-        },
-        Err(e) => {
-            hilog_error!("[Rust] cant covert {} to uuid. {}", inst_id, e);
+#[async_trait::async_trait]
+impl WebClientHooks for TrackedWebClientHooks {
+    async fn post_run_network_instance(&self, id: &Uuid) -> Result<(), String> {
+        self.instance_ids
+            .lock()
+            .map_err(|err| err.to_string())?
+            .insert(*id);
+        Ok(())
+    }
+
+    async fn post_remove_network_instances(&self, ids: &[Uuid]) -> Result<(), String> {
+        let mut guard = self.instance_ids.lock().map_err(|err| err.to_string())?;
+        for id in ids {
+            guard.remove(id);
+        }
+        Ok(())
+    }
+}
+
+fn is_config_server_config(config: &NetworkConfig) -> bool {
+    matches!(
+        NetworkingMethod::try_from(config.networking_method.unwrap_or_default())
+            .unwrap_or_default(),
+        NetworkingMethod::PublicServer
+    ) && config
+        .public_server_url
+        .as_ref()
+        .is_some_and(|url| !url.trim().is_empty())
+}
+
+fn stop_web_client(config_id: &str) -> bool {
+    let managed = match WEB_CLIENTS.lock() {
+        Ok(mut guard) => guard.remove(config_id),
+        Err(err) => {
+            hilog_error!("[Rust] stop_web_client lock failed {}", err);
+            return false;
+        }
+    };
+
+    let Some(managed) = managed else {
+        return false;
+    };
+
+    let tracked_ids = managed
+        .hooks
+        .instance_ids
+        .lock()
+        .map(|guard| guard.iter().copied().collect::<Vec<_>>())
+        .unwrap_or_default();
+    drop(managed);
+
+    if tracked_ids.is_empty() {
+        maybe_stop_local_socket_server();
+        return true;
+    }
+
+    let ret = INSTANCE_MANAGER
+        .delete_network_instance(tracked_ids)
+        .map(|_| true)
+        .unwrap_or_else(|err| {
+            hilog_error!(
+                "[Rust] stop config server instances failed {}: {}",
+                config_id,
+                err
+            );
+            false
+        });
+    maybe_stop_local_socket_server();
+    ret
+}
+
+fn ensure_local_socket_server_started() -> bool {
+    start_local_socket_server_inner()
+}
+
+fn maybe_stop_local_socket_server() {
+    let no_local_instances = INSTANCE_MANAGER.list_network_instance_ids().is_empty();
+    let no_web_clients = WEB_CLIENTS
+        .lock()
+        .map(|guard| guard.is_empty())
+        .unwrap_or(false);
+    if no_local_instances && no_web_clients {
+        let _ = stop_local_socket_server_inner();
+    }
+}
+
+fn run_config_server_instance(config_id: &str, config: &NetworkConfig) -> bool {
+    if INSTANCE_MANAGER
+        .list_network_instance_ids()
+        .iter()
+        .next()
+        .is_some()
+    {
+        hilog_error!("[Rust] there is a running instance!");
+        return false;
+    }
+
+    let Some(config_server_url) = config.public_server_url.clone() else {
+        hilog_error!("[Rust] public_server_url missing for config server mode");
+        return false;
+    };
+    let hooks = Arc::new(TrackedWebClientHooks::default());
+    let secure_mode = config
+        .secure_mode
+        .as_ref()
+        .map(|mode| mode.enabled)
+        .unwrap_or(false);
+    let hostname = config.hostname.clone();
+
+    if !ensure_local_socket_server_started() {
+        return false;
+    }
+
+    let client = ASYNC_RUNTIME.block_on(run_web_client(
+        &config_server_url,
+        MachineIdOptions::default(),
+        hostname,
+        secure_mode,
+        INSTANCE_MANAGER.clone(),
+        Some(hooks.clone()),
+    ));
+
+    let client = match client {
+        Ok(client) => client,
+        Err(err) => {
+            hilog_error!("[Rust] start config server failed {}", err);
+            return false;
+        }
+    };
+
+    match WEB_CLIENTS.lock() {
+        Ok(mut guard) => {
+            guard.insert(
+                config_id.to_string(),
+                ManagedWebClient {
+                    _client: client,
+                    hooks,
+                },
+            );
+            true
+        }
+        Err(err) => {
+            hilog_error!("[Rust] store config server client failed {}", err);
             false
         }
     }
 }
 
-#[napi]
-pub fn default_network_config() -> String {
-    match NetworkConfig::new_from_config(TomlConfigLoader::default()) {
-        Ok(result) => serde_json::to_string(&result).unwrap_or_else(|e| format!("ERROR {}", e)),
-        Err(e) => {
-            hilog_error!("[Rust] default_network_config failed {}", e);
-            format!("ERROR {}", e)
-        }
-    }
+pub(crate) fn build_default_network_config_json() -> Result<String, String> {
+    let config = NetworkConfig::new_from_config(TomlConfigLoader::default())
+        .map_err(|e| format!("default_network_config failed {}", e))?;
+    serde_json::to_string(&config).map_err(|e| format!("default_network_config failed {}", e))
 }
 
-#[napi]
-pub fn convert_toml_to_network_config(cfg_str: String) -> String {
-    match TomlConfigLoader::new_from_str(&cfg_str) {
-        Ok(cfg) => match NetworkConfig::new_from_config(cfg) {
-            Ok(result) => serde_json::to_string(&result).unwrap_or_else(|e| format!("ERROR {}", e)),
-            Err(e) => {
-                hilog_error!("[Rust] convert_toml_to_network_config failed {}", e);
-                format!("ERROR {}", e)
-            }
-        },
-        Err(e) => {
-            hilog_error!("[Rust] convert_toml_to_network_config failed {}", e);
-            format!("ERROR {}", e)
-        }
-    }
+fn convert_toml_to_network_config_inner(toml_text: &str) -> Result<String, String> {
+    let config = NetworkConfig::new_from_config(
+        TomlConfigLoader::new_from_str(toml_text).map_err(|e| e.to_string())?,
+    )
+    .map_err(|e| e.to_string())?;
+    serde_json::to_string(&config).map_err(|e| e.to_string())
 }
 
-#[napi]
-pub fn parse_network_config(cfg_json: String) -> bool {
-    match serde_json::from_str::<NetworkConfig>(&cfg_json) {
-        Ok(cfg) => match cfg.gen_config() {
-            Ok(toml) => {
-                hilog_debug!("[Rust] Convert to Toml {}", toml.dump());
-                true
-            }
-            Err(e) => {
-                hilog_error!("[Rust] parse config failed {}", e);
-                false
-            }
-        },
-        Err(e) => {
-            hilog_error!("[Rust] parse config failed {}", e);
-            false
-        }
-    }
+fn parse_network_config_inner(cfg_json: &str) -> bool {
+    serde_json::from_str::<NetworkConfig>(cfg_json)
+        .ok()
+        .and_then(|cfg| cfg.gen_config().ok())
+        .is_some()
 }
 
-#[napi]
-pub fn run_network_instance(cfg_json: String) -> bool {
-    let cfg = match serde_json::from_str::<NetworkConfig>(&cfg_json) {
-        Ok(cfg) => match cfg.gen_config() {
-            Ok(toml) => toml,
-            Err(e) => {
-                hilog_error!("[Rust] parse config failed {}", e);
-                return false;
-            }
-        },
+pub(crate) fn run_network_instance_from_json(cfg_json: &str) -> bool {
+    let config = match serde_json::from_str::<NetworkConfig>(cfg_json) {
+        Ok(cfg) => cfg,
         Err(e) => {
             hilog_error!("[Rust] parse config failed {}", e);
             return false;
         }
     };
 
-    if INSTANCE_MANAGER.list_network_instance_ids().len() > 0 {
+    if is_config_server_config(&config) {
+        let Some(config_id) = config.instance_id.as_deref() else {
+            hilog_error!("[Rust] config server config missing instance id");
+            return false;
+        };
+        return run_config_server_instance(config_id, &config);
+    }
+
+    let cfg = match config.gen_config() {
+        Ok(toml) => toml,
+        Err(e) => {
+            hilog_error!("[Rust] parse config failed {}", e);
+            return false;
+        }
+    };
+
+    if !INSTANCE_MANAGER.list_network_instance_ids().is_empty() {
         hilog_error!("[Rust] there is a running instance!");
         return false;
     }
 
+    if !ensure_local_socket_server_started() {
+        return false;
+    }
+
     let inst_id = cfg.get_id();
     if INSTANCE_MANAGER
         .list_network_instance_ids()
         .contains(&inst_id)
     {
+        hilog_error!("[Rust] instance {} already exists", inst_id);
         return false;
     }
-    INSTANCE_MANAGER
-        .run_network_instance(cfg, false, ConfigFileControl::STATIC_CONFIG)
-        .unwrap();
-    true
-}
 
-#[napi]
-pub fn stop_network_instance(inst_names: Vec<String>) {
-    INSTANCE_MANAGER
-        .delete_network_instance(
-            inst_names
-                .into_iter()
-                .filter_map(|s| Uuid::parse_str(&s).ok())
-                .collect(),
-        )
-        .unwrap();
-    hilog_debug!("[Rust] stop_network_instance");
-}
-
-#[napi]
-pub fn collect_network_infos() -> Vec<KeyValuePair> {
-    let mut result = Vec::new();
-    match INSTANCE_MANAGER.collect_network_infos_sync() {
-        Ok(map) => {
-            for (uuid, info) in map.iter() {
-                // convert value to json string
-                let value = match serde_json::to_string(&info) {
-                    Ok(value) => value,
-                    Err(e) => {
-                        hilog_error!("[Rust] failed to serialize instance {} info: {}", uuid, e);
-                        continue;
-                    }
-                };
-                result.push(KeyValuePair {
-                    key: uuid.clone().to_string(),
-                    value: value.clone(),
-                });
-            }
-        }
-        Err(_) => {}
-    }
-    result
-}
-
-#[napi]
-pub fn collect_running_network() -> Vec<String> {
-    INSTANCE_MANAGER
-        .list_network_instance_ids()
-        .clone()
-        .into_iter()
-        .map(|id| id.to_string())
-        .collect()
-}
-
-#[napi]
-pub fn is_running_network(inst_id: String) -> bool {
-    match Uuid::try_parse(&inst_id) {
-        Ok(uuid) => INSTANCE_MANAGER.list_network_instance_ids().contains(&uuid),
-        Err(e) => {
-            hilog_error!("[Rust] cant covert {} to uuid. {}", inst_id, e);
+    match INSTANCE_MANAGER.run_network_instance(cfg, false, ConfigFileControl::STATIC_CONFIG) {
+        Ok(_) => true,
+        Err(err) => {
+            hilog_error!("[Rust] start_kernel failed for {}: {}", inst_id, err);
             false
         }
     }
 }
+
+fn parse_instance_uuid(config_id: &str) -> Option<Uuid> {
+    match Uuid::parse_str(config_id) {
+        Ok(uuid) => Some(uuid),
+        Err(err) => {
+            hilog_error!("[Rust] invalid config_id {}: {}", config_id, err);
+            None
+        }
+    }
+}
+
+#[napi]
+pub fn init_config_store(root_dir: String) -> bool {
+    exports::config_api::init_config_store(root_dir)
+}
+
+#[napi]
+pub fn list_configs() -> String {
+    exports::config_api::list_configs()
+}
+
+#[napi]
+pub fn get_config_display_name_by_id(config_id: String) -> Option<String> {
+    get_config_display_name(&config_id)
+}
+
+#[napi]
+pub fn save_config(config_id: String, display_name: String, config_json: String) -> bool {
+    exports::config_api::save_config(config_id, display_name, config_json)
+}
+
+#[napi]
+pub fn create_config(config_id: String, display_name: String) -> bool {
+    exports::config_api::create_config(config_id, display_name)
+}
+
+#[napi]
+pub fn rename_stored_config(config_id: String, display_name: String) -> bool {
+    config::storage::config_meta::set_config_display_name(config_id, display_name).is_some()
+}
+
+#[napi]
+pub fn delete_stored_config_meta(config_id: String) -> bool {
+    exports::config_api::delete_stored_config_meta(config_id)
+}
+
+#[napi]
+pub fn get_config(config_id: String) -> Option<String> {
+    exports::config_api::get_config(config_id)
+}
+
+#[napi]
+pub fn get_default_config() -> Option<String> {
+    exports::config_api::get_default_config()
+}
+
+#[napi]
+pub fn get_config_field(config_id: String, field: String) -> Option<String> {
+    exports::config_api::get_config_field(config_id, field)
+}
+
+#[napi]
+pub fn set_config_field(config_id: String, field: String, json_value: String) -> bool {
+    exports::config_api::set_config_field(config_id, field, json_value)
+}
+
+#[napi]
+pub fn import_toml(toml_text: String, display_name: Option<String>) -> Option<String> {
+    exports::config_api::import_toml(toml_text, display_name)
+}
+
+#[napi]
+pub fn export_toml(config_id: String) -> Option<String> {
+    exports::config_api::export_toml(config_id)
+}
+
+#[napi]
+pub fn start_kernel(config_id: String) -> bool {
+    exports::runtime_api::start_kernel(config_id, start_kernel_with_config_id)
+}
+
+#[napi]
+pub fn stop_kernel(config_id: String) -> bool {
+    exports::runtime_api::stop_kernel(
+        config_id,
+        stop_web_client,
+        parse_instance_uuid,
+        maybe_stop_local_socket_server,
+    )
+}
+
+#[napi]
+pub fn stop_network_instance(config_ids: Vec<String>) -> bool {
+    exports::runtime_api::stop_network_instance(config_ids, stop_kernel)
+}
+
+#[napi]
+pub fn easytier_version() -> String {
+    EASYTIER_VERSION.to_string()
+}
+
+#[napi]
+pub fn default_network_config() -> String {
+    get_default_config().unwrap_or_else(|| "{}".to_string())
+}
+
+#[napi]
+pub fn convert_toml_to_network_config(toml_text: String) -> String {
+    convert_toml_to_network_config_inner(&toml_text).unwrap_or_else(|err| format!("ERROR: {err}"))
+}
+
+#[napi]
+pub fn parse_network_config(cfg_json: String) -> bool {
+    parse_network_config_inner(&cfg_json)
+}
+
+#[napi]
+pub fn run_network_instance(cfg_json: String) -> bool {
+    run_network_instance_from_json(&cfg_json)
+}
+
+#[napi]
+pub fn collect_network_infos() -> Vec<KeyValuePair> {
+    exports::runtime_api::collect_network_infos()
+}
+
+#[napi]
+pub fn set_tun_fd(config_id: String, fd: i32) -> bool {
+    exports::runtime_api::set_tun_fd(config_id, fd, parse_instance_uuid)
+}
+
+#[napi]
+pub fn get_network_config_schema() -> NetworkConfigSchema {
+    build_network_config_schema()
+}
+
+#[napi]
+pub fn get_network_config_field_mappings() -> Vec<ConfigFieldMapping> {
+    build_network_config_field_mappings()
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn exported_plain_object_schema_contains_core_networkconfig_metadata() {
+        let schema = get_network_config_schema();
+        assert_eq!(schema.name, "NetworkConfig");
+        assert_eq!(schema.node_kind, "schema");
+        assert!(
+            schema
+                .children
+                .iter()
+                .any(|field| field.name == "network_name")
+        );
+        let secure_mode = schema
+            .children
+            .iter()
+            .find(|field| field.name == "secure_mode")
+            .expect("secure_mode field");
+        assert!(
+            secure_mode
+                .children
+                .iter()
+                .any(|field| field.name == "enabled")
+        );
+    }
+}
+
+#[napi]
+pub fn get_runtime_snapshot() -> RuntimeAggregateState {
+    exports::runtime_api::get_runtime_snapshot()
+}
+
+pub(crate) fn get_runtime_snapshot_inner() -> RuntimeAggregateState {
+    exports::runtime_api::get_runtime_snapshot_inner()
+}
+
+#[napi]
+pub fn build_config_share_link(config_id: String, only_start: Option<bool>) -> Option<String> {
+    build_config_share_link_inner(&config_id, None, only_start.unwrap_or(false))
+}
+
+#[napi]
+pub fn parse_config_share_link(share_link: String) -> Option<SharedConfigLinkPayload> {
+    parse_config_share_link_inner(&share_link)
+}
+
+#[napi]
+pub fn import_config_share_link(
+    share_link: String,
+    display_name_override: Option<String>,
+) -> Option<String> {
+    import_config_share_link_inner(&share_link, display_name_override)
+}

easytier-contrib/easytier-ohrs/src/platform.rs

@@ -0,0 +1 @@
+pub(crate) mod logging;

easytier-contrib/easytier-ohrs/src/platform/logging/mod.rs

@@ -0,0 +1 @@
+pub(crate) mod native_log;

easytier-contrib/easytier-ohrs/src/runtime.rs

@@ -0,0 +1 @@
+pub(crate) mod state;

easytier-contrib/easytier-ohrs/src/runtime/state/mod.rs

@@ -0,0 +1 @@
+pub(crate) mod runtime_state;

easytier-contrib/easytier-ohrs/src/runtime/state/runtime_state.rs

@@ -0,0 +1,293 @@
+use easytier::proto::{api, common};
+use napi_derive_ohos::napi;
+use serde::Serialize;
+use std::collections::HashSet;
+use std::sync::Mutex;
+
+static ATTACHED_TUN_INSTANCE_IDS: once_cell::sync::Lazy<Mutex<HashSet<String>>> =
+    once_cell::sync::Lazy::new(|| Mutex::new(HashSet::new()));
+
+pub fn mark_tun_attached(instance_id: &str) {
+    if let Ok(mut guard) = ATTACHED_TUN_INSTANCE_IDS.lock() {
+        guard.insert(instance_id.to_string());
+    }
+}
+
+pub fn clear_tun_attached(instance_id: &str) {
+    if let Ok(mut guard) = ATTACHED_TUN_INSTANCE_IDS.lock() {
+        guard.remove(instance_id);
+    }
+}
+
+pub fn is_tun_attached(instance_id: &str) -> bool {
+    ATTACHED_TUN_INSTANCE_IDS
+        .lock()
+        .map(|guard| guard.contains(instance_id))
+        .unwrap_or(false)
+}
+
+#[derive(Serialize)]
+#[serde(rename_all = "camelCase")]
+#[napi(object)]
+pub struct PeerConnStats {
+    pub rx_bytes: i64,
+    pub tx_bytes: i64,
+    pub rx_packets: i64,
+    pub tx_packets: i64,
+    pub latency_us: i64,
+}
+
+#[derive(Serialize)]
+#[serde(rename_all = "camelCase")]
+#[napi(object)]
+pub struct PeerConnInfo {
+    pub conn_id: String,
+    pub my_peer_id: i64,
+    pub peer_id: i64,
+    pub features: Vec<String>,
+    pub tunnel_type: Option<String>,
+    pub local_addr: Option<String>,
+    pub remote_addr: Option<String>,
+    pub resolved_remote_addr: Option<String>,
+    pub stats: Option<PeerConnStats>,
+    pub loss_rate: Option<f64>,
+    pub is_client: bool,
+    pub network_name: Option<String>,
+    pub is_closed: bool,
+    pub secure_auth_level: Option<i32>,
+    pub peer_identity_type: Option<i32>,
+}
+
+#[derive(Serialize)]
+#[serde(rename_all = "camelCase")]
+#[napi(object)]
+pub struct PeerInfo {
+    pub peer_id: i64,
+    pub default_conn_id: Option<String>,
+    pub directly_connected_conns: Vec<String>,
+    pub conns: Vec<PeerConnInfo>,
+}
+
+#[derive(Serialize)]
+#[serde(rename_all = "camelCase")]
+#[napi(object)]
+pub struct RouteView {
+    pub peer_id: i64,
+    pub hostname: Option<String>,
+    pub ipv4: Option<String>,
+    pub ipv4_cidr: Option<String>,
+    pub ipv6_cidr: Option<String>,
+    pub proxy_cidrs: Vec<String>,
+    pub next_hop_peer_id: Option<i64>,
+    pub cost: Option<i32>,
+    pub path_latency: Option<i64>,
+    pub udp_nat_type: Option<i32>,
+    pub tcp_nat_type: Option<i32>,
+    pub inst_id: Option<String>,
+    pub version: Option<String>,
+    pub is_public_server: Option<bool>,
+}
+
+#[derive(Serialize)]
+#[serde(rename_all = "camelCase")]
+#[napi(object)]
+pub struct MyNodeInfo {
+    pub virtual_ipv4: Option<String>,
+    pub virtual_ipv4_cidr: Option<String>,
+    pub hostname: Option<String>,
+    pub version: Option<String>,
+    pub peer_id: Option<i64>,
+    pub listeners: Vec<String>,
+    pub vpn_portal_cfg: Option<String>,
+    pub udp_nat_type: Option<i32>,
+    pub tcp_nat_type: Option<i32>,
+}
+
+#[derive(Serialize)]
+#[serde(rename_all = "camelCase")]
+#[napi(object)]
+pub struct RuntimeInstanceState {
+    pub config_id: String,
+    pub instance_id: String,
+    pub display_name: String,
+    pub running: bool,
+    pub tun_required: bool,
+    pub tun_attached: bool,
+    pub magic_dns_enabled: bool,
+    pub need_exit_node: bool,
+    pub error_message: Option<String>,
+    pub my_node_info: Option<MyNodeInfo>,
+    pub events: Vec<String>,
+    pub routes: Vec<RouteView>,
+    pub peers: Vec<PeerInfo>,
+}
+
+#[derive(Serialize)]
+#[serde(rename_all = "camelCase")]
+#[napi(object)]
+pub struct TunAggregateState {
+    pub active: bool,
+    pub attached_instance_ids: Vec<String>,
+    pub aggregated_routes: Vec<String>,
+    pub dns_servers: Vec<String>,
+    pub need_rebuild: bool,
+}
+
+#[derive(Serialize)]
+#[serde(rename_all = "camelCase")]
+#[napi(object)]
+pub struct RuntimeAggregateState {
+    pub instances: Vec<RuntimeInstanceState>,
+    pub tun: TunAggregateState,
+    pub running_instance_count: i32,
+}
+
+fn stringify_ipv4_inet(value: Option<common::Ipv4Inet>) -> Option<String> {
+    value.map(|v| v.to_string())
+}
+
+fn stringify_ipv6_inet(value: Option<common::Ipv6Inet>) -> Option<String> {
+    value.map(|v| v.to_string())
+}
+
+fn stringify_url(value: Option<common::Url>) -> Option<String> {
+    value.map(|v| v.to_string())
+}
+
+fn stringify_uuid(value: Option<common::Uuid>) -> Option<String> {
+    value.map(|v| v.to_string())
+}
+
+fn optional_u32_to_i64(value: Option<u32>) -> Option<i64> {
+    value.map(|v| v as i64)
+}
+
+fn optional_i32_to_i64(value: Option<i32>) -> Option<i64> {
+    value.map(|v| v as i64)
+}
+
+fn route_to_view(route: api::instance::Route) -> RouteView {
+    let stun = route.stun_info;
+    let feature_flag = route.feature_flag;
+    RouteView {
+        peer_id: route.peer_id as i64,
+        hostname: (!route.hostname.is_empty()).then_some(route.hostname),
+        ipv4: route
+            .ipv4_addr
+            .as_ref()
+            .and_then(|inet| inet.address.as_ref())
+            .map(|addr| addr.to_string()),
+        ipv4_cidr: stringify_ipv4_inet(route.ipv4_addr),
+        ipv6_cidr: stringify_ipv6_inet(route.ipv6_addr),
+        proxy_cidrs: route.proxy_cidrs,
+        next_hop_peer_id: optional_u32_to_i64(route.next_hop_peer_id_latency_first)
+            .or_else(|| Some(route.next_hop_peer_id as i64)),
+        cost: Some(route.cost),
+        path_latency: optional_i32_to_i64(route.path_latency_latency_first)
+            .or_else(|| Some(route.path_latency as i64)),
+        udp_nat_type: stun.as_ref().map(|info| info.udp_nat_type),
+        tcp_nat_type: stun.as_ref().map(|info| info.tcp_nat_type),
+        inst_id: (!route.inst_id.is_empty()).then_some(route.inst_id),
+        version: (!route.version.is_empty()).then_some(route.version),
+        is_public_server: feature_flag.map(|flag| flag.is_public_server),
+    }
+}
+
+fn peer_conn_to_view(conn: api::instance::PeerConnInfo) -> PeerConnInfo {
+    let stats = conn.stats.map(|stats| PeerConnStats {
+        rx_bytes: stats.rx_bytes as i64,
+        tx_bytes: stats.tx_bytes as i64,
+        rx_packets: stats.rx_packets as i64,
+        tx_packets: stats.tx_packets as i64,
+        latency_us: stats.latency_us as i64,
+    });
+
+    PeerConnInfo {
+        conn_id: conn.conn_id,
+        my_peer_id: conn.my_peer_id as i64,
+        peer_id: conn.peer_id as i64,
+        features: conn.features,
+        tunnel_type: conn.tunnel.as_ref().map(|t| t.tunnel_type.clone()),
+        local_addr: conn
+            .tunnel
+            .as_ref()
+            .and_then(|t| stringify_url(t.local_addr.clone())),
+        remote_addr: conn
+            .tunnel
+            .as_ref()
+            .and_then(|t| stringify_url(t.remote_addr.clone())),
+        resolved_remote_addr: conn
+            .tunnel
+            .as_ref()
+            .and_then(|t| stringify_url(t.resolved_remote_addr.clone())),
+        stats,
+        loss_rate: Some(conn.loss_rate as f64),
+        is_client: conn.is_client,
+        network_name: (!conn.network_name.is_empty()).then_some(conn.network_name),
+        is_closed: conn.is_closed,
+        secure_auth_level: Some(conn.secure_auth_level),
+        peer_identity_type: Some(conn.peer_identity_type),
+    }
+}
+
+fn peer_to_view(peer: api::instance::PeerInfo) -> PeerInfo {
+    PeerInfo {
+        peer_id: peer.peer_id as i64,
+        default_conn_id: stringify_uuid(peer.default_conn_id),
+        directly_connected_conns: peer
+            .directly_connected_conns
+            .into_iter()
+            .map(|id| id.to_string())
+            .collect(),
+        conns: peer.conns.into_iter().map(peer_conn_to_view).collect(),
+    }
+}
+
+fn my_node_info_to_view(info: api::manage::MyNodeInfo) -> MyNodeInfo {
+    MyNodeInfo {
+        virtual_ipv4: info
+            .virtual_ipv4
+            .as_ref()
+            .and_then(|inet| inet.address.as_ref())
+            .map(|addr| addr.to_string()),
+        virtual_ipv4_cidr: stringify_ipv4_inet(info.virtual_ipv4),
+        hostname: (!info.hostname.is_empty()).then_some(info.hostname),
+        version: (!info.version.is_empty()).then_some(info.version),
+        peer_id: Some(info.peer_id as i64),
+        listeners: info
+            .listeners
+            .into_iter()
+            .map(|url| url.to_string())
+            .collect(),
+        vpn_portal_cfg: info.vpn_portal_cfg,
+        udp_nat_type: info.stun_info.as_ref().map(|stun| stun.udp_nat_type),
+        tcp_nat_type: info.stun_info.as_ref().map(|stun| stun.tcp_nat_type),
+    }
+}
+
+pub fn runtime_instance_from_running_info(
+    config_id: String,
+    display_name: String,
+    magic_dns_enabled: bool,
+    need_exit_node: bool,
+    info: api::manage::NetworkInstanceRunningInfo,
+) -> RuntimeInstanceState {
+    let tun_attached = info.running && is_tun_attached(&config_id);
+    let tun_required = info.running && (info.dev_name != "no_tun" || tun_attached);
+
+    RuntimeInstanceState {
+        config_id: config_id.clone(),
+        instance_id: config_id,
+        display_name,
+        running: info.running,
+        tun_required,
+        tun_attached,
+        magic_dns_enabled,
+        need_exit_node,
+        error_message: info.error_msg,
+        my_node_info: info.my_node_info.map(my_node_info_to_view),
+        events: info.events,
+        routes: info.routes.into_iter().map(route_to_view).collect(),
+        peers: info.peers.into_iter().map(peer_to_view).collect(),
+    }
+}

easytier-gui/package.json

@@ -1,7 +1,7 @@
 {
   "name": "easytier-gui",
   "type": "module",
-  "version": "2.6.3",
+  "version": "2.6.4",
   "private": true,
   "packageManager": "pnpm@9.12.1+sha512.e5a7e52a4183a02d5931057f7a0dbff9d5e9ce3161e33fa68ae392125b79282a8a8a470a51dfc8a0ed86221442eb2fb57019b0990ed24fab519bf0e1bc5ccfc4",
   "scripts": {

easytier-gui/src-tauri/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "easytier-gui"
-version = "2.6.3"
+version = "2.6.4"
 description = "EasyTier GUI"
 authors = ["you"]
 edition.workspace = true

easytier-gui/src-tauri/src/lib.rs

@@ -15,9 +15,7 @@ use easytier::rpc_service::remote_client::{
 use easytier::web_client::{self, WebClient};
 use easytier::{
     common::{
-        config::{
-            ConfigLoader, ConfigSource, FileLoggerConfig, LoggingConfigBuilder, TomlConfigLoader,
-        },
+        config::{ConfigLoader, ConfigSource, FileLoggerConfig, LoggingConfig, TomlConfigLoader},
         log,
     },
     instance_manager::NetworkInstanceManager,
@@ -490,10 +488,18 @@ async fn init_web_client(app: AppHandle, url: Option<String>) -> Result<(), Stri
         .ok_or_else(|| "Instance manager is not available".to_string())?;
 
     let hooks = Arc::new(manager::GuiHooks { app: app.clone() });
+    let machine_id_state_dir = app
+        .path()
+        .app_data_dir()
+        .with_context(|| "Failed to resolve machine id state directory")
+        .map_err(|e| format!("{:#}", e))?;
 
     let web_client = web_client::run_web_client(
         url.as_str(),
-        None,
+        easytier::common::MachineIdOptions {
+            explicit_machine_id: None,
+            state_dir: Some(machine_id_state_dir),
+        },
         None,
         false,
         instance_manager,
@@ -1318,7 +1324,7 @@ pub fn run_gui() -> std::process::ExitCode {
             let Ok(log_dir) = get_log_dir(app.app_handle()) else {
                 return Ok(());
             };
-            let config = LoggingConfigBuilder::default()
+            let config = LoggingConfig::builder()
                 .file_logger(FileLoggerConfig {
                     dir: Some(log_dir.to_string_lossy().to_string()),
                     level: None,
@@ -1326,8 +1332,7 @@ pub fn run_gui() -> std::process::ExitCode {
                     size_mb: None,
                     count: None,
                 })
-                .build()
-                .map_err(|e| e.to_string())?;
+                .build();
             let Ok(_) = log::init(&config, true) else {
                 return Ok(());
             };

easytier-gui/src-tauri/tauri.conf.json

@@ -17,7 +17,7 @@
     "createUpdaterArtifacts": false
   },
   "productName": "easytier-gui",
-  "version": "2.6.3",
+  "version": "2.6.4",
   "identifier": "com.kkrainbow.easytier",
   "plugins": {
     "shell": {

easytier-rpc-build/Cargo.toml

@@ -1,20 +0,0 @@
-[package]
-name = "easytier-rpc-build"
-description = "Protobuf RPC Service Generator for EasyTier"
-version = "0.1.0"
-edition.workspace = true
-homepage = "https://github.com/EasyTier/EasyTier"
-repository = "https://github.com/EasyTier/EasyTier"
-authors = ["kkrainbow"]
-keywords = ["vpn", "p2p", "network", "easytier"]
-categories = ["network-programming", "command-line-utilities"]
-license-file = "LICENSE"
-readme = "README.md"
-
-[dependencies]
-heck = "0.5"
-prost-build = "0.13"
-
-[features]
-default = []
-internal-namespace = []

easytier-rpc-build/LICENSE

@@ -1 +0,0 @@
-../LICENSE

easytier-rpc-build/README.md

@@ -1,3 +0,0 @@
-# Introduction
-
-This is a protobuf rpc service stub generator for [EasyTier](https://github.com/EasyTier/EasyTier) project.

easytier-rpc-build/src/lib.rs

@@ -1,449 +0,0 @@
-extern crate heck;
-extern crate prost_build;
-
-use std::fmt;
-
-#[cfg(feature = "internal-namespace")]
-const NAMESPACE: &str = "crate::proto::rpc_types";
-
-#[cfg(not(feature = "internal-namespace"))]
-const NAMESPACE: &str = "easytier::proto::rpc_types";
-
-/// The service generator to be used with `prost-build` to generate RPC implementations for
-/// `prost-simple-rpc`.
-///
-/// See the crate-level documentation for more info.
-#[allow(missing_copy_implementations)]
-#[derive(Clone, Debug, Default)]
-pub struct ServiceGenerator {
-    _private: (),
-}
-
-impl prost_build::ServiceGenerator for ServiceGenerator {
-    fn generate(&mut self, service: prost_build::Service, mut buf: &mut String) {
-        use std::fmt::Write;
-
-        let descriptor_name = format!("{}Descriptor", service.name);
-        let server_name = format!("{}Server", service.name);
-        let client_name = format!("{}Client", service.name);
-        let method_descriptor_name = format!("{}MethodDescriptor", service.name);
-
-        let mut trait_methods = String::new();
-        let mut weak_impl_methods = String::new();
-        let mut enum_methods = String::new();
-        let mut list_enum_methods = String::new();
-        let mut client_methods = String::new();
-        let mut client_own_methods = String::new();
-        let mut match_name_methods = String::new();
-        let mut match_proto_name_methods = String::new();
-        let mut match_input_type_methods = String::new();
-        let mut match_input_proto_type_methods = String::new();
-        let mut match_output_type_methods = String::new();
-        let mut match_output_proto_type_methods = String::new();
-        let mut match_handle_methods = String::new();
-        // generate trait default method Xxx::json_call_method match branch
-        let mut match_trait_json_methods = String::new();
-
-        let mut match_method_try_from = String::new();
-
-        for (idx, method) in service.methods.iter().enumerate() {
-            assert!(
-                !method.client_streaming,
-                "Client streaming not yet supported for method {}",
-                method.proto_name
-            );
-            assert!(
-                !method.server_streaming,
-                "Server streaming not yet supported for method {}",
-                method.proto_name
-            );
-
-            ServiceGenerator::write_comments(&mut trait_methods, 4, &method.comments).unwrap();
-            writeln!(
-                trait_methods,
-                r#"    async fn {name}(&self, ctrl: Self::Controller, input: {input_type}) -> {namespace}::error::Result<{output_type}>;"#,
-                name = method.name,
-                input_type = method.input_type,
-                output_type = method.output_type,
-                namespace = NAMESPACE,
-            )
-            .unwrap();
-
-            writeln!(
-                weak_impl_methods,
-                r#"    async fn {method_name}(&self, ctrl: Self::Controller, input: {input_type}) -> {namespace}::error::Result<{output_type}> {{
-        let Some(service) = self.upgrade() else {{
-            return Err({namespace}::error::Error::Shutdown);
-        }};
-        service.{method_name}(ctrl, input).await
-    }}"#,
-                method_name = method.name,
-                input_type = method.input_type,
-                output_type = method.output_type,
-                namespace = NAMESPACE,
-            )
-            .unwrap();
-
-            ServiceGenerator::write_comments(&mut enum_methods, 4, &method.comments).unwrap();
-            writeln!(
-                enum_methods,
-                "    {name} = {index},",
-                name = method.proto_name,
-                index = idx + 1
-            )
-            .unwrap();
-
-            writeln!(
-                match_method_try_from,
-                "            {index} => Ok({service_name}MethodDescriptor::{name}),",
-                service_name = service.name,
-                name = method.proto_name,
-                index = idx + 1,
-            )
-            .unwrap();
-
-            writeln!(
-                list_enum_methods,
-                "            {service_name}MethodDescriptor::{name},",
-                service_name = service.name,
-                name = method.proto_name
-            )
-            .unwrap();
-
-            writeln!(
-                client_methods,
-                r#"    async fn {name}(&self, ctrl: H::Controller, input: {input_type}) -> {namespace}::error::Result<{output_type}> {{
-        {client_name}Client::{name}_inner(self.0.clone(), ctrl, input).await
-    }}"#,
-                name = method.name,
-                input_type = method.input_type,
-                output_type = method.output_type,
-                client_name = service.name,
-                namespace = NAMESPACE,
-            )
-            .unwrap();
-
-            writeln!(
-                client_own_methods,
-                r#"    async fn {name}_inner(handler: H, ctrl: H::Controller, input: {input_type}) -> {namespace}::error::Result<{output_type}> {{
-            {namespace}::__rt::call_method(handler, ctrl, {method_descriptor_name}::{proto_name}, input).await
-    }}"#,
-                name = method.name,
-                method_descriptor_name = method_descriptor_name,
-                proto_name = method.proto_name,
-                input_type = method.input_type,
-                output_type = method.output_type,
-                namespace = NAMESPACE,
-            ).unwrap();
-
-            let case = format!(
-                "            {service_name}MethodDescriptor::{proto_name} => ",
-                service_name = service.name,
-                proto_name = method.proto_name
-            );
-
-            writeln!(match_name_methods, "{}{:?},", case, method.name).unwrap();
-            writeln!(match_proto_name_methods, "{}{:?},", case, method.proto_name).unwrap();
-            writeln!(
-                match_input_type_methods,
-                "{}::std::any::TypeId::of::<{}>(),",
-                case, method.input_type
-            )
-            .unwrap();
-            writeln!(
-                match_input_proto_type_methods,
-                "{}{:?},",
-                case, method.input_proto_type
-            )
-            .unwrap();
-            writeln!(
-                match_output_type_methods,
-                "{}::std::any::TypeId::of::<{}>(),",
-                case, method.output_type
-            )
-            .unwrap();
-            writeln!(
-                match_output_proto_type_methods,
-                "{}{:?},",
-                case, method.output_proto_type
-            )
-            .unwrap();
-            write!(
-                match_handle_methods,
-                r#"{} {{
-                    let decoded: {input_type} = {namespace}::__rt::decode(input)?;
-                    let ret = service.{name}(ctrl, decoded).await?;
-                    {namespace}::__rt::encode(ret)
-                }}
-"#,
-                case,
-                input_type = method.input_type,
-                name = method.name,
-                namespace = NAMESPACE,
-            )
-            .unwrap();
-
-            write!(
-                match_trait_json_methods,
-                r#"            "{name}" | "{proto_name}" => {{
-                let req: {input_type} = ::serde_json::from_value(json).map_err(|e| {namespace}::error::Error::MalformatRpcPacket(format!("json error: {{}}", e)))?;
-                let resp = self.{typed_method}(ctrl, req).await?;
-                Ok(::serde_json::to_value(resp).map_err(|e| {namespace}::error::Error::MalformatRpcPacket(format!("json error: {{}}", e)))?)
-            }}
-"#,
-                name = method.name,
-                proto_name = method.proto_name,
-                input_type = method.input_type,
-                typed_method = method.name,
-                namespace = NAMESPACE,
-            )
-            .unwrap();
-        }
-
-        ServiceGenerator::write_comments(&mut buf, 0, &service.comments).unwrap();
-        write!(
-            buf,
-            r#"
-#[async_trait::async_trait]
-#[auto_impl::auto_impl(&, Arc, Box)]
-pub trait {name} {{
-    type Controller: {namespace}::controller::Controller;
-
-    {trait_methods}
-
-    async fn json_call_method(
-        &self,
-        ctrl: Self::Controller,
-        method_name: &str,
-        json: ::serde_json::Value,
-    ) -> {namespace}::error::Result<::serde_json::Value> {{
-        match method_name {{
-{match_trait_json_methods}
-            _ => Err({namespace}::error::Error::InvalidMethodIndex(0, method_name.to_string())),
-        }}
-    }}
-}}
-
-#[async_trait::async_trait]
-impl<T> {name} for ::std::sync::Weak<T>
-where
-    T: Send + Sync + 'static,
-    ::std::sync::Arc<T>: {name},
-{{
-    type Controller = <::std::sync::Arc<T> as {name}>::Controller;
-
-    {weak_impl_methods}
-}}
-
-/// A service descriptor for a `{name}`.
-#[derive(Clone, Debug, Eq, Ord, PartialEq, PartialOrd, Default)]
-pub struct {descriptor_name};
-
-/// Methods available on a `{name}`.
-///
-/// This can be used as a key when routing requests for servers/clients of a `{name}`.
-#[derive(Clone, Copy, Debug, Eq, Ord, PartialEq, PartialOrd)]
-#[repr(u8)]
-pub enum {method_descriptor_name} {{
-    {enum_methods}
-}}
-
-impl std::convert::TryFrom<u8> for {method_descriptor_name} {{
-    type Error = {namespace}::error::Error;
-    fn try_from(value: u8) -> {namespace}::error::Result<Self> {{
-        match value {{
-            {match_method_try_from}
-            _ => Err({namespace}::error::Error::InvalidMethodIndex(value, "{name}".to_string())),
-        }}
-    }}
-}}
-
-/// A client for a `{name}`.
-///
-/// This implements the `{name}` trait by dispatching all method calls to the supplied `Handler`.
-#[derive(Clone, Debug)]
-pub struct {client_name}<H>(H) where H: {namespace}::handler::Handler;
-
-impl<H> {client_name}<H> where H: {namespace}::handler::Handler<Descriptor = {descriptor_name}> {{
-    /// Creates a new client instance that delegates all method calls to the supplied handler.
-    pub fn new(handler: H) -> {client_name}<H> {{
-        {client_name}(handler)
-    }}
-}}
-
-impl<H> {client_name}<H> where H: {namespace}::handler::Handler<Descriptor = {descriptor_name}> {{
-    {client_own_methods}
-}}
-
-#[async_trait::async_trait]
-impl<H> {name} for {client_name}<H> where H: {namespace}::handler::Handler<Descriptor = {descriptor_name}> {{
-    type Controller = H::Controller;
-
-    {client_methods}
-}}
-
-pub struct {client_name}Factory<C: {namespace}::controller::Controller>(std::marker::PhantomData<C>);
-
-impl<C: {namespace}::controller::Controller> Clone for {client_name}Factory<C> {{
-    fn clone(&self) -> Self {{
-        Self(std::marker::PhantomData)
-    }}
-}}
-
-impl<C> {namespace}::__rt::RpcClientFactory for {client_name}Factory<C> where C: {namespace}::controller::Controller {{
-    type Descriptor = {descriptor_name};
-    type ClientImpl = Box<dyn {name}<Controller = C> + Send + Sync + 'static>;
-    type Controller = C;
-
-    fn new(handler: impl {namespace}::handler::Handler<Descriptor = Self::Descriptor, Controller = Self::Controller>) -> Self::ClientImpl {{
-        Box::new({client_name}::new(handler))
-    }}
-}}
-
-/// A server for a `{name}`.
-///
-/// This implements the `Server` trait by handling requests and dispatch them to methods on the
-/// supplied `{name}`.
-#[derive(Clone, Debug)]
-pub struct {server_name}<A>(A) where A: {name} + Clone + Send + 'static;
-
-impl<T> {server_name}<::std::sync::Weak<T>>
-where
-    T: Send + Sync + 'static,
-    ::std::sync::Arc<T>: {name},
-{{
-    pub fn new_arc(service: ::std::sync::Arc<T>) -> {server_name}<::std::sync::Weak<T>> {{
-        {server_name}(::std::sync::Arc::downgrade(&service))
-    }}
-}}
-
-impl<A> {server_name}<A> where A: {name} + Clone + Send + 'static {{
-    /// Creates a new server instance that dispatches all calls to the supplied service.
-    pub fn new(service: A) -> {server_name}<A> {{
-        {server_name}(service)
-    }}
-
-    async fn call_inner(
-        service: A,
-        method: {method_descriptor_name},
-        ctrl: A::Controller,
-        input: ::bytes::Bytes)
-        -> {namespace}::error::Result<::bytes::Bytes> {{
-        match method {{
-            {match_handle_methods}        
-        }}
-    }}
-}}
-
-impl {namespace}::descriptor::ServiceDescriptor for {descriptor_name} {{
-    type Method = {method_descriptor_name};
-    fn name(&self) -> &'static str {{ {name:?} }}
-    fn proto_name(&self) -> &'static str {{ {proto_name:?} }}
-    fn package(&self) -> &'static str {{ {package:?} }}
-    fn methods(&self) -> &'static [Self::Method] {{
-        &[ {list_enum_methods} ]
-    }}
-}}
-
-#[async_trait::async_trait]
-impl<A> {namespace}::handler::Handler for {server_name}<A>
-where
-    A: {name} + Clone + Send + Sync + 'static {{
-    type Descriptor = {descriptor_name};
-    type Controller = A::Controller;
-
-    async fn call(
-        &self,
-        ctrl: A::Controller,
-        method: {method_descriptor_name},
-        input: ::bytes::Bytes)
-        -> {namespace}::error::Result<::bytes::Bytes> {{
-        {server_name}::call_inner(self.0.clone(), method, ctrl, input).await
-    }}
-}}
-
-impl {namespace}::descriptor::MethodDescriptor for {method_descriptor_name} {{
-    fn name(&self) -> &'static str {{
-        match *self {{
-            {match_name_methods}        
-        }}
-    }}
-
-    fn proto_name(&self) -> &'static str {{
-        match *self {{
-            {match_proto_name_methods}
-        }}
-    }}
-
-    fn input_type(&self) -> ::std::any::TypeId {{
-        match *self {{
-            {match_input_type_methods}
-        }}
-    }}
-
-    fn input_proto_type(&self) -> &'static str {{
-        match *self {{
-            {match_input_proto_type_methods}
-        }}
-    }}
-
-    fn output_type(&self) -> ::std::any::TypeId {{
-        match *self {{
-            {match_output_type_methods}
-        }}
-    }}
-
-    fn output_proto_type(&self) -> &'static str {{
-        match *self {{
-            {match_output_proto_type_methods}
-        }}
-    }}
-
-    fn index(&self) -> u8 {{
-        *self as u8
-    }}
-}}
-"#,
-            name = service.name,
-            descriptor_name = descriptor_name,
-            server_name = server_name,
-            client_name = client_name,
-            method_descriptor_name = method_descriptor_name,
-            proto_name = service.proto_name,
-            package = service.package,
-            trait_methods = trait_methods,
-            weak_impl_methods = weak_impl_methods,
-            enum_methods = enum_methods,
-            list_enum_methods = list_enum_methods,
-            client_own_methods = client_own_methods,
-            client_methods = client_methods,
-            match_name_methods = match_name_methods,
-            match_proto_name_methods = match_proto_name_methods,
-            match_input_type_methods = match_input_type_methods,
-            match_input_proto_type_methods = match_input_proto_type_methods,
-            match_output_type_methods = match_output_type_methods,
-            match_output_proto_type_methods = match_output_proto_type_methods,
-            match_handle_methods = match_handle_methods,
-            match_trait_json_methods = match_trait_json_methods,
-            namespace = NAMESPACE,
-        ).unwrap();
-    }
-}
-
-impl ServiceGenerator {
-    fn write_comments<W>(
-        mut write: W,
-        indent: usize,
-        comments: &prost_build::Comments,
-    ) -> fmt::Result
-    where
-        W: fmt::Write,
-    {
-        for comment in &comments.leading {
-            for line in comment.lines().filter(|s| !s.is_empty()) {
-                writeln!(write, "{}///{}", " ".repeat(indent), line)?;
-            }
-        }
-        Ok(())
-    }
-}

easytier-web/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "easytier-web"
-version = "2.6.3"
+version = "2.6.4"
 edition.workspace = true
 description = "Config server for easytier. easytier-core gets config from this and web frontend use it as restful api server."
 

easytier-web/frontend-lib/src/components/Config.vue

@@ -99,6 +99,7 @@ const bool_flags: BoolFlag[] = [
   { field: 'disable_encryption', help: 'disable_encryption_help' },
   { field: 'disable_tcp_hole_punching', help: 'disable_tcp_hole_punching_help' },
   { field: 'disable_udp_hole_punching', help: 'disable_udp_hole_punching_help' },
+  { field: 'enable_udp_broadcast_relay', help: 'enable_udp_broadcast_relay_help' },
   { field: 'disable_upnp', help: 'disable_upnp_help' },
   { field: 'disable_sym_hole_punching', help: 'disable_sym_hole_punching_help' },
   { field: 'enable_magic_dns', help: 'enable_magic_dns_help' },

easytier-web/frontend-lib/src/locales/cn.yaml

@@ -160,6 +160,9 @@ disable_tcp_hole_punching_help: 禁用TCP打洞功能
 disable_udp_hole_punching: 禁用UDP打洞
 disable_udp_hole_punching_help: 禁用UDP打洞功能
 
+enable_udp_broadcast_relay: UDP 广播中继
+enable_udp_broadcast_relay_help: "仅 Windows：捕获物理网卡上的本机 UDP 广播包并转发给 EasyTier 对等节点，帮助局域网游戏发现房间。需要管理员权限。"
+
 disable_upnp: 禁用 UPnP
 disable_upnp_help: 禁用符合条件监听器的运行时 UPnP/NAT-PMP 端口映射；自动端口映射默认开启。
 
@@ -260,6 +263,7 @@ event:
   DhcpIpv4Conflicted: DHCP IPv4地址冲突
   PortForwardAdded: 端口转发添加
   ProxyCidrsUpdated: 子网代理CIDR更新
+  UdpBroadcastRelayStartResult: UDP广播中继启动结果
 
 web:
   login:

easytier-web/frontend-lib/src/locales/en.yaml

@@ -159,6 +159,9 @@ disable_tcp_hole_punching_help: Disable tcp hole punching
 disable_udp_hole_punching: Disable UDP Hole Punching
 disable_udp_hole_punching_help: Disable udp hole punching
 
+enable_udp_broadcast_relay: UDP Broadcast Relay
+enable_udp_broadcast_relay_help: "Windows only: capture local UDP broadcast packets from physical interfaces and forward them to EasyTier peers. Helps games to find rooms in local network. Requires administrator privileges."
+
 disable_upnp: Disable UPnP
 disable_upnp_help: Disable runtime UPnP/NAT-PMP port mapping for eligible listeners; automatic port mapping is enabled by default.
 
@@ -260,6 +263,7 @@ event:
   DhcpIpv4Conflicted: DhcpIpv4Conflicted
   PortForwardAdded: PortForwardAdded
   ProxyCidrsUpdated: ProxyCidrsUpdated
+  UdpBroadcastRelayStartResult: UDP Broadcast Relay Start Result
 
 web:
   login:

easytier-web/frontend-lib/src/types/network.ts

@@ -134,6 +134,7 @@ export interface NetworkConfig {
   disable_tcp_hole_punching?: boolean
   disable_udp_hole_punching?: boolean
   disable_upnp?: boolean
+  enable_udp_broadcast_relay?: boolean
   disable_sym_hole_punching?: boolean
 
   enable_relay_network_whitelist?: boolean
@@ -211,6 +212,7 @@ export function DEFAULT_NETWORK_CONFIG(): NetworkConfig {
     disable_tcp_hole_punching: false,
     disable_udp_hole_punching: false,
     disable_upnp: false,
+    enable_udp_broadcast_relay: false,
     disable_sym_hole_punching: false,
     enable_relay_network_whitelist: false,
     relay_network_whitelist: [],
@@ -447,4 +449,6 @@ export enum EventType {
   PortForwardAdded = 'PortForwardAdded', // PortForwardConfigPb
 
   ProxyCidrsUpdated = 'ProxyCidrsUpdated', // string[], string[]
+
+  UdpBroadcastRelayStartResult = 'UdpBroadcastRelayStartResult', // { capture_backend?: string, error?: string }
 }

easytier-web/src/client_manager/mod.rs

@@ -365,6 +365,7 @@ mod tests {
         let _c = WebClient::new(
             connector,
             "test",
+            uuid::Uuid::new_v4(),
             "test",
             false,
             Arc::new(NetworkInstanceManager::new()),

easytier/Cargo.toml

@@ -3,7 +3,7 @@ name = "easytier"
 description = "A full meshed p2p VPN, connecting all your devices in one network with one command."
 homepage = "https://github.com/EasyTier/EasyTier"
 repository = "https://github.com/EasyTier/EasyTier"
-version = "2.6.3"
+version = "2.6.4"
 edition.workspace = true
 rust-version.workspace = true
 authors = ["kkrainbow"]
@@ -11,6 +11,7 @@ keywords = ["vpn", "p2p", "network", "easytier"]
 categories = ["network-programming", "command-line-utilities"]
 license-file = "LICENSE"
 readme = "README.md"
+build = "build/main.rs"
 
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
@@ -50,6 +51,9 @@ time = "0.3"
 toml = "0.8.12"
 chrono = { version = "0.4.37", features = ["serde"] }
 
+getset = "0.1.6"
+optionize = "0.2"
+
 guarden = "0.1"
 
 delegate = "0.13.5"
@@ -58,7 +62,7 @@ itertools = "0.14.0"
 
 strum = { version = "0.27.2", features = ["derive"] }
 
-gethostname = "0.5.0"
+hostname = "0.4.2"
 
 futures = { version = "0.3", features = ["bilock", "unstable"] }
 
@@ -69,9 +73,12 @@ tokio-util = { version = "0.7.9", features = ["codec", "net", "io", "rt"] }
 async-stream = "0.3.5"
 async-trait = "0.1.74"
 
+maplit = "1.0.2"
 dashmap = "6.0"
 timedmap = "=1.0.1"
 
+moka = { version = "0.12", features = ["future"] }
+
 # for full-path zero-copy
 zerocopy = { version = "0.7.32", features = ["derive", "simd"] }
 bytes = "1.5.0"
@@ -150,6 +157,7 @@ rand = "0.8.5"
 serde = { version = "1.0", features = ["derive"] }
 pnet = { version = "0.35.0", features = ["serde"] }
 serde_json = "1"
+serde_with = "3"
 
 clap = { version = "4.5.30", features = [
     "string",
@@ -168,6 +176,7 @@ network-interface = "2.0"
 # for ospf route
 petgraph = "0.8.1"
 ordered_hash_map = "0.5.0"
+indexmap = "2.13.1"
 
 # for wireguard
 boringtun = { package = "boringtun-easytier", version = "0.6.1", optional = true }
@@ -234,17 +243,14 @@ http_req = { git = "https://github.com/EasyTier/http_req.git", default-features
 ] }
 
 # for dns connector
-hickory-resolver = "0.25.2"
-hickory-proto = "0.25.2"
+hickory-proto = "0.26.0"
+hickory-net = { version = "0.26.0", features = ["serde"] }
+hickory-resolver = { version = "0.26.0", features = ["https-ring", "webpki-roots"] }
 
 # for magic dns
-hickory-client = { version = "0.25.2", optional = true }
-hickory-server = { version = "0.25.2", features = [
-    "resolver",
-], optional = true }
+hickory-server = { version = "0.26.0", features = ["resolver"], optional = true }
 
 bon = "3.9.1"
-derive_builder = "0.20.2"
 humantime-serde = "1.1.1"
 multimap = "0.10.1"
 version-compare = "0.2.0"
@@ -315,15 +321,14 @@ jemalloc-sys = { package = "tikv-jemalloc-sys", version = "0.6.0", features = [
 
 [build-dependencies]
 cfg_aliases = "0.2.1"
-tonic-build = "0.12"
+indoc = "2.0"
 globwalk = "0.8.1"
 regex = "1"
 prost-build = "0.13.5"
 prost-wkt-build = "0.6"
-easytier-rpc-build = { path = "../easytier-rpc-build", features = [
-    "internal-namespace",
-] }
 prost-reflect-build = { version = "0.14.0" }
+proc-macro2 = "1"
+quote = "1"
 thunk-rs = { git = "https://github.com/easytier/thunk.git", default-features = false, features = [
     "win7",
 ] }
@@ -337,7 +342,6 @@ zip = "4.0.0"
 serial_test = "3.0.0"
 rstest = "0.25.0"
 futures-util = "0.3.31"
-maplit = "1.0.2"
 tempfile = "3.22.0"
 ctor = "0.8.0"
 
@@ -399,7 +403,7 @@ jemalloc-prof = [
     "jemalloc-sys/stats",
 ]
 tracing = ["tokio/tracing", "dep:console-subscriber"]
-magic-dns = ["dep:hickory-client", "dep:hickory-server"]
+magic-dns = ["dep:hickory-server"]
 faketcp = ["dep:flume"]
 zstd = ["dep:zstd"]
 # For Network Extension on macOS

easytier/build/main.rs

@@ -1,3 +1,6 @@
+mod rpc;
+
+use crate::rpc::ServiceGenerator;
 use cfg_aliases::cfg_aliases;
 use prost_wkt_build::{FileDescriptorSet, Message as _};
 #[cfg(target_os = "windows")]
@@ -163,7 +166,7 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
         "src/proto/api_config.proto",
         "src/proto/api_manage.proto",
         "src/proto/web.proto",
-        "src/proto/magic_dns.proto",
+        "src/proto/dns.proto",
         "src/proto/acl.proto",
     ];
 
@@ -171,17 +174,14 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
         println!("cargo:rerun-if-changed={proto_file}");
     }
 
-    let out = PathBuf::from(env::var("OUT_DIR").unwrap());
-    let descriptor_file = out.join("descriptors.bin");
-
     let mut config = prost_build::Config::new();
     config
-        .type_attribute(".", "#[derive(serde::Serialize,serde::Deserialize)]")
         .extern_path(".google.protobuf.Any", "::prost_wkt_types::Any")
         .extern_path(".google.protobuf.Timestamp", "::prost_wkt_types::Timestamp")
-        .extern_path(".google.protobuf.Value", "::prost_wkt_types::Value")
-        .file_descriptor_set_path(&descriptor_file)
-        .protoc_arg("--experimental_allow_proto3_optional")
+        .extern_path(".google.protobuf.Value", "::prost_wkt_types::Value");
+
+    config
+        .type_attribute(".", "#[derive(serde::Serialize,serde::Deserialize)]")
         .type_attribute("peer_rpc.DirectConnectedPeerInfo", "#[derive(Hash)]")
         .type_attribute("peer_rpc.PeerInfoForGlobalMap", "#[derive(Hash)]")
         .type_attribute("peer_rpc.ForeignNetworkRouteInfoKey", "#[derive(Hash, Eq)]")
@@ -195,20 +195,28 @@ fn main() -> Result<(), Box<dyn std::error::Error>> {
         .type_attribute("acl.AclV1", "#[serde(default)]")
         .type_attribute("acl.Chain", "#[serde(default)]")
         .type_attribute("acl.Rule", "#[serde(default)]")
-        .type_attribute("acl.GroupInfo", "#[serde(default)]")
-        .field_attribute(".api.manage.NetworkConfig", "#[serde(default)]")
-        .service_generator(Box::new(easytier_rpc_build::ServiceGenerator::default()))
-        .btree_map(["."])
-        .skip_debug([".common.Ipv4Addr", ".common.Ipv6Addr", ".common.UUID"]);
+        .type_attribute("acl.GroupInfo", "#[serde(default)]");
 
-    config.compile_protos(&proto_files, &["src/proto/"])?;
+    config.field_attribute("api.manage.NetworkConfig", "#[serde(default)]");
+
+    config.skip_debug([".common.Ipv4Addr", ".common.Ipv6Addr", ".common.UUID"]);
+
+    let out = PathBuf::from(env::var("OUT_DIR")?);
+    let descriptor_file = out.join("descriptors.bin");
+
+    config
+        .btree_map(["."])
+        .service_generator(Box::new(ServiceGenerator::default()))
+        .protoc_arg("--experimental_allow_proto3_optional")
+        .file_descriptor_set_path(&descriptor_file)
+        .compile_protos(&proto_files, &["src/proto/"])?;
 
     prost_reflect_build::Builder::new()
         .file_descriptor_set_bytes("crate::proto::DESCRIPTOR_POOL_BYTES")
         .compile_protos_with_config(config, &proto_files_reflect, &["src/proto/"])?;
 
-    let descriptor_bytes = std::fs::read(descriptor_file).unwrap();
-    let descriptor = FileDescriptorSet::decode(&descriptor_bytes[..]).unwrap();
+    let descriptor_bytes = std::fs::read(descriptor_file)?;
+    let descriptor = FileDescriptorSet::decode(&descriptor_bytes[..])?;
     prost_wkt_build::add_serde(out, descriptor);
 
     check_locale();

easytier/build/rpc.rs

@@ -0,0 +1,720 @@
+#![allow(non_snake_case)]
+
+use indoc::formatdoc;
+use proc_macro2::{Ident, TokenStream};
+use quote::{format_ident, quote};
+use std::str::FromStr;
+
+fn parse(value: &str) -> TokenStream {
+    TokenStream::from_str(value)
+        .unwrap_or_else(|err| panic!("Failed to parse tokens: {} ({})", value, err))
+}
+
+fn doc(comments: &prost_build::Comments) -> TokenStream {
+    let doc = comments
+        .leading
+        .iter()
+        .flat_map(|c| c.lines().filter(|s| !s.is_empty()));
+    quote! { #( #[doc = #doc] )* }
+}
+
+const NAMESPACE: &str = "crate::proto::rpc_types";
+
+struct Method {
+    index: u8,
+    doc: TokenStream,
+    method: Ident,
+    method_inner: Ident,
+    method_str: String,
+    method_proto: Ident,
+    method_proto_str: String,
+    Input: TokenStream,
+    Input_proto_str: String,
+    Output: TokenStream,
+    Output_proto_str: String,
+}
+
+impl Method {
+    fn new(index: u8, method: prost_build::Method) -> Self {
+        assert!(
+            !method.client_streaming,
+            "Client streaming not yet supported for method {}",
+            method.proto_name
+        );
+        assert!(
+            !method.server_streaming,
+            "Server streaming not yet supported for method {}",
+            method.proto_name
+        );
+        Self {
+            index,
+            doc: doc(&method.comments),
+            method: format_ident!("{}", method.name),
+            method_inner: format_ident!("{}_inner", method.name),
+            method_str: method.name,
+            method_proto: format_ident!("{}", method.proto_name),
+            method_proto_str: method.proto_name,
+            Input: parse(&method.input_type),
+            Input_proto_str: method.input_proto_type,
+            Output: parse(&method.output_type),
+            Output_proto_str: method.output_proto_type,
+        }
+    }
+}
+
+struct Service {
+    namespace: TokenStream,
+    doc: TokenStream,
+    Service: Ident,
+    ServiceDescriptor: Ident,
+    ServiceServer: Ident,
+    ServiceClient: Ident,
+    ServiceClientFactory: Ident,
+    ServiceMethodDescriptor: Ident,
+    Service_str: String,
+    Service_proto_str: String,
+    Service_package_str: String,
+    methods: Vec<Method>,
+}
+
+impl Service {
+    fn new(service: prost_build::Service) -> Self {
+        let methods = service
+            .methods
+            .into_iter()
+            .enumerate()
+            .map(|(i, method)| Method::new((i + 1) as u8, method))
+            .collect();
+
+        Self {
+            namespace: parse(NAMESPACE),
+            doc: doc(&service.comments),
+            Service: format_ident!("{}", service.name),
+            ServiceDescriptor: format_ident!("{}Descriptor", service.name),
+            ServiceServer: format_ident!("{}Server", service.name),
+            ServiceClient: format_ident!("{}Client", service.name),
+            ServiceClientFactory: format_ident!("{}ClientFactory", service.name),
+            ServiceMethodDescriptor: format_ident!("{}MethodDescriptor", service.name),
+            Service_str: service.name,
+            Service_proto_str: service.proto_name,
+            Service_package_str: service.package,
+            methods,
+        }
+    }
+
+    fn trait_Service(&self) -> TokenStream {
+        let Self {
+            namespace,
+            doc,
+            Service,
+            methods,
+            ..
+        } = self;
+
+        let match_json_call_method = methods.iter().map(
+            |Method {
+                 method,
+                 method_str,
+                 method_proto_str,
+                 Input,
+                 ..
+             }| {
+                quote! {
+                    #method_str | #method_proto_str => {
+                        let req: #Input = ::serde_json::from_value(json)
+                            .map_err(|e| #namespace::error::Error::MalformatRpcPacket(format!("json error: {}", e)))?;
+                        let resp = self.#method(ctrl, req).await?;
+                        Ok(::serde_json::to_value(resp)
+                            .map_err(|e| #namespace::error::Error::MalformatRpcPacket(format!("json error: {}", e)))?)
+                    }
+                }
+            },
+        );
+
+        let methods = methods.iter().map(
+            |Method {
+                 doc,
+                 method,
+                 Input,
+                 Output,
+                 ..
+             }| {
+                quote! {
+                    #doc
+                    async fn #method(&self, ctrl: Self::Controller, input: #Input) -> #namespace::error::Result<#Output>;
+                }
+            },
+        );
+
+        quote! {
+            #doc
+            #[async_trait::async_trait]
+            #[auto_impl::auto_impl(&, Arc, Box)]
+            pub trait #Service {
+                type Controller: #namespace::controller::Controller;
+
+                #(#methods)*
+
+                async fn json_call_method(
+                    &self,
+                    ctrl: Self::Controller,
+                    method: &str,
+                    json: ::serde_json::Value,
+                ) -> #namespace::error::Result<::serde_json::Value> {
+                    match method {
+                        #(#match_json_call_method)*
+                        _ => Err(#namespace::error::Error::InvalidMethodIndex(0, method.to_string())),
+                    }
+                }
+            }
+        }
+    }
+
+    fn impl_Service_for_Weak(&self) -> TokenStream {
+        let Self {
+            namespace,
+            Service,
+            methods,
+            ..
+        } = self;
+        let methods = methods.iter().map(
+            |Method {
+                 method,
+                 Input,
+                 Output,
+                 ..
+             }| {
+                quote! {
+                    async fn #method(&self, ctrl: Self::Controller, input: #Input) -> #namespace::error::Result<#Output> {
+                        let Some(service) = self.upgrade() else {
+                            return Err(#namespace::error::Error::Shutdown);
+                        };
+                        service.#method(ctrl, input).await
+                    }
+                }
+            },
+        );
+
+        quote! {
+            #[async_trait::async_trait]
+            impl<T> #Service for ::std::sync::Weak<T>
+            where
+                T: Send + Sync + 'static,
+                ::std::sync::Arc<T>: #Service,
+            {
+                type Controller = <::std::sync::Arc<T> as #Service>::Controller;
+
+                #(#methods)*
+            }
+        }
+    }
+
+    fn struct_ServiceDescriptor(&self) -> TokenStream {
+        let Self {
+            namespace,
+            ServiceDescriptor,
+            ServiceMethodDescriptor,
+            Service_str,
+            Service_proto_str,
+            Service_package_str,
+            methods,
+            ..
+        } = self;
+
+        let doc = format!("A service descriptor for a `{}`.", Service_str);
+
+        let methods = methods.iter().map(|Method { method_proto, .. }| {
+            quote! { #ServiceMethodDescriptor::#method_proto, }
+        });
+
+        quote! {
+            #[doc = #doc]
+            #[derive(Clone, Debug, Eq, Ord, PartialEq, PartialOrd, Default)]
+            pub struct #ServiceDescriptor;
+
+            impl #namespace::descriptor::ServiceDescriptor for #ServiceDescriptor {
+                type Method = #ServiceMethodDescriptor;
+                fn name(&self) -> &'static str { #Service_str }
+                fn proto_name(&self) -> &'static str { #Service_proto_str }
+                fn package(&self) -> &'static str { #Service_package_str }
+                fn methods(&self) -> &'static [Self::Method] {
+                    &[ #(#methods)* ]
+                }
+            }
+        }
+    }
+
+    fn enum_ServiceMethodDescriptor(&self) -> TokenStream {
+        let Self {
+            ServiceMethodDescriptor,
+            Service_str,
+            methods,
+            ..
+        } = self;
+
+        let doc = formatdoc! {"
+            Methods available on a `{Service_str}`.
+
+            This can be used as a key when routing requests for servers/clients of a `{Service_str}`.
+        "};
+
+        let variants = methods.iter().map(
+            |Method {
+                 method_proto,
+                 index,
+                 ..
+             }| {
+                quote! { #method_proto = #index, }
+            },
+        );
+
+        let impl_MethodDescriptor = self.impl_MethodDescriptor_for_ServiceMethodDescriptor();
+        let impl_TryFrom = self.impl_TryFrom_for_ServiceMethodDescriptor();
+        quote! {
+            #[doc = #doc]
+            #[derive(Clone, Copy, Debug, Eq, Ord, PartialEq, PartialOrd)]
+            #[repr(u8)]
+            pub enum #ServiceMethodDescriptor {
+                #(#variants)*
+            }
+
+            #impl_MethodDescriptor
+
+            #impl_TryFrom
+        }
+    }
+
+    fn impl_MethodDescriptor_for_ServiceMethodDescriptor(&self) -> TokenStream {
+        let Self {
+            namespace,
+            ServiceMethodDescriptor,
+            methods,
+            ..
+        } = self;
+
+        let name = {
+            let arms = methods.iter().map(
+                |Method {
+                     method_proto,
+                     method_str,
+                     ..
+                 }| {
+                    quote! { #ServiceMethodDescriptor::#method_proto => #method_str, }
+                },
+            );
+
+            quote! {
+                 fn name(&self) -> &'static str {
+                    match *self {
+                        #(#arms)*
+                    }
+                }
+            }
+        };
+
+        let proto_name = {
+            let arms = methods.iter().map(
+                |Method {
+                     method_proto,
+                     method_proto_str,
+                     ..
+                 }| {
+                    quote! { #ServiceMethodDescriptor::#method_proto => #method_proto_str, }
+                },
+            );
+
+            quote! {
+                fn proto_name(&self) -> &'static str {
+                    match *self {
+                        #(#arms)*
+                    }
+                }
+            }
+        };
+
+        let input_type = {
+            let arms = methods.iter().map(|Method { method_proto, Input, .. }| {
+                quote! { #ServiceMethodDescriptor::#method_proto => ::std::any::TypeId::of::<#Input>(), }
+            });
+
+            quote! {
+                fn input_type(&self) -> ::std::any::TypeId {
+                    match *self {
+                        #(#arms)*
+                    }
+                }
+            }
+        };
+
+        let input_proto_type = {
+            let arms = methods.iter().map(
+                |Method {
+                     method_proto,
+                     Input_proto_str,
+                     ..
+                 }| {
+                    quote! { #ServiceMethodDescriptor::#method_proto => #Input_proto_str, }
+                },
+            );
+
+            quote! {
+                fn input_proto_type(&self) -> &'static str {
+                    match *self {
+                        #(#arms)*
+                    }
+                }
+            }
+        };
+
+        let output_type = {
+            let arms = methods.iter().map(|Method { method_proto, Output, .. }| {
+                quote! { #ServiceMethodDescriptor::#method_proto => ::std::any::TypeId::of::<#Output>(), }
+            });
+
+            quote! {
+                fn output_type(&self) -> ::std::any::TypeId {
+                    match *self {
+                        #(#arms)*
+                    }
+                }
+            }
+        };
+
+        let output_proto_type = {
+            let arms = methods.iter().map(
+                |Method {
+                     method_proto,
+                     Output_proto_str,
+                     ..
+                 }| {
+                    quote! { #ServiceMethodDescriptor::#method_proto => #Output_proto_str, }
+                },
+            );
+
+            quote! {
+                fn output_proto_type(&self) -> &'static str {
+                    match *self {
+                        #(#arms)*
+                    }
+                }
+            }
+        };
+
+        quote! {
+            impl #namespace::descriptor::MethodDescriptor for #ServiceMethodDescriptor {
+                #name
+
+                #proto_name
+
+                #input_type
+
+                #input_proto_type
+
+                #output_type
+
+                #output_proto_type
+
+                fn index(&self) -> u8 {
+                    *self as u8
+                }
+            }
+        }
+    }
+
+    fn impl_TryFrom_for_ServiceMethodDescriptor(&self) -> TokenStream {
+        let Self {
+            namespace,
+            ServiceMethodDescriptor,
+            Service_str,
+            methods,
+            ..
+        } = self;
+
+        let arms = methods.iter().map(
+            |Method {
+                 method_proto,
+                 index,
+                 ..
+             }| {
+                quote! { #index => Ok(#ServiceMethodDescriptor::#method_proto), }
+            },
+        );
+
+        quote! {
+            impl std::convert::TryFrom<u8> for #ServiceMethodDescriptor {
+                type Error = #namespace::error::Error;
+                fn try_from(value: u8) -> #namespace::error::Result<Self> {
+                    match value {
+                        #(#arms)*
+                        _ => Err(#namespace::error::Error::InvalidMethodIndex(value, #Service_str.to_string())),
+                    }
+                }
+            }
+        }
+    }
+
+    fn struct_ServiceClient(&self) -> TokenStream {
+        let Self {
+            namespace,
+            ServiceDescriptor,
+            ServiceClient,
+            Service_str,
+            ..
+        } = self;
+
+        let doc = formatdoc! {"
+            A client for a `{Service_str}`.
+
+            This implements the `{Service_str}` trait by dispatching all method calls to the supplied `Handler`.
+        "};
+
+        let impl_service_client = self.impl_ServiceClient();
+        let impl_service_for_client = self.impl_Service_for_ServiceClient();
+        quote! {
+            #[doc = #doc]
+            #[derive(Clone, Debug)]
+            pub struct #ServiceClient<H>(H) where H: #namespace::handler::Handler;
+
+            impl<H> #ServiceClient<H> where H: #namespace::handler::Handler<Descriptor = #ServiceDescriptor> {
+                /// Creates a new client instance that delegates all method calls to the supplied handler.
+                pub fn new(handler: H) -> Self {
+                    Self(handler)
+                }
+            }
+
+            #impl_service_client
+
+            #impl_service_for_client
+        }
+    }
+
+    fn impl_ServiceClient(&self) -> TokenStream {
+        let Self {
+            namespace,
+            ServiceClient,
+            ServiceDescriptor,
+            ServiceMethodDescriptor,
+            methods,
+            ..
+        } = self;
+
+        let methods = methods.iter().map(
+            |Method {
+                 method_inner,
+                 method_proto,
+                 Input,
+                 Output,
+                 ..
+             }| {
+                quote! {
+                    async fn #method_inner(handler: H, ctrl: H::Controller, input: #Input) -> #namespace::error::Result<#Output> {
+                        #namespace::__rt::call_method(handler, ctrl, #ServiceMethodDescriptor::#method_proto, input).await
+                    }
+                }
+            },
+        );
+
+        quote! {
+            impl<H> #ServiceClient<H> where H: #namespace::handler::Handler<Descriptor = #ServiceDescriptor> {
+                #(#methods)*
+            }
+        }
+    }
+
+    fn impl_Service_for_ServiceClient(&self) -> TokenStream {
+        let Self {
+            namespace,
+            Service,
+            ServiceClient,
+            ServiceDescriptor,
+            methods,
+            ..
+        } = self;
+
+        let methods = methods.iter().map(
+            |Method {
+                 method,
+                 method_inner,
+                 Input,
+                 Output,
+                 ..
+             }| {
+                quote! {
+                    async fn #method(&self, ctrl: H::Controller, input: #Input) -> #namespace::error::Result<#Output> {
+                        #ServiceClient::#method_inner(self.0.clone(), ctrl, input).await
+                    }
+                }
+            },
+        );
+
+        quote! {
+            #[async_trait::async_trait]
+            impl<H> #Service for #ServiceClient<H> where H: #namespace::handler::Handler<Descriptor = #ServiceDescriptor> {
+                type Controller = H::Controller;
+
+                #(#methods)*
+            }
+        }
+    }
+
+    fn struct_ServiceClientFactory(&self) -> TokenStream {
+        let Self {
+            namespace,
+            Service,
+            ServiceClient,
+            ServiceClientFactory,
+            ServiceDescriptor,
+            ..
+        } = self;
+
+        quote! {
+            pub struct #ServiceClientFactory<C: #namespace::controller::Controller>(std::marker::PhantomData<C>);
+
+            impl<C: #namespace::controller::Controller> Clone for #ServiceClientFactory<C> {
+                fn clone(&self) -> Self {
+                    Self(std::marker::PhantomData)
+                }
+            }
+
+            impl<C> #namespace::__rt::RpcClientFactory for #ServiceClientFactory<C> where C: #namespace::controller::Controller {
+                type Descriptor = #ServiceDescriptor;
+                type ClientImpl = Box<dyn #Service<Controller = C> + Send + Sync + 'static>;
+                type Controller = C;
+
+                fn new(handler: impl #namespace::handler::Handler<Descriptor = Self::Descriptor, Controller = Self::Controller>) -> Self::ClientImpl {
+                    Box::new(#ServiceClient::new(handler))
+                }
+            }
+        }
+    }
+
+    fn struct_ServiceServer(&self) -> TokenStream {
+        let Self {
+            namespace,
+            Service,
+            ServiceDescriptor,
+            ServiceServer,
+            ServiceMethodDescriptor,
+            Service_str,
+            methods,
+            ..
+        } = self;
+
+        let doc = formatdoc! {"
+            A server for a `{Service_str}`.
+
+            This implements the `Server` trait by handling requests and dispatch them to methods on the
+            supplied `{Service_str}`.
+        "};
+
+        let arms = methods.iter().map(
+            |Method {
+                 method_proto,
+                 method,
+                 Input,
+                 ..
+             }| {
+                quote! {
+                    #ServiceMethodDescriptor::#method_proto => {
+                        let decoded: #Input = #namespace::__rt::decode(input)?;
+                        let ret = service.#method(ctrl, decoded).await?;
+                        #namespace::__rt::encode(ret)
+                    }
+                }
+            },
+        );
+
+        quote! {
+            #[doc = #doc]
+            #[derive(Clone, Debug)]
+            pub struct #ServiceServer<A>(A) where A: #Service + Clone + Send + 'static;
+
+            impl<T> #ServiceServer<::std::sync::Weak<T>>
+            where
+                T: Send + Sync + 'static,
+                ::std::sync::Arc<T>: #Service,
+            {
+                pub fn new_arc(service: ::std::sync::Arc<T>) -> #ServiceServer<::std::sync::Weak<T>> {
+                    #ServiceServer(::std::sync::Arc::downgrade(&service))
+                }
+            }
+
+            impl<A> #ServiceServer<A> where A: #Service + Clone + Send + 'static {
+                /// Creates a new server instance that dispatches all calls to the supplied service.
+                pub fn new(service: A) -> #ServiceServer<A> {
+                    #ServiceServer(service)
+                }
+
+                async fn call_inner(
+                    service: A,
+                    method: #ServiceMethodDescriptor,
+                    ctrl: A::Controller,
+                    input: ::bytes::Bytes)
+                    -> #namespace::error::Result<::bytes::Bytes> {
+                    match method {
+                        #(#arms)*
+                    }
+                }
+            }
+
+            #[async_trait::async_trait]
+            impl<A> #namespace::handler::Handler for #ServiceServer<A>
+            where
+                A: #Service + Clone + Send + Sync + 'static {
+                type Descriptor = #ServiceDescriptor;
+                type Controller = A::Controller;
+
+                async fn call(
+                    &self,
+                    ctrl: A::Controller,
+                    method: #ServiceMethodDescriptor,
+                    input: ::bytes::Bytes)
+                    -> #namespace::error::Result<::bytes::Bytes> {
+                    #ServiceServer::call_inner(self.0.clone(), method, ctrl, input).await
+                }
+            }
+        }
+    }
+}
+
+/// The service generator to be used with `prost-build` to generate RPC implementations for
+/// `prost-simple-rpc`.
+///
+/// See the crate-level documentation for more info.
+#[non_exhaustive]
+#[derive(Debug, Default)]
+pub struct ServiceGenerator;
+
+impl prost_build::ServiceGenerator for ServiceGenerator {
+    fn generate(&mut self, service: prost_build::Service, buf: &mut String) {
+        let info = Service::new(service);
+
+        let trait_Service = info.trait_Service();
+        let impl_Service_for_Weak = info.impl_Service_for_Weak();
+        let struct_ServiceDescriptor = info.struct_ServiceDescriptor();
+        let enum_ServiceMethodDescriptor = info.enum_ServiceMethodDescriptor();
+        let struct_ServiceClient = info.struct_ServiceClient();
+        let struct_ServiceClientFactory = info.struct_ServiceClientFactory();
+        let struct_ServiceServer = info.struct_ServiceServer();
+
+        let tokens = quote! {
+            #trait_Service
+
+            #impl_Service_for_Weak
+
+            #struct_ServiceDescriptor
+
+            #enum_ServiceMethodDescriptor
+
+            #struct_ServiceClient
+
+            #struct_ServiceClientFactory
+
+            #struct_ServiceServer
+        };
+
+        buf.push('\n');
+        buf.push_str(&tokens.to_string());
+        buf.push('\n');
+    }
+}

easytier/locales/app.yml

@@ -12,9 +12,9 @@ core_clap:
       仅用户名：--config-server admin，将使用官方的服务器
   machine_id:
     en: |+
-      the machine id to identify this machine, used for config recovery after disconnection, must be unique and fixed. default is from system.
+      the machine id to identify this machine, used for config recovery after disconnection, must be unique and fixed. by default it is loaded from persisted local state; on first start it may be migrated from system information or generated, then remains fixed.
     zh-CN: |+
-      Web 配置服务器通过 machine id 来识别机器，用于断线重连后的配置恢复，需要保证唯一且固定不变。默认从系统获得。
+      Web 配置服务器通过 machine id 来识别机器，用于断线重连后的配置恢复，需要保证唯一且固定不变。默认从本地持久化状态读取；首次启动时可能基于系统信息迁移或生成，之后保持固定不变。
   config_file:
     en: "path to the config file, NOTE: the options set by cmdline args will override options in config file"
     zh-CN: "配置文件路径，注意：命令行中的配置的选项会覆盖配置文件中的选项"
@@ -184,6 +184,9 @@ core_clap:
   disable_upnp:
     en: "disable runtime UPnP/NAT-PMP port mapping for eligible listeners; automatic port mapping is enabled by default"
     zh-CN: "禁用符合条件监听器的运行时 UPnP/NAT-PMP 端口映射；自动端口映射默认开启"
+  enable_udp_broadcast_relay:
+    en: "Windows only: capture local UDP broadcast packets from physical interfaces and forward them to EasyTier peers. Helps games to find rooms in local network. Requires administrator privileges."
+    zh-CN: "仅 Windows：捕获物理网卡上的本机 UDP 广播包并转发给 EasyTier 对等节点，帮助局域网游戏发现房间。需要管理员权限。"
   relay_all_peer_rpc:
     en: "relay all peer rpc packets, even if the peer is not in the relay network whitelist. this can help peers not in relay network whitelist to establish p2p connection."
     zh-CN: "转发所有对等节点的RPC数据包，即使对等节点不在转发网络白名单中。这可以帮助白名单外网络中的对等节点建立P2P连接。"
@@ -217,12 +220,6 @@ core_clap:
   port_forward:
     en: "forward local port to remote port in virtual network. e.g.: udp://0.0.0.0:12345/10.126.126.1:23456, means forward local udp port 12345 to 10.126.126.1:23456 in the virtual network. can specify multiple."
     zh-CN: "将本地端口转发到虚拟网络中的远程端口。例如：udp://0.0.0.0:12345/10.126.126.1:23456，表示将本地UDP端口12345转发到虚拟网络中的10.126.126.1:23456。可以指定多个。"
-  accept_dns:
-    en: "if true, enable magic dns. with magic dns, you can access other nodes with a domain name, e.g.: <hostname>.et.net. magic dns will modify your system dns settings, enable it carefully."
-    zh-CN: "如果为true，则启用魔法DNS。使用魔法DNS，您可以使用域名访问其他节点，例如：<hostname>.et.net。魔法DNS将修改您的系统DNS设置，请谨慎启用。"
-  tld_dns_zone:
-    en: "specify the top-level domain zone for magic DNS. if not provided, defaults to the value from dns_server module (et.net.). only used when accept_dns is true."
-    zh-CN: "指定魔法DNS的顶级域名区域。如果未提供，默认使用dns_server模块中的值（et.net.）。仅在accept_dns为true时使用。"
   private_mode:
     en: "if true, foreign networks are only allowed when this node can verify they use the same network secret, or when a foreign credential node is already trusted via admin-issued credential propagation; different or missing secrets are otherwise rejected."
     zh-CN: "如果为true，则仅允许两类 foreign network 接入：本节点能验证其使用相同 network secret 的节点，或已通过 foreign network 管理节点传播而被信任的 credential 节点；否则 secret 不同或缺失时会被拒绝。"
@@ -274,6 +271,9 @@ core_clap:
   check_config:
     en: Check config validity without starting the network
     zh-CN: 检查配置文件的有效性并退出
+  daemon:
+    en: Run in daemon mode
+    zh-CN: 以守护进程模式运行
   file_log_size_mb:
     en: "per file log size in MB, default is 100MB"
     zh-CN: "单个文件日志大小，单位 MB，默认值为 100MB"

easytier/src/arch/windows.rs

@@ -11,9 +11,8 @@ use windows::{
             NET_FW_RULE_DIR_OUT,
         },
         Networking::WinSock::{
-            IP_UNICAST_IF, IPPROTO_IP, IPPROTO_IPV6, IPV6_UNICAST_IF, SIO_UDP_CONNRESET,
-            SO_EXCLUSIVEADDRUSE, SOCKET, SOCKET_ERROR, SOL_SOCKET, WSAGetLastError, WSAIoctl,
-            htonl, setsockopt,
+            IP_UNICAST_IF, IPPROTO_IP, IPPROTO_IPV6, IPV6_UNICAST_IF, SIO_UDP_CONNRESET, SOCKET,
+            SOCKET_ERROR, WSAGetLastError, WSAIoctl, htonl, setsockopt,
         },
         System::Com::{
             CLSCTX_ALL, COINIT_MULTITHREADED, CoCreateInstance, CoInitializeEx, CoUninitialize,

easytier/src/common/config.rs

@@ -1,36 +1,114 @@
-use std::{
-    hash::Hasher,
-    net::{IpAddr, SocketAddr},
-    path::PathBuf,
-    sync::{Arc, Mutex},
-};
-
-use anyhow::Context;
-use base64::{Engine as _, prelude::BASE64_STANDARD};
-use clap::ValueEnum;
-use clap::builder::PossibleValue;
-use serde::{Deserialize, Serialize};
-use strum::{Display, EnumString, VariantArray};
-use tokio::io::AsyncReadExt as _;
-
+use super::env_parser;
+use crate::utils::dns;
 use crate::{
     common::stun::StunInfoCollector,
-    instance::dns_server::DEFAULT_ET_DNS_ZONE,
     proto::{
         acl::Acl,
         api::manage::ConfigSource as RpcConfigSource,
         common::{CompressionAlgoPb, PortForwardConfigPb, SecureModeConfig, SocketType},
     },
     tunnel::{IpScheme, TunnelScheme, generate_digest_from_str},
+    utils,
 };
+use anyhow::Context;
+use base64::{Engine as _, prelude::BASE64_STANDARD};
+use bon::Builder;
+use clap::ValueEnum;
+use clap::builder::PossibleValue;
+use derivative::Derivative;
+use derive_more::{Constructor, Deref};
+use getset::Getters;
+use optionize::Optionized;
+use serde::{Deserialize, Serialize};
+use std::fmt::{Debug, Display};
+use std::{
+    hash::Hasher,
+    net::{IpAddr, SocketAddr},
+    path::PathBuf,
+    sync::{Arc, Mutex},
+};
+use strum::{Display, EnumString, VariantArray};
+use tokio::io::AsyncReadExt as _;
 
-use super::env_parser;
+#[derive(Derivative, Debug, Clone, Constructor, Getters, Deref, Deserialize)]
+#[derivative(PartialEq(bound = "Parsed: PartialEq"))]
+#[serde(try_from = "Raw")]
+#[serde(
+    bound = "Raw: Deserialize<'de>, <ConfigBase<Raw, Parsed, Data> as TryFrom<Raw>>::Error: Display"
+)]
+pub struct ConfigBase<Raw, Parsed, Data = ()>
+where
+    Raw: Optionized<Subject = Parsed>,
+    ConfigBase<Raw, Parsed, Data>: TryFrom<Raw>,
+{
+    #[deref]
+    parsed: Parsed,
+    #[getset(get)]
+    #[derivative(PartialEq = "ignore")]
+    raw: Raw,
+    #[getset(get)]
+    #[derivative(PartialEq = "ignore")]
+    data: Data,
+}
+
+impl<Raw, Parsed, Data> Serialize for ConfigBase<Raw, Parsed, Data>
+where
+    Raw: Optionized<Subject = Parsed> + Serialize,
+    ConfigBase<Raw, Parsed, Data>: TryFrom<Raw, Error: Debug>,
+{
+    fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
+    where
+        S: serde::Serializer,
+    {
+        self.raw.serialize(serializer)
+    }
+}
+
+impl<Raw, Parsed, Data> Default for ConfigBase<Raw, Parsed, Data>
+where
+    Raw: Optionized<Subject = Parsed> + Default,
+    ConfigBase<Raw, Parsed, Data>: TryFrom<Raw, Error: Debug>,
+{
+    fn default() -> Self {
+        Raw::default().try_into().unwrap()
+    }
+}
+
+impl<Raw, Parsed, Data> ConfigBase<Raw, Parsed, Data>
+where
+    Raw: Optionized<Subject = Parsed>,
+    ConfigBase<Raw, Parsed, Data>: TryFrom<Raw, Error: Debug>,
+{
+    pub fn into_parsed(self) -> Parsed {
+        self.parsed
+    }
+
+    pub fn into_raw(self) -> Raw {
+        self.raw
+    }
+
+    pub fn into_data(self) -> Data {
+        self.data
+    }
+
+    pub fn update(self, config: Raw) -> Result<Self, <Self as TryFrom<Raw>>::Error> {
+        let mut raw = self.into_raw();
+        raw.merge(config);
+        raw.try_into()
+    }
+}
 
 pub type Flags = crate::proto::common::FlagsInConfig;
 
 pub fn gen_default_flags() -> Flags {
-    #[allow(deprecated)]
     Flags {
+        #[allow(deprecated)]
+        quic_listen_port: u32::MAX,
+        #[allow(deprecated)]
+        accept_dns: false,
+        #[allow(deprecated)]
+        tld_dns_zone: "".to_string(),
+
         default_protocol: "tcp".to_string(),
         dev_name: "".to_string(),
         enable_encryption: true,
@@ -55,7 +133,6 @@ pub fn gen_default_flags() -> Flags {
         disable_kcp_input: false,
         disable_relay_kcp: false,
         enable_relay_foreign_network_kcp: false,
-        accept_dns: false,
         private_mode: false,
         enable_quic_proxy: false,
         disable_quic_input: false,
@@ -65,13 +142,11 @@ pub fn gen_default_flags() -> Flags {
         multi_thread_count: 2,
         encryption_algorithm: EncryptionAlgorithm::default().to_string(),
         disable_sym_hole_punching: false,
-        tld_dns_zone: DEFAULT_ET_DNS_ZONE.to_string(),
-
-        quic_listen_port: u32::MAX,
         need_p2p: false,
         instance_recv_bps_limit: u64::MAX,
         disable_upnp: false,
         disable_relay_data: false,
+        enable_udp_broadcast_relay: false,
     }
 }
 
@@ -151,8 +226,19 @@ impl Default for EncryptionAlgorithm {
     }
 }
 
+cfg_select! {
+    feature = "magic-dns" => {
+        use crate::dns::config::{DnsConfig, DnsConfigLoaderExt};
+    }
+
+    _ => {
+        #[auto_impl::auto_impl(Box, &)]
+        pub trait DnsConfigLoaderExt {}
+    }
+}
+
 #[auto_impl::auto_impl(Box, &)]
-pub trait ConfigLoader: Send + Sync {
+pub trait ConfigLoader: Send + Sync + DnsConfigLoaderExt {
     fn get_id(&self) -> uuid::Uuid;
     fn set_id(&self, id: uuid::Uuid);
 
@@ -239,6 +325,14 @@ pub trait ConfigLoader: Send + Sync {
     fn get_stun_servers_v6(&self) -> Option<Vec<String>>;
     fn set_stun_servers_v6(&self, servers: Option<Vec<String>>);
 
+    fn get_dns_resolvers(&self) -> Vec<String> {
+        dns::get_default_dns_resolvers()
+    }
+    fn get_dns_resolvers_config(&self) -> Option<Vec<String>> {
+        None
+    }
+    fn set_dns_resolvers(&self, _resolvers: Option<Vec<String>>) {}
+
     fn get_secure_mode(&self) -> Option<SecureModeConfig>;
     fn set_secure_mode(&self, secure_mode: Option<SecureModeConfig>);
 
@@ -426,11 +520,11 @@ pub struct ConsoleLoggerConfig {
     pub level: Option<String>,
 }
 
-#[derive(Debug, Clone, Deserialize, Serialize, PartialEq, derive_builder::Builder)]
+#[derive(Debug, Clone, Deserialize, Serialize, PartialEq, Builder)]
 pub struct LoggingConfig {
-    #[builder(setter(into, strip_option), default = None)]
+    #[builder(into)]
     pub file_logger: Option<FileLoggerConfig>,
-    #[builder(setter(into, strip_option), default = None)]
+    #[builder(into)]
     pub console_logger: Option<ConsoleLoggerConfig>,
 }
 
@@ -541,6 +635,10 @@ struct Config {
     peer: Option<Vec<PeerConfig>>,
     proxy_network: Option<Vec<ProxyNetworkConfig>>,
 
+    #[cfg(feature = "magic-dns")]
+    #[serde(default)]
+    dns: DnsConfig,
+
     vpn_portal_config: Option<VpnPortalConfig>,
 
     routes: Option<Vec<cidr::Ipv4Cidr>>,
@@ -562,6 +660,7 @@ struct Config {
     udp_whitelist: Option<Vec<String>>,
     stun_servers: Option<Vec<String>>,
     stun_servers_v6: Option<Vec<String>>,
+    dns_resolvers: Option<Vec<String>>,
 
     credential_file: Option<PathBuf>,
     source: Option<ConfigSourceConfig>,
@@ -595,6 +694,10 @@ impl TomlConfigLoader {
         Self::normalize_config_source(&mut config);
 
         config.flags_struct = Some(Self::gen_flags(config.flags.clone().unwrap_or_default()));
+        if let Some(dns_resolvers) = &config.dns_resolvers {
+            dns::validate_dns_resolvers(dns_resolvers)
+                .with_context(|| "invalid dns_resolvers config")?;
+        }
 
         let config = TomlConfigLoader {
             config: Arc::new(Mutex::new(config)),
@@ -636,6 +739,21 @@ impl TomlConfigLoader {
     }
 }
 
+impl DnsConfigLoaderExt for TomlConfigLoader {
+    cfg_select! {
+        feature = "magic-dns" => {
+            fn get_dns(&self) -> DnsConfig {
+                self.config.lock().unwrap().dns.clone()
+            }
+            fn set_dns(&self, config: DnsConfig) {
+                self.config.lock().unwrap().dns = config;
+            }
+        }
+
+        _ => {}
+    }
+}
+
 impl ConfigLoader for TomlConfigLoader {
     fn get_inst_name(&self) -> String {
         self.config
@@ -651,26 +769,17 @@ impl ConfigLoader for TomlConfigLoader {
     }
 
     fn get_hostname(&self) -> String {
-        let hostname = self.config.lock().unwrap().hostname.clone();
+        let hostname = self
+            .config
+            .lock()
+            .unwrap()
+            .hostname
+            .as_ref()
+            .map(|hostname| dns::sanitize(hostname))
+            .filter(|h| !h.is_empty());
 
-        match hostname {
-            Some(hostname) => {
-                let hostname = hostname
-                    .chars()
-                    .filter(|c| !c.is_control())
-                    .take(32)
-                    .collect::<String>();
-
-                if !hostname.is_empty() {
-                    self.set_hostname(Some(hostname.clone()));
-                    hostname
-                } else {
-                    self.set_hostname(None);
-                    gethostname::gethostname().to_string_lossy().to_string()
-                }
-            }
-            None => gethostname::gethostname().to_string_lossy().to_string(),
-        }
+        self.set_hostname(hostname.clone());
+        hostname.unwrap_or_else(|| utils::dns::sanitize(utils::hostname()))
     }
 
     fn set_hostname(&self, name: Option<String>) {
@@ -998,6 +1107,23 @@ impl ConfigLoader for TomlConfigLoader {
         self.config.lock().unwrap().stun_servers_v6 = servers;
     }
 
+    fn get_dns_resolvers(&self) -> Vec<String> {
+        self.config
+            .lock()
+            .unwrap()
+            .dns_resolvers
+            .clone()
+            .unwrap_or_else(dns::get_default_dns_resolvers)
+    }
+
+    fn get_dns_resolvers_config(&self) -> Option<Vec<String>> {
+        self.config.lock().unwrap().dns_resolvers.clone()
+    }
+
+    fn set_dns_resolvers(&self, resolvers: Option<Vec<String>>) {
+        self.config.lock().unwrap().dns_resolvers = resolvers;
+    }
+
     fn get_secure_mode(&self) -> Option<SecureModeConfig> {
         self.config.lock().unwrap().secure_mode.clone()
     }
@@ -1060,6 +1186,9 @@ impl ConfigLoader for TomlConfigLoader {
         if config.stun_servers_v6 == Some(StunInfoCollector::get_default_servers_v6()) {
             config.stun_servers_v6 = None;
         }
+        if config.dns_resolvers == Some(dns::get_default_dns_resolvers()) {
+            config.dns_resolvers = None;
+        }
         toml::to_string_pretty(&config).unwrap()
     }
 }
@@ -1291,6 +1420,51 @@ stun_servers = [
         assert_eq!(stun_servers[2], "txt:stun.easytier.cn");
     }
 
+    #[test]
+    fn test_dns_resolvers_default_and_roundtrip() {
+        let config = TomlConfigLoader::default();
+        assert_eq!(config.get_dns_resolvers_config(), None);
+        assert_eq!(config.get_dns_resolvers(), vec!["system".to_string()]);
+        assert!(!config.dump().contains("dns_resolvers"));
+
+        let config = TomlConfigLoader::new_from_str(
+            r#"
+dns_resolvers = ["system", "https://dns.alidns.com/dns-query"]
+"#,
+        )
+        .unwrap();
+        assert_eq!(
+            config.get_dns_resolvers_config().unwrap(),
+            vec![
+                "system".to_string(),
+                "https://dns.alidns.com/dns-query".to_string()
+            ]
+        );
+        assert_eq!(
+            config.get_dns_resolvers(),
+            vec![
+                "system".to_string(),
+                "https://dns.alidns.com/dns-query".to_string()
+            ]
+        );
+
+        let dumped = config.dump();
+        assert!(dumped.contains("dns_resolvers"));
+        let loaded = TomlConfigLoader::new_from_str(&dumped).unwrap();
+        assert_eq!(loaded.get_dns_resolvers(), config.get_dns_resolvers());
+    }
+
+    #[test]
+    fn test_dns_resolvers_reject_unknown_doh_without_bootstrap() {
+        let err = TomlConfigLoader::new_from_str(
+            r#"
+dns_resolvers = ["https://example.com/dns-query"]
+"#,
+        )
+        .unwrap_err();
+        assert!(err.to_string().contains("invalid dns_resolvers"));
+    }
+
     #[test]
     fn test_network_config_source_toml_roundtrip() {
         let config = TomlConfigLoader::default();

easytier/src/common/constants.rs

@@ -23,8 +23,6 @@ define_global_var!(MANUAL_CONNECTOR_RECONNECT_INTERVAL_MS, u64, 1000);
 
 define_global_var!(OSPF_UPDATE_MY_GLOBAL_FOREIGN_NETWORK_INTERVAL_SEC, u64, 10);
 
-define_global_var!(MACHINE_UID, Option<String>, None);
-
 define_global_var!(MAX_DIRECT_CONNS_PER_PEER_IN_FOREIGN_NETWORK, u32, 3);
 
 define_global_var!(DIRECT_CONNECT_TO_PUBLIC_SERVER, bool, true);

easytier/src/common/dns.rs

@@ -1,151 +0,0 @@
-use std::net::SocketAddr;
-use std::sync::Arc;
-use std::sync::atomic::AtomicBool;
-
-use anyhow::Context;
-use hickory_proto::runtime::TokioRuntimeProvider;
-use hickory_proto::xfer::Protocol;
-use hickory_resolver::config::{LookupIpStrategy, NameServerConfig, ResolverConfig, ResolverOpts};
-use hickory_resolver::name_server::{GenericConnector, TokioConnectionProvider};
-use hickory_resolver::system_conf::read_system_conf;
-use hickory_resolver::{Resolver, TokioResolver};
-use once_cell::sync::Lazy;
-use tokio::net::lookup_host;
-
-use super::error::Error;
-
-pub fn get_default_resolver_config() -> ResolverConfig {
-    let mut default_resolve_config = ResolverConfig::new();
-    default_resolve_config.add_name_server(NameServerConfig::new(
-        "223.5.5.5:53".parse().unwrap(),
-        Protocol::Udp,
-    ));
-    default_resolve_config.add_name_server(NameServerConfig::new(
-        "180.184.1.1:53".parse().unwrap(),
-        Protocol::Udp,
-    ));
-    default_resolve_config
-}
-
-pub static ALLOW_USE_SYSTEM_DNS_RESOLVER: Lazy<AtomicBool> = Lazy::new(|| AtomicBool::new(true));
-
-pub static RESOLVER: Lazy<Arc<Resolver<GenericConnector<TokioRuntimeProvider>>>> =
-    Lazy::new(|| {
-        let system_cfg = read_system_conf();
-        let mut cfg = get_default_resolver_config();
-        let mut opt = ResolverOpts::default();
-        if let Ok(s) = system_cfg {
-            for ns in s.0.name_servers() {
-                cfg.add_name_server(ns.clone());
-            }
-            opt = s.1;
-        }
-        opt.ip_strategy = LookupIpStrategy::Ipv4AndIpv6;
-        let builder = TokioResolver::builder_with_config(cfg, TokioConnectionProvider::default())
-            .with_options(opt);
-        Arc::new(builder.build())
-    });
-
-pub async fn resolve_txt_record(domain_name: &str) -> Result<String, Error> {
-    let r = RESOLVER.clone();
-    let response = r
-        .txt_lookup(domain_name)
-        .await
-        .with_context(|| format!("txt_lookup failed, domain_name: {}", domain_name))?;
-
-    let txt_record = response
-        .iter()
-        .next()
-        .with_context(|| format!("no txt record found, domain_name: {}", domain_name))?;
-
-    let txt_data = String::from_utf8_lossy(&txt_record.txt_data()[0]);
-    tracing::info!(?txt_data, ?domain_name, "get txt record");
-
-    Ok(txt_data.to_string())
-}
-
-pub async fn socket_addrs(
-    url: &url::Url,
-    default_port_number: impl Fn() -> Option<u16>,
-) -> Result<Vec<SocketAddr>, Error> {
-    let host = url.host().ok_or(Error::InvalidUrl(url.to_string()))?;
-    let port = url
-        .port()
-        .or_else(default_port_number)
-        .ok_or(Error::InvalidUrl(url.to_string()))?;
-
-    // if host is an ip address, return it directly
-    match host {
-        url::Host::Ipv4(ip) => return Ok(vec![SocketAddr::new(std::net::IpAddr::V4(ip), port)]),
-        url::Host::Ipv6(ip) => return Ok(vec![SocketAddr::new(std::net::IpAddr::V6(ip), port)]),
-        _ => {}
-    }
-    let host = host.to_string();
-
-    if ALLOW_USE_SYSTEM_DNS_RESOLVER.load(std::sync::atomic::Ordering::Relaxed) {
-        let socket_addr = format!("{}:{}", host, port);
-        match lookup_host(socket_addr).await {
-            Ok(a) => {
-                let a = a.collect();
-                tracing::debug!(?a, "system dns lookup done");
-                return Ok(a);
-            }
-            Err(e) => {
-                tracing::error!(?e, "system dns lookup failed");
-            }
-        }
-    }
-
-    // use hickory_resolver
-    let ret = RESOLVER.lookup_ip(&host).await.with_context(|| {
-        format!(
-            "hickory dns lookup_ip failed, host: {}, port: {}",
-            host, port
-        )
-    })?;
-    Ok(ret
-        .iter()
-        .map(|ip| SocketAddr::new(ip, port))
-        .collect::<Vec<_>>())
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-    use guarden::defer;
-
-    #[tokio::test]
-    async fn test_socket_addrs() {
-        let url = url::Url::parse("tcp://github-ci-test.easytier.cn:80").unwrap();
-        let addrs = socket_addrs(&url, || Some(80)).await.unwrap();
-        assert_eq!(2, addrs.len(), "addrs: {:?}", addrs);
-        println!("addrs: {:?}", addrs);
-
-        ALLOW_USE_SYSTEM_DNS_RESOLVER.store(false, std::sync::atomic::Ordering::Relaxed);
-        defer!(
-            ALLOW_USE_SYSTEM_DNS_RESOLVER.store(true, std::sync::atomic::Ordering::Relaxed);
-        );
-        let addrs = socket_addrs(&url, || Some(80)).await.unwrap();
-        assert_eq!(2, addrs.len(), "addrs: {:?}", addrs);
-        println!("addrs2: {:?}", addrs);
-    }
-
-    #[tokio::test]
-    async fn socket_addrs_preserves_explicit_zero_port() {
-        let cases = [
-            ("ws://127.0.0.1:0", 80, 0),
-            ("wss://127.0.0.1:0", 443, 0),
-            ("ws://127.0.0.1", 80, 80),
-            ("wss://127.0.0.1", 443, 443),
-        ];
-
-        for (raw_url, default_port, expected_port) in cases {
-            let url = url::Url::parse(raw_url).unwrap();
-            let addrs = socket_addrs(&url, || Some(default_port)).await.unwrap();
-            assert_eq!(
-                addrs,
-                vec![SocketAddr::from(([127, 0, 0, 1], expected_port))]
-            );
-        }
-    }
-}

easytier/src/common/error.rs

@@ -1,5 +1,4 @@
 use std::{io, result};
-
 use thiserror::Error;
 
 use crate::tunnel;
@@ -55,4 +54,6 @@ pub enum Error {
 
 pub type Result<T> = result::Result<T, Error>;
 
+pub type ErrorCollection = crate::utils::error::ErrorCollection<Error>;
+
 // impl From for std::

easytier/src/common/global_ctx.rs

@@ -1,14 +1,18 @@
+use arc_swap::ArcSwap;
+use crossbeam::atomic::AtomicCell;
+use dashmap::DashMap;
+use hmac::{Hmac, Mac};
+use sha2::Sha256;
+use socket2::Protocol;
 use std::{
     collections::{BTreeSet, HashMap, hash_map::DefaultHasher},
     hash::Hasher,
+    iter,
     net::{IpAddr, SocketAddr},
     sync::{Arc, Mutex},
     time::{SystemTime, UNIX_EPOCH},
 };
 
-use arc_swap::ArcSwap;
-use dashmap::DashMap;
-
 use super::{
     PeerId,
     config::{ConfigLoader, Flags},
@@ -31,10 +35,11 @@ use crate::{
     rpc_service::protected_port,
     tunnel::matches_protocol,
 };
-use crossbeam::atomic::AtomicCell;
-use hmac::{Hmac, Mac};
-use sha2::Sha256;
-use socket2::Protocol;
+#[cfg(feature = "magic-dns")]
+use crate::{
+    dns::config::{DnsConfigLoaderExt, DnsExportConfig, DnsGlobalCtxExt, zone::ZoneConfig},
+    utils::dns,
+};
 
 pub type NetworkIdentity = crate::common::config::NetworkIdentity;
 
@@ -48,6 +53,8 @@ pub enum GlobalCtxEvent {
     PeerConnAdded(PeerConnInfo),
     PeerConnRemoved(PeerConnInfo),
 
+    PeerInfoUpdated(Vec<PeerId>),
+
     ListenerAdded(url::Url),
     ListenerAddFailed(url::Url, String), // (url, error message)
     ListenerAcceptFailed(url::Url, String), // (url, error message)
@@ -77,6 +84,11 @@ pub enum GlobalCtxEvent {
 
     ProxyCidrsUpdated(Vec<cidr::Ipv4Cidr>, Vec<cidr::Ipv4Cidr>), // (added, removed)
 
+    UdpBroadcastRelayStartResult {
+        capture_backend: Option<String>,
+        error: Option<String>,
+    },
+
     CredentialChanged,
 }
 
@@ -250,7 +262,7 @@ impl std::fmt::Debug for GlobalCtx {
     }
 }
 
-pub type ArcGlobalCtx = std::sync::Arc<GlobalCtx>;
+pub type ArcGlobalCtx = Arc<GlobalCtx>;
 
 impl GlobalCtx {
     fn apply_disable_relay_data_flag(
@@ -282,6 +294,12 @@ impl GlobalCtx {
 
         let (event_bus, _) = tokio::sync::broadcast::channel(16);
 
+        if let Some(dns_resolvers) = config_fs.get_dns_resolvers_config()
+            && let Err(e) = crate::utils::dns::set_dns_resolvers(dns_resolvers)
+        {
+            crate::common::log::warn!("failed to set dns resolvers: {:?}", e);
+        }
+
         let stun_info_collector = StunInfoCollector::new_with_default_servers();
 
         if let Some(stun_servers) = config_fs.get_stun_servers() {
@@ -486,7 +504,7 @@ impl GlobalCtx {
     }
 
     pub fn get_hostname(&self) -> String {
-        return self.hostname.lock().unwrap().clone();
+        self.hostname.lock().unwrap().clone()
     }
 
     pub fn set_hostname(&self, hostname: String) {
@@ -781,6 +799,39 @@ impl GlobalCtx {
     }
 }
 
+#[cfg(feature = "magic-dns")]
+impl DnsGlobalCtxExt for GlobalCtx {
+    fn dns_self_zone(&self) -> ZoneConfig {
+        use hickory_proto::rr::Name;
+        let dns = self.config.get_dns();
+        let name: Name = dns
+            .name
+            .clone()
+            .unwrap_or_else(|| dns::parse(self.get_hostname()))
+            .into();
+        let fqdn = name.append_domain(&dns.domain).unwrap_or_default().into();
+        let ipv4 = self.get_ipv4().map(|ip| ip.address());
+        let ipv6 = self.get_ipv6().map(|ip| ip.address());
+        let ipv6 = ipv6.map(|a| vec![a]).unwrap_or_default();
+
+        ZoneConfig::dedicated(fqdn, ipv4, ipv6)
+    }
+
+    fn dns_export_config(&self) -> DnsExportConfig {
+        DnsExportConfig {
+            zones: self
+                .dns_iter_zones()
+                .filter(|z| z.policy.export.as_ref().is_some_and(|f| !f.disabled)) // TODO: check policies of parent zones
+                .map(ZoneConfig::into_data)
+                .collect(),
+        }
+    }
+
+    fn dns_iter_zones(&self) -> impl Iterator<Item = ZoneConfig> {
+        iter::once(self.dns_self_zone()).chain(self.config.get_dns().into_parsed().zones)
+    }
+}
+
 #[cfg(test)]
 pub mod tests {
     use crate::{

easytier/src/common/machine_id.rs

@@ -0,0 +1,594 @@
+use std::{
+    env,
+    ffi::OsString,
+    io::Write as _,
+    path::{Path, PathBuf},
+    time::{Duration, Instant},
+};
+
+use crate::utils::hostname;
+use anyhow::Context as _;
+#[cfg(unix)]
+use nix::{
+    errno::Errno,
+    fcntl::{Flock, FlockArg},
+};
+
+#[derive(Debug, Clone, Default)]
+pub struct MachineIdOptions {
+    pub explicit_machine_id: Option<String>,
+    pub state_dir: Option<PathBuf>,
+}
+
+pub fn resolve_machine_id(opts: &MachineIdOptions) -> anyhow::Result<uuid::Uuid> {
+    if let Some(explicit_machine_id) = opts.explicit_machine_id.as_deref() {
+        return Ok(parse_or_hash_machine_id(explicit_machine_id));
+    }
+
+    let state_file = resolve_machine_id_state_file(opts.state_dir.as_deref())?;
+    let allow_legacy_machine_uid_migration =
+        should_attempt_legacy_machine_uid_migration(&state_file);
+    if let Some(machine_id) = read_state_machine_id(&state_file)? {
+        return Ok(machine_id);
+    }
+
+    if let Some(machine_id) = read_legacy_machine_id_file() {
+        return persist_machine_id(&state_file, machine_id);
+    }
+
+    if allow_legacy_machine_uid_migration
+        && let Some(machine_id) = resolve_legacy_machine_uid_hash()
+    {
+        return persist_machine_id(&state_file, machine_id);
+    }
+
+    let machine_id = resolve_new_machine_id().unwrap_or_else(uuid::Uuid::new_v4);
+    persist_machine_id(&state_file, machine_id)
+}
+
+fn parse_or_hash_machine_id(raw: &str) -> uuid::Uuid {
+    if let Ok(mid) = uuid::Uuid::parse_str(raw.trim()) {
+        return mid;
+    }
+    digest_uuid_from_str(raw)
+}
+
+fn digest_uuid_from_str(raw: &str) -> uuid::Uuid {
+    let mut b = [0u8; 16];
+    crate::tunnel::generate_digest_from_str("", raw, &mut b);
+    uuid::Uuid::from_bytes(b)
+}
+
+fn resolve_machine_id_state_file(state_dir: Option<&Path>) -> anyhow::Result<PathBuf> {
+    let state_dir = match state_dir {
+        Some(dir) => dir.to_path_buf(),
+        None => default_machine_id_state_dir()?,
+    };
+    Ok(state_dir.join("machine_id"))
+}
+
+fn non_empty_os_string(value: Option<OsString>) -> Option<OsString> {
+    value.filter(|value| !value.is_empty())
+}
+
+#[cfg(target_os = "linux")]
+fn default_linux_machine_id_state_dir(
+    xdg_data_home: Option<OsString>,
+    home: Option<OsString>,
+) -> PathBuf {
+    if let Some(path) = non_empty_os_string(xdg_data_home) {
+        return PathBuf::from(path).join("easytier");
+    }
+
+    if let Some(home) = non_empty_os_string(home) {
+        return PathBuf::from(home)
+            .join(".local")
+            .join("share")
+            .join("easytier");
+    }
+
+    PathBuf::from("/var/lib/easytier")
+}
+
+fn default_machine_id_state_dir() -> anyhow::Result<PathBuf> {
+    cfg_select! {
+        target_os = "linux" => Ok(default_linux_machine_id_state_dir(
+            env::var_os("XDG_DATA_HOME"),
+            env::var_os("HOME"),
+        )),
+        all(target_os = "macos", not(feature = "macos-ne")) => {
+            let home = non_empty_os_string(env::var_os("HOME"))
+                .ok_or_else(|| anyhow::anyhow!("HOME is not set, cannot resolve machine id state directory"))?;
+            Ok(PathBuf::from(home)
+                .join("Library")
+                .join("Application Support")
+                .join("com.easytier"))
+        },
+        target_os = "windows" => {
+            let local_app_data = non_empty_os_string(env::var_os("LOCALAPPDATA")).ok_or_else(|| {
+                anyhow::anyhow!("LOCALAPPDATA is not set, cannot resolve machine id state directory")
+            })?;
+            Ok(PathBuf::from(local_app_data).join("easytier"))
+        },
+        target_os = "freebsd" => {
+            let home = non_empty_os_string(env::var_os("HOME"))
+                .ok_or_else(|| anyhow::anyhow!("HOME is not set, cannot resolve machine id state directory"))?;
+            Ok(PathBuf::from(home).join(".local").join("share").join("easytier"))
+        },
+        target_os = "android" => {
+            anyhow::bail!("machine id state directory must be provided explicitly on Android");
+        },
+        _ => anyhow::bail!("machine id state directory is unsupported on this platform"),
+    }
+}
+
+fn read_state_machine_id(path: &Path) -> anyhow::Result<Option<uuid::Uuid>> {
+    let Some(contents) = read_optional_file(path)? else {
+        return Ok(None);
+    };
+
+    let machine_id = uuid::Uuid::parse_str(contents.trim())
+        .with_context(|| format!("invalid machine id in state file {}", path.display()))?;
+    Ok(Some(machine_id))
+}
+
+fn read_legacy_machine_id_file() -> Option<uuid::Uuid> {
+    let path = legacy_machine_id_file_path()?;
+    read_legacy_machine_id_file_at(&path)
+}
+
+fn read_legacy_machine_id_file_at(path: &Path) -> Option<uuid::Uuid> {
+    let contents = match std::fs::read_to_string(path) {
+        Ok(contents) => contents,
+        Err(err) if err.kind() == std::io::ErrorKind::NotFound => return None,
+        Err(err) => {
+            tracing::warn!(
+                path = %path.display(),
+                %err,
+                "ignoring unreadable legacy machine id file"
+            );
+            return None;
+        }
+    };
+
+    match uuid::Uuid::parse_str(contents.trim()) {
+        Ok(machine_id) => Some(machine_id),
+        Err(err) => {
+            tracing::warn!(
+                path = %path.display(),
+                %err,
+                "ignoring invalid legacy machine id file"
+            );
+            None
+        }
+    }
+}
+
+fn legacy_machine_id_file_path() -> Option<PathBuf> {
+    std::env::current_exe()
+        .ok()
+        .map(|path| path.with_file_name("et_machine_id"))
+}
+
+fn read_optional_file(path: &Path) -> anyhow::Result<Option<String>> {
+    match std::fs::read_to_string(path) {
+        Ok(contents) => Ok(Some(contents)),
+        Err(err) if err.kind() == std::io::ErrorKind::NotFound => Ok(None),
+        Err(err) => Err(err).with_context(|| format!("failed to read {}", path.display())),
+    }
+}
+
+fn should_attempt_legacy_machine_uid_migration(state_file: &Path) -> bool {
+    let Some(state_dir) = state_file.parent() else {
+        return false;
+    };
+
+    let Ok(mut entries) = std::fs::read_dir(state_dir) else {
+        return false;
+    };
+    entries.any(|entry| entry.is_ok())
+}
+
+fn resolve_legacy_machine_uid_hash() -> Option<uuid::Uuid> {
+    machine_uid_seed().map(|seed| digest_uuid_from_str(seed.as_str()))
+}
+
+fn resolve_new_machine_id() -> Option<uuid::Uuid> {
+    let seed = machine_uid_seed()?;
+
+    #[cfg(target_os = "linux")]
+    {
+        let seed = linux_machine_id_seed(&seed);
+        Some(digest_uuid_from_str(&seed))
+    }
+
+    #[cfg(not(target_os = "linux"))]
+    {
+        Some(digest_uuid_from_str(&seed))
+    }
+}
+
+#[cfg(any(
+    target_os = "linux",
+    all(target_os = "macos", not(feature = "macos-ne")),
+    target_os = "windows",
+    target_os = "freebsd"
+))]
+fn machine_uid_seed() -> Option<String> {
+    machine_uid::get()
+        .ok()
+        .filter(|value| !value.trim().is_empty())
+}
+
+#[cfg(not(any(
+    target_os = "linux",
+    all(target_os = "macos", not(feature = "macos-ne")),
+    target_os = "windows",
+    target_os = "freebsd"
+)))]
+fn machine_uid_seed() -> Option<String> {
+    None
+}
+
+#[cfg(target_os = "linux")]
+fn linux_machine_id_seed(machine_uid: &str) -> String {
+    let mut seed = format!("machine_uid={machine_uid}");
+
+    let hostname = hostname();
+    if !hostname.is_empty() {
+        seed.push_str("\nhostname=");
+        seed.push_str(&hostname);
+    }
+
+    let mac_addresses = collect_linux_mac_addresses();
+    if !mac_addresses.is_empty() {
+        seed.push_str("\nmacs=");
+        seed.push_str(&mac_addresses.join(","));
+    }
+
+    seed
+}
+
+#[cfg(target_os = "linux")]
+fn collect_linux_mac_addresses() -> Vec<String> {
+    let mut macs = Vec::new();
+    let Ok(entries) = std::fs::read_dir("/sys/class/net") else {
+        return macs;
+    };
+
+    for entry in entries.flatten() {
+        let Ok(name) = entry.file_name().into_string() else {
+            continue;
+        };
+        if name == "lo" {
+            continue;
+        }
+
+        let address_path = entry.path().join("address");
+        let Ok(address) = std::fs::read_to_string(address_path) else {
+            continue;
+        };
+        let address = address.trim().to_ascii_lowercase();
+        if address.is_empty() || address == "00:00:00:00:00:00" {
+            continue;
+        }
+        macs.push(address);
+    }
+
+    macs.sort();
+    macs.dedup();
+    macs.truncate(3);
+    macs
+}
+
+fn persist_machine_id(path: &Path, machine_id: uuid::Uuid) -> anyhow::Result<uuid::Uuid> {
+    if let Some(existing) = read_state_machine_id(path)? {
+        return Ok(existing);
+    }
+
+    let _lock = MachineIdWriteLock::acquire(path)?;
+
+    if let Some(existing) = read_state_machine_id(path)? {
+        return Ok(existing);
+    }
+
+    write_uuid_file_atomically(path, machine_id)?;
+    Ok(machine_id)
+}
+
+fn write_uuid_file_atomically(path: &Path, machine_id: uuid::Uuid) -> anyhow::Result<()> {
+    let parent = path.parent().ok_or_else(|| {
+        anyhow::anyhow!(
+            "machine id state file {} has no parent directory",
+            path.display()
+        )
+    })?;
+    std::fs::create_dir_all(parent).with_context(|| {
+        format!(
+            "failed to create machine id state directory {}",
+            parent.display()
+        )
+    })?;
+
+    let tmp_path = parent.join(format!(
+        ".machine_id.tmp-{}-{}",
+        std::process::id(),
+        uuid::Uuid::new_v4()
+    ));
+    {
+        let mut file = std::fs::OpenOptions::new()
+            .write(true)
+            .create_new(true)
+            .open(&tmp_path)
+            .with_context(|| format!("failed to create {}", tmp_path.display()))?;
+        file.write_all(machine_id.to_string().as_bytes())
+            .with_context(|| format!("failed to write {}", tmp_path.display()))?;
+        file.sync_all()
+            .with_context(|| format!("failed to flush {}", tmp_path.display()))?;
+    }
+
+    if let Err(err) = std::fs::rename(&tmp_path, path) {
+        let _ = std::fs::remove_file(&tmp_path);
+        return Err(err).with_context(|| {
+            format!(
+                "failed to move machine id state file into place at {}",
+                path.display()
+            )
+        });
+    }
+
+    Ok(())
+}
+
+struct MachineIdWriteLock {
+    #[cfg(unix)]
+    _lock: Flock<std::fs::File>,
+    #[cfg(not(unix))]
+    path: PathBuf,
+}
+
+impl MachineIdWriteLock {
+    fn acquire(path: &Path) -> anyhow::Result<Self> {
+        let parent = path.parent().ok_or_else(|| {
+            anyhow::anyhow!(
+                "machine id state file {} has no parent directory",
+                path.display()
+            )
+        })?;
+        std::fs::create_dir_all(parent).with_context(|| {
+            format!(
+                "failed to create machine id state directory {}",
+                parent.display()
+            )
+        })?;
+
+        #[cfg(unix)]
+        {
+            Self::acquire_unix(path)
+        }
+
+        #[cfg(not(unix))]
+        {
+            Self::acquire_fallback(path)
+        }
+    }
+
+    #[cfg(unix)]
+    fn acquire_unix(path: &Path) -> anyhow::Result<Self> {
+        let lock_path = path.with_extension("lock");
+        let deadline = Instant::now() + Duration::from_secs(5);
+        let mut lock_file = std::fs::OpenOptions::new()
+            .read(true)
+            .write(true)
+            .create(true)
+            .truncate(false)
+            .open(&lock_path)
+            .with_context(|| format!("failed to open machine id lock {}", lock_path.display()))?;
+
+        loop {
+            match Flock::lock(lock_file, FlockArg::LockExclusiveNonblock) {
+                Ok(lock) => return Ok(Self { _lock: lock }),
+                Err((file, Errno::EAGAIN)) => {
+                    if Instant::now() >= deadline {
+                        anyhow::bail!(
+                            "timed out waiting for machine id lock {}",
+                            lock_path.display()
+                        );
+                    }
+                    lock_file = file;
+                    std::thread::sleep(Duration::from_millis(50));
+                }
+                Err((_file, err)) => {
+                    anyhow::bail!(
+                        "failed to acquire machine id lock {}: {}",
+                        lock_path.display(),
+                        err
+                    );
+                }
+            }
+        }
+    }
+
+    #[cfg(not(unix))]
+    fn acquire_fallback(path: &Path) -> anyhow::Result<Self> {
+        let lock_path = path.with_extension("lock");
+        let deadline = Instant::now() + Duration::from_secs(5);
+
+        loop {
+            match std::fs::OpenOptions::new()
+                .write(true)
+                .create_new(true)
+                .open(&lock_path)
+            {
+                Ok(mut file) => {
+                    writeln!(file, "pid={}", std::process::id()).ok();
+                    return Ok(Self { path: lock_path });
+                }
+                Err(err) if err.kind() == std::io::ErrorKind::AlreadyExists => {
+                    if should_reap_stale_lock_file(&lock_path) {
+                        let _ = std::fs::remove_file(&lock_path);
+                        continue;
+                    }
+                    if Instant::now() >= deadline {
+                        anyhow::bail!(
+                            "timed out waiting for machine id lock {}",
+                            lock_path.display()
+                        );
+                    }
+                    std::thread::sleep(Duration::from_millis(50));
+                }
+                Err(err) => {
+                    return Err(err).with_context(|| {
+                        format!("failed to acquire machine id lock {}", lock_path.display())
+                    });
+                }
+            }
+        }
+    }
+}
+
+#[cfg(not(unix))]
+fn should_reap_stale_lock_file(lock_path: &Path) -> bool {
+    const STALE_LOCK_AGE: Duration = Duration::from_secs(30);
+
+    let Ok(metadata) = std::fs::metadata(lock_path) else {
+        return false;
+    };
+    let Ok(modified) = metadata.modified() else {
+        return false;
+    };
+    modified
+        .elapsed()
+        .is_ok_and(|elapsed| elapsed >= STALE_LOCK_AGE)
+}
+
+impl Drop for MachineIdWriteLock {
+    fn drop(&mut self) {
+        #[cfg(not(unix))]
+        let _ = std::fs::remove_file(&self.path);
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_resolve_machine_id_uses_uuid_seed_verbatim() {
+        let raw = "33333333-3333-3333-3333-333333333333".to_string();
+        let opts = MachineIdOptions {
+            explicit_machine_id: Some(raw.clone()),
+            state_dir: None,
+        };
+        assert_eq!(
+            resolve_machine_id(&opts).unwrap(),
+            uuid::Uuid::parse_str(&raw).unwrap()
+        );
+    }
+
+    #[test]
+    fn test_resolve_machine_id_reads_state_file() {
+        let temp_dir = tempfile::tempdir().unwrap();
+        let expected = uuid::Uuid::new_v4();
+        std::fs::write(temp_dir.path().join("machine_id"), expected.to_string()).unwrap();
+
+        let opts = MachineIdOptions {
+            explicit_machine_id: None,
+            state_dir: Some(temp_dir.path().to_path_buf()),
+        };
+
+        assert_eq!(resolve_machine_id(&opts).unwrap(), expected);
+    }
+
+    #[test]
+    fn test_read_legacy_machine_id_file_ignores_read_errors() {
+        let temp_dir = tempfile::tempdir().unwrap();
+
+        assert_eq!(read_legacy_machine_id_file_at(temp_dir.path()), None);
+    }
+
+    #[test]
+    fn test_write_uuid_file_atomically_writes_expected_contents() {
+        let temp_dir = tempfile::tempdir().unwrap();
+        let machine_id = uuid::Uuid::new_v4();
+        let state_file = temp_dir.path().join("machine_id");
+
+        write_uuid_file_atomically(&state_file, machine_id).unwrap();
+
+        assert_eq!(
+            std::fs::read_to_string(state_file).unwrap(),
+            machine_id.to_string()
+        );
+    }
+
+    #[test]
+    fn test_non_empty_os_string_filters_empty_values() {
+        assert_eq!(non_empty_os_string(Some(OsString::new())), None);
+        assert_eq!(
+            non_empty_os_string(Some(OsString::from("foo"))),
+            Some(OsString::from("foo"))
+        );
+    }
+
+    #[cfg(target_os = "linux")]
+    #[test]
+    fn test_default_linux_machine_id_state_dir_falls_back_in_order() {
+        assert_eq!(
+            default_linux_machine_id_state_dir(
+                Some(OsString::from("/tmp/xdg")),
+                Some(OsString::from("/tmp/home"))
+            ),
+            PathBuf::from("/tmp/xdg").join("easytier")
+        );
+        assert_eq!(
+            default_linux_machine_id_state_dir(
+                Some(OsString::new()),
+                Some(OsString::from("/tmp/home"))
+            ),
+            PathBuf::from("/tmp/home")
+                .join(".local")
+                .join("share")
+                .join("easytier")
+        );
+        assert_eq!(
+            default_linux_machine_id_state_dir(Some(OsString::new()), Some(OsString::new())),
+            PathBuf::from("/var/lib/easytier")
+        );
+    }
+
+    #[test]
+    fn test_persist_machine_id_creates_missing_state_dir() {
+        let temp_dir = tempfile::tempdir().unwrap();
+        let state_file = temp_dir.path().join("nested").join("machine_id");
+        let machine_id = uuid::Uuid::new_v4();
+
+        assert_eq!(
+            persist_machine_id(&state_file, machine_id).unwrap(),
+            machine_id
+        );
+        assert_eq!(
+            std::fs::read_to_string(state_file).unwrap(),
+            machine_id.to_string()
+        );
+    }
+
+    #[test]
+    fn test_legacy_machine_uid_migration_requires_existing_state_dir_content() {
+        let temp_dir = tempfile::tempdir().unwrap();
+        let missing_state_file = temp_dir.path().join("missing").join("machine_id");
+        assert!(!should_attempt_legacy_machine_uid_migration(
+            &missing_state_file
+        ));
+
+        let empty_dir = temp_dir.path().join("empty");
+        std::fs::create_dir_all(&empty_dir).unwrap();
+        assert!(!should_attempt_legacy_machine_uid_migration(
+            &empty_dir.join("machine_id")
+        ));
+
+        std::fs::write(empty_dir.join("config.toml"), "x=1").unwrap();
+        assert!(should_attempt_legacy_machine_uid_migration(
+            &empty_dir.join("machine_id")
+        ));
+    }
+}

easytier/src/common/mod.rs

@@ -1,26 +1,23 @@
 use std::{
     fmt::Debug,
     future,
-    io::Write as _,
     sync::{Arc, Mutex},
 };
 use time::util::refresh_tz;
 use tokio::{task::JoinSet, time::timeout};
 use tracing::Instrument;
 
-use crate::{set_global_var, use_global_var};
-
 pub mod acl_processor;
 pub mod compressor;
 pub mod config;
 pub mod constants;
-pub mod dns;
 pub mod env_parser;
 pub mod error;
 pub mod global_ctx;
 pub mod idn;
 pub mod ifcfg;
 pub mod log;
+pub mod machine_id;
 pub mod netns;
 pub mod network;
 pub mod os_info;
@@ -31,6 +28,8 @@ pub mod token_bucket;
 pub mod tracing_rolling_appender;
 pub mod upnp;
 
+pub use machine_id::{MachineIdOptions, resolve_machine_id};
+
 pub fn get_logger_timer<F: time::formatting::Formattable>(
     format: F,
 ) -> tracing_subscriber::fmt::time::OffsetTime<F> {
@@ -96,71 +95,6 @@ pub fn join_joinset_background<T: Debug + Send + Sync + 'static>(
     );
 }
 
-pub fn set_default_machine_id(mid: Option<String>) {
-    set_global_var!(MACHINE_UID, mid);
-}
-
-pub fn get_machine_id() -> uuid::Uuid {
-    if let Some(default_mid) = use_global_var!(MACHINE_UID) {
-        if let Ok(mid) = uuid::Uuid::parse_str(default_mid.trim()) {
-            return mid;
-        }
-        let mut b = [0u8; 16];
-        crate::tunnel::generate_digest_from_str("", &default_mid, &mut b);
-        return uuid::Uuid::from_bytes(b);
-    }
-
-    // a path same as the binary
-    let machine_id_file = std::env::current_exe()
-        .map(|x| x.with_file_name("et_machine_id"))
-        .unwrap_or_else(|_| std::path::PathBuf::from("et_machine_id"));
-
-    // try load from local file
-    if let Ok(mid) = std::fs::read_to_string(&machine_id_file)
-        && let Ok(mid) = uuid::Uuid::parse_str(mid.trim())
-    {
-        return mid;
-    }
-
-    #[cfg(any(
-        target_os = "linux",
-        all(target_os = "macos", not(feature = "macos-ne")),
-        target_os = "windows",
-        target_os = "freebsd"
-    ))]
-    let gen_mid = machine_uid::get()
-        .map(|x| {
-            if x.is_empty() {
-                return uuid::Uuid::new_v4();
-            }
-            let mut b = [0u8; 16];
-            crate::tunnel::generate_digest_from_str("", x.as_str(), &mut b);
-            uuid::Uuid::from_bytes(b)
-        })
-        .ok();
-
-    #[cfg(not(any(
-        target_os = "linux",
-        all(target_os = "macos", not(feature = "macos-ne")),
-        target_os = "windows",
-        target_os = "freebsd"
-    )))]
-    let gen_mid = None;
-
-    if let Some(mid) = gen_mid {
-        return mid;
-    }
-
-    let gen_mid = uuid::Uuid::new_v4();
-
-    // try save to local file
-    if let Ok(mut file) = std::fs::File::create(machine_id_file) {
-        let _ = file.write_all(gen_mid.to_string().as_bytes());
-    }
-
-    gen_mid
-}
-
 pub fn shrink_dashmap<K: Eq + std::hash::Hash, V>(
     map: &dashmap::DashMap<K, V>,
     threshold: Option<usize>,
@@ -210,12 +144,4 @@ mod tests {
         assert_eq!(weak_js.weak_count(), 0);
         assert_eq!(weak_js.strong_count(), 0);
     }
-
-    #[test]
-    fn test_get_machine_id_uses_uuid_seed_verbatim() {
-        let raw = "33333333-3333-3333-3333-333333333333".to_string();
-        set_default_machine_id(Some(raw.clone()));
-        assert_eq!(get_machine_id(), uuid::Uuid::parse_str(&raw).unwrap());
-        set_default_machine_id(None);
-    }
 }

easytier/src/common/stats_manager.rs

@@ -85,6 +85,15 @@ pub enum MetricName {
     /// Traffic packets forwarded for foreign network, forward
     TrafficPacketsForeignForwardForwarded,
 
+    /// UDP broadcast relay packets captured from the raw socket
+    UdpBroadcastRelayPacketsCaptured,
+    /// UDP broadcast relay packets ignored before forwarding
+    UdpBroadcastRelayPacketsIgnored,
+    /// UDP broadcast relay packets forwarded
+    UdpBroadcastRelayPacketsForwarded,
+    /// UDP broadcast relay packets that failed to forward
+    UdpBroadcastRelayPacketsForwardFailed,
+
     /// Compression bytes before compression
     CompressionBytesRxBefore,
     /// Compression bytes after compression
@@ -167,6 +176,19 @@ impl fmt::Display for MetricName {
                 write!(f, "traffic_packets_foreign_forward_forwarded")
             }
 
+            MetricName::UdpBroadcastRelayPacketsCaptured => {
+                write!(f, "udp_broadcast_relay_packets_captured")
+            }
+            MetricName::UdpBroadcastRelayPacketsIgnored => {
+                write!(f, "udp_broadcast_relay_packets_ignored")
+            }
+            MetricName::UdpBroadcastRelayPacketsForwarded => {
+                write!(f, "udp_broadcast_relay_packets_forwarded")
+            }
+            MetricName::UdpBroadcastRelayPacketsForwardFailed => {
+                write!(f, "udp_broadcast_relay_packets_forward_failed")
+            }
+
             MetricName::CompressionBytesRxBefore => write!(f, "compression_bytes_rx_before"),
             MetricName::CompressionBytesRxAfter => write!(f, "compression_bytes_rx_after"),
             MetricName::CompressionBytesTxBefore => write!(f, "compression_bytes_tx_before"),

easytier/src/common/stun.rs

@@ -11,7 +11,7 @@ use crossbeam::atomic::AtomicCell;
 use rand::seq::IteratorRandom;
 use socket2::{SockAddr, SockRef};
 use tokio::io::{AsyncReadExt, AsyncWriteExt};
-use tokio::net::{UdpSocket, lookup_host};
+use tokio::net::UdpSocket;
 use tokio::sync::{Mutex, broadcast};
 use tokio::task::JoinSet;
 use tracing::{Instrument, Level};
@@ -20,10 +20,9 @@ use bytecodec::{DecodeExt, EncodeExt};
 use stun_codec::rfc5389::methods::BINDING;
 use stun_codec::{Message, MessageClass, MessageDecoder, MessageEncoder};
 
-use crate::common::error::Error;
-
-use super::dns::resolve_txt_record;
 use super::stun_codec_ext::*;
+use crate::common::error::Error;
+use crate::utils::dns::{resolve_host, txt_resolve};
 
 const DEFAULT_UDP_STUN_SERVERS: &[&str] = &[
     "txt:stun.easytier.cn",
@@ -61,9 +60,16 @@ impl HostResolverIter {
         }
     }
 
-    async fn get_txt_record(domain_name: &str) -> Result<Vec<String>, Error> {
-        let txt_data = resolve_txt_record(domain_name).await?;
-        Ok(txt_data.split(" ").map(|x| x.to_string()).collect())
+    fn parse_ipv6_socket_addr_without_brackets(host: &str) -> Option<SocketAddr> {
+        if host.parse::<IpAddr>().is_ok() {
+            return None;
+        }
+
+        let (ip, port) = host.rsplit_once(':')?;
+        Some(SocketAddr::new(
+            IpAddr::V6(ip.parse().ok()?),
+            port.parse().ok()?,
+        ))
     }
 
     #[async_recursion::async_recursion]
@@ -74,15 +80,10 @@ impl HostResolverIter {
             }
 
             let host = self.hostnames.remove(0);
-            let host = if host.contains(':') {
-                host
-            } else {
-                format!("{}:3478", host)
-            };
 
             if host.starts_with("txt:") {
                 let domain_name = host.trim_start_matches("txt:");
-                match Self::get_txt_record(domain_name).await {
+                match txt_resolve(domain_name).await {
                     Ok(hosts) => {
                         tracing::info!(
                             ?domain_name,
@@ -104,22 +105,53 @@ impl HostResolverIter {
             }
 
             let use_ipv6 = self.use_ipv6;
-
-            match lookup_host(&host).await {
-                Ok(ips) => {
-                    self.ips = ips
-                        .filter(|x| if use_ipv6 { x.is_ipv6() } else { x.is_ipv4() })
-                        .choose_multiple(&mut rand::thread_rng(), self.max_ip_per_domain as usize);
-
-                    if self.ips.is_empty() {
-                        return self.next().await;
-                    }
+            if let Ok(addr) = host.parse::<SocketAddr>() {
+                if (use_ipv6 && addr.is_ipv6()) || (!use_ipv6 && addr.is_ipv4()) {
+                    self.ips = vec![addr];
                 }
-                Err(e) => {
-                    tracing::warn!(?host, ?e, "lookup host for stun failed");
+                if self.ips.is_empty() {
                     return self.next().await;
                 }
-            };
+            } else if let Some(addr) = Self::parse_ipv6_socket_addr_without_brackets(&host) {
+                if use_ipv6 {
+                    self.ips = vec![addr];
+                }
+                if self.ips.is_empty() {
+                    return self.next().await;
+                }
+            } else {
+                let (host, port) = if let Ok(ip) = host.parse::<IpAddr>() {
+                    (ip.to_string(), 3478)
+                } else if let Ok(url) = url::Url::parse(&format!("stun://{}", host)) {
+                    let Some(parsed_host) = url.host_str() else {
+                        tracing::warn!(?host, "parse stun host failed");
+                        return self.next().await;
+                    };
+                    (parsed_host.to_string(), url.port().unwrap_or(3478))
+                } else {
+                    (host, 3478)
+                };
+
+                match resolve_host(&host, port).await {
+                    Ok(ips) => {
+                        self.ips = ips
+                            .into_iter()
+                            .filter(|x| if use_ipv6 { x.is_ipv6() } else { x.is_ipv4() })
+                            .choose_multiple(
+                                &mut rand::thread_rng(),
+                                self.max_ip_per_domain as usize,
+                            );
+
+                        if self.ips.is_empty() {
+                            return self.next().await;
+                        }
+                    }
+                    Err(e) => {
+                        tracing::warn!(?host, ?e, "resolve host for stun failed");
+                        return self.next().await;
+                    }
+                };
+            }
         }
 
         Some(self.ips.remove(0))
@@ -1349,6 +1381,26 @@ mod tests {
 
     use super::*;
 
+    #[test]
+    fn parse_ipv6_socket_addr_without_brackets_rejects_plain_ipv6_literals() {
+        assert_eq!(
+            HostResolverIter::parse_ipv6_socket_addr_without_brackets("2001:db8::1"),
+            None
+        );
+        assert_eq!(
+            HostResolverIter::parse_ipv6_socket_addr_without_brackets("2001:db8:0:0:0:0:0:1"),
+            None
+        );
+    }
+
+    #[test]
+    fn parse_ipv6_socket_addr_without_brackets_accepts_unambiguous_port() {
+        assert_eq!(
+            HostResolverIter::parse_ipv6_socket_addr_without_brackets("::1:55355"),
+            Some("[::1]:55355".parse().unwrap())
+        );
+    }
+
     #[tokio::test]
     async fn test_udp_nat_type_detector() {
         let collector = StunInfoCollector::new(
@@ -1563,6 +1615,6 @@ mod tests {
         });
         let stun_servers = vec!["::1:55355".to_string()];
         let ret = StunInfoCollector::get_public_ipv6(&stun_servers).await;
-        println!("{:#?}", ret);
+        assert_eq!(ret, Some(Ipv6Addr::LOCALHOST));
     }
 }

easytier/src/connector/direct.rs

@@ -12,10 +12,7 @@ use std::{
 };
 
 use crate::{
-    common::{
-        PeerId, dns::socket_addrs, error::Error, global_ctx::ArcGlobalCtx,
-        stun::StunInfoCollectorTrait,
-    },
+    common::{PeerId, error::Error, global_ctx::ArcGlobalCtx, stun::StunInfoCollectorTrait},
     connector::udp_hole_punch::handle_rpc_result,
     peers::{
         peer_conn::PeerConnId,
@@ -40,6 +37,7 @@ use super::{
     udp_hole_punch,
 };
 use crate::tunnel::{FromUrl, IpScheme, TunnelScheme, matches_scheme};
+use crate::utils::dns::socket_addrs;
 use anyhow::Context;
 use rand::Rng;
 use socket2::Protocol;

easytier/src/connector/dns_connector.rs

@@ -1,19 +1,15 @@
 use std::{net::SocketAddr, sync::Arc};
 
 use super::{create_connector_by_url, http_connector::TunnelWithInfo};
+use crate::utils::dns::{srv_lookup, txt_resolve};
 use crate::{
-    common::{
-        dns::{RESOLVER, resolve_txt_record},
-        error::Error,
-        global_ctx::ArcGlobalCtx,
-        log,
-    },
+    common::{error::Error, global_ctx::ArcGlobalCtx, log},
     proto::common::TunnelInfo,
     tunnel::{IpScheme, IpVersion, Tunnel, TunnelConnector, TunnelError, TunnelScheme},
 };
 use anyhow::Context;
 use dashmap::DashSet;
-use hickory_resolver::proto::rr::rdata::SRV;
+use hickory_proto::rr::rdata::SRV;
 use rand::{Rng as _, seq::SliceRandom};
 use strum::VariantArray;
 
@@ -58,14 +54,13 @@ impl DnsTunnelConnector {
         &self,
         domain_name: &str,
     ) -> Result<Box<dyn TunnelConnector>, Error> {
-        let txt_data = resolve_txt_record(domain_name)
+        let txt_data = txt_resolve(domain_name)
             .await
             .with_context(|| format!("resolve txt record failed, domain_name: {}", domain_name))?;
 
         let candidate_urls = txt_data
-            .split(" ")
-            .map(|s| s.to_string())
-            .filter_map(|s| url::Url::parse(s.as_str()).ok())
+            .iter()
+            .filter_map(|s| url::Url::parse(s).ok())
             .collect::<Vec<_>>();
 
         // shuffle candidate_urls and get the first one
@@ -73,7 +68,7 @@ impl DnsTunnelConnector {
             .choose(&mut rand::thread_rng())
             .with_context(|| {
                 format!(
-                    "no valid url found, txt_data: {}, expecting an url list splitted by space",
+                    "no valid url found, txt_data: {:?}, expecting an url list split by space",
                     txt_data
                 )
             })?;
@@ -83,26 +78,23 @@ impl DnsTunnelConnector {
         Ok(connector)
     }
 
-    fn handle_one_srv_record(record: &SRV, protocol: IpScheme) -> Result<(url::Url, u64), Error> {
+    fn handle_one_srv_record(record: SRV, protocol: IpScheme) -> Result<(url::Url, u64), Error> {
         // port must be non-zero
-        if record.port() == 0 {
+        if record.port == 0 {
             return Err(anyhow::anyhow!("port must be non-zero").into());
         }
 
-        let connector_dst = record.target().to_utf8();
-        let dst_url = format!("{}://{}:{}", protocol, connector_dst, record.port());
+        let connector_dst = record.target.to_utf8();
+        let dst_url = format!("{}://{}:{}", protocol, connector_dst, record.port);
 
         Ok((
             dst_url.parse().with_context(|| {
                 format!(
                     "parse dst_url failed, protocol: {}, connector_dst: {}, port: {}, dst_url: {}",
-                    protocol,
-                    connector_dst,
-                    record.port(),
-                    dst_url
+                    protocol, connector_dst, record.port, dst_url
                 )
             })?,
-            record.priority() as _,
+            record.priority as _,
         ))
     }
 
@@ -122,14 +114,9 @@ impl DnsTunnelConnector {
         let srv_lookup_tasks = srv_domains
             .iter()
             .map(|(protocol, srv_domain)| {
-                let resolver = RESOLVER.clone();
                 let responses = responses.clone();
                 async move {
-                    let response = resolver.srv_lookup(srv_domain).await.with_context(|| {
-                        format!("srv_lookup failed, srv_domain: {}", srv_domain)
-                    })?;
-                    tracing::info!(?response, ?srv_domain, "srv_lookup response");
-                    for record in response.iter() {
+                    for record in srv_lookup(srv_domain).await? {
                         let parsed_record = Self::handle_one_srv_record(record, **protocol);
                         tracing::info!(?parsed_record, ?srv_domain, "parsed_record");
                         if let Err(e) = &parsed_record {

easytier/src/connector/manual.rs

@@ -1,13 +1,15 @@
 use std::{
     collections::BTreeSet,
+    future::Future,
     sync::{Arc, Weak},
+    time::{Duration, Instant},
 };
 
 use dashmap::DashSet;
 use tokio::{sync::mpsc, task::JoinSet, time::timeout};
 
 use crate::{
-    common::{PeerId, dns::socket_addrs, join_joinset_background},
+    common::{PeerId, join_joinset_background},
     peers::peer_conn::PeerConnId,
     proto::{
         api::instance::{
@@ -16,10 +18,12 @@ use crate::{
         },
         rpc_types::{self, controller::BaseController},
     },
-    tunnel::{IpVersion, TunnelConnector},
+    tunnel::{IpVersion, TunnelConnector, TunnelScheme, matches_scheme},
     utils::weak_upgrade,
 };
 
+use super::create_connector_by_url;
+use crate::utils::dns::socket_addrs;
 use crate::{
     common::{
         error::Error,
@@ -30,8 +34,6 @@ use crate::{
     use_global_var,
 };
 
-use super::create_connector_by_url;
-
 type ConnectorMap = Arc<DashSet<url::Url>>;
 
 #[derive(Debug, Clone)]
@@ -83,6 +85,55 @@ impl ManualConnectorManager {
         ret
     }
 
+    fn reconnect_timeout(dead_url: &url::Url) -> Duration {
+        let use_long_timeout = matches_scheme!(
+            dead_url,
+            TunnelScheme::Http | TunnelScheme::Https | TunnelScheme::Txt | TunnelScheme::Srv
+        ) || matches!(dead_url.scheme(), "ws" | "wss");
+
+        Duration::from_secs(if use_long_timeout { 20 } else { 2 })
+    }
+
+    fn remaining_budget(started_at: Instant, total_timeout: Duration) -> Option<Duration> {
+        let remaining = total_timeout.checked_sub(started_at.elapsed())?;
+        (!remaining.is_zero()).then_some(remaining)
+    }
+
+    fn emit_connect_error(
+        data: &ConnectorManagerData,
+        dead_url: &url::Url,
+        ip_version: IpVersion,
+        error: &Error,
+    ) {
+        data.global_ctx.issue_event(GlobalCtxEvent::ConnectError(
+            dead_url.to_string(),
+            format!("{:?}", ip_version),
+            format!("{:#?}", error),
+        ));
+    }
+
+    fn reconnect_timeout_error(stage: &str, duration: Duration) -> Error {
+        Error::AnyhowError(anyhow::anyhow!("{} timeout after {:?}", stage, duration))
+    }
+
+    async fn with_reconnect_timeout<T, F>(
+        stage: &'static str,
+        started_at: Instant,
+        total_timeout: Duration,
+        fut: F,
+    ) -> Result<T, Error>
+    where
+        F: Future<Output = Result<T, Error>>,
+    {
+        let remaining = Self::remaining_budget(started_at, total_timeout)
+            .ok_or_else(|| Self::reconnect_timeout_error(stage, started_at.elapsed()))?;
+        timeout(remaining, fut)
+            .await
+            .map_err(|_| Self::reconnect_timeout_error(stage, remaining))?
+    }
+}
+
+impl ManualConnectorManager {
     pub fn add_connector<T>(&self, connector: T)
     where
         T: TunnelConnector + 'static,
@@ -242,11 +293,18 @@ impl ManualConnectorManager {
 
     async fn conn_reconnect_with_ip_version(
         data: Arc<ConnectorManagerData>,
-        dead_url: String,
+        dead_url: url::Url,
         ip_version: IpVersion,
+        started_at: Instant,
+        total_timeout: Duration,
     ) -> Result<ReconnResult, Error> {
-        let connector =
-            create_connector_by_url(&dead_url, &data.global_ctx.clone(), ip_version).await?;
+        let connector = Self::with_reconnect_timeout(
+            "resolve",
+            started_at,
+            total_timeout,
+            create_connector_by_url(dead_url.as_str(), &data.global_ctx, ip_version),
+        )
+        .await?;
 
         data.global_ctx
             .issue_event(GlobalCtxEvent::Connecting(connector.remote_url()));
@@ -257,10 +315,25 @@ impl ManualConnectorManager {
             )));
         };
 
-        let (peer_id, conn_id) = pm.try_direct_connect(connector).await?;
+        let tunnel = Self::with_reconnect_timeout(
+            "connect",
+            started_at,
+            total_timeout,
+            pm.connect_tunnel(connector),
+        )
+        .await?;
+
+        let (peer_id, conn_id) = Self::with_reconnect_timeout(
+            "handshake",
+            started_at,
+            total_timeout,
+            pm.add_client_tunnel_with_peer_id_hint(tunnel, true, None),
+        )
+        .await?;
+
         tracing::info!("reconnect succ: {} {} {}", peer_id, conn_id, dead_url);
         Ok(ReconnResult {
-            dead_url,
+            dead_url: dead_url.to_string(),
             peer_id,
             conn_id,
         })
@@ -273,22 +346,33 @@ impl ManualConnectorManager {
         tracing::info!("reconnect: {}", dead_url);
 
         let mut ip_versions = vec![];
-        if dead_url.scheme() == "ring" || dead_url.scheme() == "txt" || dead_url.scheme() == "srv" {
+        if matches_scheme!(
+            dead_url,
+            TunnelScheme::Ring | TunnelScheme::Txt | TunnelScheme::Srv
+        ) {
             ip_versions.push(IpVersion::Both);
         } else {
-            let converted_dead_url = crate::common::idn::convert_idn_to_ascii(dead_url.clone())?;
-            let addrs = match socket_addrs(&converted_dead_url, || Some(1000)).await {
+            let converted_dead_url =
+                match crate::common::idn::convert_idn_to_ascii(dead_url.clone()) {
+                    Ok(url) => url,
+                    Err(error) => {
+                        let error: Error = error.into();
+                        Self::emit_connect_error(&data, &dead_url, IpVersion::Both, &error);
+                        return Err(error);
+                    }
+                };
+            let addrs = match Self::with_reconnect_timeout(
+                "resolve",
+                Instant::now(),
+                Self::reconnect_timeout(&dead_url),
+                socket_addrs(&converted_dead_url, || Some(1000)),
+            )
+            .await
+            {
                 Ok(addrs) => addrs,
-                Err(e) => {
-                    data.global_ctx.issue_event(GlobalCtxEvent::ConnectError(
-                        dead_url.to_string(),
-                        format!("{:?}", IpVersion::Both),
-                        format!("{:?}", e),
-                    ));
-                    return Err(Error::AnyhowError(anyhow::anyhow!(
-                        "get ip from url failed: {:?}",
-                        e
-                    )));
+                Err(error) => {
+                    Self::emit_connect_error(&data, &dead_url, IpVersion::Both, &error);
+                    return Err(error);
                 }
             };
             tracing::info!(?addrs, ?dead_url, "get ip from url done");
@@ -313,46 +397,24 @@ impl ManualConnectorManager {
             "cannot get ip from url"
         )));
         for ip_version in ip_versions {
-            let use_long_timeout = dead_url.scheme() == "http"
-                || dead_url.scheme() == "https"
-                || dead_url.scheme() == "ws"
-                || dead_url.scheme() == "wss"
-                || dead_url.scheme() == "txt"
-                || dead_url.scheme() == "srv";
-            let ret = timeout(
-                // allow http/websocket connector to wait longer
-                std::time::Duration::from_secs(if use_long_timeout { 20 } else { 2 }),
-                Self::conn_reconnect_with_ip_version(
-                    data.clone(),
-                    dead_url.to_string(),
-                    ip_version,
-                ),
+            let started_at = Instant::now();
+            let ret = Self::conn_reconnect_with_ip_version(
+                data.clone(),
+                dead_url.clone(),
+                ip_version,
+                started_at,
+                Self::reconnect_timeout(&dead_url),
             )
             .await;
             tracing::info!("reconnect: {} done, ret: {:?}", dead_url, ret);
 
             match ret {
-                Ok(Ok(_)) => {
-                    // 外层和内层都成功：解包并跳出
-                    reconn_ret = ret.unwrap();
-                    break;
-                }
-                Ok(Err(e)) => {
-                    // 外层成功，内层失败
-                    reconn_ret = Err(e);
-                }
-                Err(e) => {
-                    // 外层失败
-                    reconn_ret = Err(e.into());
+                Ok(result) => return Ok(result),
+                Err(error) => {
+                    Self::emit_connect_error(&data, &dead_url, ip_version, &error);
+                    reconn_ret = Err(error);
                 }
             }
-
-            // 发送事件（只有在未 break 时才执行）
-            data.global_ctx.issue_event(GlobalCtxEvent::ConnectError(
-                dead_url.to_string(),
-                format!("{:?}", ip_version),
-                format!("{:?}", reconn_ret),
-            ));
         }
 
         reconn_ret
@@ -388,6 +450,54 @@ mod tests {
 
     use super::*;
 
+    #[tokio::test]
+    async fn reconnect_timeout_reports_exhausted_budget_for_stage() {
+        let started_at = Instant::now() - Duration::from_millis(50);
+        let err = ManualConnectorManager::with_reconnect_timeout(
+            "resolve",
+            started_at,
+            Duration::from_millis(1),
+            async { Ok::<(), Error>(()) },
+        )
+        .await
+        .unwrap_err();
+
+        let message = err.to_string();
+        assert!(message.contains("resolve timeout after"));
+    }
+
+    #[tokio::test]
+    async fn reconnect_timeout_reports_stage_timeout_with_remaining_budget() {
+        let err = ManualConnectorManager::with_reconnect_timeout(
+            "handshake",
+            Instant::now(),
+            Duration::from_millis(10),
+            async {
+                tokio::time::sleep(Duration::from_millis(50)).await;
+                Ok::<(), Error>(())
+            },
+        )
+        .await
+        .unwrap_err();
+
+        let message = err.to_string();
+        assert!(message.contains("handshake timeout after"));
+    }
+
+    #[tokio::test]
+    async fn reconnect_timeout_preserves_success_within_budget() {
+        let result = ManualConnectorManager::with_reconnect_timeout(
+            "connect",
+            Instant::now(),
+            Duration::from_millis(50),
+            async { Ok::<_, Error>(123_u32) },
+        )
+        .await
+        .unwrap();
+
+        assert_eq!(result, 123);
+    }
+
     #[tokio::test]
     async fn test_reconnect_with_connecting_addr() {
         set_global_var!(MANUAL_CONNECTOR_RECONNECT_INTERVAL_MS, 1);

easytier/src/connector/mod.rs

@@ -1,14 +1,14 @@
 use std::net::{IpAddr, Ipv6Addr, SocketAddr, SocketAddrV4, SocketAddrV6};
 
 use crate::{
-    common::{dns::socket_addrs, error::Error, global_ctx::ArcGlobalCtx, idn},
+    common::{error::Error, global_ctx::ArcGlobalCtx, idn},
     connector::dns_connector::DnsTunnelConnector,
     proto::common::PeerFeatureFlag,
     tunnel::{
         self, IpScheme, IpVersion, TunnelConnector, TunnelError, TunnelScheme,
         ring::RingTunnelConnector, tcp::TcpTunnelConnector, udp::UdpTunnelConnector,
     },
-    utils::BoxExt,
+    utils::{BoxExt, dns::socket_addrs},
 };
 use http_connector::HttpTunnelConnector;
 use rand::seq::SliceRandom;

easytier/src/core.rs

@@ -484,6 +484,15 @@ struct NetworkOptions {
     )]
     disable_upnp: Option<bool>,
 
+    #[arg(
+        long,
+        env = "ET_ENABLE_UDP_BROADCAST_RELAY",
+        help = t!("core_clap.enable_udp_broadcast_relay").to_string(),
+        num_args = 0..=1,
+        default_missing_value = "true"
+    )]
+    enable_udp_broadcast_relay: Option<bool>,
+
     #[arg(
         long,
         env = "ET_RELAY_ALL_PEER_RPC",
@@ -569,19 +578,6 @@ struct NetworkOptions {
     )]
     port_forward: Vec<url::Url>,
 
-    #[arg(
-        long,
-        env = "ET_ACCEPT_DNS",
-        help = t!("core_clap.accept_dns").to_string(),
-    )]
-    accept_dns: Option<bool>,
-
-    #[arg(
-        long = "tld-dns-zone",
-        env = "ET_TLD_DNS_ZONE",
-        help = t!("core_clap.tld_dns_zone").to_string())]
-    tld_dns_zone: Option<String>,
-
     #[arg(
         long,
         env = "ET_PRIVATE_MODE",
@@ -1121,7 +1117,6 @@ impl NetworkOptions {
         f.disable_kcp_input = self.disable_kcp_input.unwrap_or(f.disable_kcp_input);
         f.enable_quic_proxy = self.enable_quic_proxy.unwrap_or(f.enable_quic_proxy);
         f.disable_quic_input = self.disable_quic_input.unwrap_or(f.disable_quic_input);
-        f.accept_dns = self.accept_dns.unwrap_or(f.accept_dns);
         f.private_mode = self.private_mode.unwrap_or(f.private_mode);
         f.foreign_relay_bps_limit = self
             .foreign_relay_bps_limit
@@ -1142,10 +1137,9 @@ impl NetworkOptions {
             .disable_sym_hole_punching
             .unwrap_or(f.disable_sym_hole_punching);
         f.disable_upnp = self.disable_upnp.unwrap_or(f.disable_upnp);
-        // Configure tld_dns_zone: use provided value if set
-        if let Some(tld_dns_zone) = &self.tld_dns_zone {
-            f.tld_dns_zone = tld_dns_zone.clone();
-        }
+        f.enable_udp_broadcast_relay = self
+            .enable_udp_broadcast_relay
+            .unwrap_or(f.enable_udp_broadcast_relay);
         cfg.set_flags(f);
 
         if !self.exit_nodes.is_empty() {
@@ -1336,7 +1330,10 @@ async fn run_main(cli: Cli) -> anyhow::Result<()> {
     let _web_client = if let Some(config_server_url_s) = cli.config_server.as_ref() {
         let wc = web_client::run_web_client(
             config_server_url_s,
-            cli.machine_id.clone(),
+            crate::common::MachineIdOptions {
+                explicit_machine_id: cli.machine_id.clone(),
+                state_dir: None,
+            },
             cli.network_options.hostname.clone(),
             cli.network_options.secure_mode.unwrap_or(false),
             manager.clone(),

easytier/src/dns/config/dns.rs

@@ -0,0 +1,55 @@
+use crate::common::config::ConfigBase;
+use crate::dns::config::policy::DnsPolicyConfig;
+use crate::dns::config::zone::ZoneConfig;
+use crate::dns::config::{DNS_DEFAULT_ADDRESSES, DNS_DEFAULT_DOMAIN};
+use crate::dns::utils::addr::NameServerAddrGroup;
+use crate::proto::dns::GetExportConfigResponse;
+use hickory_proto::rr::LowerName;
+use optionize::{Optionizable, optionized};
+use serde::{Deserialize, Serialize};
+use std::collections::HashMap;
+
+#[optionized]
+#[optionize(name = "DnsConfigRaw")]
+#[derive(Debug, Clone, Default, PartialEq, Deserialize, Serialize)]
+pub struct DnsConfigParsed {
+    pub disabled: bool,
+    #[serde(rename = "zone")]
+    pub zones: Vec<ZoneConfig>,
+    #[optionize(flatten)]
+    #[serde(flatten)]
+    pub policies: HashMap<LowerName, DnsPolicyConfig>,
+    #[optionize(flatten)]
+    pub name: Option<LowerName>,
+    pub domain: LowerName,
+    pub addresses: NameServerAddrGroup,
+    pub listeners: NameServerAddrGroup,
+}
+
+pub type DnsConfig = ConfigBase<DnsConfigRaw, DnsConfigParsed, ()>;
+
+impl From<DnsConfigRaw> for DnsConfig {
+    fn from(raw: DnsConfigRaw) -> Self {
+        let mut parsed = DnsConfigParsed {
+            domain: DNS_DEFAULT_DOMAIN.clone(),
+            addresses: DNS_DEFAULT_ADDRESSES.clone(),
+            ..Default::default()
+        };
+        parsed.load(raw.clone());
+        Self::new(parsed, raw, ())
+    }
+}
+
+#[auto_impl::auto_impl(Box, &)]
+pub trait DnsConfigLoaderExt {
+    fn get_dns(&self) -> DnsConfig;
+    fn set_dns(&self, dns: DnsConfig);
+}
+
+pub type DnsExportConfig = GetExportConfigResponse;
+
+pub trait DnsGlobalCtxExt {
+    fn dns_self_zone(&self) -> ZoneConfig;
+    fn dns_export_config(&self) -> DnsExportConfig;
+    fn dns_iter_zones(&self) -> impl Iterator<Item = ZoneConfig>;
+}

easytier/src/dns/config/mod.rs

@@ -0,0 +1,29 @@
+use crate::dns::utils::addr::NameServerAddrGroup;
+use hickory_proto::rr::LowerName;
+use std::net::IpAddr;
+use std::str::FromStr;
+use std::sync::LazyLock;
+use std::time::Duration;
+use url::Url;
+
+mod dns;
+pub use dns::*;
+mod policy;
+pub mod zone;
+
+pub static DNS_DEFAULT_DOMAIN: LazyLock<LowerName> =
+    LazyLock::new(|| LowerName::from_str("et.net.").unwrap());
+pub static DNS_DEFAULT_ADDRESSES: LazyLock<NameServerAddrGroup> =
+    LazyLock::new(|| IpAddr::from_str("100.100.100.101").unwrap().into());
+
+pub static DNS_SERVER_RPC_ADDR: LazyLock<Url> =
+    LazyLock::new(|| Url::parse("tcp://127.0.0.1:49813").unwrap());
+
+pub const DNS_NODE_TTI: Duration = Duration::from_secs(5);
+
+pub const DNS_NODE_HEARTBEAT_INTERVAL: Duration = Duration::from_secs(2);
+pub const DNS_NODE_RECONCILE_INTERVAL: Duration = Duration::from_secs(10);
+pub const DNS_SERVER_ELECTION_INTERVAL: Duration = Duration::from_secs(5);
+pub const DNS_PEER_TTI: Duration = Duration::from_secs(3);
+pub const DNS_PEER_REFRESH_ATTEMPTS: usize = 3;
+pub const DNS_PEER_REFRESH_BACKOFF: Duration = Duration::from_secs(1);

easytier/src/dns/config/policy.rs

@@ -0,0 +1,46 @@
+use derive_more::{Deref, DerefMut};
+use serde::{Deserialize, Serialize};
+
+#[derive(Debug, Clone, PartialEq, Default, Deserialize, Serialize)]
+#[serde(default)]
+pub struct AclPolicy {
+    pub whitelist: Option<Vec<String>>,
+    pub blacklist: Option<Vec<String>>,
+}
+
+#[derive(Debug, Clone, PartialEq, Default, Deserialize, Serialize, Deref, DerefMut)]
+#[serde(default)]
+pub struct FunctionalityPolicy {
+    #[serde(flatten)]
+    #[deref]
+    #[deref_mut]
+    acl: AclPolicy, // TODO
+    pub disabled: bool,
+}
+
+#[derive(Debug, Clone, PartialEq, Default, Deserialize, Serialize, Deref, DerefMut)]
+#[serde(default)]
+pub struct DnsPolicy<P = FunctionalityPolicy> {
+    #[serde(flatten)]
+    #[deref]
+    #[deref_mut]
+    policy: P,
+    pub recursive: bool, // TODO
+}
+
+pub type ZoneExportPolicy = FunctionalityPolicy;
+pub type DnsExportPolicy = DnsPolicy<ZoneExportPolicy>;
+pub type DnsImportPolicy = DnsPolicy<FunctionalityPolicy>;
+
+#[derive(Debug, Clone, PartialEq, Default, Deserialize, Serialize)]
+#[serde(default)]
+pub struct DnsPolicyConfig {
+    pub import: DnsImportPolicy,
+    pub export: Option<DnsExportPolicy>,
+}
+
+#[derive(Debug, Clone, PartialEq, Default, Deserialize, Serialize)]
+#[serde(default)]
+pub struct ZonePolicyConfig {
+    pub export: Option<DnsExportPolicy>,
+}

easytier/src/dns/config/zone.rs

@@ -0,0 +1,111 @@
+use crate::common::config::ConfigBase;
+use crate::dns::config::policy::{DnsExportPolicy, ZonePolicyConfig};
+use crate::dns::utils::addr::NameServerAddrGroup;
+use crate::dns::zone::Zone;
+use crate::proto::dns::ZoneData;
+use derive_more::From;
+use hickory_proto::op::ResponseCode;
+use hickory_proto::rr::LowerName;
+use maplit::hashset;
+use optionize::{Optionizable, optionized};
+use serde::{Deserialize, Serialize};
+use std::collections::HashSet;
+use std::convert::TryFrom;
+use std::net::{Ipv4Addr, Ipv6Addr};
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Hash, From, Deserialize, Serialize)]
+#[serde(untagged)]
+pub enum Fallthrough {
+    Any,
+    ResponseCode(ResponseCode),
+}
+
+impl From<Fallthrough> for i32 {
+    fn from(value: Fallthrough) -> Self {
+        match value {
+            Fallthrough::ResponseCode(code) => u16::from(code).into(),
+            Fallthrough::Any => -1,
+        }
+    }
+}
+
+impl From<i32> for Fallthrough {
+    fn from(value: i32) -> Self {
+        match u16::try_from(value) {
+            Ok(value) => Self::ResponseCode(value.into()),
+            Err(_) => Self::Any,
+        }
+    }
+}
+
+#[optionized]
+#[optionize(name = "ZoneConfigRaw")]
+#[derive(Debug, Clone, Default, PartialEq, Deserialize, Serialize)]
+pub struct ZoneConfigParsed {
+    #[optionize(flatten)]
+    pub origin: LowerName,
+    pub ttl: u32,
+    pub records: Vec<String>,
+    pub forwarders: NameServerAddrGroup,
+    #[optionize(flatten)]
+    #[serde(flatten)]
+    pub policy: ZonePolicyConfig,
+    pub fallthrough: HashSet<Fallthrough>,
+}
+
+impl From<&ZoneConfigParsed> for ZoneData {
+    fn from(value: &ZoneConfigParsed) -> Self {
+        Self::new(
+            &value.origin,
+            value.ttl,
+            &value.records,
+            value.forwarders.iter().map(Into::into),
+            value.fallthrough.iter().copied(),
+        )
+    }
+}
+
+pub type ZoneConfig = ConfigBase<ZoneConfigRaw, ZoneConfigParsed, ZoneData>;
+
+impl TryFrom<ZoneConfigRaw> for ZoneConfig {
+    type Error = anyhow::Error;
+
+    fn try_from(raw: ZoneConfigRaw) -> Result<Self, Self::Error> {
+        let mut parsed = ZoneConfigParsed {
+            fallthrough: hashset! {Fallthrough::Any},
+            ..Default::default()
+        };
+        parsed.load(raw.clone());
+        let data = (&parsed).into();
+        let _ = Zone::try_from(&data)?; // validation
+        Ok(Self::new(parsed, raw, data))
+    }
+}
+
+impl ZoneConfig {
+    pub fn dedicated(origin: LowerName, ipv4: Option<Ipv4Addr>, ipv6: Vec<Ipv6Addr>) -> Self {
+        let mut records = Vec::new();
+
+        if let Some(ipv4) = ipv4 {
+            records.push(format!("@ IN A {}", ipv4));
+        }
+        for ipv6 in ipv6 {
+            records.push(format!("@ IN AAAA {}", ipv6));
+        }
+
+        let policy = ZonePolicyConfig {
+            export: Some(DnsExportPolicy::default()),
+        };
+
+        let parsed = ZoneConfigParsed {
+            origin,
+            records,
+            policy,
+            ..Default::default()
+        };
+
+        let data = (&parsed).into();
+
+        Self::new(parsed, Default::default(), data)
+    }
+}

easytier/src/dns/mod.rs

@@ -0,0 +1,11 @@
+pub mod config;
+pub mod node;
+mod node_mgr;
+mod peer_mgr;
+pub mod server;
+mod system;
+mod utils;
+mod zone;
+
+#[cfg(test)]
+mod tests;

easytier/src/dns/node.rs

@@ -0,0 +1,560 @@
+use crate::common::global_ctx::{ArcGlobalCtx, GlobalCtxEvent};
+use crate::dns::config::{
+    DNS_NODE_HEARTBEAT_INTERVAL, DNS_NODE_RECONCILE_INTERVAL, DNS_PEER_REFRESH_ATTEMPTS,
+    DNS_PEER_REFRESH_BACKOFF, DNS_SERVER_ELECTION_INTERVAL, DNS_SERVER_RPC_ADDR,
+};
+use crate::dns::peer_mgr::DnsPeerMgr;
+use crate::dns::server::DnsServer;
+#[cfg(feature = "tun")]
+use crate::instance::instance::ArcNicCtx;
+use crate::peers::peer_manager::PeerManager;
+use crate::proto::dns::{DnsNodeMgrRpcClientFactory, HeartbeatRequest};
+use crate::proto::rpc_impl::standalone::{StandAloneClient, StandAloneServer};
+use crate::proto::rpc_types::controller::BaseController;
+use crate::tunnel::tcp::{TcpTunnelConnector, TcpTunnelListener};
+use crate::utils::task::CancellableTask;
+use std::io;
+use std::sync::Arc;
+use tokio::sync::{Notify, broadcast};
+use tokio::task::JoinSet;
+use tokio::time::{MissedTickBehavior, interval};
+use tokio_util::sync::CancellationToken;
+use tracing::instrument;
+use uuid::Uuid;
+
+#[derive(Debug, Clone)]
+struct DnsNodeRuntime {
+    mgr: DnsPeerMgr,
+
+    #[cfg(feature = "tun")]
+    nic_ctx: ArcNicCtx, // TODO: REMOVE THIS
+
+    peer_mgr: Arc<PeerManager>,
+    global_ctx: ArcGlobalCtx,
+
+    elect: Arc<Notify>,
+}
+
+impl DnsNodeRuntime {
+    fn id(&self) -> Uuid {
+        self.global_ctx.get_id()
+    }
+
+    #[instrument(skip_all, name = "DnsNode election loop")]
+    async fn run_election(&self, token: CancellationToken) {
+        let mut election_interval = interval(DNS_SERVER_ELECTION_INTERVAL);
+        election_interval.set_missed_tick_behavior(MissedTickBehavior::Skip);
+
+        loop {
+            tokio::select! {
+                biased;
+                _ = token.cancelled() => {
+                    tracing::info!("DnsNode received shutdown signal, exiting election loop");
+                    break;
+                }
+                _ = self.elect.notified() => {}
+                _ = election_interval.tick() => {}
+            }
+
+            tracing::info!("trying to become DNS server");
+
+            let mut rpc =
+                StandAloneServer::new(TcpTunnelListener::new(DNS_SERVER_RPC_ADDR.clone()));
+
+            if rpc.serve().await.is_err() {
+                // Another node already owns the address — that's fine.
+                tracing::info!(
+                    "failed to bind RPC server, another node might have won the election"
+                );
+                continue;
+            }
+
+            tracing::info!("won DNS server election, starting DnsServer");
+
+            let server = Arc::new(DnsServer::new(
+                self.peer_mgr.clone(),
+                self.global_ctx.clone(),
+                #[cfg(feature = "tun")]
+                self.nic_ctx.clone(),
+            ));
+            server.register(&rpc);
+            server.run(token.child_token()).await;
+
+            tracing::warn!("DnsServer exited, will retry election");
+        }
+    }
+
+    #[instrument(skip_all, name = "DnsNode main loop")]
+    async fn run(&self, token: CancellationToken) {
+        let mut rpc = StandAloneClient::new(TcpTunnelConnector::new(DNS_SERVER_RPC_ADDR.clone()));
+
+        let mut heartbeat = HeartbeatRequest {
+            id: Some(self.id().into()),
+
+            ..Default::default()
+        };
+
+        let mut heartbeat_interval = interval(DNS_NODE_HEARTBEAT_INTERVAL);
+        heartbeat_interval.set_missed_tick_behavior(MissedTickBehavior::Skip);
+
+        let mut reconcile_interval = interval(DNS_NODE_RECONCILE_INTERVAL);
+        reconcile_interval.set_missed_tick_behavior(MissedTickBehavior::Skip);
+
+        let mut subscriber = self.global_ctx.subscribe();
+        let mut tasks = JoinSet::new();
+
+        loop {
+            tokio::select! {
+                biased;
+
+                _ = token.cancelled() => {
+                    tracing::info!("DnsNode received shutdown signal, exiting main loop");
+                    break;
+                }
+
+                _ = heartbeat_interval.tick() => {
+                    if let Err(error) = self.heartbeat(&mut rpc, &mut heartbeat).await {
+                        tracing::error!(?error, "heartbeat failed");
+                        self.elect.notify_one();
+                    }
+                }
+
+                _ = reconcile_interval.tick() => {
+                    let mgr = self.mgr.clone();
+                    tasks.spawn(async move {
+                        mgr.reconcile().await;
+                    });
+                }
+
+                _ = self.mgr.dirty.wait() => {}
+
+                event = subscriber.recv() => {
+                    match event {
+                        Ok(GlobalCtxEvent::PeerInfoUpdated(peer_ids)) => {
+                            for peer_id in peer_ids {
+                                let mgr = self.mgr.clone();
+                                tasks.spawn(async move {
+                                    if let Err(error) = mgr.refresh(peer_id, DNS_PEER_REFRESH_ATTEMPTS, DNS_PEER_REFRESH_BACKOFF).await {
+                                        tracing::error!(?error, ?peer_id, "failed to refresh peer");
+                                    }
+                                });
+                            }
+                            continue;
+                        }
+                        Ok(
+                            GlobalCtxEvent::DhcpIpv4Changed(..)
+                            | GlobalCtxEvent::DhcpIpv4Conflicted(..),
+                        ) => {
+                            tracing::info!(?event, "ip change detected, rebuilding snapshot");
+                        }
+                        Ok(GlobalCtxEvent::ConfigPatched(patch)) => {
+                            // TODO: inspect patch
+                            tracing::info!(?patch, "config change detected, rebuilding snapshot");
+                        }
+                        Err(broadcast::error::RecvError::Lagged(n)) => {
+                            tracing::warn!("event listener lagged, skipped {n} events, rebuilding snapshot");
+                        }
+                        Err(broadcast::error::RecvError::Closed) => {
+                            tracing::info!("event bus closed");
+                            break;
+                        }
+                        _ => continue,
+                    }
+
+                    self.mgr.dirty.mark();
+                }
+
+                result = tasks.join_next(), if !tasks.is_empty() => {
+                    if let Some(Err(error)) = result {
+                        tracing::error!(?error, "refresh task panicked");
+                    }
+                }
+            }
+        }
+    }
+
+    async fn heartbeat(
+        &self,
+        rpc: &mut StandAloneClient<TcpTunnelConnector>,
+        heartbeat: &mut HeartbeatRequest,
+    ) -> anyhow::Result<()> {
+        let request = if heartbeat.snapshot.is_none() || self.mgr.dirty.reset() {
+            heartbeat.update(self.mgr.snapshot());
+            heartbeat.clone()
+        } else {
+            let snapshot = heartbeat.snapshot.take();
+            let request = heartbeat.clone();
+            heartbeat.snapshot = snapshot;
+            request
+        };
+
+        let client = rpc
+            .scoped_client::<DnsNodeMgrRpcClientFactory<BaseController>>("".to_string())
+            .await?;
+
+        let response = client.heartbeat(BaseController::default(), request).await?;
+        if response.resync {
+            tracing::trace!("resync requested by server, sending full snapshot");
+            client
+                .heartbeat(BaseController::default(), heartbeat.clone())
+                .await?;
+        }
+
+        Ok(())
+    }
+}
+
+#[derive(Debug)]
+pub struct DnsNode {
+    runtime: DnsNodeRuntime,
+    task: Option<CancellableTask<()>>,
+}
+
+impl DnsNode {
+    pub fn new(
+        peer_mgr: Arc<PeerManager>,
+        global_ctx: ArcGlobalCtx,
+        #[cfg(feature = "tun")] nic_ctx: ArcNicCtx, // TODO: REMOVE THIS
+    ) -> Self {
+        let runtime = DnsNodeRuntime {
+            mgr: DnsPeerMgr::new(peer_mgr.clone(), global_ctx.clone()),
+            #[cfg(feature = "tun")]
+            nic_ctx,
+            peer_mgr,
+            global_ctx,
+            elect: Default::default(),
+        };
+
+        Self {
+            runtime,
+            task: None,
+        }
+    }
+
+    pub fn start(&mut self) {
+        let runtime = self.runtime.clone();
+        self.task
+            .replace(CancellableTask::spawn(|token| async move {
+                runtime.elect.notify_one();
+                tokio::join!(runtime.run_election(token.clone()), runtime.run(token));
+            }));
+        self.runtime.mgr.register();
+    }
+
+    pub async fn stop(&mut self) -> io::Result<()> {
+        self.runtime.mgr.unregister();
+        let Some(task) = self.task.take() else {
+            return Ok(());
+        };
+        task.stop(None).await
+    }
+}
+
+impl Drop for DnsNode {
+    fn drop(&mut self) {
+        self.runtime.mgr.unregister();
+    }
+}
+
+#[cfg(all(test, feature = "tun"))]
+mod tests {
+    use super::*;
+    use crate::common::global_ctx::GlobalCtxEvent;
+    use crate::peers::tests::create_mock_peer_manager;
+    use crate::proto::api::config::InstanceConfigPatch;
+    use crate::proto::dns::{DnsNodeMgrRpc, DnsNodeMgrRpcServer, HeartbeatResponse};
+    use crate::proto::rpc_impl::standalone::StandAloneServer;
+    use crate::proto::rpc_types;
+    use crate::tunnel::common::tests::wait_for_condition;
+    use std::sync::atomic::{AtomicBool, Ordering};
+    use std::time::Duration;
+    use tokio::sync::Mutex;
+    use tokio::time::sleep;
+    use url::Url;
+
+    #[derive(Debug)]
+    struct RecordingDnsNodeMgr {
+        requests: Mutex<Vec<HeartbeatRequest>>,
+        resync_on_first: AtomicBool,
+    }
+
+    impl RecordingDnsNodeMgr {
+        fn new(resync_on_first: bool) -> Self {
+            Self {
+                requests: Mutex::new(Vec::new()),
+                resync_on_first: AtomicBool::new(resync_on_first),
+            }
+        }
+
+        async fn recorded_requests(&self) -> Vec<HeartbeatRequest> {
+            self.requests.lock().await.clone()
+        }
+    }
+
+    #[async_trait::async_trait]
+    impl DnsNodeMgrRpc for RecordingDnsNodeMgr {
+        type Controller = BaseController;
+
+        async fn heartbeat(
+            &self,
+            _: Self::Controller,
+            input: HeartbeatRequest,
+        ) -> rpc_types::error::Result<HeartbeatResponse> {
+            let mut requests = self.requests.lock().await;
+            requests.push(input);
+            let is_first = requests.len() == 1;
+            let resync = is_first && self.resync_on_first.load(Ordering::Relaxed);
+            if is_first {
+                self.resync_on_first.store(false, Ordering::Relaxed);
+            }
+            Ok(HeartbeatResponse { resync })
+        }
+    }
+
+    async fn build_test_runtime() -> DnsNodeRuntime {
+        let peer_mgr = create_mock_peer_manager().await;
+        let global_ctx = peer_mgr.get_global_ctx();
+        let nic_ctx: ArcNicCtx = Arc::new(Mutex::new(None));
+        DnsNodeRuntime {
+            mgr: DnsPeerMgr::new(peer_mgr.clone(), global_ctx.clone()),
+            nic_ctx,
+            peer_mgr,
+            global_ctx,
+            elect: Default::default(),
+        }
+    }
+
+    async fn start_recording_rpc_server(
+        rpc_addr: Url,
+        resync_on_first: bool,
+    ) -> anyhow::Result<(
+        Arc<RecordingDnsNodeMgr>,
+        StandAloneServer<TcpTunnelListener>,
+    )> {
+        let mgr = Arc::new(RecordingDnsNodeMgr::new(resync_on_first));
+        let mut server = StandAloneServer::new(TcpTunnelListener::new(rpc_addr));
+        server
+            .registry()
+            .register(DnsNodeMgrRpcServer::new_arc(mgr.clone()), "");
+        server.serve().await?;
+        sleep(Duration::from_millis(50)).await;
+        Ok((mgr, server))
+    }
+
+    async fn occupy_dns_rpc_addr(rpc_addr: Url) -> StandAloneServer<TcpTunnelListener> {
+        let mut server = StandAloneServer::new(TcpTunnelListener::new(rpc_addr));
+        server.serve().await.unwrap();
+        server
+    }
+
+    #[tokio::test]
+    async fn heartbeat_first_send_includes_snapshot() {
+        let rpc_addr = Url::parse(&format!("tcp://127.0.0.1:{}", 49851)).unwrap();
+        let (_mgr, server) = start_recording_rpc_server(rpc_addr.clone(), false)
+            .await
+            .unwrap();
+        let node = build_test_runtime().await;
+
+        let mut rpc = StandAloneClient::new(TcpTunnelConnector::new(rpc_addr));
+        let mut heartbeat = HeartbeatRequest {
+            id: Some(node.id().into()),
+            ..Default::default()
+        };
+
+        node.heartbeat(&mut rpc, &mut heartbeat).await.unwrap();
+
+        drop(server);
+        sleep(Duration::from_millis(50)).await;
+
+        assert!(heartbeat.snapshot.is_some());
+        assert!(!heartbeat.digest.is_empty());
+    }
+
+    #[tokio::test]
+    async fn heartbeat_clean_send_digest_only() {
+        let rpc_addr = Url::parse(&format!("tcp://127.0.0.1:{}", 49852)).unwrap();
+        let (mgr, server) = start_recording_rpc_server(rpc_addr.clone(), false)
+            .await
+            .unwrap();
+        let node = build_test_runtime().await;
+
+        let mut rpc = StandAloneClient::new(TcpTunnelConnector::new(rpc_addr));
+        let mut heartbeat = HeartbeatRequest {
+            id: Some(node.id().into()),
+            ..Default::default()
+        };
+
+        node.heartbeat(&mut rpc, &mut heartbeat).await.unwrap();
+        let _ = node.mgr.dirty.reset();
+        node.heartbeat(&mut rpc, &mut heartbeat).await.unwrap();
+
+        let requests = mgr.recorded_requests().await;
+        drop(server);
+        sleep(Duration::from_millis(50)).await;
+
+        assert_eq!(requests.len(), 2);
+        assert!(requests[0].snapshot.is_some());
+        assert!(requests[1].snapshot.is_none());
+        assert_eq!(requests[0].digest, requests[1].digest);
+    }
+
+    #[tokio::test]
+    async fn heartbeat_dirty_forces_full_snapshot() {
+        let rpc_addr = Url::parse(&format!("tcp://127.0.0.1:{}", 49853)).unwrap();
+        let (mgr, server) = start_recording_rpc_server(rpc_addr.clone(), false)
+            .await
+            .unwrap();
+        let node = build_test_runtime().await;
+
+        let mut rpc = StandAloneClient::new(TcpTunnelConnector::new(rpc_addr));
+        let mut heartbeat = HeartbeatRequest {
+            id: Some(node.id().into()),
+            ..Default::default()
+        };
+
+        node.heartbeat(&mut rpc, &mut heartbeat).await.unwrap();
+        node.mgr.dirty.mark();
+        node.heartbeat(&mut rpc, &mut heartbeat).await.unwrap();
+
+        let requests = mgr.recorded_requests().await;
+        drop(server);
+        sleep(Duration::from_millis(50)).await;
+
+        assert_eq!(requests.len(), 2);
+        assert!(requests[0].snapshot.is_some());
+        assert!(requests[1].snapshot.is_some());
+    }
+
+    #[tokio::test]
+    async fn heartbeat_resync_triggers_second_send() {
+        let rpc_addr = Url::parse(&format!("tcp://127.0.0.1:{}", 49854)).unwrap();
+        let (mgr, server) = start_recording_rpc_server(rpc_addr.clone(), true)
+            .await
+            .unwrap();
+        let node = build_test_runtime().await;
+
+        let mut rpc = StandAloneClient::new(TcpTunnelConnector::new(rpc_addr));
+        let mut heartbeat = HeartbeatRequest {
+            id: Some(node.id().into()),
+            ..Default::default()
+        };
+
+        node.heartbeat(&mut rpc, &mut heartbeat).await.unwrap();
+
+        let requests = mgr.recorded_requests().await;
+        drop(server);
+        sleep(Duration::from_millis(50)).await;
+
+        assert_eq!(requests.len(), 2);
+        assert!(requests[0].snapshot.is_some());
+        assert!(requests[1].snapshot.is_some());
+    }
+
+    #[tokio::test]
+    #[serial_test::serial(dns_node_rpc_addr)]
+    async fn run_marks_dirty_on_dhcp_event() {
+        let node = build_test_runtime().await;
+
+        let _ = node.mgr.dirty.reset();
+        assert!(!node.mgr.dirty.peek());
+
+        let token = CancellationToken::new();
+        let handle = tokio::spawn({
+            let node = node.clone();
+            let token = token.clone();
+            async move { node.run(token).await }
+        });
+
+        sleep(Duration::from_millis(50)).await;
+        node.global_ctx
+            .issue_event(GlobalCtxEvent::DhcpIpv4Changed(None, None));
+
+        wait_for_condition(async || node.mgr.dirty.peek(), Duration::from_secs(2)).await;
+
+        token.cancel();
+        tokio::time::timeout(Duration::from_secs(2), handle)
+            .await
+            .unwrap()
+            .unwrap();
+    }
+
+    #[tokio::test]
+    async fn run_marks_dirty_on_config_patched_event() {
+        let node = build_test_runtime().await;
+
+        let _ = node.mgr.dirty.reset();
+        assert!(!node.mgr.dirty.peek());
+
+        let token = CancellationToken::new();
+        let handle = tokio::spawn({
+            let node = node.clone();
+            let token = token.clone();
+            async move { node.run(token).await }
+        });
+
+        sleep(Duration::from_millis(50)).await;
+        node.global_ctx
+            .issue_event(GlobalCtxEvent::ConfigPatched(InstanceConfigPatch::default()));
+
+        wait_for_condition(async || node.mgr.dirty.peek(), Duration::from_secs(2)).await;
+
+        token.cancel();
+        tokio::time::timeout(Duration::from_secs(2), handle)
+            .await
+            .unwrap()
+            .unwrap();
+    }
+
+    #[tokio::test]
+    async fn run_peer_info_updated_non_self_does_not_mark_dirty() {
+        let node = build_test_runtime().await;
+
+        let _ = node.mgr.dirty.reset();
+        assert!(!node.mgr.dirty.peek());
+
+        let token = CancellationToken::new();
+        let handle = tokio::spawn({
+            let node = node.clone();
+            let token = token.clone();
+            async move { node.run(token).await }
+        });
+
+        sleep(Duration::from_millis(50)).await;
+        node.global_ctx
+            .issue_event(GlobalCtxEvent::PeerInfoUpdated(vec![u32::MAX]));
+        sleep(Duration::from_millis(200)).await;
+
+        assert!(!node.mgr.dirty.peek());
+
+        token.cancel();
+        tokio::time::timeout(Duration::from_secs(2), handle)
+            .await
+            .unwrap()
+            .unwrap();
+    }
+
+    #[tokio::test]
+    async fn run_heartbeat_error_notifies_election() {
+        let node = build_test_runtime().await;
+
+        let _ = node.mgr.dirty.reset();
+
+        let token = CancellationToken::new();
+        let notified = node.elect.notified();
+        let handle = tokio::spawn({
+            let node = node.clone();
+            let token = token.clone();
+            async move { node.run(token).await }
+        });
+
+        tokio::time::timeout(2 * DNS_NODE_HEARTBEAT_INTERVAL, notified)
+            .await
+            .expect("heartbeat failure should notify election");
+
+        token.cancel();
+        tokio::time::timeout(Duration::from_secs(5), handle)
+            .await
+            .unwrap()
+            .unwrap();
+    }
+}

easytier/src/dns/node_mgr.rs

@@ -0,0 +1,561 @@
+use crate::dns::config::DNS_NODE_TTI;
+use crate::dns::utils::addr::NameServerAddr;
+use crate::dns::zone::{Zone, ZoneGroup};
+use crate::proto::dns::DnsNodeMgrRpc;
+use crate::proto::dns::{DnsSnapshot, HeartbeatRequest, HeartbeatResponse};
+use crate::proto::rpc_types;
+use crate::proto::rpc_types::controller::BaseController;
+use crate::proto::utils::TransientDigest;
+use crate::utils::dirty::DirtyFlag;
+use anyhow::Error;
+use hickory_server::zone_handler::Catalog;
+use itertools::Itertools;
+use moka::future::Cache;
+use std::collections::HashSet;
+use uuid::Uuid;
+
+#[derive(Debug, Clone, Default)]
+struct DnsNodeInfo {
+    digest: [u8; 32],
+    zones: ZoneGroup,
+    addresses: HashSet<NameServerAddr>,
+    listeners: HashSet<NameServerAddr>,
+}
+
+impl TryFrom<&DnsSnapshot> for DnsNodeInfo {
+    type Error = Error;
+
+    fn try_from(value: &DnsSnapshot) -> Result<Self, Self::Error> {
+        Ok(Self {
+            digest: value.digest(),
+            zones: value.zones.as_slice().try_into()?,
+            addresses: value
+                .addresses
+                .iter()
+                .map(TryInto::try_into)
+                .collect::<Result<_, _>>()?,
+            listeners: value
+                .listeners
+                .iter()
+                .map(TryInto::try_into)
+                .collect::<Result<_, _>>()?,
+        })
+    }
+}
+
+#[derive(Debug, Default)]
+pub struct DnsNodeMgrDirtyFlags {
+    pub catalog: DirtyFlag,
+    pub addresses: DirtyFlag,
+    pub listeners: DirtyFlag,
+}
+
+#[derive(Debug)]
+pub struct DnsNodeMgr {
+    nodes: Cache<Uuid, DnsNodeInfo>,
+    pub dirty: DnsNodeMgrDirtyFlags,
+}
+
+impl DnsNodeMgr {
+    pub fn new() -> Self {
+        Self {
+            nodes: Cache::builder().time_to_idle(DNS_NODE_TTI).build(),
+            dirty: Default::default(),
+        }
+    }
+
+    pub fn catalog(&self) -> Catalog {
+        let groups = self.collect_zones().into_groups();
+
+        tracing::trace!("building catalog with zones: {:?}", groups);
+
+        let system = Zone::system().create_forward_zone_handler();
+        groups
+            .into_iter()
+            .fold(Catalog::new(), |mut catalog, (origin, zones)| {
+                catalog.upsert(
+                    origin.clone(),
+                    zones
+                        .iter_zone_handlers()
+                        .chain(system.iter().cloned())
+                        .collect(),
+                );
+                catalog
+            })
+    }
+
+    pub fn collect_zones(&self) -> ZoneGroup {
+        let mut zones = Vec::new();
+        let mut local = HashSet::new();
+
+        for (_, info) in self.nodes.iter() {
+            zones.extend(info.zones);
+            local.extend(info.addresses);
+            local.extend(info.listeners);
+        }
+
+        zones.push(Zone::system());
+
+        for forward in zones.iter_mut().flat_map(|z| &mut z.forward) {
+            forward.name_servers.retain_mut(|ns| {
+                ns.connections
+                    .retain(|c| !local.contains(&(ns.ip, c).into()));
+                !ns.connections.is_empty()
+            });
+        }
+
+        zones.into()
+    }
+
+    pub fn iter_addresses(&self) -> impl Iterator<Item = NameServerAddr> + use<'_> {
+        self.nodes
+            .iter()
+            .flat_map(|(_, info)| info.addresses)
+            .unique()
+    }
+
+    pub fn iter_listeners(&self) -> impl Iterator<Item = NameServerAddr> + use<'_> {
+        self.nodes
+            .iter()
+            .flat_map(|(_, info)| info.listeners)
+            .unique()
+    }
+}
+
+#[async_trait::async_trait]
+impl DnsNodeMgrRpc for DnsNodeMgr {
+    type Controller = BaseController;
+
+    async fn heartbeat(
+        &self,
+        _: BaseController,
+        input: HeartbeatRequest,
+    ) -> rpc_types::error::Result<HeartbeatResponse> {
+        let id = input
+            .id
+            .ok_or(anyhow::anyhow!(
+                "missing id in heartbeat request: {:?}",
+                input
+            ))?
+            .into();
+
+        let resync = if let Some(snapshot) = input.snapshot.as_ref() {
+            let new = DnsNodeInfo::try_from(snapshot)?;
+            let old = self.nodes.get(&id).await.unwrap_or_default();
+            if new.digest != old.digest {
+                self.dirty.catalog.mark();
+                if new.addresses != old.addresses {
+                    self.dirty.addresses.mark();
+                }
+                if new.listeners != old.listeners {
+                    self.dirty.listeners.mark();
+                }
+
+                self.nodes.insert(id, new).await;
+            }
+            false
+        } else {
+            self.nodes
+                .get(&id)
+                .await
+                .is_none_or(|info| input.digest != info.digest)
+        };
+
+        Ok(HeartbeatResponse { resync })
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::dns::tests::{
+        dns_snapshot_with as snapshot_with, heartbeat_with_snapshot, new_request,
+        zone_data_a_with_forwarders as valid_zone_data,
+    };
+    use crate::dns::utils::response::ResponseHandle;
+    use hickory_proto::op::{Message, ResponseCode};
+    use hickory_proto::rr::{RData, RecordType};
+    use std::net::Ipv4Addr;
+    use tokio::time::{Duration, sleep};
+
+    fn heartbeat_digest_only(id: Uuid, digest: Vec<u8>) -> HeartbeatRequest {
+        HeartbeatRequest {
+            id: Some(id.into()),
+            digest,
+            snapshot: None,
+        }
+    }
+
+    fn reset_all_dirty(mgr: &DnsNodeMgr) {
+        let _ = mgr.dirty.catalog.reset();
+        let _ = mgr.dirty.addresses.reset();
+        let _ = mgr.dirty.listeners.reset();
+    }
+
+    async fn send_heartbeat(mgr: &DnsNodeMgr, input: HeartbeatRequest) -> HeartbeatResponse {
+        DnsNodeMgrRpc::heartbeat(mgr, BaseController::default(), input)
+            .await
+            .expect("heartbeat should succeed")
+    }
+
+    fn ns(s: &str) -> NameServerAddr {
+        s.parse().expect("invalid nameserver")
+    }
+
+    async fn lookup_a_record(mgr: &DnsNodeMgr, name: &str) -> anyhow::Result<Message> {
+        let request = new_request(name, RecordType::A)?;
+        let response = ResponseHandle::new(512);
+        let info = mgr
+            .catalog()
+            .lookup(&request, None, 0, response.clone())
+            .await;
+
+        assert_eq!(info.response_code, ResponseCode::NoError);
+
+        let response = response.into_inner().expect("response should exist");
+        Message::from_vec(&response).map_err(Into::into)
+    }
+
+    #[tokio::test]
+    async fn catalog_lookup_returns_record_after_snapshot_heartbeat() -> anyhow::Result<()> {
+        let mgr = DnsNodeMgr::new();
+        let id = Uuid::new_v4();
+        let snapshot = snapshot_with(
+            vec![valid_zone_data("catalog.test", "10.20.30.40", vec![])],
+            vec![],
+            vec![],
+        );
+
+        let _ = send_heartbeat(&mgr, heartbeat_with_snapshot(id, snapshot)).await;
+
+        let message = lookup_a_record(&mgr, "catalog.test.").await?;
+        assert!(message.answers.iter().any(|record| {
+            matches!(
+                record.data,
+                RData::A(addr) if *addr == Ipv4Addr::new(10, 20, 30, 40)
+            )
+        }));
+
+        Ok(())
+    }
+
+    #[tokio::test]
+    async fn catalog_lookup_aggregates_records_from_multiple_nodes() -> anyhow::Result<()> {
+        let mgr = DnsNodeMgr::new();
+
+        let snap_a = snapshot_with(
+            vec![valid_zone_data("node-a.test", "10.11.12.13", vec![])],
+            vec!["udp://10.0.1.1:53"],
+            vec![],
+        );
+        let snap_b = snapshot_with(
+            vec![valid_zone_data("node-b.test", "10.21.22.23", vec![])],
+            vec!["udp://10.0.2.1:53"],
+            vec![],
+        );
+
+        let _ = send_heartbeat(&mgr, heartbeat_with_snapshot(Uuid::new_v4(), snap_a)).await;
+        let _ = send_heartbeat(&mgr, heartbeat_with_snapshot(Uuid::new_v4(), snap_b)).await;
+
+        let message_a = lookup_a_record(&mgr, "node-a.test.").await?;
+        let message_b = lookup_a_record(&mgr, "node-b.test.").await?;
+
+        assert!(message_a.answers.iter().any(|record| {
+            matches!(
+                record.data,
+                RData::A(addr) if *addr == Ipv4Addr::new(10, 11, 12, 13)
+            )
+        }));
+        assert!(message_b.answers.iter().any(|record| {
+            matches!(
+                record.data,
+                RData::A(addr) if *addr == Ipv4Addr::new(10, 21, 22, 23)
+            )
+        }));
+
+        Ok(())
+    }
+
+    #[tokio::test]
+    async fn heartbeat_digest_only_resync_behavior() {
+        let mgr = DnsNodeMgr::new();
+        let id = Uuid::new_v4();
+
+        let first = send_heartbeat(&mgr, heartbeat_digest_only(id, vec![1, 2, 3])).await;
+        assert!(first.resync);
+
+        let snapshot = snapshot_with(
+            vec![valid_zone_data("resync.test", "10.0.0.10", vec![])],
+            vec!["udp://10.0.0.1:53"],
+            vec!["udp://10.0.0.2:53"],
+        );
+        let digest = snapshot.digest();
+        let full = send_heartbeat(&mgr, heartbeat_with_snapshot(id, snapshot)).await;
+        assert!(!full.resync);
+
+        let same = send_heartbeat(&mgr, heartbeat_digest_only(id, digest.into())).await;
+        assert!(!same.resync);
+
+        let different = send_heartbeat(&mgr, heartbeat_digest_only(id, vec![9, 9, 9])).await;
+        assert!(different.resync);
+    }
+
+    #[tokio::test]
+    async fn heartbeat_with_snapshot_marks_dirty_flags_by_field_changes() {
+        let mgr = DnsNodeMgr::new();
+        let id = Uuid::new_v4();
+
+        reset_all_dirty(&mgr);
+
+        let first = snapshot_with(
+            vec![valid_zone_data("dirty.test", "10.0.0.1", vec![])],
+            vec!["udp://10.10.10.1:53"],
+            vec!["udp://10.10.10.2:53"],
+        );
+        let _ = send_heartbeat(&mgr, heartbeat_with_snapshot(id, first)).await;
+        assert!(mgr.dirty.catalog.peek());
+        assert!(mgr.dirty.addresses.peek());
+        assert!(mgr.dirty.listeners.peek());
+
+        reset_all_dirty(&mgr);
+
+        let record_changed = snapshot_with(
+            vec![valid_zone_data("dirty.test", "10.0.0.2", vec![])],
+            vec!["udp://10.10.10.1:53"],
+            vec!["udp://10.10.10.2:53"],
+        );
+        let _ = send_heartbeat(&mgr, heartbeat_with_snapshot(id, record_changed)).await;
+        assert!(mgr.dirty.catalog.peek());
+        assert!(!mgr.dirty.addresses.peek());
+        assert!(!mgr.dirty.listeners.peek());
+
+        reset_all_dirty(&mgr);
+
+        let addr_listener_changed = snapshot_with(
+            vec![valid_zone_data("dirty.test", "10.0.0.2", vec![])],
+            vec!["udp://10.10.10.10:53"],
+            vec!["udp://10.10.10.20:53"],
+        );
+        let _ = send_heartbeat(&mgr, heartbeat_with_snapshot(id, addr_listener_changed)).await;
+        assert!(mgr.dirty.catalog.peek());
+        assert!(mgr.dirty.addresses.peek());
+        assert!(mgr.dirty.listeners.peek());
+    }
+
+    #[tokio::test]
+    async fn heartbeat_with_same_snapshot_digest_is_noop_for_dirty() {
+        let mgr = DnsNodeMgr::new();
+        let id = Uuid::new_v4();
+
+        let snapshot = snapshot_with(
+            vec![valid_zone_data("stable.test", "10.30.40.50", vec![])],
+            vec!["udp://10.3.0.1:53"],
+            vec!["udp://10.3.0.2:53"],
+        );
+        let _ = send_heartbeat(&mgr, heartbeat_with_snapshot(id, snapshot.clone())).await;
+
+        reset_all_dirty(&mgr);
+
+        let _ = send_heartbeat(&mgr, heartbeat_with_snapshot(id, snapshot)).await;
+        assert!(!mgr.dirty.catalog.peek());
+        assert!(!mgr.dirty.addresses.peek());
+        assert!(!mgr.dirty.listeners.peek());
+    }
+
+    #[tokio::test]
+    async fn heartbeat_missing_id_returns_error() {
+        let mgr = DnsNodeMgr::new();
+        let err =
+            DnsNodeMgrRpc::heartbeat(&mgr, BaseController::default(), HeartbeatRequest::default())
+                .await
+                .expect_err("missing id should error");
+        assert!(err.to_string().contains("missing id"));
+    }
+
+    #[tokio::test]
+    async fn iter_addresses_and_listeners_deduplicate_across_multiple_nodes() -> anyhow::Result<()>
+    {
+        let mgr = DnsNodeMgr::new();
+        let zone_a = Zone::try_from(&valid_zone_data("iter-a.test", "10.1.1.1", vec![]))?;
+        let zone_b = Zone::try_from(&valid_zone_data("iter-b.test", "10.1.1.2", vec![]))?;
+
+        mgr.nodes
+            .insert(
+                Uuid::new_v4(),
+                DnsNodeInfo {
+                    digest: [1; 32],
+                    zones: vec![zone_a].into(),
+                    addresses: [ns("udp://10.100.0.1:53"), ns("udp://10.100.0.2:53")]
+                        .into_iter()
+                        .collect(),
+                    listeners: [ns("udp://10.200.0.1:53")].into_iter().collect(),
+                },
+            )
+            .await;
+        mgr.nodes
+            .insert(
+                Uuid::new_v4(),
+                DnsNodeInfo {
+                    digest: [2; 32],
+                    zones: vec![zone_b].into(),
+                    addresses: [ns("udp://10.100.0.2:53"), ns("udp://10.100.0.3:53")]
+                        .into_iter()
+                        .collect(),
+                    listeners: [ns("udp://10.200.0.1:53"), ns("udp://10.200.0.2:53")]
+                        .into_iter()
+                        .collect(),
+                },
+            )
+            .await;
+
+        let addresses: HashSet<_> = mgr.iter_addresses().collect();
+        let listeners: HashSet<_> = mgr.iter_listeners().collect();
+
+        assert_eq!(addresses.len(), 3);
+        assert!(addresses.contains(&ns("udp://10.100.0.1:53")));
+        assert!(addresses.contains(&ns("udp://10.100.0.2:53")));
+        assert!(addresses.contains(&ns("udp://10.100.0.3:53")));
+
+        assert_eq!(listeners.len(), 2);
+        assert!(listeners.contains(&ns("udp://10.200.0.1:53")));
+        assert!(listeners.contains(&ns("udp://10.200.0.2:53")));
+
+        Ok(())
+    }
+
+    #[tokio::test]
+    async fn collect_zones_filters_out_local_forwarders() -> anyhow::Result<()> {
+        let mgr = DnsNodeMgr::new();
+        let zone = Zone::try_from(&valid_zone_data(
+            "filter-loop.test",
+            "10.2.3.4",
+            vec![
+                "udp://10.0.0.10:53",
+                "tcp://10.0.0.11:53",
+                "udp://1.1.1.1:53",
+            ],
+        ))?;
+
+        mgr.nodes
+            .insert(
+                Uuid::new_v4(),
+                DnsNodeInfo {
+                    digest: [1; 32],
+                    zones: vec![zone].into(),
+                    addresses: [ns("udp://10.0.0.10:53")].into_iter().collect(),
+                    listeners: [ns("tcp://10.0.0.11:53")].into_iter().collect(),
+                },
+            )
+            .await;
+
+        let zones: Vec<_> = mgr.collect_zones().into_iter().map(Into::into).collect();
+        let loop_zone = zones
+            .into_iter()
+            .find(|z: &crate::proto::dns::ZoneData| z.content.contains("$ORIGIN filter-loop.test"))
+            .expect("test zone should exist");
+
+        let forwarders: HashSet<NameServerAddr> = loop_zone
+            .forwarders
+            .iter()
+            .map(|u| NameServerAddr::try_from(u).expect("forwarder should be valid"))
+            .collect();
+
+        assert_eq!(forwarders.len(), 1);
+        assert!(forwarders.contains(&ns("udp://1.1.1.1:53")));
+
+        Ok(())
+    }
+
+    #[tokio::test]
+    async fn collect_zones_filters_cross_node_local_forwarders() -> anyhow::Result<()> {
+        let mgr = DnsNodeMgr::new();
+
+        let node_a = snapshot_with(
+            vec![valid_zone_data(
+                "cross-node-filter.test",
+                "10.8.8.8",
+                vec![
+                    "udp://10.50.0.1:53",
+                    "udp://10.50.0.2:53",
+                    "udp://8.8.8.8:53",
+                ],
+            )],
+            vec!["udp://10.50.0.1:53"],
+            vec![],
+        );
+        let node_b = snapshot_with(
+            vec![valid_zone_data(
+                "cross-node-helper.test",
+                "10.9.9.9",
+                vec![],
+            )],
+            vec![],
+            vec!["udp://10.50.0.2:53"],
+        );
+
+        let _ = send_heartbeat(&mgr, heartbeat_with_snapshot(Uuid::new_v4(), node_a)).await;
+        let _ = send_heartbeat(&mgr, heartbeat_with_snapshot(Uuid::new_v4(), node_b)).await;
+
+        let zones: Vec<_> = mgr.collect_zones().into_iter().map(Into::into).collect();
+        let zone = zones
+            .into_iter()
+            .find(|z: &crate::proto::dns::ZoneData| {
+                z.content.contains("$ORIGIN cross-node-filter.test")
+            })
+            .expect("test zone should exist");
+
+        let forwarders: HashSet<NameServerAddr> = zone
+            .forwarders
+            .iter()
+            .map(|u| NameServerAddr::try_from(u).expect("forwarder should be valid"))
+            .collect();
+
+        assert_eq!(forwarders.len(), 1);
+        assert!(forwarders.contains(&ns("udp://8.8.8.8:53")));
+
+        Ok(())
+    }
+
+    #[tokio::test]
+    async fn heartbeat_digest_resync_is_node_scoped() {
+        let mgr = DnsNodeMgr::new();
+        let node_a = Uuid::new_v4();
+        let node_b = Uuid::new_v4();
+
+        let snap_a = snapshot_with(
+            vec![valid_zone_data("scope-a.test", "10.60.0.1", vec![])],
+            vec!["udp://10.60.0.2:53"],
+            vec![],
+        );
+        let digest_a = snap_a.digest();
+
+        let _ = send_heartbeat(&mgr, heartbeat_with_snapshot(node_a, snap_a)).await;
+
+        let a_same = send_heartbeat(&mgr, heartbeat_digest_only(node_a, digest_a.into())).await;
+        assert!(!a_same.resync);
+
+        let b_unknown = send_heartbeat(&mgr, heartbeat_digest_only(node_b, vec![1, 2, 3])).await;
+        assert!(b_unknown.resync);
+    }
+
+    #[tokio::test]
+    async fn heartbeat_resync_after_node_idle_ttl_expiry() {
+        let mgr = DnsNodeMgr::new();
+        let id = Uuid::new_v4();
+        let snapshot = snapshot_with(
+            vec![valid_zone_data("ttl.test", "10.9.9.9", vec![])],
+            vec!["udp://10.9.0.1:53"],
+            vec![],
+        );
+        let digest = snapshot.digest();
+
+        let _ = send_heartbeat(&mgr, heartbeat_with_snapshot(id, snapshot)).await;
+        let before_expiry = send_heartbeat(&mgr, heartbeat_digest_only(id, digest.to_vec())).await;
+        assert!(!before_expiry.resync);
+
+        sleep(DNS_NODE_TTI + Duration::from_millis(300)).await;
+
+        let after_expiry = send_heartbeat(&mgr, heartbeat_digest_only(id, digest.into())).await;
+        assert!(after_expiry.resync);
+    }
+}

easytier/src/dns/peer_mgr.rs

@@ -0,0 +1,873 @@
+use crate::common::PeerId;
+use crate::common::global_ctx::ArcGlobalCtx;
+use crate::dns::config::zone::ZoneConfig;
+use crate::dns::config::{
+    DNS_PEER_REFRESH_ATTEMPTS, DNS_PEER_REFRESH_BACKOFF, DNS_PEER_TTI, DnsExportConfig,
+    DnsGlobalCtxExt,
+};
+use crate::dns::zone::ZoneGroup;
+use crate::peer_center::instance::PeerCenterPeerManagerTrait;
+use crate::peers::peer_manager::PeerManager;
+use crate::peers::route_trait::Route;
+use crate::proto::dns::{
+    DnsPeerMgrRpc, DnsPeerMgrRpcClientFactory, DnsPeerMgrRpcServer, DnsSnapshot,
+    GetExportConfigRequest, GetExportConfigResponse, ZoneData,
+};
+use crate::proto::rpc_types;
+use crate::proto::rpc_types::controller::BaseController;
+use crate::proto::utils::TransientDigest;
+use crate::utils::dirty::DirtyFlag;
+use anyhow::Context;
+use futures::StreamExt;
+use futures::stream;
+use moka::future::Cache;
+use std::ops::Deref;
+use std::sync::Arc;
+use std::time::Duration;
+use tokio::time::sleep;
+use tracing::instrument;
+
+#[derive(Debug, Clone)]
+struct DnsPeerInfo {
+    digest: [u8; 32],
+    zones: Vec<ZoneData>,
+}
+
+impl TryFrom<DnsExportConfig> for DnsPeerInfo {
+    type Error = anyhow::Error;
+
+    fn try_from(value: DnsExportConfig) -> Result<Self, Self::Error> {
+        let _ = ZoneGroup::try_from(value.zones.as_slice())?;
+        Ok(Self {
+            digest: value.digest(),
+            zones: value.zones,
+        })
+    }
+}
+
+#[derive(Debug)]
+pub struct DnsPeerMgrInner {
+    peers: Cache<PeerId, DnsPeerInfo>,
+    pub dirty: DirtyFlag,
+
+    peer_mgr: Arc<PeerManager>,
+    global_ctx: ArcGlobalCtx,
+}
+
+impl DnsPeerMgrInner {
+    pub fn snapshot(&self) -> DnsSnapshot {
+        let global_ctx = &self.global_ctx;
+
+        let zones = global_ctx
+            .dns_iter_zones()
+            .map(ZoneConfig::into_data)
+            .chain(
+                self.peers
+                    .iter()
+                    .flat_map(|(_, info)| info.zones.into_iter()),
+            )
+            .collect();
+
+        let config = global_ctx.config.get_dns().into_parsed();
+        DnsSnapshot {
+            zones,
+            addresses: config.addresses.into(),
+            listeners: config.listeners.into(),
+        }
+    }
+
+    #[instrument(skip(self), level = "trace", ret)]
+    pub async fn refresh(
+        &self,
+        peer_id: PeerId,
+        mut attempts: usize,
+        mut backoff: Duration,
+    ) -> anyhow::Result<bool> {
+        loop {
+            attempts = attempts.saturating_sub(1);
+            let result = self.try_refresh(peer_id).await;
+            match &result {
+                Ok(_) => {
+                    tracing::trace!(?peer_id, "peer info refreshed");
+                    return result;
+                }
+                Err(_) if attempts == 0 => {
+                    self.peers.invalidate(&peer_id).await;
+                    self.dirty.mark();
+                    tracing::error!(
+                        ?peer_id,
+                        "exhausted all attempts to refresh peer info, invalidating cache"
+                    );
+                    return result;
+                }
+                Err(error) => {
+                    tracing::error!(
+                        ?error,
+                        ?peer_id,
+                        "failed to refresh peer info, retrying in {:?}",
+                        backoff
+                    );
+                    sleep(backoff).await;
+                    backoff *= 2;
+                }
+            }
+        }
+    }
+
+    async fn try_refresh(&self, peer_id: PeerId) -> anyhow::Result<bool> {
+        if peer_id == self.peer_mgr.my_peer_id() {
+            self.dirty.mark();
+            return Ok(true);
+        }
+
+        let Some(route) = self.peer_mgr.get_route().get_peer_info(peer_id).await else {
+            if self.peers.remove(&peer_id).await.is_some() {
+                tracing::debug!(?peer_id, "peer route disappeared, removing from cache");
+                self.dirty.mark();
+            }
+            return Ok(true);
+        };
+
+        if self
+            .peers
+            .get(&peer_id)
+            .await
+            .is_some_and(|info| route.dns == info.digest)
+        {
+            return Ok(false);
+        }
+
+        if !route.dns.is_empty() {
+            let info = self.fetch(peer_id).await.with_context(|| {
+                format!("failed to fetch dns export config from peer {}", peer_id)
+            })?;
+            self.peers.insert(peer_id, info).await;
+        } else {
+            self.peers.invalidate(&peer_id).await;
+        }
+
+        self.dirty.mark();
+        Ok(true)
+    }
+
+    #[instrument(skip(self), level = "trace", ret)]
+    async fn fetch(&self, peer_id: PeerId) -> anyhow::Result<DnsPeerInfo> {
+        self.peer_mgr
+            .get_peer_rpc_mgr()
+            .rpc_client()
+            .scoped_client::<DnsPeerMgrRpcClientFactory<BaseController>>(
+                self.peer_mgr.my_peer_id(),
+                peer_id,
+                self.global_ctx.get_network_name(),
+            )
+            .get_export_config(BaseController::default(), GetExportConfigRequest {})
+            .await
+            .context("rpc call failed")?
+            .try_into()
+    }
+}
+
+#[async_trait::async_trait]
+impl DnsPeerMgrRpc for DnsPeerMgrInner {
+    type Controller = BaseController;
+
+    async fn get_export_config(
+        &self,
+        _: Self::Controller,
+        _: GetExportConfigRequest,
+    ) -> rpc_types::error::Result<GetExportConfigResponse> {
+        Ok(self.global_ctx.dns_export_config())
+    }
+}
+
+#[derive(Debug, Clone)]
+pub struct DnsPeerMgr(Arc<DnsPeerMgrInner>);
+
+impl DnsPeerMgr {
+    pub fn new(peer_mgr: Arc<PeerManager>, global_ctx: ArcGlobalCtx) -> Self {
+        Self(Arc::new(DnsPeerMgrInner {
+            peers: Cache::builder().time_to_idle(DNS_PEER_TTI).build(),
+            dirty: Default::default(),
+            peer_mgr,
+            global_ctx,
+        }))
+    }
+
+    pub fn register(&self) {
+        self.peer_mgr
+            .get_peer_rpc_mgr()
+            .rpc_server()
+            .registry()
+            .register(
+                DnsPeerMgrRpcServer::new_arc(self.0.clone()),
+                &self.global_ctx.get_network_name(),
+            );
+    }
+
+    pub fn unregister(&self) -> Option<()> {
+        self.peer_mgr
+            .get_peer_rpc_mgr()
+            .rpc_server()
+            .registry()
+            .unregister(
+                DnsPeerMgrRpcServer::new_arc(self.0.clone()),
+                &self.global_ctx.get_network_name(),
+            )
+    }
+
+    #[instrument(skip(self), level = "trace")]
+    pub async fn reconcile(&self) {
+        stream::iter(self.peer_mgr.list_routes().await.into_iter())
+            .map(|route| {
+                let peer_id = route.peer_id;
+                let this = self.clone();
+                async move {
+                    if let Err(error) = this
+                        .refresh(peer_id, DNS_PEER_REFRESH_ATTEMPTS, DNS_PEER_REFRESH_BACKOFF)
+                        .await
+                    {
+                        tracing::error!(?error, ?peer_id, "failed to refresh peer info");
+                    }
+                }
+            })
+            .buffer_unordered(32)
+            .collect::<Vec<_>>()
+            .await;
+    }
+}
+
+impl Deref for DnsPeerMgr {
+    type Target = DnsPeerMgrInner;
+
+    fn deref(&self) -> &Self::Target {
+        &self.0
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::common::global_ctx::tests::get_mock_global_ctx;
+    use crate::dns::config::zone::ZoneConfig;
+    use crate::dns::tests::zone_data_a as valid_zone_data;
+    use crate::peers::create_packet_recv_chan;
+    use crate::peers::peer_manager::RouteAlgoType;
+    use crate::peers::tests::{connect_peer_manager, wait_route_appear};
+    use crate::proto::dns::GetExportConfigRequest;
+
+    use std::collections::HashSet;
+    use std::net::Ipv4Addr;
+    use tokio::time::{Duration, sleep};
+
+    async fn create_peer_manager_with_zone(
+        host: &str,
+        origin: &str,
+        record_ip: Ipv4Addr,
+    ) -> Arc<PeerManager> {
+        let ctx = get_mock_global_ctx();
+        let mut dns = ctx.config.get_dns().into_raw();
+        dns.name = Some(host.parse().unwrap());
+        dns.zones
+            .get_or_insert_default()
+            .push(ZoneConfig::dedicated(
+                origin.parse().expect("invalid zone origin"),
+                Some(record_ip),
+                vec![],
+            ));
+        ctx.config.set_dns(dns.into());
+
+        let (s, _r) = create_packet_recv_chan();
+        let peer_mgr = Arc::new(PeerManager::new(RouteAlgoType::Ospf, ctx, s));
+        peer_mgr.run().await.unwrap();
+        peer_mgr
+    }
+
+    #[test]
+    fn dns_peer_info_try_from_valid_config() {
+        let cfg = DnsExportConfig {
+            zones: vec![valid_zone_data("valid.peer.test", "10.0.0.10")],
+        };
+
+        let info = DnsPeerInfo::try_from(cfg).expect("valid export config should pass");
+        assert_eq!(info.zones.len(), 1);
+        assert!(!info.digest.is_empty());
+    }
+
+    #[test]
+    fn dns_peer_info_try_from_invalid_zone_rejected() {
+        let cfg = DnsExportConfig {
+            zones: vec![ZoneData::new(&".".parse().unwrap(), 60, ["?"], [], [])],
+        };
+
+        assert!(DnsPeerInfo::try_from(cfg).is_err());
+    }
+
+    #[tokio::test]
+    async fn snapshot_merges_local_and_cached_peer_zones() {
+        let peer_mgr = create_peer_manager_with_zone(
+            "local-peer",
+            "local-custom.test",
+            Ipv4Addr::new(10, 10, 10, 10),
+        )
+        .await;
+        let global_ctx = peer_mgr.get_global_ctx();
+        let mgr = DnsPeerMgr::new(peer_mgr, global_ctx);
+
+        mgr.peers
+            .insert(
+                999_999,
+                DnsPeerInfo {
+                    digest: [9; 32],
+                    zones: vec![valid_zone_data("peer-cache.test", "10.20.30.40")],
+                },
+            )
+            .await;
+
+        let snapshot = mgr.snapshot();
+        assert!(
+            snapshot
+                .zones
+                .iter()
+                .any(|z| z.content.contains("$ORIGIN peer-cache.test"))
+        );
+        assert!(
+            snapshot
+                .zones
+                .iter()
+                .any(|z| z.content.contains("$ORIGIN local-custom.test"))
+        );
+    }
+
+    #[tokio::test]
+    async fn snapshot_includes_local_addresses_and_listeners() {
+        let peer_mgr = create_peer_manager_with_zone(
+            "local-addr-listener",
+            "local-addr-zone.test",
+            Ipv4Addr::new(10, 10, 11, 11),
+        )
+        .await;
+        let global_ctx = peer_mgr.get_global_ctx();
+        let expected = global_ctx.config.get_dns().into_parsed();
+        let mgr = DnsPeerMgr::new(peer_mgr, global_ctx);
+
+        let snapshot = mgr.snapshot();
+        let mut expected_addresses = expected
+            .addresses
+            .into_iter()
+            .map(|a| a.to_string())
+            .collect::<Vec<_>>();
+        let mut expected_listeners = expected
+            .listeners
+            .into_iter()
+            .map(|a| a.to_string())
+            .collect::<Vec<_>>();
+        let mut got_addresses = snapshot
+            .addresses
+            .into_iter()
+            .map(|a| a.to_string())
+            .collect::<Vec<_>>();
+        let mut got_listeners = snapshot
+            .listeners
+            .into_iter()
+            .map(|a| a.to_string())
+            .collect::<Vec<_>>();
+
+        expected_addresses.sort();
+        expected_listeners.sort();
+        got_addresses.sort();
+        got_listeners.sort();
+
+        assert_eq!(got_addresses, expected_addresses);
+        assert_eq!(got_listeners, expected_listeners);
+    }
+
+    #[tokio::test]
+    async fn snapshot_aggregates_zones_from_multiple_cached_peers() {
+        let peer_mgr = create_peer_manager_with_zone(
+            "local-multi",
+            "local-multi.test",
+            Ipv4Addr::new(10, 10, 12, 1),
+        )
+        .await;
+        let global_ctx = peer_mgr.get_global_ctx();
+        let mgr = DnsPeerMgr::new(peer_mgr, global_ctx);
+
+        mgr.peers
+            .insert(
+                11,
+                DnsPeerInfo {
+                    digest: [11; 32],
+                    zones: vec![valid_zone_data("peer-a.test", "10.20.30.41")],
+                },
+            )
+            .await;
+        mgr.peers
+            .insert(
+                12,
+                DnsPeerInfo {
+                    digest: [12; 32],
+                    zones: vec![valid_zone_data("peer-b.test", "10.20.30.42")],
+                },
+            )
+            .await;
+
+        let snapshot = mgr.snapshot();
+        let contents: HashSet<_> = snapshot.zones.into_iter().map(|z| z.content).collect();
+
+        assert!(contents.iter().any(|z| z.contains("$ORIGIN peer-a.test")));
+        assert!(contents.iter().any(|z| z.contains("$ORIGIN peer-b.test")));
+        assert!(
+            contents
+                .iter()
+                .any(|z| z.contains("$ORIGIN local-multi.test"))
+        );
+    }
+
+    #[tokio::test]
+    async fn snapshot_with_peer_without_zones_keeps_local_snapshot() {
+        let peer_mgr = create_peer_manager_with_zone(
+            "local-empty-peer-zone",
+            "local-empty-zone.test",
+            Ipv4Addr::new(10, 10, 13, 1),
+        )
+        .await;
+        let mgr = DnsPeerMgr::new(peer_mgr, get_mock_global_ctx());
+
+        let before = mgr.snapshot();
+
+        mgr.peers
+            .insert(
+                13,
+                DnsPeerInfo {
+                    digest: [13; 32],
+                    zones: vec![],
+                },
+            )
+            .await;
+
+        let after = mgr.snapshot();
+        assert_eq!(before.zones.len(), after.zones.len());
+        assert_eq!(before.addresses, after.addresses);
+        assert_eq!(before.listeners, after.listeners);
+    }
+
+    #[tokio::test]
+    async fn get_export_config_returns_global_ctx_export() {
+        let peer_mgr = create_peer_manager_with_zone(
+            "export-peer",
+            "exported-zone.test",
+            Ipv4Addr::new(10, 10, 20, 20),
+        )
+        .await;
+        let global_ctx = peer_mgr.get_global_ctx();
+        let mgr = DnsPeerMgr::new(peer_mgr, global_ctx.clone());
+
+        let got = DnsPeerMgrRpc::get_export_config(
+            mgr.0.as_ref(),
+            BaseController::default(),
+            GetExportConfigRequest {},
+        )
+        .await
+        .expect("get_export_config should succeed");
+
+        assert_eq!(got, global_ctx.dns_export_config());
+    }
+
+    #[tokio::test]
+    async fn refresh_self_peer_marks_dirty_only() {
+        let peer_mgr = create_peer_manager_with_zone(
+            "self-peer",
+            "self-zone.test",
+            Ipv4Addr::new(10, 0, 0, 1),
+        )
+        .await;
+        let mgr = DnsPeerMgr::new(peer_mgr.clone(), peer_mgr.get_global_ctx());
+
+        mgr.dirty.reset();
+        mgr.try_refresh(peer_mgr.my_peer_id()).await.unwrap();
+
+        assert!(mgr.dirty.peek());
+    }
+
+    #[tokio::test]
+    async fn refresh_missing_route_noop_and_not_dirty() {
+        let peer_mgr = create_peer_manager_with_zone(
+            "solo-peer",
+            "solo-zone.test",
+            Ipv4Addr::new(10, 0, 0, 2),
+        )
+        .await;
+        let mgr = DnsPeerMgr::new(peer_mgr, get_mock_global_ctx());
+
+        mgr.dirty.reset();
+        mgr.try_refresh(987_654).await.unwrap();
+
+        assert!(!mgr.dirty.peek());
+    }
+
+    #[tokio::test]
+    async fn refresh_same_digest_skips_fetch_and_not_mark_dirty() {
+        let local = create_peer_manager_with_zone(
+            "local-same",
+            "local-same.test",
+            Ipv4Addr::new(10, 0, 1, 1),
+        )
+        .await;
+        let remote = create_peer_manager_with_zone(
+            "remote-same",
+            "remote-same.test",
+            Ipv4Addr::new(10, 0, 1, 2),
+        )
+        .await;
+
+        connect_peer_manager(local.clone(), remote.clone()).await;
+        wait_route_appear(local.clone(), remote.clone())
+            .await
+            .expect("route should appear");
+
+        let remote_id = remote.my_peer_id();
+        let remote_route_dns = local
+            .get_route()
+            .get_peer_info(remote_id)
+            .await
+            .expect("remote route should exist")
+            .dns;
+
+        let mgr = DnsPeerMgr::new(local, get_mock_global_ctx());
+        mgr.peers
+            .insert(
+                remote_id,
+                DnsPeerInfo {
+                    digest: remote_route_dns
+                        .try_into()
+                        .expect("route dns digest should be 32 bytes"),
+                    zones: vec![valid_zone_data("cached-same.test", "10.0.1.9")],
+                },
+            )
+            .await;
+
+        mgr.dirty.reset();
+        mgr.try_refresh(remote_id).await.unwrap();
+        sleep(Duration::from_millis(50)).await;
+
+        assert!(!mgr.dirty.peek());
+    }
+
+    #[tokio::test]
+    async fn refresh_remote_peer_fetches_and_updates_snapshot() {
+        let local = create_peer_manager_with_zone(
+            "local-refresh",
+            "local-refresh.test",
+            Ipv4Addr::new(10, 0, 2, 1),
+        )
+        .await;
+        let remote = create_peer_manager_with_zone(
+            "remote-refresh",
+            "remote-export.test",
+            Ipv4Addr::new(10, 0, 2, 2),
+        )
+        .await;
+
+        let local_dns = DnsPeerMgr::new(local.clone(), local.get_global_ctx());
+        let remote_dns = DnsPeerMgr::new(remote.clone(), remote.get_global_ctx());
+        remote_dns.register();
+
+        connect_peer_manager(local.clone(), remote.clone()).await;
+        wait_route_appear(local.clone(), remote.clone())
+            .await
+            .expect("route should appear");
+
+        local_dns.dirty.reset();
+        local_dns.try_refresh(remote.my_peer_id()).await.unwrap();
+
+        assert!(local_dns.dirty.peek());
+        let snapshot = local_dns.snapshot();
+        assert!(
+            snapshot
+                .zones
+                .iter()
+                .any(|z| z.content.contains("$ORIGIN remote-export.test"))
+        );
+    }
+
+    #[tokio::test]
+    async fn multi_peer_refresh_updates_only_target_peer_snapshot_data() {
+        let local = create_peer_manager_with_zone(
+            "local-multi-refresh",
+            "local-multi-refresh.test",
+            Ipv4Addr::new(10, 2, 0, 1),
+        )
+        .await;
+        let peer_a =
+            create_peer_manager_with_zone("peer-a", "remote-a.test", Ipv4Addr::new(10, 2, 0, 2))
+                .await;
+        let peer_b =
+            create_peer_manager_with_zone("peer-b", "remote-b.test", Ipv4Addr::new(10, 2, 0, 3))
+                .await;
+
+        let local_dns = DnsPeerMgr::new(local.clone(), local.get_global_ctx());
+        let peer_a_dns = DnsPeerMgr::new(peer_a.clone(), peer_a.get_global_ctx());
+        peer_a_dns.register();
+
+        connect_peer_manager(local.clone(), peer_a.clone()).await;
+        connect_peer_manager(local.clone(), peer_b.clone()).await;
+        wait_route_appear(local.clone(), peer_a.clone())
+            .await
+            .expect("route to peer_a should appear");
+        wait_route_appear(local.clone(), peer_b.clone())
+            .await
+            .expect("route to peer_b should appear");
+
+        local_dns.try_refresh(peer_a.my_peer_id()).await.unwrap();
+
+        let snapshot = local_dns.snapshot();
+        assert!(
+            snapshot
+                .zones
+                .iter()
+                .any(|z| z.content.contains("$ORIGIN remote-a.test"))
+        );
+        assert!(
+            !snapshot
+                .zones
+                .iter()
+                .any(|z| z.content.contains("$ORIGIN remote-b.test"))
+        );
+    }
+
+    #[tokio::test]
+    async fn multi_peer_refresh_failure_invalidates_only_target_peer_cache() {
+        let local = create_peer_manager_with_zone(
+            "local-invalidate",
+            "local-invalidate.test",
+            Ipv4Addr::new(10, 2, 1, 1),
+        )
+        .await;
+        let fail_peer = create_peer_manager_with_zone(
+            "peer-fail",
+            "peer-fail.test",
+            Ipv4Addr::new(10, 2, 1, 2),
+        )
+        .await;
+        let keep_peer = create_peer_manager_with_zone(
+            "peer-keep",
+            "peer-keep.test",
+            Ipv4Addr::new(10, 2, 1, 3),
+        )
+        .await;
+
+        let local_dns = DnsPeerMgr::new(local.clone(), local.get_global_ctx());
+        local_dns.register();
+        let keep_dns = DnsPeerMgr::new(keep_peer.clone(), keep_peer.get_global_ctx());
+        keep_dns.register();
+
+        let fail_id = fail_peer.my_peer_id();
+        let keep_id = keep_peer.my_peer_id();
+
+        local_dns
+            .peers
+            .insert(
+                fail_id,
+                DnsPeerInfo {
+                    digest: [1; 32],
+                    zones: vec![valid_zone_data("cached-fail.test", "10.2.1.20")],
+                },
+            )
+            .await;
+        local_dns
+            .peers
+            .insert(
+                keep_id,
+                DnsPeerInfo {
+                    digest: [2; 32],
+                    zones: vec![valid_zone_data("cached-keep.test", "10.2.1.21")],
+                },
+            )
+            .await;
+
+        connect_peer_manager(local.clone(), fail_peer.clone()).await;
+        connect_peer_manager(local.clone(), keep_peer.clone()).await;
+        wait_route_appear(local.clone(), fail_peer.clone())
+            .await
+            .expect("route to fail_peer should appear");
+        wait_route_appear(local.clone(), keep_peer.clone())
+            .await
+            .expect("route to keep_peer should appear");
+
+        local_dns.dirty.reset();
+        local_dns
+            .refresh(fail_id, Default::default(), Default::default())
+            .await
+            .unwrap_err();
+
+        assert!(local_dns.dirty.peek());
+        assert!(local_dns.peers.get(&fail_id).await.is_none());
+        assert!(local_dns.peers.get(&keep_id).await.is_some());
+    }
+
+    #[tokio::test]
+    async fn multi_peer_mixed_digest_changes_only_mark_for_changed_peer() {
+        let local = create_peer_manager_with_zone(
+            "local-mixed",
+            "local-mixed.test",
+            Ipv4Addr::new(10, 2, 2, 1),
+        )
+        .await;
+        let changed_peer = create_peer_manager_with_zone(
+            "peer-changed",
+            "peer-changed.test",
+            Ipv4Addr::new(10, 2, 2, 2),
+        )
+        .await;
+        let unchanged_peer = create_peer_manager_with_zone(
+            "peer-unchanged",
+            "peer-unchanged.test",
+            Ipv4Addr::new(10, 2, 2, 3),
+        )
+        .await;
+
+        let local_dns = DnsPeerMgr::new(local.clone(), local.get_global_ctx());
+        let changed_dns = DnsPeerMgr::new(changed_peer.clone(), changed_peer.get_global_ctx());
+        let unchanged_dns =
+            DnsPeerMgr::new(unchanged_peer.clone(), unchanged_peer.get_global_ctx());
+        changed_dns.register();
+        unchanged_dns.register();
+
+        connect_peer_manager(local.clone(), changed_peer.clone()).await;
+        connect_peer_manager(local.clone(), unchanged_peer.clone()).await;
+        wait_route_appear(local.clone(), changed_peer.clone())
+            .await
+            .expect("route to changed_peer should appear");
+        wait_route_appear(local.clone(), unchanged_peer.clone())
+            .await
+            .expect("route to unchanged_peer should appear");
+
+        let unchanged_id = unchanged_peer.my_peer_id();
+        let unchanged_digest = local
+            .get_route()
+            .get_peer_info(unchanged_id)
+            .await
+            .expect("unchanged route should exist")
+            .dns;
+
+        local_dns
+            .peers
+            .insert(
+                changed_peer.my_peer_id(),
+                DnsPeerInfo {
+                    digest: [0; 32],
+                    zones: vec![valid_zone_data("stale-changed.test", "10.2.2.20")],
+                },
+            )
+            .await;
+        local_dns
+            .peers
+            .insert(
+                unchanged_id,
+                DnsPeerInfo {
+                    digest: unchanged_digest
+                        .try_into()
+                        .expect("route dns digest should be 32 bytes"),
+                    zones: vec![valid_zone_data("cached-unchanged.test", "10.2.2.21")],
+                },
+            )
+            .await;
+
+        local_dns.dirty.reset();
+        local_dns
+            .try_refresh(changed_peer.my_peer_id())
+            .await
+            .unwrap();
+        assert!(local_dns.dirty.peek());
+
+        local_dns.dirty.reset();
+        local_dns.try_refresh(unchanged_id).await.unwrap();
+        assert!(!local_dns.dirty.peek());
+
+        let unchanged_cache = local_dns
+            .peers
+            .get(&unchanged_id)
+            .await
+            .expect("unchanged peer cache should stay");
+        assert!(
+            unchanged_cache
+                .zones
+                .iter()
+                .any(|z| z.content.contains("$ORIGIN cached-unchanged.test"))
+        );
+    }
+
+    #[tokio::test]
+    async fn snapshot_removes_cached_peer_zone_after_tti_expire() {
+        let peer_mgr = create_peer_manager_with_zone(
+            "local-tti",
+            "local-tti.test",
+            Ipv4Addr::new(10, 3, 0, 1),
+        )
+        .await;
+        let global_ctx = peer_mgr.get_global_ctx();
+        let mgr = DnsPeerMgr::new(peer_mgr, global_ctx);
+
+        let cached_peer_id = 66_666;
+        mgr.peers
+            .insert(
+                cached_peer_id,
+                DnsPeerInfo {
+                    digest: [6; 32],
+                    zones: vec![valid_zone_data("cached-expire.test", "10.3.0.2")],
+                },
+            )
+            .await;
+
+        let before = mgr.snapshot();
+        assert!(
+            before
+                .zones
+                .iter()
+                .any(|z| z.content.contains("$ORIGIN cached-expire.test"))
+        );
+        assert!(
+            before
+                .zones
+                .iter()
+                .any(|z| z.content.contains("$ORIGIN local-tti.test"))
+        );
+
+        let deadline = tokio::time::Instant::now() + DNS_PEER_TTI + Duration::from_secs(3);
+        loop {
+            let now_snapshot = mgr.snapshot();
+            let expired = !now_snapshot
+                .zones
+                .iter()
+                .any(|z| z.content.contains("$ORIGIN cached-expire.test"));
+            if expired {
+                assert!(
+                    now_snapshot
+                        .zones
+                        .iter()
+                        .any(|z| z.content.contains("local-tti.test"))
+                );
+                break;
+            }
+
+            assert!(
+                tokio::time::Instant::now() < deadline,
+                "cached peer zone did not expire within expected TTI window"
+            );
+            sleep(Duration::from_millis(100)).await;
+        }
+    }
+
+    #[tokio::test]
+    async fn register_then_unregister_returns_some() {
+        let peer_mgr = create_peer_manager_with_zone(
+            "register-peer",
+            "register-zone.test",
+            Ipv4Addr::new(10, 1, 0, 1),
+        )
+        .await;
+        let mgr = DnsPeerMgr::new(peer_mgr.clone(), peer_mgr.get_global_ctx());
+        mgr.register();
+        assert!(mgr.unregister().is_some());
+    }
+}

easytier/src/dns/plan.md

@@ -0,0 +1,163 @@
+## 目标
+
+将 `instance/dns_server` 重写为单独的 `dns` 模块，为如下的配置项提供支持：
+
+```toml
+[dns]
+name = "localhost" # optional, replaces hostname, default to system hostname
+domain = "localdomain" # optional, replaces tld_dns_zone, default to et.net
+
+addresses = [
+    "100.100.100.101:53",
+] # optional, default to [ "100.100.100.101:53" ]
+# any UDP packet or ICMP packet to these addresses will be hijacked by the dns server
+# the server does *not* bind to/listen on these addresses!
+
+listeners = [
+] # optional, default to empty
+# let the dns server bind to these addresses
+# could be useful when no_tun = true
+
+# these two options supersede accept_dns
+# setting both of them to empty is equivalent to set accept_dns = false, but zones are still broadcasted
+
+# this policy applies to all zones with origin "example.com"
+[dns."example.com".import]
+whitelist = ["*"]
+blacklist = []
+disabled = true # optional, whether to reject zones with this origin from connected peers, default to false
+recursive = true # optional, apply this policy to all subzones, default to false
+
+[[dns.zone]]
+origin = "example.com" # required, name of the zone
+ttl = 3600 # optional, default to 0
+records = [
+    "www 60 IN A 123.123.123.123",
+    "app IN CNAME www",
+] # optional, custom DNS records
+
+forwarders = [
+    "1.1.1.1",
+] # optional, forward DNS requests to these servers
+
+fallthrough = false # optional, whether to fall back to next zone (with same origin) if request doesn't match any record in this zone, default to true
+
+# this policy applies to the current zone block
+[dns.zone.export] # if present, export this zone to connected peers
+whitelist = ["*"] # optional
+blacklist = [] # optional
+
+# same zone, but not exported
+[[dns.zone]]
+origin = "example.com"
+
+forwarders = [
+    "tcp://192.168.0.53:5353",
+]
+```
+
+<details>
+<summary><h2>计划和进展</h2></summary>
+
+每个 peer 会默认拥有一个专用 zone，它的 origin 是这个 peer 的 fqdn，唯一的记录是指向该 peer 的 ip 的 A、AAAA 记录
+
+## protobuf
+
+- `ZoneData`：包含 Zone 配置，以及一个 ID，该 ID 在读取 TOML 时生成
+- `GetExportConfigResponse` (`DnsExportConfig`)：包含全部 export 的 `ZoneData`（特别地，包含专用 zone）、该 peer 的 fqdn
+- `HeartbeatRequest`: DnsNode 发送的心跳，包含：id、digest、`Option<Snapshot>`
+- `DnsSnapshot`: 所有 DnsServer 需要的配置
+
+## RoutePeerInfo
+
+为预防用户提交大量自定义 DNS 记录导致 RoutePeerInfo 泛洪造成带宽压力：
+
+- 在 `RoutePeerInfo` 中只保存本地 DNS 配置的 hash
+- 收到 `RoutePeerInfo` 后读取其中 DNS 的 hash，若与本地不同，通过 RPC 拉取 Peer 的 DNS 配置
+
+## DnsNode
+
+1. - [x] 监听配置更新/IP 地址变化，重建快照
+2. - [x] (`GlobalCtx` 的扩展 trait) 使用自己的 name 和 domain 创建一个专用 zone，让 name 指向自身 IP（为 DNS 一致性避免使用 127.0.0.1 作为 IP，若没有 IP 则不创建这个 zone）
+3. - [x] 每次获得 RoutePeerInfo 时，读取其中的 dns 字段（和一些别的身份标记字段），这是远程 Peer 的 dns 配置（不含 addresses 和 listeners）的 digest
+   - [x] 接收后检查 digest 和本地配置是否一致，如果一致，不做修改，否则标记 dirty，下一次心跳时将重建快照
+4. - [x] 每隔一小段时间向 DnsServer 发送心跳和当前 digest： 
+   1. 如果没有 dirty 标记，心跳不含 snapshot；
+   2. 如果有 dirty 标记，重建 snapshot 并在心跳中包含；
+   3. 如果 DnsServer 返回 resync，立刻重新发送带有 Snapshot 的心跳
+5. - [x] 一个 RPC 接口，供 Peer 拉取 DNS 配置
+6. - [x] 一个独立循环，用于选举 DnsServer，每次循环尝试绑定 DNS_SERVER_RPC_ADDR 监听 RPC 请求
+   1. 一台机器上所有 EasyTier 实例一起尝试绑定 DNS_SERVER_RPC_ADDR，绑定成功的那个就启动 DnsServer（当然也启动 DnsNode），失败的那些就只有 DnsNode
+   2. 每隔一小段时间或者 DnsNode 心跳失败（notify）后立刻尝试 bind，如果 bind 成功就说明 DnsServer 真挂了，那就自己在这个已有的 SocketAddr 上启动 DnsServer（忽略 bind 失败或启动失败，启动失败就直接释放 socket），这样才能保证服务不断
+
+## DnsServer
+
+1.  - [x] 提供一个 RPC 接口接受 DnsNode 的心跳，如果心跳 digest 和本地不符则返回 resync
+2.  - [x] 收到含有 snapshot 的心跳时替换本地配置；如果 snapshot 中的 listeners 或者 addresses 不同则 rebind
+3.  - [x] (`moka::Cache`) 持续检查是否有过期（丢失心跳）的 DnsNode，需要把这些 DnsNode 提供的所有配置清除
+4.  - [x] 每次更新 zone 时自动添加 root zone
+    - [x] (`Zone::system`) 并把它的 forwarder 设置为系统 DNS
+5.  - [x] 使用 snapshot 更新 zone。不用合并同名 zone，直接用 Zone 结构体提供的 ChainedZoneHandler 按顺序插入 Catalog 就行，不过注意要先插入 MemoryZoneHandler，这些都是 records，后插入 ForwardZoneHandler，这都是 forwarders
+6.  - [x] 更新 zone 的时候自动去掉 forwarder 中导致回环的那些，就是把 addresses 和 listeners 去掉（root zone 也需要这个逻辑）
+7.  - [x] 内部接口，控制 DnsServer 是否 bind 到某些 socket（也就是配置中的 listeners）
+8.  - [x] Listeners 绑定失败打印日志（失败一个打印一次然后就跳过），即便这时 addresses 为空也不要停机。（否则释放 socket 绑定后会有 instance 抢占 socket 试图启动 server，然后就死循环）
+9.  - [x] 内部接口，更新 addresses。目前这些用来 hijack 的 addresses 都是只支持 udp 简单查询，就是一个 UDP 包查询，tcp 完全不管。但是可以支持除了 53 之外的端口，这个不难。
+    - [x] 并且给 tun 添加删除这些 addresses 的路由
+10. - [x] 启动时，往 packet pipeline 上挂一个 filter，和目前 magic dns 的操作一样，给 addresses 添加路由并劫持所有目的为配置中 addresses 的 UDP 包，直接作为 DNS request 读取并交给 DnsServer 解析
+    - [x] 这个 addresses 可能还得 append 到 resolv.conf 之类的地方
+11. - [x] Addresses 和 Listeners 更新时~~需要检查所有 zone 的 forwarder~~直接更新所有 zone，之前为了避免回环可能去掉了一些 forwarder，或者有新的 forwarder 要去除
+
+此外，还有以下几个设计要点：
+
+- Zone 允许只有 forwarder，这时候就是纯转发器
+- Zone 允许没有 forwarder，这时候要检查是不是有 SOA 和 NS 记录，如果没有可能需要添加？
+- 另一种方案是 DnsNode 挂 filter，自己处理 UDP 劫持，用某种方式（如 RPC）把 DNS 请求代理给 DnsServer，该方案的优势在于完全解耦 DnsServer 的实现，特别是解决了 DnsServer 所在实例可能 no_tun 的问题，缺点是：
+  - 性能更差
+  - 操作路由表或 /etc/resolv.conf 时会有多个 instance 同时修改，修改结果没有确定性
+  - DnsServer 仍然需要得知 addresses 以进行回环检测
+  - debug 更麻烦
+  - 难以实现策略 DNS，比如不同来源的 DNS 请求走不同的 zone
+
+另外任何关于系统 DNS 的操作，清理都参考现有的 magic dns。
+
+## 已知但无需/无计划解决的问题
+
+- the ttl option isn't working because of https://github.com/hickory-dns/hickory-dns/pull/3450
+- [minor] address 路由绑定必须在有 tun 的实例上做；listener 绑定则与 tun 无关，现有竞选机制无法保证有 tun 的实例能优先启动 DnsServer
+  - 不妨假设大多数情况下一台机器上所有实例的 no_tun 设置相同，这时候这个问题实际上不存在
+- [minor] DnsServer 更新 zone 的时候需要更精细的合并/去重控制，如延迟低者/本地优先
+- [minor] 更新 forwarder 时还需要检查间接回环，如 DNS 请求发送给某个 Peer，这个 Peer 又把请求转发回自己了
+- [minor] 防止死锁/挂起的 DnsServer 占用 socket
+- ~~[minor] RoutePeerInfo 可能不能过大~~
+- [minor] 增量 Zone 更新
+- DNS 策略
+
+</details>
+
+<details>
+<summary><h2>Related Issues</h2></summary>
+
+- closes https://github.com/EasyTier/EasyTier/issues/742
+- closes https://github.com/EasyTier/EasyTier/issues/771
+- closes https://github.com/EasyTier/EasyTier/issues/927
+- closes https://github.com/EasyTier/EasyTier/issues/1071
+- closes https://github.com/EasyTier/EasyTier/issues/1142
+- closes https://github.com/EasyTier/EasyTier/issues/1322
+- closes https://github.com/EasyTier/EasyTier/issues/1381
+- closes https://github.com/EasyTier/EasyTier/issues/1488
+- closes https://github.com/EasyTier/EasyTier/issues/1597
+- closes https://github.com/EasyTier/EasyTier/issues/1645
+- closes https://github.com/EasyTier/EasyTier/issues/1764
+- closes https://github.com/EasyTier/EasyTier/issues/1814
+- closes https://github.com/EasyTier/EasyTier/issues/1826
+- closes https://github.com/EasyTier/EasyTier/issues/2004
+
+---
+
+- (maybe) related to https://github.com/EasyTier/EasyTier/issues/937
+- (maybe) related to https://github.com/EasyTier/EasyTier/issues/1016
+- (maybe) related to https://github.com/EasyTier/EasyTier/issues/1348
+- (maybe) related to https://github.com/EasyTier/EasyTier/issues/1699
+- (maybe) related to https://github.com/EasyTier/EasyTier/issues/1873
+
+</details>

easytier/src/dns/readme.md

@@ -0,0 +1,562 @@
+# EasyTier DNS 模块设计说明（重构版）
+
+> 本文档基于 `easytier/src/dns` 当前代码实现与 `plan.md`。  
+> 当前先给出第一部分：**整体架构与基础逻辑**。
+
+## 1. 模块目标与定位
+
+`dns` 模块是对旧 `instance/dns_server` 方案的重构，目标是把 DNS 能力从“单点功能”升级为“可同步、可扩展、可自治”的子系统。它同时承担三类职责：
+
+1. **本机 DNS 服务能力**：
+   - 能监听配置中的 `listeners`（UDP/TCP）作为标准 DNS server。
+   - 能对配置中的 `addresses` 做流量劫持（UDP DNS + ICMP echo）。
+
+2. **多 Peer DNS 配置同步能力**：
+   - 每个实例作为 `DnsNode` 生成快照并定期心跳。
+   - 机器上被选举出的 `DnsServer` 聚合所有 Node 快照并动态重建 Catalog。
+
+3. **系统 DNS 接入能力**（tun 场景）：
+   - 把 DNS nameserver/search/match domain 写入系统配置（当前主要是 Windows/macOS，Linux 仍在演进中）。
+
+---
+
+## 2. 顶层架构（角色分层）
+
+从职责上看，模块分成 4 层：
+
+- **配置层**（`config/*`）
+  - 解析 TOML 的 `[dns]`、`[[dns.zone]]`、策略字段。
+  - 产出 `DnsConfig`、`ZoneConfig`，并提供默认值（如默认域名 `et.net`、默认地址 `100.100.100.101:53`）。
+
+- **节点层（控制面）**（`node.rs` + `peer_mgr.rs`）
+  - `DnsNode`：本实例的 DNS 控制器，负责选举、心跳、事件监听、重建 snapshot。
+  - `DnsPeerMgr`：维护远端 peer 的 DNS 摘要与配置拉取，拼装 `DnsSnapshot`。
+
+- **服务层（数据面）**（`server.rs` + `node_mgr.rs` + `zone.rs`）
+  - `DnsServer`：真正处理 DNS 请求、维护监听 socket、管理 hijack addresses。
+  - `DnsNodeMgr`：服务端的快照管理器，接收 Node 心跳，维护节点 TTL 与 dirty 状态。
+  - `Zone`/`ZoneGroup`：把 records + forwarders 变成 Hickory `ZoneHandler` 并装配 `Catalog`。
+
+- **系统集成层**（`system/*`）
+  - 将当前 DNS 配置下发到 OS（`SystemConfigurator` 抽象）。
+  - 服务退出/变更时负责清理或覆盖。
+
+---
+
+## 3. 关键对象与数据模型
+
+- **`DnsConfig`**（`config/dns.rs`）
+  - 核心字段：`zones`、`policies`、`name`、`domain`、`addresses`、`listeners`。
+  - `get_fqdn()` 用 `name + domain` 生成本机 FQDN。
+
+- **`ZoneData` / `Zone`**（`proto/dns.proto` + `zone.rs`）
+  - `ZoneData` 是网络传输模型（protobuf），含 `id/origin/ttl/records/forwarders`。
+  - `Zone` 是运行期模型：
+    - `records -> InMemoryZoneHandler`
+    - `forwarders -> ForwardZoneHandler`
+    - 同 origin 可链式共存（ChainedZoneHandler 语义）。
+
+- **`DnsSnapshot`**（`proto/dns.proto`）
+  - Node 发给 Server 的完整状态：`zones + addresses + listeners`。
+
+- **`HeartbeatRequest`**
+  - 发送 `id + digest + optional snapshot`。
+  - digest 一致时可只发轻量心跳，不带 snapshot。
+
+- **`DirtyFlag`**（`utils/dirty.rs`）
+  - 全模块统一的“脏标记 + 通知器”，用于节流和增量触发（不是每次事件都全量重建）。
+
+---
+
+## 4. 基础运行逻辑（主链路）
+
+### 4.1 本地节点启动
+
+`Instance` 在 `magic-dns` feature 下创建并启动 `DnsNode`。`DnsNode` 启动后并行跑两个循环：
+
+1. **选举循环**（`run_election`）
+   - 周期尝试绑定固定 RPC 地址 `tcp://127.0.0.1:49813`。
+   - 绑定成功者成为本机 `DnsServer` 持有者；失败者继续只做 `DnsNode`。
+
+2. **主循环**（`run`）
+   - 监听配置变更/IP 变化/PeerInfo 更新。
+   - 维护 dirty 状态并按节奏发送 heartbeat。
+
+### 4.2 快照构建与同步
+
+`DnsPeerMgr::snapshot()` 组装快照：
+
+- 本机 zones：`dns_iter_zones()`（包含“自有专用 zone” + 用户配置 zone）。
+- 远端 zones：从 peer RPC 拉取并缓存的 export zones。
+- 本机 `addresses/listeners`：来自 `DnsConfig`。
+
+Node 发送 heartbeat 时：
+
+- dirty 或首包 -> 带 `snapshot` 全量发送。
+- 未 dirty -> 只发 `digest`（轻量心跳）。
+- Server 返回 `resync=true` -> 立刻补发全量 snapshot。
+
+### 4.3 服务端聚合与生效
+
+`DnsNodeMgr` 收到 heartbeat 后：
+
+- 若 snapshot digest 改变：更新节点缓存并标记 dirty（catalog/addresses/listeners 分开标记）。
+- 若仅 digest 且本地无该节点或不一致：返回 `resync=true`。
+
+`DnsServer::run()` 有三个独立 reload 循环：
+
+- `reload_catalog`：替换 `DynamicCatalog`。
+- `reload_addresses`：更新 hijack 地址，并尝试下发系统 DNS。
+- `reload_listeners`：重绑 DNS listener socket。
+
+这三个循环彼此解耦，避免单一失败阻塞全部 DNS 功能。
+
+---
+
+## 5. 数据面请求路径（DNS/ICMP 劫持）
+
+`DnsServer` 作为 `NicPacketFilter` 挂入 packet pipeline：
+
+1. 检查目的 IP 是否命中 `addresses`。
+2. UDP：
+   - 解析 DNS 请求 -> 投递给 `catalog.handle_request()`。
+   - 用响应覆盖原 UDP payload，修正长度与校验和。
+3. ICMP：
+   - 对 EchoRequest 直接改写为 EchoReply。
+4. 最后交换源/目的 IP，并把包回注到本机 peer pipeline。
+
+这使得 `addresses` 不要求真实 bind/listen，也能作为“虚拟 DNS 入口地址”。
+
+---
+
+## 6. 可靠性与收敛机制
+
+- **服务高可用（单机维度）**：
+  - 任何实例都可竞选 Server；现任退出后其余实例会重试接管。
+
+- **配置高效同步（全网维度）**：
+  - `RoutePeerInfo` 只传播 DNS digest，不直接携带全量记录。
+  - digest 变化后才通过 RPC 拉取详情，降低路由泛洪压力。
+
+- **自动过期清理**：
+  - `DnsNodeMgr` 通过 `moka::Cache` TTL 自动淘汰失联节点配置（心跳过期）。
+
+- **回环防护**：
+  - 重建 zones 时会从 forwarders 中剔除本地 `addresses/listeners`，避免显式自环。
+
+---
+
+## 7. 当前实现状态（对应 plan.md）
+
+从代码可见，以下主干能力已经落地：
+
+- Node/Server 双角色、选举、心跳与 resync。
+- 快照机制（zone/addresses/listeners）与 digest 驱动同步。
+- 自有专用 zone 自动生成与 export。
+- Catalog 动态替换、listener/address 分离热更新。
+- UDP DNS 劫持 + ICMP 响应。
+- forwarder 的本地回环剔除。
+
+仍在计划中的重点：
+
+- 系统 DNS 配置改造（尤其 Linux 路径统一与清理语义完善）。
+- 更完整的单元测试覆盖与 CLI 状态输出。
+
+---
+
+## 8. 配置层详解（`config/*`）
+
+这一层负责把 TOML 配置映射成可校验、可传播、可计算 digest 的运行模型。
+
+### 8.1 常量与默认值（`config/mod.rs`）
+
+- `DNS_DEFAULT_TLD = et.net.`：`domain` 缺省值。
+- `DNS_DEFAULT_ADDRESS = udp://100.100.100.101:53`：`addresses` 缺省值。
+- `DNS_SERVER_RPC_ADDR = tcp://127.0.0.1:49813`：本机 DNS Server 选举地址。
+- `DNS_SERVER_ELECTION_INTERVAL = 5s`：选举重试周期。
+- `DNS_SUPPORTED_PROTOCOLS = [Udp, Tcp]`：地址/转发器协议白名单。
+
+### 8.2 `DnsConfig`（`config/dns.rs`）
+
+`DnsConfig` 是 `[dns]` 根配置，关键点如下：
+
+- `zones: Vec<ZoneConfig>` 对应 `[[dns.zone]]`。
+- `policies: HashMap<LowerName, DnsPolicyConfig>` 用 `#[serde(flatten)]` 承接 `[dns."origin".import]` 形式策略。
+- `name/domain` 组合 FQDN。
+- `addresses/listeners` 使用 `NameServerAddrGroup`（支持 `ip`、`ip:port`、`udp://`、`tcp://` 解析）。
+
+约束与语义：
+
+- `deserialize_addresses()` 强制 `addresses` 只能是 UDP（与当前 hijack 数据面能力一致）。
+- `get_name()`：若 `name` 为空，回退系统 hostname。
+- `get_fqdn()`：将 `name` 拼接 `domain` 得到完整域名。
+- `set_fqdn()`：反向拆分 FQDN 到 `name` 和 `domain`。
+
+### 8.3 `ZoneConfig` 与专用 Zone（`config/zone.rs`）
+
+`ZoneConfig` 由两部分构成：
+
+- `ZoneData`：用于 protobuf 传输（`id/origin/ttl/records/forwarders`）。
+- `ZoneConfigInner`：配置层字段（含 policy）。
+
+关键设计：
+
+- `TryFrom<ZoneConfigInner> for ZoneConfig` 会立即调用 `Zone::try_from(&ZoneData)` 做语法校验，确保“能进配置就能进运行时”。
+- `ZoneConfig::dedicated(...)` 用于自动生成“本节点专用 zone”：
+  - `origin = 节点 fqdn`
+  - records 自动填充 `@ IN A/AAAA ...`
+  - `policy.export = Some(default)`，默认可导出给 peers。
+
+### 8.4 策略结构体现状（`config/policy.rs`）
+
+策略模型已就位，但功能并未完全落实到执行路径：
+
+- `AclPolicy { whitelist, blacklist }`
+- `FunctionalityPolicy { disabled }`
+- `DnsPolicy { recursive }`
+
+目前代码中的直接使用点主要是：
+
+- `dns_export_config()` 只检查 `zone.policy.export.is_some()` 决定是否导出。
+- `import/recursive/acl` 仍处于待完整落地状态（与 `plan.md` 的 TODO 对齐）。
+
+### 8.5 `DnsGlobalCtxExt`：配置到发布面的桥（`config/dns.rs`）
+
+`GlobalCtx` 被扩展出 3 个关键方法：
+
+- `dns_self_zone()`：基于当前 IP 与 FQDN 生成专用 zone。
+- `dns_iter_zones()`：`self_zone + 用户配置 zones`。
+- `dns_export_config()`：从 `dns_iter_zones()` 中筛选可导出的 zones，并附加本机 `fqdn`。
+
+这三个方法是后续 `RoutePeerInfo.dns` digest 与 RPC 拉取的源头。
+
+---
+
+## 9. 节点控制面详解（`node.rs` + `peer_mgr.rs`）
+
+### 9.1 `DnsNode` 初始化与 RPC 注册
+
+`DnsNode::new(...)` 会创建 `DnsPeerMgr`，并把 `DnsPeerMgrRpcServer` 注册到 peer RPC registry。  
+这使“我给别人提供 DNS 导出配置”与“我向别人拉取导出配置”在同一组件闭环。
+
+### 9.2 选举循环（`DnsNode::run_election`）
+
+选举逻辑是“抢占固定地址”的单机 leader 机制：
+
+1. 周期或被 `elect.notify_one()` 触发。
+2. 尝试 `StandAloneServer(TcpTunnelListener(DNS_SERVER_RPC_ADDR)).serve()`。
+3. 绑定成功 -> 启动 `DnsServer`，注册 `DnsNodeMgrRpc`，并挂载 NIC packet pipeline。
+4. `DnsServer` 退出后清理 pipeline，回到选举循环。
+
+要点：
+
+- 不依赖外部分布式锁，仅利用本机 socket 独占。
+- 失败不是错误态，意味着“已有实例担任 Server”。
+
+### 9.3 主循环（`DnsNode::run`）
+
+主循环负责“何时重建、何时发全量、何时触发重选举”：
+
+- 维护 `HeartbeatRequest { id, digest, snapshot? }`。
+- 基于 `DirtyFlag` 动态调整心跳节奏：
+  - dirty 时更积极（`rr_interval`）
+  - clean 时更快短轮询（`rr_interval / 8`）
+- 监听 `GlobalCtxEvent`：
+  - `PeerInfoUpdated` -> 并发调用 `mgr.refresh(peer_id)`
+  - IP 变化、配置变化、事件丢失（lagged）-> `dirty.mark()`
+- 心跳失败 -> 触发一次选举通知（可能是 Server 挂了）。
+
+### 9.4 心跳协议（`DnsNode::heartbeat`）
+
+发送策略：
+
+- 首次或 dirty -> `heartbeat.update(self.mgr.snapshot())`，发送全量 snapshot。
+- 非 dirty -> 尽量只发 digest（轻量包）。
+
+服务端响应：
+
+- `resync = true` 时，客户端立刻再发一次带 snapshot 的心跳。
+
+这实现了“正常轻量保活 + 状态漂移时快速自愈”。
+
+### 9.5 `DnsPeerMgr`：远端配置拉取与去抖
+
+`DnsPeerMgr` 核心职责：
+
+- 本地缓存：`Cache<PeerId, DnsPeerInfo>`（TTL = 3s）。
+- `refresh(peer_id)`：
+  - 先读路由里的 `route.dns` digest。
+  - 若与本地缓存一致则跳过 RPC。
+  - 不一致才调用 `fetch(peer_id)` 拉取 `GetExportConfigResponse`。
+- `snapshot()`：拼接
+  - 本机 zones（`dns_iter_zones()`）
+  - 所有远端缓存 zones
+  - 本机 addresses/listeners
+
+这正是 `plan.md` 中“RoutePeerInfo 仅携带 hash，详情按需拉取”的落地实现。
+
+---
+
+## 10. 服务聚合与数据面详解（`node_mgr.rs` + `server.rs`）
+
+### 10.1 `DnsNodeMgr`：服务器侧状态机
+
+`DnsNodeMgr` 保存每个 Node 的最新状态：
+
+- `nodes: Cache<Uuid, DnsNodeInfo>`（TTL = 5s，心跳过期即自动淘汰）。
+- `DnsNodeInfo = digest + zones + addresses + listeners`。
+- `dirty` 分三类：`catalog`、`addresses`、`listeners`。
+
+`heartbeat()` 判定逻辑：
+
+- 请求带 snapshot：
+  - 反序列化为 `DnsNodeInfo`。
+  - digest 变化才更新缓存并打脏标记。
+  - 此分支返回 `resync = false`。
+- 请求不带 snapshot：
+  - 若本地没有该 node 或 digest 不一致 -> `resync = true`。
+
+### 10.2 Catalog 构建（`DnsNodeMgr::catalog/collect_zones`）
+
+构建步骤：
+
+1. 聚合全部节点 zones。
+2. 追加 `Zone::system()` 作为 root zone。
+3. 收集本地所有 `addresses + listeners` 形成 `local` 集合。
+4. 遍历每个 zone 的 forwarders，剔除命中 `local` 的 nameserver（避免显式回环）。
+5. 以 `origin -> zone_handlers[]` 方式 `upsert` 到 Hickory `Catalog`。
+
+### 10.3 `DnsServer::run`：三路热重载
+
+`DnsServer` 使用 3 个异步循环处理不同脏标记：
+
+- `reload_catalog`：`DynamicCatalog::replace(...)` 原子替换目录。
+- `reload_addresses`：更新劫持地址集合，并尝试下发系统 DNS。
+- `reload_listeners`：重建 `ServerFuture` 的 UDP/TCP 监听 socket。
+
+每路失败都会重新 `mark()` 自己，避免瞬时错误导致永久失效。
+
+### 10.4 listener/address 的行为边界
+
+- `listeners`：真正 bind 的服务地址；单个地址 bind 失败会打印错误并跳过，不导致整体停机。
+- `addresses`：仅用于劫持匹配，不需要 bind；可用于 `no_tun=false` 下的虚拟 DNS 入口。
+- `addresses` 与 `listeners` 分离，符合 `plan.md` 中“hijack 地址不等于监听地址”的设计。
+
+### 10.5 NIC 数据面处理（`NicPacketFilter`）
+
+处理链：
+
+1. `handle_ip_packet()` 解析 IPv4 头并检查目标 IP 是否属于 hijack 地址集合。
+2. UDP 分支：
+   - `MessageRequest::from_bytes` 解包 DNS 查询。
+   - 交给 `catalog.handle_request(...)` 获取响应。
+   - 回填 payload，修正 UDP/IP 长度与 checksum。
+3. ICMP 分支：
+   - EchoRequest 改写为 EchoReply。
+4. 统一收尾：交换 src/dst IP，并把包路由回本机 `peer_id`。
+
+该路径让 DNS 响应无需经过用户态 socket recv/send，直接在 packet pipeline 内完成。
+
+---
+
+## 11. Zone 组装与权威链详解（`zone.rs`）
+
+### 11.1 `Zone` 运行时模型
+
+`Zone` 包含：
+
+- `id: Uuid`（来源于配置/网络数据）
+- `origin: LowerName`
+- `records: BTreeMap<RrKey, RecordSet>`
+- `forward: Option<ForwardConfig>`
+
+`PartialEq` 对 `forward` 使用自定义比较（只比较 nameserver 序列），避免与无关字段耦合。
+
+### 11.2 反序列化与校验（`TryFrom<&ZoneData>`）
+
+转换过程：
+
+1. 必须有 `id`，否则报错。
+2. 用 Hickory `Parser` 解析 zone 文本（origin + RR）。
+3. 把 `forwarders` URL 转成 `NameServerAddr`，为空则 `forward=None`。
+
+这确保网络收到的 `ZoneData` 能直接映射成可执行 zone_handler。
+
+### 11.3 ZoneHandler 构建策略
+
+- `create_memory_zone_handler()`：仅当 records 非空时创建 `InMemoryZoneHandler`。
+- `create_forward_zone_handler()`：仅当 forward 非空时创建 `ForwardZoneHandler`。
+
+因此允许 3 种 zone 形态：
+
+1. 纯记录（权威回答）
+2. 纯转发（forward-only）
+3. 记录 + 转发（链式）
+
+### 11.4 `ZoneGroup` 与同源链式行为
+
+- `ZoneGroup::into_groups()` 按 `origin` 分组。
+- `iter_zone_handlers()` 对每个 zone 按顺序产出：先 memory，再 forward。
+- `DnsNodeMgr::catalog()` 把同 origin 的多个 zone zone_handler 以数组形式 `upsert`。
+
+结果是同 origin 下可自然形成 ChainedZoneHandler，不做“硬合并单 Zone”，与 `plan.md` 一致。
+
+### 11.5 `Zone::system()` 的作用边界
+
+`Zone::system()` 读取系统 resolver 作为 root zone forwarders。  
+在当前文档范围内仅关注它在 catalog 聚合中的语义：**兜底递归出口**。
+
+---
+
+## 12. 文档后续范围
+
+后续若继续扩写，将集中在以下主题（不再展开 `system/*`）：
+
+1. 策略执行链路补齐：`import/recursive/acl` 如何从配置走到查询路径。
+2. 测试矩阵梳理：单元测试、集成测试与故障注入测试的覆盖面。
+3. CLI 状态输出：如何观测 node/server 角色、snapshot digest、zone 来源与健康状态。
+
+---
+
+## 13. 策略执行链路现状与缺口
+
+本节专门回答一个容易误解的问题：**配置里有策略字段，不等于运行时已经完全执行**。
+
+### 13.1 已生效的策略相关行为
+
+当前代码中，和策略直接相关且已生效的路径主要有一条：
+
+- `GlobalCtx::dns_export_config()` 在导出 zones 时仅检查：
+  - `zone.policy.export.is_some()`
+
+也就是说，当前“导出/不导出”是可工作的，但粒度仍偏粗。
+
+### 13.2 已建模但尚未完整落地的策略字段
+
+以下字段在 `config/policy.rs` 已定义，但执行链路尚未完全打通：
+
+- `import.whitelist / import.blacklist`
+- `import.disabled`
+- `import.recursive`
+- `export` 内更细粒度 ACL
+
+从调用路径看：
+
+- `DnsPeerMgr::snapshot()` 只做本地 + 远端 zones 拼接，不做 import/export ACL 过滤。
+- `DnsNodeMgr::collect_zones()` 只做聚合与回环剔除，不做来源级策略裁剪。
+- `DnsServer::handle_ip_packet()` 是纯查询执行，不做请求来源与策略绑定。
+
+### 13.3 代码中的明确信号（TODO）
+
+当前有两个关键 TODO 信号：
+
+- `dns_export_config()` 里标注了 `TODO: check policies of parent zones`。
+- `policy.rs` 中 `AclPolicy`、`recursive` 旁边保留了 TODO 注释。
+
+这说明作者已经把策略模型前置到配置层，但执行面仍属于“进行中”。
+
+### 13.4 文档使用建议（给维护者）
+
+在策略彻底落地前，建议把语义按两层理解：
+
+1. **已可依赖**：`zone.policy.export.is_some()` 控制是否导出。
+2. **暂不可依赖**：import/export ACL、recursive、disabled 的全链路行为。
+
+---
+
+## 14. 测试体系与覆盖面
+
+本模块测试不是集中在一个文件，而是“按组件就地内嵌”。
+
+### 14.1 测试分布
+
+- `dns/tests.rs`：测试基建与辅助函数（构造环境、启动 `DnsNode`、DNS 查询断言工具）。
+- `dns/server.rs`：数据面与 server 行为主测试集。
+- `dns/node_mgr.rs`：聚合 catalog 的基本可用性测试。
+- `dns/zone.rs`：配置解析、记录转换、zone_handler 装配测试。
+
+> 说明：`system/*` 也有测试，但本轮文档按约定不展开。
+
+### 14.2 `server.rs` 覆盖要点
+
+`server.rs` 的测试集中验证了以下核心行为：
+
+- `DynamicCatalog::replace()` 可安全替换。
+- hijack 判定：`is_hijacked_ip` / `is_hijacked_addr`。
+- ICMP 改写：EchoRequest -> EchoReply。
+- UDP DNS 包内联处理：解析请求、生成应答、回填 payload。
+- 一个基础端到端路径：真实 UDP listener + Hickory client 查询。
+
+这些测试对应模块里最复杂、最容易回归的包处理逻辑。
+
+### 14.3 `node_mgr.rs` 覆盖要点
+
+`node_mgr.rs` 的测试重点是：
+
+- 人工插入节点 zone 后，`catalog()` 能查到预期记录。
+
+它验证了“快照聚合 -> Catalog 查询可用”的最小闭环，但尚未覆盖复杂心跳时序、TTL 过期后的清理行为。
+
+### 14.4 `zone.rs` 覆盖要点
+
+`zone.rs` 的测试覆盖了：
+
+- TOML `DnsConfig` 解析。
+- `ZoneConfig -> ZoneData -> Zone` 转换链。
+- record 解析/TTL 基本行为。
+- memory/forward zone_handler 构建，以及通过 server 查询验证。
+
+该测试更多是“模型与解析正确性”，不是策略执行链路完整验证。
+
+### 14.5 当前测试缺口
+
+结合 `plan.md` 与现有测试，仍建议补充：
+
+- `DnsNode` 心跳 + resync + 重选举的并发时序测试。
+- `DnsNodeMgr` TTL 过期淘汰与脏标记联动测试。
+- 策略字段（import/export ACL、recursive）的行为测试。
+- 多 peer、同 origin 多 zone 的优先级/去重回归测试。
+
+---
+
+## 15. CLI 与可观测性现状
+
+### 15.1 CLI 现状
+
+从当前代码看，`easytier/src/easytier-cli.rs` 没有 DNS 专用子命令。  
+因此“查看 DNS 子系统状态”主要依赖日志与通用状态接口，而非专门 CLI 面板。
+
+### 15.2 日志观测点（已存在）
+
+`dns` 子系统已经布置了较多 `tracing` 埋点，关键入口包括：
+
+- `DnsNode election loop`
+- `DnsNode main loop`
+- `DnsServer main loop`
+- `DnsNodeMgr::heartbeat`（含来源 id 与 snapshot 信息）
+
+可用于定位：
+
+- 当前实例是否赢得选举。
+- 心跳是否失败、是否触发 `resync`。
+- catalog/addresses/listeners 是否持续重载失败。
+
+### 15.3 当前可观测性短板
+
+- 缺少 DNS 专项 CLI 展示：
+  - 本机角色（Node/Server）
+  - 当前 snapshot digest
+  - zone 来源与数量
+  - 监听地址与 hijack 地址状态
+- 缺少结构化指标（metrics），目前偏日志驱动排障。
+
+### 15.4 建议的最小可观测面
+
+后续若补 CLI，可先实现一个最小 DNS 状态视图：
+
+1. 角色与选举状态（是否持有 `DNS_SERVER_RPC_ADDR`）。
+2. 最近心跳时间、`resync` 次数。
+3. 已装载 zone 数量（按本地/远端分组）。
+4. listeners 与 addresses 当前集合。
+
+该视图不改变数据面行为，但能显著降低线上排障成本。

easytier/src/dns/server.rs

@@ -0,0 +1,570 @@
+use crate::common::global_ctx::ArcGlobalCtx;
+use crate::dns::node_mgr::DnsNodeMgr;
+use crate::dns::system;
+use crate::dns::utils::addr::NameServerAddr;
+#[cfg(feature = "tun")]
+use crate::instance::instance::{ArcNicCtx, NicCtx};
+use crate::peers::peer_manager::PeerManager;
+use crate::proto::dns::DnsNodeMgrRpcServer;
+use crate::proto::rpc_impl::standalone::StandAloneServer;
+use crate::tunnel::common::bind;
+use crate::tunnel::tcp::TcpTunnelListener;
+use crate::utils::task::CancellableTask;
+use anyhow::Context;
+use derivative::Derivative;
+use guarden::guarded;
+use hickory_net::runtime::Time;
+use hickory_net::xfer::Protocol;
+use hickory_server::{
+    Server,
+    server::{Request, RequestHandler, ResponseHandler, ResponseInfo},
+    zone_handler::Catalog,
+};
+use itertools::chain;
+use parking_lot::RwLock;
+use std::collections::HashSet;
+use std::net::IpAddr;
+use std::{sync::Arc, time::Duration};
+use tokio::sync::Mutex;
+use tokio_util::sync::CancellationToken;
+use tracing::{Instrument, instrument};
+
+#[derive(Clone)]
+struct DynamicCatalog {
+    inner: Arc<tokio::sync::RwLock<Catalog>>,
+}
+
+impl DynamicCatalog {
+    fn new() -> Self {
+        Self {
+            inner: Arc::new(tokio::sync::RwLock::new(Catalog::new())),
+        }
+    }
+
+    async fn replace(&self, new: Catalog) {
+        *self.inner.write().await = new;
+    }
+}
+
+#[async_trait::async_trait]
+impl RequestHandler for DynamicCatalog {
+    async fn handle_request<R: ResponseHandler, T: Time>(
+        &self,
+        request: &Request,
+        response_handle: R,
+    ) -> ResponseInfo {
+        self.inner
+            .read()
+            .await
+            .handle_request::<_, T>(request, response_handle)
+            .await
+    }
+}
+
+#[derive(Derivative)]
+#[derivative(Debug)]
+pub struct DnsServer {
+    mgr: Arc<DnsNodeMgr>,
+
+    #[cfg(feature = "tun")]
+    nic_ctx: ArcNicCtx, // TODO: REMOVE THIS
+
+    peer_mgr: Arc<PeerManager>,
+    global_ctx: ArcGlobalCtx,
+
+    #[derivative(Debug = "ignore")]
+    catalog: DynamicCatalog,
+    runtime: Mutex<Option<CancellableTask<()>>>,
+    bindings: RwLock<HashSet<NameServerAddr>>,
+
+    addresses: RwLock<HashSet<NameServerAddr>>,
+    listeners: RwLock<HashSet<NameServerAddr>>,
+}
+
+const DNS_SERVER_TCP_TIMEOUT: Duration = Duration::from_secs(5);
+const DNS_SERVER_TCP_BUFFER_SIZE: usize = 32;
+
+impl DnsServer {
+    pub fn new(
+        peer_mgr: Arc<PeerManager>,
+        global_ctx: ArcGlobalCtx,
+        #[cfg(feature = "tun")] nic_ctx: ArcNicCtx, // TODO: REMOVE THIS
+    ) -> Self {
+        Self {
+            mgr: Arc::new(DnsNodeMgr::new()),
+            #[cfg(feature = "tun")]
+            nic_ctx,
+            peer_mgr,
+            global_ctx,
+            catalog: DynamicCatalog::new(),
+            runtime: Default::default(),
+            bindings: Default::default(),
+            listeners: Default::default(),
+            addresses: Default::default(),
+        }
+    }
+
+    pub fn register(&self, rpc: &StandAloneServer<TcpTunnelListener>) {
+        rpc.registry()
+            .register(DnsNodeMgrRpcServer::new_arc(self.mgr.clone()), "");
+    }
+
+    #[cfg(feature = "tun")]
+    async fn update_system(&self, nameservers: &HashSet<NameServerAddr>) -> anyhow::Result<()> {
+        let nic_ctx = self.nic_ctx.lock().await;
+        let nic_ctx = nic_ctx
+            .as_ref()
+            .and_then(|nic_ctx| nic_ctx.downcast_ref::<NicCtx>())
+            .with_context(|| "failed to get NicCtx")?;
+        let ifname = nic_ctx
+            .ifname()
+            .await
+            .with_context(|| "failed to get interface name from NicCtx")?;
+        let system = system::get(&ifname)?.with_context(|| "failed to get system configurator")?;
+        let config = self.global_ctx.config.get_dns();
+        let domain = vec![config.domain.to_string()];
+        system.set_dns(&system::SystemConfig {
+            nameservers: nameservers
+                .iter()
+                .filter_map(|a| {
+                    (a.protocol == Protocol::Udp && a.addr.port() == 53)
+                        .then_some(a.addr.ip().to_string())
+                })
+                .collect(),
+            search_domains: domain.clone(),
+            match_domains: domain
+                .into_iter()
+                .chain(config.zones.iter().map(|z| z.origin.to_string()))
+                .collect(),
+        })?;
+        Ok(())
+    }
+
+    #[instrument(skip_all)]
+    async fn rebind(&self) -> anyhow::Result<bool> {
+        let Ok(mut runtime) = self.runtime.try_lock() else {
+            return Ok(false);
+        };
+
+        let mut bindings = {
+            let current = self.bindings.read();
+            let bindings = chain(
+                self.addresses.read().iter().cloned(),
+                self.listeners.read().iter().cloned(),
+            )
+            .collect();
+            if *current == bindings {
+                tracing::info!("bindings unchanged, no need to rebind");
+                return Ok(false);
+            }
+            bindings
+        };
+
+        if let Some(runtime) = runtime.take() {
+            runtime.stop(None).await?;
+        }
+
+        let mut server = Server::new(self.catalog.clone());
+
+        bindings.retain(|binding| {
+            let addr = binding.addr;
+            tracing::info!(?addr, "binding");
+            match binding.protocol {
+                Protocol::Tcp => bind().addr(addr).call().map(|s| {
+                    server.register_listener(s, DNS_SERVER_TCP_TIMEOUT, DNS_SERVER_TCP_BUFFER_SIZE)
+                }),
+                Protocol::Udp => bind().addr(addr).call().map(|s| server.register_socket(s)),
+                _ => unimplemented!(),
+            }
+            .inspect_err(|error| tracing::error!(?addr, ?error, "failed to bind"))
+            .is_ok()
+        });
+
+        let token = server.shutdown_token().clone();
+        let handle = tokio::spawn(
+            async move {
+                server
+                    .block_until_done()
+                    .await
+                    .unwrap_or_else(|e| tracing::error!("DNS server exited with error: {:?}", e));
+            }
+            .instrument(tracing::info_span!("DNS server backend runtime")),
+        );
+
+        *runtime = Some(CancellableTask::with_handle(token, handle));
+
+        #[cfg(feature = "tun")]
+        if let Err(error) = self.update_system(&bindings).await {
+            tracing::error!(?error, "failed to update system DNS settings");
+        }
+
+        *self.bindings.write() = bindings;
+
+        Ok(true)
+    }
+
+    #[instrument(skip_all)]
+    async fn reload_addresses(&self) -> anyhow::Result<()> {
+        let addresses = self.mgr.iter_addresses().collect();
+
+        let removed = {
+            let current = self.addresses.read();
+            if *current == addresses {
+                tracing::info!("addresses unchanged, no need to reload");
+                return Ok(());
+            }
+            current
+                .difference(&addresses)
+                .cloned()
+                .collect::<HashSet<_>>()
+        };
+        tracing::info!(?addresses, "reloading");
+
+        #[cfg(feature = "tun")]
+        {
+            let nic_ctx = self.nic_ctx.lock().await;
+            if let Some(nic_ctx) = nic_ctx
+                .as_ref()
+                .and_then(|nic_ctx| nic_ctx.downcast_ref::<NicCtx>())
+            {
+                for addr in &addresses {
+                    let ip = addr.addr.ip();
+                    if let Err(error) = match ip {
+                        IpAddr::V4(ipv4) => nic_ctx.add_ipv4_to_tun_device(ipv4.into()).await,
+                        IpAddr::V6(ipv6) => nic_ctx.add_ipv6_to_tun_device(ipv6.into()).await,
+                    } {
+                        tracing::error!(?addr, ?error, "failed to add address to tun device");
+                    }
+                }
+
+                for addr in removed {
+                    let ip = addr.addr.ip();
+                    if let Err(error) = match ip {
+                        IpAddr::V4(ipv4) => nic_ctx.remove_ipv4_from_tun_device(ipv4.into()).await,
+                        IpAddr::V6(ipv6) => nic_ctx.remove_ipv6_from_tun_device(ipv6.into()).await,
+                    } {
+                        tracing::error!(?addr, ?error, "failed to remove address from tun device");
+                    }
+                }
+            }
+        }
+
+        *self.addresses.write() = addresses;
+
+        self.rebind().await?;
+
+        Ok(())
+    }
+
+    #[instrument(skip_all)]
+    async fn reload_listeners(&self) -> anyhow::Result<()> {
+        let listeners = self.mgr.iter_listeners().collect();
+
+        if *self.listeners.read() == listeners {
+            tracing::info!("listeners unchanged, no need to reload");
+            return Ok(());
+        }
+        tracing::info!(?listeners, "reloading");
+
+        *self.listeners.write() = listeners;
+
+        self.rebind().await?;
+
+        Ok(())
+    }
+
+    #[instrument(skip_all, name = "DnsServer main loop")]
+    pub async fn run(&self, token: CancellationToken) {
+        let dirty = &self.mgr.dirty;
+        let runtime = None::<CancellableTask<()>>;
+
+        #[cfg(feature = "tun")]
+        guarded! {
+            system_guard => [
+                nic_ctx = self.nic_ctx.clone(),
+            ]
+            async move {
+                if let Some(nic_ctx) = nic_ctx
+                    .lock()
+                    .await
+                    .as_ref()
+                    .and_then(|nic_ctx| nic_ctx.downcast_ref::<NicCtx>())
+                    && let Some(system) = nic_ctx
+                        .ifname()
+                        .await
+                        .and_then(|ifname| system::get(&ifname).ok())
+                        .flatten()
+                {
+                    let _ = system.clean();
+                }
+            }
+        }
+
+        guarded! {
+            runtime_guard => [
+                mut runtime,
+            ]
+            async move {
+                if let Some(runtime) = runtime.take() {
+                    let _ = runtime.stop(Some(Duration::from_secs(1))).await;
+                }
+            }
+        }
+
+        let reload_catalog = async {
+            loop {
+                dirty.catalog.wait().await;
+                if dirty.catalog.reset() {
+                    self.catalog.replace(self.mgr.catalog()).await;
+                }
+                tokio::time::sleep(Duration::from_secs(1)).await;
+            }
+        };
+
+        let reload_addresses = async {
+            loop {
+                dirty.addresses.wait().await;
+                if dirty.addresses.reset()
+                    && let Err(error) = self.reload_addresses().await
+                {
+                    tracing::error!(?error, "failed to reload addresses");
+                    dirty.addresses.mark();
+                }
+                tokio::time::sleep(Duration::from_secs(1)).await;
+            }
+        };
+
+        let reload_listeners = async {
+            loop {
+                dirty.listeners.wait().await;
+                if dirty.listeners.reset()
+                    && let Err(error) = self.reload_listeners().await
+                {
+                    tracing::error!(?error, "failed to reload listeners");
+                    dirty.listeners.mark();
+                }
+                tokio::time::sleep(Duration::from_secs(1)).await;
+            }
+        };
+
+        tokio::select!(
+            _ = token.cancelled() => {
+                tracing::info!("DnsServer received shutdown signal, exiting server loop");
+            }
+
+            _ = reload_catalog => {},
+            _ = reload_addresses => {},
+            _ = reload_listeners => {},
+        );
+
+        #[cfg(feature = "tun")]
+        system_guard.trigger().await;
+        runtime_guard.trigger().await;
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::peers::tests::create_mock_peer_manager;
+    use hickory_net::client::{Client, ClientHandle};
+    use hickory_net::runtime::TokioRuntimeProvider;
+    use hickory_net::udp::UdpClientStream;
+    use hickory_proto::op::{Message, MessageType, OpCode, Query};
+    use hickory_proto::rr::{DNSClass, Name, RData, Record, RecordType, rdata};
+    use hickory_proto::serialize::binary::BinEncodable;
+    use hickory_server::store::in_memory::InMemoryZoneHandler;
+    use hickory_server::zone_handler::ZoneType;
+    use hickory_server::zone_handler::{AxfrPolicy, Catalog};
+    use pnet::packet::icmp::{IcmpTypes, MutableIcmpPacket};
+    use pnet::packet::ipv4::MutableIpv4Packet;
+    use pnet::packet::udp::MutableUdpPacket;
+    use pnet::packet::{MutablePacket, icmp, ipv4, udp};
+    use std::net::Ipv4Addr;
+    use std::str::FromStr;
+    use std::time::Duration;
+    use tokio::time::sleep;
+
+    /// Build a `Catalog` containing a single A record: `test.example.com -> 1.2.3.4`.
+    fn build_test_catalog() -> Catalog {
+        let origin = Name::from_str("example.com.").unwrap();
+        let mut zone_handler = InMemoryZoneHandler::<TokioRuntimeProvider>::empty(
+            origin.clone(),
+            ZoneType::Primary,
+            AxfrPolicy::default(),
+        );
+
+        let record = Record::from_rdata(
+            Name::from_str("test.example.com.").unwrap(),
+            60,
+            RData::A(rdata::a::A(Ipv4Addr::new(1, 2, 3, 4))),
+        );
+        let rr_key =
+            hickory_proto::rr::RrKey::new(record.name.clone().into(), record.record_type());
+        let mut rr_set =
+            hickory_proto::rr::RecordSet::new(record.name.clone(), record.record_type(), 0);
+        rr_set.insert(record, 0);
+        zone_handler
+            .records_get_mut()
+            .insert(rr_key, Arc::new(rr_set));
+
+        let mut catalog = Catalog::new();
+        catalog.upsert(
+            origin.into(),
+            vec![Arc::new(zone_handler) as Arc<dyn hickory_server::zone_handler::ZoneHandler>],
+        );
+        catalog
+    }
+
+    /// Create a test `DnsServer` with `create_mock_peer_manager()`.
+    async fn create_test_server() -> Arc<DnsServer> {
+        let peer_mgr = create_mock_peer_manager().await;
+        let global_ctx = peer_mgr.get_global_ctx();
+        Arc::new(DnsServer::new(
+            peer_mgr,
+            global_ctx,
+            #[cfg(feature = "tun")]
+            ArcNicCtx::default(),
+        ))
+    }
+
+    /// Build a raw IPv4 packet (as `Vec<u8>`) carrying the given L4 payload bytes.
+    /// `protocol` selects ICMP / UDP etc.
+    fn build_ipv4_packet(
+        src: Ipv4Addr,
+        dst: Ipv4Addr,
+        protocol: pnet::packet::ip::IpNextHeaderProtocol,
+        l4_payload: &[u8],
+    ) -> Vec<u8> {
+        let ip_header_len = 20usize;
+        let total_len = ip_header_len + l4_payload.len();
+        let mut buf = vec![0u8; total_len];
+        {
+            let mut ip = MutableIpv4Packet::new(&mut buf).unwrap();
+            ip.set_version(4);
+            ip.set_header_length(5); // 20 bytes
+            ip.set_total_length(total_len as u16);
+            ip.set_ttl(64);
+            ip.set_next_level_protocol(protocol);
+            ip.set_source(src);
+            ip.set_destination(dst);
+            ip.payload_mut().copy_from_slice(l4_payload);
+            ip.set_checksum(ipv4::checksum(&ip.to_immutable()));
+        }
+        buf
+    }
+
+    /// Build ICMP Echo Request payload (8 bytes minimum).
+    fn build_icmp_echo_request() -> Vec<u8> {
+        let mut buf = vec![0u8; 8];
+        {
+            let mut icmp_pkt = MutableIcmpPacket::new(&mut buf).unwrap();
+            icmp_pkt.set_icmp_type(IcmpTypes::EchoRequest);
+            icmp_pkt.set_icmp_code(icmp::IcmpCode::new(0));
+            icmp_pkt.set_checksum(icmp::checksum(&icmp_pkt.to_immutable()));
+        }
+        buf
+    }
+
+    /// Build a minimal DNS query message for `name` and encode it to bytes.
+    fn build_dns_query_bytes(name: &str) -> Vec<u8> {
+        let mut msg = Message::new(0x1234, MessageType::Query, OpCode::Query);
+        msg.metadata.recursion_desired = true;
+        let mut query = Query::new();
+        query.set_name(Name::from_str(name).unwrap());
+        query.set_query_type(RecordType::A);
+        query.set_query_class(DNSClass::IN);
+        msg.add_query(query);
+        msg.to_bytes().unwrap().to_vec()
+    }
+
+    /// Build a UDP packet carrying `payload`, with given src/dst ports.
+    fn build_udp_packet(
+        src_port: u16,
+        dst_port: u16,
+        payload: &[u8],
+        src_ip: Ipv4Addr,
+        dst_ip: Ipv4Addr,
+    ) -> Vec<u8> {
+        let udp_len = 8 + payload.len();
+        let mut buf = vec![0u8; udp_len];
+        {
+            let mut udp_pkt = MutableUdpPacket::new(&mut buf).unwrap();
+            udp_pkt.set_source(src_port);
+            udp_pkt.set_destination(dst_port);
+            udp_pkt.set_length(udp_len as u16);
+            udp_pkt.payload_mut().copy_from_slice(payload);
+            udp_pkt.set_checksum(udp::ipv4_checksum(
+                &udp_pkt.to_immutable(),
+                &src_ip,
+                &dst_ip,
+            ));
+        }
+        buf
+    }
+
+    async fn wait_until(mut f: impl FnMut() -> bool) {
+        for _ in 0..80 {
+            if f() {
+                return;
+            }
+            sleep(Duration::from_millis(50)).await;
+        }
+        panic!("condition not met in time");
+    }
+
+    // ─── Tests ───────────────────────────────────────────────────────────
+
+    /// Full end-to-end test: start a real DNS UDP listener via `ServerFuture`,
+    /// send a query with a `hickory_client`, and verify the response.
+    #[tokio::test]
+    async fn should_resolve_record_via_real_udp_listener() {
+        use hickory_server::Server;
+        use tokio::net::UdpSocket;
+        use tokio::time::timeout;
+
+        // Build a catalog with test.example.com -> 1.2.3.4.
+        let catalog = build_test_catalog();
+
+        // Bind to a random port.
+        let socket = UdpSocket::bind("127.0.0.1:0").await.unwrap();
+        let addr = socket.local_addr().unwrap();
+
+        let mut server = Server::new(catalog);
+        server.register_socket(socket);
+
+        let shutdown_token = server.shutdown_token().clone();
+        tokio::spawn(async move {
+            server.block_until_done().await.ok();
+        });
+
+        // Send a real DNS query using hickory_client.
+        let stream = UdpClientStream::builder(addr, TokioRuntimeProvider::default()).build();
+        let (mut client, bg) = Client::<TokioRuntimeProvider>::from_sender(stream);
+
+        tokio::spawn(bg);
+
+        let response = timeout(
+            Duration::from_secs(2),
+            client.query(
+                Name::from_str("test.example.com.").unwrap(),
+                DNSClass::IN,
+                RecordType::A,
+            ),
+        )
+        .await
+        .expect("query timeout")
+        .expect("query failed");
+
+        assert!(!response.answers.is_empty(), "should get answers");
+        let a_record = &response.answers[0];
+        if let RData::A(a) = a_record.data {
+            assert_eq!(a.0, Ipv4Addr::new(1, 2, 3, 4));
+        } else {
+            panic!("expected A record, got {:?}", a_record.data);
+        }
+
+        // Shutdown the server.
+        shutdown_token.cancel();
+    }
+}

easytier/src/dns/system/macos.rs

@@ -6,13 +6,13 @@ use std::{
     path::Path,
 };
 
-use super::{OSConfig, SystemConfig};
+use super::{SystemConfig, SystemConfigurator};
 
 const MAC_RESOLVER_FILE_HEADER: &str = "# Added by easytier\n";
 const ETC_RESOLVER: &str = "/etc/resolver";
 const ETC_RESOLV_CONF: &str = "/etc/resolv.conf";
 
-#[derive(Default)]
+#[derive(Clone, Default)]
 pub struct DarwinConfigurator {}
 impl DarwinConfigurator {
     pub fn new() -> Self {
@@ -27,7 +27,7 @@ impl DarwinConfigurator {
         true
     }
 
-    pub fn do_set_dns(&self, cfg: &OSConfig) -> io::Result<()> {
+    pub fn do_set_dns(&self, cfg: &SystemConfig) -> io::Result<()> {
         fs::create_dir_all(ETC_RESOLVER)?;
         let mut keep = HashSet::new();
 
@@ -104,12 +104,12 @@ impl DarwinConfigurator {
     }
 }
 
-impl SystemConfig for DarwinConfigurator {
-    fn set_dns(&self, cfg: &OSConfig) -> io::Result<()> {
+impl SystemConfigurator for DarwinConfigurator {
+    fn set_dns(&self, cfg: &SystemConfig) -> io::Result<()> {
         self.do_set_dns(cfg)
     }
 
-    fn close(&self) -> io::Result<()> {
+    fn clean(&self) -> io::Result<()> {
         self.do_close()
     }
 }

easytier/src/dns/system/mod.rs

@@ -0,0 +1,42 @@
+#[allow(unused_imports)]
+use crate::utils::BoxExt;
+
+#[cfg(target_os = "linux")]
+pub mod linux;
+
+#[cfg(target_os = "windows")]
+pub mod windows;
+
+#[cfg(all(target_os = "macos", not(feature = "macos-ne")))]
+pub mod macos;
+
+#[derive(Default, Debug)]
+pub struct SystemConfig {
+    pub nameservers: Vec<String>,
+    pub search_domains: Vec<String>,
+    pub match_domains: Vec<String>,
+}
+
+pub trait SystemConfigurator: Send + Sync {
+    fn set_dns(&self, cfg: &SystemConfig) -> std::io::Result<()>;
+    fn clean(&self) -> std::io::Result<()>;
+}
+
+// TODO: move this to nic mod
+pub fn get(
+    #[allow(unused_variables)] interface: &str,
+) -> Result<Option<Box<dyn SystemConfigurator>>, anyhow::Error> {
+    cfg_select! {
+        target_os = "windows" => {
+            use crate::dns::system::windows::WindowsDNSManager;
+            Ok(Some(WindowsDNSManager::new(interface)?.boxed()))
+        }
+
+        all(target_os = "macos", not(feature = "macos-ne")) => {
+            use crate::dns::system::macos::DarwinConfigurator;
+            Ok(Some(DarwinConfigurator::new().boxed()))
+        }
+
+        _ => Ok(None)
+    }
+}

easytier/src/dns/system/windows.rs

@@ -6,7 +6,7 @@ use winreg::RegKey;
 
 use crate::common::ifcfg::RegistryManager;
 
-use super::{OSConfig, SystemConfig};
+use super::{SystemConfig, SystemConfigurator};
 
 pub fn is_windows_10_or_better() -> io::Result<bool> {
     let hklm = winreg::enums::HKEY_LOCAL_MACHINE;
@@ -19,6 +19,7 @@ pub fn is_windows_10_or_better() -> io::Result<bool> {
 }
 
 // 假设 interface_guid 是你的网络接口 GUID
+#[derive(Clone)]
 pub struct InterfaceControl {
     interface_guid: String,
 }
@@ -125,6 +126,7 @@ impl InterfaceControl {
     }
 }
 
+#[derive(Clone)]
 pub struct WindowsDNSManager {
     tun_dev_name: String,
     interface_control: InterfaceControl,
@@ -146,8 +148,8 @@ impl WindowsDNSManager {
     }
 }
 
-impl SystemConfig for WindowsDNSManager {
-    fn set_dns(&self, cfg: &OSConfig) -> io::Result<()> {
+impl SystemConfigurator for WindowsDNSManager {
+    fn set_dns(&self, cfg: &SystemConfig) -> io::Result<()> {
         self.set_primary_dns(
             &cfg.nameservers
                 .iter()
@@ -158,44 +160,48 @@ impl SystemConfig for WindowsDNSManager {
         Ok(())
     }
 
-    fn close(&self) -> io::Result<()> {
+    fn clean(&self) -> io::Result<()> {
         Ok(())
     }
 }
 
-#[cfg(test)]
+#[cfg(all(test, target_os = "windows", feature = "magic-dns", feature = "tun"))]
 mod tests {
     use cidr::Ipv4Inet;
+    use std::net::IpAddr;
 
-    #[cfg(target_os = "windows")]
     #[tokio::test]
     async fn test_windows_set_primary_server() {
-        use std::{net::Ipv4Addr, str::FromStr as _, time::Duration};
+        use std::{str::FromStr as _, sync::Arc, time::Duration};
 
-        use tokio_util::sync::CancellationToken;
-
-        use crate::instance::dns_server::{
-            runner::DnsRunner,
-            tests::{check_dns_record, prepare_env},
+        use crate::dns::{
+            config::DNS_DEFAULT_ADDRESSES,
+            tests::{prepare_env, start_dns_node},
         };
+        use crate::instance::proxy_cidrs_monitor::ProxyCidrsMonitor;
+        use crate::instance::virtual_nic::NicCtx;
+        use crate::peers::peer_manager::PeerManager;
 
         let tun_ip = Ipv4Inet::from_str("10.144.144.10/24").unwrap();
-        let (peer_mgr, virtual_nic) = prepare_env("test1", tun_ip).await;
+        let (peer_mgr, virtual_nic): (Arc<PeerManager>, NicCtx) =
+            prepare_env("test1", tun_ip).await;
         let tun_name = virtual_nic.ifname().await.unwrap();
 
+        // prepare_env does not run full Instance::run, so start the monitor explicitly in test.
+        let _monitor = ProxyCidrsMonitor::new(peer_mgr.clone(), peer_mgr.get_global_ctx()).start();
+
+        let mut dns_node = start_dns_node(peer_mgr, virtual_nic);
+
         println!("dev_name: {}", tun_name);
-        let fake_ip = Ipv4Addr::from_str("100.100.100.101").unwrap();
-        let mut dns_runner = DnsRunner::new(peer_mgr, Some(tun_name.clone()), tun_ip, fake_ip);
+        let fake_ip = match DNS_DEFAULT_ADDRESSES[0].addr.ip() {
+            IpAddr::V4(ip) => ip,
+            IpAddr::V6(ip) => panic!("unexpected ipv6 default dns address in test: {ip}"),
+        };
 
-        let cancel_token = CancellationToken::new();
-        let cancel_token_clone = cancel_token.clone();
-        let t = tokio::spawn(async move {
-            dns_runner.run(cancel_token_clone).await;
-        });
-
-        // windows is slow to add a ip address, wait for a longer time for dns server ready ,with ping
+        // Windows may take a while to attach the test IP; wait until ping succeeds.
         let now = std::time::Instant::now();
-        while now.elapsed() < Duration::from_secs(15) {
+        let mut ping_ready = false;
+        while now.elapsed() < Duration::from_secs(5) {
             tokio::time::sleep(Duration::from_secs(1)).await;
             if let Ok(o) = tokio::process::Command::new("ping")
                 .arg("-n")
@@ -207,11 +213,49 @@ mod tests {
                 .await
                 && o.status.success()
             {
+                ping_ready = true;
                 break;
             }
         }
+        if !ping_ready {
+            tracing::warn!(
+                "dns test endpoint {} did not respond to ping in time; continue with dns checks",
+                fake_ip
+            );
+        }
 
-        check_dns_record(&fake_ip, "test1.et.net", "10.144.144.10").await;
+        // First verify the DNS node can answer queries when explicitly targeting it.
+        let direct_lookup_deadline = std::time::Instant::now() + Duration::from_secs(3);
+        let mut direct_lookup_ok = false;
+        let mut last_direct_output = String::new();
+
+        while std::time::Instant::now() < direct_lookup_deadline {
+            let result = tokio::time::timeout(
+                Duration::from_secs(3),
+                tokio::process::Command::new("nslookup")
+                    .arg("test1.et.net")
+                    .arg(fake_ip.to_string())
+                    .output(),
+            )
+            .await;
+
+            if let Ok(Ok(ret)) = result {
+                let output = String::from_utf8_lossy(&ret.stdout).to_string();
+                println!("direct nslookup output: {}", output);
+                if ret.status.success() && output.contains("10.144.144.10") {
+                    direct_lookup_ok = true;
+                    break;
+                }
+                last_direct_output = output;
+            }
+
+            tokio::time::sleep(Duration::from_millis(500)).await;
+        }
+
+        assert!(
+            direct_lookup_ok,
+            "direct nslookup against dns server did not return expected IP in time; last output: {last_direct_output}"
+        );
 
         let dns_mgr = super::WindowsDNSManager::new(&tun_name).unwrap();
         println!("dev_name: {}", tun_name);
@@ -219,27 +263,44 @@ mod tests {
 
         dns_mgr
             .interface_control
-            .set_primary_dns(
-                &["100.100.100.101".parse().unwrap()],
-                &[".et.net.".to_string()],
-            )
+            .set_primary_dns(&[fake_ip.into()], &[".et.net.".to_string()])
             .unwrap();
         dns_mgr.interface_control.flush_dns().unwrap();
 
         tracing::info!("check dns record with nslookup");
 
-        // nslookup should return 10.144.144.10
-        let ret = tokio::process::Command::new("nslookup")
-            .arg("test1.et.net")
-            .output()
-            .await
-            .expect("failed to execute process");
-        assert!(ret.status.success());
-        let output = String::from_utf8_lossy(&ret.stdout);
-        println!("nslookup output: {}", output);
-        assert!(output.contains("10.144.144.10"));
+        // nslookup should eventually return 10.144.144.10 after system DNS setting propagation.
+        let lookup_deadline = std::time::Instant::now() + Duration::from_secs(30);
+        let mut lookup_ok = false;
+        let mut last_output = String::new();
 
-        cancel_token.cancel();
-        let _ = t.await;
+        while std::time::Instant::now() < lookup_deadline {
+            let result = tokio::time::timeout(
+                Duration::from_secs(3),
+                tokio::process::Command::new("nslookup")
+                    .arg("test1.et.net")
+                    .output(),
+            )
+            .await;
+
+            if let Ok(Ok(ret)) = result {
+                let output = String::from_utf8_lossy(&ret.stdout).to_string();
+                println!("nslookup output: {}", output);
+                if ret.status.success() && output.contains("10.144.144.10") {
+                    lookup_ok = true;
+                    break;
+                }
+                last_output = output;
+            }
+
+            tokio::time::sleep(Duration::from_millis(500)).await;
+        }
+
+        assert!(
+            lookup_ok,
+            "nslookup did not return expected IP in time; last output: {last_output}"
+        );
+
+        dns_node.stop().await.unwrap();
     }
 }

easytier/src/dns/tests.rs

@@ -0,0 +1,984 @@
+#![cfg(all(feature = "magic-dns", feature = "tun"))]
+
+use std::net::{Ipv4Addr, SocketAddr, SocketAddrV4};
+use std::str::FromStr as _;
+use std::sync::Arc;
+use std::time::{Duration, Instant};
+
+use crate::common::config::TomlConfigLoader;
+use crate::common::global_ctx::GlobalCtx;
+use crate::common::global_ctx::tests::get_mock_global_ctx;
+use crate::connector::udp_hole_punch::tests::replace_stun_info_collector;
+use crate::dns::node::DnsNode;
+use crate::dns::peer_mgr::DnsPeerMgr;
+use crate::instance::instance::ArcNicCtx;
+use crate::instance::virtual_nic::NicCtx;
+use crate::peers::create_packet_recv_chan;
+use crate::peers::peer_manager::{PeerManager, RouteAlgoType};
+use crate::peers::tests::{connect_peer_manager, wait_route_appear};
+use crate::proto::common::{NatType, Url};
+use crate::proto::dns::{DnsSnapshot, HeartbeatRequest, ZoneData};
+use cidr::Ipv4Inet;
+use hickory_net::client::{Client, ClientHandle};
+use hickory_net::runtime::TokioRuntimeProvider;
+use hickory_net::udp::UdpClientStream;
+use hickory_net::xfer::Protocol;
+use hickory_proto::op::{Message, MessageType, OpCode, Query};
+use hickory_proto::rr;
+use hickory_proto::rr::{DNSClass, Name, RData, RecordType};
+use hickory_proto::serialize::binary::{BinEncodable, BinEncoder};
+use hickory_server::server::Request;
+use maplit::hashset;
+use tokio::sync::Notify;
+use uuid::Uuid;
+
+// TODO: move to system::tests
+pub async fn prepare_env(dns_name: &str, tun_ip: Ipv4Inet) -> (Arc<PeerManager>, NicCtx) {
+    prepare_env_with_tld_dns_zone(dns_name, tun_ip, None).await
+}
+
+pub async fn prepare_env_with_tld_dns_zone(
+    dns_name: &str,
+    tun_ip: Ipv4Inet,
+    tld_dns_zone: Option<&str>,
+) -> (Arc<PeerManager>, NicCtx) {
+    let ctx = get_mock_global_ctx();
+    ctx.set_hostname(dns_name.to_owned());
+    ctx.set_ipv4(Some(tun_ip));
+
+    let mut dns_config = ctx.config.get_dns().into_raw();
+    dns_config.name = Some(dns_name.parse().unwrap());
+    if let Some(zone) = tld_dns_zone {
+        dns_config.domain = Some(zone.parse().expect("invalid test dns zone"));
+    }
+    ctx.config.set_dns(dns_config.into());
+
+    let (s, r) = create_packet_recv_chan();
+    let peer_mgr = Arc::new(PeerManager::new(RouteAlgoType::Ospf, ctx, s));
+    peer_mgr.run().await.unwrap();
+    replace_stun_info_collector(peer_mgr.clone(), NatType::PortRestricted);
+
+    let r = Arc::new(tokio::sync::Mutex::new(r));
+    let mut virtual_nic = NicCtx::new(
+        peer_mgr.get_global_ctx(),
+        &peer_mgr,
+        r,
+        Arc::new(Notify::new()),
+    );
+    virtual_nic.run(Some(tun_ip), None).await.unwrap();
+
+    (peer_mgr, virtual_nic)
+}
+
+pub fn start_dns_node(peer_mgr: Arc<PeerManager>, virtual_nic: NicCtx) -> DnsNode {
+    let global_ctx = peer_mgr.get_global_ctx();
+    let nic_ctx: ArcNicCtx = Arc::new(tokio::sync::Mutex::new(Some(Box::new(virtual_nic))));
+
+    let mut node = DnsNode::new(peer_mgr, global_ctx, nic_ctx);
+    node.start();
+    node
+}
+
+pub fn start_dns_node_without_nic(peer_mgr: Arc<PeerManager>) -> DnsNode {
+    let global_ctx = peer_mgr.get_global_ctx();
+    let nic_ctx: ArcNicCtx = Arc::new(tokio::sync::Mutex::new(None));
+
+    let mut node = DnsNode::new(peer_mgr, global_ctx, nic_ctx);
+    node.start();
+    node
+}
+
+pub async fn prepare_env_from_config_str(config_str: &str) -> Arc<PeerManager> {
+    let config = TomlConfigLoader::new_from_str(config_str).expect("invalid test config");
+    let ctx = Arc::new(GlobalCtx::new(config));
+
+    let (s, _r) = create_packet_recv_chan();
+    let peer_mgr = Arc::new(PeerManager::new(RouteAlgoType::Ospf, ctx, s));
+    peer_mgr.run().await.unwrap();
+    replace_stun_info_collector(peer_mgr.clone(), NatType::PortRestricted);
+
+    peer_mgr
+}
+
+pub fn zone_data_a(origin: &str, record: &str) -> ZoneData {
+    zone_data_a_with_forwarders(origin, record, vec![])
+}
+
+pub fn zone_data_a_with_forwarders(origin: &str, record: &str, forwarders: Vec<&str>) -> ZoneData {
+    ZoneData::new(
+        &origin.parse().unwrap(),
+        60,
+        [format!("@ IN A {record}")],
+        forwarders.into_iter().map(|f| Url::from_str(f).unwrap()),
+        hashset! {},
+    )
+}
+
+pub fn dns_snapshot_with(
+    zones: Vec<ZoneData>,
+    addresses: Vec<&str>,
+    listeners: Vec<&str>,
+) -> DnsSnapshot {
+    DnsSnapshot {
+        zones,
+        addresses: addresses
+            .into_iter()
+            .map(|a| Url::from_str(a).expect("invalid address"))
+            .collect(),
+        listeners: listeners
+            .into_iter()
+            .map(|l| Url::from_str(l).expect("invalid listener"))
+            .collect(),
+    }
+}
+
+pub fn heartbeat_with_snapshot(id: Uuid, snapshot: DnsSnapshot) -> HeartbeatRequest {
+    let mut hb = HeartbeatRequest {
+        id: Some(id.into()),
+        ..Default::default()
+    };
+    hb.update(snapshot);
+    hb
+}
+
+fn find_free_udp_port() -> u16 {
+    std::net::UdpSocket::bind("127.0.0.1:0")
+        .expect("failed to bind temp udp socket")
+        .local_addr()
+        .expect("failed to get local addr")
+        .port()
+}
+
+pub async fn check_dns_record(fake_ip: &Ipv4Addr, domain: &str, expected_ip: &str) {
+    check_dns_record_at(SocketAddr::new((*fake_ip).into(), 53), domain, expected_ip).await;
+}
+
+pub async fn check_dns_record_at(server_addr: SocketAddr, domain: &str, expected_ip: &str) {
+    let expected = expected_ip.parse::<Ipv4Addr>().unwrap();
+    let name = Name::from_str(domain).unwrap();
+    let deadline = Instant::now() + Duration::from_secs(30);
+
+    loop {
+        let stream = UdpClientStream::builder(server_addr, TokioRuntimeProvider::default()).build();
+        let (mut client, background) = Client::<TokioRuntimeProvider>::from_sender(stream);
+        let background_task = tokio::spawn(background);
+
+        let query_result = tokio::time::timeout(
+            Duration::from_secs(2),
+            client.query(name.clone(), DNSClass::IN, RecordType::A),
+        )
+        .await;
+
+        background_task.abort();
+        let _ = background_task.await;
+
+        let attempt_err = match query_result {
+            Ok(Ok(response)) => {
+                if response.answers.len() == 1
+                    && let Some(resp) = response.answers.first()
+                    && matches!(resp.clone().data, RData::A(a) if a.0 == expected)
+                {
+                    return;
+                }
+                format!("unexpected response: {:?}", response.answers)
+            }
+            Ok(Err(e)) => {
+                format!("DNS query failed for domain '{domain}': {e}")
+            }
+            Err(_) => {
+                format!("DNS query timed out for domain '{domain}'")
+            }
+        };
+
+        if Instant::now() >= deadline {
+            panic!(
+                "DNS query failed unexpectedly for domain '{domain}' after retries: {attempt_err}"
+            );
+        }
+
+        tokio::time::sleep(Duration::from_millis(500)).await;
+    }
+}
+
+pub async fn check_dns_record_missing(fake_ip: &Ipv4Addr, domain: &str) {
+    check_dns_record_missing_at(SocketAddr::new((*fake_ip).into(), 53), domain).await;
+}
+
+pub async fn check_dns_record_missing_at(server_addr: SocketAddr, domain: &str) {
+    let deadline = Instant::now() + Duration::from_secs(30);
+
+    loop {
+        let stream = UdpClientStream::builder(server_addr, TokioRuntimeProvider::default()).build();
+        let (mut client, background) = Client::<TokioRuntimeProvider>::from_sender(stream);
+        let background_task = tokio::spawn(background);
+        let query_result = tokio::time::timeout(
+            Duration::from_secs(2),
+            client.query(Name::from_str(domain).unwrap(), DNSClass::IN, RecordType::A),
+        )
+        .await;
+        background_task.abort();
+        let _ = background_task.await;
+
+        let attempt_err = match query_result {
+            Ok(Ok(response)) => {
+                if response.answers.is_empty() {
+                    return;
+                }
+                format!("unexpected non-empty response: {:?}", response.answers)
+            }
+            Ok(Err(e)) => {
+                format!("DNS query for missing record failed for domain '{domain}': {e}")
+            }
+            Err(_) => {
+                format!("DNS query for missing record timed out for domain '{domain}'")
+            }
+        };
+
+        if Instant::now() >= deadline {
+            panic!(
+                "missing-record query failed unexpectedly for domain '{domain}' after retries: {attempt_err}"
+            );
+        }
+
+        tokio::time::sleep(Duration::from_millis(500)).await;
+    }
+}
+
+pub fn new_request(name: &str, rtype: RecordType) -> anyhow::Result<Request> {
+    let mut msg = Message::new(0, MessageType::Query, OpCode::Query);
+    msg.metadata.recursion_desired = true;
+    msg.add_query(Query::query(Name::from_ascii(name)?, rtype));
+
+    let mut request = Vec::new();
+    let mut encoder = BinEncoder::new(&mut request);
+    msg.emit(&mut encoder)?;
+
+    Ok(Request::from_bytes(
+        request,
+        SocketAddrV4::new(Ipv4Addr::LOCALHOST, 0).into(),
+        Protocol::Udp,
+    )?)
+}
+
+async fn wait_route_disappear(peer_mgr: Arc<PeerManager>, target_peer_id: u32) {
+    let deadline = Instant::now() + Duration::from_secs(10);
+    loop {
+        let has_route = peer_mgr
+            .list_routes()
+            .await
+            .iter()
+            .any(|r| r.peer_id == target_peer_id);
+        if !has_route {
+            return;
+        }
+
+        assert!(
+            Instant::now() < deadline,
+            "route to peer {} did not disappear in time",
+            target_peer_id
+        );
+        tokio::time::sleep(Duration::from_millis(100)).await;
+    }
+}
+
+async fn disconnect_all_peer_conns(a: Arc<PeerManager>, b: Arc<PeerManager>) {
+    if let Some(conns) = a.get_peer_map().list_peer_conns(b.my_peer_id()).await {
+        for conn in conns {
+            let conn_id = conn.conn_id.parse().expect("invalid conn id");
+            let _ = a.close_peer_conn(b.my_peer_id(), &conn_id).await;
+        }
+    }
+    if let Some(conns) = b.get_peer_map().list_peer_conns(a.my_peer_id()).await {
+        for conn in conns {
+            let conn_id = conn.conn_id.parse().expect("invalid conn id");
+            let _ = b.close_peer_conn(a.my_peer_id(), &conn_id).await;
+        }
+    }
+}
+
+async fn check_dns_unavailable_at(server_addr: SocketAddr, domain: &str) {
+    let deadline = Instant::now() + Duration::from_secs(15);
+    let name = rr::Name::from_str(domain).unwrap();
+
+    loop {
+        let stream = UdpClientStream::builder(server_addr, TokioRuntimeProvider::default()).build();
+        let (mut client, background) = Client::<TokioRuntimeProvider>::from_sender(stream);
+
+        let background_task = tokio::spawn(background);
+        let query_result = tokio::time::timeout(
+            Duration::from_secs(1),
+            client.query(name.clone(), DNSClass::IN, RecordType::A),
+        )
+        .await;
+
+        background_task.abort();
+        let _ = background_task.await;
+
+        match query_result {
+            Ok(Ok(response)) if !response.answers.is_empty() => {
+                if Instant::now() >= deadline {
+                    panic!(
+                        "DNS endpoint {server_addr} still answered for '{domain}': {:?}",
+                        response.answers
+                    );
+                }
+            }
+            _ => return,
+        }
+
+        tokio::time::sleep(Duration::from_millis(200)).await;
+    }
+}
+
+async fn wait_peer_zone_visibility(
+    peer_mgr: Arc<PeerManager>,
+    target_peer_id: u32,
+    zone_origin_substr: &str,
+    expected_visible: bool,
+) {
+    let dns = DnsPeerMgr::new(peer_mgr.clone(), peer_mgr.get_global_ctx());
+    let deadline = Instant::now() + Duration::from_secs(20);
+
+    loop {
+        let _ = dns
+            .refresh(target_peer_id, Default::default(), Default::default())
+            .await;
+
+        let snapshot = dns.snapshot();
+
+        let visible = snapshot.zones.iter().any(|z| {
+            z.content
+                .contains(&format!("$ORIGIN {}", zone_origin_substr))
+        });
+
+        if visible == expected_visible {
+            return;
+        }
+
+        assert!(
+            Instant::now() < deadline,
+            "zone visibility mismatch for '{}': expected {}, got {}",
+            zone_origin_substr,
+            expected_visible,
+            visible,
+        );
+        tokio::time::sleep(Duration::from_millis(200)).await;
+    }
+}
+
+fn cfg_with_listener(name: &str, ipv4: &str, domain: &str, listener_port: u16) -> String {
+    format!(
+        r#"
+instance_name = "dns-int-{name}"
+ipv4 = "{ipv4}"
+
+[dns]
+name = "{name}"
+domain = "{domain}"
+listeners = ["udp://127.0.0.1:{listener_port}"]
+"#
+    )
+}
+
+#[tokio::test]
+#[serial_test::serial(dns_integration_rpc)]
+async fn config_string_single_node_resolves_self_and_custom_zone() {
+    let listener_port = find_free_udp_port();
+    let config = format!(
+        r#"
+{}
+
+[[dns.zone]]
+origin = "services.alpha.et-test"
+records = ["api IN A 10.77.0.7"]
+
+[dns.zone.export]
+whitelist = ["*"]
+"#,
+        cfg_with_listener("alpha", "10.144.144.11/24", "et-test", listener_port)
+    );
+
+    let peer_mgr = prepare_env_from_config_str(&config).await;
+    let mut dns_node = start_dns_node_without_nic(peer_mgr);
+
+    let server_addr = SocketAddr::new(Ipv4Addr::LOCALHOST.into(), listener_port);
+    check_dns_record_at(server_addr, "alpha.et-test.", "10.144.144.11").await;
+    check_dns_record_at(server_addr, "api.services.alpha.et-test.", "10.77.0.7").await;
+
+    dns_node.stop().await.unwrap();
+}
+
+#[tokio::test]
+#[serial_test::serial(dns_integration_rpc)]
+async fn config_string_two_nodes_sync_self_zone_and_exported_zone() {
+    let listener_a = find_free_udp_port();
+    let listener_b = find_free_udp_port();
+
+    let config_a = format!(
+        r#"
+{}
+
+[[dns.zone]]
+origin = "shared.mesh-test"
+records = ["app IN A 10.66.0.8"]
+
+[dns.zone.export]
+whitelist = ["*"]
+"#,
+        cfg_with_listener("node-a", "10.144.144.21/24", "mesh-test", listener_a)
+    );
+
+    let config_b = cfg_with_listener("node-b", "10.144.144.22/24", "mesh-test", listener_b);
+
+    let peer_a = prepare_env_from_config_str(&config_a).await;
+    let peer_b = prepare_env_from_config_str(&config_b).await;
+    connect_peer_manager(peer_a.clone(), peer_b.clone()).await;
+    wait_route_appear(peer_a.clone(), peer_b.clone())
+        .await
+        .expect("route should appear");
+
+    let mut node_a = start_dns_node_without_nic(peer_a);
+    let mut node_b = start_dns_node_without_nic(peer_b);
+
+    let addr_a = SocketAddr::new(Ipv4Addr::LOCALHOST.into(), listener_a);
+    let addr_b = SocketAddr::new(Ipv4Addr::LOCALHOST.into(), listener_b);
+
+    check_dns_record_at(addr_a, "node-b.mesh-test.", "10.144.144.22").await;
+    check_dns_record_at(addr_b, "node-b.mesh-test.", "10.144.144.22").await;
+    check_dns_record_at(addr_a, "app.shared.mesh-test.", "10.66.0.8").await;
+    check_dns_record_at(addr_b, "app.shared.mesh-test.", "10.66.0.8").await;
+
+    node_a.stop().await.unwrap();
+    node_b.stop().await.unwrap();
+}
+
+#[tokio::test]
+#[serial_test::serial(dns_integration_rpc)]
+async fn config_string_zone_without_export_section_is_not_synced() {
+    let listener_a = find_free_udp_port();
+    let listener_b = find_free_udp_port();
+
+    let config_a = format!(
+        r#"
+{}
+
+[[dns.zone]]
+origin = "private.mesh-test"
+records = ["secret IN A 10.99.0.9"]
+"#,
+        cfg_with_listener("local-a", "10.144.144.31/24", "mesh-test", listener_a)
+    );
+
+    let config_b = cfg_with_listener("local-b", "10.144.144.32/24", "mesh-test", listener_b);
+
+    let peer_a = prepare_env_from_config_str(&config_a).await;
+    let peer_b = prepare_env_from_config_str(&config_b).await;
+    connect_peer_manager(peer_a.clone(), peer_b.clone()).await;
+    wait_route_appear(peer_a.clone(), peer_b.clone())
+        .await
+        .expect("route should appear");
+
+    // Export behavior is determined by whether `[dns.zone.export]` exists.
+    // Verify from peer-sync view to avoid host-wide DNS-server election side effects.
+    let dns_a = DnsPeerMgr::new(peer_a.clone(), peer_a.get_global_ctx());
+    dns_a.register();
+    let dns_b = DnsPeerMgr::new(peer_b.clone(), peer_b.get_global_ctx());
+    dns_b.register();
+    dns_b
+        .refresh(peer_a.my_peer_id(), Default::default(), Default::default())
+        .await
+        .unwrap();
+
+    let snapshot = dns_b.snapshot();
+    assert!(
+        !snapshot
+            .zones
+            .iter()
+            .any(|z| z.content.contains("$ORIGIN private.mesh-test")),
+        "zone without [dns.zone.export] should not be exported to peer snapshot"
+    );
+}
+
+#[tokio::test]
+#[serial_test::serial(dns_integration_rpc)]
+async fn config_string_export_section_disabled_is_not_synced() {
+    let listener_a = find_free_udp_port();
+    let listener_b = find_free_udp_port();
+
+    let config_a = format!(
+        r#"
+{}
+
+[[dns.zone]]
+origin = "disabled.mesh-test"
+records = ["secret IN A 10.99.1.9"]
+
+[dns.zone.export]
+disabled = true
+"#,
+        cfg_with_listener("local-da", "10.144.144.41/24", "mesh-test", listener_a)
+    );
+
+    let config_b = cfg_with_listener("local-db", "10.144.144.42/24", "mesh-test", listener_b);
+
+    let peer_a = prepare_env_from_config_str(&config_a).await;
+    let peer_b = prepare_env_from_config_str(&config_b).await;
+    connect_peer_manager(peer_a.clone(), peer_b.clone()).await;
+    wait_route_appear(peer_a.clone(), peer_b.clone())
+        .await
+        .expect("route should appear");
+
+    let dns_a = DnsPeerMgr::new(peer_a.clone(), peer_a.get_global_ctx());
+    dns_a.register();
+    let dns_b = DnsPeerMgr::new(peer_b.clone(), peer_b.get_global_ctx());
+    dns_b.register();
+    dns_b
+        .refresh(peer_a.my_peer_id(), Default::default(), Default::default())
+        .await
+        .unwrap();
+
+    let snapshot = dns_b.snapshot();
+    assert!(
+        !snapshot
+            .zones
+            .iter()
+            .any(|z| z.content.contains("$ORIGIN disabled.mesh-test")),
+        "zone with [dns.zone.export] disabled=true should not be exported"
+    );
+}
+
+#[tokio::test]
+#[serial_test::serial(dns_integration_rpc)]
+async fn config_patch_updates_zone_record_visible_on_query() {
+    let listener = find_free_udp_port();
+    let config = format!(
+        r#"
+{}
+
+[[dns.zone]]
+origin = "patch.mesh-test"
+records = ["api IN A 10.80.0.1"]
+
+[dns.zone.export]
+"#,
+        cfg_with_listener("patch-node", "10.144.149.11/24", "mesh-test", listener)
+    );
+
+    let peer = prepare_env_from_config_str(&config).await;
+    let mut dns_node = start_dns_node_without_nic(peer.clone());
+    let server_addr = SocketAddr::new(Ipv4Addr::LOCALHOST.into(), listener);
+
+    check_dns_record_at(server_addr, "api.patch.mesh-test.", "10.80.0.1").await;
+
+    let mut dns = peer.get_global_ctx().config.get_dns().into_raw();
+    let mut zones = dns.zones.unwrap();
+    let zone_idx = zones
+        .iter()
+        .position(|z| z.origin.to_string().contains("patch.mesh-test"))
+        .expect("patch zone should exist");
+    let mut zone = zones[zone_idx].clone().into_raw();
+    zone.records = Some(vec!["api IN A 10.80.0.2".to_string()]);
+    zones[zone_idx] = zone.try_into().expect("patch zone update should be valid");
+    dns.zones = Some(zones);
+    peer.get_global_ctx().config.set_dns(dns.into());
+    peer.get_global_ctx()
+        .issue_event(crate::common::global_ctx::GlobalCtxEvent::ConfigPatched(
+            crate::proto::api::config::InstanceConfigPatch::default(),
+        ));
+
+    check_dns_record_at(server_addr, "api.patch.mesh-test.", "10.80.0.2").await;
+
+    dns_node.stop().await.unwrap();
+}
+
+#[tokio::test]
+#[serial_test::serial(dns_integration_rpc)]
+async fn config_patch_reloads_listener_binding() {
+    let listener_old = find_free_udp_port();
+    let listener_new = find_free_udp_port();
+    let config = cfg_with_listener(
+        "listener-patch",
+        "10.144.150.11/24",
+        "mesh-test",
+        listener_old,
+    );
+
+    let peer = prepare_env_from_config_str(&config).await;
+    let mut dns_node = start_dns_node_without_nic(peer.clone());
+
+    let old_addr = SocketAddr::new(Ipv4Addr::LOCALHOST.into(), listener_old);
+    let new_addr = SocketAddr::new(Ipv4Addr::LOCALHOST.into(), listener_new);
+    check_dns_record_at(old_addr, "listener-patch.mesh-test.", "10.144.150.11").await;
+
+    let mut dns = peer.get_global_ctx().config.get_dns().into_raw();
+    dns.listeners = Some(
+        vec![
+            format!("udp://127.0.0.1:{listener_new}")
+                .parse()
+                .expect("invalid listener"),
+        ]
+        .into(),
+    );
+    peer.get_global_ctx().config.set_dns(dns.into());
+    peer.get_global_ctx()
+        .issue_event(crate::common::global_ctx::GlobalCtxEvent::ConfigPatched(
+            crate::proto::api::config::InstanceConfigPatch::default(),
+        ));
+
+    check_dns_record_at(new_addr, "listener-patch.mesh-test.", "10.144.150.11").await;
+    check_dns_unavailable_at(old_addr, "listener-patch.mesh-test.").await;
+
+    dns_node.stop().await.unwrap();
+}
+
+#[tokio::test]
+#[serial_test::serial(dns_integration_rpc)]
+async fn config_string_three_nodes_partition_and_recover_dns_propagation() {
+    let listener_a = find_free_udp_port();
+    let listener_b = find_free_udp_port();
+    let listener_c = find_free_udp_port();
+
+    let config_a = cfg_with_listener("node-a7", "10.144.151.11/24", "mesh7-test", listener_a);
+    let config_b = cfg_with_listener("node-b7", "10.144.151.12/24", "mesh7-test", listener_b);
+    let config_c = format!(
+        r#"
+{}
+
+[[dns.zone]]
+origin = "shared-c7.mesh7-test"
+records = ["svc IN A 10.77.7.7"]
+
+[dns.zone.export]
+"#,
+        cfg_with_listener("node-c7", "10.144.151.13/24", "mesh7-test", listener_c)
+    );
+
+    let peer_a = prepare_env_from_config_str(&config_a).await;
+    let peer_b = prepare_env_from_config_str(&config_b).await;
+    let peer_c = prepare_env_from_config_str(&config_c).await;
+
+    connect_peer_manager(peer_a.clone(), peer_b.clone()).await;
+    connect_peer_manager(peer_b.clone(), peer_c.clone()).await;
+    wait_route_appear(peer_a.clone(), peer_c.clone())
+        .await
+        .expect("route a-c should appear via b");
+
+    let mut node_a = start_dns_node_without_nic(peer_a.clone());
+    let mut node_b = start_dns_node_without_nic(peer_b.clone());
+    let mut node_c = start_dns_node_without_nic(peer_c.clone());
+
+    let addr_a = SocketAddr::new(Ipv4Addr::LOCALHOST.into(), listener_a);
+    check_dns_record_at(addr_a, "node-c7.mesh7-test.", "10.144.151.13").await;
+    check_dns_record_at(addr_a, "svc.shared-c7.mesh7-test.", "10.77.7.7").await;
+    wait_peer_zone_visibility(
+        peer_a.clone(),
+        peer_c.my_peer_id(),
+        "node-c7.mesh7-test",
+        true,
+    )
+    .await;
+    wait_peer_zone_visibility(
+        peer_a.clone(),
+        peer_c.my_peer_id(),
+        "shared-c7.mesh7-test",
+        true,
+    )
+    .await;
+
+    disconnect_all_peer_conns(peer_b.clone(), peer_c.clone()).await;
+    wait_route_disappear(peer_a.clone(), peer_c.my_peer_id()).await;
+    // Validate via peer-sync snapshot to avoid process-wide DNS-server election side effects.
+    wait_peer_zone_visibility(
+        peer_a.clone(),
+        peer_c.my_peer_id(),
+        "node-c7.mesh7-test",
+        false,
+    )
+    .await;
+    wait_peer_zone_visibility(
+        peer_a.clone(),
+        peer_c.my_peer_id(),
+        "shared-c7.mesh7-test",
+        false,
+    )
+    .await;
+
+    connect_peer_manager(peer_b.clone(), peer_c.clone()).await;
+    wait_route_appear(peer_a.clone(), peer_c.clone())
+        .await
+        .expect("route a-c should recover via b");
+
+    wait_peer_zone_visibility(
+        peer_a.clone(),
+        peer_c.my_peer_id(),
+        "node-c7.mesh7-test",
+        true,
+    )
+    .await;
+    wait_peer_zone_visibility(
+        peer_a.clone(),
+        peer_c.my_peer_id(),
+        "shared-c7.mesh7-test",
+        true,
+    )
+    .await;
+
+    check_dns_record_at(addr_a, "node-c7.mesh7-test.", "10.144.151.13").await;
+    check_dns_record_at(addr_a, "svc.shared-c7.mesh7-test.", "10.77.7.7").await;
+
+    node_a.stop().await.unwrap();
+    node_b.stop().await.unwrap();
+    node_c.stop().await.unwrap();
+}
+
+#[tokio::test]
+#[serial_test::serial(dns_integration_rpc)]
+async fn config_string_three_nodes_chain_sync_self_and_exported_zone() {
+    let listener_a = find_free_udp_port();
+    let listener_b = find_free_udp_port();
+    let listener_c = find_free_udp_port();
+
+    let config_a = cfg_with_listener("node-a3", "10.144.145.11/24", "mesh3-test", listener_a);
+    let config_b = cfg_with_listener("node-b3", "10.144.145.12/24", "mesh3-test", listener_b);
+    let config_c = format!(
+        r#"
+{}
+
+[[dns.zone]]
+origin = "shared-c.mesh3-test"
+records = ["api IN A 10.66.1.8"]
+
+[dns.zone.export]
+"#,
+        cfg_with_listener("node-c3", "10.144.145.13/24", "mesh3-test", listener_c)
+    );
+
+    let peer_a = prepare_env_from_config_str(&config_a).await;
+    let peer_b = prepare_env_from_config_str(&config_b).await;
+    let peer_c = prepare_env_from_config_str(&config_c).await;
+
+    connect_peer_manager(peer_a.clone(), peer_b.clone()).await;
+    connect_peer_manager(peer_b.clone(), peer_c.clone()).await;
+    wait_route_appear(peer_a.clone(), peer_b.clone())
+        .await
+        .expect("route a-b should appear");
+    wait_route_appear(peer_b.clone(), peer_c.clone())
+        .await
+        .expect("route b-c should appear");
+    wait_route_appear(peer_a.clone(), peer_c.clone())
+        .await
+        .expect("route a-c should appear via b");
+
+    let mut node_a = start_dns_node_without_nic(peer_a);
+    let mut node_b = start_dns_node_without_nic(peer_b);
+    let mut node_c = start_dns_node_without_nic(peer_c);
+
+    let addr_a = SocketAddr::new(Ipv4Addr::LOCALHOST.into(), listener_a);
+    let addr_b = SocketAddr::new(Ipv4Addr::LOCALHOST.into(), listener_b);
+    let addr_c = SocketAddr::new(Ipv4Addr::LOCALHOST.into(), listener_c);
+
+    check_dns_record_at(addr_a, "node-c3.mesh3-test.", "10.144.145.13").await;
+    check_dns_record_at(addr_b, "node-c3.mesh3-test.", "10.144.145.13").await;
+    check_dns_record_at(addr_c, "node-a3.mesh3-test.", "10.144.145.11").await;
+    check_dns_record_at(addr_a, "api.shared-c.mesh3-test.", "10.66.1.8").await;
+    check_dns_record_at(addr_c, "api.shared-c.mesh3-test.", "10.66.1.8").await;
+
+    node_a.stop().await.unwrap();
+    node_b.stop().await.unwrap();
+    node_c.stop().await.unwrap();
+}
+
+#[tokio::test]
+#[serial_test::serial(dns_integration_rpc)]
+async fn config_string_three_nodes_late_join_propagates_dns() {
+    let listener_a = find_free_udp_port();
+    let listener_b = find_free_udp_port();
+    let listener_c = find_free_udp_port();
+
+    let config_a = cfg_with_listener("node-a4", "10.144.146.11/24", "mesh4-test", listener_a);
+    let config_b = cfg_with_listener("node-b4", "10.144.146.12/24", "mesh4-test", listener_b);
+    let config_c = format!(
+        r#"
+{}
+
+[[dns.zone]]
+origin = "joined.mesh4-test"
+records = ["svc IN A 10.66.2.8"]
+
+[dns.zone.export]
+"#,
+        cfg_with_listener("node-c4", "10.144.146.13/24", "mesh4-test", listener_c)
+    );
+
+    let peer_a = prepare_env_from_config_str(&config_a).await;
+    let peer_b = prepare_env_from_config_str(&config_b).await;
+    let peer_c = prepare_env_from_config_str(&config_c).await;
+
+    connect_peer_manager(peer_a.clone(), peer_b.clone()).await;
+    wait_route_appear(peer_a.clone(), peer_b.clone())
+        .await
+        .expect("route a-b should appear");
+
+    let mut node_a = start_dns_node_without_nic(peer_a.clone());
+    let mut node_b = start_dns_node_without_nic(peer_b.clone());
+
+    let addr_a = SocketAddr::new(Ipv4Addr::LOCALHOST.into(), listener_a);
+    check_dns_record_missing_at(addr_a, "node-c4.mesh4-test.").await;
+
+    connect_peer_manager(peer_b.clone(), peer_c.clone()).await;
+    wait_route_appear(peer_b.clone(), peer_c.clone())
+        .await
+        .expect("route b-c should appear");
+    wait_route_appear(peer_a.clone(), peer_c.clone())
+        .await
+        .expect("route a-c should appear after c joins");
+
+    let mut node_c = start_dns_node_without_nic(peer_c);
+
+    check_dns_record_at(addr_a, "node-c4.mesh4-test.", "10.144.146.13").await;
+    check_dns_record_at(addr_a, "svc.joined.mesh4-test.", "10.66.2.8").await;
+
+    node_a.stop().await.unwrap();
+    node_b.stop().await.unwrap();
+    node_c.stop().await.unwrap();
+}
+
+#[tokio::test]
+#[serial_test::serial(dns_integration_rpc)]
+async fn config_string_three_nodes_zone_without_export_not_synced_across_hop() {
+    let listener_a = find_free_udp_port();
+    let listener_b = find_free_udp_port();
+    let listener_c = find_free_udp_port();
+
+    let config_a = cfg_with_listener("node-a5", "10.144.147.11/24", "mesh5-test", listener_a);
+    let config_b = cfg_with_listener("node-b5", "10.144.147.12/24", "mesh5-test", listener_b);
+    let config_c = format!(
+        r#"
+{}
+
+[[dns.zone]]
+origin = "private-c.mesh5-test"
+records = ["secret IN A 10.66.3.8"]
+"#,
+        cfg_with_listener("node-c5", "10.144.147.13/24", "mesh5-test", listener_c)
+    );
+
+    let peer_a = prepare_env_from_config_str(&config_a).await;
+    let peer_b = prepare_env_from_config_str(&config_b).await;
+    let peer_c = prepare_env_from_config_str(&config_c).await;
+
+    connect_peer_manager(peer_a.clone(), peer_b.clone()).await;
+    connect_peer_manager(peer_b.clone(), peer_c.clone()).await;
+    wait_route_appear(peer_a.clone(), peer_c.clone())
+        .await
+        .expect("route a-c should appear via b");
+
+    let dns_c = DnsPeerMgr::new(peer_c.clone(), peer_c.get_global_ctx());
+    dns_c.register();
+    let dns_a = DnsPeerMgr::new(peer_a.clone(), peer_a.get_global_ctx());
+    dns_a.register();
+    dns_a
+        .refresh(peer_c.my_peer_id(), Default::default(), Default::default())
+        .await
+        .unwrap();
+
+    let snapshot = dns_a.snapshot();
+    assert!(
+        !snapshot
+            .zones
+            .iter()
+            .any(|z| z.content.contains("$ORIGIN private-c.mesh5-test")),
+        "zone without [dns.zone.export] should not sync over multi-hop"
+    );
+}
+
+#[tokio::test]
+#[serial_test::serial(dns_integration_rpc)]
+async fn config_string_two_nodes_peer_dns_offline_then_rejoin() {
+    let listener_a = find_free_udp_port();
+    let listener_b = find_free_udp_port();
+
+    let config_a = cfg_with_listener("node-a6", "10.144.148.11/24", "mesh6-test", listener_a);
+    let config_b = cfg_with_listener("node-b6", "10.144.148.12/24", "mesh6-test", listener_b);
+
+    let peer_a = prepare_env_from_config_str(&config_a).await;
+    let peer_b = prepare_env_from_config_str(&config_b).await;
+
+    connect_peer_manager(peer_a.clone(), peer_b.clone()).await;
+    wait_route_appear(peer_a.clone(), peer_b.clone())
+        .await
+        .expect("route should appear");
+
+    let dns_b = DnsPeerMgr::new(peer_b.clone(), peer_b.get_global_ctx());
+    dns_b.register();
+    let dns_a_online = DnsPeerMgr::new(peer_a.clone(), peer_a.get_global_ctx());
+    dns_a_online.register();
+    dns_a_online
+        .refresh(peer_b.my_peer_id(), Default::default(), Default::default())
+        .await
+        .unwrap();
+    assert!(
+        dns_a_online
+            .snapshot()
+            .zones
+            .iter()
+            .any(|z| z.content.contains("$ORIGIN node-b6.mesh6-test")),
+        "peer B self zone should be visible after initial refresh"
+    );
+
+    // Simulate peer offline by closing all direct connections and waiting route withdrawal.
+    if let Some(conns) = peer_a
+        .get_peer_map()
+        .list_peer_conns(peer_b.my_peer_id())
+        .await
+    {
+        for conn in conns {
+            let conn_id = conn.conn_id.parse().expect("invalid conn id");
+            let _ = peer_a.close_peer_conn(peer_b.my_peer_id(), &conn_id).await;
+        }
+    }
+    if let Some(conns) = peer_b
+        .get_peer_map()
+        .list_peer_conns(peer_a.my_peer_id())
+        .await
+    {
+        for conn in conns {
+            let conn_id = conn.conn_id.parse().expect("invalid conn id");
+            let _ = peer_b.close_peer_conn(peer_a.my_peer_id(), &conn_id).await;
+        }
+    }
+
+    wait_route_disappear(peer_a.clone(), peer_b.my_peer_id()).await;
+    wait_route_disappear(peer_b.clone(), peer_a.my_peer_id()).await;
+
+    // Cached remote zones should be purged after peer cache idle timeout.
+    tokio::time::sleep(Duration::from_secs(4)).await;
+    let dns_a_offline = DnsPeerMgr::new(peer_a.clone(), peer_a.get_global_ctx());
+    assert!(
+        !dns_a_offline
+            .snapshot()
+            .zones
+            .iter()
+            .any(|z| z.content.contains("$ORIGIN node-b6.mesh6-test")),
+        "peer B self zone should disappear after route withdrawal and cache expiry"
+    );
+
+    connect_peer_manager(peer_a.clone(), peer_b.clone()).await;
+    wait_route_appear(peer_a.clone(), peer_b.clone())
+        .await
+        .expect("route should re-appear");
+
+    let dns_a_rejoin = DnsPeerMgr::new(peer_a.clone(), peer_a.get_global_ctx());
+    dns_a_rejoin
+        .refresh(peer_b.my_peer_id(), Default::default(), Default::default())
+        .await
+        .unwrap();
+    assert!(
+        dns_a_rejoin
+            .snapshot()
+            .zones
+            .iter()
+            .any(|z| z.content.contains("$ORIGIN node-b6.mesh6-test")),
+        "peer B self zone should be restored after DNS RPC rejoins"
+    );
+}

easytier/src/dns/utils/addr.rs

@@ -0,0 +1,183 @@
+use crate::proto;
+use crate::proto::utils::{RepeatedDeserialize, RepeatedMessageModel, RepeatedSerialize};
+use anyhow::{Error, anyhow};
+use hickory_net::xfer::Protocol;
+use hickory_resolver::config::{ConnectionConfig, NameServerConfig, ProtocolConfig};
+use serde::de::IntoDeserializer;
+use serde::{Deserialize, Deserializer, de};
+use serde_with::{DeserializeFromStr, SerializeDisplay};
+use std::fmt::{Display, Formatter};
+use std::net::{IpAddr, Ipv6Addr, SocketAddr};
+use std::str::FromStr;
+use url::Url;
+
+#[derive(Debug, Copy, Clone, PartialEq, Eq, Hash, SerializeDisplay, DeserializeFromStr)]
+pub struct NameServerAddr {
+    pub protocol: Protocol,
+    pub addr: SocketAddr,
+}
+
+impl From<NameServerAddr> for NameServerConfig {
+    fn from(value: NameServerAddr) -> Self {
+        let mut config = match value.protocol {
+            Protocol::Udp => ConnectionConfig::udp(),
+            Protocol::Tcp => ConnectionConfig::tcp(),
+            _ => unimplemented!(),
+        };
+        config.port = value.addr.port();
+        Self::new(value.addr.ip(), true, vec![config])
+    }
+}
+
+impl From<(IpAddr, &ConnectionConfig)> for NameServerAddr {
+    fn from(value: (IpAddr, &ConnectionConfig)) -> Self {
+        let (ip, config) = value;
+        Self {
+            protocol: config.protocol.to_protocol(),
+            addr: SocketAddr::new(ip, config.port),
+        }
+    }
+}
+
+impl TryFrom<&Url> for NameServerAddr {
+    type Error = Error;
+
+    fn try_from(url: &Url) -> Result<Self, Self::Error> {
+        let protocol = match Protocol::deserialize(url.scheme().into_deserializer())
+            .map_err(|e: de::value::Error| anyhow!("invalid protocol '{}': {}", url.scheme(), e))?
+        {
+            Protocol::Udp => ProtocolConfig::Udp,
+            Protocol::Tcp => ProtocolConfig::Tcp,
+            p => return Err(anyhow!("unsupported protocol: {}", p)),
+        };
+        let host = url.host_str().ok_or(anyhow!("host not found"))?;
+        let port = url.port().unwrap_or(protocol.default_port());
+        let addr = if let Ok(addr) = IpAddr::from_str(host) {
+            SocketAddr::new(addr, port)
+        } else {
+            return Err(anyhow!("invalid address: {}", host));
+        };
+        Ok(Self {
+            protocol: protocol.to_protocol(),
+            addr,
+        })
+    }
+}
+
+impl TryFrom<&proto::common::Url> for NameServerAddr {
+    type Error = Error;
+
+    fn try_from(value: &proto::common::Url) -> Result<Self, Self::Error> {
+        (&Url::try_from(value)?).try_into()
+    }
+}
+
+impl From<&NameServerAddr> for Url {
+    fn from(value: &NameServerAddr) -> Self {
+        Url::parse(&format!("{}://{}", value.protocol, value.addr)).unwrap()
+    }
+}
+
+impl From<&NameServerAddr> for proto::common::Url {
+    fn from(value: &NameServerAddr) -> Self {
+        Url::from(value).into()
+    }
+}
+
+impl From<NameServerAddr> for Url {
+    fn from(value: NameServerAddr) -> Self {
+        (&value).into()
+    }
+}
+
+impl From<NameServerAddr> for proto::common::Url {
+    fn from(value: NameServerAddr) -> Self {
+        (&value).into()
+    }
+}
+
+impl FromStr for NameServerAddr {
+    type Err = Error;
+    fn from_str(s: &str) -> Result<Self, Self::Err> {
+        (&Url::parse(s)?).try_into()
+    }
+}
+
+impl Display for NameServerAddr {
+    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+        f.write_str(Url::from(*self).as_str())
+    }
+}
+
+pub type NameServerAddrGroup = RepeatedMessageModel<NameServerAddr>;
+
+impl From<&NameServerConfig> for NameServerAddrGroup {
+    fn from(value: &NameServerConfig) -> Self {
+        value
+            .connections
+            .iter()
+            .map(|c| (value.ip, c).into())
+            .collect()
+    }
+}
+
+impl From<SocketAddr> for NameServerAddrGroup {
+    fn from(value: SocketAddr) -> Self {
+        vec![
+            NameServerAddr {
+                protocol: Protocol::Udp,
+                addr: value,
+            },
+            NameServerAddr {
+                protocol: Protocol::Tcp,
+                addr: value,
+            },
+        ]
+        .into()
+    }
+}
+
+impl From<IpAddr> for NameServerAddrGroup {
+    fn from(value: IpAddr) -> Self {
+        SocketAddr::new(value, 53).into()
+    }
+}
+
+impl From<u16> for NameServerAddrGroup {
+    fn from(value: u16) -> Self {
+        SocketAddr::new(Ipv6Addr::UNSPECIFIED.into(), value).into()
+    }
+}
+
+impl RepeatedSerialize for NameServerAddr {}
+
+impl<'de> RepeatedDeserialize<'de> for NameServerAddr {
+    fn deserialize<D>(deserializer: D) -> Result<NameServerAddrGroup, D::Error>
+    where
+        D: Deserializer<'de>,
+    {
+        #[derive(Deserialize)]
+        #[serde(untagged)]
+        enum Candidate {
+            NameServerAddr(NameServerAddr),
+            U16(u16),
+            IpAddr(IpAddr),
+            SocketAddr(SocketAddr),
+        }
+
+        let items = Vec::<Candidate>::deserialize(deserializer)?;
+        let items = items
+            .into_iter()
+            .flat_map(|item| -> NameServerAddrGroup {
+                match item {
+                    Candidate::NameServerAddr(addr) => vec![addr].into(),
+                    Candidate::U16(port) => port.into(),
+                    Candidate::IpAddr(ip) => ip.into(),
+                    Candidate::SocketAddr(addr) => addr.into(),
+                }
+            })
+            .collect();
+
+        Ok(items)
+    }
+}

easytier/src/dns/utils/mod.rs

@@ -0,0 +1,3 @@
+pub mod addr;
+pub mod response;
+pub mod zone_handler;

easytier/src/dns/utils/response.rs

@@ -0,0 +1,59 @@
+use hickory_net::NetError;
+use hickory_proto::rr::Record;
+use hickory_proto::serialize::binary::BinEncoder;
+use hickory_server::server::{ResponseHandler, ResponseInfo};
+use hickory_server::zone_handler::MessageResponse;
+use parking_lot::Mutex;
+use std::sync::Arc;
+
+// ResponseWrapper for serializing DNS responses into a byte buffer.
+// Used by the address hijacking NIC packet filter to produce DNS replies in-place.
+#[derive(Debug, Clone)]
+pub struct ResponseHandle {
+    inner: Arc<Mutex<Vec<u8>>>,
+}
+
+impl ResponseHandle {
+    pub fn new(capacity: usize) -> Self {
+        Self {
+            inner: Arc::new(Mutex::new(Vec::with_capacity(capacity))),
+        }
+    }
+
+    pub fn into_inner(self) -> Option<Vec<u8>> {
+        Arc::into_inner(self.inner).map(Mutex::into_inner)
+    }
+}
+
+pub trait RecordIter<'r>: Iterator<Item = &'r Record> + Send + 'r {}
+
+impl<'r, T> RecordIter<'r> for T where T: Iterator<Item = &'r Record> + Send + 'r {}
+
+#[async_trait::async_trait]
+impl ResponseHandler for ResponseHandle {
+    async fn send_response<'r>(
+        &mut self,
+        response: MessageResponse<
+            '_,
+            'r,
+            impl RecordIter<'r>,
+            impl RecordIter<'r>,
+            impl RecordIter<'r>,
+            impl RecordIter<'r>,
+        >,
+    ) -> Result<ResponseInfo, NetError> {
+        let max_size = if let Some(edns) = response.edns() {
+            edns.max_payload()
+        } else {
+            hickory_net::udp::MAX_RECEIVE_BUFFER_SIZE as u16
+        };
+
+        let mut inner = self.inner.lock();
+        inner.clear();
+        let mut encoder = BinEncoder::new(inner.as_mut());
+        encoder.set_max_size(max_size);
+        response
+            .destructive_emit(&mut encoder)
+            .map_err(NetError::Proto)
+    }
+}

easytier/src/dns/utils/zone_handler.rs

@@ -0,0 +1,85 @@
+use crate::dns::config::zone::Fallthrough;
+use delegate::delegate;
+use derive_more::{Constructor, Deref, DerefMut};
+use hickory_proto::op::ResponseCode;
+use hickory_proto::rr::{LowerName, RecordType, TSigResponseContext};
+use hickory_server::server::{Request, RequestInfo};
+use hickory_server::zone_handler::{
+    AuthLookup, AxfrPolicy, LookupControlFlow, LookupError, LookupOptions, ZoneHandler, ZoneType,
+};
+use std::collections::HashSet;
+use std::sync::Arc;
+
+pub type ArcZoneHandler = Arc<dyn ZoneHandler>;
+
+#[derive(Constructor, Deref, DerefMut)]
+pub struct ChainedZoneHandler<H>
+where
+    H: ZoneHandler,
+{
+    #[deref]
+    #[deref_mut]
+    handler: H,
+    fallthrough: HashSet<Fallthrough>,
+}
+
+#[async_trait::async_trait]
+impl<H> ZoneHandler for ChainedZoneHandler<H>
+where
+    H: ZoneHandler,
+{
+    delegate! {
+        to self.handler {
+            fn zone_type(&self) -> ZoneType;
+            fn axfr_policy(&self) -> AxfrPolicy;
+            fn origin(&self) -> &LowerName;
+        }
+    }
+
+    #[inline]
+    async fn update(
+        &self,
+        update: &Request,
+        now: u64,
+    ) -> (Result<bool, ResponseCode>, Option<TSigResponseContext>) {
+        self.handler.update(update, now).await
+    }
+    #[inline]
+    async fn lookup(
+        &self,
+        name: &LowerName,
+        rtype: RecordType,
+        request_info: Option<&RequestInfo<'_>>,
+        lookup_options: LookupOptions,
+    ) -> LookupControlFlow<AuthLookup> {
+        self.handler
+            .lookup(name, rtype, request_info, lookup_options)
+            .await
+    }
+    #[inline]
+    async fn search(
+        &self,
+        request: &Request,
+        lookup_options: LookupOptions,
+    ) -> (LookupControlFlow<AuthLookup>, Option<TSigResponseContext>) {
+        let (result, tsig) = self.handler.search(request, lookup_options).await;
+
+        match &result {
+            LookupControlFlow::Continue(Err(e)) | LookupControlFlow::Break(Err(e))
+                if self.fallthrough.contains(&Fallthrough::Any)
+                    || matches!(e, LookupError::ResponseCode(c) if self.fallthrough.contains(&(*c).into())) =>
+            {
+                (LookupControlFlow::Skip, None)
+            }
+            _ => (result, tsig),
+        }
+    }
+    #[inline]
+    async fn nsec_records(
+        &self,
+        name: &LowerName,
+        lookup_options: LookupOptions,
+    ) -> LookupControlFlow<AuthLookup> {
+        self.handler.nsec_records(name, lookup_options).await
+    }
+}

easytier/src/dns/zone.rs

@@ -0,0 +1,572 @@
+use crate::dns::config::zone::Fallthrough;
+use crate::dns::utils::addr::{NameServerAddr, NameServerAddrGroup};
+use crate::dns::utils::zone_handler::{ArcZoneHandler, ChainedZoneHandler};
+use crate::proto::dns::ZoneData;
+use crate::proto::utils::RepeatedMessageModel;
+use crate::utils::dns::resolver_conf;
+use hickory_net::runtime::TokioRuntimeProvider;
+use hickory_proto::rr::{LowerName, RecordSet, RrKey};
+use hickory_proto::serialize::txt::Parser;
+use hickory_server::store::forwarder::{ForwardConfig, ForwardZoneHandler};
+use hickory_server::store::in_memory::InMemoryZoneHandler;
+use hickory_server::zone_handler::{AxfrPolicy, ZoneType};
+use indexmap::IndexMap;
+use itertools::chain;
+use maplit::hashset;
+use std::collections::{BTreeMap, HashSet};
+use std::sync::Arc;
+
+#[derive(Debug, Clone, Default)]
+pub struct Zone {
+    origin: LowerName,
+    records: BTreeMap<RrKey, RecordSet>,
+    pub forward: Option<ForwardConfig>,
+    fallthrough: HashSet<Fallthrough>,
+}
+
+impl Zone {
+    pub fn system() -> Self {
+        let (config, opts) = resolver_conf();
+        let forward = ForwardConfig {
+            name_servers: config.name_servers().to_vec(),
+            options: Some(opts),
+        };
+        Self {
+            origin: ".".parse().unwrap(),
+            forward: Some(forward),
+            fallthrough: hashset! {},
+            ..Default::default()
+        }
+    }
+}
+
+impl Zone {
+    pub fn create_memory_zone_handler(&self) -> Option<ArcZoneHandler> {
+        (!self.records.is_empty()).then(|| {
+            let mut memory = InMemoryZoneHandler::<TokioRuntimeProvider>::empty(
+                self.origin.clone().into(),
+                ZoneType::External,
+                AxfrPolicy::default(),
+            );
+
+            memory.records_get_mut().extend(
+                self.records
+                    .clone()
+                    .into_iter()
+                    .map(|(k, v)| (k, Arc::new(v))),
+            );
+
+            Arc::new(ChainedZoneHandler::new(memory, self.fallthrough.clone())) as _
+        })
+    }
+
+    pub fn create_forward_zone_handler(&self) -> Option<ArcZoneHandler> {
+        self.forward.as_ref().and_then(|forward| {
+            ForwardZoneHandler::builder_with_config(
+                forward.clone(),
+                TokioRuntimeProvider::default(),
+            )
+            .build()
+            .inspect_err(|error| tracing::error!(?error, "failed to create forward zone_handler"))
+            .ok()
+            .map(|handler| {
+                Arc::new(ChainedZoneHandler::new(handler, self.fallthrough.clone())) as _
+            })
+        })
+    }
+}
+
+impl TryFrom<&ZoneData> for Zone {
+    type Error = anyhow::Error;
+
+    fn try_from(value: &ZoneData) -> Result<Self, Self::Error> {
+        let (origin, records) = Parser::new(&value.content, None, None)
+            .parse()
+            .map_err(|e| anyhow::anyhow!("failed to parse zone data: {e}"))?;
+
+        let name_servers = value
+            .forwarders
+            .iter()
+            .map(NameServerAddr::try_from)
+            .map(|a| a.map(Into::into))
+            .collect::<Result<Vec<_>, _>>()?;
+        let forward = (!name_servers.is_empty()).then_some(ForwardConfig {
+            name_servers,
+            options: None,
+        });
+
+        let fallthrough = value.fallthrough.iter().copied().map(Into::into).collect();
+
+        Ok(Self {
+            origin: origin.into(),
+            records,
+            forward,
+            fallthrough,
+        })
+    }
+}
+
+impl From<Zone> for ZoneData {
+    fn from(value: Zone) -> Self {
+        let records = value
+            .records
+            .values()
+            .flat_map(RecordSet::records_without_rrsigs)
+            .map(ToString::to_string);
+
+        let forwarders = value
+            .forward
+            .into_iter()
+            .flat_map(|f| f.name_servers.into_iter())
+            .map(|ns| (&ns).into())
+            .flat_map(NameServerAddrGroup::into_iter)
+            .map(Into::into);
+
+        Self::new(&value.origin, 0, records, forwarders, value.fallthrough)
+    }
+}
+
+pub type ZoneGroup = RepeatedMessageModel<Zone>;
+
+impl ZoneGroup {
+    pub fn into_groups(self) -> IndexMap<LowerName, ZoneGroup> {
+        self.into_iter().fold(IndexMap::new(), |mut map, zone| {
+            map.entry(zone.origin.clone()).or_default().push(zone);
+            map
+        })
+    }
+
+    pub fn iter_zone_handlers(&self) -> impl Iterator<Item = ArcZoneHandler> + use<'_> {
+        self.iter().flat_map(|zone| {
+            chain(
+                zone.create_memory_zone_handler(),
+                zone.create_forward_zone_handler(),
+            )
+        })
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::dns::tests::new_request;
+    use crate::dns::utils::response::ResponseHandle;
+    use crate::proto::common::Url;
+    use crate::proto::dns::ZoneData;
+    use hickory_proto::op::{Message, ResponseCode};
+    use hickory_proto::rr::{RData, Record, RecordType, RrsetRecords};
+    use hickory_server::Server;
+    use hickory_server::zone_handler::Catalog;
+    use maplit::hashset;
+    use std::collections::HashSet;
+    use std::net::{Ipv4Addr, Ipv6Addr, SocketAddr};
+    use std::str::FromStr;
+    use tokio::net::UdpSocket;
+    use tokio::task::JoinHandle;
+
+    impl Zone {
+        // Test-only record iterator for precise assertions.
+        pub fn iter_records(&self) -> impl Iterator<Item = &Record> {
+            self.records
+                .values()
+                .filter(|set| !set.is_empty())
+                .flat_map(|set| {
+                    let RrsetRecords::RecordsOnly(records) = set.records_without_rrsigs() else {
+                        unreachable!()
+                    };
+                    records
+                })
+        }
+    }
+
+    fn zone_data_with_fallthrough(
+        origin: &str,
+        records: Vec<&str>,
+        forwarders: Vec<&str>,
+        fallthrough: HashSet<Fallthrough>,
+    ) -> ZoneData {
+        ZoneData::new(
+            &origin.parse().unwrap(),
+            60,
+            records,
+            forwarders
+                .into_iter()
+                .map(|url| Url::from_str(url).unwrap()),
+            fallthrough,
+        )
+    }
+
+    fn zone_data(origin: &str, records: Vec<&str>, forwarders: Vec<&str>) -> ZoneData {
+        zone_data_with_fallthrough(origin, records, forwarders, hashset! {Fallthrough::Any})
+    }
+
+    fn build_catalog(zones: ZoneGroup) -> Catalog {
+        zones
+            .into_groups()
+            .into_iter()
+            .fold(Catalog::new(), |mut catalog, (origin, group)| {
+                catalog.upsert(origin, group.iter_zone_handlers().collect());
+                catalog
+            })
+    }
+
+    async fn lookup_message(
+        catalog: &Catalog,
+        name: &str,
+        record_type: RecordType,
+    ) -> anyhow::Result<(ResponseCode, Option<Message>)> {
+        let request = new_request(name, record_type)?;
+        let response = ResponseHandle::new(1024);
+        let info = catalog.lookup(&request, None, 0, response.clone()).await;
+        let message = response
+            .into_inner()
+            .map(|raw| Message::from_vec(&raw))
+            .transpose()?;
+        Ok((info.response_code, message))
+    }
+
+    fn has_a_answer(message: &Message, expected: Ipv4Addr) -> bool {
+        message
+            .answers
+            .iter()
+            .any(|record| matches!(record.data, RData::A(addr) if *addr == expected))
+    }
+
+    fn has_aaaa_answer(message: &Message, expected: Ipv6Addr) -> bool {
+        message
+            .answers
+            .iter()
+            .any(|record| matches!(record.data, RData::AAAA(addr) if *addr == expected))
+    }
+
+    async fn start_upstream_server() -> anyhow::Result<(SocketAddr, JoinHandle<()>)> {
+        let upstream = Zone::try_from(&zone_data(
+            "upstream.test",
+            vec!["from-forward 60 IN A 203.0.113.9"],
+            vec![],
+        ))?;
+
+        let mut catalog = Catalog::new();
+        catalog.upsert(
+            upstream.origin.clone(),
+            vec![upstream.create_memory_zone_handler().unwrap()],
+        );
+
+        let socket = UdpSocket::bind("127.0.0.1:0").await?;
+        let addr = socket.local_addr()?;
+
+        let mut server = Server::new(catalog);
+        server.register_socket(socket);
+        let handle = tokio::spawn(async move {
+            let _ = server.block_until_done().await;
+        });
+
+        Ok((addr, handle))
+    }
+
+    #[test]
+    fn zone_try_from_rejects_invalid_record() {
+        let data = zone_data("invalid-record.test", vec!["this is not a record"], vec![]);
+
+        let err = Zone::try_from(&data).expect_err("invalid record should fail");
+        assert!(err.to_string().contains("failed to parse zone data"));
+    }
+
+    #[test]
+    fn zone_try_from_rejects_invalid_forwarder_protocol() {
+        let data = zone_data("invalid-forwarder.test", vec![], vec!["http://1.1.1.1:53"]);
+
+        let err = Zone::try_from(&data).expect_err("unsupported forwarder should fail");
+        assert!(err.to_string().contains("unsupported") || err.to_string().contains("protocol"));
+    }
+
+    #[test]
+    fn empty_zone_creates_no_zone_handler() -> anyhow::Result<()> {
+        let zone = Zone::try_from(&zone_data("empty.test", vec![], vec![]))?;
+
+        assert!(zone.create_memory_zone_handler().is_none());
+        assert!(zone.create_forward_zone_handler().is_none());
+
+        Ok(())
+    }
+
+    #[test]
+    fn zone_roundtrip_preserves_records_and_forwarders() -> anyhow::Result<()> {
+        let zone = Zone::try_from(&zone_data(
+            "roundtrip.test",
+            vec!["www 0 IN A 123.123.123.123", "app IN CNAME www"],
+            vec!["udp://1.1.1.1:53", "tcp://8.8.8.8:53"],
+        ))?;
+
+        assert_eq!(zone.iter_records().count(), 2);
+        assert_eq!(zone.forward.as_ref().unwrap().name_servers.len(), 2);
+
+        let serialized = ZoneData::from(zone.clone());
+        let reparsed = Zone::try_from(&serialized)?;
+        assert_eq!(reparsed.origin.to_string(), "roundtrip.test.");
+        assert_eq!(reparsed.iter_records().count(), 2);
+        assert_eq!(reparsed.forward.as_ref().unwrap().name_servers.len(), 2);
+
+        Ok(())
+    }
+
+    #[test]
+    fn zone_group_into_groups_merges_same_origin() -> anyhow::Result<()> {
+        let zones: ZoneGroup = vec![
+            Zone::try_from(&zone_data("same.test", vec!["@ IN A 10.0.0.1"], vec![]))?,
+            Zone::try_from(&zone_data("other.test", vec!["@ IN A 10.0.0.2"], vec![]))?,
+            Zone::try_from(&zone_data("same.test", vec![], vec!["udp://1.1.1.1:53"]))?,
+        ]
+        .into();
+
+        let groups = zones.into_groups();
+        assert_eq!(groups.len(), 2);
+        assert_eq!(
+            groups
+                .get(&LowerName::from_str("same.test.")?)
+                .unwrap()
+                .len(),
+            2
+        );
+        assert_eq!(
+            groups
+                .get(&LowerName::from_str("other.test.")?)
+                .unwrap()
+                .len(),
+            1
+        );
+
+        Ok(())
+    }
+
+    #[test]
+    fn zone_group_iter_zone_handlers_returns_memory_and_forward() -> anyhow::Result<()> {
+        let zones: ZoneGroup = vec![Zone::try_from(&zone_data(
+            "zone-handler.test",
+            vec!["@ IN A 10.0.0.10"],
+            vec!["udp://1.1.1.1:53"],
+        ))?]
+        .into();
+
+        let zone_handlers = zones.iter_zone_handlers().collect::<Vec<_>>();
+        assert_eq!(zone_handlers.len(), 2);
+
+        Ok(())
+    }
+
+    #[test]
+    fn zone_system_builds_root_forwarder() {
+        let zone = Zone::system();
+        assert_eq!(zone.origin.to_string(), ".");
+        assert!(zone.forward.is_some());
+        assert!(zone.create_forward_zone_handler().is_some());
+    }
+
+    #[tokio::test]
+    async fn catalog_lookup_returns_a_record_from_memory_zone_handler() -> anyhow::Result<()> {
+        let zones: ZoneGroup = vec![Zone::try_from(&zone_data(
+            "memory.test",
+            vec!["@ IN A 10.20.30.40"],
+            vec![],
+        ))?]
+        .into();
+        let catalog = build_catalog(zones);
+
+        let (rcode, message) = lookup_message(&catalog, "memory.test.", RecordType::A).await?;
+        assert_eq!(rcode, ResponseCode::NoError);
+        let message = message.expect("response should exist");
+        assert!(has_a_answer(&message, Ipv4Addr::new(10, 20, 30, 40)));
+
+        Ok(())
+    }
+
+    #[tokio::test]
+    async fn catalog_lookup_returns_refused_when_zone_is_missing() -> anyhow::Result<()> {
+        let zones: ZoneGroup = vec![Zone::try_from(&zone_data(
+            "present.test",
+            vec!["@ IN A 10.20.30.41"],
+            vec![],
+        ))?]
+        .into();
+        let catalog = build_catalog(zones);
+
+        let (rcode, _message) = lookup_message(&catalog, "absent.test.", RecordType::A).await?;
+        assert_eq!(rcode, ResponseCode::Refused);
+
+        Ok(())
+    }
+
+    #[tokio::test]
+    async fn catalog_lookup_forwards_on_nameexists() -> anyhow::Result<()> {
+        let upstream = Zone::try_from(&zone_data(
+            "forward-aaaa.test",
+            vec!["host 60 IN AAAA 2001:db8::1"],
+            vec![],
+        ))?;
+
+        let mut upstream_catalog = Catalog::new();
+        upstream_catalog.upsert(
+            upstream.origin.clone(),
+            vec![upstream.create_memory_zone_handler().unwrap()],
+        );
+
+        let socket = UdpSocket::bind("127.0.0.1:0").await?;
+        let upstream_addr = socket.local_addr()?;
+
+        let mut server = Server::new(upstream_catalog);
+        server.register_socket(socket);
+        let upstream_handle = tokio::spawn(async move {
+            let _ = server.block_until_done().await;
+        });
+
+        let zones: ZoneGroup = vec![Zone::try_from(&zone_data(
+            "forward-aaaa.test",
+            vec!["host IN A 10.20.30.40"],
+            vec![&format!("udp://{}", upstream_addr)],
+        ))?]
+        .into();
+        let catalog = build_catalog(zones);
+
+        let (rcode, message) =
+            lookup_message(&catalog, "host.forward-aaaa.test.", RecordType::AAAA).await?;
+        assert_eq!(rcode, ResponseCode::NoError);
+
+        let message = message.expect("response should exist");
+        assert!(has_aaaa_answer(&message, "2001:db8::1".parse()?));
+
+        upstream_handle.abort();
+        let _ = upstream_handle.await;
+
+        Ok(())
+    }
+
+    #[tokio::test]
+    async fn catalog_lookup_forwards_on_nxdomain() -> anyhow::Result<()> {
+        let upstream = Zone::try_from(&zone_data(
+            "forward-nxdomain.test",
+            vec!["missing 60 IN A 203.0.113.55"],
+            vec![],
+        ))?;
+
+        let mut upstream_catalog = Catalog::new();
+        upstream_catalog.upsert(
+            upstream.origin.clone(),
+            vec![upstream.create_memory_zone_handler().unwrap()],
+        );
+
+        let socket = UdpSocket::bind("127.0.0.1:0").await?;
+        let upstream_addr = socket.local_addr()?;
+
+        let mut server = Server::new(upstream_catalog);
+        server.register_socket(socket);
+        let upstream_handle = tokio::spawn(async move {
+            let _ = server.block_until_done().await;
+        });
+
+        let zones: ZoneGroup = vec![Zone::try_from(&zone_data(
+            "forward-nxdomain.test",
+            vec!["present IN A 10.20.30.41"],
+            vec![&format!("udp://{}", upstream_addr)],
+        ))?]
+        .into();
+        let catalog = build_catalog(zones);
+
+        let (rcode, message) =
+            lookup_message(&catalog, "missing.forward-nxdomain.test.", RecordType::A).await?;
+        assert_eq!(rcode, ResponseCode::NoError);
+
+        let message = message.expect("response should exist");
+        assert!(has_a_answer(&message, Ipv4Addr::new(203, 0, 113, 55)));
+
+        upstream_handle.abort();
+        let _ = upstream_handle.await;
+
+        Ok(())
+    }
+
+    #[tokio::test]
+    async fn catalog_lookup_falls_back_to_later_zone_handler_with_same_origin() -> anyhow::Result<()>
+    {
+        let zones: ZoneGroup = vec![
+            // First matching zone exists but does not contain the queried name.
+            Zone::try_from(&zone_data(
+                "fallback.test",
+                vec!["first IN A 10.20.30.1"],
+                vec![],
+            ))?,
+            // Second matching zone should be queried as fallback and answer.
+            Zone::try_from(&zone_data(
+                "fallback.test",
+                vec!["target IN A 10.20.30.2"],
+                vec![],
+            ))?,
+        ]
+        .into();
+        let catalog = build_catalog(zones);
+
+        let (rcode, message) =
+            lookup_message(&catalog, "target.fallback.test.", RecordType::A).await?;
+        assert_eq!(rcode, ResponseCode::NoError);
+        assert!(has_a_answer(
+            &message.expect("response should exist"),
+            Ipv4Addr::new(10, 20, 30, 2)
+        ));
+
+        Ok(())
+    }
+
+    #[tokio::test]
+    async fn catalog_lookup_does_not_fall_back_when_fallthrough_disabled() -> anyhow::Result<()> {
+        let zones: ZoneGroup = vec![
+            Zone::try_from(&zone_data_with_fallthrough(
+                "fallback-disabled.test",
+                vec!["first IN A 10.20.31.1"],
+                vec![],
+                hashset! {},
+            ))?,
+            Zone::try_from(&zone_data_with_fallthrough(
+                "fallback-disabled.test",
+                vec!["target IN A 10.20.31.2"],
+                vec![],
+                hashset! {},
+            ))?,
+        ]
+        .into();
+        let catalog = build_catalog(zones);
+
+        let (rcode, message) =
+            lookup_message(&catalog, "target.fallback-disabled.test.", RecordType::A).await?;
+
+        assert_ne!(rcode, ResponseCode::NoError);
+        if let Some(message) = message.as_ref() {
+            assert!(!has_a_answer(message, Ipv4Addr::new(10, 20, 31, 2)));
+        }
+
+        Ok(())
+    }
+
+    #[tokio::test]
+    async fn catalog_forward_only_zone_queries_upstream() -> anyhow::Result<()> {
+        let (upstream_addr, upstream_handle) = start_upstream_server().await?;
+
+        let forward_zone = Zone::try_from(&zone_data(
+            "upstream.test",
+            vec![],
+            vec![&format!("udp://{}", upstream_addr)],
+        ))?;
+        let catalog = build_catalog(vec![forward_zone].into());
+
+        let (rcode, message) =
+            lookup_message(&catalog, "from-forward.upstream.test.", RecordType::A).await?;
+        assert_eq!(rcode, ResponseCode::NoError);
+        assert!(has_a_answer(
+            &message.expect("response should exist"),
+            Ipv4Addr::new(203, 0, 113, 9)
+        ));
+
+        upstream_handle.abort();
+        let _ = upstream_handle.await;
+
+        Ok(())
+    }
+}

easytier/src/easytier-cli.rs

@@ -193,8 +193,11 @@ struct PeerArgs {
 
 #[derive(Subcommand, Debug)]
 enum PeerSubCommand {
+    /// List connected peers
     List,
+    /// Show public IPv6 address information
     Ipv6,
+    /// List foreign networks discovered by this instance
     ListForeign {
         #[arg(
             long,
@@ -203,6 +206,7 @@ enum PeerSubCommand {
         )]
         trusted_keys: bool,
     },
+    /// List global foreign networks from the peer center
     ListGlobalForeign,
 }
 
@@ -214,16 +218,18 @@ struct RouteArgs {
 
 #[derive(Subcommand, Debug)]
 enum RouteSubCommand {
+    /// List routes propagated by peers
     List,
+    /// Dump routes in CIDR format
     Dump,
 }
 
 #[derive(Args, Debug)]
 struct ConnectorArgs {
-    #[arg(short, long)]
+    #[arg(short, long, help = "filter connectors by virtual IPv4 address")]
     ipv4: Option<String>,
 
-    #[arg(short, long)]
+    #[arg(short, long, help = "filter connectors by peer URL")]
     peers: Vec<String>,
 
     #[command(subcommand)]
@@ -242,6 +248,7 @@ enum ConnectorSubCommand {
         #[arg(help = "connector url, e.g., tcp://1.2.3.4:11010")]
         url: String,
     },
+    /// List connectors
     List,
 }
 
@@ -283,6 +290,7 @@ struct AclArgs {
 
 #[derive(Subcommand, Debug)]
 enum AclSubCommand {
+    /// Show ACL rule hit statistics
     Stats,
 }
 
@@ -450,19 +458,25 @@ struct InstallArgs {
     #[arg(long, default_value = env!("CARGO_PKG_DESCRIPTION"), help = "service description")]
     description: String,
 
-    #[arg(long)]
+    #[arg(long, help = "display name shown by the service manager")]
     display_name: Option<String>,
 
-    #[arg(long)]
+    #[arg(
+        long,
+        help = "whether to disable starting the service automatically on boot (true/false)"
+    )]
     disable_autostart: Option<bool>,
 
-    #[arg(long)]
+    #[arg(
+        long,
+        help = "whether to disable automatic restart when the service fails (true/false)"
+    )]
     disable_restart_on_failure: Option<bool>,
 
     #[arg(long, help = "path to easytier-core binary")]
     core_path: Option<PathBuf>,
 
-    #[arg(long)]
+    #[arg(long, help = "working directory for the easytier-core service")]
     service_work_dir: Option<PathBuf>,
 
     #[arg(

easytier/src/gateway/kcp_proxy.rs

@@ -4,7 +4,7 @@ use std::{
     time::Duration,
 };
 
-use anyhow::Context;
+use anyhow::{Context, anyhow, bail};
 use bytes::Bytes;
 use dashmap::DashMap;
 use guarden::defer;
@@ -15,12 +15,13 @@ use kcp_sys::{
     stream::KcpStream,
 };
 use prost::Message;
-use tokio::{select, task::JoinSet};
+use tokio::task::JoinSet;
 
 use super::{
     CidrSet,
     tcp_proxy::{NatDstConnector, NatDstTcpConnector, TcpProxy},
 };
+use crate::utils::task::HedgeExt;
 use crate::{
     common::{
         acl_processor::PacketInfo,
@@ -114,72 +115,57 @@ pub struct NatDstKcpConnector {
 impl NatDstConnector for NatDstKcpConnector {
     type DstStream = KcpStream;
 
-    async fn connect(&self, src: SocketAddr, nat_dst: SocketAddr) -> Result<Self::DstStream> {
+    async fn connect(
+        &self,
+        src: SocketAddr,
+        nat_dst: SocketAddr,
+    ) -> anyhow::Result<Self::DstStream> {
+        let peer_mgr = self
+            .peer_mgr
+            .upgrade()
+            .ok_or_else(|| anyhow!("peer manager is not available"))?;
+
+        let dst_peer = {
+            let SocketAddr::V4(addr) = nat_dst else {
+                bail!("ipv6 is not supported");
+            };
+            peer_mgr
+                .get_peer_map()
+                .get_peer_id_by_ipv4(addr.ip())
+                .await
+                .ok_or_else(|| anyhow!("no peer found for nat dst: {}", nat_dst))?
+        };
+
+        tracing::trace!(?nat_dst, ?dst_peer, "kcp nat");
+
         let conn_data = KcpConnData {
             src: Some(src.into()),
             dst: Some(nat_dst.into()),
         };
 
-        let Some(peer_mgr) = self.peer_mgr.upgrade() else {
-            return Err(anyhow::anyhow!("peer manager is not available").into());
-        };
+        let stream = (0..5)
+            .map(|_| {
+                let kcp_endpoint = self.kcp_endpoint.clone();
+                let my_peer_id = peer_mgr.my_peer_id();
 
-        let dst_peer_id = match nat_dst {
-            SocketAddr::V4(addr) => peer_mgr.get_peer_map().get_peer_id_by_ipv4(addr.ip()).await,
-            SocketAddr::V6(_) => return Err(anyhow::anyhow!("ipv6 is not supported").into()),
-        };
+                async move {
+                    let conn_id = kcp_endpoint
+                        .connect(
+                            Duration::from_secs(10),
+                            my_peer_id,
+                            dst_peer,
+                            Bytes::from(conn_data.encode_to_vec()),
+                        )
+                        .await?;
 
-        let Some(dst_peer) = dst_peer_id else {
-            return Err(anyhow::anyhow!("no peer found for nat dst: {}", nat_dst).into());
-        };
-
-        tracing::trace!("kcp nat dst: {:?}, dst peers: {:?}", nat_dst, dst_peer);
-
-        let mut connect_tasks: JoinSet<std::result::Result<ConnId, anyhow::Error>> = JoinSet::new();
-        let mut retry_remain = 5;
-        loop {
-            select! {
-                Some(Ok(Ok(ret))) = connect_tasks.join_next() => {
-                    // just wait for the previous connection to finish
-                    let stream = KcpStream::new(&self.kcp_endpoint, ret)
-                        .ok_or(anyhow::anyhow!("failed to create kcp stream"))?;
-                    return Ok(stream);
+                    KcpStream::new(&kcp_endpoint, conn_id).context("failed to create kcp stream")
                 }
-                _ = tokio::time::sleep(Duration::from_millis(200)), if !connect_tasks.is_empty() && retry_remain > 0 => {
-                    // no successful connection yet, trigger another connection attempt
-                }
-                else => {
-                    // got error in connect_tasks, continue to retry
-                    if retry_remain == 0 && connect_tasks.is_empty() {
-                        break;
-                    }
-                }
-            }
+            })
+            .hedge(Duration::from_millis(200))
+            .await
+            .context("failed to connect to peer")?;
 
-            // create a new connection task
-            if retry_remain == 0 {
-                continue;
-            }
-            retry_remain -= 1;
-
-            let kcp_endpoint = self.kcp_endpoint.clone();
-            let my_peer_id = peer_mgr.my_peer_id();
-            let conn_data_clone = conn_data;
-
-            connect_tasks.spawn(async move {
-                kcp_endpoint
-                    .connect(
-                        Duration::from_secs(10),
-                        my_peer_id,
-                        dst_peer,
-                        Bytes::from(conn_data_clone.encode_to_vec()),
-                    )
-                    .await
-                    .with_context(|| format!("failed to connect to nat dst: {}", nat_dst))
-            });
-        }
-
-        Err(anyhow::anyhow!("failed to connect to nat dst: {}", nat_dst).into())
+        Ok(stream)
     }
 
     fn check_packet_from_peer_fast(&self, _cidr_set: &CidrSet, _global_ctx: &GlobalCtx) -> bool {

easytier/src/gateway/quic_proxy.rs

@@ -18,17 +18,20 @@ use crate::tunnel::packet_def::{
     PacketType, PeerManagerHeader, TAIL_RESERVED_SIZE, ZCPacket, ZCPacketType,
 };
 use crate::tunnel::quic::{client_config, endpoint_config, server_config};
-use anyhow::{Context, Error, anyhow};
+use crate::utils::task::HedgeExt;
+use anyhow::{Context, Error, anyhow, bail, ensure};
 use atomic_refcell::AtomicRefCell;
 use bytes::{BufMut, Bytes, BytesMut};
 use dashmap::DashMap;
 use derivative::Derivative;
 use derive_more::{Constructor, Deref, DerefMut, From, Into};
 use guarden::defer;
+use moka::future::Cache;
 use prost::Message;
 use quinn::udp::{EcnCodepoint, RecvMeta, Transmit};
 use quinn::{
-    AsyncUdpSocket, Endpoint, RecvStream, SendStream, StreamId, UdpPoller, default_runtime,
+    AsyncUdpSocket, Connection, ConnectionError, Endpoint, RecvStream, SendStream, StreamId,
+    UdpPoller, WriteError, default_runtime,
 };
 use std::cmp::min;
 use std::future::Future;
@@ -43,8 +46,8 @@ use tokio::io::{AsyncReadExt, Join, join};
 use tokio::sync::mpsc::error::TrySendError;
 use tokio::sync::mpsc::{Receiver, Sender, channel};
 use tokio::task::JoinSet;
-use tokio::time::{Instant, timeout};
-use tokio::{join, pin, select};
+use tokio::time::timeout;
+use tokio::{join, select};
 use tokio_util::sync::PollSender;
 use tracing::{debug, error, info, instrument, trace, warn};
 
@@ -279,6 +282,7 @@ impl From<(SendStream, RecvStream)> for QuicStream {
 pub struct NatDstQuicConnector {
     pub(crate) endpoint: Endpoint,
     pub(crate) peer_mgr: Weak<PeerManager>,
+    pub(crate) conn_map: Cache<PeerId, Connection>,
 }
 
 #[async_trait::async_trait]
@@ -289,21 +293,25 @@ impl NatDstConnector for NatDstQuicConnector {
         &self,
         src: SocketAddr,
         nat_dst: SocketAddr,
-    ) -> crate::common::error::Result<Self::DstStream> {
-        let Some(peer_mgr) = self.peer_mgr.upgrade() else {
-            return Err(anyhow::anyhow!("peer manager is not available").into());
+    ) -> anyhow::Result<Self::DstStream> {
+        let peer_mgr = self
+            .peer_mgr
+            .upgrade()
+            .ok_or_else(|| anyhow!("peer manager is not available"))?;
+
+        let dst_peer = {
+            let SocketAddr::V4(addr) = nat_dst else {
+                bail!("ipv6 is not supported");
+            };
+            peer_mgr
+                .get_peer_map()
+                .get_peer_id_by_ipv4(addr.ip())
+                .await
+                .ok_or_else(|| anyhow!("no peer found for nat dst: {}", nat_dst))?
         };
 
-        let Some(dst_peer_id) = (match nat_dst {
-            SocketAddr::V4(addr) => peer_mgr.get_peer_map().get_peer_id_by_ipv4(addr.ip()).await,
-            SocketAddr::V6(_) => return Err(anyhow::anyhow!("ipv6 is not supported").into()),
-        }) else {
-            return Err(anyhow::anyhow!("no peer found for nat dst: {}", nat_dst).into());
-        };
+        tracing::trace!(?nat_dst, ?dst_peer, "quic nat");
 
-        trace!("quic nat dst: {:?}, dst peers: {:?}", nat_dst, dst_peer_id);
-
-        let addr = QuicAddr::new(dst_peer_id, PacketType::QuicSrc).into();
         let header = {
             let conn_data = QuicConnData {
                 src: Some(src.into()),
@@ -311,61 +319,90 @@ impl NatDstConnector for NatDstQuicConnector {
             };
 
             let len = conn_data.encoded_len();
-            if len > (u16::MAX as usize) {
-                return Err(anyhow!("conn data too large: {:?}", len).into());
-            }
+            ensure!(len <= u16::MAX as usize, "conn data too large: {len}");
 
             let mut buf = BytesMut::with_capacity(2 + len);
 
             buf.put_u16(len as u16);
-            conn_data.encode(&mut buf).unwrap();
+            conn_data.encode(&mut buf)?;
 
             buf.freeze()
         };
 
-        let mut connect_tasks = JoinSet::<Result<QuicStream, Error>>::new();
-        let connect = |tasks: &mut JoinSet<_>| {
-            let endpoint = self.endpoint.clone();
-            let header = header.clone();
+        let reconnect = || async move {
+            self.conn_map.invalidate(&dst_peer).await;
 
-            tasks.spawn(async move {
-                let connection = endpoint.connect(addr, "")?.await?;
-                let mut stream: QuicStream = connection.open_bi().await?.into();
-                stream.writer_mut().write_chunk(header).await?;
-                Ok(stream)
-            });
+            let connect = (0..5)
+                .map(|_| {
+                    let endpoint = self.endpoint.clone();
+                    async move {
+                        endpoint
+                            .connect(QuicAddr::new(dst_peer, PacketType::QuicSrc).into(), "")
+                            .context("failed to create connection")?
+                            .await
+                            .context("connection failed")
+                    }
+                })
+                .hedge(Duration::from_millis(200));
+
+            self.conn_map
+                .try_get_with(dst_peer, connect)
+                .await
+                .context("failed to connect to peer")
         };
 
-        connect(&mut connect_tasks);
+        let mut reconnected = false;
 
-        let timer = tokio::time::sleep(Duration::from_millis(200));
-        pin!(timer);
+        let mut connection = if let Some(connection) = self.conn_map.get(&dst_peer).await
+            && connection.close_reason().is_none()
+        {
+            connection
+        } else {
+            reconnected = true;
+            reconnect().await?
+        };
 
-        let mut retry_remain = 5;
         loop {
-            select! {
-                Some(result) = connect_tasks.join_next() => {
-                    match result {
-                        Ok(Ok(stream)) => return Ok(stream.into()),
-                        _ => {
-                            if connect_tasks.is_empty() {
-                                if retry_remain == 0 {
-                                    return Err(anyhow!("failed to connect to nat dst: {:?}", nat_dst).into())
-                                }
+            let is_retryable = |error: &ConnectionError| {
+                matches!(
+                    error,
+                    ConnectionError::ConnectionClosed(_)
+                        | ConnectionError::ApplicationClosed(_)
+                        | ConnectionError::Reset
+                        | ConnectionError::TimedOut
+                )
+            };
+            let mut retry = !reconnected;
+            let header = header.clone();
+            let result = async {
+                let mut stream: QuicStream = connection
+                    .open_bi()
+                    .await
+                    .inspect_err(|error| retry &= is_retryable(error))?
+                    .into();
+                stream
+                    .writer_mut()
+                    .write_chunk(header)
+                    .await
+                    .inspect_err(|error| {
+                        retry &= matches!(error, WriteError::ConnectionLost(error) if is_retryable(error))
+                    })?;
+                Ok(stream.into())
+            }
+                .await;
 
-                                retry_remain -= 1;
-                                connect(&mut connect_tasks);
-                                timer.as_mut().reset(Instant::now() + Duration::from_millis(200))
-                            }
-                        }
-                    }
-                }
-                _ = &mut timer, if retry_remain > 0 => {
-                    retry_remain -= 1;
-                    connect(&mut connect_tasks);
-                    timer.as_mut().reset(Instant::now() + Duration::from_millis(200));
+            if let Err(error) = &result {
+                if retry {
+                    debug!(?error, "failed to open quic stream, retrying...");
+                    reconnected = true;
+                    connection = reconnect().await?;
+                    continue;
+                } else {
+                    self.conn_map.invalidate(&dst_peer).await;
                 }
             }
+
+            break result;
         }
     }
 
@@ -595,10 +632,17 @@ impl QuicStreamReceiver {
                                         }
                                     };
 
-                                    match Self::establish_stream(stream, ctx.clone()).await {
-                                        Ok(stream) => drop(tasks.spawn(stream)),
-                                        Err(e) => warn!("failed to establish quic stream from {:?}: {:?}", connection.remote_address(), e),
-                                    }
+                                    let ctx = ctx.clone();
+                                    tasks.spawn(async move {
+                                        match Self::establish_stream(stream, ctx).await {
+                                            Ok(transfer_fut) => {
+                                                if let Err(e) = transfer_fut.await {
+                                                    warn!("quic stream transfer error: {:?}", e);
+                                                }
+                                            }
+                                            Err(e) => warn!("failed to establish quic stream: {:?}", e),
+                                        }
+                                    });
                                 }
 
                                 res = tasks.join_next(), if !tasks.is_empty() => {
@@ -816,7 +860,7 @@ impl QuicProxy {
             Arc::new(socket),
             default_runtime().unwrap(),
         )
-        .unwrap();
+        .unwrap(); // TODO: maybe a different transport config
         endpoint.set_default_client_config(client_config());
         self.endpoint = Some(endpoint.clone());
 
@@ -845,6 +889,10 @@ impl QuicProxy {
                 NatDstQuicConnector {
                     endpoint: endpoint.clone(),
                     peer_mgr: Arc::downgrade(&peer_mgr),
+                    conn_map: Cache::builder()
+                        .max_capacity(u8::MAX.into()) // cf. quinn transport config (max_concurrent_bidi_streams)
+                        .time_to_idle(Duration::from_secs(600)) // cf. quinn transport config (max_idle_timeout)
+                        .build(),
                 },
             ));
 

easytier/src/gateway/socks5.rs

@@ -240,7 +240,7 @@ impl AsyncTcpConnector for Socks5KcpConnector {
         let ret = c
             .connect(self.src_addr, addr)
             .await
-            .map_err(|e| super::fast_socks5::SocksError::Other(e.into()))?;
+            .map_err(super::fast_socks5::SocksError::Other)?;
         Ok(SocksTcpStream::Kcp(ret))
     }
 }

easytier/src/gateway/tcp_proxy.rs

@@ -44,7 +44,7 @@ use super::tokio_smoltcp::{self, Net, NetConfig, channel_device};
 pub(crate) trait NatDstConnector: Send + Sync + Clone + 'static {
     type DstStream: AsyncRead + AsyncWrite + Unpin + Send;
 
-    async fn connect(&self, src: SocketAddr, dst: SocketAddr) -> Result<Self::DstStream>;
+    async fn connect(&self, src: SocketAddr, dst: SocketAddr) -> anyhow::Result<Self::DstStream>;
     fn check_packet_from_peer_fast(&self, cidr_set: &CidrSet, global_ctx: &GlobalCtx) -> bool;
     fn check_packet_from_peer(
         &self,
@@ -63,14 +63,13 @@ pub struct NatDstTcpConnector;
 #[async_trait::async_trait]
 impl NatDstConnector for NatDstTcpConnector {
     type DstStream = TcpStream;
-    async fn connect(&self, _src: SocketAddr, nat_dst: SocketAddr) -> Result<Self::DstStream> {
-        let socket = match TcpSocket::new_v4() {
-            Ok(s) => s,
-            Err(error) => {
-                log::error!(?error, "create v4 socket failed");
-                return Err(error.into());
-            }
-        };
+    async fn connect(
+        &self,
+        _src: SocketAddr,
+        nat_dst: SocketAddr,
+    ) -> anyhow::Result<Self::DstStream> {
+        let socket = TcpSocket::new_v4()
+            .inspect_err(|error| log::error!(?error, "create v4 socket failed"))?;
 
         let stream = timeout(Duration::from_secs(10), socket.connect(nat_dst))
             .await?

easytier/src/instance/dns_server/client_instance.rs

@@ -1,111 +0,0 @@
-use std::{sync::Arc, time::Duration};
-
-use tokio::task::JoinSet;
-
-use crate::{
-    peers::peer_manager::PeerManager,
-    proto::{
-        api::instance::Route,
-        common::Void,
-        magic_dns::{
-            HandshakeRequest, MagicDnsServerRpc, MagicDnsServerRpcClientFactory,
-            UpdateDnsRecordRequest,
-        },
-        rpc_impl::standalone::StandAloneClient,
-        rpc_types::controller::BaseController,
-    },
-    tunnel::tcp::TcpTunnelConnector,
-};
-
-use super::MAGIC_DNS_INSTANCE_ADDR;
-
-pub struct MagicDnsClientInstance {
-    rpc_client: StandAloneClient<TcpTunnelConnector>,
-    rpc_stub: Option<Box<dyn MagicDnsServerRpc<Controller = BaseController> + Send>>,
-    peer_mgr: Arc<PeerManager>,
-    tasks: JoinSet<()>,
-}
-
-impl MagicDnsClientInstance {
-    pub async fn new(peer_mgr: Arc<PeerManager>) -> Result<Self, anyhow::Error> {
-        let tcp_connector = TcpTunnelConnector::new(MAGIC_DNS_INSTANCE_ADDR.parse().unwrap());
-        let mut rpc_client = StandAloneClient::new(tcp_connector);
-        let rpc_stub = rpc_client
-            .scoped_client::<MagicDnsServerRpcClientFactory<BaseController>>("".to_string())
-            .await?;
-        Ok(MagicDnsClientInstance {
-            rpc_client,
-            rpc_stub: Some(rpc_stub),
-            peer_mgr,
-            tasks: JoinSet::new(),
-        })
-    }
-
-    async fn update_dns_task(
-        peer_mgr: Arc<PeerManager>,
-        rpc_stub: Box<dyn MagicDnsServerRpc<Controller = BaseController> + Send>,
-    ) -> Result<(), anyhow::Error> {
-        let mut prev_last_update = None;
-        rpc_stub
-            .handshake(BaseController::default(), HandshakeRequest::default())
-            .await?;
-        loop {
-            rpc_stub
-                .heartbeat(BaseController::default(), Void::default())
-                .await?;
-
-            let last_update = peer_mgr.get_route_peer_info_last_update_time().await;
-            if Some(last_update) == prev_last_update {
-                tokio::time::sleep(Duration::from_millis(500)).await;
-                continue;
-            }
-
-            let mut routes = peer_mgr.list_routes().await;
-            // add self as a route
-            let ctx = peer_mgr.get_global_ctx();
-            routes.push(Route {
-                hostname: ctx.get_hostname(),
-                ipv4_addr: ctx.get_ipv4().map(Into::into),
-                ..Default::default()
-            });
-            // Use configured tld_dns_zone (always set by default)
-            let flags = ctx.config.get_flags();
-            let req = UpdateDnsRecordRequest {
-                routes,
-                zone: flags.tld_dns_zone.clone(),
-            };
-            tracing::debug!(
-                "MagicDnsClientInstance::update_dns_task: update dns records: {:?}",
-                req
-            );
-            rpc_stub
-                .update_dns_record(BaseController::default(), req)
-                .await?;
-
-            let last_update_after_rpc = peer_mgr.get_route_peer_info_last_update_time().await;
-            if last_update_after_rpc == last_update {
-                prev_last_update = Some(last_update);
-            }
-        }
-    }
-
-    pub async fn run_and_wait(&mut self) {
-        let rpc_stub = self.rpc_stub.take().unwrap();
-        let peer_mgr = self.peer_mgr.clone();
-        self.tasks.spawn(async move {
-            let ret = Self::update_dns_task(peer_mgr, rpc_stub).await;
-            if let Err(e) = ret {
-                tracing::error!("MagicDnsServerInstanceData::run_and_wait: {:?}", e);
-            }
-        });
-
-        tokio::select! {
-            _ = self.tasks.join_next() => {
-                tracing::warn!("MagicDnsServerInstanceData::run_and_wait: dns record update task exited");
-            }
-            _ = self.rpc_client.wait() => {
-                tracing::warn!("MagicDnsServerInstanceData::run_and_wait: rpc client exited");
-            }
-        }
-    }
-}

easytier/src/instance/dns_server/config.rs

@@ -1,193 +0,0 @@
-use hickory_proto::rr;
-use hickory_proto::rr::RData;
-use serde::{Deserialize, Serialize};
-use std::collections::HashMap;
-use std::net::{IpAddr, Ipv4Addr};
-use std::str::FromStr;
-use std::time::Duration;
-
-#[derive(Serialize, Deserialize, Debug, Clone, derive_builder::Builder)]
-pub struct RunConfig {
-    general: GeneralConfig,
-
-    #[builder(default = HashMap::new())]
-    zones: Zone,
-
-    #[builder(default = Vec::new())]
-    #[serde(default)]
-    excluded_forward_nameservers: Vec<IpAddr>,
-}
-
-impl RunConfig {
-    pub fn general(&self) -> &GeneralConfig {
-        &self.general
-    }
-
-    pub fn zones(&self) -> &Zone {
-        &self.zones
-    }
-
-    pub fn excluded_forward_nameservers(&self) -> &Vec<IpAddr> {
-        &self.excluded_forward_nameservers
-    }
-}
-
-#[derive(Serialize, Deserialize, Debug, Clone, derive_builder::Builder)]
-pub struct GeneralConfig {
-    #[builder(setter(into, strip_option), default = None)]
-    listen_tcp: Option<String>,
-
-    #[builder(setter(into, strip_option), default = None)]
-    listen_udp: Option<String>,
-}
-
-impl GeneralConfig {
-    pub fn listen_tcp(&self) -> &Option<String> {
-        &self.listen_tcp
-    }
-
-    pub fn listen_udp(&self) -> &Option<String> {
-        &self.listen_udp
-    }
-}
-
-pub type Zone = HashMap<String, Vec<Record>>; // domain -> records
-
-pub type RecordType = rr::RecordType;
-
-#[derive(Serialize, Deserialize, Debug, Clone, derive_builder::Builder)]
-pub struct Record {
-    #[serde(rename = "type")]
-    rr_type: RecordType,
-
-    name: String,
-    value: String,
-
-    #[serde(with = "humantime_serde")]
-    ttl: Duration,
-}
-
-impl Record {
-    pub fn name(&self) -> anyhow::Result<rr::Name> {
-        let name = rr::Name::from_str(self.name.as_str())?;
-        Ok(name)
-    }
-
-    fn rr_type(&self) -> rr::RecordType {
-        self.rr_type
-    }
-}
-
-impl TryFrom<Record> for rr::Record {
-    type Error = anyhow::Error;
-
-    fn try_from(value: Record) -> Result<Self, Self::Error> {
-        let r: rr::Record = (&value).try_into()?;
-        Ok(r)
-    }
-}
-
-impl TryFrom<&Record> for rr::Record {
-    type Error = anyhow::Error;
-
-    fn try_from(value: &Record) -> Result<Self, Self::Error> {
-        let name = value.name()?;
-        let mut record = Self::update0(name, value.ttl.as_secs() as u32, value.rr_type());
-        record.set_dns_class(rr::DNSClass::IN);
-        match value.rr_type {
-            RecordType::A => {
-                let addr: Ipv4Addr = value.value.parse()?;
-                record.set_data(RData::A(rr::rdata::a::A(addr)));
-            }
-            RecordType::SOA => {
-                let soa = value.value.split_whitespace().collect::<Vec<_>>();
-                if soa.len() != 7 {
-                    return Err(anyhow::anyhow!("invalid SOA record"));
-                }
-                let mname = rr::Name::from_str(soa[0])?;
-                let rname = rr::Name::from_str(soa[1])?;
-                let serial: u32 = soa[2].parse()?;
-                let refresh: u32 = soa[3].parse()?;
-                let retry: u32 = soa[4].parse()?;
-                let expire: u32 = soa[5].parse()?;
-                let minimum: u32 = soa[6].parse()?;
-                record.set_data(RData::SOA(rr::rdata::soa::SOA::new(
-                    mname,
-                    rname,
-                    serial,
-                    refresh.try_into().unwrap(),
-                    retry.try_into().unwrap(),
-                    expire.try_into().unwrap(),
-                    minimum,
-                )));
-            }
-            _ => todo!(),
-        }
-        Ok(record)
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-    use anyhow::anyhow;
-
-    #[tokio::test]
-    async fn it_works() -> anyhow::Result<()> {
-        let text = r#"
-[general]
-listen_tcp = "127.0.0.1:5300"
-listen_udp = "127.0.0.1:5353"
-
-[[zones."et.internal"]]
-type = "A"
-name = "www"
-value = "123.123.123.123"
-ttl = "60s"
-
-[[zones."et.top"]]
-type = "A"
-name = "@"
-value = "100.100.100.100"
-ttl = "61s"
-
-"#;
-
-        let config = toml::from_str::<RunConfig>(text)?;
-        assert_eq!(
-            config.general.listen_tcp().clone().unwrap(),
-            "127.0.0.1:5300"
-        );
-        assert_eq!(
-            config.general.listen_udp().clone().unwrap(),
-            "127.0.0.1:5353"
-        );
-        assert_eq!(config.zones.len(), 2);
-
-        let (domain, records) = config
-            .zones
-            .get_key_value("et.internal")
-            .ok_or(anyhow!("et.internal not found"))?;
-        assert_eq!(domain, "et.internal");
-        assert_eq!(records.len(), 1);
-        let record = &records[0];
-        assert_eq!(record.rr_type, RecordType::A);
-        assert_eq!(record.name, "www");
-        assert_eq!(record.value, "123.123.123.123");
-        assert_eq!(record.ttl.as_secs(), 60);
-
-        let (domain, records) = config
-            .zones
-            .get_key_value("et.top")
-            .ok_or(anyhow!("et.top not found"))?;
-        assert_eq!(domain, "et.top");
-        assert_eq!(records.len(), 1);
-        let record = &records[0];
-        assert_eq!(record.rr_type, RecordType::A);
-        assert_eq!(record.name, "@");
-        assert_eq!(record.value, "100.100.100.100");
-        assert_eq!(record.ttl.as_secs(), 61);
-
-        Ok(())
-    }
-}

easytier/src/instance/dns_server/mod.rs

@@ -1,21 +0,0 @@
-// This module is copy and modified from https://github.com/fanyang89/libdns
-#[cfg(feature = "magic-dns")]
-pub(crate) mod config;
-#[cfg(feature = "magic-dns")]
-pub(crate) mod server;
-
-#[cfg(feature = "magic-dns")]
-pub mod client_instance;
-#[cfg(feature = "magic-dns")]
-pub mod runner;
-#[cfg(feature = "magic-dns")]
-pub mod server_instance;
-#[cfg(feature = "magic-dns")]
-pub mod system_config;
-
-#[cfg(all(test, feature = "tun", feature = "magic-dns"))]
-mod tests;
-
-pub static MAGIC_DNS_INSTANCE_ADDR: &str = "tcp://127.0.0.1:49813";
-pub static MAGIC_DNS_FAKE_IP: &str = "100.100.100.101";
-pub static DEFAULT_ET_DNS_ZONE: &str = "et.net.";

easytier/src/instance/dns_server/runner.rs

@@ -1,93 +0,0 @@
-use cidr::Ipv4Inet;
-use tokio_util::sync::CancellationToken;
-
-use crate::peers::peer_manager::PeerManager;
-use std::{net::Ipv4Addr, sync::Arc, time::Duration};
-
-use super::{client_instance::MagicDnsClientInstance, server_instance::MagicDnsServerInstance};
-
-static DEFAULT_ET_DNS_ZONE: &str = "et.net.";
-
-pub struct DnsRunner {
-    client: Option<MagicDnsClientInstance>,
-    server: Option<MagicDnsServerInstance>,
-    peer_mgr: Arc<PeerManager>,
-    tun_dev: Option<String>,
-    tun_inet: Ipv4Inet,
-    fake_ip: Ipv4Addr,
-}
-
-impl DnsRunner {
-    pub fn new(
-        peer_mgr: Arc<PeerManager>,
-        tun_dev: Option<String>,
-        tun_inet: Ipv4Inet,
-        fake_ip: Ipv4Addr,
-    ) -> Self {
-        Self {
-            client: None,
-            server: None,
-            peer_mgr,
-            tun_dev,
-            tun_inet,
-            fake_ip,
-        }
-    }
-
-    async fn clean_env(&mut self) {
-        if let Some(server) = self.server.take() {
-            server.clean_env().await;
-        }
-        self.client.take();
-    }
-
-    async fn run_once(&mut self) -> anyhow::Result<()> {
-        // try server first
-        match MagicDnsServerInstance::new(
-            self.peer_mgr.clone(),
-            self.tun_dev.clone(),
-            self.tun_inet,
-            self.fake_ip,
-        )
-        .await
-        {
-            Ok(server) => {
-                self.server = Some(server);
-                tracing::info!("DnsRunner::run_once: server started");
-            }
-            Err(e) => {
-                tracing::error!("DnsRunner::run_once: {:?}", e);
-            }
-        }
-
-        // every runner must run a client
-        let client = MagicDnsClientInstance::new(self.peer_mgr.clone()).await?;
-        self.client = Some(client);
-        self.client.as_mut().unwrap().run_and_wait().await;
-
-        Err(anyhow::anyhow!("Client instance exit"))
-    }
-
-    pub async fn run(&mut self, canel_token: CancellationToken) {
-        loop {
-            tracing::info!("DnsRunner::run: start");
-            tokio::select! {
-                _ = canel_token.cancelled() => {
-                    self.clean_env().await;
-                    tracing::info!("DnsRunner::run: cancelled");
-                    return;
-                }
-
-                ret = self.run_once() => {
-                    self.clean_env().await;
-                    if let Err(e) = ret {
-                        tracing::error!("DnsRunner::run: {:?}", e);
-                    } else {
-                        tracing::info!("DnsRunner::run: unexpected exit, server may be down");
-                    }
-                    tokio::time::sleep(Duration::from_millis(500)).await;
-                }
-            }
-        }
-    }
-}

easytier/src/instance/dns_server/server.rs

@@ -1,338 +0,0 @@
-use anyhow::{Context, Result};
-use hickory_proto::op::Edns;
-use hickory_proto::rr;
-use hickory_proto::rr::LowerName;
-use hickory_resolver::config::ResolverOpts;
-use hickory_resolver::name_server::TokioConnectionProvider;
-use hickory_resolver::system_conf::read_system_conf;
-use hickory_server::ServerFuture;
-use hickory_server::authority::{AuthorityObject, Catalog, ZoneType};
-use hickory_server::server::{Request, RequestHandler, ResponseHandler, ResponseInfo};
-use hickory_server::store::forwarder::ForwardConfig;
-use hickory_server::store::{forwarder::ForwardAuthority, in_memory::InMemoryAuthority};
-use std::io;
-use std::net::SocketAddr;
-use std::str::FromStr;
-use std::sync::Arc;
-use std::time::Duration;
-use tokio::net::{TcpListener, UdpSocket};
-use tokio::sync::{RwLock, RwLockReadGuard, RwLockWriteGuard};
-use tokio::task::JoinSet;
-
-use crate::common::dns::get_default_resolver_config;
-
-use super::config::{GeneralConfig, Record, RunConfig};
-
-pub struct Server {
-    server: ServerFuture<CatalogRequestHandler>,
-    catalog: Arc<RwLock<Catalog>>,
-    general_config: GeneralConfig,
-    udp_local_addr: Option<SocketAddr>,
-    tcp_local_addr: Option<SocketAddr>,
-    tasks: JoinSet<()>,
-}
-
-struct CatalogRequestHandler {
-    catalog: Arc<RwLock<Catalog>>,
-}
-
-impl CatalogRequestHandler {
-    fn new(catalog: Arc<RwLock<Catalog>>) -> CatalogRequestHandler {
-        // let system_conf = read_system_conf();
-        // let recursor = match system_conf {
-        //     Ok((conf, _)) => RecursorBuilder::default().build(conf),
-        //     Err(_) => RecursorBuilder::default().build(get_default_resolver_config()),
-        // }
-        // // policy is security unware, this will never return an error
-        // .unwrap();
-
-        Self { catalog }
-    }
-}
-
-#[async_trait::async_trait]
-impl RequestHandler for CatalogRequestHandler {
-    async fn handle_request<R: ResponseHandler>(
-        &self,
-        request: &Request,
-        response_handle: R,
-    ) -> ResponseInfo {
-        self.catalog
-            .read()
-            .await
-            .handle_request(request, response_handle)
-            .await
-    }
-}
-
-pub fn build_authority(domain: &str, records: &[Record]) -> Result<InMemoryAuthority> {
-    let zone = rr::Name::from_str(domain)?;
-    let mut authority = InMemoryAuthority::empty(zone, ZoneType::Primary, false);
-    for record in records.iter() {
-        let r = record.try_into()?;
-        authority.upsert_mut(r, 0);
-    }
-    Ok(authority)
-}
-
-impl Server {
-    pub fn new(config: RunConfig) -> Self {
-        Self::try_new(config).unwrap()
-    }
-
-    fn try_new(config: RunConfig) -> Result<Self> {
-        let mut catalog = Catalog::new();
-        for (domain, records) in config.zones().iter() {
-            let zone = rr::Name::from_str(domain.as_str())?;
-            let authroty = build_authority(domain, records)?;
-            catalog.upsert(zone.clone().into(), vec![Arc::new(authroty)]);
-        }
-
-        // use forwarder authority for the root zone
-        let system_conf =
-            read_system_conf().unwrap_or((get_default_resolver_config(), ResolverOpts::default()));
-        let forward_config = ForwardConfig {
-            name_servers: system_conf
-                .0
-                .name_servers()
-                .iter()
-                .filter(|&x| {
-                    !config
-                        .excluded_forward_nameservers()
-                        .contains(&x.socket_addr.ip())
-                })
-                .cloned()
-                .collect::<Vec<_>>()
-                .into(),
-            options: Some(system_conf.1),
-        };
-        let auth = ForwardAuthority::builder_with_config(
-            forward_config,
-            TokioConnectionProvider::default(),
-        )
-        .build()
-        .unwrap();
-
-        catalog.upsert(rr::Name::from_str(".")?.into(), vec![Arc::new(auth)]);
-
-        let catalog = Arc::new(RwLock::new(catalog));
-        let handler = CatalogRequestHandler::new(catalog.clone());
-        let server = ServerFuture::new(handler);
-
-        Ok(Self {
-            server,
-            catalog,
-            general_config: config.general().clone(),
-            udp_local_addr: None,
-            tcp_local_addr: None,
-            tasks: JoinSet::new(),
-        })
-    }
-
-    pub fn udp_local_addr(&self) -> Option<SocketAddr> {
-        self.udp_local_addr
-    }
-
-    pub fn tcp_local_addr(&self) -> Option<SocketAddr> {
-        self.tcp_local_addr
-    }
-
-    pub async fn register_udp_socket(&mut self, address: String) -> Result<SocketAddr> {
-        let bind_addr = SocketAddr::from_str(&address)
-            .with_context(|| format!("DNS Server failed to parse address {}", address))?;
-        let socket = socket2::Socket::new(
-            socket2::Domain::IPV4,
-            socket2::Type::DGRAM,
-            Some(socket2::Protocol::UDP),
-        )
-        .with_context(|| {
-            format!(
-                "DNS Server failed to create UDP socket for address {}",
-                address
-            )
-        })?;
-        socket2::SockRef::from(&socket)
-            .set_reuse_address(true)
-            .with_context(|| {
-                format!(
-                    "DNS Server failed to set reuse address on socket {}",
-                    address
-                )
-            })?;
-        socket.bind(&bind_addr.into()).with_context(|| {
-            format!("DNS Server failed to bind socket to address {}", bind_addr)
-        })?;
-        socket
-            .set_nonblocking(true)
-            .with_context(|| "DNS Server failed to set socket to non-blocking".to_string())?;
-        let socket = UdpSocket::from_std(socket.into()).with_context(|| {
-            format!(
-                "DNS Server failed to convert socket to UdpSocket for address {}",
-                address
-            )
-        })?;
-
-        let local_addr = socket
-            .local_addr()
-            .with_context(|| "DNS Server failed to get local address".to_string())?;
-        self.server.register_socket(socket);
-
-        Ok(local_addr)
-    }
-
-    pub async fn run(&mut self) -> Result<()> {
-        if let Some(address) = self.general_config.listen_tcp() {
-            let tcp_listener = TcpListener::bind(address.clone())
-                .await
-                .with_context(|| format!("DNS Server failed to bind TCP address {}", address))?;
-            self.tcp_local_addr = Some(tcp_listener.local_addr()?);
-            self.server
-                .register_listener(tcp_listener, Duration::from_secs(5));
-        }
-
-        if let Some(address) = self.general_config.listen_udp() {
-            let local_addr = self.register_udp_socket(address.clone()).await?;
-            self.udp_local_addr = Some(local_addr);
-        };
-
-        Ok(())
-    }
-
-    pub async fn shutdown(&mut self) -> Result<()> {
-        self.server.shutdown_gracefully().await?;
-        Ok(())
-    }
-
-    pub async fn upsert(&self, name: LowerName, authority: Arc<dyn AuthorityObject>) {
-        self.catalog.write().await.upsert(name, vec![authority]);
-    }
-
-    pub async fn remove(&self, name: &LowerName) -> Option<Vec<Arc<dyn AuthorityObject>>> {
-        self.catalog.write().await.remove(name)
-    }
-
-    pub async fn update<R: ResponseHandler>(
-        &self,
-        update: &Request,
-        response_edns: Option<Edns>,
-        response_handle: R,
-    ) -> io::Result<ResponseInfo> {
-        self.catalog
-            .write()
-            .await
-            .update(update, response_edns, response_handle)
-            .await
-    }
-
-    pub async fn contains(&self, name: &LowerName) -> bool {
-        self.catalog.read().await.contains(name)
-    }
-
-    pub async fn lookup<R: ResponseHandler>(
-        &self,
-        request: &Request,
-        response_edns: Option<Edns>,
-        response_handle: R,
-    ) -> ResponseInfo {
-        self.catalog
-            .read()
-            .await
-            .lookup(request, response_edns, response_handle)
-            .await
-    }
-
-    pub async fn read_catalog(&self) -> RwLockReadGuard<'_, Catalog> {
-        self.catalog.read().await
-    }
-
-    pub async fn write_catalog(&self) -> RwLockWriteGuard<'_, Catalog> {
-        self.catalog.write().await
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-    use crate::instance::dns_server::config::{
-        GeneralConfigBuilder, RecordBuilder, RecordType, RunConfigBuilder,
-    };
-    use anyhow::Result;
-    use hickory_client::client::{Client, ClientHandle};
-    use hickory_proto::rr;
-    use hickory_proto::runtime::TokioRuntimeProvider;
-    use hickory_proto::udp::UdpClientStream;
-    use maplit::hashmap;
-    use std::time::Duration;
-
-    #[tokio::test]
-    async fn it_works() -> Result<()> {
-        let mut server = Server::new(
-            RunConfigBuilder::default()
-                .general(GeneralConfigBuilder::default().build()?)
-                .build()?,
-        );
-        server.run().await?;
-        server.shutdown().await?;
-        Ok(())
-    }
-
-    #[tokio::test]
-    async fn can_resolve_records() -> Result<()> {
-        let configured_record = RecordBuilder::default()
-            .rr_type(RecordType::A)
-            .name("www.et.internal.".to_string())
-            .value("123.123.123.123".to_string())
-            .ttl(Duration::from_secs(60))
-            .build()?;
-        let configured_record2 = RecordBuilder::default()
-            .rr_type(RecordType::A)
-            .name("中文.et.internal.".to_string())
-            .value("123.123.123.123".to_string())
-            .ttl(Duration::from_secs(60))
-            .build()?;
-        let soa_record = RecordBuilder::default()
-            .rr_type(RecordType::SOA)
-            .name("et.internal.".to_string())
-            .value(
-                "ns.et.internal. hostmaster.et.internal. 2023101001 7200 3600 1209600 86400"
-                    .to_string(),
-            )
-            .ttl(Duration::from_secs(60))
-            .build()?;
-        let config = RunConfigBuilder::default()
-            .general(
-                GeneralConfigBuilder::default()
-                    .listen_udp("127.0.0.1:0")
-                    .build()?,
-            )
-            .zones(hashmap! {
-                "et.internal.".to_string() => vec![configured_record.clone(), soa_record.clone(), configured_record2.clone()],
-            })
-            .build()?;
-
-        let mut server = Server::new(config);
-        server.run().await?;
-
-        let local_addr = server.udp_local_addr().unwrap();
-        let stream = UdpClientStream::builder(local_addr, TokioRuntimeProvider::default()).build();
-        let (mut client, background) = Client::connect(stream).await?;
-        let background_task = tokio::spawn(background);
-        let response = client
-            .query(
-                rr::Name::from_str("www.et.internal")?,
-                rr::DNSClass::IN,
-                rr::RecordType::A,
-            )
-            .await?;
-        drop(background_task);
-
-        println!("Response: {:?}", response);
-
-        assert_eq!(response.answers().len(), 1);
-        let expected_record: rr::Record = configured_record.try_into()?;
-        assert_eq!(response.answers().first().unwrap(), &expected_record);
-
-        server.shutdown().await?;
-        Ok(())
-    }
-}