chore: release v0.4.4 on master

A Scheduled Task stores an absolute path. Checking that a task named
"OSTP TUN Helper" exists said nothing about whether its <Command> still points
at the helper we are about to run, and the paths do drift: a dev build
registers target\debug\ostp-tun-helper.exe, an installer registers Program
Files, and moving or reinstalling the app leaves the old path behind.

That failed silently in the worst way. schtasks /Run reports success for
merely ACCEPTING the launch request — a task whose exe no longer exists fails
afterwards, out of band, with nothing returned to us. So launch_as_admin
returned Ok, and the caller then sat in its 60-second connect loop before
reporting "Timeout connecting to helper." On every connect, permanently, with
no way out except deleting the task by hand.

The check now reads the registered <Command> back and compares it to the exe,
re-registering through the existing /F overwrite when they differ: one consent
prompt, once, instead of a permanent silent breakage.

The path is read via /Query /XML rather than /FO LIST /V because the list
format's field labels are localized — "Task To Run" is "Задача для запуска" on
a Russian Windows — while XML tag names are not. schtasks emits UTF-16LE with
a BOM there, which is decoded explicitly, with UTF-8 tolerated as a fallback.
Both paths are canonicalized before comparison so casing, `..` and 8.3 short
names do not read as a mismatch; a path that cannot be canonicalized no longer
exists, which is itself grounds to re-register.

.github/workflows/release.yml

@@ -1,29 +1,151 @@
-name: Universal CI/CD Release Matrix
+name: CI/CD
+
+
+# `run-name` is evaluated at workflow-start, BEFORE any job runs - it cannot
+# see resolve-channel's computed tag_name (e.g. "0.4.3-alpha"), only the
+# `github.*` context. The old "release version ${{ github.ref_name }}" showed
+# the bare branch name ("alpha"/"beta") for every run, which reads
+# exactly like a literal release tag and caused real confusion - the actual
+# release tag has been correct (versioned) all along; only this label lied
+# about it. Spell out "channel" so nobody mistakes one for the other again.
+# NOTE: this value MUST be quoted. The GHA string literal below contains
+# "Release build: {0}" - an unquoted YAML plain scalar treats ": " (colon
+# then space) as starting a nested mapping, which is exactly what broke every
+# single push since this line was introduced: GitHub rejected the whole
+# workflow file at parse time (before any job runs), silently burning an
+# Actions-minutes-billed run per push for nothing.
+run-name: "${{ startsWith(github.ref, 'refs/tags/') && (contains(github.ref_name, 'beta') && format('CI/CD: beta version {0}', github.ref_name) || contains(github.ref_name, 'alpha') && format('CI/CD: alpha version {0}', github.ref_name) || format('CI/CD: release version {0}', github.ref_name)) || format('CI/CD: {0} channel build', github.ref_name) }}"
 
 on:
   push:
     tags:
       - "v*"
   workflow_dispatch:
+    inputs:
+      channel:
+        description: >-
+          Manually build+release just this rolling channel. Stable releases
+          are NEVER picked here on purpose - cut those only via a real
+          "vX.Y.Z" tag push, so a manual dispatch can't accidentally publish
+          a "stable" release.
+        type: choice
+        required: true
+        default: alpha
+        options:
+          - alpha
+          - beta
 
 permissions:
   contents: write
 
-# ── Global defaults ─────────────────────────────────────────────────────────
+# -- Global defaults ---------------------------------------------------------
 env:
   CARGO_TERM_COLOR: always
   CARGO_INCREMENTAL: 0
   RUST_BACKTRACE: short
 
 jobs:
+  # Computes ONE channel + release tag for this whole run, so every build
+  # job (native matrix + all 3 GUI platforms + Android) uploads to the exact
+  # same release under the exact same tag, instead of repeating this logic
+  # (and risking it drifting out of sync) in five separate places.
+  #
+  # Tag shape:
+  #   - real "vX.Y.Z" / "vX.Y.Z-beta.N" tag push -> tag used as-is (stable promotion)
+  #   - push to `alpha`                        -> "{version}-alpha" (rolling, same tag every push)
+  #   - push to `beta`                            -> "{version}-beta"    (rolling, same tag every push)
+  #   - workflow_dispatch                         -> forced by the `channel` input (alpha|beta only)
+  resolve-channel:
+    name: Resolve release channel
+    runs-on: ubuntu-latest
+    outputs:
+      channel: ${{ steps.resolve.outputs.channel }}
+      tag_name: ${{ steps.resolve.outputs.tag_name }}
+      prerelease: ${{ steps.resolve.outputs.prerelease }}
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Resolve channel, version, and release tag
+        id: resolve
+        shell: bash
+        run: |
+          set -euo pipefail
+          BASE_VERSION=$(grep -m1 '^version' Cargo.toml | sed -E 's/version *= *"([^"]+)"/\1/')
+
+          if [[ "${{ github.ref }}" == refs/tags/v* ]]; then
+            # A pushed tag is authoritative — use it AS-IS (never recompute it
+            # from Cargo.toml, or the release would upload to a different tag than
+            # the one that triggered this run). The channel, and thus prerelease,
+            # is decided by the tag's suffix: v0.4.7-beta / v0.4.7-alpha are
+            # prereleases; a bare vX.Y.Z is the only thing that becomes stable.
+            TAG="${{ github.ref_name }}"
+            case "$TAG" in
+              *-alpha*) CHANNEL="alpha" ;;
+              *-beta*)  CHANNEL="beta"  ;;
+              *)        CHANNEL="stable" ;;
+            esac
+          else
+            # No tag (workflow_dispatch, or a legacy branch push): pick the
+            # channel, then synthesize the rolling tag from Cargo.toml's version.
+            if [ "${{ github.event_name }}" = "workflow_dispatch" ]; then
+              CHANNEL="${{ github.event.inputs.channel }}"
+            elif [ "${{ github.ref_name }}" = "beta" ]; then
+              CHANNEL="beta"
+            else
+              CHANNEL="alpha"
+            fi
+            TAG="v${BASE_VERSION}-${CHANNEL}"
+          fi
+
+          echo "Resolved channel=$CHANNEL tag=$TAG (base version $BASE_VERSION)"
+          echo "channel=$CHANNEL" >> "$GITHUB_OUTPUT"
+          echo "tag_name=$TAG" >> "$GITHUB_OUTPUT"
+          echo "prerelease=$([ "$CHANNEL" = "stable" ] && echo false || echo true)" >> "$GITHUB_OUTPUT"
+
+  check-and-test:
+    name: Check & Test
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+
+      - name: Setup Rust toolchain
+        uses: dtolnay/rust-toolchain@stable
+        with:
+          toolchain: stable
+
+      - name: Restore Cargo cache
+        uses: actions/cache@v4
+        with:
+          path: |
+            ~/.cargo/registry/index/
+            ~/.cargo/registry/cache/
+            ~/.cargo/git/db/
+            target/
+          key: cargo-check-${{ hashFiles('**/Cargo.lock') }}
+          restore-keys: cargo-check-
+
+      - name: Install musl-tools
+        run: sudo apt-get update && sudo apt-get install -y musl-tools
+
+      - name: Create dummy dist for rust-embed
+        run: mkdir -p ostp-control/dist && touch ostp-control/dist/index.html
+
+      - name: cargo check
+        run: cargo check --workspace
+
+      - name: cargo test
+        run: cargo test --workspace --lib
+
   publish-release-matrix:
     name: Release for ${{ matrix.target }}
+    needs: [check-and-test, resolve-channel]
     runs-on: ${{ matrix.os }}
     strategy:
       fail-fast: false
       matrix:
         include:
-          # ── Windows ──────────────────────────────────────────────────────
+          # -- Windows ------------------------------------------------------
           - os: windows-latest
             target: x86_64-pc-windows-msvc
             artifact_name: ostp.exe
@@ -42,7 +164,7 @@ jobs:
             release_name: ostp-windows-arm64.zip
             wintun_arch: arm64
 
-          # ── macOS ─────────────────────────────────────────────────────────
+          # -- macOS ---------------------------------------------------------
           - os: macos-latest
             target: x86_64-apple-darwin
             artifact_name: ostp
@@ -53,7 +175,7 @@ jobs:
             artifact_name: ostp
             release_name: ostp-darwin-arm64.tar.gz
 
-          # ── Linux native ──────────────────────────────────────────────────
+          # -- Linux native --------------------------------------------------
           - os: ubuntu-latest
             target: x86_64-unknown-linux-musl
             artifact_name: ostp
@@ -65,7 +187,7 @@ jobs:
             release_name: ostp-linux-386.tar.gz
             use_cross: true
 
-          # ── Linux cross ───────────────────────────────────────────────────
+          # -- Linux cross ---------------------------------------------------
           - os: ubuntu-latest
             target: aarch64-unknown-linux-musl
             artifact_name: ostp
@@ -103,25 +225,31 @@ jobs:
       - name: Checkout code
         uses: actions/checkout@v4
 
-      # ── Frontend Build ─────────────────────────────────────────────────────
+      # -- Frontend Build -----------------------------------------------------
       - name: Setup Node.js
         uses: actions/setup-node@v4
         with:
           node-version: 20
-      - name: Build Web Panel
-        working-directory: ostp-control
+      - name: Build Web Panel (skip if no source; use committed dist/)
+        shell: bash
         run: |
-          npm install
-          npm run build
+          mkdir -p ostp-control/dist
+          cd ostp-control
+          if [ -f package.json ]; then
+            npm install && npm run build
+          else
+            echo "ostp-control has no package.json - using committed dist/"
+            [ -f dist/index.html ] || echo '<!doctype html><title>OSTP</title>' > dist/index.html
+          fi
 
-      # ── Rust toolchain ─────────────────────────────────────────────────────
+      # -- Rust toolchain -----------------------------------------------------
       - name: Setup Rust toolchain
         uses: dtolnay/rust-toolchain@stable
         with:
           toolchain: ${{ matrix.toolchain || 'stable' }}
           targets: ${{ !matrix.use_cross && matrix.target || '' }}
 
-      # ── Cargo cache (shared per target) ───────────────────────────────────
+      # -- Cargo cache (shared per target) -----------------------------------
       - name: Restore Cargo cache
         uses: actions/cache@v4
         with:
@@ -134,18 +262,18 @@ jobs:
           restore-keys: |
             cargo-${{ matrix.target }}-
 
-      # ── MUSL tools for native Linux musl builds ────────────────────────────
+      # -- MUSL tools for native Linux musl builds ----------------------------
       - name: Install musl-tools
         if: ${{ matrix.os == 'ubuntu-latest' && !matrix.use_cross }}
         run: sudo apt-get update && sudo apt-get install -y musl-tools
 
-      # ── Native build ───────────────────────────────────────────────────────
+      # -- Native build -------------------------------------------------------
       - name: Build (native)
         if: ${{ !matrix.use_cross }}
         shell: bash
         run: cargo build --release --target ${{ matrix.target }} --bin ostp
 
-      # ── Cross build ────────────────────────────────────────────────────────
+      # -- Cross build --------------------------------------------------------
       - name: Restore cross binary cache
         if: ${{ matrix.use_cross }}
         id: cross-cache
@@ -156,13 +284,21 @@ jobs:
 
       - name: Install cross (if not cached)
         if: ${{ matrix.use_cross && steps.cross-cache.outputs.cache-hit != 'true' }}
-        run: cargo install cross --git https://github.com/cross-rs/cross.git --locked
+        # cross-rs's own source (not ours, not a dependency of ours) uses a
+        # macro-at-end-of-block pattern that trips rustc's
+        # semicolon_in_expressions_from_macros lint on current toolchains -
+        # harmless in cross's actual behavior, but `cargo install` compiles
+        # the installed package as the "local" crate, so dependency lint
+        # capping doesn't shield it. --cap-lints=warn is the standard escape
+        # hatch for building a third-party tool against a newer compiler than
+        # its own lint config assumed; it doesn't touch our own build.
+        run: RUSTFLAGS="--cap-lints=warn" cargo install cross --git https://github.com/cross-rs/cross.git --locked
 
       - name: Build (cross)
         if: ${{ matrix.use_cross }}
         run: cross build --release --target ${{ matrix.target }} --bin ostp
 
-      # ── Driver dependencies ────────────────────────────────────────────────
+      # -- Driver dependencies ------------------------------------------------
       - name: Download wintun (Windows)
         if: ${{ matrix.os == 'windows-latest' }}
         shell: pwsh
@@ -174,7 +310,7 @@ jobs:
           Get-ChildItem "$dir/wt_tmp" -Filter "wintun.dll" -Recurse | Where-Object { $_.FullName -match 'bin[\\/]${{ matrix.wintun_arch }}[\\/]' } | Copy-Item -Destination "$dir/"
           Remove-Item "$dir/wt.zip","$dir/wt_tmp" -Recurse -Force
 
-      # ── Package ────────────────────────────────────────────────────────────
+      # -- Package ------------------------------------------------------------
       - name: Package (Windows)
         if: ${{ matrix.os == 'windows-latest' }}
         shell: pwsh
@@ -193,17 +329,23 @@ jobs:
           FILES="${{ matrix.artifact_name }}"
           tar -czf "${{ matrix.release_name }}" -C "$dir" $FILES
 
-      # ── Upload ─────────────────────────────────────────────────────────────
+      # -- Upload -------------------------------------------------------------
       - name: Upload to GitHub Release
-        if: ${{ startsWith(github.ref, 'refs/tags/') }}
         uses: softprops/action-gh-release@v2
         with:
+          # Computed once in resolve-channel so every platform/job in this run
+          # lands on the exact same tag: "{version}-alpha" / "{version}-beta"
+          # for rolling channel pushes, or the pushed "vX.Y.Z" tag as-is for a
+          # real stable release.
+          tag_name: ${{ needs.resolve-channel.outputs.tag_name }}
+          prerelease: ${{ needs.resolve-channel.outputs.prerelease }}
           files: ${{ matrix.release_name }}
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
 
   build-windows-gui:
     name: Build Windows GUI (Tauri) - ${{ matrix.arch }}
+    needs: [check-and-test, resolve-channel]
     runs-on: windows-latest
     strategy:
       matrix:
@@ -236,7 +378,15 @@ jobs:
             ~/.cargo/registry/cache/
             ~/.cargo/git/db/
             target/
+            ostp-gui/src-tauri/target/
           key: cargo-windows-gui-${{ matrix.target }}-${{ hashFiles('**/Cargo.lock') }}
+          # Without a prefix fallback this cache NEVER restored on a release:
+          # cutting a release rewrites every Cargo.lock (version bump), which
+          # changes hashFiles(), which misses the exact key — so each release
+          # rebuilt every dependency from scratch. That is why the GUI jobs ran
+          # 2-4x longer than the plain release targets, which had this all along.
+          restore-keys: |
+            cargo-windows-gui-${{ matrix.target }}-
 
       - name: Download wintun
         shell: pwsh
@@ -268,15 +418,21 @@ jobs:
           Compress-Archive -Path "$dir/*" -DestinationPath "ostp-windows-gui-${{ matrix.arch }}.zip" -Force
 
       - name: Upload to GitHub Release
-        if: ${{ startsWith(github.ref, 'refs/tags/') }}
         uses: softprops/action-gh-release@v2
         with:
+          # Computed once in resolve-channel so every platform/job in this run
+          # lands on the exact same tag: "{version}-alpha" / "{version}-beta"
+          # for rolling channel pushes, or the pushed "vX.Y.Z" tag as-is for a
+          # real stable release.
+          tag_name: ${{ needs.resolve-channel.outputs.tag_name }}
+          prerelease: ${{ needs.resolve-channel.outputs.prerelease }}
           files: ostp-windows-gui-${{ matrix.arch }}.zip
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
 
   build-linux-gui:
     name: Build Linux GUI (Tauri) - ${{ matrix.arch }}
+    needs: [check-and-test, resolve-channel]
     runs-on: ubuntu-latest
     strategy:
       matrix:
@@ -312,30 +468,46 @@ jobs:
             ~/.cargo/registry/cache/
             ~/.cargo/git/db/
             target/
+            ostp-gui/src-tauri/target/
           key: cargo-linux-gui-${{ matrix.target }}-${{ hashFiles('**/Cargo.lock') }}
+          restore-keys: |
+            cargo-linux-gui-${{ matrix.target }}-
 
       - name: Build Tauri App
         working-directory: ostp-gui
         run: |
           npm install
+          # TUN mode shells out to this helper, elevated via pkexec. Only the
+          # Windows job used to build it, so the Linux package shipped without
+          # it and TUN could never start.
+          cargo build -p ostp-tun-helper --release --target ${{ matrix.target }} --manifest-path ../Cargo.toml
           npx tauri build --no-bundle --target ${{ matrix.target }}
 
       - name: Package Portable Tarball
         run: |
+          set -euo pipefail
           mkdir ostp-linux-gui-${{ matrix.arch }}
           cp ostp-gui/src-tauri/target/${{ matrix.target }}/release/ostp-gui ostp-linux-gui-${{ matrix.arch }}/
+          # The GUI looks for the helper next to its own executable first.
+          cp target/${{ matrix.target }}/release/ostp-tun-helper ostp-linux-gui-${{ matrix.arch }}/
           tar -czf ostp-linux-gui-${{ matrix.arch }}.tar.gz ostp-linux-gui-${{ matrix.arch }}
 
       - name: Upload to GitHub Release
-        if: ${{ startsWith(github.ref, 'refs/tags/') }}
         uses: softprops/action-gh-release@v2
         with:
+          # Computed once in resolve-channel so every platform/job in this run
+          # lands on the exact same tag: "{version}-alpha" / "{version}-beta"
+          # for rolling channel pushes, or the pushed "vX.Y.Z" tag as-is for a
+          # real stable release.
+          tag_name: ${{ needs.resolve-channel.outputs.tag_name }}
+          prerelease: ${{ needs.resolve-channel.outputs.prerelease }}
           files: ostp-linux-gui-${{ matrix.arch }}.tar.gz
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
 
   build-macos-gui:
     name: Build macOS GUI (Tauri) - ${{ matrix.arch }}
+    needs: [check-and-test, resolve-channel]
     runs-on: macos-latest
     strategy:
       matrix:
@@ -368,7 +540,10 @@ jobs:
             ~/.cargo/registry/cache/
             ~/.cargo/git/db/
             target/
+            ostp-gui/src-tauri/target/
           key: cargo-macos-gui-${{ matrix.target }}-${{ hashFiles('**/Cargo.lock') }}
+          restore-keys: |
+            cargo-macos-gui-${{ matrix.target }}-
 
       - name: Build Tauri App
         working-directory: ostp-gui
@@ -383,15 +558,21 @@ jobs:
           tar -czf ostp-macos-gui-${{ matrix.arch }}.tar.gz ostp-macos-gui-${{ matrix.arch }}
 
       - name: Upload to GitHub Release
-        if: ${{ startsWith(github.ref, 'refs/tags/') }}
         uses: softprops/action-gh-release@v2
         with:
+          # Computed once in resolve-channel so every platform/job in this run
+          # lands on the exact same tag: "{version}-alpha" / "{version}-beta"
+          # for rolling channel pushes, or the pushed "vX.Y.Z" tag as-is for a
+          # real stable release.
+          tag_name: ${{ needs.resolve-channel.outputs.tag_name }}
+          prerelease: ${{ needs.resolve-channel.outputs.prerelease }}
           files: ostp-macos-gui-${{ matrix.arch }}.tar.gz
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
 
   build-android:
     name: Build Android Client (Flutter) - ${{ matrix.arch }}
+    needs: [check-and-test, resolve-channel]
     runs-on: ubuntu-latest
     strategy:
       matrix:
@@ -402,7 +583,6 @@ jobs:
           - arch: armeabi-v7a
             rust_target: armv7-linux-androideabi
             flutter_target: android-arm
-            tun2socks_arch: linux-armv7
     steps:
       - uses: actions/checkout@v4
 
@@ -428,32 +608,117 @@ jobs:
         with:
           ndk-version: r26b
 
-      - name: Install cargo-ndk
-        run: cargo install cargo-ndk
+      # The Android jobs had no Rust caching at all, so every release recompiled
+      # the whole ostp-jni dependency graph from scratch — the main reason these
+      # were among the slowest jobs in the matrix.
+      - name: Cache cargo
+        uses: actions/cache@v4
+        with:
+          path: |
+            ~/.cargo/registry/index/
+            ~/.cargo/registry/cache/
+            ~/.cargo/git/db/
+            target/
+          key: cargo-android-${{ matrix.arch }}-${{ hashFiles('**/Cargo.lock') }}
+          restore-keys: |
+            cargo-android-${{ matrix.arch }}-
+
+      # cargo-ndk was built from source on every run. Cache the binary the same
+      # way the cross-compilation jobs already cache `cross`.
+      - name: Restore cargo-ndk binary cache
+        id: cargo-ndk-cache
+        uses: actions/cache@v4
+        with:
+          path: ~/.cargo/bin/cargo-ndk
+          key: cargo-ndk-bin-${{ runner.os }}-v1
+
+      - name: Install cargo-ndk (if not cached)
+        if: steps.cargo-ndk-cache.outputs.cache-hit != 'true'
+        run: cargo install cargo-ndk --locked
 
       - name: Build Android APK
         shell: bash
         working-directory: ostp-flutter
+        env:
+          OSTP_KEYSTORE_B64: ${{ secrets.ANDROID_KEYSTORE_BASE64 }}
+          OSTP_KEYSTORE_PASSWORD: ${{ secrets.ANDROID_KEYSTORE_PASSWORD }}
+          OSTP_KEY_ALIAS: ${{ secrets.ANDROID_KEY_ALIAS }}
+          OSTP_KEY_PASSWORD: ${{ secrets.ANDROID_KEY_PASSWORD }}
         run: |
-          # 1. Compile JNI
+          set -euo pipefail
+
+          # 1. Materialise the upload keystore from secrets. Android keys an app
+          #    by applicationId + signing key and refuses to update across a key
+          #    change, so every published build MUST use this one key. Releases
+          #    used to fall through to the per-machine debug keystore, which on
+          #    ephemeral CI runners meant a different random key every build -
+          #    hence "App not installed" on upgrade.
+          if [ -z "${OSTP_KEYSTORE_B64:-}" ]; then
+            echo "::error::ANDROID_KEYSTORE_BASE64 secret is not set. Refusing to publish a"
+            echo "::error::debug-signed APK: users could not update over it and the key is"
+            echo "::error::not reproducible. See docs for the one-time keystore setup."
+            exit 1
+          fi
+          export OSTP_KEYSTORE_PATH="$RUNNER_TEMP/ostp-upload.jks"
+          # Strip any stray CR/LF before decoding: the secret is pasted from a
+          # shell whose line endings we don't control, and a single trailing \r
+          # is enough to corrupt the decode.
+          printf '%s' "$OSTP_KEYSTORE_B64" | tr -d '\r\n' | base64 -d > "$OSTP_KEYSTORE_PATH"
+
+          # Verify the keystore opens BEFORE spending four minutes on Gradle only
+          # to fail at the packaging step. The size/SHA-256 are safe to print (a
+          # hash reveals nothing) and let the operator compare against the local
+          # file to tell a transport problem apart from a wrong password.
+          echo "keystore: $(stat -c%s "$OSTP_KEYSTORE_PATH") bytes, sha256 $(sha256sum "$OSTP_KEYSTORE_PATH" | cut -d' ' -f1)"
+          if ! keytool -list -keystore "$OSTP_KEYSTORE_PATH" \
+                 -storepass "$OSTP_KEYSTORE_PASSWORD" >/dev/null 2>&1; then
+            echo "::error::The keystore did not open with ANDROID_KEYSTORE_PASSWORD."
+            echo "::error::If the SHA-256 above matches your local ostp-upload.jks, the file"
+            echo "::error::arrived intact and the password secret itself is wrong - note that"
+            echo "::error::PowerShell expands \$ inside double quotes, so a password containing"
+            echo "::error::one gets mangled unless it was set with single quotes."
+            exit 1
+          fi
+          if ! keytool -list -keystore "$OSTP_KEYSTORE_PATH" \
+                 -storepass "$OSTP_KEYSTORE_PASSWORD" -alias "$OSTP_KEY_ALIAS" >/dev/null 2>&1; then
+            echo "::error::Keystore opened, but it has no key under ANDROID_KEY_ALIAS."
+            echo "::error::Aliases present in the keystore:"
+            keytool -list -keystore "$OSTP_KEYSTORE_PATH" -storepass "$OSTP_KEYSTORE_PASSWORD" \
+              | grep -i "PrivateKeyEntry" || true
+            exit 1
+          fi
+
+          # 2. Compile JNI
           mkdir -p android/app/src/main/jniLibs/${{ matrix.arch }}
-          
+
           cd ../ostp-jni
           cargo ndk -t ${{ matrix.arch }} -o "../ostp-flutter/android/app/src/main/jniLibs" build --release
           cd ../ostp-flutter
-          
 
-          
           # 3. Build Flutter APK
           flutter build apk --release --target-platform ${{ matrix.flutter_target }}
-          
-          # 4. Copy to output
-          cp build/app/outputs/flutter-apk/app-release.apk ostp-android-${{ matrix.arch }}.apk
+
+          # 4. Fail loudly if the APK somehow still came out debug-signed, rather
+          #    than shipping another un-updatable build.
+          APK=build/app/outputs/flutter-apk/app-release.apk
+          if "$ANDROID_HOME"/build-tools/*/apksigner verify --print-certs "$APK" 2>/dev/null \
+               | grep -qi "CN=Android Debug"; then
+            echo "::error::APK is signed with the Android debug certificate - aborting."
+            exit 1
+          fi
+
+          # 5. Copy to output
+          cp "$APK" ostp-android-${{ matrix.arch }}.apk
 
       - name: Upload to GitHub Release
-        if: ${{ startsWith(github.ref, 'refs/tags/') }}
         uses: softprops/action-gh-release@v2
         with:
+          # Computed once in resolve-channel so every platform/job in this run
+          # lands on the exact same tag: "{version}-alpha" / "{version}-beta"
+          # for rolling channel pushes, or the pushed "vX.Y.Z" tag as-is for a
+          # real stable release.
+          tag_name: ${{ needs.resolve-channel.outputs.tag_name }}
+          prerelease: ${{ needs.resolve-channel.outputs.prerelease }}
           files: ostp-flutter/ostp-android-${{ matrix.arch }}.apk
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.gitignore

@@ -5,6 +5,7 @@
 **/*.rs.bk
 .idea/
 .vscode/
+**/node_modules/
 
 # Binaries & libraries
 *.exe
@@ -25,6 +26,17 @@ test_route.ps1
 config.json
 wintun.dll
 
+# Android signing keys. The upload keystore is the ONE key every published APK
+# must be signed with (Android refuses to update an app across a key change),
+# so losing or leaking it is unrecoverable — it can never be committed.
+*.jks
+*.keystore
+key.properties
+
+# Server runtime cache (public IP autodetect) — must never be committed,
+# it's regenerated locally and leaks whatever host it ran on last.
+.ostp_public_ip
+
 # Logs
 *.log
 
@@ -34,5 +46,14 @@ turn-harvesting-idea.md
 
 # Private tooling (closed-source)
 ostp-prober/
+ostp-lab/
 
 ostp-brain/
+
+# Management panel built assets (built separately; dummy dist created for rust-embed build)
+ostp-control/
+
+.agents/
+netstack-smoltcp/
+dnstt/
+ostp-web/

.ostp_public_ip

@@ -1 +0,0 @@
-127.0.0.1

.release-state.json

@@ -0,0 +1,6 @@
+{
+  "target_version": "0.4.4",
+  "branch": "master",
+  "alpha_iteration": 0,
+  "beta_iteration": 0
+}

CONTRIBUTING.md

@@ -10,10 +10,12 @@ By contributing to this project, you agree to abide by our code of conduct and l
 
 1. [Development Setup](#development-setup)
 2. [Project Structure](#project-structure)
-3. [Development Workflow](#development-workflow)
-4. [Coding Guidelines](#coding-guidelines)
-5. [Submitting Pull Requests](#submitting-pull-requests)
-6. [Security Vulnerabilities](#security-vulnerabilities)
+3. [Branch Strategy](#branch-strategy)
+4. [Development Workflow](#development-workflow)
+5. [Commit Message Conventions](#commit-message-conventions)
+6. [Coding Guidelines](#coding-guidelines)
+7. [Submitting Pull Requests](#submitting-pull-requests)
+8. [Security Vulnerabilities](#security-vulnerabilities)
 
 ---
 
@@ -33,20 +35,19 @@ To build and test OSTP locally, you will need:
     cd ostp
     ```
 
-2.  **Build the control panel frontend**:
-    ```bash
-    cd ostp-control
-    npm install
-    npm run build
-    cd ..
-    ```
-
-3.  **Build the entire Cargo workspace**:
+2.  **Build the entire Cargo workspace**:
     ```bash
     cargo build
     ```
+    `ostp-control` (the web panel) is only needed if you're working on it
+    specifically - the server build embeds a dummy `dist/` via `rust-embed`
+    otherwise, so this step is not required for day-to-day core/client/server
+    work. If you *are* touching the panel:
+    ```bash
+    cd ostp-control && npm install && npm run build && cd ..
+    ```
 
-4.  **Run tests**:
+3.  **Run tests**:
     ```bash
     cargo test --workspace
     ```
@@ -58,7 +59,7 @@ To build and test OSTP locally, you will need:
 The repository is organized as a Cargo workspace containing the following crates:
 
 *   [`ostp-core/`](file:///d:/ospab-projects/ostp/ostp-core): Core protocol logic, including packet formatting, serialization, selective ACK/NACK (ARQ) state machine, and the Noise protocol (`Noise_NNpsk0_25519_ChaChaPoly_BLAKE2s`) handshake.
-*   [`ostp-client/`](file:///d:/ospab-projects/ostp/ostp-client): Client implementations, including SOCKS5/HTTP local proxies, `tun2socks` integration, native TUN interface routing, and split-tunneling bypass mechanisms.
+*   [`ostp-client/`](file:///d:/ospab-projects/ostp/ostp-client): Client implementations, including SOCKS5/HTTP local proxies, the native OSTP TUN interface routing, and split-tunneling bypass mechanisms.
 *   [`ostp-server/`](file:///d:/ospab-projects/ostp/ostp-server): Server logic, session dispatcher, anti-probing fallback server proxying, access key database, and the REST API for control panel communication.
 *   [`ostp-control/`](file:///d:/ospab-projects/ostp/ostp-control): A modern web dashboard for server administration (user management, real-time metrics, bandwidth limits).
 *   [`ostp-gui/`](file:///d:/ospab-projects/ostp/ostp-gui): Tauri-based desktop GUI application for Windows and Linux.
@@ -66,11 +67,28 @@ The repository is organized as a Cargo workspace containing the following crates
 
 ---
 
+## Branch Strategy
+
+The repository runs three long-lived branches, in increasing order of stability:
+
+| Branch | Role |
+|---|---|
+| `alpha` | Active development. All feature work and fixes land here first. |
+| `beta` | Periodically fast-forwarded from `alpha` once it's had some soak time. Ships as the `{version}-beta` release channel. |
+| `master` | Fast-forwarded from `beta` when it's proven stable. Real, tagged releases (`vX.Y.Z`) are cut from here. |
+
+`beta` and `master` are **never** committed to directly - they only ever move forward by fast-forwarding from the branch below them. This means promotion is always a plain `git merge` with zero conflicts by construction: don't `git merge`/rebase feature work directly onto `beta` or `master`.
+
+**Contributor PRs target `alpha`**, not `master`.
+
+---
+
 ## Development Workflow
 
 1.  **Check for existing issues** or open a new one to discuss proposed changes before starting work.
-2.  **Fork the repository** and create a new branch from `master`:
+2.  **Fork the repository** and create a new branch from `alpha`:
     ```bash
+    git checkout alpha
     git checkout -b feat/your-feature-name
     ```
 3.  **Implement your changes**, ensuring you write appropriate unit or integration tests.
@@ -89,6 +107,32 @@ The repository is organized as a Cargo workspace containing the following crates
 
 ---
 
+## Commit Message Conventions
+
+```
+<type>(<scope>): <short, imperative summary>
+
+<optional body - explain WHY, not what; the diff already shows what changed>
+```
+
+- **Type** - one of: `feat` (new capability), `fix` (bug fix), `docs`, `refactor` (no behavior change), `perf`, `test`, `chore` (deps/tooling/version bumps), `ci`, `security`.
+- **Scope** (optional) - the crate or area touched: `client`, `server`, `core`, `gui`, `flutter`, `ci`, `docs`, etc. e.g. `fix(client): ...`.
+- **Summary** - imperative mood ("add", not "added"/"adds"), no trailing period, ideally under ~70 characters.
+- **Body** - only when the *why* isn't obvious from the diff: a prior bug this fixes, a constraint that shaped the approach, a tradeoff you made. Don't restate what the diff already shows. Wrap at ~72 columns.
+
+```
+fix(server): drop junk frames by per-key marker instead of a global one
+
+A fixed 4-byte marker on every junk packet is itself a DPI signature any
+observer can filter on across every OSTP deployment. Derive the marker
+from the access key (HKDF, same scheme as obfuscation_key/psk) so it's
+per-user and indistinguishable from the packet's own random payload.
+```
+
+Multiple unrelated changes belong in separate commits, not one bundled commit - it keeps `git bisect` and review useful. Squash-merge is fine for a PR with a few "fix typo" / "address review" commits, but don't squash logically distinct changes together.
+
+---
+
 ## Coding Guidelines
 
 *   **Safety**: Avoid using `unsafe` blocks unless absolutely necessary for low-level system bindings (e.g., FFI configurations like `setsockopt`). When using `unsafe`, add safety doc comments explaining why it is safe.
@@ -104,7 +148,7 @@ The repository is organized as a Cargo workspace containing the following crates
     ```bash
     git push origin feat/your-feature-name
     ```
-2.  Open a Pull Request (PR) targeting the `master` branch.
+2.  Open a Pull Request (PR) targeting the `alpha` branch (see [Branch Strategy](#branch-strategy) - `master` only receives fast-forwards from `beta`, never direct PRs).
 3.  In your PR description, explain the rationale behind your changes, what was fixed/added, and how it was tested.
 4.  Verify that GitHub Actions CI runs successfully on your PR.
 

CONTRIBUTING.ru.md

@@ -10,10 +10,12 @@
 
 1. [Подготовка окружения](#подготовка-окружения)
 2. [Структура проекта](#структура-проекта)
-3. [Процесс разработки](#процесс-разработки)
-4. [Правила оформления кода](#правила-оформления-кода)
-5. [Создание Pull Request](#создание-pull-request)
-6. [Уязвимости безопасности](#уязвимости-безопасности)
+3. [Стратегия веток](#стратегия-веток)
+4. [Процесс разработки](#процесс-разработки)
+5. [Оформление коммитов](#оформление-коммитов)
+6. [Правила оформления кода](#правила-оформления-кода)
+7. [Создание Pull Request](#создание-pull-request)
+8. [Уязвимости безопасности](#уязвимости-безопасности)
 
 ---
 
@@ -33,20 +35,19 @@
     cd ostp
     ```
 
-2.  **Соберите веб-интерфейс панели управления**:
-    ```bash
-    cd ostp-control
-    npm install
-    npm run build
-    cd ..
-    ```
-
-3.  **Соберите весь Cargo-workspace**:
+2.  **Соберите весь Cargo-workspace**:
     ```bash
     cargo build
     ```
+    `ostp-control` (веб-панель) нужна только если вы работаете конкретно над
+    ней - в остальных случаях сервер собирается с пустым `dist/` через
+    `rust-embed`, и этот шаг не нужен для повседневной работы над
+    core/client/server. Если вы всё же трогаете панель:
+    ```bash
+    cd ostp-control && npm install && npm run build && cd ..
+    ```
 
-4.  **Запустите тесты**:
+3.  **Запустите тесты**:
     ```bash
     cargo test --workspace
     ```
@@ -58,7 +59,7 @@
 Репозиторий представляет собой единый Cargo-workspace со следующими компонентами:
 
 *   [`ostp-core/`](file:///d:/ospab-projects/ostp/ostp-core): Базовая логика протокола: форматирование пакетов, сериализация, конечный автомат выборочного подтверждения (ARQ/ACK/NACK) и рукопожатие Noise (`Noise_NNpsk0_25519_ChaChaPoly_BLAKE2s`).
-*   [`ostp-client/`](file:///d:/ospab-projects/ostp/ostp-client): Клиентская часть: локальные SOCKS5/HTTP прокси-серверы, интеграция с драйвером `wintun` / `tun2socks` и реализация раздельного туннелирования для прямого обхода трафика.
+*   [`ostp-client/`](file:///d:/ospab-projects/ostp/ostp-client): Клиентская часть: локальные SOCKS5/HTTP прокси-серверы, нативный OSTP TUN-интерфейс (через драйвер `wintun`) и реализация раздельного туннелирования для прямого обхода трафика.
 *   [`ostp-server/`](file:///d:/ospab-projects/ostp/ostp-server): Серверная часть: диспетчеризация сессий, маскировка под классические веб-серверы при активном сканировании, база данных ключей доступа и REST API панели управления.
 *   [`ostp-control/`](file:///d:/ospab-projects/ostp/ostp-control): Панель администратора (пользователи, статистика трафика в реальном времени, лимиты скорости и объема данных).
 *   [`ostp-gui/`](file:///d:/ospab-projects/ostp/ostp-gui): Настольное приложение-клиент для Windows и Linux на платформе Tauri.
@@ -66,11 +67,28 @@
 
 ---
 
+## Стратегия веток
+
+В репозитории три долгоживущие ветки, по возрастанию стабильности:
+
+| Ветка | Роль |
+|---|---|
+| `alpha` | Активная разработка. Вся новая работа и фиксы попадают сюда первыми. |
+| `beta` | Периодически перематывается вперёд (fast-forward) от `alpha`, когда та немного «отлежалась». Собирается в канал релиза `{версия}-beta`. |
+| `master` | Перематывается вперёд от `beta`, когда та доказала стабильность. Настоящие тегированные релизы (`vX.Y.Z`) режутся отсюда. |
+
+В `beta` и `master` **никогда** не коммитят напрямую - они только перематываются вперёд от ветки уровнем ниже. Это значит, что промоушен - всегда обычный `git merge` без единого конфликта по построению: не мержите/не ребейзьте свою фичу прямо в `beta` или `master`.
+
+**PR от контрибьюторов нацелены на `alpha`**, не на `master`.
+
+---
+
 ## Процесс разработки
 
 1.  **Проверьте существующие задачи** или откройте новую тему (Issue) для обсуждения предлагаемых изменений.
-2.  **Сделайте fork репозитория** и создайте новую ветку от `master`:
+2.  **Сделайте fork репозитория** и создайте новую ветку от `alpha`:
     ```bash
+    git checkout alpha
     git checkout -b feat/имя-вашей-фичи
     ```
 3.  **Внесите необходимые изменения** и добавьте соответствующие модульные или интеграционные тесты.
@@ -89,6 +107,33 @@
 
 ---
 
+## Оформление коммитов
+
+```
+<тип>(<область>): <краткое описание в повелительном наклонении>
+
+<опционально: тело - объясняет ПОЧЕМУ, а не что; диф и так показывает что изменилось>
+```
+
+- **Тип** - один из: `feat` (новая функциональность), `fix` (исправление бага), `docs`, `refactor` (без изменения поведения), `perf`, `test`, `chore` (зависимости/тулинг/версии), `ci`, `security`.
+- **Область** (опционально) - крейт или часть проекта: `client`, `server`, `core`, `gui`, `flutter`, `ci`, `docs` и т.д., например `fix(client): ...`.
+- **Краткое описание** - повелительное наклонение ("добавь", а не "добавил"/"добавляет"), без точки в конце, желательно до ~70 символов.
+- **Тело** - только когда причина не очевидна из дифа: какой баг это чинит, какое ограничение определило подход, на какой trade-off вы пошли. Не пересказывайте то, что и так видно в дифе. Перенос строк на ~72 символах.
+
+```
+fix(server): отбрасывать junk-фреймы по маркеру для каждого ключа, а не глобальному
+
+Фиксированный 4-байтовый маркер на каждом junk-пакете сам по себе - сигнатура
+DPI, по которой можно фильтровать любого наблюдателя во всех деплойментах OSTP
+сразу. Выводим маркер из access_key (HKDF, та же схема что у
+obfuscation_key/psk), чтобы он был индивидуальным для ключа и неотличимым от
+случайной полезной нагрузки пакета.
+```
+
+Несколько несвязанных изменений - это несколько отдельных коммитов, а не один сборный. Это сохраняет пользу от `git bisect` и код-ревью. Squash-merge подходит для PR с парой коммитов вроде "fix typo" / "address review", но не сквошьте вместе логически разные изменения.
+
+---
+
 ## Правила оформления кода
 
 *   **Безопасность (Safety)**: Избегайте использования блоков `unsafe` везде, где это возможно. Допускается их использование только для низкоуровневых системных вызовов (например, FFI-настройки сокетов `setsockopt`). Любой блок `unsafe` должен сопровождаться комментарием `// SAFETY: ...`.
@@ -104,7 +149,7 @@
     ```bash
     git push origin feat/имя-вашей-фичи
     ```
-2.  Создайте Pull Request (PR) в ветку `master` основного репозитория.
+2.  Создайте Pull Request (PR) в ветку `alpha` основного репозитория (см. [Стратегия веток](#стратегия-веток) - `master` получает только fast-forward от `beta`, PR туда не принимаются напрямую).
 3.  Подробно опишите внесенные изменения: какая проблема решается, как проводилось тестирование и на каких платформах проверялась сборка.
 4.  Убедитесь, что автоматическое тестирование (GitHub Actions CI) завершилось успешно.
 

Cargo.lock

@@ -432,6 +432,15 @@ dependencies = [
  "libc",
 ]
 
+[[package]]
+name = "crossbeam-channel"
+version = "0.5.15"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "82b8f8f868b36967f9606790d1903570de9ceaf870a7bf9fbbd3016d636a2cb2"
+dependencies = [
+ "crossbeam-utils",
+]
+
 [[package]]
 name = "crossbeam-utils"
 version = "0.8.21"
@@ -525,6 +534,15 @@ dependencies = [
  "thiserror 2.0.18",
 ]
 
+[[package]]
+name = "deranged"
+version = "0.5.8"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7cd812cc2bc1d69d4764bd80df88b4317eaef9e773c75226407d9bc0876b211c"
+dependencies = [
+ "powerfmt",
+]
+
 [[package]]
 name = "digest"
 version = "0.10.7"
@@ -1298,7 +1316,9 @@ dependencies = [
 
 [[package]]
 name = "netstack-smoltcp"
-version = "0.2.2"
+version = "0.2.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4c38f66cdd673ff0e760752f27c6d34a7e3a140f0b1eea9efae3c46d8867c83d"
 dependencies = [
  "etherparse",
  "futures",
@@ -1331,6 +1351,12 @@ dependencies = [
  "windows-sys 0.61.2",
 ]
 
+[[package]]
+name = "num-conv"
+version = "0.2.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "521739c6d2bac4aa25192232afe6841231376b2b26d4d9fae5ecf8ca5772e441"
+
 [[package]]
 name = "num-traits"
 version = "0.2.19"
@@ -1360,7 +1386,7 @@ checksum = "c08d65885ee38876c4f86fa503fb49d7b507c2b62552df7c70b2fce627e06381"
 
 [[package]]
 name = "ostp"
-version = "0.2.83"
+version = "0.4.4"
 dependencies = [
  "anyhow",
  "base64",
@@ -1371,8 +1397,10 @@ dependencies = [
  "ostp-core",
  "ostp-server",
  "rand 0.8.5",
+ "rlimit",
  "serde",
  "serde_json",
+ "sha2",
  "tokio",
  "tracing",
  "tracing-subscriber",
@@ -1381,21 +1409,20 @@ dependencies = [
 
 [[package]]
 name = "ostp-client"
-version = "0.2.83"
+version = "0.4.4"
 dependencies = [
  "anyhow",
  "base64",
  "bytes",
- "chacha20poly1305",
  "chrono",
  "futures",
  "futures-util",
- "hex",
  "hmac",
  "json_comments",
  "libc",
  "netstack-smoltcp",
  "ostp-core",
+ "ostp-tun",
  "portable-atomic",
  "rand 0.8.5",
  "serde",
@@ -1404,15 +1431,16 @@ dependencies = [
  "socket2",
  "tokio",
  "tracing",
+ "tracing-appender",
+ "tracing-subscriber",
  "tun",
  "webpki-roots 0.26.11",
  "winapi",
- "x25519-dalek",
 ]
 
 [[package]]
 name = "ostp-core"
-version = "0.2.83"
+version = "0.4.4"
 dependencies = [
  "anyhow",
  "bytes",
@@ -1446,7 +1474,7 @@ dependencies = [
 
 [[package]]
 name = "ostp-server"
-version = "0.2.83"
+version = "0.4.4"
 dependencies = [
  "anyhow",
  "axum",
@@ -1469,6 +1497,7 @@ dependencies = [
  "sha2",
  "simple-dns",
  "socket2",
+ "subtle",
  "tokio",
  "tower-http",
  "tracing",
@@ -1476,9 +1505,21 @@ dependencies = [
  "x25519-dalek",
 ]
 
+[[package]]
+name = "ostp-tun"
+version = "0.4.4"
+dependencies = [
+ "anyhow",
+ "libc",
+ "tokio",
+ "tracing",
+ "tun",
+ "winapi",
+]
+
 [[package]]
 name = "ostp-tun-helper"
-version = "0.2.83"
+version = "0.4.4"
 dependencies = [
  "anyhow",
  "chrono",
@@ -1557,6 +1598,12 @@ dependencies = [
  "zerovec",
 ]
 
+[[package]]
+name = "powerfmt"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "439ee305def115ba05938db6eb1644ff94165c5ab5e9420d1c1bcedbba909391"
+
 [[package]]
 name = "ppv-lite86"
 version = "0.2.21"
@@ -1811,6 +1858,15 @@ dependencies = [
  "windows-sys 0.52.0",
 ]
 
+[[package]]
+name = "rlimit"
+version = "0.11.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f35ee2729c56bb610f6dba436bf78135f728b7373bdffae2ec815b2d3eb98cc3"
+dependencies = [
+ "libc",
+]
+
 [[package]]
 name = "rust-embed"
 version = "8.11.0"
@@ -2119,6 +2175,12 @@ version = "2.6.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "13c2bddecc57b384dee18652358fb23172facb8a2c51ccc10d74c157bdea3292"
 
+[[package]]
+name = "symlink"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a7973cce6668464ea31f176d85b13c7ab3bba2cb3b77a2ed26abd7801688010a"
+
 [[package]]
 name = "syn"
 version = "2.0.117"
@@ -2199,6 +2261,37 @@ dependencies = [
  "cfg-if",
 ]
 
+[[package]]
+name = "time"
+version = "0.3.47"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "743bd48c283afc0388f9b8827b976905fb217ad9e647fae3a379a9283c4def2c"
+dependencies = [
+ "deranged",
+ "itoa",
+ "num-conv",
+ "powerfmt",
+ "serde_core",
+ "time-core",
+ "time-macros",
+]
+
+[[package]]
+name = "time-core"
+version = "0.1.8"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7694e1cfe791f8d31026952abf09c69ca6f6fa4e1a1229e18988f06a04a12dca"
+
+[[package]]
+name = "time-macros"
+version = "0.2.27"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2e70e4c5a0e0a8a4823ad65dfe1a6930e4f4d756dcd9dd7939022b5e8c501215"
+dependencies = [
+ "num-conv",
+ "time-core",
+]
+
 [[package]]
 name = "tinystr"
 version = "0.8.3"
@@ -2341,6 +2434,19 @@ dependencies = [
  "tracing-core",
 ]
 
+[[package]]
+name = "tracing-appender"
+version = "0.2.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "050686193eb999b4bb3bc2acfa891a13da00f79734704c4b8b4ef1a10b368a3c"
+dependencies = [
+ "crossbeam-channel",
+ "symlink",
+ "thiserror 2.0.18",
+ "time",
+ "tracing-subscriber",
+]
+
 [[package]]
 name = "tracing-attributes"
 version = "0.1.31"

Cargo.toml

@@ -5,27 +5,24 @@ members = [
     "ostp-server",
     "ostp-jni", "ostp",
     "ostp-tun-helper"
-]
+, "ostp-tun"]
 exclude = ["ostp-gui/src-tauri", "ostp-brain", "ostp-prober"]
 resolver = "2"
 
 [workspace.package]
 edition = "2021"
-license = "BSL 1.1"
-version = "0.2.83"
+license = "AGPL-3.0"
+version = "0.4.4"
 
 [workspace.dependencies]
 anyhow = "1.0"
 bytes = "1.6"
 chacha20poly1305 = "0.10"
 rand = "0.8"
-snow = "0.9"
+snow = { version = "0.9", features = ["risky-raw-split"] }
 thiserror = "1.0"
 tokio = { version = "1.37", features = ["rt-multi-thread", "macros", "net", "time", "io-util", "sync", "signal"] }
 tracing = "0.1"
 sha2 = "0.10"
 hmac = "0.12"
 portable-atomic = "1.10"
-
-[patch.crates-io]
-netstack-smoltcp = { path = "netstack-smoltcp" }

LICENSE

@@ -1,74 +1,661 @@
-Business Source License 1.1
+                    GNU AFFERO GENERAL PUBLIC LICENSE
+                       Version 3, 19 November 2007
 
-Parameters
+ Copyright (C) 2007 Free Software Foundation, Inc. <https://fsf.org/>
+ Everyone is permitted to copy and distribute verbatim copies
+ of this license document, but changing it is not allowed.
 
-Licensor:                     Ospab Foundation (represented by Syralev Georgiy)
-Licensed Work:                The Ospab Stealth Transport Protocol (OSTP) and all
-                              associated workspace crates, utilities, and documents.
-Additional Use Grant:         The Licensor hereby grants you the right to copy,
-                              modify, create derivative works, redistribute, and
-                              make non-production and non-commercial use of the
-                              Licensed Work. You are also permitted to use the
-                              Licensed Work in production for personal, private
-                              utility and non-profit organizations.
-Change Date:                  May 14, 2030
-Change License:               MIT License (as defined below)
+                            Preamble
 
------------------------------------------------------------------------------------
+  The GNU Affero General Public License is a free, copyleft license for
+software and other kinds of works, specifically designed to ensure
+cooperation with the community in the case of network server software.
 
-Terms
+  The licenses for most software and other practical works are designed
+to take away your freedom to share and change the works.  By contrast,
+our General Public Licenses are intended to guarantee your freedom to
+share and change all versions of a program--to make sure it remains free
+software for all its users.
 
-1. The Licensor hereby grants you the right to copy, modify, create derivative works, 
-   redistribute, and make use of the Licensed Work only as permitted by the 
-   Additional Use Grant.
+  When we speak of free software, we are referring to freedom, not
+price.  Our General Public Licenses are designed to make sure that you
+have the freedom to distribute copies of free software (and charge for
+them if you wish), that you receive source code or can get it if you
+want it, that you can change the software or use pieces of it in new
+free programs, and that you know you can do these things.
 
-2. The Licensor hereby grants you the right to copy, modify, create derivative works, 
-   redistribute, and make use of the Licensed Work under the terms of the Change 
-   License on and after the Change Date.
+  Developers that use our General Public Licenses protect your rights
+with two steps: (1) assert copyright on the software, and (2) offer
+you this License which gives you legal permission to copy, distribute
+and/or modify the software.
 
-3. To the extent that any term of this License (including the Additional Use Grant 
-   and the Change License) is in conflict with the Terms of this License, these 
-   Terms shall take precedence.
+  A secondary benefit of defending all users' freedom is that
+improvements made in alternate versions of the program, if they
+receive widespread use, become available for other developers to
+incorporate.  Many developers of free software are heartened and
+encouraged by the resulting cooperation.  However, in the case of
+software used on network servers, this result may fail to come about.
+The GNU General Public License permits making a modified version and
+letting the public access it on a server without ever releasing its
+source code to the public.
 
-4. Every copy of the Licensed Work and any derivative work must include this 
-   License and all other copyright, trademark, and proprietary notices included 
-   with the Licensed Work.
+  The GNU Affero General Public License is designed specifically to
+ensure that, in such cases, the modified source code becomes available
+to the community.  It requires the operator of a network server to
+provide the source code of the modified version running there to the
+users of that server.  Therefore, public use of a modified version, on
+a publicly accessible server, gives the public access to the source
+code of the modified version.
 
-5. Any use of the Licensed Work that is not permitted by this License is a breach 
-   of this License and may terminate your rights under this License.
+  An older license, called the Affero General Public License and
+published by Affero, was designed to accomplish similar goals.  This is
+a different license, not a version of the Affero GPL, but Affero has
+released a new version of the Affero GPL which permits relicensing under
+this license.
 
-6. DISCLAIMER OF WARRANTY. TO THE EXTENT PERMITTED BY APPLICABLE LAW, THE LICENSED 
-   WORK IS PROVIDED ON AN "AS IS" BASIS. THE LICENSOR MAKES NO REPRESENTATIONS OR 
-   WARRANTIES OF ANY KIND CONCERNING THE LICENSED WORK, EXPRESS OR IMPLIED, STATUTORY 
-   OR OTHERWISE, INCLUDING, WITHOUT LIMITATION, WARRANTIES OF TITLE, 
-   MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, OR NONINFRINGEMENT.
+  The precise terms and conditions for copying, distribution and
+modification follow.
 
-7. LIMITATION OF LIABILITY. TO THE EXTENT PERMITTED BY APPLICABLE LAW, IN NO EVENT 
-   WILL THE LICENSOR BE LIABLE TO YOU ON ANY LEGAL THEORY FOR ANY SPECIAL, INCIDENTAL, 
-   CONSEQUENTIAL, PUNITIVE, OR EXEMPLARY DAMAGES ARISING OUT OF THIS LICENSE OR THE 
-   USE OF THE LICENSED WORK, EVEN IF THE LICENSOR HAS BEEN ADVISED OF THE 
-   POSSIBILITY OF SUCH DAMAGES.
+                       TERMS AND CONDITIONS
 
------------------------------------------------------------------------------------
+  0. Definitions.
 
-Change License Text (MIT License)
+  "This License" refers to version 3 of the GNU Affero General Public License.
 
-Copyright (c) 2026 Syralev Georgiy (Ospab Foundation)
+  "Copyright" also means copyright-like laws that apply to other kinds of
+works, such as semiconductor masks.
 
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
+  "The Program" refers to any copyrightable work licensed under this
+License.  Each licensee is addressed as "you".  "Licensees" and
+"recipients" may be individuals or organizations.
 
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
+  To "modify" a work means to copy from or adapt all or part of the work
+in a fashion requiring copyright permission, other than the making of an
+exact copy.  The resulting work is called a "modified version" of the
+earlier work or a work "based on" the earlier work.
 
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
+  A "covered work" means either the unmodified Program or a work based
+on the Program.
+
+  To "propagate" a work means to do anything with it that, without
+permission, would make you directly or secondarily liable for
+infringement under applicable copyright law, except executing it on a
+computer or modifying a private copy.  Propagation includes copying,
+distribution (with or without modification), making available to the
+public, and in some countries other activities as well.
+
+  To "convey" a work means any kind of propagation that enables other
+parties to make or receive copies.  Mere interaction with a user through
+a computer network, with no transfer of a copy, is not conveying.
+
+  An interactive user interface displays "Appropriate Legal Notices"
+to the extent that it includes a convenient and prominently visible
+feature that (1) displays an appropriate copyright notice, and (2)
+tells the user that there is no warranty for the work (except to the
+extent that warranties are provided), that licensees may convey the
+work under this License, and how to view a copy of this License.  If
+the interface presents a list of user commands or options, such as a
+menu, a prominent item in the list meets this criterion.
+
+  1. Source Code.
+
+  The "source code" for a work means the preferred form of the work
+for making modifications to it.  "Object code" means any non-source
+form of a work.
+
+  A "Standard Interface" means an interface that either is an official
+standard defined by a recognized standards body, or, in the case of
+interfaces specified for a particular programming language, one that
+is widely used among developers working in that language.
+
+  The "System Libraries" of an executable work include anything, other
+than the work as a whole, that (a) is included in the normal form of
+packaging a Major Component, but which is not part of that Major
+Component, and (b) serves only to enable use of the work with that
+Major Component, or to implement a Standard Interface for which an
+implementation is available to the public in source code form.  A
+"Major Component", in this context, means a major essential component
+(kernel, window system, and so on) of the specific operating system
+(if any) on which the executable work runs, or a compiler used to
+produce the work, or an object code interpreter used to run it.
+
+  The "Corresponding Source" for a work in object code form means all
+the source code needed to generate, install, and (for an executable
+work) run the object code and to modify the work, including scripts to
+control those activities.  However, it does not include the work's
+System Libraries, or general-purpose tools or generally available free
+programs which are used unmodified in performing those activities but
+which are not part of the work.  For example, Corresponding Source
+includes interface definition files associated with source files for
+the work, and the source code for shared libraries and dynamically
+linked subprograms that the work is specifically designed to require,
+such as by intimate data communication or control flow between those
+subprograms and other parts of the work.
+
+  The Corresponding Source need not include anything that users
+can regenerate automatically from other parts of the Corresponding
+Source.
+
+  The Corresponding Source for a work in source code form is that
+same work.
+
+  2. Basic Permissions.
+
+  All rights granted under this License are granted for the term of
+copyright on the Program, and are irrevocable provided the stated
+conditions are met.  This License explicitly affirms your unlimited
+permission to run the unmodified Program.  The output from running a
+covered work is covered by this License only if the output, given its
+content, constitutes a covered work.  This License acknowledges your
+rights of fair use or other equivalent, as provided by copyright law.
+
+  You may make, run and propagate covered works that you do not
+convey, without conditions so long as your license otherwise remains
+in force.  You may convey covered works to others for the sole purpose
+of having them make modifications exclusively for you, or provide you
+with facilities for running those works, provided that you comply with
+the terms of this License in conveying all material for which you do
+not control copyright.  Those thus making or running the covered works
+for you must do so exclusively on your behalf, under your direction
+and control, on terms that prohibit them from making any copies of
+your copyrighted material outside their relationship with you.
+
+  Conveying under any other circumstances is permitted solely under
+the conditions stated below.  Sublicensing is not allowed; section 10
+makes it unnecessary.
+
+  3. Protecting Users' Legal Rights From Anti-Circumvention Law.
+
+  No covered work shall be deemed part of an effective technological
+measure under any applicable law fulfilling obligations under article
+11 of the WIPO copyright treaty adopted on 20 December 1996, or
+similar laws prohibiting or restricting circumvention of such
+measures.
+
+  When you convey a covered work, you waive any legal power to forbid
+circumvention of technological measures to the extent such circumvention
+is effected by exercising rights under this License with respect to
+the covered work, and you disclaim any intention to limit operation or
+modification of the work as a means of enforcing, against the work's
+users, your or third parties' legal rights to forbid circumvention of
+technological measures.
+
+  4. Conveying Verbatim Copies.
+
+  You may convey verbatim copies of the Program's source code as you
+receive it, in any medium, provided that you conspicuously and
+appropriately publish on each copy an appropriate copyright notice;
+keep intact all notices stating that this License and any
+non-permissive terms added in accord with section 7 apply to the code;
+keep intact all notices of the absence of any warranty; and give all
+recipients a copy of this License along with the Program.
+
+  You may charge any price or no price for each copy that you convey,
+and you may offer support or warranty protection for a fee.
+
+  5. Conveying Modified Source Versions.
+
+  You may convey a work based on the Program, or the modifications to
+produce it from the Program, in the form of source code under the
+terms of section 4, provided that you also meet all of these conditions:
+
+    a) The work must carry prominent notices stating that you modified
+    it, and giving a relevant date.
+
+    b) The work must carry prominent notices stating that it is
+    released under this License and any conditions added under section
+    7.  This requirement modifies the requirement in section 4 to
+    "keep intact all notices".
+
+    c) You must license the entire work, as a whole, under this
+    License to anyone who comes into possession of a copy.  This
+    License will therefore apply, along with any applicable section 7
+    additional terms, to the whole of the work, and all its parts,
+    regardless of how they are packaged.  This License gives no
+    permission to license the work in any other way, but it does not
+    invalidate such permission if you have separately received it.
+
+    d) If the work has interactive user interfaces, each must display
+    Appropriate Legal Notices; however, if the Program has interactive
+    interfaces that do not display Appropriate Legal Notices, your
+    work need not make them do so.
+
+  A compilation of a covered work with other separate and independent
+works, which are not by their nature extensions of the covered work,
+and which are not combined with it such as to form a larger program,
+in or on a volume of a storage or distribution medium, is called an
+"aggregate" if the compilation and its resulting copyright are not
+used to limit the access or legal rights of the compilation's users
+beyond what the individual works permit.  Inclusion of a covered work
+in an aggregate does not cause this License to apply to the other
+parts of the aggregate.
+
+  6. Conveying Non-Source Forms.
+
+  You may convey a covered work in object code form under the terms
+of sections 4 and 5, provided that you also convey the
+machine-readable Corresponding Source under the terms of this License,
+in one of these ways:
+
+    a) Convey the object code in, or embodied in, a physical product
+    (including a physical distribution medium), accompanied by the
+    Corresponding Source fixed on a durable physical medium
+    customarily used for software interchange.
+
+    b) Convey the object code in, or embodied in, a physical product
+    (including a physical distribution medium), accompanied by a
+    written offer, valid for at least three years and valid for as
+    long as you offer spare parts or customer support for that product
+    model, to give anyone who possesses the object code either (1) a
+    copy of the Corresponding Source for all the software in the
+    product that is covered by this License, on a durable physical
+    medium customarily used for software interchange, for a price no
+    more than your reasonable cost of physically performing this
+    conveying of source, or (2) access to copy the
+    Corresponding Source from a network server at no charge.
+
+    c) Convey individual copies of the object code with a copy of the
+    written offer to provide the Corresponding Source.  This
+    alternative is allowed only occasionally and noncommercially, and
+    only if you received the object code with such an offer, in accord
+    with subsection 6b.
+
+    d) Convey the object code by offering access from a designated
+    place (gratis or for a charge), and offer equivalent access to the
+    Corresponding Source in the same way through the same place at no
+    further charge.  You need not require recipients to copy the
+    Corresponding Source along with the object code.  If the place to
+    copy the object code is a network server, the Corresponding Source
+    may be on a different server (operated by you or a third party)
+    that supports equivalent copying facilities, provided you maintain
+    clear directions next to the object code saying where to find the
+    Corresponding Source.  Regardless of what server hosts the
+    Corresponding Source, you remain obligated to ensure that it is
+    available for as long as needed to satisfy these requirements.
+
+    e) Convey the object code using peer-to-peer transmission, provided
+    you inform other peers where the object code and Corresponding
+    Source of the work are being offered to the general public at no
+    charge under subsection 6d.
+
+  A separable portion of the object code, whose source code is excluded
+from the Corresponding Source as a System Library, need not be
+included in conveying the object code work.
+
+  A "User Product" is either (1) a "consumer product", which means any
+tangible personal property which is normally used for personal, family,
+or household purposes, or (2) anything designed or sold for incorporation
+into a dwelling.  In determining whether a product is a consumer product,
+doubtful cases shall be resolved in favor of coverage.  For a particular
+product received by a particular user, "normally used" refers to a
+typical or common use of that class of product, regardless of the status
+of the particular user or of the way in which the particular user
+actually uses, or expects or is expected to use, the product.  A product
+is a consumer product regardless of whether the product has substantial
+commercial, industrial or non-consumer uses, unless such uses represent
+the only significant mode of use of the product.
+
+  "Installation Information" for a User Product means any methods,
+procedures, authorization keys, or other information required to install
+and execute modified versions of a covered work in that User Product from
+a modified version of its Corresponding Source.  The information must
+suffice to ensure that the continued functioning of the modified object
+code is in no case prevented or interfered with solely because
+modification has been made.
+
+  If you convey an object code work under this section in, or with, or
+specifically for use in, a User Product, and the conveying occurs as
+part of a transaction in which the right of possession and use of the
+User Product is transferred to the recipient in perpetuity or for a
+fixed term (regardless of how the transaction is characterized), the
+Corresponding Source conveyed under this section must be accompanied
+by the Installation Information.  But this requirement does not apply
+if neither you nor any third party retains the ability to install
+modified object code on the User Product (for example, the work has
+been installed in ROM).
+
+  The requirement to provide Installation Information does not include a
+requirement to continue to provide support service, warranty, or updates
+for a work that has been modified or installed by the recipient, or for
+the User Product in which it has been modified or installed.  Access to a
+network may be denied when the modification itself materially and
+adversely affects the operation of the network or violates the rules and
+protocols for communication across the network.
+
+  Corresponding Source conveyed, and Installation Information provided,
+in accord with this section must be in a format that is publicly
+documented (and with an implementation available to the public in
+source code form), and must require no special password or key for
+unpacking, reading or copying.
+
+  7. Additional Terms.
+
+  "Additional permissions" are terms that supplement the terms of this
+License by making exceptions from one or more of its conditions.
+Additional permissions that are applicable to the entire Program shall
+be treated as though they were included in this License, to the extent
+that they are valid under applicable law.  If additional permissions
+apply only to part of the Program, that part may be used separately
+under those permissions, but the entire Program remains governed by
+this License without regard to the additional permissions.
+
+  When you convey a copy of a covered work, you may at your option
+remove any additional permissions from that copy, or from any part of
+it.  (Additional permissions may be written to require their own
+removal in certain cases when you modify the work.)  You may place
+additional permissions on material, added by you to a covered work,
+for which you have or can give appropriate copyright permission.
+
+  Notwithstanding any other provision of this License, for material you
+add to a covered work, you may (if authorized by the copyright holders of
+that material) supplement the terms of this License with terms:
+
+    a) Disclaiming warranty or limiting liability differently from the
+    terms of sections 15 and 16 of this License; or
+
+    b) Requiring preservation of specified reasonable legal notices or
+    author attributions in that material or in the Appropriate Legal
+    Notices displayed by works containing it; or
+
+    c) Prohibiting misrepresentation of the origin of that material, or
+    requiring that modified versions of such material be marked in
+    reasonable ways as different from the original version; or
+
+    d) Limiting the use for publicity purposes of names of licensors or
+    authors of the material; or
+
+    e) Declining to grant rights under trademark law for use of some
+    trade names, trademarks, or service marks; or
+
+    f) Requiring indemnification of licensors and authors of that
+    material by anyone who conveys the material (or modified versions of
+    it) with contractual assumptions of liability to the recipient, for
+    any liability that these contractual assumptions directly impose on
+    those licensors and authors.
+
+  All other non-permissive additional terms are considered "further
+restrictions" within the meaning of section 10.  If the Program as you
+received it, or any part of it, contains a notice stating that it is
+governed by this License along with a term that is a further
+restriction, you may remove that term.  If a license document contains
+a further restriction but permits relicensing or conveying under this
+License, you may add to a covered work material governed by the terms
+of that license document, provided that the further restriction does
+not survive such relicensing or conveying.
+
+  If you add terms to a covered work in accord with this section, you
+must place, in the relevant source files, a statement of the
+additional terms that apply to those files, or a notice indicating
+where to find the applicable terms.
+
+  Additional terms, permissive or non-permissive, may be stated in the
+form of a separately written license, or stated as exceptions;
+the above requirements apply either way.
+
+  8. Termination.
+
+  You may not propagate or modify a covered work except as expressly
+provided under this License.  Any attempt otherwise to propagate or
+modify it is void, and will automatically terminate your rights under
+this License (including any patent licenses granted under the third
+paragraph of section 11).
+
+  However, if you cease all violation of this License, then your
+license from a particular copyright holder is reinstated (a)
+provisionally, unless and until the copyright holder explicitly and
+finally terminates your license, and (b) permanently, if the copyright
+holder fails to notify you of the violation by some reasonable means
+prior to 60 days after the cessation.
+
+  Moreover, your license from a particular copyright holder is
+reinstated permanently if the copyright holder notifies you of the
+violation by some reasonable means, this is the first time you have
+received notice of violation of this License (for any work) from that
+copyright holder, and you cure the violation prior to 30 days after
+your receipt of the notice.
+
+  Termination of your rights under this section does not terminate the
+licenses of parties who have received copies or rights from you under
+this License.  If your rights have been terminated and not permanently
+reinstated, you do not qualify to receive new licenses for the same
+material under section 10.
+
+  9. Acceptance Not Required for Having Copies.
+
+  You are not required to accept this License in order to receive or
+run a copy of the Program.  Ancillary propagation of a covered work
+occurring solely as a consequence of using peer-to-peer transmission
+to receive a copy likewise does not require acceptance.  However,
+nothing other than this License grants you permission to propagate or
+modify any covered work.  These actions infringe copyright if you do
+not accept this License.  Therefore, by modifying or propagating a
+covered work, you indicate your acceptance of this License to do so.
+
+  10. Automatic Licensing of Downstream Recipients.
+
+  Each time you convey a covered work, the recipient automatically
+receives a license from the original licensors, to run, modify and
+propagate that work, subject to this License.  You are not responsible
+for enforcing compliance by third parties with this License.
+
+  An "entity transaction" is a transaction transferring control of an
+organization, or substantially all assets of one, or subdividing an
+organization, or merging organizations.  If propagation of a covered
+work results from an entity transaction, each party to that
+transaction who receives a copy of the work also receives whatever
+licenses to the work the party's predecessor in interest had or could
+give under the previous paragraph, plus a right to possession of the
+Corresponding Source of the work from the predecessor in interest, if
+the predecessor has it or can get it with reasonable efforts.
+
+  You may not impose any further restrictions on the exercise of the
+rights granted or affirmed under this License.  For example, you may
+not impose a license fee, royalty, or other charge for exercise of
+rights granted under this License, and you may not initiate litigation
+(including a cross-claim or counterclaim in a lawsuit) alleging that
+any patent claim is infringed by making, using, selling, offering for
+sale, or importing the Program or any portion of it.
+
+  11. Patents.
+
+  A "contributor" is a copyright holder who authorizes use under this
+License of the Program or a work on which the Program is based.  The
+work thus licensed is called the contributor's "contributor version".
+
+  A contributor's "essential patent claims" are all patent claims
+owned or controlled by the contributor, whether already acquired or
+hereafter acquired, that would be infringed by some manner, permitted
+by this License, of making, using, or selling its contributor version,
+but do not include claims that would be infringed only as a
+consequence of further modification of the contributor version.  For
+purposes of this definition, "control" includes the right to grant
+patent sublicenses in a manner consistent with the requirements of
+this License.
+
+  Each contributor grants you a non-exclusive, worldwide, royalty-free
+patent license under the contributor's essential patent claims, to
+make, use, sell, offer for sale, import and otherwise run, modify and
+propagate the contents of its contributor version.
+
+  In the following three paragraphs, a "patent license" is any express
+agreement or commitment, however denominated, not to enforce a patent
+(such as an express permission to practice a patent or covenant not to
+sue for patent infringement).  To "grant" such a patent license to a
+party means to make such an agreement or commitment not to enforce a
+patent against the party.
+
+  If you convey a covered work, knowingly relying on a patent license,
+and the Corresponding Source of the work is not available for anyone
+to copy, free of charge and under the terms of this License, through a
+publicly available network server or other readily accessible means,
+then you must either (1) cause the Corresponding Source to be so
+available, or (2) arrange to deprive yourself of the benefit of the
+patent license for this particular work, or (3) arrange, in a manner
+consistent with the requirements of this License, to extend the patent
+license to downstream recipients.  "Knowingly relying" means you have
+actual knowledge that, but for the patent license, your conveying the
+covered work in a country, or your recipient's use of the covered work
+in a country, would infringe one or more identifiable patents in that
+country that you have reason to believe are valid.
+
+  If, pursuant to or in connection with a single transaction or
+arrangement, you convey, or propagate by procuring conveyance of, a
+covered work, and grant a patent license to some of the parties
+receiving the covered work authorizing them to use, propagate, modify
+or convey a specific copy of the covered work, then the patent license
+you grant is automatically extended to all recipients of the covered
+work and works based on it.
+
+  A patent license is "discriminatory" if it does not include within
+the scope of its coverage, prohibits the exercise of, or is
+conditioned on the non-exercise of one or more of the rights that are
+specifically granted under this License.  You may not convey a covered
+work if you are a party to an arrangement with a third party that is
+in the business of distributing software, under which you make payment
+to the third party based on the extent of your activity of conveying
+the work, and under which the third party grants, to any of the
+parties who would receive the covered work from you, a discriminatory
+patent license (a) in connection with copies of the covered work
+conveyed by you (or copies made from those copies), or (b) primarily
+for and in connection with specific products or compilations that
+contain the covered work, unless you entered into that arrangement,
+or that patent license was granted, prior to 28 March 2007.
+
+  Nothing in this License shall be construed as excluding or limiting
+any implied license or other defenses to infringement that may
+otherwise be available to you under applicable patent law.
+
+  12. No Surrender of Others' Freedom.
+
+  If conditions are imposed on you (whether by court order, agreement or
+otherwise) that contradict the conditions of this License, they do not
+excuse you from the conditions of this License.  If you cannot convey a
+covered work so as to satisfy simultaneously your obligations under this
+License and any other pertinent obligations, then as a consequence you may
+not convey it at all.  For example, if you agree to terms that obligate you
+to collect a royalty for further conveying from those to whom you convey
+the Program, the only way you could satisfy both those terms and this
+License would be to refrain entirely from conveying the Program.
+
+  13. Remote Network Interaction; Use with the GNU General Public License.
+
+  Notwithstanding any other provision of this License, if you modify the
+Program, your modified version must prominently offer all users
+interacting with it remotely through a computer network (if your version
+supports such interaction) an opportunity to receive the Corresponding
+Source of your version by providing access to the Corresponding Source
+from a network server at no charge, through some standard or customary
+means of facilitating copying of software.  This Corresponding Source
+shall include the Corresponding Source for any work covered by version 3
+of the GNU General Public License that is incorporated pursuant to the
+following paragraph.
+
+  Notwithstanding any other provision of this License, you have
+permission to link or combine any covered work with a work licensed
+under version 3 of the GNU General Public License into a single
+combined work, and to convey the resulting work.  The terms of this
+License will continue to apply to the part which is the covered work,
+but the work with which it is combined will remain governed by version
+3 of the GNU General Public License.
+
+  14. Revised Versions of this License.
+
+  The Free Software Foundation may publish revised and/or new versions of
+the GNU Affero General Public License from time to time.  Such new versions
+will be similar in spirit to the present version, but may differ in detail to
+address new problems or concerns.
+
+  Each version is given a distinguishing version number.  If the
+Program specifies that a certain numbered version of the GNU Affero General
+Public License "or any later version" applies to it, you have the
+option of following the terms and conditions either of that numbered
+version or of any later version published by the Free Software
+Foundation.  If the Program does not specify a version number of the
+GNU Affero General Public License, you may choose any version ever published
+by the Free Software Foundation.
+
+  If the Program specifies that a proxy can decide which future
+versions of the GNU Affero General Public License can be used, that proxy's
+public statement of acceptance of a version permanently authorizes you
+to choose that version for the Program.
+
+  Later license versions may give you additional or different
+permissions.  However, no additional obligations are imposed on any
+author or copyright holder as a result of your choosing to follow a
+later version.
+
+  15. Disclaimer of Warranty.
+
+  THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
+APPLICABLE LAW.  EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
+HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
+OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
+THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+PURPOSE.  THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
+IS WITH YOU.  SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF
+ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
+
+  16. Limitation of Liability.
+
+  IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
+WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
+THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
+GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
+USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
+DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
+PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
+EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
+SUCH DAMAGES.
+
+  17. Interpretation of Sections 15 and 16.
+
+  If the disclaimer of warranty and limitation of liability provided
+above cannot be given local legal effect according to their terms,
+reviewing courts shall apply local law that most closely approximates
+an absolute waiver of all civil liability in connection with the
+Program, unless a warranty or assumption of liability accompanies a
+copy of the Program in return for a fee.
+
+                     END OF TERMS AND CONDITIONS
+
+            How to Apply These Terms to Your New Programs
+
+  If you develop a new program, and you want it to be of the greatest
+possible use to the public, the best way to achieve this is to make it
+free software which everyone can redistribute and change under these terms.
+
+  To do so, attach the following notices to the program.  It is safest
+to attach them to the start of each source file to most effectively
+state the exclusion of warranty; and each file should have at least
+the "copyright" line and a pointer to where the full notice is found.
+
+    <one line to give the program's name and a brief idea of what it does.>
+    Copyright (C) <year>  <name of author>
+
+    This program is free software: you can redistribute it and/or modify
+    it under the terms of the GNU Affero General Public License as published by
+    the Free Software Foundation, either version 3 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU Affero General Public License for more details.
+
+    You should have received a copy of the GNU Affero General Public License
+    along with this program.  If not, see <https://www.gnu.org/licenses/>.
+
+Also add information on how to contact you by electronic and paper mail.
+
+  If your software can interact with users remotely through a computer
+network, you should also make sure that it provides a way for users to
+get its source.  For example, if your program is a web application, its
+interface could display a "Source" link that leads users to an archive
+of the code.  There are many ways you could offer source, and different
+solutions will be better for different programs; see section 13 for the
+specific requirements.
+
+  You should also get your employer (if you work as a programmer) or school,
+if any, to sign a "copyright disclaimer" for the program, if necessary.
+For more information on this, and how to apply and follow the GNU AGPL, see
+<https://www.gnu.org/licenses/>.

README.md

@@ -1,14 +1,16 @@
-# OSTP — Ospab Stealth Transport Protocol
+# OSTP - Ospab Stealth Transport Protocol
 
 [Русский язык](README.ru.md) · [Wiki](https://github.com/ospab/ostp/wiki) · [Contributing](CONTRIBUTING.md) · [Releases](https://github.com/ospab/ostp/releases)
 
-![GitHub Release](https://img.shields.io/github/v/release/ospab/ostp?style=flat-square&color=blue)
-![License: BSL 1.1](https://img.shields.io/badge/License-BSL%201.1-orange.svg?style=flat-square)
-![Platform: Windows | Linux | macOS | Android](https://img.shields.io/badge/Platform-Windows%20%7C%20Linux%20%7C%20macOS%20%7C%20Android-green.svg?style=flat-square)
-![Crypto](https://img.shields.io/badge/Crypto-Noise__NNpsk0-blueviolet?style=flat-square)
-![Transport](https://img.shields.io/badge/Transport-UDP%20ARQ-informational?style=flat-square)
+![GitHub Release](https://img.shields.io/github/v/release/ospab/ostp?style=for-the-badge&color=blue)
+![License: AGPL v3](https://img.shields.io/badge/License-AGPL%20v3-blue.svg?style=for-the-badge)
+![Platform: Windows | Linux | macOS | Android](https://img.shields.io/badge/Platform-Windows%20%7C%20Linux%20%7C%20macOS%20%7C%20Android-green.svg?style=for-the-badge)
+![Crypto](https://img.shields.io/badge/Crypto-Noise__NNpsk0-blueviolet?style=for-the-badge)
+![Transport](https://img.shields.io/badge/Transport-UDP%20ARQ-informational?style=for-the-badge)
 
-**OSTP** is a high-performance, censorship-resistant transport protocol designed to tunnel TCP traffic over UDP with full traffic obfuscation. Every byte on the wire — including packet headers — is cryptographically indistinguishable from random noise. Resistant to Deep Packet Inspection (DPI), active probing, and statistical traffic analysis.
+> A fast, custom encrypted transport protocol written in Rust.
+
+**OSTP** (Ospab Stealth Transport Protocol) is a high-performance transport protocol. It implements a custom ARQ transport over UDP, as well as a UoT (UDP-over-TCP) mode. Every byte on the wire - including packet headers - is cryptographically indistinguishable from random noise, making it highly resistant to Deep Packet Inspection (DPI).
 
 ---
 
@@ -33,17 +35,17 @@ Download pre-built binaries for your platform from [GitHub Releases](https://git
 
 | Feature | Description |
 |---------|-------------|
-| **Full Traffic Obfuscation** | Every packet — including headers — is indistinguishable from random noise. Session IDs and nonces are masked with per-packet HMAC-derived keys. |
-| **Noise Protocol Handshake** | `Noise_NNpsk0_25519_ChaChaPoly_BLAKE2s` — PSK-authenticated, forward-secret key exchange with no static identity exposure. |
+| **Full Traffic Obfuscation** | Every packet - including headers - is indistinguishable from random noise. Session IDs and nonces are masked with per-packet HMAC-derived keys. |
+| **Noise Protocol Handshake** | `Noise_NNpsk0_25519_ChaChaPoly_BLAKE2s` - PSK-authenticated, forward-secret key exchange with no static identity exposure. |
 | **Reliable UDP (ARQ)** | Selective ACK/NACK with rate-limited retransmission, configurable reorder buffer, and exponential backoff. |
 | **Multiplexed Streams** | Multiple logical TCP streams over a single encrypted UDP session with per-stream flow control. |
-| **Seamless Roaming** | Clients can switch networks (WiFi ↔ LTE) without session interruption — tracked by session-ID, not IP. |
+| **Seamless Roaming** | Clients can switch networks (WiFi ↔ LTE) without session interruption - tracked by session-ID, not IP. |
 | **Management API** | Built-in REST API for third-party panels (3x-ui, custom dashboards). Per-user stats, traffic limits, key CRUD. |
-| **Fallback Server** | TCP fallback proxy to a web server — makes OSTP indistinguishable from nginx during active probing. |
+| **Fallback Server** | TCP fallback proxy to a web server - makes OSTP indistinguishable from nginx during active probing. |
 | **Multi-Listener** | Bind to multiple addresses simultaneously (dual-stack IPv4/IPv6, multi-port). |
-| **TUN Mode** | Full-system VPN via `tun2socks` integration. All traffic transparently routed through the tunnel. |
-| **xHTTP Stealth (UoT)** | UDP-over-TCP tunnel disguised as standard HTTP/1.1 or TLS traffic to bypass Level 1 Deep Packet Inspection (DPI) whitelists. |
-| **XTLS-Reality** | Custom, dependency-free implementation of the Reality protocol using ChaCha20Poly1305 and X25519 for perfect TLS 1.3 impersonation. |
+| **TUN Mode** | Full-system VPN via native `smoltcp` network stack without external dependencies. All traffic transparently routed through the tunnel. |
+| **UoT (UDP-over-TCP)** | Bare UDP-over-TCP tunnel, no protocol mimicry. Since all data is fully encrypted and length-prefixed, it bypasses DPI filters that block unknown UDP traffic by riding over a plain TCP connection. |
+| **Mobile & Web Apps** | Beautiful cross-platform mobile client (Flutter) and a modern Web Control Panel (React/Vite) for effortless server and client management. |
 | **TURN Relay** | RFC 5766 TURN support for environments where direct UDP is blocked. |
 | **Hot-Reload** | Runtime config reload without restart (access keys, exclusions, mux settings). |
 | **Structured Logging** | `tracing`-based logging with `RUST_LOG` filtering. JSON/file/syslog output support. |
@@ -53,40 +55,43 @@ Download pre-built binaries for your platform from [GitHub Releases](https://git
 
 ## Architecture
 
-```
-┌─────────────────────────────────────────────────────────────┐
-│  Client                                                     │
-│  ┌──────────┐   ┌──────────┐   ┌────────────────────────┐   │
-│  │ Browser  │──▸│ SOCKS5/  │──▸│    Bridge (Mux)        │   │
-│  │ / Apps   │   │ HTTP     │   │  ┌─────────────────┐   │   │
-│  │          │   │ Proxy    │   │  │ ProtocolMachine │   │   │
-│  └──────────┘   └──────────┘   │  │ (Noise + AEAD)  │   │   │
-│                                │  └────────┬────────┘   │   │
-│  ┌──────────┐                  │           │            │   │
-│  │ TUN Mode │──────────────────┤      UDP Socket        │   │
-│  │tun2socks │                  │  (32MB buffers,        │   │
-│  └──────────┘                  │   obfuscated wire)     │   │
-│                                └───────────┬────────────┘   │
-└────────────────────────────────────────────┼────────────────┘
-                                             │ UDP
-┌────────────────────────────────────────────┼────────────────┐
-│  Server                                    │                │
-│  ┌─────────────────────────────────────────┴───────────┐    │
-│  │              Dispatcher                             │    │
-│  │  (Session lookup, roaming, replay guard, per-user   │    │
-│  │   traffic accounting, limit enforcement)            │    │
-│  └──┬──────────────────────┬───────────────────────────┘    │
-│     │                      │                                │
-│  ┌──▾──────────────────┐ ┌─▾──────────────────────────┐     │
-│  │ Relay Loop          │ │ Management API (REST)      │     │
-│  │ (per-stream TCP)    │ │ /api/users, /api/stats     │     │
-│  │ ──▸ Internet        │ │ Bearer token auth          │     │
-│  └─────────────────────┘ └────────────────────────────┘     │
-│                                                             │
-│  ┌──────────────────────────────────────────────────────┐   │
-│  │ Fallback TCP Proxy ──▸ nginx/caddy (anti-DPI)        │   │
-│  └──────────────────────────────────────────────────────┘   │
-└─────────────────────────────────────────────────────────────┘
+```mermaid
+flowchart LR
+    %% Styles
+    classDef userApp fill:#e1f5fe,stroke:#01579b,stroke-width:2px,color:#01579b
+    classDef ostpCore fill:#e8f5e9,stroke:#2e7d32,stroke-width:2px,color:#2e7d32
+    classDef network fill:#fff3e0,stroke:#e65100,stroke-width:2px,color:#e65100,stroke-dasharray: 5 5
+    classDef external fill:#f3e5f5,stroke:#4a148c,stroke-width:2px,color:#4a148c
+    classDef fallback fill:#ffebee,stroke:#c62828,stroke-width:2px,color:#c62828
+
+    subgraph Local["💻 Client Device"]
+        Apps["Web Browser / Apps"]:::userApp
+        Socks["SOCKS5 / HTTP Proxy"]:::ostpCore
+        Tun["Global TUN (VPN)"]:::ostpCore
+        Client["OSTP Client Protocol Engine\n(Noise + ChaCha20 + ARQ)"]:::ostpCore
+
+        Apps -->|TCP/UDP| Socks
+        Apps -->|IP Packets| Tun
+        Socks --> Client
+        Tun --> Client
+    end
+
+    subgraph Internet["🌐 Hostile Network (DPI/Firewall)"]
+        Tunnel{"Fully Obfuscated\nEncrypted UDP\n(Looks like noise)"}:::network
+    end
+
+    subgraph Remote["🖥️ Remote VPS (Server)"]
+        Server["OSTP Server Protocol Engine\n(Authentication & Decryption)"]:::ostpCore
+        Relay["Connection Multiplexer"]:::ostpCore
+        Fallback["Fake Website\n(Nginx/Caddy)"]:::fallback
+        Target["Open Internet\n(YouTube, Google, etc)"]:::external
+
+        Server -->|Decrypted Traffic| Relay
+        Server -->|Active Probe / Scanner| Fallback
+        Relay -->|Clear Traffic| Target
+    end
+
+    Client <==> Tunnel <==> Server
 ```
 
 ---
@@ -97,15 +102,15 @@ Download pre-built binaries for your platform from [GitHub Releases](https://git
 
 ```bash
 # On your VPS (server):
-./ostp --init server
+./ostp init server
 
 # On your machine (client):
-./ostp --init client
+./ostp init client
 ```
 
 ### 2. Edit config
 
-**Server** — set your access keys:
+**Server** - set your access keys:
 ```jsonc
 {
   "mode": "server",
@@ -116,14 +121,14 @@ Download pre-built binaries for your platform from [GitHub Releases](https://git
 }
 ```
 
-**Client** — point to your server:
+**Client** - point to your server:
 ```jsonc
 {
   "mode": "client",
   "server": "YOUR_SERVER_IP:50000",
   "access_key": "YOUR_SECRET_KEY",
   "socks5_bind": "127.0.0.1:1088",
-  "transport": { "mode": "udp", "stealth_sni": "vk.com", "stealth_port": 443 },
+  "transport": { "mode": "udp" },
   "tun": { "enable": false, "dns": "1.1.1.1" }
 }
 ```
@@ -131,18 +136,20 @@ Download pre-built binaries for your platform from [GitHub Releases](https://git
 ### 3. Run
 
 ```bash
-./ostp                        # Uses config.json in current directory
-./ostp --config /path/to.json # Custom config path
-./ostp --check                # Validate config without running
-./ostp --generate-key         # Generate a new access key
-./ostp --links                # Print client share links
+./ostp                         # Uses config.json in current directory
+./ostp --config /path/to.json  # Custom config path
+./ostp check                   # Validate config without running
+./ostp gk                      # Generate a new access key
+./ostp links                   # Print client share links
 ```
 
 ### 4. Connect via share link (one-liner)
 ```bash
-./ostp "ostp://ACCESS_KEY@server.com:50000?..."
+./ostp connect "ostp://ACCESS_KEY@server.com:50000?..."
 ```
-> **Note**: Always wrap the `ostp://...` link in quotes (`"`) so your terminal doesn't misinterpret special characters like `&` or `?`.
+
+> [!WARNING]
+> Always wrap the `ostp://...` link in quotes (`"`) so your terminal doesn't misinterpret special characters like `&` or `?`.
 
 ---
 
@@ -171,29 +178,40 @@ Full API reference: [Management API](https://github.com/ospab/ostp/wiki/Manageme
 ## CLI Reference
 
 ```
-ostp [OPTIONS] [URL]
+ostp [--config <PATH>] [COMMAND]
 
-Options:
+Commands:
+  run                    Run the daemon using the config file (default when no command is given)
+  connect <URL>          Connect once using a share link: ostp://KEY@HOST:PORT
+  setup                  Interactive setup wizard
+  init <MODE>            Generate a template config (server/client/relay)
+  check                  Validate the configuration file and exit
+  gk                     Generate a secure access key (alias: generate-key)
+    --format <FMT>         Key format: hex, base64 (default: hex)
+    -n, --count <N>        Number of keys to generate (default: 1)
+  links                  Print client share links from the server config
+  import <URL>           Import a share link into the config file
+  update                 Update OSTP to the latest release
+    -b, --branch <NAME>    Release channel: stable, beta, alpha (default: stable)
+    -v, --version <VER>    Update to an exact version instead of the channel's latest
+  migrate                Force-migrate the configuration file to the current format
+  proxy-env              Print shell export commands for the local SOCKS proxy
+  proxy-env-clear        Print shell export commands to unset it
+  uninstall              Stop the service and remove the binary and config
+
+Global options:
   --config <PATH>        Config file path (default: config.json)
-  --init <MODE>          Generate template config (server/client)
-  --check                Validate configuration and exit
-  -g, --generate-key     Generate a secure access key
-  -c, --count <N>        Number of keys to generate (default: 1)
-  --format <FMT>         Key format: hex, base64 (default: hex)
-  --links                Print client share links from server config
-
-Arguments:
-  [URL]                  Connect via share link: ostp://KEY@HOST:PORT
 ```
 
+Every subcommand also accepts `-h`/`--help` for its own option list.
+
 ---
 
 ## Protocol Summary
 
 | Layer | Mechanism |
 |-------|-----------|
-| XTLS-Reality | Spoofed TLS 1.3 ClientHello, X25519 Key Exchange, ChaCha20-Poly1305 AEAD |
-| Key Exchange | Noise NNpsk0 (X25519 + ChaChaPoly + BLAKE2s) |
+| Key Exchange | Noise NNpsk0 (X25519 + ChaChaPoly + BLAKE2s) zero-RTT |
 | Encryption | ChaCha20-Poly1305 AEAD per-packet |
 | Header Obfuscation | HMAC-SHA256 derived per-packet mask |
 | Reliability | Selective ACK with cumulative + SACK ranges |
@@ -219,7 +237,7 @@ cargo test -p ostp-core -p ostp-server
 
 ## Documentation
 
-- **[Wiki](https://github.com/ospab/ostp/wiki)** — Full documentation
+- **[Wiki](https://github.com/ospab/ostp/wiki)** - Full documentation
 - [Installation](https://github.com/ospab/ostp/wiki/Installation)
 - [Configuration Reference](https://github.com/ospab/ostp/wiki/Configuration)
 - [Management API](https://github.com/ospab/ostp/wiki/Management-API)
@@ -231,8 +249,7 @@ cargo test -p ostp-core -p ostp-server
 
 ## License
 
-Business Source License 1.1. Free for personal and non-commercial use.  
-Converts to MIT License on May 14, 2030.
+GNU Affero General Public License v3.0 (AGPL-3.0). See [LICENSE](LICENSE) for the full text.
 
 ---
 

README.ru.md

@@ -1,12 +1,16 @@
-# OSTP — Ospab Stealth Transport Protocol
+# OSTP - Ospab Stealth Transport Protocol
 
 [English](README.md) · [Contributing](CONTRIBUTING.ru.md)
 
-![GitHub Release](https://img.shields.io/github/v/release/ospab/ostp?style=flat-square&color=blue)
-![License: BSL 1.1](https://img.shields.io/badge/License-BSL%201.1-orange.svg?style=flat-square)
-![Platform: Windows | Linux | macOS | Android](https://img.shields.io/badge/Platform-Windows%20%7C%20Linux%20%7C%20macOS%20%7C%20Android-green.svg?style=flat-square)
+![GitHub Release](https://img.shields.io/github/v/release/ospab/ostp?style=for-the-badge&color=blue)
+![License: AGPL v3](https://img.shields.io/badge/License-AGPL%20v3-blue.svg?style=for-the-badge)
+![Platform: Windows | Linux | macOS | Android](https://img.shields.io/badge/Platform-Windows%20%7C%20Linux%20%7C%20macOS%20%7C%20Android-green.svg?style=for-the-badge)
+![Crypto](https://img.shields.io/badge/Crypto-Noise__NNpsk0-blueviolet?style=for-the-badge)
+![Transport](https://img.shields.io/badge/Transport-UDP%20ARQ-informational?style=for-the-badge)
 
-OSTP — высокопроизводительный транспортный протокол, устойчивый к цензуре. Туннелирует TCP-трафик поверх UDP с полной обфускацией. Устойчив к Deep Packet Inspection (DPI), активному зондированию и статистическому анализу трафика.
+> Быстрый кастомный зашифрованный транспортный протокол на Rust.
+
+**OSTP** (Ospab Stealth Transport Protocol) - кастомный транспортный протокол. Реализует собственный ARQ-транспорт поверх UDP, а также режим UoT (UDP-over-TCP). Каждый байт, включая заголовки пакетов, криптографически неотличим от случайного шума, что делает его устойчивым к системам глубокого анализа трафика (DPI).
 
 ---
 
@@ -15,13 +19,13 @@ OSTP — высокопроизводительный транспортный
 | Возможность | Описание |
 |-------------|----------|
 | **Обфускация трафика** | Каждый пакет, включая заголовки, неотличим от случайного шума. Session ID и nonce маскируются HMAC-ключами, уникальными для каждого пакета. |
-| **Noise Protocol** | `Noise_NNpsk0_25519_ChaChaPoly_BLAKE2s` — аутентификация через PSK, forward secrecy, без раскрытия идентичности. |
+| **Noise Protocol** | `Noise_NNpsk0_25519_ChaChaPoly_BLAKE2s` - аутентификация через PSK, forward secrecy, без раскрытия идентичности. |
 | **Reliable UDP (ARQ)** | Selective ACK/NACK с rate-limited ретрансмиссией, настраиваемым reorder-буфером и exponential backoff. Разработан для 10 Гбит/с. |
 | **Мультиплексирование** | Несколько логических TCP-потоков поверх одной зашифрованной UDP-сессии с per-stream flow control. |
-| **Бесшовный роуминг** | Клиент может менять сети (WiFi ↔ 4G) без разрыва сессии — сервер отслеживает session-ID, а не IP-адрес. |
-| **TUN-режим** | Полносистемный VPN через интеграцию с `tun2socks` на Windows и Linux. |
-| **xHTTP Стелс (UoT)** | Туннель UDP-over-TCP, замаскированный под обычный HTTP/1.1 или TLS трафик для обхода белых списков ТСПУ (DPI). |
-| **XTLS-Reality** | Собственная реализация протокола Reality (без зависимостей) с использованием ChaCha20Poly1305 и X25519 для идеальной маскировки под TLS 1.3. |
+| **Бесшовный роуминг** | Клиент может менять сети (WiFi ↔ 4G) без разрыва сессии - сервер отслеживает session-ID, а не IP-адрес. |
+| **TUN-режим** | Полносистемный VPN без внешних зависимостей (встроенный network stack на базе `smoltcp`). |
+| **UoT (UDP-over-TCP)** | Голый туннель UDP-over-TCP, без имитации протоколов. Поскольку все данные полностью зашифрованы и имеют префикс длины, он обходит DPI фильтры, блокирующие неизвестный UDP трафик, передавая всё по обычному TCP соединению. |
+| **Мобильные и Web приложения** | Красивый кроссплатформенный мобильный клиент (Flutter) и современная Web панель управления (React/Vite) для удобного администрирования. |
 | **TURN Relay** | RFC 5766 TURN для окружений, где прямой UDP заблокирован. |
 | **Hot-Reload** | Перезагрузка конфига в рантайме без перезапуска (ключи, исключения, mux, TURN). |
 | **Кросс-платформа** | Windows, Linux, macOS, Android. Один бинарник, без зависимостей. |
@@ -30,33 +34,43 @@ OSTP — высокопроизводительный транспортный
 
 ## Архитектура
 
-```
-┌────────────────────────────────────────────────────────────┐
-│  Клиент                                                     │
-│  ┌──────────┐   ┌──────────┐   ┌───────────────────────┐   │
-│  │ Браузер  │──▸│ SOCKS5/  │──▸│    Bridge (Mux)       │   │
-│  │ / Прил.  │   │ HTTP     │   │  ┌─────────────────┐   │   │
-│  │          │   │ Прокси   │   │  │ ProtocolMachine │   │   │
-│  └──────────┘   └──────────┘   │  │ (Noise + AEAD)  │   │   │
-│                                │  └────────┬────────┘   │   │
-│  ┌──────────┐                  │           │            │   │
-│  │ TUN Mode │──────────────────┤      UDP-сокет         │   │
-│  │tun2socks │                  │  (32МБ буферы,         │   │
-│  └──────────┘                  │   обфускация)          │   │
-│                                └───────────┬────────────┘   │
-└────────────────────────────────────────────┼────────────────┘
-                                             │ UDP
-┌────────────────────────────────────────────┼────────────────┐
-│  Сервер                                    │                │
-│  ┌─────────────────────────────────────────┴──────────┐     │
-│  │              Dispatcher                             │     │
-│  │  (Поиск сессий, роуминг, защита от replay)          │     │
-│  └──────────────┬──────────────────────────────────────┘     │
-│                 │                                            │
-│  ┌──────────────▾──────────────────┐                         │
-│  │   Relay Loop (TCP per-stream)   │──▸ Интернет / Backend   │
-│  └─────────────────────────────────┘                         │
-└──────────────────────────────────────────────────────────────┘
+```mermaid
+flowchart LR
+    %% Styles
+    classDef userApp fill:#e1f5fe,stroke:#01579b,stroke-width:2px,color:#01579b
+    classDef ostpCore fill:#e8f5e9,stroke:#2e7d32,stroke-width:2px,color:#2e7d32
+    classDef network fill:#fff3e0,stroke:#e65100,stroke-width:2px,color:#e65100,stroke-dasharray: 5 5
+    classDef external fill:#f3e5f5,stroke:#4a148c,stroke-width:2px,color:#4a148c
+    classDef fallback fill:#ffebee,stroke:#c62828,stroke-width:2px,color:#c62828
+
+    subgraph Local["💻 Устройство клиента"]
+        Apps["Браузер / Приложения"]:::userApp
+        Socks["SOCKS5 / HTTP Прокси"]:::ostpCore
+        Tun["Global TUN (VPN)"]:::ostpCore
+        Client["OSTP Клиент\n(Noise + ChaCha20 + ARQ)"]:::ostpCore
+
+        Apps -->|TCP/UDP| Socks
+        Apps -->|IP Пакеты| Tun
+        Socks --> Client
+        Tun --> Client
+    end
+
+    subgraph Internet["🌐 Сеть с цензурой (DPI)"]
+        Tunnel{"Зашифрованный UDP\n(Выглядит как белый шум)"}:::network
+    end
+
+    subgraph Remote["🖥️ Удаленный сервер (VPS)"]
+        Server["OSTP Сервер\n(Аутентификация)"]:::ostpCore
+        Relay["Мультиплексор соединений"]:::ostpCore
+        Fallback["Фейковый сайт\n(Nginx/Caddy)"]:::fallback
+        Target["Свободный интернет\n(YouTube, Google и т.д.)"]:::external
+
+        Server -->|Расшифрованный трафик| Relay
+        Server -->|Сканеры цензоров| Fallback
+        Relay -->|Чистый трафик| Target
+    end
+
+    Client <==> Tunnel <==> Server
 ```
 
 ---
@@ -79,8 +93,8 @@ irm https://raw.githubusercontent.com/ospab/ostp/master/scripts/install.ps1 | ie
 
 Создать конфиг по умолчанию:
 ```bash
-./ostp --init server   # VPS
-./ostp --init client   # Локальная машина
+./ostp init server   # VPS
+./ostp init client   # Локальная машина
 ```
 
 ### Сервер (`config.json`)
@@ -111,9 +125,7 @@ irm https://raw.githubusercontent.com/ospab/ostp/master/scripts/install.ps1 | ie
   "debug": false,
   // Настройки транспорта (udp или uot)
   "transport": {
-    "mode": "udp",
-    "stealth_sni": "vk.com",
-    "stealth_port": 443
+    "mode": "udp"
   },
   // TUN-режим (полносистемный VPN)
   "tun": {
@@ -152,11 +164,41 @@ irm https://raw.githubusercontent.com/ospab/ostp/master/scripts/install.ps1 | ie
 ./ostp
 ```
 
+### Справка по командам
+
+```
+ostp [--config <PATH>] [КОМАНДА]
+
+Команды:
+  run                    Запустить демон по конфигу (по умолчанию, если команда не указана)
+  connect <URL>          Подключиться по share-ссылке: ostp://KEY@HOST:PORT
+  setup                  Интерактивный мастер настройки
+  init <MODE>            Сгенерировать шаблон конфига (server/client/relay)
+  check                  Проверить конфиг и выйти
+  gk                     Сгенерировать access-key (алиас: generate-key)
+    --format <FMT>         Формат ключа: hex, base64 (по умолчанию hex)
+    -n, --count <N>        Количество ключей (по умолчанию 1)
+  links                  Вывести client-share-ссылки из серверного конфига
+  import <URL>           Импортировать share-ссылку в конфиг
+  update                 Обновить OSTP до актуального релиза
+    -b, --branch <NAME>    Канал релиза: stable, beta, alpha (по умолчанию stable)
+    -v, --version <VER>    Обновиться на точную версию вместо последней в канале
+  migrate                Принудительно мигрировать конфиг к текущему формату
+  proxy-env              Вывести shell-команды для локального SOCKS-прокси
+  proxy-env-clear        Вывести shell-команды для их отмены
+  uninstall              Остановить сервис и удалить бинарник с конфигом
+
+Глобальные опции:
+  --config <PATH>        Путь к конфигу (по умолчанию config.json)
+```
+
+У каждой подкоманды есть своя справка через `-h`/`--help`.
+
 ### TUN-режим (Windows)
-Требуется `tun2socks.exe` в той же директории. Автоматически запрашивает права Администратора.
+Использует встроенный сетевой стек `smoltcp` и виртуальный адаптер `wintun` (необходима `wintun.dll`). Требует запуска с правами Администратора.
 
 ### TUN-режим (Linux)
-Требуется root. Нужен бинарник `tun2socks` (рядом или в `$PATH`).
+Использует встроенный сетевой стек `smoltcp` и `/dev/net/tun`. Требует запуска от имени `root` (или наличия `CAP_NET_ADMIN`).
 
 ---
 
@@ -164,8 +206,7 @@ irm https://raw.githubusercontent.com/ospab/ostp/master/scripts/install.ps1 | ie
 
 | Уровень | Механизм |
 |---------|----------|
-| XTLS-Reality | Поддельный TLS 1.3 ClientHello, X25519 обмен ключами, ChaCha20-Poly1305 AEAD |
-| Обмен ключами | Noise NNpsk0 (X25519 + ChaChaPoly + BLAKE2s) |
+| Обмен ключами | Noise NNpsk0 (X25519 + ChaChaPoly + BLAKE2s) zero-RTT |
 | Шифрование | ChaCha20-Poly1305 AEAD на каждый пакет |
 | Обфускация заголовков | HMAC-SHA256 маска session_id + nonce, уникальная для каждого пакета |
 | Надёжность | Selective ACK с cumulative + SACK диапазонами |
@@ -201,5 +242,4 @@ cross build --release --target x86_64-unknown-linux-gnu
 
 ## Лицензия
 
-Business Source License 1.1. Бесплатно для личного и некоммерческого использования.
-Переходит в MIT License 14 мая 2030 года.
+GNU Affero General Public License v3.0 (AGPL-3.0). Полный текст - в файле [LICENSE](LICENSE).

REBUILD_PLAN.md

@@ -0,0 +1,185 @@
+# Чистая переборка на базе v0.2.98
+
+База: `v0.2.98` (commit `31d0020`) — последняя версия, которая **стабильно работает**.
+Ветка: `clean-rebuild`. Всё, что появилось после (0.3.1 … 0.3.21), переносим
+**выборочно и с чистой головой**, а не копируем рефактор целиком.
+
+Принцип: 0.3.1 принёс «модульный multi-server рефактор» + лавину фич — и вместе с
+ними нестабильность. Берём только проверенное и нужное.
+
+---
+
+## Решения (зафиксировано пользователем)
+- **Junk-пакеты + TCP-фрагментация — ОСТАВЛЯЕМ** (нравятся). НО починить вредную
+  часть: junk по UDP не должен выглядеть для сервера как `Unauthorized probe`
+  (rate-limit/гейт на сервере), иначе флуд лога и риск самобана клиента. Фича
+  остаётся — чиним поведение, а не выпиливаем. Тонкая настройка — §E.
+- **Версия переборки — 0.4.0** (решено; 0.3.x сожжены в pre-release).
+- **WSS и Reality (TLS-мимикрия) — ВЫКИНУТЬ.** Путь проекта — **zapret-like**:
+  обфускация/DPI-evasion на уровне пакетов (junk, фрагментация, обфускация), а НЕ
+  мимикрия под TLS. Reality с нуля тяжела и не вписывается.
+- **Multi-server — НЕ НУЖЕН.** Режем до одного сервера → уходит urltest-группа и
+  половина сложности 0.3.1.
+- **Конфиг — ПЛОСКИЙ по сути, но оформлен красиво/секционно как сейчас** (решено).
+  Сохраняем читаемую секционную структуру (server / transport / tun / dns / exclude
+  и т.п.), но **выпиливаем модульную машинерию**: массивы `inbounds[]`/`outbounds[]`,
+  `routing.rules[]` с тегами, `default_outbound`, urltest, мульти-сервер. Один сервер
+  на конфиг. Исключения = плоский список внутри секции `exclude`.
+- **Профили — ОСТАВЛЯЕМ, single-select, в UI-слое** (решено). Профиль = сохранённый
+  конфиг одного сервера; активен ровно один (radio). Список/выбор/share живут во
+  фронте (prefs GUI / Flutter); **ядро о профилях не знает** — на «Подключить» из
+  выбранного профиля генерится плоский конфиг на один сервер. Никаких чекбоксов/
+  мульти-актив/urltest.
+- **Derived-secrets — ОСТАВЛЯЕМ, но ОБЯЗАТЕЛЬНО проверить, что он РЕАЛЬНО работает:**
+  старый клиент НЕ должен подключаться к новому серверу. В прошлой реализации это
+  НЕ соблюдалось (старый клиент → новый сервер подключался) — значит сервер всё ещё
+  принимал старый формат handshake / obfuscation-key. Это **баг**, закрыть в первую
+  очередь: сервер обязан отвергать всё, что не прошло derived-secrets.
+- **Лицензия — AGPLv3.**
+- **Брендинг — ОСТАВЛЯЕМ**: тёмная тема + орёл на фоне (watermark/логотип).
+- **Стелс-философия (north-star): zapret-like** — «нет узнаваемого заголовка +
+  манипуляции пакетами» (обфускация, junk, фрагментация, DNS/UoT-транспорты), а НЕ
+  «притворись известным протоколом» (Reality/WSS — выкинуты).
+
+---
+
+## 0. Корневая причина нестабильности 0.3.x
+**Модульный multi-server рефактор (0.3.1)** — `580faf6`, `8ed66f9`, `67f9c06`.
+Сменил формат конфига (inbounds/outbounds/routing/urltest), session-модель,
+hot-reload. Источник большинства багов (мёртвые маршруты, фейк-коннект,
+рассинхрон конфига). **НЕ копировать целиком.** Если multi-server реально нужен —
+добавлять минимально и поверх рабочей одно-серверной модели 0.2.98.
+
+---
+
+## A. ВЫКИНУТЬ / не переносить
+1. **WSS-фрейминг и Reality (TLS-мимикрия)** — оба выкинуть. Путь zapret-like, а не
+   маскировка под TLS-сайт; Reality (`reality.rs`) к тому же сложно сделать корректно
+   с нуля. Удалить из базы 0.2.98 целиком.
+2. **Multi-server / urltest-группа** — не нужен. Один сервер на конфиг.
+3. Остатки **tun2socks** на Android (`libtun2socks.so`, `tun2socks-arm64`,
+   `tun_child`, `t2sBinPath`) — давно мёртвый код, только раздувает APK. Не тащить.
+
+> ⚠️ Junk-пакеты и TCP-фрагментация **ОСТАЮТСЯ** (см. Решения и §E) — это уже не
+> «мусор». Но junk по UDP нужно сделать так, чтобы сервер его не считал
+> `Unauthorized probe` (rate-limit/гейт), иначе лог-флуд и риск самобана.
+
+---
+
+## B. ОБЯЗАТЕЛЬНО перенести (фиксы стабильности)
+- **fd limits / EMFILE** — `922cf0b`.
+- **Lifecycle хелпера**: принудительный `std::process::exit` после остановки, чтобы
+  не оставался зомби-процесс, держащий адаптер `ostp_tun` и дефолтный маршрут — `b6e78c1`.
+- **Bypass-маршрут сервера через `route.exe` по шлюзу** (а не legacy
+  `CreateIpForwardEntry`, который падал с err 160 из-за рассинхрона индексов
+  интерфейсов) — `b6e78c1`.
+- **IPC хелпера** (ChaCha20Poly1305 + hex) + **единый формат логов** — `ee38b15`.
+- **Closing-state fix** + `sent_history` на `BTreeMap` (O(log n) NACK) — `47d44fa`.
+- **Handshake timeout fixes** — `d65af35`, `6eb7b36` (ждать ответ до отправки данных).
+- **Buffer / UDP handler** — `b5e830a`.
+- **Логи**: UoT и unauthorized-probe → debug; rate-limit probe-лога — `1151726`, `fc339b3`.
+
+---
+
+## C. Протокол / крипто — решить и перенести
+- **Derived secrets handshake** — `f8f27d3`. PSK и obfuscation-key выводятся из
+  access-key через HKDF; handshake-payload = `[timestamp][session_id][access_key]`;
+  параметры паддинга деривируются; timestamp anti-replay (±300с).
+  ⚠️ **Ломает совместимость с 0.2.98 wire** (старый клиент не подключится).
+  Безопаснее старого (raw-PSK + нулевой obfuscation-key). **РЕШЕНИЕ:** переносим ли
+  (тогда нужен ребилд всех клиентов) — ДА, скорее всего, но осознанно.
+- **l4_protocol** для server outbound — `2997bfd`, `ad3a8cb`, `aae9d22`.
+
+---
+
+## D. Транспорты — перенести аккуратно (большие куски)
+- **DNS transport (dnstt)** как fallback — `3f1adbc`, `3ced4a1`, `d031b15`,
+  `10c1772`, `b31da29`. Полезно против блокировок, но объёмно и со своей
+  фрагментацией/reassembly. Переносить отдельным изолированным модулем.
+- UoT (UDP-over-TCP) — уже есть в 0.2.98, проверить что не сломан.
+
+---
+
+## E. Тонкая настройка junk/фрагментации (как в AmneziaWG)
+Junk и фрагментацию **оставляем** (Решения), а это — их параметризация. Главное
+условие: **координация клиент↔сервер**, иначе junk превращается в probe-флуд.
+- `Jc` — кол-во junk-пакетов, `Jmin`/`Jmax` — размеры; **сервер знает и молча отбрасывает**.
+- `S1`/`S2` — размеры init/response подгоняются.
+- Магические заголовки/сигнатуры пакетов (`H1..H4`).
+Реализовать как явные настраиваемые поля (не хардкод). Сервер ОБЯЗАН их понимать.
+Сам факт junk/frag — в базе; это «желание» — сделать их настраиваемыми. Можно потом.
+
+---
+
+## F. GUI (desktop) — перенести нужное, без хаоса
+- Профили на странице **настроек** (пусто + «Create a new profile» + «+» когда нет
+  профиля; «+» → меню «из ссылки / вручную»). Главный экран не усложнять.
+- **Share** профиля: QR (генерить локально, ключ наружу не отдавать — крейт `qrcode`)
+  + копируемая `ostp://` ссылка.
+- **Метрики**: байты считать в TUN-инбаунде; rtt брать из round-trip handshake
+  (а не отдельным TCP-probe).
+- **Health/состояние**: «connected» по реальной достижимости сервера на ПРАВИЛЬНОМ
+  порту (не хардкод :443), а не по факту «процесс запустился».
+- **routing**: всегда задавать `default_outbound: "proxy"`; ключи правил —
+  `domain_suffix` / `ip_cidr` / `process_name` (не `domains/ips/processes`).
+- Смена сервера = полный **stop+start**, а не hot-reload (иначе остаётся старый сервер).
+- Никаких непрогарженных `addEventListener` на удалённые элементы (краш init).
+
+---
+
+## G. Мобилка (Flutter + JNI) — перенести нужное
+- **routing**: тот же `default_outbound` + правильные ключи правил
+  (без них трафик шёл мимо туннеля — реальный IP).
+- **Байты на Android**: считать в обеих задачах fd-пути (read=upload, write=download).
+- **rtt**: из handshake (health-probe сокет на Android не protected → до сервера не доходит).
+- **Смена сети (WiFi↔LTE)**: реальный reconnect (сейчас `notifyNetworkChanged` — no-op).
+- **fd ownership**: НЕ двойное закрытие (Rust `OwnedFd` + Kotlin `close()`) → `detachFd()`.
+- **Share** профиля: QR (`qr_flutter`) + ссылка.
+- Выкинуть tun2socks (см. §A.3).
+
+---
+
+## H. Инфра / лицензия / брендинг
+- Лицензия: **AGPLv3** ✅ (зафиксировано). В 0.2.98 был BSL 1.1 → заменить (`9ce9e6d`).
+- **Брендинг — ОСТАВЛЯЕМ** ✅: тёмная тема + орёл на фоне (watermark/логотип) в GUI.
+  Перенести из текущего `ostp-gui` (assets/logo.svg, тёмная палитра) в чистую переборку.
+- Панель/license-check: open-source без license-check — `5782107`, `99ff76d` (если нужно).
+- Версионирование/CI build-script — `774d926` и пр.
+
+---
+
+## Инвентаризация базы 0.2.98 (что уже есть / что портировать)
+- **Есть в 0.2.98**: WSS (→ удалить), Reality/`reality.rs` (→ удалить),
+  инфра derived-secrets (`derive_all_secrets`, `obfuscation_key`) — но клиент юзал
+  dummy-ключи до `f8f27d3`.
+- **Нет в 0.2.98 — портировать из пост-0.2.98 кода**: junk-пакеты, TCP-фрагментация,
+  DNS-transport (dnstt), фикс derived-secrets `f8f27d3`, все фиксы §B, GUI/мобилка §F/§G.
+
+## Что легко упустить (решить до старта)
+1. **Версия переборки — 0.4.0** (решено). Сожжённые 0.3.x не переиспользуем.
+2. **Серверный конфиг — тоже плоский** и согласован с клиентским. Сервер обязан
+   поддерживать всё оставленное: derived-secrets (и **отвергать** старый формат),
+   корректную обработку junk (не probe-флуд), UoT, DNS-transport, management API.
+3. **Версия/магический байт протокола ДО крипто-слоя.** Сейчас нельзя отличить старый
+   handshake от нового — отсюда баг «старый клиент → новый сервер подключился».
+   Добавить версию в wire → будущие изменения управляемы, сервер чётко режет
+   несовместимое. Это системный фикс проблемы derived-secrets.
+4. **Клиент и сервер обновляются ВМЕСТЕ** — derived-secrets ломает совместимость,
+   смешивать старое и новое нельзя. Координировать выкладку.
+5. **Reality — выкинуть** (решено; в базе 0.2.98 есть `reality.rs` → удалить целиком).
+6. **Verify-loop = критерий «готово».** Каждая фича проверяется реальным тестом, не
+   «на словах»: connect → `curl` показывает IP **сервера**; старый клиент к новому
+   серверу **не** подключается; смена сети на мобилке восстанавливает туннель.
+
+## Порядок переборки (предложение, 1 сессия)
+1. §B (фиксы стабильности) — на чистый 0.2.98.
+2. §C (derived-secrets) — и СРАЗУ проверить: старый клиент к новому серверу НЕ
+   подключается (в прошлый раз был баг — подключался).
+3. **Junk + TCP-фрагментация** — перенести (оставляем), но junk по UDP не должен
+   читаться сервером как `Unauthorized probe` (rate-limit/гейт на сервере).
+4. §F/§G по минимуму (routing, метрики, состояние, share) **+ брендинг** (тёмная
+   тема, орёл на фоне).
+5. §D (DNS transport) — если нужно.
+6. ВЫКИНУТЬ: **WSS, multi-server, tun2socks** (§A). §E (тюнинг junk) — позже.
+7. Конфиг: плоский по сути, секционно-оформленный, один сервер (РЕШЕНО — без
+   inbounds/outbounds/routing-движка).

docs/banner.txt

@@ -0,0 +1,7 @@
+
+  ____   _____ _______ _____  
+ / __ \ / ____|__   __|  __ \ 
+| |  | | (___    | |  | |__) |
+| |  | |\___ \   | |  |  ___/ 
+| |__| |____) |  | |  | |     
+ \____/|_____/   |_|  |_|     

docs/en/specification.md

@@ -90,11 +90,22 @@ Because the `Nonce` is unique per packet, the mask is cryptographically independ
 
 OSTP executes a Noise Protocol Framework exchange utilizing the `Noise_NNpsk0_25519_ChaChaPoly_BLAKE2s` pattern.
 
-1.  The Registration Key (`access_key`) is converted to a 32-octet strong pre-shared key (PSK) via SHA-256.
+1.  The Registration Key (`access_key`) is converted to a 32-octet strong pre-shared key (PSK) via HKDF-SHA-256.
 2.  The PSK is integrated into the state at pattern position zero, authorizing and encrypting the very first handshaking datagram.
-3.  Ephemeral Curve25519 key exchange is evaluated to synthesize autonomous symmetric keys for subsequent read/write channels.
+3.  Ephemeral Curve25519 key exchange (`ee`) is evaluated, and the two directional transport keys are taken from Noise's `Split()` over the final chaining key `ck`.
 
-The initial handshake payload includes a Unix timestamp to mitigate replay attacks. The server enforces a strict ±30-second synchronization window.
+> **Forward secrecy.** The transport keys are derived from the chaining key
+> `ck`, which absorbs the ephemeral `ee` Diffie-Hellman result. They are **not**
+> derived from the Noise handshake hash `h` — `h` only ever absorbs public
+> transcript data (ephemeral public keys and on-wire ciphertexts) and never the
+> DH secret, so keys derived from it would give an access-key holder the ability
+> to decrypt any recorded session. Deriving from `ck` binds each session to its
+> ephemeral private keys, which are discarded after the handshake: an adversary
+> who later compromises the PSK still cannot decrypt past traffic. This is a
+> wire-breaking property gated by the internal protocol version (currently 5);
+> peers on an older version derive different keys and cannot interoperate.
+
+The initial handshake payload includes a Unix timestamp to mitigate replay attacks. The server enforces a ±300-second (5-minute) synchronization window and additionally records accepted handshakes in an anti-replay set for that window.
 
 ---
 
@@ -126,4 +137,5 @@ The server supports seamless network handoffs (e.g., transitioning from Wi-Fi to
 
 *   **Nonce Exhaustion:** The Nonce field is 64 bits. Implementations MUST terminate and re-key a session before the Nonce overflows to prevent AEAD keystream reuse.
 *   **Session Exhaustion (DoS):** Servers MUST enforce a strict cap on concurrent sessions (e.g., 1024) and silently drop handshake attempts exceeding this limit to prevent memory exhaustion attacks.
+*   **Handshake-trial CPU DoS:** Because there is no cleartext key identifier on the wire (a deliberate stealth property), a datagram from an unknown source must be trial-decrypted against every registered key. Servers MUST bound this work: OSTP caches each key's derived secrets and time-windowed junk markers (so a trial is a cheap comparison plus one AEAD attempt per key, not a fresh HKDF/HMAC), and gates the trial path behind a global token bucket (default 100/s) so a spoofed-source flood cannot force unbounded per-packet crypto. The established-session fast path and IP-roaming path are not subject to this bucket.
 *   **Header Authentication:** The header obfuscation mechanism provides privacy, not integrity. Header integrity is mathematically guaranteed by the Poly1305 Authentication Tag, which covers the entire 12-byte header as Additional Authenticated Data (AAD).

docs/relay-config-example.json

@@ -20,9 +20,15 @@
   // Адрес следующего узла в цепочке — UDP
   "upstream_udp": "TARGET_SERVER_IP:50000",
 
-  // URL API конечного (целевого) сервера для синхронизации access_keys
-  // Должен быть доступен с этого relay-сервера (можно через SSH-туннель)
-  "upstream_api_url": "http://TARGET_SERVER_IP:9090",
+  // URL API конечного (целевого) сервера для синхронизации access_keys.
+  // Должен быть доступен с этого relay-сервера (можно через SSH-туннель).
+  //
+  // ВАЖНО: URL обязан включать секретный путь панели (api.webpath целевого
+  // сервера). Management API смонтирован ВНУТРИ этого пути — именно он скрывает
+  // панель от сканеров, — поэтому голый host:port попадает в несуществующий
+  // маршрут, и синхронизация падает с 404 ещё до проверки токена.
+  // Это тот же адрес, по которому вы открываете веб-панель.
+  "upstream_api_url": "http://TARGET_SERVER_IP:9090/TARGET_SERVER_WEBPATH",
 
   // Bearer-токен для доступа к API целевого сервера
   // Должен совпадать с api.token в конфиге target-сервера

docs/ru/specification.md

@@ -90,11 +90,23 @@ OSTP поддерживает **внутреннее криптографиче
 
 OSTP использует Noise Protocol Framework с паттерном `Noise_NNpsk0_25519_ChaChaPoly_BLAKE2s`.
 
-1.  Регистрационный ключ доступа (`access_key`) преобразуется в 32-байтный строгий предварительно распределенный ключ (PSK) через SHA-256.
-2.  PSK применяется на нулевой позиции паттерна, обеспечивая авторизацию и шифрование самой первой датаграммы рукопожатия (Zero-RTT авторизация).
-3.  Выполняется эфемерный обмен ключами Curve25519 для создания симметричных ключей передачи данных.
+1.  Регистрационный ключ доступа (`access_key`) преобразуется в 32-байтный строгий предварительно распределенный ключ (PSK) через HKDF-SHA-256.
+2.  PSK применяется на нулевой позиции паттерна, обеспечивая авторизацию и шифрование самой первой датаграммы рукопожатия.
+3.  Выполняется эфемерный обмен ключами Curve25519 (`ee`), и два однонаправленных транспортных ключа берутся из `Split()` протокола Noise над финальным chaining key `ck`.
 
-Первичная полезная нагрузка рукопожатия содержит Unix-отметку времени для защиты от атак повторного воспроизведения (Replay Attacks). Сервер строго контролирует окно синхронизации (±30 секунд).
+> **Прямая секретность (Forward Secrecy).** Транспортные ключи выводятся из
+> chaining key `ck`, который вбирает результат эфемерного обмена Диффи-Хеллмана
+> `ee`. Они **не** выводятся из handshake hash `h` протокола Noise: `h` вбирает
+> только публичные данные транскрипта (эфемерные публичные ключи и шифртексты с
+> провода) и никогда — сам DH-секрет, поэтому ключи, выведенные из `h`, дали бы
+> держателю PSK возможность расшифровать любую записанную сессию. Вывод из `ck`
+> привязывает каждую сессию к её эфемерным приватным ключам, которые
+> уничтожаются после рукопожатия: злоумышленник, скомпрометировавший PSK позже,
+> всё равно не сможет расшифровать прошлый трафик. Это свойство ломает
+> совместимость и защищено внутренней версией протокола (сейчас 5): узлы более
+> старой версии выводят другие ключи и не могут взаимодействовать.
+
+Первичная полезная нагрузка рукопожатия содержит Unix-отметку времени для защиты от атак повторного воспроизведения (Replay Attacks). Сервер контролирует окно синхронизации (±300 секунд, 5 минут) и дополнительно фиксирует принятые рукопожатия в множестве защиты от повтора на время этого окна.
 
 ---
 
@@ -119,4 +131,5 @@ OSTP обеспечивает надежную доставку поверх UDP
 
 *   **Исчерпание Nonce:** Поле Nonce имеет размер 64 бита. Реализации ОБЯЗАНЫ разрывать сессию до переполнения Nonce, чтобы предотвратить катастрофическое повторное использование гаммы AEAD-шифра.
 *   **DDoS и исчерпание ресурсов:** Серверы ДОЛЖНЫ применять жесткий лимит на количество одновременных сессий (например, 1024) и молча отбрасывать запросы на рукопожатие при превышении лимита, предотвращая атаки на исчерпание памяти.
+*   **CPU-DoS на пути перебора рукопожатия:** Поскольку на проводе нет открытого идентификатора ключа (намеренное свойство скрытности), датаграмму от неизвестного источника приходится пробно расшифровывать каждым зарегистрированным ключом. Серверы ОБЯЗАНЫ ограничивать эту работу: OSTP кэширует производные секреты каждого ключа и его junk-маркеры для текущего временно́го окна (поэтому одна попытка — это дешёвое сравнение плюс одна попытка AEAD на ключ, а не новые HKDF/HMAC), и ограничивает путь перебора глобальным token bucket (по умолчанию 100/с), так что флуд с подменённых адресов не может навязать неограниченную криптографию на пакет. Быстрый путь установленных сессий и путь IP-роуминга под этот лимит не попадают.
 *   **Целостность заголовка:** Механизм маскирования обеспечивает только скрытность, а не целостность. Целостность заголовков математически гарантируется 16-байтным тегом аутентификации Poly1305, который покрывает 12-байтный заголовок как присоединенные данные (AAD).

icons/sqare.svg

@@ -0,0 +1,15 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 512 512">
+  <defs>
+    <linearGradient id="g2" x1="0%" y1="0%" x2="100%" y2="100%">
+      <stop offset="0%" stop-color="#111827" />
+      <stop offset="100%" stop-color="#374151" />
+    </linearGradient>
+    <linearGradient id="g2_path" x1="0%" y1="0%" x2="100%" y2="100%">
+      <stop offset="0%" stop-color="#3B82F6" />
+      <stop offset="100%" stop-color="#14B8A6" />
+    </linearGradient>
+  </defs>
+  <rect width="512" height="512" rx="120" fill="url(#g2)" />
+  <path d="M144 256c0-61.9 50.1-112 112-112s112 50.1 112 112-50.1 112-112 112S144 317.9 144 256zm-48 0c0 88.4 71.6 160 160 160s160-71.6 160-160S344.4 96 256 96 96 167.6 96 256z" fill="url(#g2_path)"/>
+  <circle cx="256" cy="256" r="40" fill="#F59E0B" />
+</svg>

netstack-smoltcp/.cargo-ok

@@ -1 +0,0 @@
-{"v":1}

netstack-smoltcp/.cargo_vcs_info.json

@@ -1,6 +0,0 @@
-{
-  "git": {
-    "sha1": "702f6dfe124c5e4d343cfd3ca5a3efe0446cf6f0"
-  },
-  "path_in_vcs": ""
-}

netstack-smoltcp/.github/actions/ndk-dev-rs/action.yml

@@ -1,37 +0,0 @@
-name: Setup Android NDK and Rust compiler ENV
-description: Setup an Android_NDK_HOME environment by downloading and Rust compiler environment.
-inputs:
-  rust-target:
-    description: Rust target to build
-    required: true
-  sdk-version:
-    description: Exact SDK version to use
-    default: "33"
-  ndk-version:
-    description: Exact NDK version to use
-    default: "25"
-  ndk-platform:
-    description: Which host platform to use
-    default: "linux"
-runs:
-  using: "composite"
-  steps:
-    - name: Download Android NDK
-      run: curl --http1.1 -O https://dl.google.com/android/repository/android-ndk-r${{ inputs.ndk-version }}-${{ inputs.ndk-platform }}.zip
-      shell: bash
-    - name: Extract Android NDK
-      run: unzip -q android-ndk-r${{ inputs.ndk-version }}-${{ inputs.ndk-platform }}.zip
-      shell: bash
-    - name: Set Rust compiler ENV
-      run: |
-        ndk_home=${{ github.workspace }}/android-ndk-r${{ inputs.ndk-version }}
-        platform=$(ls ${ndk_home}/toolchains/llvm/prebuilt/ | head -1)
-        ndk_tool=${ndk_home}/toolchains/llvm/prebuilt/${platform}/bin
-        envvar_suffix=$(echo ${{ inputs.rust-target }} | sed "s/-/_/g")
-        upper_suffix=$(echo ${envvar_suffix} | tr '[:lower:]' '[:upper:]')
-        tool_prefix=${{ inputs.rust-target }}${{ inputs.sdk-version }}
-        echo "ANDROID_NDK_HOME=${ndk_home}" >> $GITHUB_ENV
-        echo "CC_${envvar_suffix}=${ndk_tool}/${tool_prefix}-clang" >> $GITHUB_ENV
-        echo "AR_${envvar_suffix}=${ndk_tool}/llvm-ar" >> $GITHUB_ENV
-        echo "CARGO_TARGET_${upper_suffix}_LINKER=${ndk_tool}/${tool_prefix}-clang" >> $GITHUB_ENV
-      shell: bash

netstack-smoltcp/.github/workflows/main.yml

@@ -1,80 +0,0 @@
-name: CI
-
-on:
-  push:
-    branches:
-      - '**'
-  pull_request:
-    branches:
-      - '**'
-
-env:
-  CARGO_INCREMENTAL: 0
-  CARGO_REGISTRIES_CRATES_IO_PROTOCOL: sparse
-
-jobs:
-  test:
-    name: Test
-    runs-on: ${{ matrix.os }}
-    strategy:
-      matrix:
-        include:
-          - build: linux-amd64
-            os: ubuntu-latest
-            target: x86_64-unknown-linux-gnu
-          - build: android-arm64
-            os: ubuntu-latest
-            target: aarch64-linux-android
-            no_run: --no-run
-          - build: android-amd64
-            os: ubuntu-latest
-            target: x86_64-linux-android
-            no_run: --no-run
-          - build: macos-amd64
-            os: macos-latest
-            target: x86_64-apple-darwin
-          - build: macos-arm64
-            os: macos-14
-            target: aarch64-apple-darwin
-          - build: ios-arm64
-            os: macos-latest
-            target: aarch64-apple-ios
-            no_run: --no-run
-          - build: windows-amd64
-            os: windows-latest
-            target: x86_64-pc-windows-msvc
-          - build: windows-arm64
-            os: windows-latest
-            target: aarch64-pc-windows-msvc
-            no_run: --no-run
-    steps:
-      - uses: actions/checkout@v4
-      - name: Install Rust (rustup)
-        run: |
-          set -euxo pipefail
-          rustup toolchain install stable --no-self-update --profile minimal --target ${{ matrix.target }}
-          rustup default stable
-        shell: bash
-      - uses: Swatinem/rust-cache@v2
-      - name: Setup android environment
-        if: contains(matrix.build, 'android')
-        uses: ./.github/actions/ndk-dev-rs
-        with:
-          rust-target: ${{ matrix.target }}
-      - run: cargo test ${{ matrix.no_run }} --workspace --target ${{ matrix.target }}
-      - run: cargo test ${{ matrix.no_run }} --workspace --target ${{ matrix.target }} --release
-
-  msrv_n_clippy:
-    name: MSRV & Clippy & Rustfmt
-    runs-on: ${{ matrix.os }}
-    strategy:
-      fail-fast: false
-      matrix:
-        os: [ubuntu-latest, macos-latest, windows-latest]
-    steps:
-      - uses: actions/checkout@v4
-      - uses: dtolnay/rust-toolchain@stable
-      - run: cargo fmt -- --check
-      - run: cargo clippy --all-features -- -D warnings
-      - run: cargo check --lib -p netstack-smoltcp
-      - run: cargo check --lib -p netstack-smoltcp --all-features

netstack-smoltcp/.github/workflows/publish.yml

@@ -1,15 +0,0 @@
-on:
-  push:
-    tags:        
-      - '*'
-
-jobs:
-  publish:
-    runs-on: ubuntu-latest
-    steps:
-    - uses: actions/checkout@v4
-    - name: Publish to crates.io
-      run: |
-        cargo publish
-      env:
-        CARGO_REGISTRY_TOKEN: ${{ secrets.CARGO_REGISTRY_TOKEN }}

netstack-smoltcp/.gitignore

@@ -1,9 +0,0 @@
-/target
-/Cargo.lock
-
-.idea
-.VSCodeCounter/
-.vscode
-.DS_Store
-*.iml
-**/*.log

netstack-smoltcp/Cargo.toml

@@ -1,136 +0,0 @@
-# THIS FILE IS AUTOMATICALLY GENERATED BY CARGO
-#
-# When uploading crates to the registry Cargo will automatically
-# "normalize" Cargo.toml files for maximal compatibility
-# with all versions of Cargo and also rewrite `path` dependencies
-# to registry (e.g., crates.io) dependencies.
-#
-# If you are reading this file be aware that the original Cargo.toml
-# will likely look very different (and much more reasonable).
-# See Cargo.toml.orig for the original contents.
-
-[package]
-edition = "2021"
-rust-version = "1.75.0"
-name = "netstack-smoltcp"
-version = "0.2.2"
-authors = ["cavivie <cavivie@gmail.com>"]
-build = false
-autolib = false
-autobins = false
-autoexamples = false
-autotests = false
-autobenches = false
-description = """
-A netstack for the special purpose of turning packets from/to a TUN interface
-into TCP streams and UDP packets. It uses smoltcp-rs as the backend netstack.
-"""
-homepage = "https://github.com/cavivie/netstack-smoltcp"
-documentation = "https://docs.rs/netstack-smoltcp"
-readme = "README.md"
-keywords = [
-    "netstack",
-    "smoltcp",
-    "network",
-    "ip",
-    "tun",
-]
-categories = ["network-programming"]
-license = "MIT OR Apache-2.0"
-repository = "https://github.com/cavivie/netstack-smoltcp"
-
-[lib]
-name = "netstack_smoltcp"
-path = "src/lib.rs"
-
-[[example]]
-name = "forward"
-path = "examples/forward.rs"
-
-[[example]]
-name = "forward-offload-linux"
-path = "examples/forward-offload-linux.rs"
-
-[[test]]
-name = "regression"
-path = "tests/regression.rs"
-
-[dependencies.etherparse]
-version = "0.16"
-
-[dependencies.futures]
-version = "0.3"
-
-[dependencies.rand]
-version = "0.8"
-
-[dependencies.smoltcp]
-version = "0.12"
-features = [
-    "std",
-    "log",
-    "medium-ip",
-    "proto-ipv4",
-    "proto-ipv6",
-    "socket-icmp",
-    "socket-udp",
-    "socket-tcp",
-]
-default-features = false
-
-[dependencies.spin]
-version = "0.9"
-
-[dependencies.tokio]
-version = "1"
-features = [
-    "sync",
-    "time",
-    "rt",
-    "macros",
-]
-
-[dependencies.tokio-util]
-version = "0.7.10"
-
-[dependencies.tracing]
-version = "0.1"
-features = ["std"]
-default-features = false
-
-[dev-dependencies.socket2]
-version = "0.5.6"
-
-[dev-dependencies.socket2-ext]
-version = "0.1"
-
-[dev-dependencies.structopt]
-version = "0.3"
-
-[dev-dependencies.tokio]
-version = "1"
-features = [
-    "rt",
-    "macros",
-    "rt-multi-thread",
-    "io-util",
-]
-
-[dev-dependencies.tracing]
-version = "0.1"
-features = ["std"]
-default-features = false
-
-[dev-dependencies.tracing-subscriber]
-version = "0.3.18"
-
-[dev-dependencies.tun-rs]
-version = "2"
-features = [
-    "async",
-    "async_framed",
-]
-
-[dev-dependencies.tun2]
-version = "3"
-features = ["async"]

netstack-smoltcp/Cargo.toml.orig

@@ -1,51 +0,0 @@
-[package]
-name = "netstack-smoltcp"
-version = "0.2.2"
-edition = "2021"
-authors = ["cavivie <cavivie@gmail.com>"]
-license = "MIT OR Apache-2.0"
-repository = "https://github.com/cavivie/netstack-smoltcp"
-homepage = "https://github.com/cavivie/netstack-smoltcp"
-documentation = "https://docs.rs/netstack-smoltcp"
-keywords = ["netstack", "smoltcp", "network", "ip", "tun"]
-categories = ["network-programming"]
-description = """
-A netstack for the special purpose of turning packets from/to a TUN interface
-into TCP streams and UDP packets. It uses smoltcp-rs as the backend netstack.
-"""
-rust-version = "1.75.0"
-
-[dependencies]
-tracing = { version = "0.1", default-features = false, features = ["std"] }
-tokio = { version = "1", features = ["sync", "time", "rt", "macros"] }
-tokio-util = "0.7.10"
-etherparse = "0.16"
-futures = "0.3"
-rand = "0.8"
-spin = "0.9"
-smoltcp = { version = "0.12", default-features = false, features = [
-    "std",
-    "log",
-    "medium-ip",
-    "proto-ipv4",
-    "proto-ipv6",
-    "socket-icmp",
-    "socket-udp",
-    "socket-tcp",
-] }
-
-[dev-dependencies]
-tun2 = { version = "3", features = ["async"] }
-# has better performance on linux than tun2
-tun-rs = { version = "2", features = ["async", "async_framed"] } 
-tokio = { version = "1", features = [
-    "rt",
-    "macros",
-    "rt-multi-thread",
-    "io-util",
-] }
-tracing = { version = "0.1", default-features = false, features = ["std"] }
-tracing-subscriber = "0.3.18"
-structopt = "0.3"
-socket2 = "0.5.6"
-socket2-ext = { version = "0.1" }

netstack-smoltcp/LICENSE-APACHE

@@ -1,201 +0,0 @@
-                                 Apache License
-                           Version 2.0, January 2004
-                        http://www.apache.org/licenses/
-
-   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
-
-   1. Definitions.
-
-      "License" shall mean the terms and conditions for use, reproduction,
-      and distribution as defined by Sections 1 through 9 of this document.
-
-      "Licensor" shall mean the copyright owner or entity authorized by
-      the copyright owner that is granting the License.
-
-      "Legal Entity" shall mean the union of the acting entity and all
-      other entities that control, are controlled by, or are under common
-      control with that entity. For the purposes of this definition,
-      "control" means (i) the power, direct or indirect, to cause the
-      direction or management of such entity, whether by contract or
-      otherwise, or (ii) ownership of fifty percent (50%) or more of the
-      outstanding shares, or (iii) beneficial ownership of such entity.
-
-      "You" (or "Your") shall mean an individual or Legal Entity
-      exercising permissions granted by this License.
-
-      "Source" form shall mean the preferred form for making modifications,
-      including but not limited to software source code, documentation
-      source, and configuration files.
-
-      "Object" form shall mean any form resulting from mechanical
-      transformation or translation of a Source form, including but
-      not limited to compiled object code, generated documentation,
-      and conversions to other media types.
-
-      "Work" shall mean the work of authorship, whether in Source or
-      Object form, made available under the License, as indicated by a
-      copyright notice that is included in or attached to the work
-      (an example is provided in the Appendix below).
-
-      "Derivative Works" shall mean any work, whether in Source or Object
-      form, that is based on (or derived from) the Work and for which the
-      editorial revisions, annotations, elaborations, or other modifications
-      represent, as a whole, an original work of authorship. For the purposes
-      of this License, Derivative Works shall not include works that remain
-      separable from, or merely link (or bind by name) to the interfaces of,
-      the Work and Derivative Works thereof.
-
-      "Contribution" shall mean any work of authorship, including
-      the original version of the Work and any modifications or additions
-      to that Work or Derivative Works thereof, that is intentionally
-      submitted to Licensor for inclusion in the Work by the copyright owner
-      or by an individual or Legal Entity authorized to submit on behalf of
-      the copyright owner. For the purposes of this definition, "submitted"
-      means any form of electronic, verbal, or written communication sent
-      to the Licensor or its representatives, including but not limited to
-      communication on electronic mailing lists, source code control systems,
-      and issue tracking systems that are managed by, or on behalf of, the
-      Licensor for the purpose of discussing and improving the Work, but
-      excluding communication that is conspicuously marked or otherwise
-      designated in writing by the copyright owner as "Not a Contribution."
-
-      "Contributor" shall mean Licensor and any individual or Legal Entity
-      on behalf of whom a Contribution has been received by Licensor and
-      subsequently incorporated within the Work.
-
-   2. Grant of Copyright License. Subject to the terms and conditions of
-      this License, each Contributor hereby grants to You a perpetual,
-      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
-      copyright license to reproduce, prepare Derivative Works of,
-      publicly display, publicly perform, sublicense, and distribute the
-      Work and such Derivative Works in Source or Object form.
-
-   3. Grant of Patent License. Subject to the terms and conditions of
-      this License, each Contributor hereby grants to You a perpetual,
-      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
-      (except as stated in this section) patent license to make, have made,
-      use, offer to sell, sell, import, and otherwise transfer the Work,
-      where such license applies only to those patent claims licensable
-      by such Contributor that are necessarily infringed by their
-      Contribution(s) alone or by combination of their Contribution(s)
-      with the Work to which such Contribution(s) was submitted. If You
-      institute patent litigation against any entity (including a
-      cross-claim or counterclaim in a lawsuit) alleging that the Work
-      or a Contribution incorporated within the Work constitutes direct
-      or contributory patent infringement, then any patent licenses
-      granted to You under this License for that Work shall terminate
-      as of the date such litigation is filed.
-
-   4. Redistribution. You may reproduce and distribute copies of the
-      Work or Derivative Works thereof in any medium, with or without
-      modifications, and in Source or Object form, provided that You
-      meet the following conditions:
-
-      (a) You must give any other recipients of the Work or
-          Derivative Works a copy of this License; and
-
-      (b) You must cause any modified files to carry prominent notices
-          stating that You changed the files; and
-
-      (c) You must retain, in the Source form of any Derivative Works
-          that You distribute, all copyright, patent, trademark, and
-          attribution notices from the Source form of the Work,
-          excluding those notices that do not pertain to any part of
-          the Derivative Works; and
-
-      (d) If the Work includes a "NOTICE" text file as part of its
-          distribution, then any Derivative Works that You distribute must
-          include a readable copy of the attribution notices contained
-          within such NOTICE file, excluding those notices that do not
-          pertain to any part of the Derivative Works, in at least one
-          of the following places: within a NOTICE text file distributed
-          as part of the Derivative Works; within the Source form or
-          documentation, if provided along with the Derivative Works; or,
-          within a display generated by the Derivative Works, if and
-          wherever such third-party notices normally appear. The contents
-          of the NOTICE file are for informational purposes only and
-          do not modify the License. You may add Your own attribution
-          notices within Derivative Works that You distribute, alongside
-          or as an addendum to the NOTICE text from the Work, provided
-          that such additional attribution notices cannot be construed
-          as modifying the License.
-
-      You may add Your own copyright statement to Your modifications and
-      may provide additional or different license terms and conditions
-      for use, reproduction, or distribution of Your modifications, or
-      for any such Derivative Works as a whole, provided Your use,
-      reproduction, and distribution of the Work otherwise complies with
-      the conditions stated in this License.
-
-   5. Submission of Contributions. Unless You explicitly state otherwise,
-      any Contribution intentionally submitted for inclusion in the Work
-      by You to the Licensor shall be under the terms and conditions of
-      this License, without any additional terms or conditions.
-      Notwithstanding the above, nothing herein shall supersede or modify
-      the terms of any separate license agreement you may have executed
-      with Licensor regarding such Contributions.
-
-   6. Trademarks. This License does not grant permission to use the trade
-      names, trademarks, service marks, or product names of the Licensor,
-      except as required for reasonable and customary use in describing the
-      origin of the Work and reproducing the content of the NOTICE file.
-
-   7. Disclaimer of Warranty. Unless required by applicable law or
-      agreed to in writing, Licensor provides the Work (and each
-      Contributor provides its Contributions) on an "AS IS" BASIS,
-      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
-      implied, including, without limitation, any warranties or conditions
-      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
-      PARTICULAR PURPOSE. You are solely responsible for determining the
-      appropriateness of using or redistributing the Work and assume any
-      risks associated with Your exercise of permissions under this License.
-
-   8. Limitation of Liability. In no event and under no legal theory,
-      whether in tort (including negligence), contract, or otherwise,
-      unless required by applicable law (such as deliberate and grossly
-      negligent acts) or agreed to in writing, shall any Contributor be
-      liable to You for damages, including any direct, indirect, special,
-      incidental, or consequential damages of any character arising as a
-      result of this License or out of the use or inability to use the
-      Work (including but not limited to damages for loss of goodwill,
-      work stoppage, computer failure or malfunction, or any and all
-      other commercial damages or losses), even if such Contributor
-      has been advised of the possibility of such damages.
-
-   9. Accepting Warranty or Additional Liability. While redistributing
-      the Work or Derivative Works thereof, You may choose to offer,
-      and charge a fee for, acceptance of support, warranty, indemnity,
-      or other liability obligations and/or rights consistent with this
-      License. However, in accepting such obligations, You may act only
-      on Your own behalf and on Your sole responsibility, not on behalf
-      of any other Contributor, and only if You agree to indemnify,
-      defend, and hold each Contributor harmless for any liability
-      incurred by, or claims asserted against, such Contributor by reason
-      of your accepting any such warranty or additional liability.
-
-   END OF TERMS AND CONDITIONS
-
-   APPENDIX: How to apply the Apache License to your work.
-
-      To apply the Apache License to your work, attach the following
-      boilerplate notice, with the fields enclosed by brackets "[]"
-      replaced with your own identifying information. (Don't include
-      the brackets!)  The text should be enclosed in the appropriate
-      comment syntax for the file format. We also recommend that a
-      file or class name and description of purpose be included on the
-      same "printed page" as the copyright notice for easier
-      identification within third-party archives.
-
-   Copyright [yyyy] [name of copyright owner]
-
-   Licensed under the Apache License, Version 2.0 (the "License");
-   you may not use this file except in compliance with the License.
-   You may obtain a copy of the License at
-
-       http://www.apache.org/licenses/LICENSE-2.0
-
-   Unless required by applicable law or agreed to in writing, software
-   distributed under the License is distributed on an "AS IS" BASIS,
-   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-   See the License for the specific language governing permissions and
-   limitations under the License.

netstack-smoltcp/LICENSE-MIT

@@ -1,25 +0,0 @@
-Copyright (c) 2024 cavivie and netstack-smoltcp Contributors
-
-Permission is hereby granted, free of charge, to any
-person obtaining a copy of this software and associated
-documentation files (the "Software"), to deal in the
-Software without restriction, including without
-limitation the rights to use, copy, modify, merge,
-publish, distribute, sublicense, and/or sell copies of
-the Software, and to permit persons to whom the Software
-is furnished to do so, subject to the following
-conditions:
-
-The above copyright notice and this permission notice
-shall be included in all copies or substantial portions
-of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF
-ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED
-TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A
-PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT
-SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
-CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR
-IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
-DEALINGS IN THE SOFTWARE.

netstack-smoltcp/README.md

@@ -1,136 +0,0 @@
-# Netstack Smoltcp
-
-A netstack for the special purpose of turning packets from/to a TUN interface into TCP streams and UDP packets. It uses smoltcp-rs as the backend netstack.
-
-[![Crates.io][crates-badge]][crates-url]
-[![MIT licensed][mit-badge]][mit-url]
-[![Apache licensed, Version 2.0][apache-badge]][apache-url]
-[![Build Status][actions-badge]][actions-url]
-
-[crates-badge]: https://img.shields.io/crates/v/netstack-smoltcp.svg
-[crates-url]: https://crates.io/crates/netstack-smoltcp
-[mit-badge]: https://img.shields.io/badge/license-MIT-blue.svg
-[mit-url]: https://github.com/automesh-network/netstack-smoltcp/blob/master/LICENSE-MIT
-[apache-badge]: https://img.shields.io/badge/license-APACHE2.0-blue.svg
-[apache-url]: https://github.com/automesh-network/netstack-smoltcp/blob/master/LICENSE-APACHE
-[actions-badge]: https://github.com/automesh-network/netstack-smoltcp/workflows/CI/badge.svg
-[actions-url]: https://github.com/automesh-network/netstack-smoltcp/actions?query=workflow%3ACI+branch%3Amain
-
-## Features
-
-- Supports Future Send and non-Send, mostly pepole use Send.
-- Supports ICMP protocol drive by TCP runner to use ICMP ping.
-- Supports filtering packets by source and destination IP addresses.
-- Can read IP packets from netstack, write IP packets to netstack.
-- Can receive TcpStream from TcpListener exposed from netstack.
-- Can receive UDP datagram from UdpSocket exposed from netstack.
-- Implements popular future streaming traits and asynchronous IO traits:
-    * TcpListener implements futures Stream/Sink trait
-    * TcpStream implements tokio AsyncRead/AsyncWrite trait
-    * UdpSocket(ReadHalf/WriteHalf) implements futures Stream/Sink trait.
-
-## Platforms
-
-This crate provides lightweight netstack support for Linux, iOS, macOS, Android and Windows.
-Currently, it works on most targets, but mainly tested the popular platforms which includes:
-- linux-amd64: x86_64-unknown-linux-gnu
-- android-arm64: aarch64-linux-android
-- android-amd64: x86_64-linux-android
-- macos-amd64: x86_64-apple-darwin
-- macos-arm64: aarch64-apple-darwin
-- ios-arm64: aarch64-apple-ios
-- windows-amd64: x86_64-pc-windows-msvc
-- windows-arm64: aarch64-pc-windows-msvc
-
-## Example
-
-```rust
-// let device = tun2::create_as_async(&cfg)?;
-// let framed = device.into_framed();
-
-let (stack, runner, udp_socket, tcp_listener) = netstack_smoltcp::StackBuilder::default()
-    .stack_buffer_size(512)
-    .tcp_buffer_size(4096)
-    .enable_udp(true)
-    .enable_tcp(true)
-    .enable_icmp(true)
-    .mtu(9000) // virtual device usually benefits from larger MTU
-    .build()
-    .unwrap();
-let mut udp_socket = udp_socket.unwrap(); // udp enabled
-let mut tcp_listener = tcp_listener.unwrap(); // tcp/icmp enabled
-if let Some(runner) = runner {
-    tokio::spawn(runner);
-}
-
-let (mut stack_sink, mut stack_stream) = stack.split();
-let (mut tun_sink, mut tun_stream) = framed.split();
-
-// Reads packet from stack and sends to TUN.
-tokio::spawn(async move {
-    while let Some(pkt) = stack_stream.next().await {
-        if let Ok(pkt) = pkt {
-            tun_sink.send(pkt).await.unwrap();
-        }
-    }
-});
-
-// Reads packet from TUN and sends to stack.
-tokio::spawn(async move {
-    while let Some(pkt) = tun_stream.next().await {
-        if let Ok(pkt) = pkt {
-            stack_sink.send(pkt).await.unwrap();
-        }
-    }
-});
-
-// Extracts TCP connections from stack and sends them to the dispatcher.
-tokio::spawn(async move {
-    handle_inbound_stream(tcp_listener).await;
-});
-
-// Receive and send UDP packets between netstack and NAT manager. The NAT
-// manager would maintain UDP sessions and send them to the dispatcher.
-tokio::spawn(async move {
-    handle_inbound_datagram(udp_socket).await;
-});
-```
-
-## Performance
-
-Typically, `netstack-smoltcp` will be used with an tun device, so a careful choice of TUN crate matters.
-
-[tun-rs](https://github.com/tun-rs/tun-rs) have better performance on **Linux** than [rust-tun](https://github.com/meh/rust-tun/) due to GSO/GRO which allow you to process the packets in batches.
-
-`bash scripts/bench-offload.sh` could tell that `tun-rs` boosts the performance by 4x. Try it out on your Linux machine!
-
-The example for using `tun-rs` with `netstack-smoltcp` could be found at [forward-offload-linux.rs](examples/forward-offload-linux.rs)
-
-For further tuning, refer to `tun-rs`'s detailed [README](https://github.com/tun-rs/tun-rs/blob/main/README.md)
-
-## License
-
-This project is licensed under either of
-
- * Apache License, Version 2.0, ([LICENSE-APACHE](LICENSE-APACHE) or
-   https://www.apache.org/licenses/LICENSE-2.0)
- * MIT license ([LICENSE-MIT](LICENSE-MIT) or
-   https://opensource.org/licenses/MIT)
-
-at your option.
-
-### Contribution
-
-Unless you explicitly state otherwise, any contribution intentionally submitted
-for inclusion in netstack-smoltcp by you, as defined in the Apache-2.0 license,
-shall be dual licensed as above, without any additional terms or conditions.
-
-## Inspired By
-
-Special thanks to these amazing projects that inspired netstack-smoltcp (in no particular order):
-- [shadowsocks-rust](https://github.com/shadowsocks/shadowsocks-rust/)
-- [netstack-lwip](https://github.com/eycorsican/netstack-lwip/)
-- [rust-tun-active](https://github.com/tun2proxy/rust-tun)
-- [rust-tun](https://github.com/meh/rust-tun/)
-- [tun-rs](https://github.com/tun-rs/tun-rs)
-- [smoltcp](https://github.com/smoltcp-rs/smoltcp)

netstack-smoltcp/examples/forward-offload-linux.rs

@@ -1,239 +0,0 @@
-#[cfg(target_os = "linux")]
-mod inner {
-    use futures::{SinkExt, StreamExt};
-    use netstack_smoltcp::{StackBuilder, TcpListener, UdpSocket};
-    use std::{net::SocketAddr, sync::Arc};
-    use structopt::StructOpt;
-    use tokio::net::{TcpSocket, TcpStream};
-    use tracing::{error, info, warn};
-    use tun_rs::{DeviceBuilder, IDEAL_BATCH_SIZE, VIRTIO_NET_HDR_LEN};
-
-    // Patched forward example: tun2 → tun-rs with Linux GRO/GSO offload.
-    // For further reading, check out https://blog.cloudflare.com/virtual-networking-101-understanding-tap
-    //
-    // Key changes vs forward.rs:
-    //   1. Use tun-rs DeviceBuilder with .offload(true) on Linux (enables
-    //      IFF_VNET_HDR + TUN_F_CSUM/TSO4/TSO6/USO4/USO6).
-    //   2. TX (stack → TUN): prepend 10-byte zero virtio_net_hdr (GSO_NONE)
-    //      so the kernel accepts the write when IFF_VNET_HDR is set.
-    //   3. RX (TUN → stack): use recv_multiple() for batch GSO splitting;
-    //      buffers sized to 1600 to fit smoltcp's 1504-byte MTU segments.
-    #[derive(Debug, StructOpt)]
-    #[structopt(name = "forward", about = "Simply forward tun tcp/udp traffic.")]
-    struct Opt {
-        /// Outbound interface to bind forwarded connections to.
-        #[structopt(short = "i", long = "interface")]
-        interface: String,
-        /// Name of the TUN device.
-        #[structopt(short = "n", long = "name", default_value = "utun8")]
-        name: String,
-        /// Tracing log level.
-        #[structopt(long = "log-level", default_value = "debug")]
-        log_level: tracing::Level,
-        /// Use current-thread Tokio runtime (default: multi-thread).
-        #[structopt(long = "current-thread")]
-        current_thread: bool,
-        /// Use spawn_local instead of spawn.
-        #[structopt(long = "local-task")]
-        local_task: bool,
-    }
-
-    pub(super) fn main() {
-        let opt = Opt::from_args();
-        let rt = if opt.current_thread {
-            tokio::runtime::Builder::new_current_thread()
-        } else {
-            tokio::runtime::Builder::new_multi_thread()
-        }
-        .enable_all()
-        .build()
-        .unwrap();
-        rt.block_on(main_exec(opt));
-    }
-
-    async fn main_exec(opt: Opt) {
-        macro_rules! tokio_spawn {
-            ($fut:expr) => {
-                if opt.local_task {
-                    tokio::task::spawn_local($fut)
-                } else {
-                    tokio::task::spawn($fut)
-                }
-            };
-        }
-        tracing::subscriber::set_global_default(
-            tracing_subscriber::FmtSubscriber::builder()
-                .with_max_level(opt.log_level)
-                .finish(),
-        )
-        .unwrap();
-
-        // Build TUN device with GRO/GSO offload on Linux.
-        let builder = DeviceBuilder::new()
-            .name(opt.name)
-            .ipv4("10.10.10.2", 24, Some("10.10.10.1"))
-            .mtu(9000);
-        let builder = builder.offload(true);
-        let dev = Arc::new(builder.build_async().unwrap());
-
-        let (stack, runner, udp_socket, tcp_listener) = StackBuilder::default()
-            .enable_tcp(true)
-            .enable_udp(true)
-            .enable_icmp(true)
-            .build()
-            .unwrap();
-        let udp_socket = udp_socket.unwrap();
-        let tcp_listener = tcp_listener.unwrap();
-        if let Some(runner) = runner {
-            tokio_spawn!(runner);
-        }
-        let (mut stack_sink, mut stack_stream) = stack.split();
-
-        let mut futs = vec![];
-
-        // stack → TUN
-        // With IFF_VNET_HDR every write must start with a virtio_net_hdr.
-        // We use all-zero (gso_type = GSO_NONE, flags = 0): plain packet,
-        // checksum already valid (smoltcp always computes checksums itself).
-        let dev1 = dev.clone();
-        futs.push(tokio_spawn!(async move {
-            while let Some(pkt) = stack_stream.next().await {
-                if let Ok(pkt) = pkt {
-                    let result = {
-                        let mut buf = vec![0u8; VIRTIO_NET_HDR_LEN + pkt.len()];
-                        buf[VIRTIO_NET_HDR_LEN..].copy_from_slice(&pkt);
-                        dev1.send(&buf).await
-                    };
-                    if let Err(e) = result {
-                        warn!("failed to send packet to TUN: {:?}", e);
-                    }
-                }
-            }
-        }));
-
-        // TUN → stack
-        // recv_multiple() does one read() syscall and returns N individual IP
-        // packets after splitting any incoming GRO super-packet.
-        // Buffer size 1600 > smoltcp MTU (1504) to avoid an out-of-bounds panic
-        // when the kernel segments at MSS=1464 with 40-byte IP+TCP headers.
-        futs.push(tokio_spawn!(async move {
-            let mut orig = vec![0u8; VIRTIO_NET_HDR_LEN + 65535];
-            let mut bufs = vec![vec![0u8; 1600]; IDEAL_BATCH_SIZE];
-            let mut sizes = vec![0usize; IDEAL_BATCH_SIZE];
-            while let Ok(n) = dev.recv_multiple(&mut orig, &mut bufs, &mut sizes, 0).await {
-                for i in 0..n {
-                    let pkt = &bufs[i][..sizes[i]];
-                    if let Err(e) = stack_sink.send(pkt.to_vec()).await {
-                        warn!("failed to send packet to stack: {:?}", e);
-                    }
-                }
-            }
-        }));
-
-        futs.push(tokio_spawn!({
-            let iface = opt.interface.clone();
-            async move {
-                handle_inbound_stream(tcp_listener, iface).await;
-            }
-        }));
-
-        futs.push(tokio_spawn!(async move {
-            handle_inbound_datagram(udp_socket, opt.interface).await;
-        }));
-
-        futures::future::join_all(futs).await.iter().for_each(|r| {
-            if let Err(e) = r {
-                error!("{:?}", e);
-            }
-        });
-    }
-
-    async fn handle_inbound_stream(mut tcp_listener: TcpListener, interface: String) {
-        while let Some((mut stream, local, remote)) = tcp_listener.next().await {
-            let interface = interface.clone();
-            tokio::spawn(async move {
-                info!("tcp: {:?} => {:?}", local, remote);
-                match new_tcp_stream(remote, &interface).await {
-                    Ok(mut r) => {
-                        if let Err(e) = tokio::io::copy_bidirectional(&mut stream, &mut r).await {
-                            warn!(
-                                "failed to copy tcp stream {:?}=>{:?}: {:?}",
-                                local, remote, e
-                            );
-                        }
-                    }
-                    Err(e) => warn!(
-                        "failed to open tcp stream {:?}=>{:?}: {:?}",
-                        local, remote, e
-                    ),
-                }
-            });
-        }
-    }
-
-    async fn handle_inbound_datagram(udp_socket: UdpSocket, interface: String) {
-        let (tx, mut rx) = tokio::sync::mpsc::unbounded_channel();
-        let (mut read_half, mut write_half) = udp_socket.split();
-        tokio::spawn(async move {
-            while let Some((data, local, remote)) = rx.recv().await {
-                let _ = write_half.send((data, remote, local)).await;
-            }
-        });
-        while let Some((data, local, remote)) = read_half.next().await {
-            let tx = tx.clone();
-            let interface = interface.clone();
-            tokio::spawn(async move {
-                match new_udp_packet(remote, &interface).await {
-                    Ok(sock) => {
-                        let _ = sock.send(&data).await;
-                        loop {
-                            let mut buf = vec![0; 1024];
-                            match sock.recv_from(&mut buf).await {
-                                Ok((n, _)) => {
-                                    let _ = tx.send((buf[..n].to_vec(), local, remote));
-                                }
-                                Err(e) => {
-                                    warn!("udp recv {:?}: {:?}", remote, e);
-                                    break;
-                                }
-                            }
-                        }
-                    }
-                    Err(e) => warn!("failed to open udp socket {:?}: {:?}", remote, e),
-                }
-            });
-        }
-    }
-
-    async fn new_tcp_stream(addr: SocketAddr, iface: &str) -> std::io::Result<TcpStream> {
-        use socket2_ext::{AddressBinding, BindDeviceOption};
-        let s = socket2::Socket::new(socket2::Domain::IPV4, socket2::Type::STREAM, None)?;
-        s.bind_to_device(BindDeviceOption::v4(iface))?;
-        s.set_keepalive(true)?;
-        s.set_nodelay(true)?;
-        s.set_nonblocking(true)?;
-        Ok(TcpSocket::from_std_stream(s.into()).connect(addr).await?)
-    }
-
-    async fn new_udp_packet(
-        addr: SocketAddr,
-        iface: &str,
-    ) -> std::io::Result<tokio::net::UdpSocket> {
-        use socket2_ext::{AddressBinding, BindDeviceOption};
-        let s = socket2::Socket::new(socket2::Domain::IPV4, socket2::Type::DGRAM, None)?;
-        s.bind_to_device(BindDeviceOption::v4(iface))?;
-        s.set_nonblocking(true)?;
-        let sock = tokio::net::UdpSocket::from_std(s.into())?;
-        sock.connect(addr).await?;
-        Ok(sock)
-    }
-}
-
-#[cfg(not(target_os = "linux"))]
-mod inner {
-    pub(super) fn main() {}
-}
-
-fn main() {
-    inner::main();
-}

netstack-smoltcp/examples/forward.rs

@@ -1,326 +0,0 @@
-use std::net::{IpAddr, SocketAddr};
-
-use futures::{SinkExt, StreamExt};
-use netstack_smoltcp::{StackBuilder, TcpListener, UdpSocket};
-use structopt::StructOpt;
-use tokio::net::{TcpSocket, TcpStream};
-use tracing::{error, info, warn};
-
-// to run this example, you should set the policy routing **after the start of the main program**
-//
-// linux:
-// with bind device:
-// `curl 1.1.1.1 --interface utun8`
-// with default route:
-// `bash scripts/route-linux.sh add`
-// `curl 1.1.1.1`
-// with single route:
-// `ip rule add to 1.1.1.1 table 200`
-// `ip route add default dev utun8 table 200`
-// `curl 1.1.1.1`
-//
-// macos:
-// with default route:
-// `bash scripts/route-macos.sh add`
-// `curl 1.1.1.1`
-//
-// windows:
-// with default route:
-// tun2 set default route automatically, won't set agian
-// # `powershell.exe scripts/route-windows.ps1 add`
-// `curl 1.1.1.1`
-//
-// currently, the example only supports the TCP stream, and the UDP packet will be dropped.
-
-#[derive(Debug, StructOpt)]
-#[structopt(name = "forward", about = "Simply forward tun tcp/udp traffic.")]
-struct Opt {
-    /// Default binding interface, default by guessed.
-    /// Specify but doesn't exist, no device is bound.
-    #[structopt(short = "i", long = "interface")]
-    interface: String,
-
-    /// name of the tun device, default to rtun8.
-    #[structopt(short = "n", long = "name", default_value = "utun8")]
-    name: String,
-
-    /// Tracing subscriber log level.
-    #[structopt(long = "log-level", default_value = "debug")]
-    log_level: tracing::Level,
-
-    /// Tokio current-thread runtime, default to multi-thread.
-    #[structopt(long = "current-thread")]
-    current_thread: bool,
-
-    /// Tokio task spawn_local, default to spwan.
-    #[structopt(long = "local-task")]
-    local_task: bool,
-}
-
-fn main() {
-    let opt = Opt::from_args();
-
-    let rt = if opt.current_thread {
-        tokio::runtime::Builder::new_current_thread()
-    } else {
-        tokio::runtime::Builder::new_multi_thread()
-    }
-    .enable_all()
-    .build()
-    .unwrap();
-
-    rt.block_on(main_exec(opt));
-}
-
-async fn main_exec(opt: Opt) {
-    macro_rules! tokio_spawn {
-        ($fut: expr) => {
-            if opt.local_task {
-                tokio::task::spawn_local($fut)
-            } else {
-                tokio::task::spawn($fut)
-            }
-        };
-    }
-
-    tracing::subscriber::set_global_default(
-        tracing_subscriber::FmtSubscriber::builder()
-            .with_max_level(opt.log_level)
-            .finish(),
-    )
-    .unwrap();
-
-    let mut cfg = tun2::Configuration::default();
-    cfg.layer(tun2::Layer::L3);
-    let fd = -1;
-    if fd >= 0 {
-        cfg.raw_fd(fd);
-    } else {
-        cfg.tun_name(&opt.name)
-            .address("10.10.10.2")
-            .destination("10.10.10.1")
-            .mtu(tun2::DEFAULT_MTU);
-        #[cfg(not(any(target_arch = "mips", target_arch = "mips64",)))]
-        {
-            cfg.netmask("255.255.255.0");
-        }
-        cfg.up();
-    }
-
-    let device = tun2::create_as_async(&cfg).unwrap();
-    let mut builder = StackBuilder::default()
-        .enable_tcp(true)
-        .enable_udp(true)
-        .enable_icmp(true)
-        .mtu(9000);
-    if let Some(device_broadcast) = get_device_broadcast(&device) {
-        builder = builder
-            // .add_ip_filter(Box::new(move |src, dst| *src != device_broadcast && *dst != device_broadcast));
-            .add_ip_filter_fn(move |src, dst| *src != device_broadcast && *dst != device_broadcast);
-    }
-
-    let (stack, runner, udp_socket, tcp_listener) = builder.build().unwrap();
-    let udp_socket = udp_socket.unwrap(); // udp enabled
-    let tcp_listener = tcp_listener.unwrap(); // tcp enabled or icmp enabled
-
-    if let Some(runner) = runner {
-        tokio_spawn!(runner);
-    }
-
-    let framed = device.into_framed();
-    let (mut tun_sink, mut tun_stream) = framed.split();
-    let (mut stack_sink, mut stack_stream) = stack.split();
-
-    let mut futs = vec![];
-
-    // Reads packet from stack and sends to TUN.
-    futs.push(tokio_spawn!(async move {
-        while let Some(pkt) = stack_stream.next().await {
-            if let Ok(pkt) = pkt {
-                match tun_sink.send(pkt).await {
-                    Ok(_) => {}
-                    Err(e) => warn!("failed to send packet to TUN, err: {:?}", e),
-                }
-            }
-        }
-    }));
-
-    // Reads packet from TUN and sends to stack.
-    futs.push(tokio_spawn!(async move {
-        while let Some(pkt) = tun_stream.next().await {
-            if let Ok(pkt) = pkt {
-                match stack_sink.send(pkt).await {
-                    Ok(_) => {}
-                    Err(e) => warn!("failed to send packet to stack, err: {:?}", e),
-                };
-            }
-        }
-    }));
-
-    // Extracts TCP connections from stack and sends them to the dispatcher.
-    futs.push(tokio_spawn!({
-        let interface = opt.interface.clone();
-        async move {
-            handle_inbound_stream(tcp_listener, interface).await;
-        }
-    }));
-
-    // Receive and send UDP packets between netstack and NAT manager. The NAT
-    // manager would maintain UDP sessions and send them to the dispatcher.
-    futs.push(tokio_spawn!(async move {
-        handle_inbound_datagram(udp_socket, opt.interface).await;
-    }));
-
-    futures::future::join_all(futs)
-        .await
-        .iter()
-        .for_each(|res| {
-            if let Err(e) = res {
-                error!("error: {:?}", e);
-            }
-        });
-}
-
-/// simply forward tcp stream
-async fn handle_inbound_stream(mut tcp_listener: TcpListener, interface: String) {
-    while let Some((mut stream, local, remote)) = tcp_listener.next().await {
-        let interface = interface.clone();
-        tokio::spawn(async move {
-            info!("new tcp connection: {:?} => {:?}", local, remote);
-            match new_tcp_stream(remote, &interface).await {
-                Ok(mut remote_stream) => {
-                    // pipe between two tcp stream
-                    match tokio::io::copy_bidirectional(&mut stream, &mut remote_stream).await {
-                        Ok(_) => {}
-                        Err(e) => warn!(
-                            "failed to copy tcp stream {:?}=>{:?}, err: {:?}",
-                            local, remote, e
-                        ),
-                    }
-                }
-                Err(e) => warn!(
-                    "failed to new tcp stream {:?}=>{:?}, err: {:?}",
-                    local, remote, e
-                ),
-            }
-        });
-    }
-}
-
-/// simply forward udp datagram
-async fn handle_inbound_datagram(udp_socket: UdpSocket, interface: String) {
-    let (tx, mut rx) = tokio::sync::mpsc::unbounded_channel();
-    let (mut read_half, mut write_half) = udp_socket.split();
-    tokio::spawn(async move {
-        while let Some((data, local, remote)) = rx.recv().await {
-            let _ = write_half.send((data, remote, local)).await;
-        }
-    });
-
-    while let Some((data, local, remote)) = read_half.next().await {
-        let tx = tx.clone();
-        let interface = interface.clone();
-        tokio::spawn(async move {
-            info!("new udp datagram: {:?} => {:?}", local, remote);
-            match new_udp_packet(remote, &interface).await {
-                Ok(remote_socket) => {
-                    // pipe between two udp sockets
-                    let _ = remote_socket.send(&data).await;
-                    loop {
-                        let mut buf = vec![0; 1024];
-                        match remote_socket.recv_from(&mut buf).await {
-                            Ok((len, _)) => {
-                                let _ = tx.send((buf[..len].to_vec(), local, remote));
-                            }
-                            Err(e) => {
-                                warn!(
-                                    "failed to recv udp datagram {:?}<->{:?}: {:?}",
-                                    local, remote, e
-                                );
-                                break;
-                            }
-                        }
-                    }
-                }
-                Err(e) => warn!(
-                    "failed to new udp socket {:?}=>{:?}, err: {:?}",
-                    local, remote, e
-                ),
-            }
-        });
-    }
-}
-
-async fn new_tcp_stream<'a>(addr: SocketAddr, iface: &str) -> std::io::Result<TcpStream> {
-    use socket2_ext::{AddressBinding, BindDeviceOption};
-    let socket = socket2::Socket::new(socket2::Domain::IPV4, socket2::Type::STREAM, None)?;
-    socket.bind_to_device(BindDeviceOption::v4(iface))?;
-    socket.set_keepalive(true)?;
-    socket.set_nodelay(true)?;
-    socket.set_nonblocking(true)?;
-
-    let stream = TcpSocket::from_std_stream(socket.into())
-        .connect(addr)
-        .await?;
-
-    Ok(stream)
-}
-
-async fn new_udp_packet(addr: SocketAddr, iface: &str) -> std::io::Result<tokio::net::UdpSocket> {
-    use socket2_ext::{AddressBinding, BindDeviceOption};
-    let socket = socket2::Socket::new(socket2::Domain::IPV4, socket2::Type::DGRAM, None)?;
-    socket.bind_to_device(BindDeviceOption::v4(iface))?;
-    socket.set_nonblocking(true)?;
-
-    let socket = tokio::net::UdpSocket::from_std(socket.into());
-    if let Ok(ref socket) = socket {
-        socket.connect(addr).await?;
-    }
-    socket
-}
-
-fn get_device_broadcast(device: &tun2::AsyncDevice) -> Option<std::net::Ipv4Addr> {
-    use tun2::AbstractDevice;
-
-    let mtu = device.mtu().unwrap_or(tun2::DEFAULT_MTU);
-
-    let address = match device.address() {
-        Ok(a) => match a {
-            IpAddr::V4(v4) => v4,
-            IpAddr::V6(_) => return None,
-        },
-        Err(_) => return None,
-    };
-
-    let netmask = match device.netmask() {
-        Ok(n) => match n {
-            IpAddr::V4(v4) => v4,
-            IpAddr::V6(_) => return None,
-        },
-        Err(_) => return None,
-    };
-
-    match smoltcp::wire::Ipv4Cidr::from_netmask(address, netmask) {
-        Ok(address_net) => match address_net.broadcast() {
-            Some(broadcast) => {
-                info!(
-                    "tun device network: {} (address: {}, netmask: {}, broadcast: {}, mtu: {})",
-                    address_net, address, netmask, broadcast, mtu,
-                );
-
-                Some(broadcast)
-            }
-            None => {
-                error!("invalid tun address {}, netmask {}", address, netmask);
-                None
-            }
-        },
-        Err(err) => {
-            error!(
-                "invalid tun address {}, netmask {}, error: {}",
-                address, netmask, err
-            );
-            None
-        }
-    }
-}

netstack-smoltcp/scripts/bench-offload.sh

@@ -1,174 +0,0 @@
-#!/usr/bin/env bash
-# bench-offload.sh
-#
-# Benchmarks netstack-smoltcp's forward examples with 2-stream iperf3.
-# Compares:
-#   - examples/forward               (tun2, no GRO/GSO offload)
-#   - examples/forward-offload-linux (tun-rs, Linux GRO/GSO offload via IFF_VNET_HDR)
-#
-# Setup: creates a veth pair + network namespace; iperf3 server runs inside
-# the namespace, the forward proxy bridges traffic through a TUN device.
-#
-# Requirements: cargo, iperf3, ip (iproute2), root/CAP_NET_ADMIN
-#
-# Usage:
-#   sudo bash scripts/bench-offload.sh
-#
-# Run from the root of the netstack-smoltcp repository.
-
-set -euo pipefail
-
-# ── config ────────────────────────────────────────────────────────────────────
-REPO_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
-NS=bench
-VETH_HOST=veth-host
-VETH_NS=veth-bench
-HOST_IP=172.19.0.1
-NS_IP=172.19.0.2
-PREFIX=24
-TUN_NAME=utun8
-TUN_IP=10.10.10.2
-IPERF_PORT=5201
-DURATION=15
-STREAMS=2
-
-# ── helpers ───────────────────────────────────────────────────────────────────
-die()     { echo "ERROR: $*" >&2; exit 1; }
-require() { command -v "$1" &>/dev/null || die "'$1' not found"; }
-
-cleanup() {
-    pkill -f "forward-" 2>/dev/null || true
-    ip netns exec "$NS" pkill iperf3 2>/dev/null || true
-    ip route del "${NS_IP}/32" dev "$TUN_NAME" 2>/dev/null || true
-    ip tuntap del dev "$TUN_NAME" mode tun 2>/dev/null || true
-    ip link del "$VETH_HOST" 2>/dev/null || true
-    ip netns del "$NS" 2>/dev/null || true
-}
-trap cleanup EXIT
-
-# ── preflight ─────────────────────────────────────────────────────────────────
-require cargo
-require iperf3
-require ip
-[[ $EUID -eq 0 ]] || die "run as root (needs CAP_NET_ADMIN for TUN + netns)"
-[[ -f "$REPO_DIR/Cargo.toml" ]] || die "run from the netstack-smoltcp repo root"
-grep -q 'name = "netstack-smoltcp"' "$REPO_DIR/Cargo.toml" \
-    || die "Cargo.toml does not look like netstack-smoltcp"
-
-# ── network setup ─────────────────────────────────────────────────────────────
-echo "[net] setting up namespace '$NS' and veth pair..."
-cleanup 2>/dev/null || true
-sleep 0.5
-
-ip netns add "$NS"
-ip link add "$VETH_HOST" type veth peer name "$VETH_NS"
-ip link set "$VETH_NS" netns "$NS"
-ip addr add "${HOST_IP}/${PREFIX}" dev "$VETH_HOST"
-ip link set "$VETH_HOST" up
-ip netns exec "$NS" ip addr add "${NS_IP}/${PREFIX}" dev "$VETH_NS"
-ip netns exec "$NS" ip link set "$VETH_NS" up
-ip netns exec "$NS" ip link set lo up
-echo "[net] ${HOST_IP} <──veth──> ${NS_IP} (ns:${NS})"
-
-# ── build: forward (tun2, no offload) ────────────────────────────────────────
-echo ""
-echo "[build] examples/forward (tun2, no GRO/GSO offload)..."
-(
-    cd "$REPO_DIR"
-    cargo build --example forward --release --quiet
-    cp target/release/examples/forward /tmp/forward-tun2
-)
-echo "[build] done → /tmp/forward-tun2"
-
-# ── build: forward-offload-linux (tun-rs, GRO/GSO offload) ───────────────────
-echo ""
-echo "[build] examples/forward-offload-linux (tun-rs, GRO/GSO offload)..."
-(
-    cd "$REPO_DIR"
-    cargo build --example forward-offload-linux --release --quiet
-    cp target/release/examples/forward-offload-linux /tmp/forward-tun-rs
-)
-echo "[build] done → /tmp/forward-tun-rs"
-
-# ── benchmark runner ──────────────────────────────────────────────────────────
-run_bench() {
-    local label="$1" binary="$2"
-
-    # clean any leftover state
-    pkill -f "forward-" 2>/dev/null || true
-    ip netns exec "$NS" pkill iperf3 2>/dev/null || true
-    ip route del "${NS_IP}/32" dev "$TUN_NAME" 2>/dev/null || true
-    ip tuntap del dev "$TUN_NAME" mode tun 2>/dev/null || true
-    sleep 0.8
-
-    # start iperf3 server inside namespace
-    ip netns exec "$NS" iperf3 -s -p "$IPERF_PORT" -D \
-        --logfile /tmp/iperf3-bench-server.log
-
-    # start proxy
-    "$binary" -i "$VETH_HOST" -n "$TUN_NAME" --log-level warn &
-    sleep 2
-
-    ip link show "$TUN_NAME" &>/dev/null \
-        || { echo "  [!] TUN not up, skipping"; return 1; }
-
-    # route iperf3 traffic through TUN (more-specific /32 overrides /24 via veth)
-    ip route add "${NS_IP}/32" dev "$TUN_NAME"
-
-    echo "  running iperf3: ${STREAMS} streams × ${DURATION}s …"
-    local out
-    out=$(iperf3 -c "$NS_IP" -p "$IPERF_PORT" \
-              -t "$DURATION" -P "$STREAMS" 2>&1)
-
-    local sender receiver
-    sender=$(echo "$out"   | grep "SUM.*sender"   | awk '{print $6, $7}')
-    receiver=$(echo "$out" | grep "SUM.*receiver" | awk '{print $6, $7}')
-
-    if [[ -z "$sender" ]]; then
-        echo "  result: FAILED"
-        echo "$out" | tail -5 | sed 's/^/    /'
-    else
-        printf "  sender:   %s\n" "$sender"
-        printf "  receiver: %s\n" "$receiver"
-    fi
-
-    pkill -f "forward-" 2>/dev/null || true
-    ip netns exec "$NS" pkill iperf3 2>/dev/null || true
-    ip route del "${NS_IP}/32" dev "$TUN_NAME" 2>/dev/null || true
-    ip tuntap del dev "$TUN_NAME" mode tun 2>/dev/null || true
-    sleep 0.8
-}
-
-# ── direct baseline ───────────────────────────────────────────────────────────
-echo ""
-echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
-echo " BASELINE: direct veth (no TUN, no proxy)"
-echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
-ip netns exec "$NS" pkill iperf3 2>/dev/null || true; sleep 0.3
-ip netns exec "$NS" iperf3 -s -p "$IPERF_PORT" -D \
-    --logfile /tmp/iperf3-bench-server.log; sleep 0.3
-echo "  running iperf3: ${STREAMS} streams × ${DURATION}s …"
-baseline_out=$(iperf3 -c "$NS_IP" -p "$IPERF_PORT" \
-                   -t "$DURATION" -P "$STREAMS" 2>&1)
-echo "$baseline_out" | grep "SUM.*sender"   | awk '{printf "  sender:   %s %s\n", $6, $7}'
-echo "$baseline_out" | grep "SUM.*receiver" | awk '{printf "  receiver: %s %s\n", $6, $7}'
-ip netns exec "$NS" pkill iperf3 2>/dev/null || true; sleep 0.5
-
-# ── tun2 ─────────────────────────────────────────────────────────────────────
-echo ""
-echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
-echo " tun2  (main branch — no GRO/GSO offload)"
-echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
-run_bench "tun2" /tmp/forward-tun2
-
-# ── tun-rs + offload ──────────────────────────────────────────────────────────
-echo ""
-echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
-echo " tun-rs (patched — GRO/GSO offload via IFF_VNET_HDR)"
-echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
-run_bench "tun-rs+offload" /tmp/forward-tun-rs
-
-echo ""
-echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
-echo " done."
-echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"

netstack-smoltcp/scripts/route-linux.sh

@@ -1,26 +0,0 @@
-#!/bin/bash
-#__author__: cavivie
-
-DEFAULT_TUN_NAME="utun8"
-
-function do_route() {
-    local route_op="${1}"
-    local tun_name="${2:-$DEFAULT_TUN_NAME}"
-    ip route ${route_op} 0.0.0.0/1 dev ${tun_name}
-    ip route ${route_op} 128.0.0.0/1 dev ${tun_name}
-}
-
-function usage(){
-    echo "Usage:
-    route add    add tun routes to system route table
-    route del    delete routes from system route table
-    route help   display all usages of the shell script"
-}
-
-# START MAIN-OPTIONS
-case $1 in
-    add) do_route add $2;;
-    del) do_route delete $2;;
-    *) usage ;;
-esac
-# END MAIN-OPTIONS

netstack-smoltcp/scripts/route-macos.sh

@@ -1,36 +0,0 @@
-#!/bin/bash
-#__author__: cavivie
-
-DEFAULT_TUN_ADDR="10.10.10.2/24"
-DEFAULT_TUN_DEST="10.10.10.1"
-
-function do_route() {
-    local route_op="${1}"
-    local tun_addr="${2:-$DEFAULT_TUN_ADDR}"
-    local tun_dest="${3:-$DEFAULT_TUN_DEST}"
-    sudo route ${route_op} -net 1.0.0.0/8 ${tun_dest}
-    sudo route ${route_op} -net 2.0.0.0/7 ${tun_dest}
-    sudo route ${route_op} -net 4.0.0.0/6 ${tun_dest}
-    sudo route ${route_op} -net 8.0.0.0/5 ${tun_dest}
-    sudo route ${route_op} -net 16.0.0.0/4 ${tun_dest}
-    sudo route ${route_op} -net 32.0.0.0/3 ${tun_dest}
-    sudo route ${route_op} -net 64.0.0.0/2 ${tun_dest}
-    sudo route ${route_op} -net 128.0.0.0/1 ${tun_dest}
-    # tun2 do like this automatically
-    sudo route ${route_op} -net ${tun_addr} ${tun_dest}
-}
-
-function usage(){
-    echo "Usage:
-    route add    add tun routes to system route table
-    route del    delete routes from system route table
-    route help   display all usages of the shell script"
-}
-
-# START MAIN-OPTIONS
-case $1 in
-    add) do_route add $2 $3;;
-    del) do_route delete $2 $3;;
-    *) usage ;;
-esac
-# END MAIN-OPTIONS

netstack-smoltcp/scripts/route-windows.ps1

@@ -1,30 +0,0 @@
-#__author__: cavivie
-
-param(
-    [string]$Cmd = "help",
-    [string]$TunName = "utun8",
-    [string]$TunGateway = "10.10.10.1"
-)
-
-$ErrorActionPreference = "Stop"
-
-# START MAIN-OPTIONS
-switch ($Cmd) {
-    "add" {
-        # tun2 do like this automatically
-        New-NetRoute -DestinationPrefix "0.0.0.0/1" -InterfaceAlias $TunName -NextHop "$TunGateway"
-        New-NetRoute -DestinationPrefix "128.0.0.0/1" -InterfaceAlias $TunName -NextHop "$TunGateway"
-    }
-    "del" {
-        # tun2 do like this automatically
-        Get-NetRoute -DestinationPrefix "0.0.0.0/1" -InterfaceAlias $TunName | Remove-NetRoute
-        Get-NetRoute -DestinationPrefix "128.0.0.0/1" -InterfaceAlias $TunName | Remove-NetRoute
-    }
-    default {
-        Write-Host "Usage:
-    route add    add tun routes to system route table
-    route del    delete routes from system route table
-    route help   display all usages of the shell script"
-    }
-}
-# END MAIN-OPTIONS

netstack-smoltcp/src/device.rs

@@ -1,109 +0,0 @@
-use std::sync::{
-    atomic::{AtomicBool, Ordering},
-    Arc,
-};
-
-use smoltcp::{
-    phy::{Device, DeviceCapabilities, Medium, RxToken, TxToken},
-    time::Instant,
-};
-use tokio::sync::mpsc::{unbounded_channel, Permit, Sender, UnboundedReceiver, UnboundedSender};
-
-use crate::packet::AnyIpPktFrame;
-
-pub(super) struct VirtualDevice {
-    in_buf_avail: Arc<AtomicBool>,
-    in_buf: UnboundedReceiver<Vec<u8>>,
-    out_buf: Sender<AnyIpPktFrame>,
-    mtu: usize,
-    cached_packet: Option<Vec<u8>>,
-}
-
-impl VirtualDevice {
-    pub(super) fn new(
-        iface_egress_tx: Sender<AnyIpPktFrame>,
-        mtu: usize,
-    ) -> (Self, UnboundedSender<Vec<u8>>, Arc<AtomicBool>) {
-        let iface_ingress_tx_avail = Arc::new(AtomicBool::new(false));
-        let (iface_ingress_tx, iface_ingress_rx) = unbounded_channel();
-        (
-            Self {
-                in_buf_avail: iface_ingress_tx_avail.clone(),
-                in_buf: iface_ingress_rx,
-                out_buf: iface_egress_tx,
-                mtu,
-                cached_packet: None,
-            },
-            iface_ingress_tx,
-            iface_ingress_tx_avail,
-        )
-    }
-}
-
-impl Device for VirtualDevice {
-    type RxToken<'a> = VirtualRxToken;
-    type TxToken<'a> = VirtualTxToken<'a>;
-
-    fn receive(&mut self, _timestamp: Instant) -> Option<(Self::RxToken<'_>, Self::TxToken<'_>)> {
-        let buffer = if let Some(buf) = self.cached_packet.take() {
-            buf
-        } else {
-            let Ok(buf) = self.in_buf.try_recv() else {
-                self.in_buf_avail.store(false, Ordering::Release);
-                return None;
-            };
-            buf
-        };
-
-        let Ok(permit) = self.out_buf.try_reserve() else {
-            self.cached_packet = Some(buffer);
-            self.in_buf_avail.store(false, Ordering::Release);
-            return None;
-        };
-
-        Some((Self::RxToken { buffer }, Self::TxToken { permit }))
-    }
-
-    fn transmit(&mut self, _timestamp: Instant) -> Option<Self::TxToken<'_>> {
-        match self.out_buf.try_reserve() {
-            Ok(permit) => Some(Self::TxToken { permit }),
-            Err(_) => None,
-        }
-    }
-
-    fn capabilities(&self) -> DeviceCapabilities {
-        let mut capabilities = DeviceCapabilities::default();
-        capabilities.medium = Medium::Ip;
-        capabilities.max_transmission_unit = self.mtu;
-        capabilities
-    }
-}
-
-pub(super) struct VirtualRxToken {
-    buffer: Vec<u8>,
-}
-
-impl RxToken for VirtualRxToken {
-    fn consume<R, F>(self, f: F) -> R
-    where
-        F: FnOnce(&[u8]) -> R,
-    {
-        f(&self.buffer[..])
-    }
-}
-
-pub(super) struct VirtualTxToken<'a> {
-    permit: Permit<'a, Vec<u8>>,
-}
-
-impl<'a> TxToken for VirtualTxToken<'a> {
-    fn consume<R, F>(self, len: usize, f: F) -> R
-    where
-        F: FnOnce(&mut [u8]) -> R,
-    {
-        let mut buffer = vec![0u8; len];
-        let result = f(&mut buffer);
-        self.permit.send(buffer);
-        result
-    }
-}

netstack-smoltcp/src/filter.rs

@@ -1,56 +0,0 @@
-use std::net::IpAddr;
-
-pub type IpFilter<'a> = Box<dyn Fn(&IpAddr, &IpAddr) -> bool + Send + Sync + 'a>;
-
-pub struct IpFilters<'a> {
-    filters: Vec<IpFilter<'a>>,
-}
-
-impl<'a> Default for IpFilters<'a> {
-    fn default() -> Self {
-        Self::new()
-    }
-}
-
-impl<'a> IpFilters<'a> {
-    pub fn new() -> Self {
-        Self {
-            filters: Default::default(),
-        }
-    }
-
-    pub fn with_non_broadcast() -> Self {
-        macro_rules! non_broadcast {
-            ($addr:ident) => {
-                match $addr {
-                    IpAddr::V4(a) => !(a.is_broadcast() || a.is_multicast() || a.is_unspecified()),
-                    IpAddr::V6(a) => !(a.is_multicast() || a.is_unspecified()),
-                }
-            };
-        }
-        Self {
-            filters: vec![Box::new(|src, dst| {
-                non_broadcast!(src) && non_broadcast!(dst)
-            })],
-        }
-    }
-
-    pub fn add(&mut self, filter: IpFilter<'a>) {
-        self.filters.push(filter);
-    }
-
-    pub fn add_fn<F>(&mut self, filter: F)
-    where
-        F: Fn(&IpAddr, &IpAddr) -> bool + Send + Sync + 'a,
-    {
-        self.filters.push(Box::new(filter));
-    }
-
-    pub fn add_all<I: IntoIterator<Item = IpFilter<'a>>>(&mut self, filters: I) {
-        self.filters.extend(filters);
-    }
-
-    pub fn is_allowed(&self, src: &IpAddr, dst: &IpAddr) -> bool {
-        self.filters.iter().all(|filter| filter(src, dst))
-    }
-}

netstack-smoltcp/src/lib.rs

@@ -1,22 +0,0 @@
-mod device;
-
-mod runner;
-pub use runner::Runner;
-
-mod packet;
-pub use packet::AnyIpPktFrame;
-
-mod filter;
-pub use filter::{IpFilter, IpFilters};
-
-pub mod udp;
-pub use udp::UdpSocket;
-
-pub mod tcp;
-pub use tcp::{TcpListener, TcpStream};
-
-pub mod stack;
-pub use stack::{Stack, StackBuilder};
-
-/// Re-export
-pub use smoltcp;

netstack-smoltcp/src/packet.rs

@@ -1,53 +0,0 @@
-use std::net::IpAddr;
-
-use smoltcp::wire::{IpProtocol, IpVersion, Ipv4Packet, Ipv6Packet};
-
-pub type AnyIpPktFrame = Vec<u8>;
-
-#[derive(Debug)]
-pub(super) enum IpPacket<T: AsRef<[u8]>> {
-    Ipv4(Ipv4Packet<T>),
-    Ipv6(Ipv6Packet<T>),
-}
-
-impl<T: AsRef<[u8]> + Copy> IpPacket<T> {
-    pub fn new_checked(packet: T) -> smoltcp::wire::Result<IpPacket<T>> {
-        let buffer = packet.as_ref();
-        match IpVersion::of_packet(buffer)? {
-            IpVersion::Ipv4 => Ok(IpPacket::Ipv4(Ipv4Packet::new_checked(packet)?)),
-            IpVersion::Ipv6 => Ok(IpPacket::Ipv6(Ipv6Packet::new_checked(packet)?)),
-        }
-    }
-
-    pub fn src_addr(&self) -> IpAddr {
-        match *self {
-            IpPacket::Ipv4(ref packet) => IpAddr::from(packet.src_addr()),
-            IpPacket::Ipv6(ref packet) => IpAddr::from(packet.src_addr()),
-        }
-    }
-
-    pub fn dst_addr(&self) -> IpAddr {
-        match *self {
-            IpPacket::Ipv4(ref packet) => IpAddr::from(packet.dst_addr()),
-            IpPacket::Ipv6(ref packet) => IpAddr::from(packet.dst_addr()),
-        }
-    }
-
-    pub fn protocol(&self) -> IpProtocol {
-        match *self {
-            IpPacket::Ipv4(ref packet) => packet.next_header(),
-            IpPacket::Ipv6(ref packet) => packet.next_header(),
-        }
-    }
-}
-
-impl<'a, T: AsRef<[u8]> + ?Sized> IpPacket<&'a T> {
-    /// Return a pointer to the payload.
-    #[inline]
-    pub fn payload(&self) -> &'a [u8] {
-        match *self {
-            IpPacket::Ipv4(ref packet) => packet.payload(),
-            IpPacket::Ipv6(ref packet) => packet.payload(),
-        }
-    }
-}

netstack-smoltcp/src/runner.rs

@@ -1,42 +0,0 @@
-use std::{
-    future::{Future, IntoFuture},
-    pin::Pin,
-    task::{Context, Poll},
-};
-
-/// BoxFuture acts the same as the [BoxFuture in crate futures utils],
-/// which is an owned dynamically typed Future for use in cases where you
-/// can’t statically type your result or need to add some indirection.
-/// But the difference of this structure is that it will conditionally
-/// implement Send according to the properties of type T, which does not
-/// require two sets of API interfaces in single-threaded and multi-threaded.
-///
-/// [BoxFuture in crate futures utils]: https://docs.rs/futures-util/latest/futures_util/future/type.BoxFuture.html
-pub struct BoxFuture<'a, T>(Pin<Box<dyn Future<Output = T> + Send + 'a>>);
-
-impl<'a, T> BoxFuture<'a, T> {
-    pub fn new<F>(f: F) -> BoxFuture<'a, T>
-    where
-        F: IntoFuture<Output = T> + Send + 'a,
-        F::IntoFuture: Send + 'a,
-    {
-        BoxFuture(Box::pin(f.into_future()))
-    }
-
-    #[allow(unused)]
-    pub fn wrap(f: Pin<Box<dyn Future<Output = T> + Send + 'a>>) -> BoxFuture<'a, T> {
-        BoxFuture(f)
-    }
-}
-
-
-
-impl<T> Future for BoxFuture<'_, T> {
-    type Output = T;
-
-    fn poll(mut self: Pin<&mut Self>, context: &mut Context<'_>) -> Poll<Self::Output> {
-        self.0.as_mut().poll(context)
-    }
-}
-
-pub type Runner = BoxFuture<'static, std::io::Result<()>>;

netstack-smoltcp/src/stack.rs

@@ -1,279 +0,0 @@
-use std::{
-    net::IpAddr,
-    pin::Pin,
-    task::{ready, Context, Poll},
-};
-
-use futures::{Sink, Stream};
-use smoltcp::wire::IpProtocol;
-use tokio::sync::mpsc::{channel, Receiver};
-use tokio_util::sync::PollSender;
-use tracing::{debug, trace};
-
-use crate::{
-    filter::{IpFilter, IpFilters},
-    packet::{AnyIpPktFrame, IpPacket},
-    runner::Runner,
-    tcp::TcpListener,
-    udp::UdpSocket,
-};
-
-pub struct StackBuilder {
-    enable_udp: bool,
-    enable_tcp: bool,
-    enable_icmp: bool,
-    stack_buffer_size: usize,
-    udp_buffer_size: usize,
-    tcp_buffer_size: usize,
-    mtu: usize,
-    ip_filters: IpFilters<'static>,
-}
-
-impl Default for StackBuilder {
-    fn default() -> Self {
-        Self {
-            enable_udp: false,
-            enable_tcp: false,
-            enable_icmp: false,
-            stack_buffer_size: 1024,
-            udp_buffer_size: 512,
-            tcp_buffer_size: 512,
-            mtu: 1504, // 1500 for Ethernet + 4 for VLAN
-            ip_filters: IpFilters::with_non_broadcast(),
-        }
-    }
-}
-
-#[allow(unused)]
-impl StackBuilder {
-    pub fn enable_udp(mut self, enable: bool) -> Self {
-        self.enable_udp = enable;
-        self
-    }
-
-    pub fn enable_tcp(mut self, enable: bool) -> Self {
-        self.enable_tcp = enable;
-        self
-    }
-
-    pub fn enable_icmp(mut self, enable: bool) -> Self {
-        self.enable_icmp = enable;
-        self
-    }
-
-    pub fn stack_buffer_size(mut self, size: usize) -> Self {
-        self.stack_buffer_size = size;
-        self
-    }
-
-    pub fn udp_buffer_size(mut self, size: usize) -> Self {
-        self.udp_buffer_size = size;
-        self
-    }
-
-    pub fn tcp_buffer_size(mut self, size: usize) -> Self {
-        self.tcp_buffer_size = size;
-        self
-    }
-
-    pub fn set_ip_filters(mut self, filters: IpFilters<'static>) -> Self {
-        self.ip_filters = filters;
-        self
-    }
-
-    pub fn add_ip_filter(mut self, filter: IpFilter<'static>) -> Self {
-        self.ip_filters.add(filter);
-        self
-    }
-
-    pub fn add_ip_filter_fn<F>(mut self, filter: F) -> Self
-    where
-        F: Fn(&IpAddr, &IpAddr) -> bool + Send + Sync + 'static,
-    {
-        self.ip_filters.add_fn(filter);
-        self
-    }
-
-    pub fn mtu(mut self, mtu: usize) -> Self {
-        self.mtu = mtu;
-        self
-    }
-
-    #[allow(clippy::type_complexity)]
-    pub fn build(
-        self,
-    ) -> std::io::Result<(
-        Stack,
-        Option<Runner>,
-        Option<UdpSocket>,
-        Option<TcpListener>,
-    )> {
-        let (stack_tx, stack_rx) = channel(self.stack_buffer_size);
-
-        let (udp_tx, udp_rx) = if self.enable_udp {
-            let (udp_tx, udp_rx) = channel(self.udp_buffer_size);
-            (Some(PollSender::new(udp_tx)), Some(udp_rx))
-        } else {
-            (None, None)
-        };
-
-        let (tcp_tx, tcp_rx) = if self.enable_tcp {
-            let (tcp_tx, tcp_rx) = channel(self.tcp_buffer_size);
-            (Some(PollSender::new(tcp_tx)), Some(tcp_rx))
-        } else {
-            (None, None)
-        };
-
-        // ICMP is handled by TCP's Interface.
-        // smoltcp's interface will always send replies to EchoRequest
-        if self.enable_icmp && !self.enable_tcp {
-            use std::io::{Error, ErrorKind::InvalidInput};
-            return Err(Error::new(InvalidInput, "ICMP requires TCP"));
-        }
-        let icmp_tx = if self.enable_icmp {
-            tcp_tx.clone()
-        } else {
-            None
-        };
-
-        let udp_socket = udp_rx.map(|udp_rx| UdpSocket::new(udp_rx, stack_tx.clone()));
-
-        let (tcp_runner, tcp_listener) = if let Some(tcp_rx) = tcp_rx {
-            let (tcp_runner, tcp_listener) = TcpListener::new(tcp_rx, stack_tx, self.mtu)?;
-            (Some(tcp_runner), Some(tcp_listener))
-        } else {
-            (None, None)
-        };
-
-        let stack = Stack {
-            ip_filters: self.ip_filters,
-            stack_rx,
-            sink_buf: None,
-            udp_tx,
-            tcp_tx,
-            icmp_tx,
-        };
-
-        Ok((stack, tcp_runner, udp_socket, tcp_listener))
-    }
-}
-
-pub struct Stack {
-    ip_filters: IpFilters<'static>,
-    sink_buf: Option<(AnyIpPktFrame, IpProtocol)>,
-    udp_tx: Option<PollSender<AnyIpPktFrame>>,
-    tcp_tx: Option<PollSender<AnyIpPktFrame>>,
-    icmp_tx: Option<PollSender<AnyIpPktFrame>>,
-    stack_rx: Receiver<AnyIpPktFrame>,
-}
-
-impl Stack {
-    fn poll_send(&mut self, cx: &mut Context<'_>) -> Poll<Result<(), std::io::Error>> {
-        let (item, proto) = match self.sink_buf.take() {
-            Some(val) => val,
-            None => return Poll::Ready(Ok(())),
-        };
-
-        let tx = match proto {
-            IpProtocol::Tcp => self.tcp_tx.as_mut(),
-            IpProtocol::Udp => self.udp_tx.as_mut(),
-            IpProtocol::Icmp | IpProtocol::Icmpv6 => self.icmp_tx.as_mut(),
-            _ => unreachable!(),
-        };
-
-        let Some(tx) = tx else {
-            return Poll::Ready(Ok(()));
-        };
-
-        match tx.poll_reserve(cx) {
-            Poll::Pending => {
-                self.sink_buf = Some((item, proto));
-                Poll::Pending
-            }
-            Poll::Ready(Err(_)) => Poll::Ready(Err(channel_closed_err("channel is closed"))),
-            Poll::Ready(Ok(_)) => match tx.send_item(item) {
-                Ok(()) => Poll::Ready(Ok(())),
-                Err(_) => Poll::Ready(Err(channel_closed_err("channel is closed"))),
-            },
-        }
-    }
-}
-
-// Recv from stack.
-impl Stream for Stack {
-    type Item = std::io::Result<AnyIpPktFrame>;
-
-    fn poll_next(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Option<Self::Item>> {
-        match self.stack_rx.poll_recv(cx) {
-            Poll::Ready(Some(pkt)) => Poll::Ready(Some(Ok(pkt))),
-            Poll::Ready(None) => Poll::Ready(None),
-            Poll::Pending => Poll::Pending,
-        }
-    }
-}
-
-// Send to stack.
-impl Sink<AnyIpPktFrame> for Stack {
-    type Error = std::io::Error;
-
-    fn poll_ready(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Result<(), Self::Error>> {
-        // If a buffered item exists, try to flush it first. This also properly
-        // registers the waker via poll_reserve so we get woken when the channel
-        // has capacity. Without this, returning Pending here with _cx unused
-        // means the task never gets rescheduled.
-        if self.sink_buf.is_some() {
-            ready!(self.poll_send(cx))?;
-        }
-        Poll::Ready(Ok(()))
-    }
-
-    fn start_send(mut self: Pin<&mut Self>, item: AnyIpPktFrame) -> Result<(), Self::Error> {
-        if item.is_empty() {
-            return Ok(());
-        }
-
-        use std::io::{Error, ErrorKind::InvalidInput};
-        let packet = IpPacket::new_checked(item.as_slice())
-            .map_err(|err| Error::new(InvalidInput, format!("invalid IP packet: {err}")))?;
-
-        let src_ip = packet.src_addr();
-        let dst_ip = packet.dst_addr();
-
-        let addr_allowed = self.ip_filters.is_allowed(&src_ip, &dst_ip);
-        if !addr_allowed {
-            trace!("IP packet {src_ip} -> {dst_ip} (allowed? {addr_allowed}) throwing away",);
-            return Ok(());
-        }
-
-        let protocol = packet.protocol();
-        if matches!(
-            protocol,
-            IpProtocol::Tcp | IpProtocol::Udp | IpProtocol::Icmp | IpProtocol::Icmpv6
-        ) {
-            self.sink_buf.replace((item, protocol));
-        } else {
-            debug!("tun IP packet ignored (protocol: {:?})", protocol);
-        }
-
-        Ok(())
-    }
-
-    fn poll_flush(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Result<(), Self::Error>> {
-        self.poll_send(cx)
-    }
-
-    fn poll_close(
-        mut self: Pin<&mut Self>,
-        _cx: &mut Context<'_>,
-    ) -> Poll<Result<(), Self::Error>> {
-        self.stack_rx.close();
-        Poll::Ready(Ok(()))
-    }
-}
-
-fn channel_closed_err<E>(err: E) -> std::io::Error
-where
-    E: Into<Box<dyn std::error::Error + Send + Sync>>,
-{
-    std::io::Error::new(std::io::ErrorKind::BrokenPipe, err)
-}

netstack-smoltcp/src/tcp.rs

@@ -1,561 +0,0 @@
-use std::{
-    collections::HashMap,
-    net::SocketAddr,
-    pin::Pin,
-    sync::{
-        atomic::{AtomicBool, Ordering},
-        Arc,
-    },
-    task::{Context, Poll, Waker},
-};
-
-use futures::Stream;
-use smoltcp::{
-    iface::{Config as InterfaceConfig, Interface, SocketHandle, SocketSet},
-    phy::Device,
-    socket::tcp::{Socket as TcpSocket, SocketBuffer as TcpSocketBuffer, State as TcpState},
-    storage::RingBuffer,
-    time::{Duration, Instant},
-    wire::{HardwareAddress, IpAddress, IpCidr, IpProtocol, Ipv4Address, Ipv6Address, TcpPacket},
-};
-use spin::Mutex as SpinMutex;
-use tokio::{
-    io::{AsyncRead, AsyncWrite, ReadBuf},
-    sync::{
-        mpsc::{unbounded_channel, Receiver, Sender, UnboundedReceiver, UnboundedSender},
-        Notify,
-    },
-};
-use tracing::{error, trace};
-
-use crate::{
-    device::VirtualDevice,
-    packet::{AnyIpPktFrame, IpPacket},
-    Runner,
-};
-
-// NOTE: Default buffer could contain 20 AEAD packets
-const DEFAULT_TCP_SEND_BUFFER_SIZE: u32 = 0x3FFF * 20;
-const DEFAULT_TCP_RECV_BUFFER_SIZE: u32 = 0x3FFF * 20;
-
-#[derive(Debug, Clone, Copy, Eq, PartialEq)]
-enum TcpSocketState {
-    Normal,
-    Close,
-    Closing,
-    Closed,
-}
-
-struct TcpSocketControl {
-    send_buffer: RingBuffer<'static, u8>,
-    send_waker: Option<Waker>,
-    recv_buffer: RingBuffer<'static, u8>,
-    recv_waker: Option<Waker>,
-    recv_state: TcpSocketState,
-    send_state: TcpSocketState,
-}
-
-struct TcpSocketCreation {
-    control: SharedControl,
-    socket: TcpSocket<'static>,
-}
-
-type SharedNotify = Arc<Notify>;
-type SharedControl = Arc<SpinMutex<TcpSocketControl>>;
-
-struct TcpListenerRunner;
-
-impl TcpListenerRunner {
-    fn create(
-        device: VirtualDevice,
-        iface: Interface,
-        iface_ingress_tx: UnboundedSender<Vec<u8>>,
-        iface_ingress_tx_avail: Arc<AtomicBool>,
-        tcp_rx: Receiver<AnyIpPktFrame>,
-        stream_tx: UnboundedSender<TcpStream>,
-        sockets: HashMap<SocketHandle, SharedControl>,
-    ) -> Runner {
-        Runner::new(async move {
-            let notify = Arc::new(Notify::new());
-            let (socket_tx, socket_rx) = unbounded_channel::<TcpSocketCreation>();
-            let res = tokio::select! {
-                v = Self::handle_packet(notify.clone(), iface_ingress_tx, iface_ingress_tx_avail.clone(), tcp_rx, stream_tx, socket_tx) => v,
-                v = Self::handle_socket(notify, device, iface, iface_ingress_tx_avail, sockets, socket_rx) => v,
-            };
-            res?;
-            trace!("VirtDevice::poll thread exited");
-            Ok(())
-        })
-    }
-
-    async fn handle_packet(
-        notify: SharedNotify,
-        iface_ingress_tx: UnboundedSender<Vec<u8>>,
-        iface_ingress_tx_avail: Arc<AtomicBool>,
-        mut tcp_rx: Receiver<AnyIpPktFrame>,
-        stream_tx: UnboundedSender<TcpStream>,
-        socket_tx: UnboundedSender<TcpSocketCreation>,
-    ) -> std::io::Result<()> {
-        while let Some(frame) = tcp_rx.recv().await {
-            let packet = match IpPacket::new_checked(frame.as_slice()) {
-                Ok(p) => p,
-                Err(err) => {
-                    error!("invalid TCP IP packet: {:?}", err,);
-                    continue;
-                }
-            };
-
-            // Specially handle icmp packet by TCP interface.
-            if matches!(packet.protocol(), IpProtocol::Icmp | IpProtocol::Icmpv6) {
-                iface_ingress_tx
-                    .send(frame)
-                    .map_err(|e| std::io::Error::new(std::io::ErrorKind::BrokenPipe, e))?;
-                iface_ingress_tx_avail.store(true, Ordering::Release);
-                notify.notify_one();
-                continue;
-            }
-
-            let src_ip = packet.src_addr();
-            let dst_ip = packet.dst_addr();
-            let payload = packet.payload();
-
-            let packet = match TcpPacket::new_checked(payload) {
-                Ok(p) => p,
-                Err(err) => {
-                    error!("invalid TCP err: {err}, src_ip: {src_ip}, dst_ip: {dst_ip}, payload: {payload:?}");
-                    continue;
-                }
-            };
-            let src_port = packet.src_port();
-            let dst_port = packet.dst_port();
-
-            let src_addr = SocketAddr::new(src_ip, src_port);
-            let dst_addr = SocketAddr::new(dst_ip, dst_port);
-
-            // TCP first handshake packet, create a new Connection
-            if packet.syn() && !packet.ack() {
-                let mut socket = TcpSocket::new(
-                    TcpSocketBuffer::new(vec![0u8; DEFAULT_TCP_RECV_BUFFER_SIZE as usize]),
-                    TcpSocketBuffer::new(vec![0u8; DEFAULT_TCP_SEND_BUFFER_SIZE as usize]),
-                );
-                socket.set_keep_alive(Some(Duration::from_secs(28)));
-                // FIXME: It should follow system's setting. 7200 is Linux's default.
-                socket.set_timeout(Some(Duration::from_secs(7200)));
-                // NO ACK delay
-                // socket.set_ack_delay(None);
-
-                if let Err(err) = socket.listen(dst_addr) {
-                    error!("listen error: {:?}", err);
-                    continue;
-                }
-
-                trace!("created TCP connection for {} <-> {}", src_addr, dst_addr);
-
-                let control = Arc::new(SpinMutex::new(TcpSocketControl {
-                    send_buffer: RingBuffer::new(vec![0u8; DEFAULT_TCP_SEND_BUFFER_SIZE as usize]),
-                    send_waker: None,
-                    recv_buffer: RingBuffer::new(vec![0u8; DEFAULT_TCP_RECV_BUFFER_SIZE as usize]),
-                    recv_waker: None,
-                    recv_state: TcpSocketState::Normal,
-                    send_state: TcpSocketState::Normal,
-                }));
-
-                stream_tx
-                    .send(TcpStream {
-                        src_addr,
-                        dst_addr,
-                        notify: notify.clone(),
-                        control: control.clone(),
-                    })
-                    .map_err(|e| std::io::Error::new(std::io::ErrorKind::BrokenPipe, e))?;
-                socket_tx
-                    .send(TcpSocketCreation { control, socket })
-                    .map_err(|e| std::io::Error::new(std::io::ErrorKind::BrokenPipe, e))?;
-            }
-
-            // Pipeline tcp stream packet
-            iface_ingress_tx
-                .send(frame)
-                .map_err(|e| std::io::Error::new(std::io::ErrorKind::BrokenPipe, e))?;
-            iface_ingress_tx_avail.store(true, Ordering::Release);
-            notify.notify_one();
-        }
-        Ok(())
-    }
-
-    async fn handle_socket(
-        notify: SharedNotify,
-        mut device: VirtualDevice,
-        mut iface: Interface,
-        iface_ingress_tx_avail: Arc<AtomicBool>,
-        mut sockets: HashMap<SocketHandle, SharedControl>,
-        mut socket_rx: UnboundedReceiver<TcpSocketCreation>,
-    ) -> std::io::Result<()> {
-        let mut socket_set = SocketSet::new(vec![]);
-        loop {
-            while let Ok(TcpSocketCreation { control, socket }) = socket_rx.try_recv() {
-                let handle = socket_set.add(socket);
-                sockets.insert(handle, control);
-            }
-
-            let before_poll = Instant::now();
-            let updated_sockets = iface.poll(before_poll, &mut device, &mut socket_set);
-            if matches!(
-                updated_sockets,
-                smoltcp::iface::PollResult::SocketStateChanged
-            ) {
-                trace!("VirtDevice::poll costed {}", Instant::now() - before_poll);
-            }
-
-            // Check all the sockets' status
-            let mut sockets_to_remove = Vec::new();
-
-            for (socket_handle, control) in sockets.iter() {
-                let socket_handle = *socket_handle;
-                let socket = socket_set.get_mut::<TcpSocket>(socket_handle);
-                let mut control = control.lock();
-
-                // Remove the socket only when it is in the closed state.
-                if socket.state() == TcpState::Closed {
-                    sockets_to_remove.push(socket_handle);
-
-                    control.send_state = TcpSocketState::Closed;
-                    control.recv_state = TcpSocketState::Closed;
-
-                    if let Some(waker) = control.send_waker.take() {
-                        waker.wake();
-                    }
-                    if let Some(waker) = control.recv_waker.take() {
-                        waker.wake();
-                    }
-
-                    trace!("closed TCP connection");
-                    continue;
-                }
-
-                // SHUT_WR — only close once the send_buffer has been fully
-                // drained into the smoltcp socket.  Closing earlier transitions
-                // the socket to FIN_WAIT_1, making can_send() return false, so
-                // the send loop below never runs and the remaining data is lost.
-                if matches!(control.send_state, TcpSocketState::Close)
-                    && control.send_buffer.is_empty()
-                {
-                    trace!("closing TCP Write Half, {:?}", socket.state());
-
-                    socket.close();
-                    control.send_state = TcpSocketState::Closing;
-                }
-
-                // Check if readable
-                let mut wake_receiver = false;
-                while socket.can_recv() && !control.recv_buffer.is_full() {
-                    let result = socket.recv(|buffer| {
-                        let n = control.recv_buffer.enqueue_slice(buffer);
-                        (n, ())
-                    });
-
-                    match result {
-                        Ok(..) => wake_receiver = true,
-                        Err(err) => {
-                            error!("socket recv error: {:?}, {:?}", err, socket.state());
-
-                            // Don't know why. Abort the connection.
-                            socket.abort();
-
-                            if matches!(control.recv_state, TcpSocketState::Normal) {
-                                control.recv_state = TcpSocketState::Closed;
-                            }
-                            wake_receiver = true;
-
-                            // The socket will be recycled in the next poll.
-                            break;
-                        }
-                    }
-                }
-
-                // If socket is not in ESTABLISH, FIN-WAIT-1, FIN-WAIT-2,
-                // the local client have closed our receiver.
-                let states = [
-                    TcpState::Listen,
-                    TcpState::SynReceived,
-                    TcpState::Established,
-                    TcpState::FinWait1,
-                    TcpState::FinWait2,
-                ];
-                if matches!(control.recv_state, TcpSocketState::Normal)
-                    && !socket.may_recv()
-                    && !states.contains(&socket.state())
-                {
-                    trace!("closed TCP Read Half, {:?}", socket.state());
-
-                    // Let TcpStream::poll_read returns EOF.
-                    control.recv_state = TcpSocketState::Closed;
-                    wake_receiver = true;
-                }
-
-                if wake_receiver && control.recv_waker.is_some() {
-                    if let Some(waker) = control.recv_waker.take() {
-                        waker.wake();
-                    }
-                }
-
-                // Check if writable
-                let mut wake_sender = false;
-                while socket.can_send() && !control.send_buffer.is_empty() {
-                    let result = socket.send(|buffer| {
-                        let n = control.send_buffer.dequeue_slice(buffer);
-                        (n, ())
-                    });
-
-                    match result {
-                        Ok(..) => wake_sender = true,
-                        Err(err) => {
-                            error!("socket send error: {:?}, {:?}", err, socket.state());
-
-                            // Don't know why. Abort the connection.
-                            socket.abort();
-
-                            if matches!(control.send_state, TcpSocketState::Normal) {
-                                control.send_state = TcpSocketState::Closed;
-                            }
-                            wake_sender = true;
-
-                            // The socket will be recycled in the next poll.
-                            break;
-                        }
-                    }
-                }
-
-                if wake_sender && control.send_waker.is_some() {
-                    if let Some(waker) = control.send_waker.take() {
-                        waker.wake();
-                    }
-                }
-            }
-
-            for socket_handle in sockets_to_remove {
-                sockets.remove(&socket_handle);
-                socket_set.remove(socket_handle);
-            }
-
-            if !iface_ingress_tx_avail.load(Ordering::Acquire) {
-                let next_duration = iface
-                    .poll_delay(before_poll, &socket_set)
-                    .unwrap_or(Duration::from_millis(5));
-                if next_duration != Duration::ZERO {
-                    let _ = tokio::time::timeout(
-                        tokio::time::Duration::from(next_duration),
-                        notify.notified(),
-                    )
-                    .await;
-                }
-            }
-        }
-    }
-}
-
-pub struct TcpListener {
-    stream_rx: UnboundedReceiver<TcpStream>,
-}
-
-impl TcpListener {
-    pub(super) fn new(
-        tcp_rx: Receiver<AnyIpPktFrame>,
-        stack_tx: Sender<AnyIpPktFrame>,
-        mtu: usize,
-    ) -> std::io::Result<(Runner, Self)> {
-        let (mut device, iface_ingress_tx, iface_ingress_tx_avail) =
-            VirtualDevice::new(stack_tx, mtu);
-        let iface = Self::create_interface(&mut device)?;
-
-        let (stream_tx, stream_rx) = unbounded_channel();
-
-        let runner = TcpListenerRunner::create(
-            device,
-            iface,
-            iface_ingress_tx,
-            iface_ingress_tx_avail,
-            tcp_rx,
-            stream_tx,
-            HashMap::new(),
-        );
-
-        Ok((runner, Self { stream_rx }))
-    }
-
-    fn create_interface<D>(device: &mut D) -> std::io::Result<Interface>
-    where
-        D: Device + ?Sized,
-    {
-        let mut iface_config = InterfaceConfig::new(HardwareAddress::Ip);
-        iface_config.random_seed = rand::random();
-        let mut iface = Interface::new(iface_config, device, Instant::now());
-        iface.update_ip_addrs(|ip_addrs| {
-            ip_addrs
-                .push(IpCidr::new(IpAddress::v4(0, 0, 0, 1), 0))
-                .expect("iface IPv4");
-            ip_addrs
-                .push(IpCidr::new(IpAddress::v6(0, 0, 0, 0, 0, 0, 0, 1), 0))
-                .expect("iface IPv6");
-        });
-        iface
-            .routes_mut()
-            .add_default_ipv4_route(Ipv4Address::new(0, 0, 0, 1))
-            .map_err(|e| std::io::Error::new(std::io::ErrorKind::AddrNotAvailable, e))?;
-        iface
-            .routes_mut()
-            .add_default_ipv6_route(Ipv6Address::new(0, 0, 0, 0, 0, 0, 0, 1))
-            .map_err(|e| std::io::Error::new(std::io::ErrorKind::AddrNotAvailable, e))?;
-        iface.set_any_ip(true);
-        Ok(iface)
-    }
-}
-
-impl Stream for TcpListener {
-    type Item = (TcpStream, SocketAddr, SocketAddr);
-
-    fn poll_next(
-        mut self: std::pin::Pin<&mut Self>,
-        cx: &mut std::task::Context<'_>,
-    ) -> std::task::Poll<Option<Self::Item>> {
-        self.stream_rx.poll_recv(cx).map(|stream| {
-            stream.map(|stream| {
-                let local_addr = *stream.local_addr();
-                let remote_addr: SocketAddr = *stream.remote_addr();
-                (stream, local_addr, remote_addr)
-            })
-        })
-    }
-}
-
-pub struct TcpStream {
-    src_addr: SocketAddr,
-    dst_addr: SocketAddr,
-    notify: SharedNotify,
-    control: SharedControl,
-}
-
-impl Drop for TcpStream {
-    fn drop(&mut self) {
-        let mut control = self.control.lock();
-
-        if matches!(control.recv_state, TcpSocketState::Normal) {
-            control.recv_state = TcpSocketState::Close;
-        }
-
-        if matches!(control.send_state, TcpSocketState::Normal) {
-            control.send_state = TcpSocketState::Close;
-        }
-
-        self.notify.notify_one();
-    }
-}
-
-impl TcpStream {
-    pub fn local_addr(&self) -> &SocketAddr {
-        &self.src_addr
-    }
-
-    pub fn remote_addr(&self) -> &SocketAddr {
-        &self.dst_addr
-    }
-}
-
-impl AsyncRead for TcpStream {
-    fn poll_read(
-        self: Pin<&mut Self>,
-        cx: &mut Context<'_>,
-        buf: &mut ReadBuf<'_>,
-    ) -> Poll<std::io::Result<()>> {
-        let mut control = self.control.lock();
-
-        // Read from buffer
-        if control.recv_buffer.is_empty() {
-            // If socket is already closed / half closed, just return EOF directly.
-            if matches!(control.recv_state, TcpSocketState::Closed) {
-                return Ok(()).into();
-            }
-
-            // Nothing could be read. Wait for notify.
-            if let Some(old_waker) = control.recv_waker.replace(cx.waker().clone()) {
-                if !old_waker.will_wake(cx.waker()) {
-                    old_waker.wake();
-                }
-            }
-
-            return Poll::Pending;
-        }
-
-        let recv_buf = buf.initialize_unfilled();
-        let n = control.recv_buffer.dequeue_slice(recv_buf);
-        buf.advance(n);
-
-        if n > 0 {
-            self.notify.notify_one();
-        }
-
-        Ok(()).into()
-    }
-}
-
-impl AsyncWrite for TcpStream {
-    fn poll_write(
-        self: Pin<&mut Self>,
-        cx: &mut Context<'_>,
-        buf: &[u8],
-    ) -> Poll<std::io::Result<usize>> {
-        let mut control = self.control.lock();
-
-        // If state == Close | Closing | Closed, the TCP stream WR half is closed.
-        if !matches!(control.send_state, TcpSocketState::Normal) {
-            return Err(std::io::ErrorKind::BrokenPipe.into()).into();
-        }
-
-        // Write to buffer
-
-        if control.send_buffer.is_full() {
-            if let Some(old_waker) = control.send_waker.replace(cx.waker().clone()) {
-                if !old_waker.will_wake(cx.waker()) {
-                    old_waker.wake();
-                }
-            }
-
-            return Poll::Pending;
-        }
-
-        let n = control.send_buffer.enqueue_slice(buf);
-
-        if n > 0 {
-            self.notify.notify_one();
-        }
-
-        Ok(n).into()
-    }
-
-    fn poll_flush(self: Pin<&mut Self>, _cx: &mut Context<'_>) -> Poll<std::io::Result<()>> {
-        Ok(()).into()
-    }
-
-    fn poll_shutdown(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<std::io::Result<()>> {
-        let mut control = self.control.lock();
-
-        if matches!(control.send_state, TcpSocketState::Closed) {
-            return Ok(()).into();
-        }
-
-        // SHUT_WR
-        if matches!(control.send_state, TcpSocketState::Normal) {
-            control.send_state = TcpSocketState::Close;
-        }
-
-        if let Some(old_waker) = control.send_waker.replace(cx.waker().clone()) {
-            if !old_waker.will_wake(cx.waker()) {
-                old_waker.wake();
-            }
-        }
-
-        self.notify.notify_one();
-
-        Poll::Pending
-    }
-}

netstack-smoltcp/src/udp.rs

@@ -1,155 +0,0 @@
-use std::{
-    net::SocketAddr,
-    pin::Pin,
-    task::{Context, Poll},
-};
-
-use etherparse::PacketBuilder;
-use futures::{ready, Sink, SinkExt, Stream};
-use smoltcp::wire::UdpPacket;
-use tokio::sync::mpsc::{Receiver, Sender};
-use tokio_util::sync::PollSender;
-use tracing::{error, trace};
-
-use crate::packet::{AnyIpPktFrame, IpPacket};
-
-pub type UdpMsg = (
-    Vec<u8>,    /* payload */
-    SocketAddr, /* local */
-    SocketAddr, /* remote */
-);
-
-pub struct UdpSocket {
-    udp_rx: Receiver<AnyIpPktFrame>,
-    stack_tx: PollSender<AnyIpPktFrame>,
-}
-
-impl UdpSocket {
-    pub(super) fn new(udp_rx: Receiver<AnyIpPktFrame>, stack_tx: Sender<AnyIpPktFrame>) -> Self {
-        Self {
-            udp_rx,
-            stack_tx: PollSender::new(stack_tx),
-        }
-    }
-
-    pub fn split(self) -> (ReadHalf, WriteHalf) {
-        (
-            ReadHalf {
-                udp_rx: self.udp_rx,
-            },
-            WriteHalf {
-                stack_tx: self.stack_tx,
-            },
-        )
-    }
-}
-
-pub struct ReadHalf {
-    udp_rx: Receiver<AnyIpPktFrame>,
-}
-
-pub struct WriteHalf {
-    stack_tx: PollSender<AnyIpPktFrame>,
-}
-
-impl Stream for ReadHalf {
-    type Item = UdpMsg;
-
-    fn poll_next(mut self: Pin<&mut Self>, cx: &mut Context) -> Poll<Option<Self::Item>> {
-        loop {
-            match ready!(self.udp_rx.poll_recv(cx)) {
-                Some(frame) => {
-                    let packet = match IpPacket::new_checked(frame.as_slice()) {
-                        Ok(p) => p,
-                        Err(err) => {
-                            error!("invalid IP packet: {}", err);
-                            continue;
-                        }
-                    };
-
-                    let src_ip = packet.src_addr();
-                    let dst_ip = packet.dst_addr();
-                    let payload = packet.payload();
-
-                    let packet = match UdpPacket::new_checked(payload) {
-                        Ok(p) => p,
-                        Err(err) => {
-                            error!("invalid err: {err}, src_ip: {src_ip}, dst_ip: {dst_ip}, payload: {payload:?}");
-                            continue;
-                        }
-                    };
-                    let src_port = packet.src_port();
-                    let dst_port = packet.dst_port();
-
-                    let src_addr = SocketAddr::new(src_ip, src_port);
-                    let dst_addr = SocketAddr::new(dst_ip, dst_port);
-
-                    trace!("created UDP socket for {} <-> {}", src_addr, dst_addr);
-
-                    return Poll::Ready(Some((packet.payload().to_vec(), src_addr, dst_addr)));
-                }
-                None => return Poll::Ready(None),
-            }
-        }
-    }
-}
-
-impl Sink<UdpMsg> for WriteHalf {
-    type Error = std::io::Error;
-
-    fn poll_ready(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Result<(), Self::Error>> {
-        match ready!(self.stack_tx.poll_ready_unpin(cx)) {
-            Ok(()) => Poll::Ready(Ok(())),
-            Err(err) => Poll::Ready(Err(std::io::Error::other(err))),
-        }
-    }
-
-    fn start_send(mut self: Pin<&mut Self>, item: UdpMsg) -> Result<(), Self::Error> {
-        use std::io::{Error, ErrorKind::InvalidData};
-        let (data, src_addr, dst_addr) = item;
-
-        if data.is_empty() {
-            return Ok(());
-        }
-
-        let builder = match (src_addr, dst_addr) {
-            (SocketAddr::V4(src), SocketAddr::V4(dst)) => {
-                PacketBuilder::ipv4(src.ip().octets(), dst.ip().octets(), 20)
-                    .udp(src_addr.port(), dst_addr.port())
-            }
-            (SocketAddr::V6(src), SocketAddr::V6(dst)) => {
-                PacketBuilder::ipv6(src.ip().octets(), dst.ip().octets(), 20)
-                    .udp(src_addr.port(), dst_addr.port())
-            }
-            _ => {
-                return Err(Error::new(InvalidData, "src or destination type unmatch"));
-            }
-        };
-
-        let mut ip_packet_writer = Vec::with_capacity(builder.size(data.len()));
-        builder
-            .write(&mut ip_packet_writer, &data)
-            .map_err(|err| Error::other(format!("PacketBuilder::write: {err}")))?;
-
-        match self.stack_tx.start_send_unpin(ip_packet_writer) {
-            Ok(()) => Ok(()),
-            Err(err) => Err(Error::other(format!("send error: {err}"))),
-        }
-    }
-
-    fn poll_flush(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Result<(), Self::Error>> {
-        use std::io::Error;
-        match ready!(self.stack_tx.poll_flush_unpin(cx)) {
-            Ok(()) => Poll::Ready(Ok(())),
-            Err(err) => Poll::Ready(Err(Error::other(format!("flush error: {err}")))),
-        }
-    }
-
-    fn poll_close(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Result<(), Self::Error>> {
-        use std::io::Error;
-        match ready!(self.stack_tx.poll_close_unpin(cx)) {
-            Ok(()) => Poll::Ready(Ok(())),
-            Err(err) => Poll::Ready(Err(Error::other(format!("close error: {err}")))),
-        }
-    }
-}

netstack-smoltcp/tests/regression.rs

@@ -1,75 +0,0 @@
-//! Regression tests that reproduce the bugs found in the static analysis.
-
-use std::time::Duration;
-
-use etherparse::{IpNumber, Ipv4Header, UdpHeader};
-use futures::SinkExt;
-use tokio::time::timeout;
-
-use netstack_smoltcp::StackBuilder;
-
-fn make_udp_ipv4(
-    src_ip: [u8; 4],
-    src_port: u16,
-    dst_ip: [u8; 4],
-    dst_port: u16,
-    payload: &[u8],
-) -> Vec<u8> {
-    let udp_hdr = UdpHeader::with_ipv4_checksum(
-        src_port,
-        dst_port,
-        &Ipv4Header::new(
-            (UdpHeader::LEN + payload.len()) as u16,
-            64,
-            IpNumber::UDP,
-            src_ip,
-            dst_ip,
-        )
-        .unwrap(),
-        payload,
-    )
-    .unwrap();
-
-    let ip_hdr = Ipv4Header::new(
-        (UdpHeader::LEN + payload.len()) as u16,
-        64,
-        IpNumber::UDP,
-        src_ip,
-        dst_ip,
-    )
-    .unwrap();
-
-    let mut buf = Vec::with_capacity(Ipv4Header::MIN_LEN + UdpHeader::LEN + payload.len());
-    ip_hdr.write(&mut buf).unwrap();
-    udp_hdr.write(&mut buf).unwrap();
-    buf.extend_from_slice(payload);
-    buf
-}
-
-/// before(include) a15e0b72bfc72cb032e67138070da01e325d66f8
-/// sink_buf is used in `Stack` to hold a slot for sending any pkt
-///
-/// the original assumption is that the `poll_ready` -> `start_send` -> `poll_flush`
-/// are called sequentially so the slot could be reused and will never get blocked.
-///
-/// but once the user calls `send_all` on `Stack`, which will not immediate flush the pkt(call `poll_flush`),
-/// then `sink_buf` is could be Some(pkt), then it will trigger `Poll::Pending` branch in `Stack::poll_ready`,
-/// who did not register the waker correctly, so it will got hanged forever.
-#[tokio::test(flavor = "current_thread")]
-async fn bug1_poll_ready_waker_registered_via_send_all() {
-    let (mut stack, _runner, _udp_socket, _tcp) = StackBuilder::default()
-        .enable_udp(true)
-        .udp_buffer_size(64)
-        .stack_buffer_size(64)
-        .build()
-        .unwrap();
-
-    let pkt1 = make_udp_ipv4([1, 2, 3, 4], 1111, [5, 6, 7, 8], 9999, b"first");
-    let pkt2 = make_udp_ipv4([1, 2, 3, 4], 1111, [5, 6, 7, 8], 9999, b"second");
-
-    let mut stream = futures::stream::iter([Ok(pkt1), Ok(pkt2)]);
-
-    let result = timeout(Duration::from_secs(1), stack.send_all(&mut stream)).await;
-    // should be ok after the fix
-    assert!(result.is_ok());
-}

ostp-client/Cargo.toml

@@ -9,7 +9,10 @@ anyhow.workspace = true
 bytes.workspace = true
 tokio.workspace = true
 tracing.workspace = true
+tracing-subscriber = { version = "0.3", features = ["env-filter"] }
+tracing-appender = "0.2"
 ostp-core = { path = "../ostp-core" }
+ostp-tun = { path = "../ostp-tun" }
 rand.workspace = true
 serde = { version = "1.0", features = ["derive"] }
 serde_json = "1.0"
@@ -26,7 +29,4 @@ tun = { version = "0.8.9", features = ["async"] }
 netstack-smoltcp = "0.2.2"
 futures = "0.3.32"
 libc = "0.2.186"
-x25519-dalek = "2.0.1"
-chacha20poly1305.workspace = true
-hex = "0.4.3"
-winapi = { version = "0.3.9", features = ["iphlpapi", "tcpmib", "processthreadsapi", "psapi", "handleapi", "winerror", "minwindef", "winnt"] }
+winapi = { version = "0.3.9", features = ["iphlpapi", "tcpmib", "processthreadsapi", "psapi", "handleapi", "winerror", "minwindef", "winnt", "iptypes", "ws2def"] }

ostp-client/src/bin/test_route.rs

@@ -0,0 +1,21 @@
+fn main() {
+    let socket = std::net::UdpSocket::bind("0.0.0.0:0").unwrap();
+    let port = socket.local_addr().unwrap().port();
+    println!("Bound UDP to port {}", port);
+    
+    if let Some(name) = ostp_client::tunnel::process_lookup::get_process_name_from_port_udp(port) {
+        println!("Found process for UDP port {}: {}", port, name);
+    } else {
+        println!("Process not found for UDP port {}", port);
+    }
+
+    let tcp_socket = std::net::TcpListener::bind("0.0.0.0:0").unwrap();
+    let tcp_port = tcp_socket.local_addr().unwrap().port();
+    println!("Bound TCP to port {}", tcp_port);
+
+    if let Some(name) = ostp_client::tunnel::process_lookup::get_process_name_from_port(tcp_port) {
+        println!("Found process for TCP port {}: {}", tcp_port, name);
+    } else {
+        println!("Process not found for TCP port {}", tcp_port);
+    }
+}

ostp-client/src/config.rs

@@ -12,7 +12,6 @@ pub struct ClientConfig {
     pub debug: bool,
     pub ostp: OstpConfig,
     pub local_proxy: LocalProxyConfig,
-    pub reality: RealityConfig,
     #[serde(default)]
     pub transport: TransportConfig,
     #[serde(default)]
@@ -22,6 +21,10 @@ pub struct ClientConfig {
     pub dns_server: Option<String>,
     #[serde(default = "default_tun_stack")]
     pub tun_stack: String,
+    #[serde(default)]
+    pub kill_switch: bool,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub gui: Option<serde_json::Value>,
 }
 
 fn default_tun_stack() -> String { "system".to_string() }
@@ -67,53 +70,51 @@ pub struct LocalProxyConfig {
 }
 
 /// Transport layer configuration.
-/// `mode` = "udp" (default) or "uot" (UDP over TCP with xHTTP stealth).
+/// `mode` = "udp" (default) or "uot" (UDP over TCP, no protocol mimicry —
+/// zapret-like: no recognizable header at all, not a fake TLS/HTTP shell).
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct TransportConfig {
     /// "udp" or "uot"
     #[serde(default = "default_transport_mode")]
     pub mode: String,
-    /// TLS SNI and HTTP Host for stealth routing
-    #[serde(default)]
-    pub stealth_sni: String,
-    /// TCP Port for the stealth connection
-    #[serde(default = "default_stealth_port")]
-    pub stealth_port: u16,
-    /// Enable strict RFC 6455 WebSocket framing
-    #[serde(default)]
-    pub wss: bool,
+    /// Split the first UoT/TCP packet (handshake) into tiny TCP segments to
+    /// break DPI that inspects the first packet. UoT/TCP only; ignored for UDP.
+    pub tcp_fragmentation: bool,
+    /// TCP chunk size (bytes)
+    #[serde(default = "default_frag_chunk")]
+    pub frag_chunk: usize,
+    /// TCP sleep duration between chunks (ms)
+    #[serde(default = "default_frag_sleep")]
+    pub frag_sleep: u64,
+    /// [min, max] junk packet count
+    #[serde(default = "default_junk_count")]
+    pub junk_pc: [usize; 2],
+    /// [min, max] junk packet size in bytes
+    #[serde(default = "default_junk_size")]
+    pub junk_ps: [usize; 2],
 }
 
 fn default_transport_mode() -> String { "udp".to_string() }
-fn default_stealth_port() -> u16 { 443 }
+fn default_frag_chunk() -> usize { 2 }
+fn default_frag_sleep() -> u64 { 2 }
+fn default_junk_count() -> [usize; 2] { [2, 5] }
+fn default_junk_size() -> [usize; 2] { [100, 1000] }
 
 impl Default for TransportConfig {
     fn default() -> Self {
         Self {
             mode: default_transport_mode(),
-            stealth_sni: String::new(),
-            stealth_port: default_stealth_port(),
-            wss: false,
+            tcp_fragmentation: false,
+            frag_chunk: default_frag_chunk(),
+            frag_sleep: default_frag_sleep(),
+            junk_pc: default_junk_count(),
+            junk_ps: default_junk_size(),
         }
     }
 }
 
 
-#[derive(Debug, Clone, Serialize, Deserialize, Default)]
-pub struct RealityConfig {
-    #[serde(default)]
-    pub enabled: bool,
-    #[serde(default)]
-    pub sni: String,
-    #[serde(default)]
-    pub fp: String,
-    #[serde(default)]
-    pub pbk: String,
-    #[serde(default)]
-    pub sid: String,
-    #[serde(default)]
-    pub spx: String,
-}
+
 
 
 impl Default for OstpConfig {
@@ -147,12 +148,13 @@ impl Default for ClientConfig {
             debug: false,
             ostp: OstpConfig::default(),
             local_proxy: LocalProxyConfig::default(),
-            reality: RealityConfig::default(),
             transport: TransportConfig::default(),
             exclusions: ExclusionConfig::default(),
             multiplex: MultiplexConfig::default(),
             dns_server: None,
             tun_stack: "system".to_string(),
+            kill_switch: false,
+            gui: None,
         }
     }
 }
@@ -180,16 +182,18 @@ struct RawUnifiedConfig {
     tun: Option<RawTunSection>,
     exclude: Option<RawExcludeSection>,
     mux: Option<RawMuxSection>,
-    reality: Option<RawRealitySection>,
     transport: Option<RawTransportSection>,
+    gui: Option<serde_json::Value>,
 }
 
 #[derive(Debug, Deserialize)]
 struct RawTransportSection {
     mode: Option<String>,
-    stealth_sni: Option<String>,
-    stealth_port: Option<u16>,
-    wss: Option<bool>,
+    tcp_fragmentation: Option<bool>,
+    frag_chunk: Option<usize>,
+    frag_sleep: Option<u64>,
+    junk_pc: Option<[usize; 2]>,
+    junk_ps: Option<[usize; 2]>,
 }
 
 #[derive(Debug, Deserialize)]
@@ -197,6 +201,7 @@ struct RawTunSection {
     enable: Option<bool>,
     dns: Option<String>,
     stack: Option<String>,
+    kill_switch: Option<bool>,
 }
 
 #[derive(Debug, Deserialize)]
@@ -212,15 +217,7 @@ struct RawMuxSection {
     sessions: Option<usize>,
 }
 
-#[derive(Debug, Deserialize)]
-struct RawRealitySection {
-    enabled: Option<bool>,
-    sni: Option<String>,
-    fp: Option<String>,
-    pbk: Option<String>,
-    sid: Option<String>,
-    spx: Option<String>,
-}
+
 
 impl ClientConfig {
     /// Hot-reload from `config.json` placed next to the running binary.
@@ -267,19 +264,13 @@ impl ClientConfig {
                 bind_addr: socks5,
                 connect_timeout_ms: 15000,
             },
-            reality: RealityConfig {
-                enabled: raw.reality.as_ref().and_then(|t| t.enabled).unwrap_or(false),
-                sni: raw.reality.as_ref().and_then(|t| t.sni.clone()).unwrap_or_default(),
-                fp: raw.reality.as_ref().and_then(|t| t.fp.clone()).unwrap_or_default(),
-                pbk: raw.reality.as_ref().and_then(|t| t.pbk.clone()).unwrap_or_default(),
-                sid: raw.reality.as_ref().and_then(|t| t.sid.clone()).unwrap_or_default(),
-                spx: raw.reality.as_ref().and_then(|t| t.spx.clone()).unwrap_or_default(),
-            },
             transport: TransportConfig {
-                mode: raw.transport.as_ref().and_then(|t| t.mode.clone()).unwrap_or_else(|| "udp".to_string()),
-                stealth_sni: raw.transport.as_ref().and_then(|t| t.stealth_sni.clone()).unwrap_or_else(|| "microsoft.com".to_string()),
-                stealth_port: raw.transport.as_ref().and_then(|t| t.stealth_port).unwrap_or(443),
-                wss: raw.transport.as_ref().and_then(|t| t.wss).unwrap_or(false),
+                mode: raw.transport.as_ref().and_then(|t| t.mode.clone()).unwrap_or_else(default_transport_mode),
+                tcp_fragmentation: raw.transport.as_ref().and_then(|t| t.tcp_fragmentation).unwrap_or(false),
+                frag_chunk: raw.transport.as_ref().and_then(|t| t.frag_chunk).unwrap_or_else(default_frag_chunk),
+                frag_sleep: raw.transport.as_ref().and_then(|t| t.frag_sleep).unwrap_or_else(default_frag_sleep),
+                junk_pc: raw.transport.as_ref().and_then(|t| t.junk_pc).unwrap_or_else(default_junk_count),
+                junk_ps: raw.transport.as_ref().and_then(|t| t.junk_ps).unwrap_or_else(default_junk_size),
             },
             exclusions: ExclusionConfig {
                 domains: exclusions.domains.unwrap_or_default(),
@@ -292,7 +283,256 @@ impl ClientConfig {
             },
             dns_server: raw.tun.as_ref().and_then(|t| t.dns.clone()),
             tun_stack: raw.tun.as_ref().and_then(|t| t.stack.clone()).unwrap_or_else(|| "system".to_string()),
+            kill_switch: raw.tun.as_ref().and_then(|t| t.kill_switch).unwrap_or(false),
+            gui: raw.gui,
         })
-
     }
 }
+
+// ═══════════════════════════════════════════════════════════════════════
+// On-disk config.json shapes — client, server, and relay.
+//
+// This is the ONE place these are defined. They used to be declared locally
+// inside ostp/src/main.rs (the CLI binary) with no other consumer able to
+// see them, which is exactly how ostp-client::migrate ended up working
+// against loosely-typed serde_json::Value instead of a real schema, and how
+// the CLI, the migrator, and this crate's own hot-reload path could each
+// silently drift out of sync with what a config.json actually looks like.
+// main.rs now imports these instead of re-declaring them (see the `use
+// ostp_client::config::{...}` at its top).
+//
+// These are DELIBERATELY separate from ClientConfig/OstpConfig/etc. above:
+// this section is the friendly, minimal shape a user actually edits by
+// hand; the types above are what the running engine needs internally
+// (handshake/io timeouts, resolved addresses, ...) and are built FROM one
+// of these via the mapping in ostp/src/main.rs::run_client_directly. Only
+// `ClientConfig` collides by name with the runtime type above, so the
+// on-disk one is `ClientFileConfig` — everything else keeps its natural name.
+// ═══════════════════════════════════════════════════════════════════════
+
+#[derive(Debug, Deserialize, Serialize)]
+#[serde(tag = "mode", rename_all = "lowercase")]
+pub enum AppMode {
+    Server(ServerConfig),
+    Client(ClientFileConfig),
+    Relay(RelayServerConfig),
+}
+
+#[derive(Debug, Deserialize, Serialize)]
+pub struct UnifiedConfig {
+    #[serde(flatten)]
+    pub mode: AppMode,
+    pub log_level: Option<String>,
+}
+
+impl UnifiedConfig {
+    pub fn validate(&self) -> Result<()> {
+        match &self.mode {
+            AppMode::Server(cfg) => {
+                if cfg.access_keys.is_empty() {
+                    anyhow::bail!("Server configuration must contain at least one access_key.");
+                }
+                if let Some(outbound) = &cfg.outbound {
+                    if outbound.enabled {
+                        let action = outbound.default_action.as_deref().unwrap_or("direct");
+                        if action == "direct" && outbound.rules.is_empty() {
+                            println!("\n[WARNING] Server outbound proxy is ENABLED, but default_action is 'direct' and there are no rules!");
+                            println!("          This means ALL traffic will bypass the proxy and go out directly from the server IP.");
+                            println!("          If you want all traffic to be proxied, change 'default_action' to 'proxy'.\n");
+                        }
+                    }
+                }
+            }
+            AppMode::Client(cfg) => {
+                if cfg.access_key.is_empty() {
+                    anyhow::bail!("Client configuration must contain an access_key.");
+                }
+            }
+            AppMode::Relay(cfg) => {
+                if cfg.upstream_tcp.is_empty() {
+                    anyhow::bail!("Relay configuration must specify upstream_tcp address.");
+                }
+                if cfg.upstream_api_url.is_empty() {
+                    anyhow::bail!("Relay configuration must specify upstream_api_url.");
+                }
+            }
+        }
+        Ok(())
+    }
+}
+
+#[derive(Debug, Deserialize, Serialize, Clone)]
+#[serde(untagged)]
+pub enum UserConfig {
+    Detailed {
+        access_key: String,
+        name: Option<String>,
+        limit_bytes: Option<u64>,
+    },
+    KeyOnly(String),
+}
+
+impl UserConfig {
+    pub fn key(&self) -> String {
+        match self {
+            UserConfig::KeyOnly(k) => k.clone(),
+            UserConfig::Detailed { access_key, .. } => access_key.clone(),
+        }
+    }
+    pub fn name(&self) -> Option<String> {
+        match self {
+            UserConfig::KeyOnly(_) => None,
+            UserConfig::Detailed { name, .. } => name.clone(),
+        }
+    }
+    pub fn limit(&self) -> Option<u64> {
+        match self {
+            UserConfig::KeyOnly(_) => None,
+            UserConfig::Detailed { limit_bytes, .. } => *limit_bytes,
+        }
+    }
+}
+
+#[derive(Debug, Deserialize, Serialize)]
+pub struct ServerConfig {
+    pub listen: ListenConfig,
+    pub access_keys: Vec<UserConfig>,
+    pub debug: Option<bool>,
+    pub outbound: Option<OutboundConfig>,
+    pub api: Option<ApiConfig>,
+    pub fallback: Option<FallbackCfg>,
+    pub transport: Option<TransportConfigRaw>,
+    // Left untyped: ostp-client does not (and should not) depend on
+    // ostp-server just to name its DnsConfig type. The CLI binary — which
+    // already depends on both crates — deserializes this into
+    // ostp_server::dns::DnsConfig right before handing it to run_server().
+    pub dns: Option<serde_json::Value>,
+}
+
+/// Relay-node config.json shape.
+#[derive(Debug, Deserialize, Serialize)]
+pub struct RelayServerConfig {
+    /// Listen address(es) (UDP + TCP UoT)
+    pub listen: ListenConfig,
+    /// Upstream address for TCP (UoT) traffic
+    pub upstream_tcp: String,
+    /// Upstream address for UDP traffic
+    pub upstream_udp: String,
+    // ── Deprecated ──────────────────────────────────────────────────────────
+    // The relay used to authenticate clients itself and pulled the access-key
+    // list from the target server's management API to do it. It no longer does:
+    // sessions are authenticated end-to-end by the target server, and a relay
+    // that re-checks credentials only adds a weaker second gate plus a copy of
+    // the key list on a machine that does not need one. These are kept solely
+    // so existing relay configs still parse; they are ignored.
+    #[serde(default)]
+    pub upstream_api_url: String,
+    #[serde(default)]
+    pub upstream_api_token: String,
+    #[serde(default)]
+    pub sync_interval_secs: u64,
+    pub debug: Option<bool>,
+}
+
+/// Supports both a single string "0.0.0.0:50000" and an array
+/// ["0.0.0.0:50000", "[::]:50000"].
+#[derive(Debug, Deserialize, Serialize, Clone)]
+#[serde(untagged)]
+pub enum ListenConfig {
+    Single(String),
+    Multiple(Vec<String>),
+}
+
+impl ListenConfig {
+    pub fn addresses(&self) -> Vec<String> {
+        match self {
+            ListenConfig::Single(s) => vec![s.clone()],
+            ListenConfig::Multiple(v) => v.clone(),
+        }
+    }
+
+    pub fn primary(&self) -> String {
+        match self {
+            ListenConfig::Single(s) => s.clone(),
+            ListenConfig::Multiple(v) => v.first().cloned().unwrap_or_default(),
+        }
+    }
+}
+
+#[derive(Debug, Deserialize, Serialize)]
+pub struct ApiConfig {
+    pub enabled: Option<bool>,
+    pub bind: Option<String>,
+    pub token: Option<String>,
+    pub webpath: Option<String>,
+    pub username: Option<String>,
+    pub password_hash: Option<String>,
+}
+
+#[derive(Debug, Deserialize, Serialize)]
+pub struct FallbackCfg {
+    pub enabled: Option<bool>,
+    pub listen: Option<String>,
+    pub target: Option<String>,
+}
+
+#[derive(Debug, Deserialize, Serialize)]
+pub struct ClientFileConfig {
+    pub server: String,
+    pub access_key: String,
+    pub mtu: Option<usize>,
+    pub socks5_bind: Option<String>,
+    pub tun: Option<TunConfig>,
+    pub debug: Option<bool>,
+    pub exclude: Option<ExcludeConfig>,
+    pub mux: Option<MuxConfig>,
+    pub transport: Option<TransportConfigRaw>,
+    pub gui: Option<serde_json::Value>,
+}
+
+#[derive(Debug, Deserialize, Serialize, Clone)]
+pub struct TransportConfigRaw {
+    pub mode: Option<String>,
+    pub tcp_fragmentation: Option<bool>,
+}
+
+#[derive(Debug, Deserialize, Serialize, Clone)]
+pub struct TunConfig {
+    pub enable: bool,
+    pub wintun_path: Option<String>,
+    pub ipv4_address: Option<String>,
+    pub dns: Option<String>,
+    pub kill_switch: Option<bool>,
+}
+
+#[derive(Debug, Deserialize, Serialize)]
+pub struct OutboundConfig {
+    pub enabled: bool,
+    pub protocol: String,
+    pub address: String,
+    pub port: u16,
+    #[serde(default)]
+    pub rules: Vec<OutboundRule>,
+    pub default_action: Option<String>,
+}
+
+#[derive(Debug, Deserialize, Serialize)]
+pub struct OutboundRule {
+    pub domain_suffix: Option<Vec<String>>,
+    pub ip_cidr: Option<Vec<String>>,
+    pub protocol: Option<String>,
+    pub action: Option<String>,
+}
+
+#[derive(Debug, Deserialize, Serialize)]
+pub struct ExcludeConfig {
+    pub domains: Option<Vec<String>>,
+    pub ips: Option<Vec<String>>,
+    pub processes: Option<Vec<String>>,
+}
+
+#[derive(Debug, Deserialize, Serialize)]
+pub struct MuxConfig {
+    pub enabled: Option<bool>,
+    pub sessions: Option<usize>,
+}

ostp-client/src/lib.rs

@@ -1,6 +1,7 @@
 pub mod app;
 pub mod bridge;
 pub mod config;
+pub mod migrate;
 pub mod signal;
 pub mod sysproxy;
 pub mod transport;
@@ -8,3 +9,4 @@ pub mod tunnel;
 
 
 pub mod runner;
+pub mod logging;

ostp-client/src/logging.rs

@@ -0,0 +1,189 @@
+use std::fs::OpenOptions;
+use std::io::Write;
+use std::path::PathBuf;
+use tracing_subscriber::{layer::SubscriberExt, util::SubscriberInitExt, EnvFilter};
+
+/// The single canonical log file for the whole core. Every process (CLI daemon,
+/// GUI, TUN helper) and every subsystem (tracing, the core event logger, the
+/// helper IPC, panics) writes here — no more per-binary / per-subsystem sprawl
+/// (`ostp-cli.log` + `ostp-core.log` + `ostp-helper.log` + `ostp-crash.log`).
+pub const LOG_FILE_NAME: &str = "ostp.log";
+
+/// Absolute path to the shared log file, next to the running executable.
+pub fn log_file_path() -> PathBuf {
+    std::env::current_exe()
+        .ok()
+        .and_then(|p| p.parent().map(|d| d.join(LOG_FILE_NAME)))
+        .unwrap_or_else(|| PathBuf::from(LOG_FILE_NAME))
+}
+
+/// True if this invocation is the long-running daemon (a client/server run),
+/// as opposed to a one-shot subcommand (`gk`, `check`, `init`, `-V`, ...).
+///
+/// Used to gate log truncation: only the daemon clears the log at startup, so a
+/// one-shot command run while a daemon is live can never wipe the daemon's log.
+/// A daemon invocation is simply one that carries none of the one-shot tokens
+/// (`ostp`, `ostp run`, `ostp connect <url>` → daemon; everything else → one-shot).
+pub fn invocation_is_daemon<I: IntoIterator<Item = String>>(args: I) -> bool {
+    const ONE_SHOT: &[&str] = &[
+        "gk", "generate-key", "check", "init", "setup", "links", "import",
+        "update", "migrate", "prober", "proxy-env", "proxy-env-clear",
+        "uninstall", "-V", "--version", "-h", "--help", "help",
+    ];
+    !args
+        .into_iter()
+        .skip(1) // program name
+        .any(|a| ONE_SHOT.contains(&a.as_str()))
+}
+
+/// Append a single timestamped line to the shared log file. Used by the manual
+/// writers (core event logger, TUN helper IPC) so their output lands in the same
+/// `ostp.log` as the tracing subscriber instead of a separate file.
+pub fn append_line(msg: &str) {
+    if let Ok(mut file) = OpenOptions::new().create(true).append(true).open(log_file_path()) {
+        let _ = writeln!(
+            file,
+            "[{}] {}",
+            chrono::Local::now().format("%Y-%m-%d %H:%M:%S"),
+            msg
+        );
+    }
+}
+
+pub fn setup_panic_hook() {
+    std::panic::set_hook(Box::new(|info| {
+        let payload = info.payload();
+        let msg = if let Some(s) = payload.downcast_ref::<&str>() {
+            *s
+        } else if let Some(s) = payload.downcast_ref::<String>() {
+            s.as_str()
+        } else {
+            "Box<dyn Any>"
+        };
+
+        let location = info.location().unwrap_or_else(|| std::panic::Location::caller());
+        let backtrace = std::backtrace::Backtrace::force_capture();
+
+        let crash_msg = format!(
+            "[{}] PANIC at {}:{}\nMessage: {}\nBacktrace:\n{:?}",
+            chrono::Local::now().format("%Y-%m-%d %H:%M:%S"),
+            location.file(),
+            location.line(),
+            msg,
+            backtrace
+        );
+
+        eprintln!("{}", crash_msg);
+        tracing::error!("{}", crash_msg);
+
+        // Crashes land in the same shared log file (append — a crash must never
+        // truncate, and the tracing worker may already be dead so we write direct).
+        if let Ok(mut file) = OpenOptions::new().create(true).append(true).open(log_file_path()) {
+            let _ = file.write_all(crash_msg.as_bytes());
+            let _ = file.write_all(b"\n===================================================\n");
+        }
+    }));
+}
+
+/// Initialises tracing and writes to the shared `ostp.log` next to the executable.
+///
+/// The `level` parameter controls the minimum log level:
+/// - `"error"` — only errors
+/// - `"warn"`  — warnings and errors
+/// - `"info"`  — informational messages (default)
+/// - `"debug"` — detailed debug messages (use when `debug: true` in config)
+/// - `"trace"` — all messages including very verbose internal state
+///
+/// The environment variable `RUST_LOG` overrides this value if set.
+///
+/// `truncate`: clear the log at startup. Honoured **only on Windows** — Linux
+/// servers keep their history (OS-rotated). Pass `true` only from the daemon's
+/// own entrypoint; one-shot commands and child processes (the TUN helper) pass
+/// `false` so they append instead of wiping a running daemon's log.
+pub fn init_tracing(
+    level: &str,
+    app_name: &str,
+    version: &str,
+    truncate: bool,
+) -> Option<tracing_appender::non_blocking::WorkerGuard> {
+    // RUST_LOG overrides the config-derived level
+    let env_filter = EnvFilter::try_from_default_env()
+        .unwrap_or_else(|_| {
+            // When debug or trace is requested, enable for all ostp crates
+            if level == "debug" || level == "trace" {
+                // Enable the requested level for ostp crates, but keep noisy deps at warn
+                EnvFilter::new(format!(
+                    "warn,ostp_client={level},ostp_core={level},ostp_jni={level},ostp_gui_lib={level}"
+                ))
+            } else {
+                EnvFilter::new(level)
+            }
+        });
+
+    let path = log_file_path();
+
+    let mut open_opts = OpenOptions::new();
+    open_opts.create(true);
+    // Truncate-on-startup is Windows-only and daemon-only. Everywhere else append:
+    // Linux keeps server history, and one-shot commands / the TUN helper must not
+    // wipe a running daemon's log.
+    if truncate && cfg!(windows) {
+        open_opts.write(true).truncate(true);
+    } else {
+        open_opts.append(true);
+    }
+
+    if let Ok(mut file) = open_opts.open(&path) {
+        // Write the startup banner directly to the log file, bypassing the
+        // tracing subscriber entirely. Emitting it via tracing::info!() hits
+        // BOTH layers below (file AND stderr), so every one-shot CLI command
+        // (`ostp -V`, `ostp gk`, `ostp check`, ...) printed this banner to the
+        // terminal on every single invocation — pure noise for anything that
+        // isn't the long-running daemon. It's still genuinely useful for
+        // whoever's reading the log file later, so keep it there, just not on
+        // screen for commands that aren't the daemon.
+        let _ = writeln!(
+            file,
+            "{} v{} | OS: {} | Arch: {} | log_level: {} | log_file: {}",
+            app_name,
+            version,
+            std::env::consts::OS,
+            std::env::consts::ARCH,
+            level,
+            path.display(),
+        );
+
+        let (file_writer, guard) = tracing_appender::non_blocking(file);
+
+        let fmt_layer = tracing_subscriber::fmt::layer()
+            .with_target(true)
+            .with_line_number(true)
+            .with_thread_ids(false)
+            .with_thread_names(false)
+            .with_ansi(false)
+            .with_writer(file_writer);
+
+        let stderr_layer = tracing_subscriber::fmt::layer()
+            .with_target(true)
+            .with_writer(std::io::stderr);
+
+        let _ = tracing_subscriber::registry()
+            .with(env_filter)
+            .with(fmt_layer)
+            .with(stderr_layer)
+            .try_init();
+
+        Some(guard)
+    } else {
+        // Fallback: stderr only
+        let stderr_layer = tracing_subscriber::fmt::layer()
+            .with_target(true)
+            .with_writer(std::io::stderr);
+        let _ = tracing_subscriber::registry()
+            .with(EnvFilter::new(level))
+            .with(stderr_layer)
+            .try_init();
+        eprintln!("[WARN] Could not open log file at {}. Logging to stderr only.", path.display());
+        None
+    }
+}

ostp-client/src/migrate.rs

@@ -0,0 +1,559 @@
+//! The ONE authoritative place that upgrades an old `config.json` to the
+//! current schema. Reachable only via the explicit `ostp migrate` command —
+//! nothing else in this codebase silently rewrites a user's config on their
+//! behalf (the old 0.3.x line used to auto-migrate on every load with just a
+//! log warning; that's exactly the kind of "invisible until something looks
+//! wrong" behavior this module replaces).
+//!
+//! Every field this module cannot map forward is reported explicitly in
+//! `MigrationReport.notes`, never silently dropped without a trace.
+
+use serde_json::{json, Value};
+
+#[derive(Debug, Default)]
+pub struct MigrationReport {
+    /// Whether anything was actually different from the current schema.
+    pub changed: bool,
+    /// Human-readable line per field added, converted, or dropped.
+    pub notes: Vec<String>,
+}
+
+impl MigrationReport {
+    fn note(&mut self, msg: impl Into<String>) {
+        self.changed = true;
+        self.notes.push(msg.into());
+    }
+}
+
+/// Which config this file is (mirrors `AppMode`'s `"mode"` tag). Old configs
+/// from before that tag existed are sniffed structurally as a fallback.
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum ConfigKind {
+    Client,
+    Server,
+    Relay,
+}
+
+pub fn detect_kind(json: &Value) -> Option<ConfigKind> {
+    match json.get("mode").and_then(|v| v.as_str()) {
+        Some("client") => return Some(ConfigKind::Client),
+        Some("server") => return Some(ConfigKind::Server),
+        Some("relay") => return Some(ConfigKind::Relay),
+        _ => {}
+    }
+    // No (or unrecognized) "mode" tag — this is an older config from before
+    // it was mandatory. Sniff by the fields that have been present on each
+    // shape since the earliest surviving config format.
+    if json.get("upstream_tcp").is_some() || json.get("upstream_api_url").is_some() {
+        Some(ConfigKind::Relay)
+    } else if json.get("access_keys").is_some() || json.get("listen").is_some() {
+        Some(ConfigKind::Server)
+    } else if json.get("access_key").is_some() || json.get("server").is_some() {
+        Some(ConfigKind::Client)
+    } else {
+        None
+    }
+}
+
+/// Migrates a client config of any known past shape to the current flat
+/// schema. Returns the migrated JSON and a report of every change made.
+///
+/// Known input shapes, oldest first:
+/// - **v0.3.1–v0.3.21 "modular multi-server"**: `inbounds`/`outbounds` arrays
+///   + `routing.rules`. Only the first `ostp`-type outbound is kept (this
+///   line no longer supports multiple simultaneous servers); every other
+///   `ostp` outbound is reported by tag+address so nothing vanishes
+///   invisibly. `urltest`/`selector`/`direct`/`block` outbounds have no
+///   equivalent and are dropped (reported).
+/// - **pre-0.3.1 flat (up to v0.2.98)**: same field names as today
+///   (`server`, `access_key`, `tun`, `exclude`, `mux`, `transport`, ...)
+///   except `tun.wintun_path`/`tun.ipv4_address` (internal driver detail,
+///   never user-meaningful data) and `transport.wss` (the WSS framing
+///   feature removed entirely in the 0.4.0 rebuild) — both dropped with an
+///   explicit note; everything else maps 1:1, nothing to convert.
+/// - **configs carrying a leftover `transport.stealth_sni`**: dropped with a
+///   note, same reasoning as `wss` — it never fed into anything on the wire
+///   (no TLS/HTTP mimicry exists in this project), so there is no successor
+///   field. Not tied to a specific version: it lingered in the schema well
+///   past when the mimicry work it was meant for got removed.
+/// - **current flat schema**: no-op, `changed = false`.
+pub fn migrate_client_json(json: Value) -> (Value, MigrationReport) {
+    let mut report = MigrationReport::default();
+
+    let has_inbounds = json.get("inbounds").and_then(|v| v.as_array()).is_some();
+    let has_outbounds = json.get("outbounds").and_then(|v| v.as_array()).is_some();
+
+    if has_inbounds && has_outbounds {
+        return migrate_client_from_modular(json, report);
+    }
+
+    // Flat shape already (current or pre-0.3.1) — normalize obsolete fields
+    // in place rather than rebuilding the whole document from scratch, so
+    // any field this module doesn't know about yet still survives untouched.
+    let mut out = json;
+
+    if let Some(tun) = out.get_mut("tun").and_then(|t| t.as_object_mut()) {
+        for dead_field in ["wintun_path", "ipv4_address"] {
+            if tun.remove(dead_field).is_some() {
+                report.note(format!(
+                    "Dropped tun.{dead_field} — internal driver detail from an older WinTun \
+                     integration, not applicable to the current TUN implementation."
+                ));
+            }
+        }
+    }
+    if let Some(transport) = out.get_mut("transport").and_then(|t| t.as_object_mut()) {
+        if transport.remove("wss").is_some() {
+            report.note(
+                "Dropped transport.wss — WSS framing was removed in the 0.4.0 rebuild \
+                 (the project follows a zapret-like approach: no protocol mimicry, \
+                 just packet-level obfuscation/manipulation, so there is no successor field)."
+                    .to_string(),
+            );
+        }
+        if transport.remove("stealth_sni").is_some() {
+            report.note(
+                "Dropped transport.stealth_sni — never actually used to construct any wire \
+                 bytes (no TLS/HTTP mimicry exists in this project — same zapret-like \
+                 reasoning as transport.wss), so it was unused config plumbing with no effect."
+                    .to_string(),
+            );
+        }
+    }
+
+    (out, report)
+}
+
+fn migrate_client_from_modular(json: Value, mut report: MigrationReport) -> (Value, MigrationReport) {
+    report.changed = true; // the shape itself is being replaced regardless of field-level detail
+
+    let inbounds = json.get("inbounds").and_then(|v| v.as_array()).cloned().unwrap_or_default();
+    let outbounds = json.get("outbounds").and_then(|v| v.as_array()).cloned().unwrap_or_default();
+    let routing = json.get("routing").cloned().unwrap_or(json!({}));
+    let default_outbound = routing.get("default_outbound").and_then(|v| v.as_str()).map(String::from);
+
+    // ── Pick the primary "ostp" outbound ────────────────────────────────
+    // Prefer the one routing.default_outbound points at (directly, or via a
+    // urltest/selector group that references it); otherwise take the first
+    // ostp outbound in file order. Every other ostp outbound is reported by
+    // tag+address, not silently discarded.
+    let ostp_outbounds: Vec<&Value> = outbounds
+        .iter()
+        .filter(|o| o.get("type").and_then(|t| t.as_str()) == Some("ostp"))
+        .collect();
+
+    // default_outbound might name an ostp outbound directly, OR name a
+    // urltest/selector GROUP whose first member is the one to actually use —
+    // check both, since a plain `.or_else` here would never even attempt the
+    // group lookup while default_outbound is Some(_) (which it almost always
+    // is), silently falling through to "just take the first ostp outbound in
+    // file order" instead — exactly the kind of silent wrong answer this
+    // migrator exists to avoid.
+    let primary_tag: Option<String> = default_outbound.as_deref().and_then(|def_tag| {
+        if ostp_outbounds.iter().any(|o| o.get("tag").and_then(|t| t.as_str()) == Some(def_tag)) {
+            return Some(def_tag.to_string());
+        }
+        outbounds.iter().find_map(|o| {
+            let is_group = matches!(o.get("type").and_then(|t| t.as_str()), Some("urltest") | Some("selector"));
+            let tag_matches = o.get("tag").and_then(|t| t.as_str()) == Some(def_tag);
+            if is_group && tag_matches {
+                o.get("outbounds")
+                    .and_then(|v| v.as_array())
+                    .and_then(|arr| arr.first())
+                    .and_then(|v| v.as_str())
+                    .map(String::from)
+            } else {
+                None
+            }
+        })
+    });
+
+    let primary = primary_tag
+        .as_deref()
+        .and_then(|tag| ostp_outbounds.iter().find(|o| o.get("tag").and_then(|t| t.as_str()) == Some(tag)))
+        .copied()
+        .or_else(|| ostp_outbounds.first().copied());
+
+    let Some(primary) = primary else {
+        report.note(
+            "No 'ostp'-type outbound found in the old modular config — nothing to migrate \
+             the server connection from. Wrote a placeholder; you MUST fill in server/access_key \
+             by hand or re-import a share link."
+                .to_string(),
+        );
+        return (
+            json!({
+                "server": "127.0.0.1:50000",
+                "access_key": "",
+            }),
+            report,
+        );
+    };
+
+    for other in &ostp_outbounds {
+        if !std::ptr::eq(*other, primary) {
+            let tag = other.get("tag").and_then(|t| t.as_str()).unwrap_or("?");
+            let addr = other.get("server").and_then(|t| t.as_str()).unwrap_or("?");
+            let port = other.get("port").and_then(|t| t.as_u64()).unwrap_or(0);
+            report.note(format!(
+                "Dropped additional server '{tag}' ({addr}:{port}) — multi-server / urltest \
+                 failover is no longer supported; only one server per config now. Kept the \
+                 one from routing.default_outbound (or the first one if that wasn't set)."
+            ));
+        }
+    }
+
+    let server = primary.get("server").and_then(|v| v.as_str()).unwrap_or("127.0.0.1").to_string();
+    let port = primary.get("port").and_then(|v| v.as_u64()).unwrap_or(50000);
+    let access_key = primary.get("access_key").and_then(|v| v.as_str()).unwrap_or("").to_string();
+    let transport_type = primary
+        .get("transport")
+        .and_then(|t| t.get("type").or_else(|| t.get("mode")))
+        .and_then(|v| v.as_str())
+        .unwrap_or("udp")
+        .to_string();
+    if let Some(sni) = primary.get("transport").and_then(|t| t.get("stealth_sni")).and_then(|v| v.as_str()) {
+        if !sni.is_empty() {
+            report.note(format!(
+                "Dropped transport.stealth_sni ({sni:?}) — never actually used to construct \
+                 any wire bytes; unused config plumbing with no successor field."
+            ));
+        }
+    }
+    let tcp_fragmentation = primary
+        .get("transport")
+        .and_then(|t| t.get("tcp_fragmentation"))
+        .and_then(|v| v.as_bool())
+        .unwrap_or(false);
+    let mux_enabled = primary.get("multiplex").and_then(|m| m.get("enabled")).and_then(|v| v.as_bool()).unwrap_or(false);
+    let mux_sessions = primary.get("multiplex").and_then(|m| m.get("sessions")).and_then(|v| v.as_u64()).unwrap_or(1);
+
+    // ── TUN + local proxy inbounds ───────────────────────────────────────
+    let tun_inbound = inbounds.iter().find(|i| i.get("type").and_then(|t| t.as_str()) == Some("tun"));
+    let proxy_inbound = inbounds.iter().find(|i| i.get("type").and_then(|t| t.as_str()) == Some("local_proxy"));
+
+    let tun_enable = tun_inbound.is_some();
+    let mtu = tun_inbound.and_then(|t| t.get("mtu")).and_then(|v| v.as_u64());
+
+    let socks5_bind = proxy_inbound
+        .map(|p| {
+            let listen = p.get("listen").and_then(|v| v.as_str()).unwrap_or("127.0.0.1");
+            let port = p.get("port").and_then(|v| v.as_u64()).unwrap_or(1088);
+            format!("{listen}:{port}")
+        })
+        .unwrap_or_else(|| "127.0.0.1:1088".to_string());
+
+    // ── Exclusions from routing.rules → direct ──────────────────────────
+    let mut ex_domains: Vec<String> = Vec::new();
+    let mut ex_ips: Vec<String> = Vec::new();
+    let mut ex_processes: Vec<String> = Vec::new();
+    if let Some(rules) = routing.get("rules").and_then(|v| v.as_array()) {
+        for rule in rules {
+            if rule.get("outbound").and_then(|v| v.as_str()) != Some("direct") {
+                continue; // only "route to direct" rules were ever exclusions in the old format
+            }
+            if let Some(v) = rule.get("domain_suffix").and_then(|v| v.as_array()) {
+                ex_domains.extend(v.iter().filter_map(|s| s.as_str().map(String::from)));
+            }
+            if let Some(v) = rule.get("ip_cidr").and_then(|v| v.as_array()) {
+                ex_ips.extend(v.iter().filter_map(|s| s.as_str().map(String::from)));
+            }
+            if let Some(v) = rule.get("process_name").and_then(|v| v.as_array()) {
+                ex_processes.extend(v.iter().filter_map(|s| s.as_str().map(String::from)));
+            }
+        }
+    }
+    for other_rule_outbound in routing
+        .get("rules")
+        .and_then(|v| v.as_array())
+        .into_iter()
+        .flatten()
+        .filter_map(|r| r.get("outbound").and_then(|v| v.as_str()))
+        .filter(|o| *o != "direct")
+    {
+        report.note(format!(
+            "Dropped a routing rule targeting outbound '{other_rule_outbound}' — only \
+             \"route to direct\" rules map to today's exclusions; anything else \
+             (custom per-domain outbound selection) has no equivalent anymore."
+        ));
+    }
+
+    let debug = json.get("log").and_then(|l| l.get("level")).and_then(|v| v.as_str()) == Some("debug");
+
+    let mut client = json!({
+        "server": server,
+        "port": port,
+        "access_key": access_key,
+        "socks5_bind": socks5_bind,
+        "debug": debug,
+        "tun": {
+            "enable": tun_enable,
+            "dns": null,
+            "kill_switch": false,
+        },
+        "exclude": {
+            "domains": ex_domains,
+            "ips": ex_ips,
+            "processes": ex_processes,
+        },
+        "mux": {
+            "enabled": mux_enabled,
+            "sessions": mux_sessions,
+        },
+        "transport": {
+            "mode": transport_type,
+            "tcp_fragmentation": tcp_fragmentation,
+        },
+    });
+    if let Some(mtu) = mtu {
+        client["mtu"] = json!(mtu);
+    }
+    if let Some(gui) = json.get("gui") {
+        client["gui"] = gui.clone();
+    }
+
+    (client, report)
+}
+
+/// Migrates a server config. The server shape has stayed structurally
+/// identical since the earliest surviving version — this only backfills the
+/// `api` section (added after some configs already existed) and drops the
+/// legacy `api.token` field. Ported from the ad-hoc Python snippet that used
+/// to live in `scripts/install.sh` and only ran at install/update time.
+pub fn migrate_server_json(json: Value) -> (Value, MigrationReport) {
+    let mut report = MigrationReport::default();
+    let mut out = json;
+
+    let obj = match out.as_object_mut() {
+        Some(o) => o,
+        None => return (out, report),
+    };
+
+    let api = obj.entry("api").or_insert_with(|| json!({}));
+    if let Some(api_obj) = api.as_object_mut() {
+        let defaults: [(&str, Value); 5] = [
+            ("enabled", json!(false)),
+            ("bind", json!("0.0.0.0:9090")),
+            ("webpath", json!("")),
+            ("username", json!("")),
+            ("password_hash", json!("")),
+        ];
+        for (key, default) in defaults {
+            if !api_obj.contains_key(key) {
+                report.note(format!("Added api.{key} = {default} (missing default)"));
+                api_obj.insert(key.to_string(), default);
+            }
+        }
+        if api_obj.remove("token").is_some() {
+            report.note(
+                "Dropped legacy api.token — superseded by api.password_hash; \
+                 set a new admin password with the management API or panel."
+                    .to_string(),
+            );
+        }
+    }
+
+    (out, report)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    /// A realistic v0.3.21-shaped modular config (TUN + local_proxy inbounds,
+    /// a single ostp outbound, exclusion rules, mux) — mirrors the actual
+    /// shape from that tag, field for field.
+    #[test]
+    fn modular_single_server_preserves_every_field() {
+        let old = json!({
+            "version": "0.3.21",
+            "log": { "level": "debug" },
+            "inbounds": [
+                { "type": "tun", "tag": "tun-in", "auto_route": true, "mtu": 1350 },
+                { "type": "local_proxy", "tag": "socks-in", "protocol": "socks", "listen": "127.0.0.1", "port": 1088 }
+            ],
+            "outbounds": [
+                {
+                    "type": "ostp", "tag": "proxy",
+                    "server": "203.0.113.5", "port": 50000, "access_key": "sekrit123",
+                    "transport": { "type": "uot", "stealth_sni": "vk.com", "tcp_fragmentation": true },
+                    "multiplex": { "enabled": true, "sessions": 4 }
+                },
+                { "type": "direct", "tag": "direct" },
+                { "type": "block", "tag": "block" }
+            ],
+            "routing": {
+                "rules": [
+                    { "domain_suffix": ["local.lan", "internal.corp"], "outbound": "direct" },
+                    { "ip_cidr": ["192.168.0.0/16"], "outbound": "direct" },
+                    { "process_name": ["steam.exe"], "outbound": "direct" }
+                ],
+                "default_outbound": "proxy"
+            }
+        });
+
+        let (new, report) = migrate_client_json(old);
+        assert!(report.changed);
+        assert_eq!(new["server"], "203.0.113.5");
+        assert_eq!(new["port"], 50000);
+        assert_eq!(new["access_key"], "sekrit123");
+        assert_eq!(new["socks5_bind"], "127.0.0.1:1088");
+        assert_eq!(new["mtu"], 1350);
+        assert_eq!(new["debug"], true);
+        assert_eq!(new["tun"]["enable"], true);
+        assert_eq!(new["transport"]["mode"], "uot");
+        assert_eq!(new["transport"]["tcp_fragmentation"], true);
+        assert_eq!(new["mux"]["enabled"], true);
+        assert_eq!(new["mux"]["sessions"], 4);
+        assert_eq!(new["exclude"]["domains"], json!(["local.lan", "internal.corp"]));
+        assert_eq!(new["exclude"]["ips"], json!(["192.168.0.0/16"]));
+        assert_eq!(new["exclude"]["processes"], json!(["steam.exe"]));
+        // stealth_sni never fed into any wire bytes — dropped, not carried forward.
+        assert!(new["transport"].get("stealth_sni").is_none());
+        assert!(report.notes.iter().any(|n| n.contains("stealth_sni") && n.contains("vk.com")));
+    }
+
+    /// Old modular configs that had MULTIPLE ostp outbounds (multi-server) —
+    /// must keep the one routing.default_outbound points at and report every
+    /// other one by name/address rather than picking silently.
+    #[test]
+    fn modular_multi_server_keeps_default_and_reports_the_rest() {
+        let old = json!({
+            "inbounds": [],
+            "outbounds": [
+                { "type": "ostp", "tag": "proxy-0", "server": "1.1.1.1", "port": 50000, "access_key": "k1" },
+                { "type": "ostp", "tag": "proxy-1", "server": "2.2.2.2", "port": 50000, "access_key": "k2" },
+                {
+                    "type": "urltest", "tag": "proxy",
+                    "outbounds": ["proxy-1", "proxy-0"], "url": "http://cp.cloudflare.com"
+                }
+            ],
+            "routing": { "rules": [], "default_outbound": "proxy" }
+        });
+
+        let (new, report) = migrate_client_json(old);
+        // urltest's first member (proxy-1 / 2.2.2.2) is the one actually picked.
+        assert_eq!(new["server"], "2.2.2.2");
+        assert_eq!(new["access_key"], "k2");
+        assert!(report.notes.iter().any(|n| n.contains("proxy-0") && n.contains("1.1.1.1")));
+    }
+
+    /// Pre-0.3.1 flat config carrying fields that no longer exist
+    /// (tun.wintun_path, tun.ipv4_address, transport.wss, transport.stealth_sni)
+    /// — those get dropped with a note; every field that's still meaningful
+    /// passes through untouched, byte for byte.
+    #[test]
+    fn flat_legacy_drops_only_dead_fields() {
+        let old = json!({
+            "server": "198.51.100.9:50000",
+            "access_key": "oldkey",
+            "mtu": 1200,
+            "socks5_bind": "127.0.0.1:1090",
+            "tun": {
+                "enable": true,
+                "wintun_path": "C:\\Program Files\\wintun\\wintun.dll",
+                "ipv4_address": "10.0.0.2",
+                "dns": "1.1.1.1",
+                "kill_switch": true
+            },
+            "exclude": { "domains": ["a.com"], "ips": null, "processes": null },
+            "mux": { "enabled": false, "sessions": 1 },
+            "transport": { "mode": "udp", "stealth_sni": "bing.com", "wss": true }
+        });
+
+        let (new, report) = migrate_client_json(old);
+        assert!(report.changed);
+        // Untouched fields survive exactly as they were.
+        assert_eq!(new["server"], "198.51.100.9:50000");
+        assert_eq!(new["access_key"], "oldkey");
+        assert_eq!(new["mtu"], 1200);
+        assert_eq!(new["tun"]["enable"], true);
+        assert_eq!(new["tun"]["dns"], "1.1.1.1");
+        assert_eq!(new["tun"]["kill_switch"], true);
+        assert_eq!(new["exclude"]["domains"], json!(["a.com"]));
+        // Dead fields are gone...
+        assert!(new["tun"].get("wintun_path").is_none());
+        assert!(new["tun"].get("ipv4_address").is_none());
+        assert!(new["transport"].get("wss").is_none());
+        assert!(new["transport"].get("stealth_sni").is_none());
+        // ...and their removal was reported, not silent.
+        assert!(report.notes.iter().any(|n| n.contains("wintun_path")));
+        assert!(report.notes.iter().any(|n| n.contains("ipv4_address")));
+        assert!(report.notes.iter().any(|n| n.contains("wss")));
+        assert!(report.notes.iter().any(|n| n.contains("stealth_sni")));
+    }
+
+    /// A config already in the current shape must be a true no-op: report
+    /// says nothing changed, and every field is untouched.
+    #[test]
+    fn current_flat_config_is_a_no_op() {
+        let current = json!({
+            "server": "example.com:50000",
+            "access_key": "k",
+            "tun": { "enable": false, "dns": null, "kill_switch": false },
+            "exclude": { "domains": [], "ips": [], "processes": [] },
+            "mux": { "enabled": false, "sessions": 1 },
+            "transport": { "mode": "udp", "tcp_fragmentation": false }
+        });
+        let (new, report) = migrate_client_json(current.clone());
+        assert!(!report.changed);
+        assert_eq!(new, current);
+    }
+
+    /// Every migrated output must actually deserialize into the ONE
+    /// canonical schema (`crate::config`) — this is the same check
+    /// `cmd_migrate` runs at runtime before ever touching a user's file,
+    /// exercised here directly so a schema/migrator drift fails a fast unit
+    /// test instead of surfacing as "your migrated config won't load".
+    #[test]
+    fn every_migrated_output_matches_the_canonical_schema() {
+        let modular = json!({
+            "inbounds": [{ "type": "tun", "tag": "tun-in", "mtu": 1350 }],
+            "outbounds": [
+                { "type": "ostp", "tag": "proxy", "server": "1.2.3.4", "port": 50000, "access_key": "k" },
+                { "type": "direct", "tag": "direct" }
+            ],
+            "routing": { "rules": [], "default_outbound": "proxy" }
+        });
+        let (new, _) = migrate_client_json(modular);
+        serde_json::from_value::<crate::config::ClientFileConfig>(new)
+            .expect("modular->flat migration output must match ClientFileConfig");
+
+        let legacy_flat = json!({
+            "server": "1.2.3.4:50000",
+            "access_key": "k",
+            "tun": { "enable": true, "wintun_path": "x", "ipv4_address": "y" }
+        });
+        let (new, _) = migrate_client_json(legacy_flat);
+        serde_json::from_value::<crate::config::ClientFileConfig>(new)
+            .expect("legacy-flat migration output must match ClientFileConfig");
+
+        let server = json!({ "listen": "0.0.0.0:50000", "access_keys": ["k"] });
+        let (new, _) = migrate_server_json(server);
+        serde_json::from_value::<crate::config::ServerConfig>(new)
+            .expect("server migration output must match ServerConfig");
+    }
+
+    #[test]
+    fn server_config_backfills_api_defaults_and_drops_legacy_token() {
+        let old = json!({
+            "listen": "0.0.0.0:50000",
+            "access_keys": ["k1"],
+            "api": { "token": "old-plain-token" }
+        });
+        let (new, report) = migrate_server_json(old);
+        assert!(report.changed);
+        assert_eq!(new["api"]["enabled"], false);
+        assert_eq!(new["api"]["bind"], "0.0.0.0:9090");
+        assert!(new["api"].get("token").is_none());
+        assert!(report.notes.iter().any(|n| n.contains("api.token")));
+    }
+
+    #[test]
+    fn detect_kind_falls_back_to_structural_sniffing_without_mode_tag() {
+        assert_eq!(detect_kind(&json!({"access_key": "x", "server": "y"})), Some(ConfigKind::Client));
+        assert_eq!(detect_kind(&json!({"access_keys": ["x"], "listen": "y"})), Some(ConfigKind::Server));
+        assert_eq!(detect_kind(&json!({"upstream_tcp": "x", "upstream_api_url": "y"})), Some(ConfigKind::Relay));
+        assert_eq!(detect_kind(&json!({"mode": "client", "server": "x"})), Some(ConfigKind::Client));
+    }
+}

ostp-client/src/runner.rs

@@ -10,10 +10,9 @@ use std::fs::OpenOptions;
 use std::io::Write as _;
 
 fn log_to_core_file(msg: &str) {
-    let path = std::env::current_exe()
-        .ok()
-        .and_then(|p| p.parent().map(|d| d.join("ostp-core.log")))
-        .unwrap_or_else(|| std::path::PathBuf::from("ostp-core.log"));
+    // Writes into the single shared ostp.log (same file as the tracing appender),
+    // not a separate ostp-core.log — see logging::LOG_FILE_NAME.
+    let path = crate::logging::log_file_path();
     if let Ok(mut file) = OpenOptions::new().create(true).append(true).open(path) {
         let _ = writeln!(file, "[{}] {}", chrono::Local::now().format("%Y-%m-%d %H:%M:%S"), msg);
     }
@@ -180,13 +179,71 @@ pub async fn run_client(config: crate::config::ClientConfig) -> Result<()> {
         }
     });
 
-    run_client_core(config, metrics, shutdown_rx).await
+    run_client_core(config, metrics, shutdown_rx, None).await
 }
 
+/// Runs the client with auto-reconnect: any subsystem ending — a network
+/// change stranding the TUN adapter/UDP socket on a dead interface, the OSTP
+/// protocol connection dropping in a way the inner Bridge-level retry (see
+/// `UiEvent::TunnelStopped` below) couldn't recover from, or a proxy/TUN task
+/// crashing outright — triggers a full clean restart (fresh DNS resolution,
+/// fresh Bridge, fresh TUN/proxy) with exponential backoff, instead of the
+/// client just dying. Only an explicit shutdown request stops this loop.
 pub async fn run_client_core(
+    config: crate::config::ClientConfig,
+    metrics: Arc<BridgeMetrics>,
+    mut shutdown_rx_ext: watch::Receiver<bool>,
+    config_rx: Option<watch::Receiver<crate::config::ClientConfig>>,
+) -> Result<()> {
+    use portable_atomic::Ordering;
+
+    const BACKOFF_SCHEDULE_SECS: [u64; 6] = [1, 2, 5, 10, 20, 30];
+    // A run that stayed up at least this long counts as "was actually
+    // connected", so a later drop restarts the backoff from the top instead
+    // of inheriting a long delay from a previous flaky stretch.
+    const STABLE_UPTIME: std::time::Duration = std::time::Duration::from_secs(60);
+    let mut backoff_idx = 0usize;
+
+    loop {
+        if *shutdown_rx_ext.borrow() {
+            return Ok(());
+        }
+
+        let attempt_start = std::time::Instant::now();
+        let result = run_client_once(config.clone(), metrics.clone(), shutdown_rx_ext.clone(), config_rx.clone()).await;
+
+        if *shutdown_rx_ext.borrow() {
+            // Shutdown was requested during (or right after) this attempt — honor it, don't retry.
+            return result;
+        }
+        if let Err(ref e) = result {
+            tracing::warn!("client run ended unexpectedly, will auto-reconnect: {e}");
+        }
+
+        if attempt_start.elapsed() >= STABLE_UPTIME {
+            backoff_idx = 0;
+        }
+        let delay = BACKOFF_SCHEDULE_SECS[backoff_idx.min(BACKOFF_SCHEDULE_SECS.len() - 1)];
+        backoff_idx += 1;
+
+        // Reflect the retry wait as "connecting" rather than "disconnected".
+        metrics.connection_state.store(1, Ordering::Relaxed);
+        tokio::select! {
+            _ = tokio::time::sleep(std::time::Duration::from_secs(delay)) => {}
+            _ = shutdown_rx_ext.changed() => {
+                if *shutdown_rx_ext.borrow() {
+                    return Ok(());
+                }
+            }
+        }
+    }
+}
+
+async fn run_client_once(
     mut config: crate::config::ClientConfig,
     metrics: Arc<BridgeMetrics>,
     mut shutdown_rx_ext: watch::Receiver<bool>,
+    mut config_rx: Option<watch::Receiver<crate::config::ClientConfig>>,
 ) -> Result<()> {
     #[cfg(target_os = "windows")]
     if config.mode == "tun" && !is_admin() {
@@ -238,19 +295,25 @@ pub async fn run_client_core(
     }
 
     let _sysproxy_guard = if config.mode == "proxy" {
-        Some(crate::sysproxy::SystemProxyGuard::enable(&config.local_proxy.bind_addr))
+        // Enable system proxy and set initial ProxyOverride with user exclusions
+        let guard = Some(crate::sysproxy::SystemProxyGuard::enable(&config.local_proxy.bind_addr));
+        crate::sysproxy::update_proxy_bypass_list(
+            &config.exclusions.domains,
+            &config.exclusions.ips,
+        );
+        guard
     } else {
         None
     };
 
-    if config.mode == "tun" && !config.exclusions.processes.is_empty() {
-        println!("[ostp] Process exclusions are not supported in TUN mode");
-    }
-
     let (proxy_events_tx, proxy_events_rx) = mpsc::channel(256);
     let (client_msgs_tx, client_msgs_rx) = mpsc::unbounded_channel();
+    
+    // Setup exclusions hot-reload channel
+    let (reload_tx, reload_rx) = watch::channel(config.exclusions.clone());
 
-    let bridge = Bridge::new(&config, metrics)?;
+    let mut bridge = Bridge::new(&config, metrics)?;
+    bridge.reload_tx = Some(reload_tx.clone());
 
     let (ui_tx, mut ui_rx) = mpsc::channel(512);
     let (cmd_tx, cmd_rx) = mpsc::channel(128);
@@ -305,11 +368,12 @@ pub async fn run_client_core(
     });
 
     let config_clone = config.clone();
+    let proxy_exclusions_rx = reload_rx.clone();
     let mut proxy_task = tokio::spawn(async move {
         tunnel::run_local_proxy(
             config.local_proxy,
             config.ostp,
-            config.exclusions,
+            proxy_exclusions_rx,
             config.debug,
             proxy_shutdown_rx,
             proxy_events_tx,
@@ -319,14 +383,49 @@ pub async fn run_client_core(
     });
 
     let wintun_shutdown_rx = shutdown_tx.subscribe();
+    let wintun_exclusions_rx = reload_rx.clone();
     let mut wintun_task = if config_clone.mode == "tun" {
         Some(tokio::spawn(async move {
-            tunnel::run_tun_tunnel(config_clone, wintun_shutdown_rx).await
+            tunnel::run_tun_tunnel(config_clone, wintun_shutdown_rx, wintun_exclusions_rx).await
         }))
     } else {
         None
     };
 
+    // Wait for local_shutdown
+    let mut local_shutdown = shutdown_rx_ext.clone();
+    let cmd_tx_loop = cmd_tx.clone();
+    tokio::spawn(async move {
+        loop {
+            tokio::select! {
+                _ = local_shutdown.changed() => {
+                    if *local_shutdown.borrow() {
+                        let _ = cmd_tx_loop.send(BridgeCommand::Shutdown).await;
+                        break;
+                    }
+                }
+                Some(Ok(_)) = async {
+                    if let Some(ref mut rx) = config_rx {
+                        Some(rx.changed().await)
+                    } else {
+                        std::future::pending().await
+                    }
+                } => {
+                    if let Some(ref rx) = config_rx {
+                        let new_cfg = rx.borrow().clone();
+                        // Update Windows ProxyOverride so excluded domains/IPs
+                        // bypass the system proxy immediately (proxy mode only).
+                        crate::sysproxy::update_proxy_bypass_list(
+                            &new_cfg.exclusions.domains,
+                            &new_cfg.exclusions.ips,
+                        );
+                        let _ = reload_tx.send(new_cfg.exclusions);
+                    }
+                }
+            }
+        }
+    });
+
     // Wait for either external shutdown OR any task to fail
     tokio::select! {
         _ = shutdown_rx_ext.changed() => {

ostp-client/src/sysproxy.rs

@@ -64,7 +64,79 @@ pub fn enable_windows_proxy(proxy_addr: &str) {
         _ => {}
     }
 
-    // Set bypass list to prevent proxy loop for localhost traffic
+    // Set initial bypass list (will be expanded by update_proxy_bypass_list)
+    update_proxy_bypass_list_windows(&[], &[]);
+
+    refresh_wininet();
+    tracing::info!("System proxy enabled successfully");
+}
+
+/// Update the Windows ProxyOverride registry value to include user-configured
+/// excluded domains and IPs. This makes excluded hosts bypass the OSTP proxy
+/// entirely at the OS level — the most reliable split-tunneling mechanism.
+///
+/// For each domain `d`, adds both `d` and `*.d` so both the root and all
+/// subdomains bypass the proxy.
+/// For IPs, adds them verbatim (Windows supports exact IPs and wildcards like
+/// `192.168.*`).
+#[cfg(target_os = "windows")]
+pub fn update_proxy_bypass_list(domains: &[String], ips: &[String]) {
+    update_proxy_bypass_list_windows(domains, ips);
+    refresh_wininet();
+}
+
+#[cfg(not(target_os = "windows"))]
+pub fn update_proxy_bypass_list(_domains: &[String], _ips: &[String]) {
+    // Linux/macOS: no-op (gnome/kde proxy bypass list update not implemented)
+}
+
+#[cfg(target_os = "windows")]
+fn update_proxy_bypass_list_windows(domains: &[String], ips: &[String]) {
+    // Base list: always bypass local addresses
+    let mut parts: Vec<String> = vec![
+        "localhost".into(),
+        "127.*".into(),
+        "10.*".into(),
+        "172.16.*".into(),
+        "172.17.*".into(),
+        "172.18.*".into(),
+        "172.19.*".into(),
+        "172.20.*".into(),
+        "172.21.*".into(),
+        "172.22.*".into(),
+        "172.23.*".into(),
+        "172.24.*".into(),
+        "172.25.*".into(),
+        "172.26.*".into(),
+        "172.27.*".into(),
+        "172.28.*".into(),
+        "172.29.*".into(),
+        "172.30.*".into(),
+        "172.31.*".into(),
+        "192.168.*".into(),
+        "<local>".into(),
+    ];
+
+    // Add excluded domains: both exact and wildcard subdomain form
+    for d in domains {
+        let d = d.trim().trim_start_matches('.').to_lowercase();
+        if d.is_empty() { continue; }
+        parts.push(d.clone());
+        parts.push(format!("*.{}", d));
+    }
+
+    // Add excluded IPs verbatim
+    for ip in ips {
+        let ip = ip.trim();
+        if ip.is_empty() { continue; }
+        // Strip CIDR suffix if present — Windows ProxyOverride doesn't support CIDR
+        let host = ip.split('/').next().unwrap_or(ip);
+        parts.push(host.to_string());
+    }
+
+    let override_value = parts.join(";");
+    tracing::info!("Updating ProxyOverride: {}", override_value);
+
     let _ = Command::new("reg")
         .creation_flags(CREATE_NO_WINDOW)
         .args([
@@ -72,13 +144,10 @@ pub fn enable_windows_proxy(proxy_addr: &str) {
             "HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Internet Settings",
             "/v", "ProxyOverride",
             "/t", "REG_SZ",
-            "/d", "localhost;127.*;10.*;192.168.*;<local>",
+            "/d", &override_value,
             "/f",
         ])
         .output();
-
-    refresh_wininet();
-    tracing::info!("System proxy enabled successfully");
 }
 
 #[cfg(target_os = "windows")]
@@ -120,7 +189,7 @@ fn refresh_wininet() {
 #[cfg(not(target_os = "windows"))]
 pub fn enable_system_proxy(proxy_addr: &str) {
     let parts: Vec<&str> = proxy_addr.split(':').collect();
-    let host = parts.get(0).unwrap_or(&"127.0.0.1");
+    let host = parts.first().unwrap_or(&"127.0.0.1");
     let port = parts.get(1).unwrap_or(&"1088");
 
     let is_gui = std::env::var("DISPLAY").is_ok() || std::env::var("WAYLAND_DISPLAY").is_ok();
@@ -166,7 +235,7 @@ pub fn enable_system_proxy(proxy_addr: &str) {
     println!("OSTP Local Proxy is running at socks5://{}", proxy_addr);
     println!("Since you are in a headless/terminal environment, OSTP cannot automatically");
     println!("configure your system proxy. To route traffic from this terminal, run:");
-    println!("\n    eval $(ostp --proxy-env)\n");
+    println!("\n    eval $(ostp proxy-env)\n");
     println!("Or configure your application (e.g. curl -x socks5://{})", proxy_addr);
     println!("===================================================================\n");
 }

ostp-client/src/transport/mod.rs

@@ -1,4 +1,3 @@
-pub mod xhttp;
 
 use std::sync::Arc;
 use tokio::net::UdpSocket;

ostp-client/src/transport/xhttp.rs

@@ -1,384 +0,0 @@
-use std::net::IpAddr;
-use std::sync::Arc;
-use tokio::io::{AsyncReadExt, AsyncWriteExt};
-use tokio::net::TcpStream;
-use bytes::{Buf, BufMut, Bytes, BytesMut};
-use anyhow::{Result, Context};
-use tokio::sync::mpsc;
-use hmac::Hmac;
-use sha2::Sha256;
-use base64::Engine;
-use std::pin::Pin;
-use std::task::{Context as TaskContext, Poll};
-use x25519_dalek::PublicKey;
-use chacha20poly1305::{aead::Aead, ChaCha20Poly1305, Nonce};
-
-use ostp_core::crypto::reality::{build_client_hello, derive_keys, generate_session_id, generate_x25519_keypair, REALITY_SERVER_HANDSHAKE_RECORDS};
-use ostp_core::framing::wss::{encode_wss_frame, decode_wss_frame, WssFrameResult};
-
-type HmacSha256 = Hmac<Sha256>;
-
-pub async fn connect_xhttp(
-    target_ip: IpAddr,
-    port: u16,
-    sni: &str,
-    access_key: &[u8],
-    reality_enabled: bool,
-    wss: bool,
-    reality_pbk: &str,
-    reality_sid: &str,
-) -> Result<(mpsc::Sender<Bytes>, Arc<tokio::sync::Mutex<mpsc::Receiver<Bytes>>>)> {
-    let addr = std::net::SocketAddr::new(target_ip, port);
-    
-    #[cfg(not(target_os = "android"))]
-    let mut tcp_stream = tokio::net::TcpStream::connect(addr).await
-        .with_context(|| format!("failed to connect to {}", addr))?;
-
-    #[cfg(target_os = "android")]
-    let mut tcp_stream = {
-        let domain = if target_ip.is_ipv6() { socket2::Domain::IPV6 } else { socket2::Domain::IPV4 };
-        let sock = socket2::Socket::new(domain, socket2::Type::STREAM, Some(socket2::Protocol::TCP))?;
-        
-        use std::os::unix::io::AsRawFd;
-        crate::bridge::protect_socket(sock.as_raw_fd());
-        
-        sock.set_nonblocking(true)?;
-        let tcp_socket = tokio::net::TcpSocket::from_std_stream(sock.into());
-        tcp_socket.connect(addr).await
-            .with_context(|| format!("failed to connect to {}", addr))?
-    };
-
-    tcp_stream.set_nodelay(true)?;
-
-    if reality_enabled {
-        let pbk_bytes = base64::engine::general_purpose::URL_SAFE_NO_PAD.decode(reality_pbk)
-            .context("invalid reality_pbk base64")?;
-        if pbk_bytes.len() != 32 {
-            anyhow::bail!("reality_pbk must be 32 bytes");
-        }
-        let pbk = PublicKey::from(<[u8; 32]>::try_from(pbk_bytes.as_slice()).unwrap());
-        
-        let sid_bytes_vec = hex::decode(reality_sid).context("invalid reality_sid hex")?;
-        if sid_bytes_vec.len() != 8 {
-            anyhow::bail!("reality_sid must be 8 bytes");
-        }
-        let sid: [u8; 8] = sid_bytes_vec.try_into().unwrap();
-
-        let (c_priv, c_pub) = generate_x25519_keypair();
-        let shared_secret = c_priv.diffie_hellman(&pbk);
-        let (auth_key, data_key) = derive_keys(shared_secret.as_bytes());
-
-        let session_id = generate_session_id(&auth_key, &sid);
-        let client_hello = build_client_hello(if sni.is_empty() { "www.microsoft.com" } else { sni }, &session_id, &c_pub);
-        
-        tcp_stream.write_all(&client_hello).await?;
-        
-        // Drain all server handshake records (ServerHello, CCS, fake encrypted records).
-        // The server sends exactly REALITY_SERVER_HANDSHAKE_RECORDS records before data starts.
-        // Reading them explicitly prevents RealityStream from seeing non-AppData bytes.
-        for i in 0..REALITY_SERVER_HANDSHAKE_RECORDS {
-            let mut head = [0u8; 5];
-            tcp_stream.read_exact(&mut head).await
-                .with_context(|| format!("reality handshake: failed reading record {} header", i))?;
-            if i == 0 && head[0] != 0x16 {
-                anyhow::bail!("expected ServerHello (0x16), got 0x{:02x}", head[0]);
-            }
-            let record_len = u16::from_be_bytes([head[3], head[4]]) as usize;
-            if record_len > 16384 {
-                anyhow::bail!("reality handshake: record {} too large: {} bytes", i, record_len);
-            }
-            let mut _payload = vec![0u8; record_len];
-            tcp_stream.read_exact(&mut _payload).await
-                .with_context(|| format!("reality handshake: failed reading record {} payload", i))?;
-        }
-
-        let reality_stream = RealityStream::new(tcp_stream, data_key);
-        xhttp_handshake_and_loop(reality_stream, target_ip, sni, access_key, wss).await
-    } else {
-        xhttp_handshake_and_loop(tcp_stream, target_ip, sni, access_key, wss).await
-    }
-}
-
-// -----------------------------------------------------------------------
-// RealityStream: Wraps a TCP stream in fake TLS Application Data Records
-// -----------------------------------------------------------------------
-struct RealityStream {
-    inner: TcpStream,
-    data_key: ChaCha20Poly1305,
-    rx_nonce: u64,
-    tx_nonce: u64,
-    rx_buf: BytesMut,
-    plaintext_buf: BytesMut,
-    tx_buf: BytesMut,
-}
-
-impl RealityStream {
-    fn new(inner: TcpStream, data_key: ChaCha20Poly1305) -> Self {
-        Self {
-            inner,
-            data_key,
-            rx_nonce: 0,
-            tx_nonce: 0,
-            rx_buf: BytesMut::with_capacity(16384),
-            plaintext_buf: BytesMut::new(),
-            tx_buf: BytesMut::new(),
-        }
-    }
-    
-    fn make_nonce(seq: u64) -> [u8; 12] {
-        let mut nonce = [0u8; 12];
-        nonce[4..12].copy_from_slice(&seq.to_le_bytes());
-        nonce
-    }
-}
-
-impl tokio::io::AsyncRead for RealityStream {
-    fn poll_read(mut self: Pin<&mut Self>, cx: &mut TaskContext<'_>, buf: &mut tokio::io::ReadBuf<'_>) -> Poll<std::io::Result<()>> {
-        loop {
-            if !self.plaintext_buf.is_empty() {
-                let out_len = std::cmp::min(buf.remaining(), self.plaintext_buf.len());
-                buf.put_slice(&self.plaintext_buf[..out_len]);
-                self.plaintext_buf.advance(out_len);
-                return Poll::Ready(Ok(()));
-            }
-
-            if self.rx_buf.len() >= 5 {
-                let len = u16::from_be_bytes([self.rx_buf[3], self.rx_buf[4]]) as usize;
-                if self.rx_buf.len() >= 5 + len {
-                    if self.rx_buf[0] != 0x17 {
-                        return Poll::Ready(Err(std::io::Error::new(std::io::ErrorKind::InvalidData, "expected application data record")));
-                    }
-                    
-                    let ciphertext = &self.rx_buf[5..5+len];
-                    let nonce_bytes = Self::make_nonce(self.rx_nonce);
-                    let nonce = Nonce::from_slice(&nonce_bytes);
-                    
-                    match self.data_key.decrypt(nonce, ciphertext) {
-                        Ok(plaintext) => {
-                            self.rx_nonce += 1;
-                            self.plaintext_buf.put_slice(&plaintext);
-                            self.rx_buf.advance(5 + len);
-                            continue;
-                        }
-                        Err(_) => {
-                            return Poll::Ready(Err(std::io::Error::new(std::io::ErrorKind::InvalidData, "reality decrypt failed")));
-                        }
-                    }
-                }
-            }
-            
-            let mut read_buf = [0u8; 8192];
-            let mut tokio_buf = tokio::io::ReadBuf::new(&mut read_buf);
-            match Pin::new(&mut self.inner).poll_read(cx, &mut tokio_buf) {
-                Poll::Ready(Ok(())) => {
-                    if tokio_buf.filled().is_empty() {
-                        return Poll::Ready(Ok(()));
-                    }
-                    self.rx_buf.put_slice(tokio_buf.filled());
-                }
-                Poll::Ready(Err(e)) => return Poll::Ready(Err(e)),
-                Poll::Pending => return Poll::Pending,
-            }
-        }
-    }
-}
-
-impl tokio::io::AsyncWrite for RealityStream {
-    fn poll_write(self: Pin<&mut Self>, cx: &mut TaskContext<'_>, buf: &[u8]) -> Poll<std::io::Result<usize>> {
-        let this = self.get_mut();
-        while !this.tx_buf.is_empty() {
-            match Pin::new(&mut this.inner).poll_write(cx, &this.tx_buf) {
-                Poll::Ready(Ok(n)) => this.tx_buf.advance(n),
-                Poll::Ready(Err(e)) => return Poll::Ready(Err(e)),
-                Poll::Pending => return Poll::Pending,
-            }
-        }
-
-        let nonce_bytes = Self::make_nonce(this.tx_nonce);
-        let nonce = Nonce::from_slice(&nonce_bytes);
-        
-        match this.data_key.encrypt(nonce, buf) {
-            Ok(ciphertext) => {
-                this.tx_nonce += 1;
-                this.tx_buf.reserve(5 + ciphertext.len());
-                this.tx_buf.put_u8(0x17);
-                this.tx_buf.put_u16(0x0303);
-                this.tx_buf.put_u16(ciphertext.len() as u16);
-                this.tx_buf.put_slice(&ciphertext);
-                
-                match Pin::new(&mut this.inner).poll_write(cx, &this.tx_buf) {
-                    Poll::Ready(Ok(n)) => this.tx_buf.advance(n),
-                    Poll::Ready(Err(e)) => return Poll::Ready(Err(e)),
-                    Poll::Pending => {}
-                }
-                Poll::Ready(Ok(buf.len()))
-            }
-            Err(_) => Poll::Ready(Err(std::io::Error::new(std::io::ErrorKind::Other, "reality encrypt failed"))),
-        }
-    }
-
-    fn poll_flush(self: Pin<&mut Self>, cx: &mut TaskContext<'_>) -> Poll<std::io::Result<()>> {
-        let this = self.get_mut();
-        while !this.tx_buf.is_empty() {
-            match Pin::new(&mut this.inner).poll_write(cx, &this.tx_buf) {
-                Poll::Ready(Ok(n)) => this.tx_buf.advance(n),
-                Poll::Ready(Err(e)) => return Poll::Ready(Err(e)),
-                Poll::Pending => return Poll::Pending,
-            }
-        }
-        Pin::new(&mut this.inner).poll_flush(cx)
-    }
-
-    fn poll_shutdown(self: Pin<&mut Self>, cx: &mut TaskContext<'_>) -> Poll<std::io::Result<()>> {
-        let this = self.get_mut();
-        while !this.tx_buf.is_empty() {
-            match Pin::new(&mut this.inner).poll_write(cx, &this.tx_buf) {
-                Poll::Ready(Ok(n)) => this.tx_buf.advance(n),
-                Poll::Ready(Err(e)) => return Poll::Ready(Err(e)),
-                Poll::Pending => return Poll::Pending,
-            }
-        }
-        Pin::new(&mut this.inner).poll_shutdown(cx)
-    }
-}
-
-async fn xhttp_handshake_and_loop<S>(
-    mut stream: S,
-    target_ip: IpAddr,
-    sni: &str,
-    access_key: &[u8],
-    wss: bool,
-) -> Result<(mpsc::Sender<Bytes>, Arc<tokio::sync::Mutex<mpsc::Receiver<Bytes>>>)>
-where
-    S: tokio::io::AsyncRead + tokio::io::AsyncWrite + Unpin + Send + 'static,
-{
-    // 1. Generate auth token: [8-byte timestamp BE] ++ [HMAC-SHA256]
-    let timestamp = std::time::SystemTime::now().duration_since(std::time::UNIX_EPOCH)?.as_secs();
-    let ts_bytes = timestamp.to_be_bytes();
-    use hmac::Mac;
-    let mut mac = <HmacSha256 as Mac>::new_from_slice(access_key).unwrap_or_else(|_| <HmacSha256 as Mac>::new_from_slice(b"").unwrap());
-    mac.update(&ts_bytes);
-    let mac_bytes = mac.finalize().into_bytes();
-
-    let mut sig_bytes = Vec::with_capacity(8 + mac_bytes.len());
-    sig_bytes.extend_from_slice(&ts_bytes);
-    sig_bytes.extend_from_slice(&mac_bytes);
-
-    let auth_token = base64::engine::general_purpose::STANDARD_NO_PAD.encode(&sig_bytes);
-
-    let http_host = if sni.is_empty() { target_ip.to_string() } else { sni.to_string() };
-
-    let req = if wss {
-        format!(
-            "GET /wss HTTP/1.1\r\n\
-             Host: {}\r\n\
-             Upgrade: websocket\r\n\
-             Connection: upgrade\r\n\
-             Sec-WebSocket-Key: dGhlIHNhbXBsZSBub25jZQ==\r\n\
-             Sec-WebSocket-Version: 13\r\n\
-             Authorization: Bearer {}\r\n\
-             \r\n",
-            http_host, auth_token
-        )
-    } else {
-        format!(
-            "GET /stream HTTP/1.1\r\n\
-             Host: {}\r\n\
-             Authorization: Bearer {}\r\n\
-             \r\n",
-            http_host, auth_token
-        )
-    };
-
-    stream.write_all(req.as_bytes()).await?;
-
-    // Wait for HTTP 200 OK or 101 Switching Protocols
-    let mut header_buf = Vec::new();
-    let mut temp = [0u8; 1];
-    loop {
-        let n = stream.read(&mut temp).await?;
-        if n == 0 {
-            anyhow::bail!("connection closed by server during handshake");
-        }
-        header_buf.push(temp[0]);
-        if header_buf.ends_with(b"\r\n\r\n") {
-            break;
-        }
-        if header_buf.len() > 8192 {
-            anyhow::bail!("server response too long");
-        }
-    }
-
-    let resp_str = String::from_utf8_lossy(&header_buf);
-    if wss {
-        if !resp_str.starts_with("HTTP/1.1 101 ") {
-            anyhow::bail!("failed to switch protocols: {}", resp_str.lines().next().unwrap_or(""));
-        }
-    } else {
-        if !resp_str.starts_with("HTTP/1.1 200 OK") {
-            anyhow::bail!("server rejected stream: {}", resp_str.lines().next().unwrap_or(""));
-        }
-    }
-
-    let (tx, mut rx) = mpsc::channel::<Bytes>(16384);
-    let (mut read_half, mut write_half) = tokio::io::split(stream);
-
-    let writer_task = tokio::spawn(async move {
-        while let Some(packet) = rx.recv().await {
-            if wss {
-                let header = encode_wss_frame(&packet, true);
-                if write_half.write_all(&header).await.is_err() { break; }
-            } else {
-                let mut out = BytesMut::with_capacity(2 + packet.len());
-                out.put_u16(packet.len() as u16);
-                out.put_slice(&packet);
-                if write_half.write_all(&out).await.is_err() { break; }
-            }
-        }
-    });
-
-    let (in_tx, in_rx) = mpsc::channel::<Bytes>(16384);
-    let in_rx_arc = Arc::new(tokio::sync::Mutex::new(in_rx));
-
-    let in_tx_clone = in_tx.clone();
-    let reader_task = tokio::spawn(async move {
-        if wss {
-            let mut read_buf = BytesMut::with_capacity(65536);
-            let mut tmp = [0u8; 8192];
-            loop {
-                match read_half.read(&mut tmp).await {
-                    Ok(0) => break,
-                    Ok(n) => {
-                        read_buf.put_slice(&tmp[..n]);
-                        loop {
-                            match decode_wss_frame(&mut read_buf) {
-                                WssFrameResult::Frame { payload, total_len } => {
-                                    if in_tx_clone.send(Bytes::from(payload)).await.is_err() { return; }
-                                    read_buf.advance(total_len);
-                                }
-                                WssFrameResult::Incomplete => break,
-                            }
-                        }
-                    }
-                    Err(_) => break,
-                }
-            }
-        } else {
-            let mut len_buf = [0u8; 2];
-            loop {
-                if read_half.read_exact(&mut len_buf).await.is_err() { break; }
-                let len = u16::from_be_bytes(len_buf) as usize;
-                if len > 65535 { break; }
-                let mut data = vec![0u8; len];
-                if read_half.read_exact(&mut data).await.is_err() { break; }
-                if in_tx_clone.send(Bytes::from(data)).await.is_err() { break; }
-            }
-        }
-    });
-
-    tokio::spawn(async move {
-        let _ = tokio::join!(writer_task, reader_task);
-    });
-
-    Ok((tx, in_rx_arc))
-}

ostp-client/src/tunnel/mod.rs

@@ -1,12 +1,14 @@
 mod proxy;
 pub mod native_handler;
+
 mod udp_nat;
 
 pub async fn run_tun_tunnel(
     config: crate::config::ClientConfig,
-    shutdown: watch::Receiver<bool>,
+    shutdown: tokio::sync::watch::Receiver<bool>,
+    exclusions_rx: tokio::sync::watch::Receiver<crate::config::ExclusionConfig>,
 ) -> anyhow::Result<()> {
-    native_handler::run_native_tunnel(config, shutdown).await
+    native_handler::run_native_tunnel(config, shutdown, exclusions_rx).await
 }
 
 use tokio::sync::{mpsc, watch};
@@ -51,17 +53,15 @@ pub enum ProxyToClientMsg {
 pub async fn run_local_proxy(
     cfg: LocalProxyConfig,
     ostp: OstpConfig,
-    exclusions: ExclusionConfig,
+    exclusions_rx: watch::Receiver<ExclusionConfig>,
     debug: bool,
     shutdown: watch::Receiver<bool>,
     proxy_events_tx: mpsc::Sender<ProxyEvent>,
     client_msgs_rx: mpsc::UnboundedReceiver<(u16, ProxyToClientMsg)>,
 ) -> anyhow::Result<()> {
-    run_local_socks5_proxy(cfg, ostp, exclusions, debug, shutdown, proxy_events_tx, client_msgs_rx).await
+    run_local_socks5_proxy(cfg, ostp, exclusions_rx, debug, shutdown, proxy_events_tx, client_msgs_rx).await
 }
 
-
-
 pub mod exclusion;
 pub mod process_lookup;
 pub mod sni_sniff;

ostp-client/src/tunnel/native_handler.rs

@@ -1,20 +1,21 @@
 use anyhow::{anyhow, Result};
 use tokio::sync::watch;
 
+// ──────────────────────────────────────────────────────────────────────────────
+// Windows / Linux desktop TUN
+// ──────────────────────────────────────────────────────────────────────────────
+
 #[cfg(any(target_os = "windows", target_os = "linux"))]
 pub async fn run_native_tunnel(
     config: crate::config::ClientConfig,
     mut shutdown: watch::Receiver<bool>,
+    mut exclusions_rx: watch::Receiver<crate::config::ExclusionConfig>,
 ) -> Result<()> {
     use std::net::ToSocketAddrs;
-    use std::process::Command;
     use netstack_smoltcp::StackBuilder;
     use tokio::io::{AsyncReadExt, AsyncWriteExt};
     use futures::{StreamExt, SinkExt};
 
-    #[cfg(target_os = "windows")]
-    use std::os::windows::process::CommandExt;
-
     #[cfg(target_os = "linux")]
     {
         use std::io::{self, IsTerminal, Write};
@@ -26,12 +27,12 @@ pub async fn run_native_tunnel(
             println!("===================================================================\n");
             print!("Are you sure you want to initialize the TUN interface? [yes/no]: ");
             io::stdout().flush().unwrap();
-            
+
             let mut input = String::new();
             io::stdin().read_line(&mut input).unwrap();
             let ans = input.trim().to_lowercase();
             if ans != "y" && ans != "yes" {
-                return Err(anyhow!("TUN initialization aborted by user. Run without TUN to use as a local proxy."));
+                return Err(anyhow!("TUN initialization aborted by user."));
             }
         }
     }
@@ -39,111 +40,73 @@ pub async fn run_native_tunnel(
     let debug = config.debug;
     tracing::info!("Initializing NATIVE TUN tunnel (smoltcp)...");
 
-    let server_ip = config.ostp.server_addr.to_socket_addrs()
-        .map_err(|e| anyhow!("Failed to resolve remote server IP: {}", e))?
+    // Capture physical interface index for bypass BEFORE we create the TUN device and alter routes.
+    #[cfg(target_os = "windows")]
+    let phys_if_for_bypass: Option<u32> = ostp_tun::windows::windows_route::sys::get_default_ipv4_route().map(|(_, idx)| idx);
+    #[cfg(not(target_os = "windows"))]
+    let phys_if_for_bypass: Option<u32> = None;
+
+    // ── 1. Resolve server IP ──────────────────────────────────────────────────
+    let server_ip = config
+        .ostp
+        .server_addr
+        .to_socket_addrs()
+        .map_err(|e| anyhow!("Failed to resolve server IP: {}", e))?
         .next()
-        .map(|addr| addr.ip())
-        .ok_or_else(|| anyhow!("Could not resolve host IP for routing exclusion"))?;
-    
+        .map(|a| a.ip())
+        .ok_or_else(|| anyhow!("Could not resolve server host"))?;
+    #[allow(unused_variables)]
     let server_ip_str = server_ip.to_string();
 
-    let mut tun_cfg = tun::Configuration::default();
-    tun_cfg.tun_name("ostp_tun")
-           .address((10, 1, 0, 2))
-           .netmask((255, 255, 255, 0))
-           .destination((10, 1, 0, 1))
-           .mtu(config.ostp.mtu as u16)
-           .up();
+    // ── 2. Resolve excluded domains → IP addresses for bypass routing ─────────
+    let mut bypass_ips: Vec<std::net::IpAddr> = Vec::new();
 
-    #[cfg(target_os = "linux")]
-    tun_cfg.platform_config(|config| {
-        config.packet_information(false);
-    });
+    // Server IP always bypasses TUN
+    bypass_ips.push(server_ip);
 
-    let dev = tun::create(&tun_cfg)
-        .map_err(|e| anyhow!("Failed to create TUN device: {}", e))?;
-    let dev = tun::AsyncDevice::new(dev)
-        .map_err(|e| anyhow!("Failed to make TUN device async: {}", e))?;
+    for ip_str in &config.exclusions.ips {
+        let host = ip_str.split('/').next().unwrap_or(ip_str);
+        if let Ok(ip) = host.parse() {
+            bypass_ips.push(ip);
+        }
+    }
 
-    tracing::info!("TUN device created natively.");
-
-    #[cfg(target_os = "windows")]
-    {
-        const CREATE_NO_WINDOW: u32 = 0x08000000;
-        let current_exe = std::env::current_exe()?.to_string_lossy().into_owned();
-
-        let setup_script = format!(
-            "$remote_ip = '{}'\n\
-            $exe_path = '{}'\n\
-            $route = Get-NetRoute -DestinationPrefix '0.0.0.0/0' | Where-Object {{ $_.InterfaceAlias -notmatch 'ostp' -and $_.InterfaceAlias -notmatch 'tun' -and $_.InterfaceAlias -notmatch 'wintun' }} | Sort-Object RouteMetric | Select-Object -First 1\n\
-            if ($route) {{\n\
-                $gw = $route.NextHop\n\
-                $ifIndex = $route.InterfaceIndex\n\
-                if ($gw -eq '0.0.0.0' -or $gw -eq '::') {{\n\
-                    New-NetRoute -DestinationPrefix \"$remote_ip/32\" -InterfaceIndex $ifIndex -RouteMetric 1 -ErrorAction SilentlyContinue\n\
-                }} else {{\n\
-                    New-NetRoute -DestinationPrefix \"$remote_ip/32\" -NextHop $gw -InterfaceIndex $ifIndex -RouteMetric 1 -ErrorAction SilentlyContinue\n\
-                }}\n\
-                if ($gw -ne '0.0.0.0') {{\n\
-                    New-NetRoute -DestinationPrefix \"$gw/32\" -NextHop '0.0.0.0' -InterfaceIndex $ifIndex -RouteMetric 1 -ErrorAction SilentlyContinue\n\
-                }}\n\
-            }}\n\
-            New-NetFirewallRule -DisplayName 'OSTP Tunnel In' -Direction Inbound -Program $exe_path -Action Allow -Enabled True -ErrorAction SilentlyContinue\n\
-            New-NetFirewallRule -DisplayName 'OSTP Tunnel Out' -Direction Outbound -Program $exe_path -Action Allow -Enabled True -ErrorAction SilentlyContinue\n\
-            netsh interface ipv4 set interface name=\"ostp_tun\" metric=1\n\
-            New-NetRoute -DestinationPrefix '0.0.0.0/0' -InterfaceAlias 'ostp_tun' -NextHop '10.1.0.1' -RouteMetric 1 -ErrorAction SilentlyContinue\n",
-            server_ip_str, current_exe
-        );
-        let _ = tokio::task::spawn_blocking(move || {
-            Command::new("powershell")
-                .creation_flags(CREATE_NO_WINDOW)
-                .args(["-NoProfile", "-Command", &setup_script])
-                .output()
-        }).await.unwrap()?;
-        
-        if let Some(ref dns) = config.dns_server {
-            if !dns.is_empty() {
-                let net_setup = format!("netsh interface ipv4 set dnsservers name=\"ostp_tun\" static {} primary\n", dns);
-                let _ = tokio::task::spawn_blocking(move || {
-                    Command::new("powershell")
-                        .creation_flags(CREATE_NO_WINDOW)
-                        .args(["-NoProfile", "-Command", &net_setup])
-                        .output()
-                }).await.unwrap()?;
+    for domain in &config.exclusions.domains {
+        match tokio::net::lookup_host((domain.as_str(), 443u16)).await {
+            Ok(addrs) => {
+                for addr in addrs {
+                    bypass_ips.push(addr.ip());
+                }
+            }
+            Err(e) => {
+                tracing::warn!("Failed to pre-resolve excluded domain {domain}: {e}");
             }
         }
     }
 
-    #[cfg(target_os = "linux")]
-    {
-        // Get real gateway before routing through TUN
-        let gw_out = Command::new("ip")
-            .args(["route", "show", "default"])
-            .output()
-            .ok()
-            .and_then(|o| String::from_utf8(o.stdout).ok());
+
+    // ── 3. Create TUN device via ostp-tun crate ───────────────────────────────
+    let opts = ostp_tun::OstpTunOptions {
+        server_ip,
+        bypass_ips,
+        dns_server: config.dns_server.clone(),
+        kill_switch: config.kill_switch,
+        mtu: config.ostp.mtu as u16,
+        wintun_path: None,
+    };
+
+    let tun_interface = ostp_tun::OstpTunInterface::create(opts)
+        .await
+        .map_err(|e| anyhow!("Failed to create OstpTunInterface: {}", e))?;
         
-        let real_gw = gw_out.as_deref().and_then(|s| {
-            // "default via 192.168.1.1 dev eth0" -> "192.168.1.1"
-            s.split_whitespace().skip_while(|w| *w != "via").nth(1).map(|s| s.to_string())
-        });
-        let real_dev = gw_out.as_deref().and_then(|s| {
-            s.split_whitespace().skip_while(|w| *w != "dev").nth(1).map(|s| s.to_string())
-        });
-
-        // Add exclusion route for server IP via real gateway (bypass TUN)
-        if let (Some(ref gw), Some(ref dev)) = (&real_gw, &real_dev) {
-            let _ = Command::new("ip").args(["route", "add", &format!("{}/32", server_ip_str), "via", gw, "dev", dev]).output();
-        }
-
-        // Add default route through TUN (lower metric to take priority)
-        let _ = Command::new("ip").args(["route", "add", "default", "via", "10.1.0.1", "dev", "ostp_tun", "metric", "10"]).output();
-    }
+    let dev = tun_interface.device;
+    let _route_guard = tun_interface.guard;
 
+    // ── 7. Build smoltcp network stack ────────────────────────────────────────
     let (stack, tcp_runner, udp_socket, tcp_listener) = StackBuilder::default()
-        .stack_buffer_size(100000)
-        .tcp_buffer_size(100000)
-        .udp_buffer_size(100000)
+        .stack_buffer_size(1024)
+        .tcp_buffer_size(1024)
+        .udp_buffer_size(1024)
         .enable_tcp(true)
         .enable_udp(true)
         .mtu(config.ostp.mtu)
@@ -155,6 +118,7 @@ pub async fn run_native_tunnel(
         }
     });
 
+    // ── 8. Wire TUN ↔ smoltcp stack ───────────────────────────────────────────
     let (mut stack_sink, mut stack_stream) = stack.split();
     let (mut tun_read, mut tun_write) = tokio::io::split(dev);
 
@@ -172,8 +136,7 @@ pub async fn run_native_tunnel(
                     }
                 }
                 Err(e) => {
-                    tracing::debug!("tun_read error: {}", e);
-                    // continue reading
+                    tracing::debug!("tun_read error: {e}");
                 }
             }
         }
@@ -182,55 +145,245 @@ pub async fn run_native_tunnel(
     let mut stack_to_tun = tokio::spawn(async move {
         while let Some(Ok(frame)) = stack_stream.next().await {
             if let Err(e) = tun_write.write(&frame).await {
-                tracing::debug!("tun_write error: {}", e);
+                tracing::debug!("tun_write error: {e}");
             }
         }
     });
 
-    let udp_proxy_addr = config.local_proxy.bind_addr.clone();
-    let debug_udp = config.debug;
+    // ── 9. UDP: forward everything through OSTP proxy ─────────────────────────
+    //   UDP exclusions are handled at the routing table level (step 5), so
+    //   UDP packets for excluded IPs never reach smoltcp at all.
+    let udp_proxy_addr = {
+        let mut a = config.local_proxy.bind_addr.clone();
+        if a.starts_with("0.0.0.0:") {
+            a = a.replace("0.0.0.0:", "127.0.0.1:");
+        }
+        a
+    };
+    // Build exclusion matcher for dynamic bypass
+    let current_exclusions = exclusions_rx.borrow().clone();
+    let matcher = crate::tunnel::exclusion::ExclusionMatcher::new(&current_exclusions, None, None);
+    let matcher_arc = std::sync::Arc::new(tokio::sync::RwLock::new(matcher));
+    
+    let matcher_clone = matcher_arc.clone();
+    tokio::spawn(async move {
+        while let Ok(_) = exclusions_rx.changed().await {
+            let current = exclusions_rx.borrow().clone();
+            let new_matcher = crate::tunnel::exclusion::ExclusionMatcher::new(&current, None, None);
+            *matcher_clone.write().await = new_matcher;
+            if true {
+                tracing::debug!("Desktop TUN exclusions hot-reloaded");
+            }
+        }
+    });
+
+    // Linux: physical interface name for SO_BINDTODEVICE
+    #[cfg(target_os = "linux")]
+    let linux_phys_name = crate::tunnel::proxy::get_linux_physical_if_name();
+    #[cfg(not(target_os = "linux"))]
+    let linux_phys_name: Option<String> = None;
+    let _ = &linux_phys_name; // suppress unused warning on Windows
+
+    let debug_udp = debug;
+    let udp_matcher = matcher_arc.clone();
+    #[cfg(target_os = "linux")]
+    let udp_lin_name = linux_phys_name.clone();
+
     let mut udp_proxy_task = tokio::spawn(async move {
         if let Some(udp_sock) = udp_socket {
-            super::udp_nat::run_udp_nat(udp_sock, udp_proxy_addr, debug_udp).await;
+            #[cfg(target_os = "linux")]
+            super::udp_nat::run_udp_nat(udp_sock, udp_proxy_addr, debug_udp, udp_matcher, phys_if_for_bypass, udp_lin_name).await;
+            #[cfg(not(target_os = "linux"))]
+            super::udp_nat::run_udp_nat(udp_sock, udp_proxy_addr, debug_udp, udp_matcher, phys_if_for_bypass, None).await;
         }
     });
 
-    let proxy_addr = config.local_proxy.bind_addr.clone();
+    // ── 10. TCP: forward to OSTP proxy (with domain-level bypass via SNI) ─────
+    //
+    //   For IP-based exclusions:  handled by routing table → packets never arrive here.
+    //   For domain-based exclusions: The IP is already in routing table (pre-resolved in
+    //     step 3), so most traffic won't arrive. As a belt-and-suspenders fallback,
+    //     we also sniff TLS SNI and bypass if it matches — this covers CDN cases where
+    //     the IP wasn't known at startup.
+    //
+    //   For bypassed connections we bind the outgoing socket to the physical interface
+    //   (IP_UNICAST_IF) so it goes out via the real NIC, not TUN.
+
+    let proxy_addr_tcp = {
+        let mut a = config.local_proxy.bind_addr.clone();
+        if a.starts_with("0.0.0.0:") {
+            a = a.replace("0.0.0.0:", "127.0.0.1:");
+        }
+        a
+    };
+
+    // Physical interface index was captured at the start of the function.
+
     let mut tcp_accept_task = tokio::spawn(async move {
-        if let Some(mut listener) = tcp_listener {
-            while let Some((mut stream, _local, remote)) = listener.next().await {
-                let proxy_addr = proxy_addr.clone();
-                tokio::spawn(async move {
-                    if debug { tracing::info!("Native TUN intercepted TCP to {}", remote); }
-                    if let Ok(mut socks) = tokio::net::TcpStream::connect(&proxy_addr).await {
-                        // SOCKS5 bypass handshake locally (loopback)
-                        if socks.write_all(&[5, 1, 0]).await.is_err() { return; }
-                        let mut buf = [0u8; 2];
-                        if socks.read_exact(&mut buf).await.is_err() || buf[0] != 5 || buf[1] != 0 { return; }
-                        
-                        let ip = remote.ip();
-                        let port = remote.port();
-                        let mut req = vec![5, 1, 0];
-                        match ip {
-                            std::net::IpAddr::V4(v4) => {
-                                req.push(1);
-                                req.extend_from_slice(&v4.octets());
+        let Some(mut listener) = tcp_listener else { return; };
+
+        while let Some((mut stream, local, remote)) = listener.next().await {
+            let proxy_addr = proxy_addr_tcp.clone();
+            let matcher_arc = matcher_arc.clone();
+            #[cfg(target_os = "linux")]
+            let lin_name = linux_phys_name.clone();
+
+            tokio::spawn(async move {
+                let matcher = matcher_arc.read().await.clone();
+                if debug {
+                    tracing::debug!("TUN TCP {local} → {remote}");
+                }
+
+                // ── Sniff TLS ClientHello for SNI ─────────────────────────────
+                let mut sniff_buf = [0u8; 2048];
+                let sniff_len =
+                    match tokio::time::timeout(
+                        std::time::Duration::from_millis(100),
+                        stream.read(&mut sniff_buf),
+                    )
+                    .await
+                    {
+                        Ok(Ok(n)) => n,
+                        _ => 0,
+                    };
+
+                // ── Decide: bypass or tunnel? ─────────────────────────────────
+                let mut should_bypass = false;
+
+                // 1. Process match via OS Extended TCP Table (Windows)
+                #[cfg(target_os = "windows")]
+                if !should_bypass {
+                    if let Some(proc_name) = crate::tunnel::process_lookup::get_process_name_from_port(local.port()) {
+                        if debug {
+                            tracing::debug!("TUN TCP lookup: port {} -> process {}", local.port(), proc_name);
+                        }
+                        if matcher.match_process(&proc_name) {
+                            if debug {
+                                tracing::debug!("TUN TCP BYPASS (Process match): {} → {remote}", proc_name);
                             }
-                            std::net::IpAddr::V6(v6) => {
-                                req.push(4);
-                                req.extend_from_slice(&v6.octets());
+                            should_bypass = true;
+                        }
+                    } else {
+                        if debug {
+                            tracing::debug!("TUN TCP lookup: port {} -> no process found", local.port());
+                        }
+                    }
+                }
+
+                // 2. SNI domain check (belt-and-suspenders for CDNs / late-resolved IPs)
+                if !should_bypass && sniff_len > 0 {
+                    if let Some(sni) =
+                        crate::tunnel::sni_sniff::extract_sni(&sniff_buf[..sniff_len])
+                    {
+                        if debug {
+                            tracing::debug!("TUN SNI: {sni}");
+                        }
+                        if matcher.match_domain(&sni) {
+                            if debug {
+                                tracing::info!("TUN TCP BYPASS (SNI domain): {sni} → {remote}");
+                            }
+                            should_bypass = true;
+                        }
+                    }
+                }
+
+                // 3. Destination IP CIDR check (for IPs not in routing table / IPv6)
+                if !should_bypass && matcher.match_ip(&remote.ip()) {
+                    if debug {
+                        tracing::info!("TUN TCP BYPASS (IP match): {remote}");
+                    }
+                    should_bypass = true;
+                }
+
+                // ── Bypass path: direct TCP bypassing TUN ─────────────────────
+                if should_bypass {
+                    let socket = match remote {
+                        std::net::SocketAddr::V4(_) => tokio::net::TcpSocket::new_v4(),
+                        std::net::SocketAddr::V6(_) => tokio::net::TcpSocket::new_v6(),
+                    };
+                    let Ok(socket) = socket else { return; };
+
+                    // Bind to physical interface so packets don't loop back into TUN
+
+                    #[cfg(target_os = "windows")]
+                    if let Some(idx) = phys_if_for_bypass {
+                        if let Err(e) = crate::tunnel::proxy::bind_socket_to_interface(
+                            &socket,
+                            remote.is_ipv6(),
+                            idx,
+                        ) {
+                            tracing::error!("TUN TCP BYPASS failed to bind to physical interface {}: {}", idx, e);
+                        } else {
+                            if debug {
+                                tracing::info!("TUN TCP BYPASS bound to physical interface {}", idx);
                             }
                         }
-                        req.extend_from_slice(&port.to_be_bytes());
-                        if socks.write_all(&req).await.is_err() { return; }
-                        
-                        let mut rep = [0u8; 10];
-                        if socks.read_exact(&mut rep).await.is_err() || rep[1] != 0 { return; }
-
-                        let _ = tokio::io::copy_bidirectional(&mut stream, &mut socks).await;
+                    } else {
+                        tracing::warn!("TUN TCP BYPASS has no physical interface index!");
                     }
-                });
-            }
+                    #[cfg(target_os = "linux")]
+                    if let Some(ref name) = lin_name {
+                        let _ = crate::tunnel::proxy::bind_socket_to_interface(&socket, name);
+                    }
+
+                    match tokio::time::timeout(
+                        std::time::Duration::from_secs(10),
+                        socket.connect(remote),
+                    )
+                    .await
+                    {
+                        Ok(Ok(mut direct)) => {
+                            if sniff_len > 0 {
+                                if direct.write_all(&sniff_buf[..sniff_len]).await.is_err() {
+                                    return;
+                                }
+                            }
+                            let _ = tokio::io::copy_bidirectional(&mut stream, &mut direct).await;
+                        }
+                        _ => {
+                            tracing::debug!("Direct bypass connect to {remote} failed");
+                        }
+                    }
+                    return;
+                }
+
+                // ── Tunnel path: forward via local OSTP SOCKS5 proxy ──────────
+                let Ok(mut socks) = tokio::net::TcpStream::connect(&proxy_addr).await else {
+                    return;
+                };
+
+                // SOCKS5 handshake (no auth)
+                if socks.write_all(&[5, 1, 0]).await.is_err() { return; }
+                let mut buf2 = [0u8; 2];
+                if socks.read_exact(&mut buf2).await.is_err() || buf2[0] != 5 || buf2[1] != 0 {
+                    return;
+                }
+
+                // CONNECT request
+                let mut req = vec![5u8, 1, 0];
+                match remote.ip() {
+                    std::net::IpAddr::V4(v4) => {
+                        req.push(1);
+                        req.extend_from_slice(&v4.octets());
+                    }
+                    std::net::IpAddr::V6(v6) => {
+                        req.push(4);
+                        req.extend_from_slice(&v6.octets());
+                    }
+                }
+                req.extend_from_slice(&remote.port().to_be_bytes());
+                if socks.write_all(&req).await.is_err() { return; }
+
+                let mut rep = [0u8; 10];
+                if socks.read_exact(&mut rep).await.is_err() || rep[1] != 0 { return; }
+
+                // Replay sniffed bytes
+                if sniff_len > 0 && socks.write_all(&sniff_buf[..sniff_len]).await.is_err() {
+                    return;
+                }
+
+                let _ = tokio::io::copy_bidirectional(&mut stream, &mut socks).await;
+            });
         }
     });
 
@@ -246,45 +399,35 @@ pub async fn run_native_tunnel(
     }
 
     tracing::info!("Deactivating NATIVE TUN tunnel...");
-    // Cleanup routes
-    #[cfg(target_os = "windows")]
-    {
-        const CREATE_NO_WINDOW: u32 = 0x08000000;
-        let cleanup_script = format!(
-            "$remote_ip = '{}'\n\
-             Remove-NetRoute -DestinationPrefix \"$remote_ip/32\" -Confirm:$false -ErrorAction SilentlyContinue\n\
-             Remove-NetFirewallRule -DisplayName 'OSTP Tunnel*' -ErrorAction SilentlyContinue\n\
-             netsh interface ipv4 set dnsservers name=\"ostp_tun\" source=dhcp 2>$null\n",
-            server_ip_str
-        );
-        let _ = Command::new("powershell")
-            .creation_flags(CREATE_NO_WINDOW)
-            .args(["-NoProfile", "-Command", &cleanup_script])
-            .output();
-    }
 
-    #[cfg(target_os = "linux")]
-    {
-        // Remove default route via TUN and server exclusion route
-        let _ = Command::new("ip").args(["route", "del", "default", "dev", "ostp_tun"]).output();
-        let _ = Command::new("ip").args(["route", "del", &format!("{}/32", server_ip_str)]).output();
-    }
+    // ── Cleanup ───────────────────────────────────────────────────────────────
+    // Cleanup is handled automatically by the _route_guard Drop trait in ostp-tun
 
     Ok(())
 }
 
+// ──────────────────────────────────────────────────────────────────────────────
+// Stub for unsupported platforms
+// ──────────────────────────────────────────────────────────────────────────────
+
 #[cfg(not(any(target_os = "windows", target_os = "linux")))]
 pub async fn run_native_tunnel(
     _config: crate::config::ClientConfig,
     _shutdown: watch::Receiver<bool>,
+    _exclusions_rx: watch::Receiver<crate::config::ExclusionConfig>,
 ) -> Result<()> {
-    Err(anyhow!("Native TUN tunnel is only supported on Windows/Linux currently"))
+    Err(anyhow!("Native TUN tunnel is only supported on Windows/Linux"))
 }
 
+// ──────────────────────────────────────────────────────────────────────────────
+// Android: TUN from file-descriptor (opened by VpnService)
+// ──────────────────────────────────────────────────────────────────────────────
+
 #[cfg(target_os = "android")]
 pub async fn run_native_tunnel_from_fd(
     config: crate::config::ClientConfig,
     mut shutdown: watch::Receiver<bool>,
+    mut exclusions_rx: watch::Receiver<crate::config::ExclusionConfig>,
     fd: i32,
 ) -> Result<()> {
     use netstack_smoltcp::StackBuilder;
@@ -304,16 +447,16 @@ pub async fn run_native_tunnel_from_fd(
 
     let read_fd = unsafe { libc::dup(fd) };
     if read_fd < 0 {
-        return Err(anyhow::anyhow!("Failed to dup tun fd for reading"));
+        return Err(anyhow!("Failed to dup tun fd for reading"));
     }
-    
+
     let file = unsafe { std::fs::File::from_raw_fd(read_fd) };
     let tun_stream = tokio::io::unix::AsyncFd::new(file)?;
 
     let (stack, tcp_runner, udp_socket, tcp_listener) = StackBuilder::default()
-        .stack_buffer_size(100000)
-        .tcp_buffer_size(100000)
-        .udp_buffer_size(100000)
+        .stack_buffer_size(1024)
+        .tcp_buffer_size(1024)
+        .udp_buffer_size(1024)
         .enable_tcp(true)
         .enable_udp(true)
         .mtu(config.ostp.mtu)
@@ -334,25 +477,29 @@ pub async fn run_native_tunnel_from_fd(
                 Ok(g) => g,
                 Err(_) => break,
             };
-
             let n = match guard.try_io(|inner| {
-                let res = unsafe { libc::read(inner.as_raw_fd(), buf.as_mut_ptr() as *mut libc::c_void, buf.len()) };
+                let res = unsafe {
+                    libc::read(
+                        inner.as_raw_fd(),
+                        buf.as_mut_ptr() as *mut libc::c_void,
+                        buf.len(),
+                    )
+                };
                 if res < 0 {
                     let err = std::io::Error::last_os_error();
                     if err.kind() == std::io::ErrorKind::WouldBlock {
                         Err(err)
                     } else {
-                        // EINTR or other transient error — treat as zero (will continue)
                         Ok(0_isize)
                     }
                 } else {
-                    Ok(res as isize)
+                    Ok(res)
                 }
             }) {
                 Ok(Ok(n)) if n > 0 => n as usize,
-                Ok(Ok(_)) => continue, // 0 = EINTR or transient error, try again
-                Ok(Err(_)) => continue, // WouldBlock retry
-                Err(_would_block) => continue,
+                Ok(Ok(_)) => continue,
+                Ok(Err(_)) => continue,
+                Err(_) => continue,
             };
 
             let frame = buf[..n].to_vec();
@@ -366,7 +513,7 @@ pub async fn run_native_tunnel_from_fd(
 
     let write_fd = unsafe { libc::dup(fd) };
     if write_fd < 0 {
-        return Err(anyhow!("Failed to dup tun fd"));
+        return Err(anyhow!("Failed to dup tun fd for writing"));
     }
     unsafe {
         let flags = libc::fcntl(write_fd, libc::F_GETFL);
@@ -385,9 +532,14 @@ pub async fn run_native_tunnel_from_fd(
                     Ok(g) => g,
                     Err(_) => break,
                 };
-                
                 let res = guard.try_io(|inner| {
-                    let res = unsafe { libc::write(inner.as_raw_fd(), frame[written..].as_ptr() as *const libc::c_void, frame.len() - written) };
+                    let res = unsafe {
+                        libc::write(
+                            inner.as_raw_fd(),
+                            frame[written..].as_ptr() as *const libc::c_void,
+                            frame.len() - written,
+                        )
+                    };
                     if res < 0 {
                         let err = std::io::Error::last_os_error();
                         if err.kind() == std::io::ErrorKind::WouldBlock {
@@ -399,11 +551,9 @@ pub async fn run_native_tunnel_from_fd(
                         Ok(res)
                     }
                 });
-
                 match res {
                     Ok(Ok(n)) if n > 0 => written += n as usize,
-                    Ok(Ok(n)) if n == 0 => break,
-                    Ok(Ok(_)) => break, // n < 0, error writing, drop this frame
+                    Ok(Ok(_)) => break,
                     Ok(Err(_)) => break,
                     Err(_) => continue,
                 }
@@ -411,111 +561,160 @@ pub async fn run_native_tunnel_from_fd(
         }
     });
 
-
-
     let mut proxy_addr = config.local_proxy.bind_addr.clone();
     if proxy_addr.starts_with("0.0.0.0:") {
         proxy_addr = proxy_addr.replace("0.0.0.0:", "127.0.0.1:");
     }
 
-    let udp_proxy_addr = proxy_addr.clone();
-    let debug_udp = config.debug;
-    let mut udp_proxy_task = tokio::spawn(async move {
-        if let Some(udp_sock) = udp_socket {
-            super::udp_nat::run_udp_nat(udp_sock, udp_proxy_addr, debug_udp).await;
+    let current_exclusions = exclusions_rx.borrow().clone();
+    let matcher = crate::tunnel::exclusion::ExclusionMatcher::new(&current_exclusions, None, None);
+    let matcher_arc = std::sync::Arc::new(tokio::sync::RwLock::new(matcher));
+    
+    let matcher_clone = matcher_arc.clone();
+    tokio::spawn(async move {
+        while let Ok(_) = exclusions_rx.changed().await {
+            let current = exclusions_rx.borrow().clone();
+            let new_matcher = crate::tunnel::exclusion::ExclusionMatcher::new(&current, None, None);
+            *matcher_clone.write().await = new_matcher;
+            if true {
+                tracing::debug!("Android TUN exclusions hot-reloaded");
+            }
         }
     });
 
-    let matcher = crate::tunnel::exclusion::ExclusionMatcher::new(&config.exclusions, None, None);
+    let udp_proxy_addr = proxy_addr.clone();
+    let debug_udp = debug;
+    let udp_matcher = matcher_arc.clone();
+    let mut udp_proxy_task = tokio::spawn(async move {
+        if let Some(udp_sock) = udp_socket {
+            super::udp_nat::run_udp_nat(udp_sock, udp_proxy_addr, debug_udp, udp_matcher, None, None).await;
+        }
+    });
+
+
 
     let mut tcp_accept_task = tokio::spawn(async move {
-        if let Some(mut listener) = tcp_listener {
-            while let Some((mut stream, local, remote)) = listener.next().await {
-                let proxy_addr = proxy_addr.clone();
-                let matcher = matcher.clone();
-                tokio::spawn(async move {
-                    if debug { tracing::info!("Native TUN intercepted TCP {local} -> {remote}"); }
+        let Some(mut listener) = tcp_listener else { return; };
 
-                    // Peak first chunk to see SNI
-                    let mut sniff_buf = [0u8; 1500];
-                    let sniff_len = match tokio::time::timeout(std::time::Duration::from_millis(50), stream.read(&mut sniff_buf)).await {
+        while let Some((mut stream, local, remote)) = listener.next().await {
+            let proxy_addr = proxy_addr.clone();
+            let matcher_arc = matcher_arc.clone();
+
+            tokio::spawn(async move {
+                let matcher = matcher_arc.read().await.clone();
+
+                if true {
+                    tracing::debug!("Android TUN TCP {local} → {remote}");
+                }
+
+                // Sniff SNI
+                let mut sniff_buf = [0u8; 2048];
+                let sniff_len =
+                    match tokio::time::timeout(
+                        std::time::Duration::from_millis(100),
+                        stream.read(&mut sniff_buf),
+                    )
+                    .await
+                    {
                         Ok(Ok(n)) => n,
-                        _ => 0, // Timeout or error
+                        _ => 0,
                     };
 
-                    let mut should_bypass = false;
+                let mut should_bypass = false;
 
-                    // 1. Check SNI
-                    if sniff_len > 0 {
-                        if let Some(sni) = crate::tunnel::sni_sniff::extract_sni(&sniff_buf[..sniff_len]) {
-                            if debug { tracing::info!("Native TUN sniffed SNI: {}", sni); }
-                            if matcher.match_domain(&sni) {
-                                should_bypass = true;
-                            }
-                        }
-                    }
-
-                    // 2. Check Process
-                    if !should_bypass {
-                        if let Some(exe) = crate::tunnel::process_lookup::get_process_name_from_port(local.port()) {
-                            if debug { tracing::info!("Native TUN source port {} maps to EXE: {}", local.port(), exe); }
-                            if matcher.match_process(&exe) {
-                                should_bypass = true;
-                            }
-                        }
-                    }
-
-                    // 3. Check Target IP
-                    if !should_bypass {
-                        if matcher.match_ip(&remote.ip()) {
+                // 1. SNI domain
+                if sniff_len > 0 {
+                    if let Some(sni) =
+                        crate::tunnel::sni_sniff::extract_sni(&sniff_buf[..sniff_len])
+                    {
+                        if true { tracing::debug!("Android TUN SNI: {sni}"); }
+                        if matcher.match_domain(&sni) {
                             should_bypass = true;
                         }
                     }
+                }
 
-                    if should_bypass {
-                        if debug { tracing::info!("Native TUN BYPASS matched for {}", remote); }
-                        if let Ok(mut direct) = tokio::time::timeout(std::time::Duration::from_secs(5), tokio::net::TcpStream::connect(remote)).await.unwrap_or(Err(std::io::Error::new(std::io::ErrorKind::TimedOut, "Direct connect timeout"))) {
+                // 2. Process (Android: /proc/net lookup)
+                if !should_bypass {
+                    if let Some(exe) =
+                        crate::tunnel::process_lookup::get_process_name_from_port(local.port())
+                    {
+                        if true {
+                            tracing::debug!("Android TUN port {} → EXE: {}", local.port(), exe);
+                        }
+                        if matcher.match_process(&exe) {
+                            should_bypass = true;
+                        }
+                    }
+                }
+
+                // 3. IP CIDR
+                if !should_bypass && matcher.match_ip(&remote.ip()) {
+                    should_bypass = true;
+                }
+
+                // Bypass: connect directly (Android VPN service already protects the socket
+                // from re-entering the TUN through VpnService.protect())
+                if should_bypass {
+                    if true {
+                        tracing::debug!("Android TUN BYPASS: {remote}");
+                    }
+                    let socket = match remote {
+                        std::net::SocketAddr::V4(_) => tokio::net::TcpSocket::new_v4(),
+                        std::net::SocketAddr::V6(_) => tokio::net::TcpSocket::new_v6(),
+                    };
+                    let Ok(socket) = socket else { return; };
+
+                    match tokio::time::timeout(
+                        std::time::Duration::from_secs(10),
+                        socket.connect(remote),
+                    )
+                    .await
+                    {
+                        Ok(Ok(mut direct)) => {
                             if sniff_len > 0 {
-                                let _ = direct.write_all(&sniff_buf[..sniff_len]).await;
+                                if direct.write_all(&sniff_buf[..sniff_len]).await.is_err() {
+                                    return;
+                                }
                             }
                             let _ = tokio::io::copy_bidirectional(&mut stream, &mut direct).await;
                         }
-                        return;
-                    }
-
-                    if let Ok(mut socks) = tokio::net::TcpStream::connect(&proxy_addr).await {
-                        if socks.write_all(&[5, 1, 0]).await.is_err() { return; }
-                        let mut buf = [0u8; 2];
-                        if socks.read_exact(&mut buf).await.is_err() || buf[0] != 5 || buf[1] != 0 { return; }
-                        
-                        let ip = remote.ip();
-                        let port = remote.port();
-                        let mut req = vec![5, 1, 0];
-                        match ip {
-                            std::net::IpAddr::V4(v4) => {
-                                req.push(1);
-                                req.extend_from_slice(&v4.octets());
-                            }
-                            std::net::IpAddr::V6(v6) => {
-                                req.push(4);
-                                req.extend_from_slice(&v6.octets());
-                            }
+                        _ => {
+                            tracing::debug!("Android bypass connect to {remote} failed");
                         }
-                        req.extend_from_slice(&port.to_be_bytes());
-                        if socks.write_all(&req).await.is_err() { return; }
-                        
-                        let mut rep = [0u8; 10];
-                        if socks.read_exact(&mut rep).await.is_err() || rep[1] != 0 { return; }
-
-                        // Write sniffed buffer to socks
-                        if sniff_len > 0 {
-                            if socks.write_all(&sniff_buf[..sniff_len]).await.is_err() { return; }
-                        }
-
-                        let _ = tokio::io::copy_bidirectional(&mut stream, &mut socks).await;
                     }
-                });
-            }
+                    return;
+                }
+
+                // Tunnel via SOCKS5 proxy
+                let Ok(mut socks) = tokio::net::TcpStream::connect(&proxy_addr).await else {
+                    return;
+                };
+                if socks.write_all(&[5, 1, 0]).await.is_err() { return; }
+                let mut buf2 = [0u8; 2];
+                if socks.read_exact(&mut buf2).await.is_err() || buf2[0] != 5 || buf2[1] != 0 {
+                    return;
+                }
+                let mut req = vec![5u8, 1, 0];
+                match remote.ip() {
+                    std::net::IpAddr::V4(v4) => {
+                        req.push(1);
+                        req.extend_from_slice(&v4.octets());
+                    }
+                    std::net::IpAddr::V6(v6) => {
+                        req.push(4);
+                        req.extend_from_slice(&v6.octets());
+                    }
+                }
+                req.extend_from_slice(&remote.port().to_be_bytes());
+                if socks.write_all(&req).await.is_err() { return; }
+                let mut rep = [0u8; 10];
+                if socks.read_exact(&mut rep).await.is_err() || rep[1] != 0 { return; }
+                if sniff_len > 0 && socks.write_all(&sniff_buf[..sniff_len]).await.is_err() {
+                    return;
+                }
+                let _ = tokio::io::copy_bidirectional(&mut stream, &mut socks).await;
+            });
         }
     });
 
@@ -530,7 +729,7 @@ pub async fn run_native_tunnel_from_fd(
         _ = &mut tcp_accept_task => {}
     }
 
-    tracing::info!("Deactivating NATIVE TUN tunnel...");
+    tracing::info!("NATIVE TUN (Android) deactivated.");
     Ok(())
 }
 
@@ -538,8 +737,8 @@ pub async fn run_native_tunnel_from_fd(
 pub async fn run_native_tunnel_from_fd(
     _config: crate::config::ClientConfig,
     _shutdown: watch::Receiver<bool>,
+    _exclusions_rx: watch::Receiver<crate::config::ExclusionConfig>,
     _fd: i32,
 ) -> Result<()> {
     Err(anyhow!("Native TUN from FD is only supported on Android"))
 }
-

ostp-client/src/tunnel/process_lookup.rs

@@ -1,6 +1,6 @@
 #[cfg(target_os = "windows")]
 pub fn get_process_name_from_port(port: u16) -> Option<String> {
-    use winapi::shared::minwindef::{DWORD, ULONG};
+    use winapi::shared::minwindef::ULONG;
     use winapi::shared::winerror::ERROR_INSUFFICIENT_BUFFER;
     use winapi::um::iphlpapi::GetExtendedTcpTable;
     use winapi::shared::tcpmib::{MIB_TCPTABLE_OWNER_PID, MIB_TCPROW_OWNER_PID};
@@ -47,6 +47,54 @@ pub fn get_process_name_from_port(port: u16) -> Option<String> {
     None
 }
 
+#[cfg(target_os = "windows")]
+pub fn get_process_name_from_port_udp(port: u16) -> Option<String> {
+    use winapi::shared::minwindef::ULONG;
+    use winapi::shared::winerror::ERROR_INSUFFICIENT_BUFFER;
+    use winapi::um::iphlpapi::GetExtendedUdpTable;
+    use winapi::shared::udpmib::{MIB_UDPTABLE_OWNER_PID, MIB_UDPROW_OWNER_PID};
+
+    let mut size: ULONG = 0;
+    let table_class = 1; // UDP_TABLE_OWNER_PID
+    let mut table = vec![0u8; 1024];
+
+    unsafe {
+        let mut ret = GetExtendedUdpTable(
+            table.as_mut_ptr() as *mut _,
+            &mut size,
+            0,
+            2, // AF_INET
+            table_class,
+            0,
+        );
+
+        if ret == ERROR_INSUFFICIENT_BUFFER {
+            table.resize(size as usize, 0);
+            ret = GetExtendedUdpTable(
+                table.as_mut_ptr() as *mut _,
+                &mut size,
+                0,
+                2, // AF_INET
+                table_class,
+                0,
+            );
+        }
+
+        if ret == 0 {
+            let udp_table = &*(table.as_ptr() as *const MIB_UDPTABLE_OWNER_PID);
+            let row_ptr = &udp_table.table[0] as *const MIB_UDPROW_OWNER_PID;
+            for i in 0..udp_table.dwNumEntries {
+                let row = &*row_ptr.add(i as usize);
+                let local_port = u16::from_be(row.dwLocalPort as u16);
+                if local_port == port {
+                    return get_process_name_from_pid(row.dwOwningPid);
+                }
+            }
+        }
+    }
+    None
+}
+
 #[cfg(target_os = "windows")]
 fn get_process_name_from_pid(pid: u32) -> Option<String> {
     use winapi::um::processthreadsapi::OpenProcess;
@@ -78,7 +126,6 @@ pub fn get_process_name_from_port(port: u16) -> Option<String> {
     use std::fs;
     use std::io::{BufRead, BufReader};
 
-    let mut target_inode = None;
     let hex_port = format!("{:04X}", port);
 
     let check_net_file = |path: &str| -> Option<u64> {
@@ -98,7 +145,7 @@ pub fn get_process_name_from_port(port: u16) -> Option<String> {
         None
     };
 
-    target_inode = check_net_file("/proc/net/tcp")
+    let target_inode = check_net_file("/proc/net/tcp")
         .or_else(|| check_net_file("/proc/net/tcp6"))
         .or_else(|| check_net_file("/proc/net/udp"))
         .or_else(|| check_net_file("/proc/net/udp6"));
@@ -140,3 +187,8 @@ pub fn get_process_name_from_port(port: u16) -> Option<String> {
 pub fn get_process_name_from_port(_port: u16) -> Option<String> {
     None
 }
+
+#[cfg(not(target_os = "windows"))]
+pub fn get_process_name_from_port_udp(port: u16) -> Option<String> {
+    get_process_name_from_port(port)
+}

ostp-client/src/tunnel/proxy.rs

@@ -1,5 +1,5 @@
 use std::collections::HashMap;
-use crate::tunnel::exclusion::{ExclusionMatcher, Cidr};
+use crate::tunnel::exclusion::ExclusionMatcher;
 use anyhow::{anyhow, Context, Result};
 use tokio::io::{AsyncReadExt, AsyncWriteExt};
 use tokio::net::{TcpListener, TcpStream, UdpSocket};
@@ -29,9 +29,10 @@ extern "system" {
 }
 
 #[cfg(target_os = "windows")]
-fn bind_socket_to_interface(socket: &impl AsRawSocket, is_ipv6: bool, if_index: u32) -> std::io::Result<()> {
+pub fn bind_socket_to_interface(socket: &impl AsRawSocket, is_ipv6: bool, if_index: u32) -> std::io::Result<()> {
     let s = socket.as_raw_socket() as usize;
     if is_ipv6 {
+        // IPV6_UNICAST_IF expects interface index in host byte order
         let optval = if_index;
         let ret = unsafe {
             setsockopt(
@@ -46,6 +47,7 @@ fn bind_socket_to_interface(socket: &impl AsRawSocket, is_ipv6: bool, if_index:
             return Err(std::io::Error::last_os_error());
         }
     } else {
+        // IP_UNICAST_IF expects interface index in NETWORK byte order (big-endian)
         let optval = if_index.to_be();
         let ret = unsafe {
             setsockopt(
@@ -64,7 +66,7 @@ fn bind_socket_to_interface(socket: &impl AsRawSocket, is_ipv6: bool, if_index:
 }
 
 #[cfg(target_os = "linux")]
-fn bind_socket_to_interface(socket: &impl AsRawFd, if_name: &str) -> std::io::Result<()> {
+pub fn bind_socket_to_interface(socket: &impl AsRawFd, if_name: &str) -> std::io::Result<()> {
     let fd = socket.as_raw_fd();
     let mut if_name_bytes = if_name.as_bytes().to_vec();
     if_name_bytes.push(0);
@@ -83,31 +85,18 @@ fn bind_socket_to_interface(socket: &impl AsRawFd, if_name: &str) -> std::io::Re
     Ok(())
 }
 
-fn get_windows_physical_if_index() -> Option<u32> {
+pub fn get_windows_physical_if_index() -> Option<u32> {
     #[cfg(target_os = "windows")]
     {
-        use std::os::windows::process::CommandExt;
-        const CREATE_NO_WINDOW: u32 = 0x08000000;
-        let output = std::process::Command::new("powershell")
-            .creation_flags(CREATE_NO_WINDOW)
-            .args([
-                "-NoProfile",
-                "-Command",
-                "Get-NetRoute -DestinationPrefix '0.0.0.0/0' | Where-Object { $_.InterfaceAlias -notmatch 'ostp' -and $_.InterfaceAlias -notmatch 'tun' -and $_.InterfaceAlias -notmatch 'wintun' } | Sort-Object RouteMetric | Select-Object -ExpandProperty InterfaceIndex -First 1"
-            ])
-            .output()
-            .ok()?;
-        if output.status.success() {
-            let s = String::from_utf8_lossy(&output.stdout);
-            if let Ok(index) = s.trim().parse::<u32>() {
-                return Some(index);
-            }
-        }
+        return ostp_tun::windows::windows_route::sys::get_default_ipv4_route().map(|(_, idx)| idx);
+    }
+    #[cfg(not(target_os = "windows"))]
+    {
+        None
     }
-    None
 }
 
-fn get_linux_physical_if_name() -> Option<String> {
+pub fn get_linux_physical_if_name() -> Option<String> {
     #[cfg(target_os = "linux")]
     {
         let output = std::process::Command::new("ip")
@@ -208,7 +197,7 @@ async fn create_udp_socket_bypassing_tun(
 pub async fn run_local_socks5_proxy(
     cfg: LocalProxyConfig,
     ostp: OstpConfig,
-    exclusions: ExclusionConfig,
+    mut exclusions_rx: watch::Receiver<ExclusionConfig>,
     debug: bool,
     mut shutdown: watch::Receiver<bool>,
     proxy_events_tx: mpsc::Sender<ProxyEvent>,
@@ -219,10 +208,7 @@ pub async fn run_local_socks5_proxy(
         .await
         .with_context(|| format!("failed to bind local HTTP/SOCKS5 proxy at {}", cfg.bind_addr))?;
 
-    if debug {
-        tracing::info!("local HTTP/SOCKS5 proxy listening at {}", cfg.bind_addr);
-        tracing::info!("Windows system proxy: set HTTP proxy to {}. tun2socks: SOCKS5 on same address.", cfg.bind_addr);
-    }
+    tracing::info!("local HTTP/SOCKS5 proxy listening at {}", cfg.bind_addr);
 
     let physical_if_index = tokio::task::spawn_blocking(get_windows_physical_if_index).await.unwrap_or(None);
     let physical_if_name = tokio::task::spawn_blocking(get_linux_physical_if_name).await.unwrap_or(None);
@@ -234,7 +220,8 @@ pub async fn run_local_socks5_proxy(
         tracing::info!("Local proxy physical interface name: {:?}", physical_if_name);
     }
 
-    let matcher = ExclusionMatcher::new(&exclusions, physical_if_index, physical_if_name.clone());
+    let mut current_exclusions = exclusions_rx.borrow().clone();
+    let mut matcher = ExclusionMatcher::new(&current_exclusions, physical_if_index, physical_if_name.clone());
     let (connect_tx, mut connect_rx) = mpsc::channel(128);
     let max_chunk = ostp.mtu.saturating_sub(150).max(512);
 
@@ -248,6 +235,13 @@ pub async fn run_local_socks5_proxy(
                     break;
                 }
             }
+            Ok(_) = exclusions_rx.changed() => {
+                current_exclusions = exclusions_rx.borrow().clone();
+                matcher = ExclusionMatcher::new(&current_exclusions, physical_if_index, physical_if_name.clone());
+                if true {
+                    tracing::info!("Local proxy exclusions hot-reloaded");
+                }
+            }
             accepted = listener.accept() => {
                 let (socket, _) = accepted?;
                 let stream_id = next_stream_id;
@@ -291,7 +285,7 @@ pub async fn run_local_socks5_proxy(
             Some((stream_id, msg)) = client_msgs_rx.recv() => {
                 if stream_id == 0 {
                     if let ProxyToClientMsg::Close = msg {
-                        if debug {
+                        if true {
                             tracing::info!("Resetting all active proxy streams on reconnect");
                         }
                         for (_, tx) in active_streams.drain() {
@@ -367,6 +361,10 @@ async fn handle_udp_associate(
 
     let mut direct_udp_v4: Option<Arc<UdpSocket>> = None;
     let mut direct_udp_v6: Option<Arc<UdpSocket>> = None;
+    // Held only to keep the direct-UDP readers' cancellation senders alive;
+    // dropping this (on every return path from this function) is what tells
+    // spawn_direct_udp_reader's tasks to stop. See its doc comment.
+    let mut direct_udp_cancel_txs: Vec<tokio::sync::oneshot::Sender<()>> = Vec::new();
 
     let mut tcp_buf = [0u8; 1];
     loop {
@@ -426,8 +424,8 @@ async fn handle_udp_associate(
                 let target_port = match split_host_port(&target) { Some((_, p)) => p, None => 0 };
                 // Check if target should bypass the tunnel
                 if matcher.should_bypass_target(&target_host, target_port, connect_timeout).await {
-                    if debug {
-                        tracing::info!("proxy UDP BYPASS target={}", target);
+                    if true {
+                        tracing::debug!("proxy UDP BYPASS target={}", target);
                     }
                     // Resolve target to find if it is IPv4 or IPv6
                     if let Ok(resolved_addrs) = tokio::net::lookup_host(&target).await {
@@ -438,7 +436,9 @@ async fn handle_udp_associate(
                                     match create_udp_socket_bypassing_tun(true, matcher.physical_if_index, &matcher.physical_if_name).await {
                                         Ok(s) => {
                                             let s_arc = Arc::new(s);
-                                            spawn_direct_udp_reader(s_arc.clone(), sock_tx.clone(), client_udp_addr.clone(), debug);
+                                            let (cancel_tx, cancel_rx) = tokio::sync::oneshot::channel();
+                                            spawn_direct_udp_reader(s_arc.clone(), sock_tx.clone(), client_udp_addr.clone(), debug, cancel_rx);
+                                            direct_udp_cancel_txs.push(cancel_tx);
                                             direct_udp_v6 = Some(s_arc);
                                         }
                                         Err(e) => {
@@ -452,7 +452,9 @@ async fn handle_udp_associate(
                                     match create_udp_socket_bypassing_tun(false, matcher.physical_if_index, &matcher.physical_if_name).await {
                                         Ok(s) => {
                                             let s_arc = Arc::new(s);
-                                            spawn_direct_udp_reader(s_arc.clone(), sock_tx.clone(), client_udp_addr.clone(), debug);
+                                            let (cancel_tx, cancel_rx) = tokio::sync::oneshot::channel();
+                                            spawn_direct_udp_reader(s_arc.clone(), sock_tx.clone(), client_udp_addr.clone(), debug, cancel_rx);
+                                            direct_udp_cancel_txs.push(cancel_tx);
                                             direct_udp_v4 = Some(s_arc);
                                         }
                                         Err(e) => {
@@ -465,7 +467,7 @@ async fn handle_udp_associate(
 
                             if let Some(s) = direct_socket {
                                 if let Err(e) = s.send_to(&payload, target_addr).await {
-                                    if debug {
+                                    if true {
                                         tracing::warn!("failed to send bypass UDP packet to {}: {}", target_addr, e);
                                     }
                                 }
@@ -525,12 +527,25 @@ fn spawn_direct_udp_reader(
     direct_socket: Arc<UdpSocket>,
     sock_tx: Arc<UdpSocket>,
     client_udp_addr: Arc<std::sync::Mutex<Option<std::net::SocketAddr>>>,
-    debug: bool,
+    _debug: bool,
+    mut cancel_rx: tokio::sync::oneshot::Receiver<()>,
 ) {
     tokio::spawn(async move {
         let mut buf = vec![0u8; 65536];
         loop {
-            match direct_socket.recv_from(&mut buf).await {
+            let recv_result = tokio::select! {
+                // Fires as soon as the sender half (held by handle_udp_associate
+                // for exactly this reason) is dropped - which happens the
+                // instant that function returns, on every exit path, with no
+                // explicit signaling needed. Without this, a UDP-associate
+                // session that ever bypassed traffic direct (excluded IP/
+                // domain) leaked this socket + task for the rest of the
+                // process's life once the session ended: nothing else ever
+                // stopped this loop.
+                _ = &mut cancel_rx => break,
+                res = direct_socket.recv_from(&mut buf) => res,
+            };
+            match recv_result {
                 Ok((len, target_addr)) => {
                     let client_addr = {
                         let guard = client_udp_addr.lock().unwrap();
@@ -550,14 +565,14 @@ fn spawn_direct_udp_reader(
                         packet.extend_from_slice(&target_addr.port().to_be_bytes());
                         packet.extend_from_slice(&buf[..len]);
                         if let Err(e) = sock_tx.send_to(&packet, client_addr).await {
-                            if debug {
+                            if true {
                                 tracing::warn!("failed to send direct UDP response to client: {e}");
                             }
                         }
                     }
                 }
                 Err(e) => {
-                    if debug {
+                    if true {
                         tracing::debug!("direct UDP socket read loop exiting: {e}");
                     }
                     break;
@@ -647,7 +662,7 @@ async fn handle_proxy_client(
         };
 
         if is_udp {
-            if debug { tracing::info!("proxy UDP ASSOCIATE stream_id={stream_id}"); }
+            if true { tracing::debug!("proxy UDP ASSOCIATE stream_id={stream_id}"); }
             let udp_socket = UdpSocket::bind("127.0.0.1:0").await?;
             let port = udp_socket.local_addr()?.port();
             let mut reply = vec![0x05, 0x00, 0x00, 0x01, 127, 0, 0, 1];
@@ -668,9 +683,7 @@ async fn handle_proxy_client(
             ).await;
         }
 
-        if debug {
-            tracing::info!("proxy CONNECT stream_id={stream_id} target={target}");
-        }
+        tracing::debug!("proxy CONNECT stream_id={stream_id} target={target}");
         let target_host = if let Some((host, _)) = split_host_port(&target) { host } else { target.clone() };
         let target_port = match split_host_port(&target) { Some((_, p)) => p, None => 0 };
         if matcher.should_bypass_target(&target_host, target_port, connect_timeout).await {
@@ -755,7 +768,7 @@ async fn handle_proxy_client(
             extract_host_port(raw_uri, default_port)
         };
 
-        if debug {
+        if true {
             tracing::info!("proxy CONNECT stream_id={stream_id} target={target}");
         }
         let target_host = if let Some((host, _)) = split_host_port(&target) { host } else { target.clone() };
@@ -815,7 +828,7 @@ async fn handle_proxy_client(
                 match read_res {
                     Ok(0) => {
                         let _ = event_tx.send(ProxyEvent::Close { stream_id }).await;
-                        if debug {
+                        if true {
                             tracing::info!("proxy CLOSE stream_id={stream_id}");
                         }
                         break;
@@ -833,7 +846,7 @@ async fn handle_proxy_client(
                     }
                     Err(_) => {
                         let _ = event_tx.send(ProxyEvent::Close { stream_id }).await;
-                        if debug {
+                        if true {
                             tracing::info!("proxy CLOSE stream_id={stream_id}");
                         }
                         break;
@@ -885,9 +898,9 @@ async fn direct_connect_socks5(
     physical_if_index: Option<u32>,
     physical_if_name: &Option<String>,
     close_tx: mpsc::Sender<u16>,
-    debug: bool,
+    _debug: bool,
 ) -> Result<()> {
-    if debug {
+    if true {
         tracing::info!("proxy BYPASS stream_id={stream_id} target={target}");
     }
     let mut remote = connect_bypassing_tun(target, physical_if_index, physical_if_name).await?;
@@ -907,9 +920,9 @@ async fn direct_connect_http(
     physical_if_index: Option<u32>,
     physical_if_name: &Option<String>,
     close_tx: mpsc::Sender<u16>,
-    debug: bool,
+    _debug: bool,
 ) -> Result<()> {
-    if debug {
+    if true {
         tracing::info!("proxy BYPASS stream_id={stream_id} target={target}");
     }
     let mut remote = connect_bypassing_tun(target, physical_if_index, physical_if_name).await?;

ostp-client/src/tunnel/sni_sniff.rs

@@ -52,7 +52,7 @@ pub fn extract_sni(data: &[u8]) -> Option<String> {
 
         if ext_type == 0x0000 { // Server Name Indication (SNI)
             if pos + 5 <= extensions_end {
-                let list_len = ((data[pos] as usize) << 8) | (data[pos + 1] as usize);
+                let _list_len = ((data[pos] as usize) << 8) | (data[pos + 1] as usize);
                 let name_type = data[pos + 2];
                 if name_type == 0 { // Hostname
                     let name_len = ((data[pos + 3] as usize) << 8) | (data[pos + 4] as usize);

ostp-client/src/tunnel/udp_nat.rs

@@ -10,6 +10,9 @@ pub async fn run_udp_nat(
     udp_socket: netstack_smoltcp::UdpSocket,
     proxy_addr: String,
     debug: bool,
+    matcher: std::sync::Arc<tokio::sync::RwLock<crate::tunnel::exclusion::ExclusionMatcher>>,
+    phys_if_index: Option<u32>,
+    phys_if_name: Option<String>,
 ) {
     let (mut rx, tx) = udp_socket.split();
     let tx = Arc::new(Mutex::new(tx));
@@ -17,33 +20,158 @@ pub async fn run_udp_nat(
     // map from internal client src to a channel that sends (payload, external_dst)
     let mut sessions: HashMap<SocketAddr, mpsc::Sender<(Vec<u8>, SocketAddr)>> = HashMap::new();
 
-    while let Some((payload, src, dst)) = rx.next().await {
-        if payload.is_empty() { continue; }
+    let mut cleanup_tick = tokio::time::interval(std::time::Duration::from_secs(60));
 
-        if !sessions.contains_key(&src) {
-            let (session_tx, mut session_rx) = mpsc::channel::<(Vec<u8>, SocketAddr)>(100000);
-            sessions.insert(src, session_tx);
+    loop {
+        tokio::select! {
+            packet = rx.next() => {
+                match packet {
+                    Some((payload, src, dst)) => {
+                        if payload.is_empty() { continue; }
 
-            let proxy_addr_clone = proxy_addr.clone();
-            let tx_clone = tx.clone();
-            
-            tokio::spawn(async move {
-                if debug { tracing::info!("Starting UDP NAT session for {}", src); }
-                let res = start_udp_session(src, proxy_addr_clone, &mut session_rx, tx_clone).await;
-                if debug && res.is_err() {
-                    tracing::info!("UDP NAT session for {} ended: {:?}", src, res.err());
+                        if !sessions.contains_key(&src) {
+                            let (session_tx, mut session_rx) = mpsc::channel::<(Vec<u8>, SocketAddr)>(1024);
+                            sessions.insert(src, session_tx);
+
+                            let proxy_addr_clone = proxy_addr.clone();
+                            let tx_clone = tx.clone();
+                            
+                            let mut should_bypass = false;
+                            {
+                                let matcher_guard = matcher.read().await;
+                                if matcher_guard.match_ip(&dst.ip()) {
+                                    should_bypass = true;
+                                    if debug {
+                                        tracing::info!("TUN UDP BYPASS (IP match): {} → {}", src, dst);
+                                    }
+                                }
+
+                                #[cfg(target_os = "windows")]
+                                if !should_bypass {
+                                    if let Some(proc_name) = crate::tunnel::process_lookup::get_process_name_from_port_udp(src.port()) {
+                                        if debug {
+                                            tracing::debug!("TUN UDP lookup: port {} -> process {}", src.port(), proc_name);
+                                        }
+                                        if matcher_guard.match_process(&proc_name) {
+                                            should_bypass = true;
+                                            if debug {
+                                                tracing::debug!("TUN UDP BYPASS (Process match): {} ({} → {})", proc_name, src, dst);
+                                            }
+                                        }
+                                    } else {
+                                        if debug {
+                                            tracing::debug!("TUN UDP lookup: port {} -> no process found", src.port());
+                                        }
+                                    }
+                                }
+                            }
+
+                            let p_if_idx = phys_if_index;
+                            let p_if_name = phys_if_name.clone();
+
+                            tokio::spawn(async move {
+                                if should_bypass {
+                                    if debug {
+                                        tracing::info!("Starting UDP BYPASS session for {}", src);
+                                    }
+                                    let res = start_udp_bypass_session(src, p_if_idx, p_if_name, &mut session_rx, tx_clone).await;
+                                    if res.is_err() {
+                                        tracing::debug!("UDP BYPASS session for {} ended: {:?}", src, res.err());
+                                    }
+                                } else {
+                                    tracing::debug!("Starting UDP NAT session for {}", src);
+                                    let res = start_udp_session(src, proxy_addr_clone, &mut session_rx, tx_clone).await;
+                                    if res.is_err() {
+                                        tracing::debug!("UDP NAT session for {} ended: {:?}", src, res.err());
+                                    }
+                                }
+                            });
+                        }
+
+                        if let Some(sender) = sessions.get(&src) {
+                            match sender.try_send((payload, dst)) {
+                                Err(mpsc::error::TrySendError::Closed(_)) => {
+                                    sessions.remove(&src);
+                                }
+                                Err(mpsc::error::TrySendError::Full(_)) => {
+                                    // Drop packet to avoid blocking the TUN interface loop
+                                }
+                                Ok(_) => {}
+                            }
+                        }
+                    }
+                    None => break,
                 }
-            });
-        }
-
-        if let Some(sender) = sessions.get(&src) {
-            if sender.send((payload, dst)).await.is_err() {
-                sessions.remove(&src);
+            }
+            _ = cleanup_tick.tick() => {
+                sessions.retain(|_, sender| !sender.is_closed());
             }
         }
     }
 }
 
+async fn start_udp_bypass_session(
+    client_src: SocketAddr,
+    phys_if_index: Option<u32>,
+    _phys_if_name: Option<String>,
+    session_rx: &mut mpsc::Receiver<(Vec<u8>, SocketAddr)>,
+    smoltcp_tx: Arc<Mutex<netstack_smoltcp::udp::WriteHalf>>,
+) -> anyhow::Result<()> {
+    let socket = match client_src {
+        SocketAddr::V4(_) => UdpSocket::bind("0.0.0.0:0").await?,
+        SocketAddr::V6(_) => UdpSocket::bind("[::]:0").await?,
+    };
+
+    #[cfg(target_os = "windows")]
+    if let Some(idx) = phys_if_index {
+        if let Err(e) = crate::tunnel::proxy::bind_socket_to_interface(&socket, client_src.is_ipv6(), idx) {
+            tracing::error!("TUN UDP BYPASS failed to bind to physical interface {}: {}", idx, e);
+        } else {
+            // Keep debug log
+        }
+    } else {
+        tracing::warn!("TUN UDP BYPASS has no physical interface index!");
+    }
+    
+    #[cfg(target_os = "linux")]
+    if let Some(ref name) = _phys_if_name {
+        let _ = crate::tunnel::proxy::bind_socket_to_interface(&socket, name);
+    }
+
+    // A single select! loop over both directions, rather than spawning a
+    // separate task for the read side, so the whole session - physical
+    // socket included - is torn down the moment this function returns
+    // (e.g. when session_rx closes). The previous spawned-task version left
+    // that task (and its Arc<UdpSocket> clone, keeping the OS socket fd
+    // alive) running forever after this function returned: nothing ever
+    // cancelled it, so every bypassed UDP flow (any excluded app/IP in TUN
+    // mode) leaked one socket + one task for the lifetime of the process.
+    use futures::SinkExt;
+    let mut buf = [0u8; 65536];
+    loop {
+        tokio::select! {
+            outbound = session_rx.recv() => {
+                match outbound {
+                    Some((payload, dst)) => { socket.send_to(&payload, dst).await?; }
+                    None => break,
+                }
+            }
+            inbound = socket.recv_from(&mut buf) => {
+                match inbound {
+                    Ok((n, peer)) => {
+                        let mut lock = smoltcp_tx.lock().await;
+                        let _ = lock.send((buf[..n].to_vec(), peer, client_src)).await;
+                    }
+                    Err(_) => break,
+                }
+            }
+        }
+    }
+
+    Ok(())
+}
+
+
 async fn start_udp_session(
     client_src: SocketAddr,
     proxy_addr: String,
@@ -98,6 +226,15 @@ async fn start_udp_session(
 
     // Local SOCKS5 proxy always returns 127.0.0.1 (IPv4), so always bind IPv4
     let udp = UdpSocket::bind("127.0.0.1:0").await?;
+
+    // CRITICAL for Android: protect this UDP socket so it goes out via the
+    // real physical interface, not back into the TUN (which would cause an
+    // infinite routing loop for DNS and all other UDP traffic).
+    #[cfg(target_os = "android")]
+    {
+        use std::os::unix::io::AsRawFd;
+        crate::bridge::protect_socket(udp.as_raw_fd());
+    }
     
     let mut buf = vec![0u8; 65536];
     

ostp-core/src/congestion.rs

@@ -4,6 +4,12 @@
 //! bandwidth and minimum RTT to determine the optimal sending rate.
 //! This replaces the fixed `retransmit_budget = 8` with an adaptive
 //! congestion window that responds to network conditions.
+//!
+//! RTO calculation follows RFC 6298:
+//!   SRTT = (1 - α) * SRTT + α * RTT       (α = 1/8)
+//!   RTTVAR = (1 - β) * RTTVAR + β * |SRTT - RTT|  (β = 1/4)
+//!   RTO = SRTT + 4 * RTTVAR
+//!   clamped to [RTO_MIN, RTO_MAX]
 
 use std::time::{Duration, Instant};
 
@@ -15,8 +21,14 @@ pub struct CongestionController {
     ssthresh: u64,
     /// Current phase
     phase: Phase,
-    /// Minimum RTT observed
+    /// Minimum RTT observed (for BBR-style bandwidth estimation)
     min_rtt: Duration,
+    /// Smoothed RTT (RFC 6298 SRTT)
+    srtt: Duration,
+    /// RTT variance (RFC 6298 RTTVAR)
+    rttvar: Duration,
+    /// Whether we have received a first RTT sample
+    rtt_initialized: bool,
     /// Bytes currently in flight (unacknowledged)
     bytes_in_flight: u64,
     /// Total bytes acknowledged (for bandwidth estimation)
@@ -27,46 +39,141 @@ pub struct CongestionController {
     loss_count: u32,
     /// Pacing rate: bytes per second
     pacing_rate: u64,
+    /// Token-bucket allowance for pacing, in bytes.
+    pacing_tokens: f64,
+    pacing_last_refill: Instant,
     /// MTU estimate (used for cwnd → packet count conversion)
     mtu: u64,
     /// Min RTT expiry: re-probe after 10 seconds
     min_rtt_stamp: Instant,
+    /// Loss events counted toward SLOW_START_LOSS_TOLERANCE within the
+    /// current SLOW_START_LOSS_WINDOW (see on_loss's SlowStart arm).
+    slow_start_losses: u32,
+    /// Start of the current loss-tolerance window.
+    slow_start_loss_window_start: Instant,
 }
 
 #[derive(Debug, Clone, Copy, PartialEq, Eq)]
 enum Phase {
     /// Exponential growth until loss or ssthresh
     SlowStart,
-    /// Probe bandwidth: cycle through pacing gains
+    /// Probe bandwidth: additive increase
     ProbeBandwidth,
 }
 
-/// Initial congestion window: 10 packets × MTU
-const INITIAL_CWND_PACKETS: u64 = 10;
+/// Initial congestion window: 32 packets × MTU (IW10 is too conservative for modern links)
+const INITIAL_CWND_PACKETS: u64 = 32;
 /// Minimum cwnd: 2 packets
 const MIN_CWND_PACKETS: u64 = 2;
 /// Min RTT expiry window (after which we re-probe)
 const MIN_RTT_EXPIRY: Duration = Duration::from_secs(10);
+/// Minimum RTO (RFC 6298: 1s in TCP; we use 50ms since we own the protocol)
+/// Absolute ceiling on the congestion window, in packets. At a ~1200-byte MTU
+/// this is roughly 1.2 MB in flight — already far above the bandwidth-delay
+/// product of any link this protocol realistically runs over, so anything
+/// beyond it is standing queue, not throughput. The client previously allowed
+/// up to 16384 packets (~20 MB), which on a mobile uplink is minutes of buffer.
+const MAX_CWND_PACKETS: u64 = 1024;
+/// SRTT/min_rtt ratio at which slow start stops. Doubling is what fills a deep
+/// buffer fastest, so growth must end when the queue starts building rather
+/// than waiting for a loss that a deep buffer may never produce.
+const RTT_INFLATION_EXIT_SLOW_START: f64 = 2.0;
+/// SRTT/min_rtt ratio treated as a standing queue that must be actively drained.
+const RTT_INFLATION_BACKOFF: f64 = 4.0;
+/// How much pacing allowance may accumulate, expressed as time-at-rate.
+const PACING_BURST: Duration = Duration::from_millis(10);
+const RTO_MIN: Duration = Duration::from_millis(50);
+/// Maximum RTO
+const RTO_MAX: Duration = Duration::from_secs(16);
+/// Initial RTT estimate — 30 ms is reasonable for a well-connected VPN server.
+/// Will be replaced by first real measurement within milliseconds.
+const INITIAL_RTT: Duration = Duration::from_millis(30);
+
+/// Isolated packet loss during slow start (a single dropped frame from
+/// wireless noise, a brief LTE handover blip, etc.) is normal on real
+/// mobile/Wi-Fi links and does NOT mean the link is congested. The previous
+/// behavior exited slow start and halved cwnd on the very FIRST loss, which
+/// on any link with a non-zero background loss rate permanently downgrades
+/// the session from exponential growth to linear (+1 MTU/RTT) ProbeBandwidth
+/// growth within the first few RTTs - turning what should be a sub-second
+/// ramp-up into tens of seconds to minutes before throughput opens up
+/// (observed as: a trickle of KB/s, then a sudden jump once cwnd finally
+/// claws back up). Only treat loss as a real congestion signal - and pay
+/// the full slow-start-exit + halving cost - once this many losses land
+/// within SLOW_START_LOSS_WINDOW.
+const SLOW_START_LOSS_TOLERANCE: u32 = 3;
+/// Window within which SLOW_START_LOSS_TOLERANCE losses must land to count
+/// as sustained (rather than isolated) loss. Roughly a few RTTs on a
+/// well-connected link, generous on a slow one.
+const SLOW_START_LOSS_WINDOW: Duration = Duration::from_millis(500);
 
 impl CongestionController {
     pub fn new(mtu: u64) -> Self {
         let now = Instant::now();
         let initial_cwnd = INITIAL_CWND_PACKETS * mtu;
+        // Initial pacing: deliver cwnd in ~2 RTTs to fill the pipe quickly
+        let initial_pacing = initial_cwnd * 1_000_000 / INITIAL_RTT.as_micros().max(1) as u64;
         Self {
             cwnd: initial_cwnd,
             ssthresh: u64::MAX,
             phase: Phase::SlowStart,
-            min_rtt: Duration::from_millis(100), // Conservative initial estimate
+            min_rtt: INITIAL_RTT,
+            srtt: INITIAL_RTT,
+            rttvar: INITIAL_RTT / 2,
+            rtt_initialized: false,
             bytes_in_flight: 0,
             total_acked: 0,
             last_ack_time: now,
             loss_count: 0,
-            pacing_rate: initial_cwnd * 10, // initial: ~10 windows/sec
+            pacing_rate: initial_pacing,
             mtu,
             min_rtt_stamp: now,
+            slow_start_losses: 0,
+            slow_start_loss_window_start: now,
+            pacing_tokens: (INITIAL_CWND_PACKETS * mtu) as f64,
+            pacing_last_refill: now,
         }
     }
 
+    /// Bytes of pacing allowance available right now, without consuming any.
+    ///
+    /// Read-only so the send path can use it as an admission check before it
+    /// commits to building a datagram.
+    pub fn pacing_available(&self) -> f64 {
+        let elapsed = self.pacing_last_refill.elapsed().as_secs_f64();
+        (self.pacing_tokens + elapsed * self.pacing_rate as f64).min(self.pacing_burst())
+    }
+
+    /// Whether at least one full-size packet may be released right now.
+    pub fn can_pace_packet(&self) -> bool {
+        self.pacing_available() >= self.mtu as f64
+    }
+
+    /// Ceiling on accumulated allowance.
+    ///
+    /// Pacing intervals here are fractions of a millisecond, so releasing
+    /// strictly one packet at a time would need a sub-millisecond timer per
+    /// packet. Instead we allow a short burst — the same trade every real
+    /// pacing implementation makes — sized so the loop's existing ~10ms wakeups
+    /// can still saturate the configured rate, with a small floor so a
+    /// cold/low estimate can never wedge sending entirely.
+    fn pacing_burst(&self) -> f64 {
+        let by_rate = self.pacing_rate as f64 * PACING_BURST.as_secs_f64();
+        by_rate.max((self.mtu * 4) as f64)
+    }
+
+    /// Refill from elapsed time and deduct `bytes`. Called on the real send
+    /// path; allowance is permitted to go negative so an oversized packet still
+    /// pays for itself rather than being released for free.
+    fn consume_pacing(&mut self, bytes: u64) {
+        let now = Instant::now();
+        let elapsed = now.duration_since(self.pacing_last_refill).as_secs_f64();
+        self.pacing_last_refill = now;
+        self.pacing_tokens =
+            (self.pacing_tokens + elapsed * self.pacing_rate as f64).min(self.pacing_burst())
+                - bytes as f64;
+    }
+
     /// Returns the current congestion window in bytes.
     pub fn cwnd(&self) -> u64 {
         self.cwnd
@@ -82,9 +189,20 @@ impl CongestionController {
         self.pacing_rate
     }
 
-    /// Returns the smoothed RTT estimate.
+    /// Returns the smoothed RTT estimate (SRTT).
     pub fn smoothed_rtt(&self) -> Duration {
-        self.min_rtt
+        self.srtt
+    }
+
+    /// Returns the adaptive RTO computed per RFC 6298:
+    ///   RTO = SRTT + 4 * RTTVAR, clamped to [RTO_MIN, RTO_MAX].
+    ///
+    /// This replaces the static `rto_ms` field in ProtocolMachine so that
+    /// retransmit timers automatically track changing network conditions.
+    pub fn rto(&self) -> Duration {
+        let rttvar4 = self.rttvar.saturating_mul(4);
+        let rto = self.srtt.saturating_add(rttvar4);
+        rto.clamp(RTO_MIN, RTO_MAX)
     }
 
     /// Returns how many bytes can still be sent.
@@ -107,6 +225,24 @@ impl CongestionController {
     /// Record that we sent `bytes` of data.
     pub fn on_send(&mut self, bytes: u64) {
         self.bytes_in_flight = self.bytes_in_flight.saturating_add(bytes);
+        // Charge the pacing bucket here rather than at the admission check, so
+        // every byte that actually reaches the wire is paid for exactly once —
+        // including retransmits, which are precisely what must not be allowed
+        // to bypass the rate limit and pile into an already-full queue.
+        self.consume_pacing(bytes);
+    }
+
+    /// Record that `bytes` were acknowledged but WITHOUT a usable RTT sample
+    /// (e.g. every acked frame was retransmitted, so Karn's algorithm forbids
+    /// measuring RTT from it). The window still advances; only the RTT estimator
+    /// is left untouched.
+    pub fn on_ack_no_rtt(&mut self, bytes: u64) {
+        let now = Instant::now();
+        self.bytes_in_flight = self.bytes_in_flight.saturating_sub(bytes);
+        self.total_acked = self.total_acked.saturating_add(bytes);
+        self.grow_window(bytes);
+        self.update_pacing_rate();
+        self.last_ack_time = now;
     }
 
     /// Record that `bytes` were acknowledged with the given RTT sample.
@@ -115,16 +251,57 @@ impl CongestionController {
         self.bytes_in_flight = self.bytes_in_flight.saturating_sub(bytes);
         self.total_acked = self.total_acked.saturating_add(bytes);
 
-        // Update RTT
+        // Update RTT measurements
         self.update_rtt(rtt, now);
 
-        // Update bandwidth estimate
-        self.update_bandwidth(bytes, now);
+        self.grow_window(bytes);
+        self.update_pacing_rate();
+        self.last_ack_time = now;
+    }
+
+    /// Congestion-window growth shared by both ACK paths (slow start / probe).
+    fn grow_window(&mut self, bytes: u64) {
+        // ── Delay-based congestion signal ────────────────────────────────────
+        // A loss-only controller is blind on a deeply-buffered path, and mobile
+        // carrier buffers are very deep: they absorb a burst instead of dropping
+        // it, so no loss is ever signalled and cwnd keeps growing. The queue —
+        // not the link — is what grows, and the standing delay it adds shows up
+        // as RTT inflating far above the path's floor. Left unchecked this is a
+        // positive feedback loop: bigger queue -> larger RTT samples -> larger
+        // SRTT -> larger RTO -> retransmits pile on -> bigger queue, which is
+        // how a session ends up reporting multi-second (even multi-minute) RTT
+        // and stalls video until the buffer finally drains or the user
+        // reconnects. Treat sustained RTT inflation as congestion in its own
+        // right, exactly as it is.
+        let inflation = if self.rtt_initialized && !self.min_rtt.is_zero() {
+            self.srtt.as_secs_f64() / self.min_rtt.as_secs_f64()
+        } else {
+            1.0
+        };
+
+        if inflation >= RTT_INFLATION_BACKOFF {
+            // Standing queue is severe — actively drain it.
+            self.cwnd = (self.cwnd / 2).max(MIN_CWND_PACKETS * self.mtu);
+            self.ssthresh = self.cwnd;
+            self.phase = Phase::ProbeBandwidth;
+            tracing::debug!(cwnd = self.cwnd, inflation, "congestion: draining standing queue");
+            self.clamp_cwnd();
+            return;
+        }
 
-        // State machine
         match self.phase {
             Phase::SlowStart => {
-                // Exponential growth: increase cwnd by acked bytes
+                // Exponential doubling is what fills a deep buffer fastest, so
+                // leave slow start as soon as the queue starts to build rather
+                // than waiting for the loss that may never come.
+                if inflation >= RTT_INFLATION_EXIT_SLOW_START {
+                    self.ssthresh = self.cwnd;
+                    self.phase = Phase::ProbeBandwidth;
+                    tracing::debug!(cwnd = self.cwnd, inflation, "congestion: RTT inflation ended slow start");
+                    self.clamp_cwnd();
+                    return;
+                }
+                // Exponential growth: increase cwnd by acked bytes (doubles per RTT)
                 self.cwnd = self.cwnd.saturating_add(bytes);
                 if self.cwnd >= self.ssthresh {
                     self.phase = Phase::ProbeBandwidth;
@@ -137,8 +314,20 @@ impl CongestionController {
             }
         }
 
-        self.update_pacing_rate();
-        self.last_ack_time = now;
+        self.clamp_cwnd();
+    }
+
+    /// Hard ceiling on the congestion window.
+    ///
+    /// Independent of any estimate: no real path this protocol runs over has a
+    /// bandwidth-delay product anywhere near this, so a window above it is
+    /// buffered queue rather than data in transit. Without it, slow start on a
+    /// buffer that never drops could grow the window into the tens of megabytes.
+    fn clamp_cwnd(&mut self) {
+        let ceiling = MAX_CWND_PACKETS.saturating_mul(self.mtu);
+        if self.cwnd > ceiling {
+            self.cwnd = ceiling;
+        }
     }
 
     /// Record a loss event.
@@ -148,11 +337,28 @@ impl CongestionController {
 
         match self.phase {
             Phase::SlowStart => {
-                // Exit slow start, set ssthresh to half of cwnd
-                self.ssthresh = self.cwnd / 2;
-                self.cwnd = self.ssthresh.max(MIN_CWND_PACKETS * self.mtu);
-                self.phase = Phase::ProbeBandwidth;
-                tracing::debug!(cwnd = self.cwnd, ssthresh = self.ssthresh, "congestion: loss during slow start");
+                let now = Instant::now();
+                if now.duration_since(self.slow_start_loss_window_start) > SLOW_START_LOSS_WINDOW {
+                    // Previous window's losses have aged out - this loss starts a fresh count.
+                    self.slow_start_losses = 0;
+                    self.slow_start_loss_window_start = now;
+                }
+                self.slow_start_losses += 1;
+
+                if self.slow_start_losses >= SLOW_START_LOSS_TOLERANCE {
+                    // Sustained loss within the window: treat as real congestion.
+                    // Exit slow start, set ssthresh to half of cwnd.
+                    self.ssthresh = self.cwnd / 2;
+                    self.cwnd = self.ssthresh.max(MIN_CWND_PACKETS * self.mtu);
+                    self.phase = Phase::ProbeBandwidth;
+                    tracing::debug!(cwnd = self.cwnd, ssthresh = self.ssthresh, "congestion: sustained loss during slow start, exiting");
+                } else {
+                    // Isolated loss: likely non-congestive noise. Take a mild,
+                    // temporary haircut but keep exponential growth going -
+                    // don't throw away slow start over a single dropped frame.
+                    self.cwnd = (self.cwnd * 8 / 10).max(MIN_CWND_PACKETS * self.mtu);
+                    tracing::debug!(cwnd = self.cwnd, count = self.slow_start_losses, "congestion: isolated loss during slow start, staying in slow start");
+                }
             }
             Phase::ProbeBandwidth => {
                 // Multiplicative decrease: cwnd *= 0.7 (BBR-style, less aggressive than Cubic's 0.5)
@@ -164,32 +370,49 @@ impl CongestionController {
         self.update_pacing_rate();
     }
 
-    /// Called periodically to update state.
-    pub fn on_tick(&mut self) {
-        // Nothing special needed per-tick -- state updates happen on ACK/loss
-    }
-
     // ── Private ──────────────────────────────────────────────────────────────
 
     fn update_rtt(&mut self, rtt: Duration, now: Instant) {
-        // Track windowed minimum RTT
+        // Update windowed minimum RTT (for pacing)
         if rtt < self.min_rtt || now.duration_since(self.min_rtt_stamp) >= MIN_RTT_EXPIRY {
             self.min_rtt = rtt;
             self.min_rtt_stamp = now;
         }
-    }
 
-    fn update_bandwidth(&mut self, _acked_bytes: u64, now: Instant) {
-        let elapsed = now.duration_since(self.last_ack_time);
-        if elapsed.as_micros() > 0 {
-        // Removed bw_samples tracking
+        // Update SRTT and RTTVAR per RFC 6298
+        if !self.rtt_initialized {
+            // First measurement: initialize directly
+            self.srtt = rtt;
+            self.rttvar = rtt / 2;
+            self.rtt_initialized = true;
+        } else {
+            // RTTVAR = (3/4) * RTTVAR + (1/4) * |SRTT - R|
+            let diff = if rtt > self.srtt {
+                rtt - self.srtt
+            } else {
+                self.srtt - rtt
+            };
+            // Integer-safe: RTTVAR = RTTVAR - RTTVAR/4 + diff/4
+            self.rttvar = self.rttvar
+                .saturating_sub(self.rttvar / 4)
+                .saturating_add(diff / 4);
+
+            // SRTT = (7/8) * SRTT + (1/8) * R
+            self.srtt = self.srtt
+                .saturating_sub(self.srtt / 8)
+                .saturating_add(rtt / 8);
         }
+
+        tracing::trace!(
+            srtt_ms = self.srtt.as_millis(),
+            rttvar_ms = self.rttvar.as_millis(),
+            rto_ms = self.rto().as_millis(),
+            "congestion: RTT updated"
+        );
     }
 
-
-
     fn update_pacing_rate(&mut self) {
-        // Pacing rate = cwnd / min_rtt (with gain)
+        // Pacing rate = cwnd / min_rtt (delivery rate target)
         let rtt_us = self.min_rtt.as_micros().max(1) as u64;
         self.pacing_rate = self.cwnd * 1_000_000 / rtt_us;
     }
@@ -202,19 +425,18 @@ mod tests {
     #[test]
     fn test_initial_state() {
         let cc = CongestionController::new(1200);
-        assert_eq!(cc.cwnd(), 12000); // 10 * 1200
+        assert_eq!(cc.cwnd(), 32 * 1200); // 32 * 1200
         assert!(cc.can_send());
-        assert_eq!(cc.cwnd_packets(), 10);
+        assert_eq!(cc.cwnd_packets(), 32);
     }
 
     #[test]
     fn test_slow_start_growth() {
         let mut cc = CongestionController::new(1200);
-        // Simulate sending and ACKing
+        let initial = cc.cwnd();
         cc.on_send(1200);
         cc.on_ack(1200, Duration::from_millis(50));
-        // cwnd should grow
-        assert!(cc.cwnd() > 12000);
+        assert!(cc.cwnd() > initial);
     }
 
     #[test]
@@ -225,11 +447,143 @@ mod tests {
         assert!(cc.cwnd() < initial);
     }
 
+    /// The bufferbloat case: a deep buffer absorbs everything, so NOTHING is
+    /// ever lost, but the standing queue inflates RTT. A loss-only controller
+    /// grows cwnd forever here — which is how a session ends up reporting
+    /// multi-second RTT and stalling video.
+    #[test]
+    fn test_rtt_inflation_halts_growth_without_any_loss() {
+        let mut cc = CongestionController::new(1200);
+
+        // Establish a low path floor; this becomes min_rtt.
+        for _ in 0..4 {
+            cc.on_send(1200);
+            cc.on_ack(1200, Duration::from_millis(20));
+        }
+        let cwnd_before = cc.cwnd();
+
+        // Queue builds: RTT climbs far above the floor, still zero loss.
+        for _ in 0..20 {
+            cc.on_send(1200);
+            cc.on_ack(1200, Duration::from_millis(400));
+        }
+
+        assert!(
+            cc.cwnd() <= cwnd_before,
+            "cwnd kept growing while the queue was inflating RTT ({} -> {})",
+            cwnd_before,
+            cc.cwnd()
+        );
+    }
+
+    /// Pacing must actually bound the release rate: draining the bucket has to
+    /// deny the next packet. Without this the congestion window alone decides,
+    /// and a whole window leaves back-to-back.
+    #[test]
+    fn test_pacing_bucket_denies_once_drained() {
+        let mut cc = CongestionController::new(1200);
+        assert!(cc.can_pace_packet(), "a fresh controller must allow sending");
+
+        // Spend well beyond one burst allowance.
+        let burst_bytes = cc.pacing_available();
+        let mut spent = 0.0;
+        while spent <= burst_bytes + 1200.0 {
+            cc.on_send(1200);
+            spent += 1200.0;
+        }
+
+        assert!(
+            !cc.can_pace_packet(),
+            "pacing allowed unbounded sending: {} bytes still available after spending {}",
+            cc.pacing_available(),
+            spent
+        );
+    }
+
+    /// The allowance must refill over time, or sending would stall permanently
+    /// once the first burst is spent.
+    #[test]
+    fn test_pacing_bucket_refills_over_time() {
+        let mut cc = CongestionController::new(1200);
+        while cc.can_pace_packet() {
+            cc.on_send(1200);
+        }
+        assert!(!cc.can_pace_packet());
+
+        std::thread::sleep(Duration::from_millis(25));
+        assert!(
+            cc.can_pace_packet(),
+            "pacing bucket never refilled; sending would be stuck forever"
+        );
+    }
+
+    /// cwnd must never exceed the absolute ceiling, however long slow start
+    /// runs unopposed — above it the window is buffered queue, not throughput.
+    #[test]
+    fn test_cwnd_never_exceeds_absolute_ceiling() {
+        let mut cc = CongestionController::new(1200);
+        // Constant RTT: no inflation signal, so only the hard cap can stop this.
+        for _ in 0..5000 {
+            cc.on_send(1200);
+            cc.on_ack(1200, Duration::from_millis(30));
+        }
+        assert!(
+            cc.cwnd() <= MAX_CWND_PACKETS * 1200,
+            "cwnd {} exceeded the {}-packet ceiling",
+            cc.cwnd(),
+            MAX_CWND_PACKETS
+        );
+    }
+
+    #[test]
+    fn test_isolated_slow_start_loss_does_not_exit_slow_start() {
+        // A single dropped packet (wireless noise, a brief handover blip) is
+        // normal on real links and must not permanently downgrade the
+        // session from exponential to linear growth.
+        let mut cc = CongestionController::new(1200);
+        cc.on_loss(1200);
+        assert_eq!(cc.phase, Phase::SlowStart, "one isolated loss must not exit slow start");
+
+        // It should still shrink the window somewhat (not ignored entirely),
+        // just far less punishing than the sustained-congestion case.
+        let after_one = cc.cwnd();
+        assert!(after_one < INITIAL_CWND_PACKETS * 1200);
+    }
+
+    #[test]
+    fn test_sustained_slow_start_loss_exits_slow_start() {
+        // Losses landing close together (within SLOW_START_LOSS_WINDOW) are
+        // a real congestion signal and must still trigger the harsher
+        // exit-slow-start + halve response.
+        let mut cc = CongestionController::new(1200);
+        for _ in 0..SLOW_START_LOSS_TOLERANCE {
+            cc.on_loss(1200);
+        }
+        assert_eq!(cc.phase, Phase::ProbeBandwidth, "sustained loss must exit slow start");
+    }
+
+    #[test]
+    fn test_slow_start_loss_window_resets_after_expiry() {
+        // Two losses far enough apart (window expired between them) must
+        // each be treated as isolated, not accumulated toward the sustained-
+        // loss threshold.
+        let mut cc = CongestionController::new(1200);
+        cc.on_loss(1200);
+        assert_eq!(cc.phase, Phase::SlowStart);
+
+        // Simulate the window having expired by resetting its start
+        // directly (std::thread::sleep in a unit test would be flaky/slow).
+        cc.slow_start_loss_window_start = Instant::now() - SLOW_START_LOSS_WINDOW - Duration::from_millis(1);
+        cc.on_loss(1200);
+        assert_eq!(cc.phase, Phase::SlowStart, "a loss after the window expired must restart the count, not accumulate");
+        assert_eq!(cc.slow_start_losses, 1);
+    }
+
     #[test]
     fn test_can_send_limits() {
         let mut cc = CongestionController::new(1200);
         // Send until cwnd is exhausted
-        for _ in 0..10 {
+        for _ in 0..32 {
             cc.on_send(1200);
         }
         assert!(!cc.can_send()); // cwnd exhausted
@@ -244,10 +598,63 @@ mod tests {
     }
 
     #[test]
-    fn test_rtt_tracking() {
+    fn test_rtt_tracking_first_sample() {
         let mut cc = CongestionController::new(1200);
         cc.on_send(1200);
         cc.on_ack(1200, Duration::from_millis(25));
+        // After first sample: SRTT = 25ms, RTTVAR = 12ms
         assert_eq!(cc.smoothed_rtt(), Duration::from_millis(25));
     }
+
+    #[test]
+    fn test_rto_rfc6298() {
+        let mut cc = CongestionController::new(1200);
+        // After first sample with RTT=50ms: SRTT=50ms, RTTVAR=25ms, RTO=150ms
+        cc.on_send(1200);
+        cc.on_ack(1200, Duration::from_millis(50));
+        let rto = cc.rto();
+        // RTO = 50 + 4*25 = 150ms; clamped to [50ms, 16s]
+        assert!(rto >= RTO_MIN);
+        assert!(rto <= RTO_MAX);
+        assert_eq!(rto, Duration::from_millis(150));
+    }
+
+    #[test]
+    fn test_on_ack_no_rtt_grows_window_without_touching_srtt() {
+        let mut cc = CongestionController::new(1200);
+        // Establish a known SRTT with a real sample.
+        cc.on_send(1200);
+        cc.on_ack(1200, Duration::from_millis(40));
+        let srtt_before = cc.smoothed_rtt();
+        let cwnd_before = cc.cwnd();
+
+        // A Karn's-algorithm ACK (all acked frames were retransmitted): window
+        // must advance, RTT estimate must be untouched.
+        cc.on_send(1200);
+        cc.on_ack_no_rtt(1200);
+        assert!(cc.cwnd() > cwnd_before, "cwnd should still grow on a no-RTT ack");
+        assert_eq!(cc.smoothed_rtt(), srtt_before, "SRTT must not move on a no-RTT ack");
+    }
+
+    #[test]
+    fn test_rto_clamp_min() {
+        let cc = CongestionController::new(1200);
+        // Even with no RTT samples, RTO should not go below RTO_MIN
+        assert!(cc.rto() >= RTO_MIN);
+    }
+
+    #[test]
+    fn test_rto_adapts_after_multiple_samples() {
+        let mut cc = CongestionController::new(1200);
+        // Feed several consistent RTT samples
+        for _ in 0..8 {
+            cc.on_send(1200);
+            cc.on_ack(1200, Duration::from_millis(20));
+        }
+        // After convergence, RTTVAR should be small → RTO close to SRTT + small margin
+        let rto = cc.rto();
+        // Should be well below 100ms (the old hardcoded default)
+        assert!(rto < Duration::from_millis(200));
+        assert!(rto >= RTO_MIN);
+    }
 }

ostp-core/src/crypto/mod.rs

@@ -1,11 +1,12 @@
 pub mod aead;
 pub mod noise;
 pub mod obfuscation;
-pub mod reality;
+
 
 pub use aead::SessionCipher;
 pub use noise::{NoiseRole, NoiseSession};
 pub use obfuscation::{
     deobfuscate_header_inplace, deobfuscate_packet_inplace, obfuscate_packet_inplace,
     derive_obfuscation_key, derive_psk, derive_all_secrets, DerivedSecrets,
+    derive_junk_marker, current_junk_window, JUNK_MARKER_WINDOW_SECS,
 };

ostp-core/src/crypto/noise.rs

@@ -1,4 +1,4 @@
-use snow::{Builder, HandshakeState, TransportState};
+use snow::{Builder, HandshakeState};
 
 use crate::protocol::ProtocolError;
 
@@ -10,9 +10,15 @@ pub enum NoiseRole {
     Responder,
 }
 
-pub enum NoiseSession {
-    Handshake(Box<HandshakeState>),
-    Transport(TransportState),
+/// A Noise handshake in progress. OSTP does not use snow's transport mode: once
+/// the handshake finishes we extract the raw Split() keys (see [`raw_split`])
+/// and drive our own out-of-order AEAD (see `crypto::aead`), because the wire
+/// protocol needs explicit per-frame nonces for reordering that snow's internal
+/// nonce counter can't express.
+///
+/// [`raw_split`]: NoiseSession::raw_split
+pub struct NoiseSession {
+    handshake: Box<HandshakeState>,
 }
 
 impl NoiseSession {
@@ -36,50 +42,92 @@ impl NoiseSession {
                 .map_err(|_| ProtocolError::Crypto("noise-responder".to_string()))?,
         };
 
-        Ok(Self::Handshake(Box::new(handshake)))
+        Ok(Self { handshake: Box::new(handshake) })
     }
 
     pub fn write_handshake(&mut self, payload: &[u8], out: &mut [u8]) -> Result<usize, ProtocolError> {
-        match self {
-            NoiseSession::Handshake(hs) => hs
-                .write_message(payload, out)
-                .map_err(|_| ProtocolError::Crypto("noise-write".to_string())),
-            NoiseSession::Transport(_) => Err(ProtocolError::State("noise already in transport".to_string())),
-        }
+        self.handshake
+            .write_message(payload, out)
+            .map_err(|_| ProtocolError::Crypto("noise-write".to_string()))
     }
 
     pub fn read_handshake(&mut self, input: &[u8], out: &mut [u8]) -> Result<usize, ProtocolError> {
-        match self {
-            NoiseSession::Handshake(hs) => hs
-                .read_message(input, out)
-                .map_err(|e| ProtocolError::Crypto(format!("noise-read: {:?}", e))),
-            NoiseSession::Transport(_) => Err(ProtocolError::State("noise already in transport".to_string())),
-        }
+        self.handshake
+            .read_message(input, out)
+            .map_err(|e| ProtocolError::Crypto(format!("noise-read: {:?}", e)))
     }
 
-    pub fn handshake_hash(&self, out: &mut [u8]) -> Result<(), ProtocolError> {
-        match self {
-            NoiseSession::Handshake(hs) => {
-                let hash = hs.get_handshake_hash();
-                if out.len() != hash.len() {
-                    return Err(ProtocolError::Crypto("handshake hash length mismatch".to_string()));
-                }
-                out.copy_from_slice(hash);
-                Ok(())
-            }
-            NoiseSession::Transport(_) => Err(ProtocolError::State("noise already in transport".to_string())),
-        }
-    }
-
-    pub fn into_transport(self) -> Result<Self, ProtocolError> {
-        match self {
-            NoiseSession::Handshake(hs) => {
-                let transport = hs
-                    .into_transport_mode()
-                    .map_err(|_| ProtocolError::Crypto("noise-transport".to_string()))?;
-                Ok(NoiseSession::Transport(transport))
-            }
-            NoiseSession::Transport(_) => Ok(self),
+    /// Derive the two directional transport keys via Noise's Split().
+    ///
+    /// SECURITY: keys are taken from the final chaining key `ck` (which absorbs
+    /// the ephemeral `ee` DH result via MixKey), NOT from the handshake hash `h`
+    /// (which only absorbs public transcript data — ephemeral pubkeys and
+    /// ciphertexts — and never the DH secret). Deriving from `ck` is what gives
+    /// the session forward secrecy: an adversary who later learns the PSK still
+    /// cannot recompute these keys without the ephemeral private keys, which are
+    /// discarded after the handshake.
+    ///
+    /// Must only be called once the handshake is finished (both messages of the
+    /// NNpsk0 exchange processed); at that point `ck` is final. Returns
+    /// `(send_key, recv_key)` for the given role, matching snow's TransportState
+    /// direction mapping: split output `.0` is initiator→responder, `.1` is
+    /// responder→initiator.
+    pub fn raw_split(&mut self, role: NoiseRole) -> Result<([u8; 32], [u8; 32]), ProtocolError> {
+        if !self.handshake.is_handshake_finished() {
+            return Err(ProtocolError::State("handshake not finished at key split".to_string()));
         }
+        let (k0, k1) = self.handshake.dangerously_get_raw_split();
+        Ok(match role {
+            // Initiator sends on .0 (i→r), receives on .1 (r→i).
+            NoiseRole::Initiator => (k0, k1),
+            // Responder is the mirror image.
+            NoiseRole::Responder => (k1, k0),
+        })
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    /// Drive a full NNpsk0 handshake and confirm both sides derive matching
+    /// directional keys. This guards the .0/.1 → send/recv role mapping in
+    /// `raw_split`: if it were wrong, the two sides' send/recv keys wouldn't
+    /// cross-match and the transport channel would silently fail to decrypt.
+    #[test]
+    fn raw_split_keys_agree_across_roles() {
+        let psk = [7u8; 32];
+        let mut initiator = NoiseSession::new(NoiseRole::Initiator, &psk).unwrap();
+        let mut responder = NoiseSession::new(NoiseRole::Responder, &psk).unwrap();
+
+        // msg1: initiator -> responder
+        let mut buf1 = [0u8; 1024];
+        let n1 = initiator.write_handshake(&[], &mut buf1).unwrap();
+        let mut tmp = [0u8; 1024];
+        responder.read_handshake(&buf1[..n1], &mut tmp).unwrap();
+
+        // msg2: responder -> initiator
+        let mut buf2 = [0u8; 1024];
+        let n2 = responder.write_handshake(&[], &mut buf2).unwrap();
+        initiator.read_handshake(&buf2[..n2], &mut tmp).unwrap();
+
+        let (i_send, i_recv) = initiator.raw_split(NoiseRole::Initiator).unwrap();
+        let (r_send, r_recv) = responder.raw_split(NoiseRole::Responder).unwrap();
+
+        // What the initiator sends with, the responder must receive with.
+        assert_eq!(i_send, r_recv, "initiator send key must equal responder recv key");
+        assert_eq!(r_send, i_recv, "responder send key must equal initiator recv key");
+        // The two directions use distinct keys.
+        assert_ne!(i_send, i_recv, "the two directions must not share a key");
+    }
+
+    /// raw_split must refuse to hand out keys before the handshake is complete —
+    /// keys taken from a half-mixed chaining key would be wrong and insecure.
+    #[test]
+    fn raw_split_rejected_before_handshake_finishes() {
+        let psk = [9u8; 32];
+        let mut initiator = NoiseSession::new(NoiseRole::Initiator, &psk).unwrap();
+        // No messages exchanged yet: handshake not finished.
+        assert!(initiator.raw_split(NoiseRole::Initiator).is_err());
     }
 }

ostp-core/src/crypto/obfuscation.rs

@@ -54,14 +54,41 @@ fn hkdf_expand(prk: &[u8; 32], info: &[u8], len: usize) -> Vec<u8> {
 /// The derivation uses the access key as both IKM and salt material,
 /// split into two halves. No fixed strings are used — the access key
 /// alone determines all derived values.
+#[derive(Clone)]
 pub struct DerivedSecrets {
     pub obfuscation_key: [u8; 8],
     pub psk: [u8; 32],
     pub handshake_pad_min: usize,
     pub handshake_pad_max: usize,
 }
+// NOTE: the junk marker is NOT part of DerivedSecrets — it is time-rotating and
+// derived separately per window via `derive_junk_marker` (see below), so it
+// carries no static per-user signature.
+
+/// OSTP wire protocol version. Mixed into key derivation (NOT sent on the
+/// wire) so peers running incompatible versions derive entirely different
+/// secrets and therefore cannot deobfuscate / decrypt each other's traffic.
+///
+/// This is a hard, deterministic version gate that needs NO plaintext version
+/// byte on the wire — a constant marker would defeat the project's stealth
+/// north-star ("no recognizable header"). A pre-0.4.0 client (which derived
+/// without a version) produces a different obfuscation key, so a 0.4.0 server
+/// cannot recover its handshake header and rejects it as an unauthorized probe.
+///
+/// Bump this on any wire-breaking protocol change. 0.4.0 = version 4;
+/// version 5 (0.4.x hardening) moved transport keys from the handshake hash to
+/// Noise's Split() output — a wire-breaking crypto change, so old peers must not
+/// interop (they would derive different session keys and fail decryption).
+pub const PROTOCOL_VERSION: u8 = 5;
 
 pub fn derive_all_secrets(access_key: &[u8]) -> DerivedSecrets {
+    derive_all_secrets_versioned(access_key, PROTOCOL_VERSION)
+}
+
+/// Version-parameterised derivation. `derive_all_secrets` always pins the
+/// current `PROTOCOL_VERSION`; this form exists so tests can prove that a
+/// different version yields incompatible secrets (the version gate).
+pub(crate) fn derive_all_secrets_versioned(access_key: &[u8], version: u8) -> DerivedSecrets {
     // Split the key hash into two halves for salt/info separation.
     // This avoids using any hardcoded strings while still providing
     // domain separation between the derived values.
@@ -70,8 +97,16 @@ pub fn derive_all_secrets(access_key: &[u8]) -> DerivedSecrets {
     let salt = &key_hash[..16];
     let info_base = &key_hash[16..];
 
-    // Extract PRK from access key using its own hash as salt
-    let prk = hkdf_extract(salt, access_key);
+    // Mix the protocol version into the IKM so a different version produces a
+    // completely different PRK → different obf_key / psk / padding. This is the
+    // wire-version gate: it is invisible on the wire (only the derived output,
+    // which is already indistinguishable from random, ever leaves the host).
+    let mut ikm = Vec::with_capacity(access_key.len() + 1);
+    ikm.extend_from_slice(access_key);
+    ikm.push(version);
+
+    // Extract PRK from version-tagged access key using its hash as salt
+    let prk = hkdf_extract(salt, &ikm);
 
     // Derive obfuscation key (8 bytes) — info = key_hash[16..] || 0x01
     let mut obf_info = info_base.to_vec();
@@ -105,6 +140,53 @@ pub fn derive_all_secrets(access_key: &[u8]) -> DerivedSecrets {
     }
 }
 
+/// Window length (seconds) for the rotating junk marker. The marker changes
+/// every window, so junk carries no static per-user fingerprint on the wire;
+/// the server checks the current and previous window to absorb clock skew.
+pub const JUNK_MARKER_WINDOW_SECS: u64 = 60;
+
+/// The current junk-marker time window (unix seconds / window length).
+pub fn current_junk_window() -> u64 {
+    std::time::SystemTime::now()
+        .duration_since(std::time::UNIX_EPOCH)
+        .map(|d| d.as_secs() / JUNK_MARKER_WINDOW_SECS)
+        .unwrap_or(0)
+}
+
+/// Derive the 4-byte junk marker for a given time `window`.
+///
+/// Uses the same version-gated HKDF scheme as [`derive_all_secrets`], with the
+/// window folded into the `info` (label byte `0x04`). Folding in the window
+/// makes the marker rotate: to an on-path observer the junk prefix changes every
+/// window (no fixed signature), and a captured marker is only valid for ~1
+/// window. Only a holder of the access key can compute it, so an outsider cannot
+/// forge a silently-dropped junk packet.
+pub fn derive_junk_marker(access_key: &[u8], window: u64) -> [u8; 4] {
+    derive_junk_marker_versioned(access_key, window, PROTOCOL_VERSION)
+}
+
+pub(crate) fn derive_junk_marker_versioned(access_key: &[u8], window: u64, version: u8) -> [u8; 4] {
+    use sha2::Digest;
+    let key_hash = sha2::Sha256::digest(access_key);
+    let salt = &key_hash[..16];
+    let info_base = &key_hash[16..];
+
+    let mut ikm = Vec::with_capacity(access_key.len() + 1);
+    ikm.extend_from_slice(access_key);
+    ikm.push(version);
+    let prk = hkdf_extract(salt, &ikm);
+
+    // info = key_hash[16..] || 0x04 || window(LE) — same label byte as before,
+    // now parameterised by the time window.
+    let mut info = info_base.to_vec();
+    info.push(0x04);
+    info.extend_from_slice(&window.to_le_bytes());
+    let bytes = hkdf_expand(&prk, &info, 4);
+    let mut marker = [0u8; 4];
+    marker.copy_from_slice(&bytes);
+    marker
+}
+
 // ── Legacy API (delegates to derive_all_secrets) ─────────────────────────────
 
 pub fn derive_obfuscation_key(access_key: &[u8]) -> [u8; 8] {

ostp-core/src/crypto/obfuscation_tests.rs

@@ -127,6 +127,37 @@ mod tests {
         assert_eq!(correct_sid, session_id, "correct key must recover session_id");
     }
 
+    /// §C version gate: a peer on a different PROTOCOL_VERSION derives
+    /// different secrets, so a handshake obfuscated with the OLD version's key
+    /// does NOT deobfuscate to a valid session_id under the current version.
+    /// This is exactly what makes an old (pre-0.4.0) client fail to connect to
+    /// a new server — with no plaintext version marker on the wire.
+    #[test]
+    fn test_protocol_version_gates_old_clients() {
+        let key = b"shared_access_key_across_versions";
+        let new = derive_all_secrets(key); // == derive_all_secrets_versioned(key, PROTOCOL_VERSION)
+        let old = derive_all_secrets_versioned(key, PROTOCOL_VERSION.wrapping_sub(1));
+
+        // Different protocol version → different derived secrets.
+        assert_ne!(new.obfuscation_key, old.obfuscation_key, "version must change obf_key");
+        assert_ne!(new.psk, old.psk, "version must change psk");
+
+        // Concretely: a handshake the old client obfuscated with its key does
+        // not recover a valid session_id when the new server deobfuscates it.
+        let session_id: u32 = 0x11223344;
+        let noise = [0x33u8; 48];
+        let mut pkt = Vec::new();
+        pkt.extend_from_slice(&session_id.to_be_bytes());
+        pkt.extend_from_slice(&(noise.len() as u16).to_be_bytes());
+        pkt.extend_from_slice(&noise);
+        pkt.extend_from_slice(&[0u8; 32]);
+
+        obfuscate_packet_inplace(&mut pkt, &old.obfuscation_key, true);   // old client
+        deobfuscate_packet_inplace(&mut pkt, &new.obfuscation_key, true); // new server
+        let recovered = u32::from_be_bytes([pkt[0], pkt[1], pkt[2], pkt[3]]);
+        assert_ne!(recovered, session_id, "old-version client must NOT be accepted by new server");
+    }
+
     /// Verifies data packet obfuscation round-trip (non-handshake path).
     #[test]
     fn test_data_packet_obfuscation_roundtrip() {
@@ -160,4 +191,29 @@ mod tests {
         assert_eq!(recovered_nonce, nonce);
         assert_eq!(&packet[12..], &ciphertext);
     }
+
+    /// The junk marker must: be stable within a window (client and server agree),
+    /// rotate across windows (no static on-wire fingerprint), and differ per key
+    /// (one user's marker never silently-drops on another user's flow).
+    #[test]
+    fn test_junk_marker_rotation() {
+        let key_a = b"access-key-alpha";
+        let key_b = b"access-key-bravo";
+
+        // Stable within a window.
+        assert_eq!(derive_junk_marker(key_a, 1000), derive_junk_marker(key_a, 1000));
+
+        // Rotates across adjacent windows.
+        assert_ne!(derive_junk_marker(key_a, 1000), derive_junk_marker(key_a, 1001));
+        assert_ne!(derive_junk_marker(key_a, 1000), derive_junk_marker(key_a, 999));
+
+        // Distinct per key within the same window.
+        assert_ne!(derive_junk_marker(key_a, 1000), derive_junk_marker(key_b, 1000));
+
+        // A different protocol version yields a different marker (version gate).
+        assert_ne!(
+            derive_junk_marker_versioned(key_a, 1000, PROTOCOL_VERSION),
+            derive_junk_marker_versioned(key_a, 1000, PROTOCOL_VERSION.wrapping_add(1)),
+        );
+    }
 }

ostp-core/src/crypto/reality.rs

@@ -1,279 +0,0 @@
-use bytes::{Buf, BufMut, Bytes, BytesMut};
-use chacha20poly1305::{aead::{Aead, KeyInit}, ChaCha20Poly1305, Nonce};
-use hkdf::Hkdf;
-use sha2::Sha256;
-use x25519_dalek::{PublicKey, StaticSecret};
-use rand::{rngs::OsRng, RngCore};
-use std::time::{SystemTime, UNIX_EPOCH};
-
-const REALITY_INFO: &[u8] = b"ostp-reality-v1";
-const RECORD_HEADER_LEN: usize = 5;
-const HANDSHAKE_HEADER_LEN: usize = 4;
-
-/// Number of TLS records sent by the server during the fake handshake phase.
-/// Client must read and discard this many records before starting RealityStream.
-/// Layout: 1× ServerHello (0x16) + 1× CCS (0x14) + 3× fake encrypted records (0x17)
-pub const REALITY_SERVER_HANDSHAKE_RECORDS: usize = 5;
-
-/// Generates an X25519 keypair
-pub fn generate_x25519_keypair() -> (StaticSecret, PublicKey) {
-    let secret = StaticSecret::random_from_rng(OsRng);
-    let public = PublicKey::from(&secret);
-    (secret, public)
-}
-
-/// Derives the Auth Key and Data Key from the X25519 shared secret
-pub fn derive_keys(shared_secret: &[u8; 32]) -> (ChaCha20Poly1305, ChaCha20Poly1305) {
-    let hk = Hkdf::<Sha256>::new(None, shared_secret);
-    let mut okm = [0u8; 64];
-    hk.expand(REALITY_INFO, &mut okm).expect("HKDF expand failed");
-    
-    let auth_key = ChaCha20Poly1305::new_from_slice(&okm[0..32]).unwrap();
-    let data_key = ChaCha20Poly1305::new_from_slice(&okm[32..64]).unwrap();
-    (auth_key, data_key)
-}
-
-/// Creates an authenticated Session ID payload (32 bytes)
-/// sid: 8 bytes, timestamp: 8 bytes. Encrypted with ChaCha20Poly1305 (16 byte tag). Total = 32 bytes.
-pub fn generate_session_id(auth_aead: &ChaCha20Poly1305, sid: &[u8; 8]) -> [u8; 32] {
-    let ts = SystemTime::now().duration_since(UNIX_EPOCH).unwrap().as_secs();
-    let mut plaintext = [0u8; 16];
-    plaintext[0..8].copy_from_slice(sid);
-    plaintext[8..16].copy_from_slice(&ts.to_be_bytes());
-    
-    let nonce = Nonce::from_slice(&[0u8; 12]); // Fixed nonce since auth key is ephemeral per connection
-    let ciphertext = auth_aead.encrypt(nonce, plaintext.as_ref()).expect("encryption failed");
-    
-    let mut session_id = [0u8; 32];
-    session_id.copy_from_slice(&ciphertext);
-    session_id
-}
-
-/// Verifies and decrypts the Session ID payload. Returns (sid, timestamp)
-pub fn verify_session_id(auth_aead: &ChaCha20Poly1305, session_id: &[u8; 32]) -> Option<([u8; 8], u64)> {
-    let nonce = Nonce::from_slice(&[0u8; 12]);
-    let plaintext = auth_aead.decrypt(nonce, session_id.as_ref()).ok()?;
-    
-    if plaintext.len() != 16 {
-        return None;
-    }
-    
-    let mut sid = [0u8; 8];
-    sid.copy_from_slice(&plaintext[0..8]);
-    let mut ts_bytes = [0u8; 8];
-    ts_bytes.copy_from_slice(&plaintext[8..16]);
-    let ts = u64::from_be_bytes(ts_bytes);
-    
-    let now = SystemTime::now().duration_since(UNIX_EPOCH).unwrap().as_secs();
-    // Allow up to 60 seconds of clock drift
-    if ts > now + 60 || ts < now.saturating_sub(60) {
-        return None; // Replay protection / stale connection
-    }
-    
-    Some((sid, ts))
-}
-
-/// Builds a fake TLS 1.3 ClientHello matching Chrome's fingerprint
-pub fn build_client_hello(sni: &str, session_id: &[u8; 32], c_pub: &PublicKey) -> Bytes {
-    let mut ext = BytesMut::new();
-    
-    // SNI Extension
-    let sni_bytes = sni.as_bytes();
-    ext.put_u16(0x0000); // Type: server_name
-    ext.put_u16((sni_bytes.len() + 5) as u16);
-    ext.put_u16((sni_bytes.len() + 3) as u16); // Server Name list length
-    ext.put_u8(0x00); // Name Type: host_name
-    ext.put_u16(sni_bytes.len() as u16);
-    ext.put_slice(sni_bytes);
-    
-    // Supported Groups
-    ext.put_u16(0x000a); // Type
-    ext.put_u16(8); // Length
-    ext.put_u16(6); // List length
-    ext.put_u16(0x001d); // x25519
-    ext.put_u16(0x0017); // secp256r1
-    ext.put_u16(0x0018); // secp384r1
-
-    // Key Share
-    let pub_bytes = c_pub.as_bytes();
-    ext.put_u16(0x0033); // Type
-    ext.put_u16((pub_bytes.len() + 6) as u16); // Length
-    ext.put_u16((pub_bytes.len() + 4) as u16); // ClientShares length
-    ext.put_u16(0x001d); // Group: x25519
-    ext.put_u16(pub_bytes.len() as u16);
-    ext.put_slice(pub_bytes);
-    
-    // Supported Versions
-    ext.put_u16(0x002b); // Type
-    ext.put_u16(5); // Length
-    ext.put_u8(4); // List length
-    ext.put_u16(0x0304); // TLS 1.3
-    ext.put_u16(0x0303); // TLS 1.2
-    
-    // ALPN
-    let alpn = b"\x02h2\x08http/1.1";
-    ext.put_u16(0x0010); // Type
-    ext.put_u16((alpn.len() + 2) as u16);
-    ext.put_u16(alpn.len() as u16);
-    ext.put_slice(alpn);
-
-    // Signature Algorithms
-    ext.put_u16(0x000d); // Type
-    ext.put_u16(10); // Length
-    ext.put_u16(8); // List length
-    ext.put_u16(0x0403); // ecdsa_secp256r1_sha256
-    ext.put_u16(0x0804); // rsa_pss_rsae_sha256
-    ext.put_u16(0x0401); // rsa_pkcs1_sha256
-    ext.put_u16(0x0503); // ecdsa_secp384r1_sha384
-
-    let mut handshake = BytesMut::new();
-    handshake.put_u16(0x0303); // Client Version
-    let mut random = [0u8; 32];
-    OsRng.fill_bytes(&mut random);
-    handshake.put_slice(&random); // Random
-    
-    handshake.put_u8(32); // Session ID length
-    handshake.put_slice(session_id); // Session ID
-    
-    // Cipher Suites
-    handshake.put_u16(6); // Length
-    handshake.put_u16(0x1301); // TLS_AES_128_GCM_SHA256
-    handshake.put_u16(0x1303); // TLS_CHACHA20_POLY1305_SHA256
-    handshake.put_u16(0x1302); // TLS_AES_256_GCM_SHA384
-    
-    // Compression
-    handshake.put_u8(1); // Length
-    handshake.put_u8(0); // null
-    
-    // Extensions
-    handshake.put_u16(ext.len() as u16);
-    handshake.put_slice(&ext);
-    
-    let handshake_len = handshake.len();
-    
-    let mut record = BytesMut::new();
-    record.put_u8(0x16); // Handshake
-    record.put_u16(0x0301); // TLS 1.0 (Compatibility)
-    record.put_u16((handshake_len + HANDSHAKE_HEADER_LEN) as u16); // Length
-    
-    record.put_u8(0x01); // ClientHello
-    record.put_u8((handshake_len >> 16) as u8);
-    record.put_u8((handshake_len >> 8) as u8);
-    record.put_u8(handshake_len as u8);
-    record.put_slice(&handshake);
-
-    // Append ChangeCipherSpec for TLS 1.3 middlebox compatibility (RFC 8446 §D.4)
-    // This makes the flow look like: ClientHello → ServerHello → CCS → AppData
-    // instead of the DPI-suspicious: ClientHello → AppData directly.
-    let mut out = BytesMut::new();
-    out.put_slice(&record);
-    out.put_slice(&[0x14, 0x03, 0x03, 0x00, 0x01, 0x01]);
-    out.freeze()
-}
-
-pub struct ParsedClientHello {
-    pub sni: String,
-    pub session_id: [u8; 32],
-    pub c_pub: PublicKey,
-}
-
-/// Parses a TLS ClientHello. Returns None if invalid or missing required fields.
-pub fn parse_client_hello(mut buf: &[u8]) -> Option<ParsedClientHello> {
-    if buf.len() < RECORD_HEADER_LEN + HANDSHAKE_HEADER_LEN {
-        return None;
-    }
-    
-    // Record Header
-    let typ = buf.get_u8();
-    if typ != 0x16 { return None; } // Not a handshake
-    let _version = buf.get_u16();
-    let record_len = buf.get_u16() as usize;
-    
-    if buf.len() < record_len {
-        return None; // Incomplete record
-    }
-    
-    let mut payload = &buf[..record_len];
-    
-    // Handshake Header
-    let hs_type = payload.get_u8();
-    if hs_type != 0x01 { return None; } // Not ClientHello
-    let hs_len_hi = payload.get_u8() as usize;
-    let hs_len_mid = payload.get_u8() as usize;
-    let hs_len_lo = payload.get_u8() as usize;
-    let hs_len = (hs_len_hi << 16) | (hs_len_mid << 8) | hs_len_lo;
-    
-    if payload.len() < hs_len { return None; }
-    
-    let mut ch = &payload[..hs_len];
-    let _client_version = ch.get_u16();
-    if ch.len() < 32 { return None; }
-    ch.advance(32); // Skip Random
-    
-    let sid_len = ch.get_u8() as usize;
-    if sid_len != 32 || ch.len() < 32 { return None; }
-    
-    let mut session_id = [0u8; 32];
-    session_id.copy_from_slice(&ch[..32]);
-    ch.advance(32);
-    
-    let ciphers_len = ch.get_u16() as usize;
-    if ch.len() < ciphers_len { return None; }
-    ch.advance(ciphers_len);
-    
-    let comp_len = ch.get_u8() as usize;
-    if ch.len() < comp_len { return None; }
-    ch.advance(comp_len);
-    
-    let ext_len = ch.get_u16() as usize;
-    if ch.len() < ext_len { return None; }
-    
-    let mut exts = &ch[..ext_len];
-    
-    let mut parsed_sni = None;
-    let mut parsed_c_pub = None;
-    
-    while exts.len() >= 4 {
-        let ext_type = exts.get_u16();
-        let ext_len = exts.get_u16() as usize;
-        if exts.len() < ext_len { break; }
-        
-        let mut ext_data = &exts[..ext_len];
-        
-        if ext_type == 0x0000 { // SNI
-            let _list_len = ext_data.get_u16() as usize;
-            if ext_data.len() >= 3 {
-                let name_type = ext_data.get_u8();
-                if name_type == 0x00 { // Hostname
-                    let name_len = ext_data.get_u16() as usize;
-                    if ext_data.len() >= name_len {
-                        if let Ok(name) = std::str::from_utf8(&ext_data[..name_len]) {
-                            parsed_sni = Some(name.to_string());
-                        }
-                    }
-                }
-            }
-        } else if ext_type == 0x0033 { // Key Share
-            let _client_shares_len = ext_data.get_u16() as usize;
-            while ext_data.len() >= 4 {
-                let group = ext_data.get_u16();
-                let key_ex_len = ext_data.get_u16() as usize;
-                if ext_data.len() < key_ex_len { break; }
-                
-                if group == 0x001d && key_ex_len == 32 { // X25519
-                    let mut pub_bytes = [0u8; 32];
-                    pub_bytes.copy_from_slice(&ext_data[..32]);
-                    parsed_c_pub = Some(PublicKey::from(pub_bytes));
-                }
-                ext_data.advance(key_ex_len);
-            }
-        }
-        
-        exts.advance(ext_len);
-    }
-    
-    match (parsed_sni, parsed_c_pub) {
-        (Some(sni), Some(c_pub)) => Some(ParsedClientHello { sni, session_id, c_pub }),
-        _ => None,
-    }
-}

ostp-core/src/framing/frame.rs

@@ -13,8 +13,6 @@ pub enum FrameKind {
     KeepAlive = 4,
     Nack = 5,
     Ack = 6,
-    /// 0-RTT session resumption: client sends ticket + early data
-    Resume = 7,
 }
 
 impl TryFrom<u8> for FrameKind {
@@ -28,7 +26,6 @@ impl TryFrom<u8> for FrameKind {
             4 => Ok(Self::KeepAlive),
             5 => Ok(Self::Nack),
             6 => Ok(Self::Ack),
-            7 => Ok(Self::Resume),
             _ => Err(ProtocolError::Framing("unknown frame kind".to_string())),
         }
     }
@@ -104,7 +101,15 @@ impl FramedPacket {
         let payload_len = header.payload_len as usize;
         let pad_len = header.pad_len as usize;
 
-        let expected = FRAME_HEADER_LEN + payload_len + pad_len;
+        // Use checked arithmetic: payload_len is a u32 from the (decrypted, but
+        // still to-be-trusted) header, and on 32-bit targets — MIPS/ARMv7
+        // routers are supported build targets — header+payload+pad can overflow
+        // usize and wrap to a small value that spuriously passes the length
+        // check, causing an out-of-range slice below.
+        let expected = FRAME_HEADER_LEN
+            .checked_add(payload_len)
+            .and_then(|v| v.checked_add(pad_len))
+            .ok_or_else(|| ProtocolError::Framing("frame length overflow".to_string()))?;
         if buf.len() < expected {
             return Err(ProtocolError::Framing("frame body truncated".to_string()));
         }

ostp-core/src/framing/mod.rs

@@ -1,7 +1,5 @@
 pub mod frame;
 pub mod padding;
-pub mod wss;
 
 pub use frame::{FrameHeader, FrameKind, FramedPacket};
 pub use padding::{AdaptivePadder, PaddingStrategy, TrafficProfile};
-pub use wss::{encode_wss_frame, decode_wss_frame, WssFrameResult};

ostp-core/src/framing/wss.rs

@@ -1,74 +0,0 @@
-use rand::RngCore;
-
-pub enum WssFrameResult {
-    Incomplete,
-    Frame { payload: Vec<u8>, total_len: usize },
-}
-
-pub fn encode_wss_frame(payload: &[u8], masked: bool) -> Vec<u8> {
-    let len = payload.len();
-    let mut header = Vec::with_capacity(14 + len);
-    header.push(0x82); // FIN + Binary
-    
-    let mask_bit = if masked { 0x80 } else { 0x00 };
-    
-    if len <= 125 {
-        header.push(mask_bit | (len as u8));
-    } else if len <= 65535 {
-        header.push(mask_bit | 126);
-        header.extend_from_slice(&(len as u16).to_be_bytes());
-    } else {
-        header.push(mask_bit | 127);
-        header.extend_from_slice(&(len as u64).to_be_bytes());
-    }
-    
-    if masked {
-        let mut mask = [0u8; 4];
-        rand::thread_rng().fill_bytes(&mut mask);
-        header.extend_from_slice(&mask);
-        
-        for (i, &b) in payload.iter().enumerate() {
-            header.push(b ^ mask[i % 4]);
-        }
-    } else {
-        header.extend_from_slice(payload);
-    }
-    
-    header
-}
-
-pub fn decode_wss_frame(buffer: &[u8]) -> WssFrameResult {
-    if buffer.len() < 2 {
-        return WssFrameResult::Incomplete;
-    }
-    let is_masked = (buffer[1] & 0x80) != 0;
-    let payload_len_7 = (buffer[1] & 0x7F) as usize;
-    
-    let (header_len, payload_len) = if payload_len_7 == 126 {
-        if buffer.len() < 4 { return WssFrameResult::Incomplete; }
-        (4, u16::from_be_bytes([buffer[2], buffer[3]]) as usize)
-    } else if payload_len_7 == 127 {
-        if buffer.len() < 10 { return WssFrameResult::Incomplete; }
-        (10, u64::from_be_bytes([buffer[2], buffer[3], buffer[4], buffer[5], buffer[6], buffer[7], buffer[8], buffer[9]]) as usize)
-    } else {
-        (2, payload_len_7)
-    };
-    
-    let mask_offset = header_len;
-    let full_header_len = header_len + if is_masked { 4 } else { 0 };
-    let total_frame_len = full_header_len + payload_len;
-    
-    if buffer.len() < total_frame_len {
-        return WssFrameResult::Incomplete;
-    }
-    
-    let mut payload = buffer[full_header_len..total_frame_len].to_vec();
-    if is_masked {
-        let mask = [buffer[mask_offset], buffer[mask_offset+1], buffer[mask_offset+2], buffer[mask_offset+3]];
-        for (i, b) in payload.iter_mut().enumerate() {
-            *b ^= mask[i % 4];
-        }
-    }
-    
-    WssFrameResult::Frame { payload, total_len: total_frame_len }
-}

ostp-core/src/lib.rs

@@ -3,7 +3,6 @@ pub mod crypto;
 pub mod framing;
 pub mod protocol;
 pub mod relay;
-pub mod resumption;
 
 pub use crypto::NoiseRole;
 pub use framing::{TrafficProfile, PaddingStrategy};

ostp-core/src/protocol.rs

@@ -1,10 +1,14 @@
 use bytes::Bytes;
 use rand::Rng;
-use sha2::{Digest, Sha256};
 use thiserror::Error;
 use std::collections::{BTreeMap, VecDeque};
 use std::time::{Duration, Instant};
 
+/// Upper bound on a single frame's retransmit timer, after exponential backoff
+/// is applied to the adaptive RTO. Past this the session is dead from the
+/// user's point of view, and waiting longer only delays recovery.
+const MAX_EFFECTIVE_RTO: Duration = Duration::from_secs(8);
+
 use crate::congestion::CongestionController;
 use crate::crypto::{NoiseRole, NoiseSession, SessionCipher};
 use crate::framing::{AdaptivePadder, FrameHeader, FrameKind, FramedPacket, PaddingStrategy};
@@ -103,6 +107,17 @@ pub struct ProtocolMachine {
     _mtu: usize,
 }
 
+// ── Gap recovery (see `ProtocolMachine::recover_stalled_gap`) ────────────────
+// How long the receive sequence may sit stuck behind a missing frame, with
+// later frames already buffered, before that frame is declared unrecoverable
+// and skipped. Derived from the live RTO so it scales with the path instead of
+// guessing, then clamped: the floor keeps a fast link from discarding a frame
+// that is merely late, the ceiling bounds how long a stall can be visible to
+// the user before the tunnel unblocks itself.
+const GAP_RECOVERY_RTO_MULTIPLIER: u32 = 8;
+const GAP_RECOVERY_MIN: Duration = Duration::from_secs(2);
+const GAP_RECOVERY_MAX: Duration = Duration::from_secs(10);
+
 #[derive(Debug, Clone)]
 struct SentFrame {
     nonce: u64,
@@ -156,10 +171,37 @@ impl ProtocolMachine {
         self.sent_history.iter().filter(|f| f.is_retransmittable).count()
     }
 
+    /// Sum of retry counters across in-flight frames. Test-only: lets a test
+    /// assert the core retransmit invariant (a retry is only ever charged to a
+    /// frame that was actually put on the wire) without needing to advance the
+    /// clock through several seconds of exponential backoff.
+    #[cfg(test)]
+    fn total_retries(&self) -> usize {
+        self.sent_history
+            .iter()
+            .filter(|f| f.is_retransmittable)
+            .map(|f| f.retries as usize)
+            .sum()
+    }
+
     pub fn cwnd_packets(&self) -> usize {
         self.cc.cwnd_packets() as usize
     }
 
+    /// Whether the pacing bucket currently allows releasing another packet.
+    ///
+    /// The congestion window bounds how much may be UNACKNOWLEDGED; it says
+    /// nothing about how fast that window is emptied onto the wire. Sending a
+    /// whole window back-to-back is what drives a deep buffer into standing
+    /// queue, so admission is gated on both.
+    pub fn can_pace_packet(&self) -> bool {
+        self.cc.can_pace_packet()
+    }
+
+    pub fn on_send(&mut self, bytes: u64) {
+        self.cc.on_send(bytes);
+    }
+
     pub fn state(&self) -> OstpState {
         self.state
     }
@@ -203,13 +245,16 @@ impl ProtocolMachine {
                     .map(ProtocolAction::SendDatagram)
             }
             (OstpState::Closing, OstpEvent::Inbound(raw)) => {
-                // Process final in-flight packets to prevent data loss during teardown.
-                // The remote may still have data or ACKs in transit when we initiated Close.
-                let result = self.handle_inbound(raw);
-                self.state = OstpState::Closed;
-                result
+                // The remote may still have data or ACKs in transit when we initiated
+                // Close. Stay in Closing and process them; handle_inbound transitions to
+                // Closed only when it actually receives the peer's Close frame — the old
+                // code force-closed after a single inbound packet, losing in-flight data.
+                // (Ported from 0.3.x 47d44fa.)
+                self.handle_inbound(raw)
             }
             (OstpState::Established, OstpEvent::Tick) => self.handle_tick(),
+            // Retransmit our Close frame (and drain pending) while waiting for teardown.
+            (OstpState::Closing, OstpEvent::Tick) => self.handle_tick(),
             (OstpState::Closed, _) => Ok(ProtocolAction::Noop),
             (_, OstpEvent::Close) => {
                 self.state = OstpState::Closed;
@@ -230,7 +275,9 @@ impl ProtocolMachine {
 
         let session_id = u32::from_be_bytes([raw_vec[0], raw_vec[1], raw_vec[2], raw_vec[3]]);
         if session_id != self.session_id {
-            tracing::error!("session id mismatch! expected={:#010x}, got={:#010x}, is_handshake={}, raw_len={}", self.session_id, session_id, is_handshake, raw_vec.len());
+            // Per-packet, attacker-triggerable event: keep at debug and don't
+            // dump internal session ids (log-flood + info-leak surface).
+            tracing::debug!("session id mismatch (is_handshake={})", is_handshake);
             return Err(ProtocolError::State("session id mismatch".to_string()));
         }
 
@@ -256,8 +303,7 @@ impl ProtocolMachine {
                 noise_len, raw_vec.len() - 6
             )));
         }
-        tracing::info!("handle_inbound: raw_vec.len()={}, noise_len={}, raw_vec[0..6]={:?}", raw_vec.len(), noise_len, &raw_vec[0..6]);
-        
+
         let mut read_out = vec![0_u8; 1024];
         let n = self.noise.read_handshake(&raw_vec[6..6 + noise_len], &mut read_out).map_err(|e| {
             ProtocolError::Crypto(format!("noise-read: {:?} (raw_len={}, noise_len={})", e, raw_vec.len(), noise_len))
@@ -274,9 +320,12 @@ impl ProtocolMachine {
             NoiseRole::Initiator => None,
         };
 
-        let mut key = [0_u8; 32];
-        self.noise.handshake_hash(&mut key)?;
-        let (send_key, recv_key) = derive_split_keys(&key, self.role);
+        // Transport keys come from Noise's Split() over the final chaining key,
+        // so they depend on the ephemeral `ee` DH secret and give the session
+        // forward secrecy. (Previously these were derived from the handshake
+        // hash, which never absorbs the DH result — see raw_split's SECURITY
+        // note. That is the wire-breaking change gated by PROTOCOL_VERSION.)
+        let (send_key, recv_key) = self.noise.raw_split(self.role)?;
         self.send_cipher = Some(SessionCipher::new(&send_key));
         self.recv_cipher = Some(SessionCipher::new(&recv_key));
         self.state = OstpState::Established;
@@ -286,7 +335,107 @@ impl ProtocolMachine {
         Ok(ProtocolAction::HandshakePayload(Bytes::from(extracted_payload), response))
     }
 
+    /// Restores liveness when the receive sequence is stuck behind a frame that
+    /// can never arrive.
+    ///
+    /// Delivery is gated on `expected_recv_nonce`, so a single missing frame
+    /// holds back every later frame. That is correct *while the sender can still
+    /// retransmit* — but the sender drops a frame from `sent_history` once it
+    /// exceeds `max_retries + 2` attempts (see the zombie eviction in
+    /// `handle_tick`). After that the frame is gone for good and the two sides
+    /// deadlock: the receiver buffers forever and NACKs a nonce nobody can
+    /// resend.
+    ///
+    /// That deadlock is invisible to the keepalive watchdog, which is why it
+    /// presented as a hard freeze rather than a reconnect: retransmits, ACKs and
+    /// NACKs keep flowing, so the client's `last_valid_recv` keeps refreshing and
+    /// its stall detector never fires. The RTT readout freezes at its last value
+    /// for the same reason — Pong rides in a Data frame stuck behind the gap.
+    ///
+    /// So: once we have been stuck long enough that retransmission has provably
+    /// given up, skip to the lowest buffered nonce and drain. This drops the
+    /// missing frame's payload (one RelayMessage — a chunk of one stream), which
+    /// is a real cost, but the alternative is a permanently dead tunnel.
+    fn recover_stalled_gap(&mut self) -> Vec<ProtocolAction> {
+        let mut recovered = Vec::new();
+        if self.reorder_buffer.is_empty() {
+            return recovered;
+        }
+
+        // Wait out the sender's full retransmit budget before giving up, so a
+        // frame that is merely late is never discarded. The sender backs off
+        // exponentially, so key this off the live RTO estimate rather than a
+        // flat constant, with a floor that keeps low-RTT links from skipping
+        // too eagerly and a ceiling that bounds the visible freeze.
+        let timeout = self
+            .cc
+            .rto()
+            .saturating_mul(GAP_RECOVERY_RTO_MULTIPLIER)
+            .clamp(GAP_RECOVERY_MIN, GAP_RECOVERY_MAX);
+        if self.last_recv_advance.elapsed() < timeout {
+            return recovered;
+        }
+
+        let Some(&resume_at) = self.reorder_buffer.keys().next() else {
+            return recovered;
+        };
+        let skipped = resume_at.saturating_sub(self.expected_recv_nonce);
+        tracing::warn!(
+            "Gap recovery: no progress for {:?}; skipping {} unrecoverable frame(s) \
+             (nonce {} -> {}) to unblock the session",
+            self.last_recv_advance.elapsed(),
+            skipped,
+            self.expected_recv_nonce,
+            resume_at
+        );
+
+        self.expected_recv_nonce = resume_at;
+        while let Some(buffered) = self.reorder_buffer.remove(&self.expected_recv_nonce) {
+            recovered.push(buffered);
+            match self.expected_recv_nonce.checked_add(1) {
+                Some(next) => self.expected_recv_nonce = next,
+                // u64 nonce space exhausted: stop draining rather than wrap.
+                // The session is finished either way; the caller's next decrypt
+                // will fail and tear it down.
+                None => break,
+            }
+        }
+        self.last_recv_advance = Instant::now();
+        // The peer must learn the sequence moved on, or it will keep
+        // retransmitting into the void.
+        self.ack_pending = true;
+
+        recovered
+    }
+
     fn handle_data_inbound(&mut self, raw_vec: &[u8]) -> Result<ProtocolAction, ProtocolError> {
+        // Check for a stalled gap before classifying this frame, so the rest of
+        // the function sees an already-advanced `expected_recv_nonce`. Runs here
+        // rather than on Tick because both tick handlers discard DeliverApp
+        // actions, and because inbound frames keep arriving throughout the stall
+        // (retransmits/ACKs/NACKs/keepalives) — so this path is reliably reached.
+        let recovered = self.recover_stalled_gap();
+        let result = self.handle_data_inbound_frame(raw_vec)?;
+        if recovered.is_empty() {
+            return Ok(result);
+        }
+
+        // Recovered payloads are older than anything this frame produces, so
+        // they go first to preserve delivery order.
+        let mut all = recovered;
+        match result {
+            ProtocolAction::Noop => {}
+            ProtocolAction::Multiple(list) => all.extend(list),
+            single => all.push(single),
+        }
+        Ok(if all.len() == 1 {
+            all.pop().unwrap()
+        } else {
+            ProtocolAction::Multiple(all)
+        })
+    }
+
+    fn handle_data_inbound_frame(&mut self, raw_vec: &[u8]) -> Result<ProtocolAction, ProtocolError> {
         if raw_vec.len() < 12 {
             return Err(ProtocolError::Framing("data datagram too short".to_string()));
         }
@@ -351,13 +500,8 @@ impl ProtocolMachine {
             FrameKind::Data => {
                 ProtocolAction::DeliverApp(packet.header.stream_id, packet.payload)
             }
-            FrameKind::Resume => {
-                // 0-RTT: treat early data as application data
-                tracing::info!("0-RTT Resume frame received, processing early data");
-                ProtocolAction::DeliverApp(packet.header.stream_id, packet.payload)
-            }
             FrameKind::Close => {
-                tracing::info!("Received Close frame, terminating session");
+                tracing::debug!("Received Close frame, terminating session");
                 self.state = OstpState::Closed;
                 ProtocolAction::Noop
             }
@@ -388,18 +532,20 @@ impl ProtocolMachine {
             self.last_recv_advance = Instant::now();
         } else {
             // Gap detected
-            if self.reorder_buffer.len() < self.max_reorder_buffer {
-                self.reorder_buffer.insert(nonce, action);
+            if nonce >= self.expected_recv_nonce {
+                if self.reorder_buffer.len() < self.max_reorder_buffer {
+                    self.reorder_buffer.insert(nonce, action);
+                } else {
+                    tracing::warn!("Reorder buffer still full after gap recovery, dropping frame nonce={}", nonce);
+                }
             } else {
-                tracing::warn!("Reorder buffer full ({}/{}), dropping frame nonce={}",
-                    self.reorder_buffer.len(), self.max_reorder_buffer, nonce
-                );
+                tracing::debug!("Frame nonce={} arrived too late after gap recovery, dropping", nonce);
             }
 
-            // Rate-limited NACK: send at most once per 30ms to prevent retransmit storms.
-            // Under high load with natural UDP reordering, sending a NACK per packet
-            // causes exponential retransmit explosion that saturates the channel.
-            let nack_cooldown = Duration::from_millis(30);
+            // Rate-limited NACK: send at most once per (rto/2) to prevent retransmit storms.
+            // Using rto/2 means we send a NACK before the sender's timer fires, prompting
+            // fast retransmit without flooding. Floor at 10ms to handle very low-RTT links.
+            let nack_cooldown = (self.cc.rto() / 2).max(Duration::from_millis(10));
             if self.last_nack_sent.elapsed() >= nack_cooldown {
                 self.last_nack_sent = Instant::now();
                 let nack_payload = self.expected_recv_nonce.to_be_bytes();
@@ -507,44 +653,18 @@ impl ProtocolMachine {
     fn handle_tick(&mut self) -> Result<ProtocolAction, ProtocolError> {
         let mut actions = Vec::new();
 
-        // ── Gap Recovery ──────────────────────────────────────────────
-        // If expected_recv_nonce hasn't advanced for 500ms+ and there
-        // are buffered frames waiting, the sender likely evicted the lost
-        // frame from sent_history. Skip the gap to restore data flow.
-        // This trades a small amount of data loss for connection liveness.
-        if !self.reorder_buffer.is_empty()
-            && self.last_recv_advance.elapsed() > Duration::from_millis(500)
-        {
-            if let Some(&first_buffered) = self.reorder_buffer.keys().next() {
-                let skipped = first_buffered.saturating_sub(self.expected_recv_nonce);
-                self.expected_recv_nonce = first_buffered;
-                self.last_recv_advance = Instant::now();
-
-                let mut delivered = 0u64;
-                while let Some(buffered_action) = self.reorder_buffer.remove(&self.expected_recv_nonce) {
-                    actions.push(buffered_action);
-                    self.expected_recv_nonce = self.expected_recv_nonce.saturating_add(1);
-                    delivered += 1;
-                }
-                self.ack_pending = true;
-                tracing::debug!("Gap recovery: skipped {} lost frames, delivered {} buffered frames (reorder_buf={})",
-                    skipped, delivered, self.reorder_buffer.len()
-                );
-            }
-        }
-
         // ── Pending ACK flush ─────────────────────────────────────────
         if let Some(ack_frame) = self.build_ack_if_due()? {
             actions.push(ProtocolAction::SendDatagram(ack_frame));
         }
 
         let now = Instant::now();
-        let base_rto_ms = self.rto.as_millis().max(1) as u64;
+        // Use the adaptive RTO from the congestion controller (RFC 6298 SRTT + 4*RTTVAR).
+        // Falls back to rto_initial before the first ACK is received.
+        let base_rto_ms = self.cc.rto().max(self.rto).as_millis().max(1) as u64;
 
         // ── Zombie frame eviction ────────────────────────────────────
         // Evict frames that exceeded max_retries + 2 grace retries.
-        // Shorter grace period than before (was +4) to free memory faster
-        // after high-throughput bursts.
         let grace = self.max_retries.saturating_add(2);
         let before = self.sent_history.len();
         self.sent_history.retain(|f| !f.is_retransmittable || f.retries <= grace);
@@ -555,24 +675,46 @@ impl ProtocolMachine {
 
         // ── Retransmit expired frames ────────────────────────────────
         // Limit retransmits per tick to prevent bandwidth saturation
+        // Backoff starts from retry #0 (immediately effective):
+        //   effective_rto = base_rto * 2^retries, capped at 2^6 = 64×
         let mut retransmit_budget: usize = self.cc.retransmit_budget();
         for frame in self.sent_history.iter_mut() {
             if !frame.is_retransmittable {
                 continue;
             }
+            // Out of budget for this tick — stop scanning rather than walking the
+            // rest of the queue. sent_history is in send order, so everything we
+            // skip is strictly newer than what we already handled; deferring it to
+            // the next tick preserves oldest-first retransmit priority.
+            if retransmit_budget == 0 {
+                break;
+            }
 
-            let retry_over = frame.retries.saturating_sub(self.max_retries);
-            let backoff_factor = 1u64 << retry_over.min(6);
-            let effective_rto = Duration::from_millis(base_rto_ms.saturating_mul(backoff_factor));
+            // Exponential backoff, but bounded in absolute terms. base_rto is
+            // itself adaptive and can reach RTO_MAX (16s) on a congested path;
+            // multiplying that by the 64x backoff cap yields a frame that sits
+            // unretransmitted for ~17 MINUTES, long past the point where the
+            // session is simply dead to the user. Cap the product so backoff
+            // stays a backoff rather than an outage.
+            let backoff_factor = 1u64 << (frame.retries as u64).min(6);
+            let effective_rto = Duration::from_millis(base_rto_ms.saturating_mul(backoff_factor))
+                .min(MAX_EFFECTIVE_RTO);
 
             if now.duration_since(frame.last_sent) >= effective_rto {
+                // Only burn the retry counter and reset the RTO timer when the
+                // frame is ACTUALLY put on the wire. Doing it unconditionally
+                // meant that whenever the per-tick budget ran out — which is
+                // exactly when loss is heavy and retransmits matter most —
+                // frames accumulated "phantom retries" they never actually got,
+                // and the zombie eviction above then silently dropped them after
+                // `grace` such rounds. The peer never received that data and
+                // never would: that stream stalls forever while the session
+                // itself stays healthy, which is precisely the reported "tunnel
+                // frozen at 0 b/s but the session still up" symptom.
                 frame.last_sent = now;
                 frame.retries = frame.retries.saturating_add(1);
-                
-                if retransmit_budget > 0 {
-                    actions.push(ProtocolAction::SendDatagram(frame.bytes.clone()));
-                    retransmit_budget -= 1;
-                }
+                actions.push(ProtocolAction::SendDatagram(frame.bytes.clone()));
+                retransmit_budget -= 1;
             }
         }
 
@@ -677,6 +819,9 @@ impl ProtocolMachine {
     }
 
     fn push_sent_frame(&mut self, nonce: u64, bytes: Bytes, is_retransmittable: bool) {
+        if is_retransmittable {
+            self.cc.on_send(bytes.len() as u64);
+        }
         self.sent_history.push_back(SentFrame {
             nonce,
             bytes,
@@ -698,24 +843,34 @@ impl ProtocolMachine {
     fn drop_acked_frames(&mut self, ranges: &[(u64, u64)]) {
         let now = Instant::now();
         let mut acked_bytes = 0u64;
-        let mut min_rtt = Duration::from_secs(60);
+        let mut min_rtt: Option<Duration> = None;
 
-        // Compute RTT from the oldest acked frame's send timestamp
         for frame in self.sent_history.iter() {
             if nonce_in_ranges(frame.nonce, ranges) {
                 acked_bytes += frame.bytes.len() as u64;
-                let rtt = now.duration_since(frame.last_sent);
-                if rtt < min_rtt {
-                    min_rtt = rtt;
+                // Karn's algorithm: never take an RTT sample from a frame that
+                // was retransmitted. `last_sent` is bumped on every retransmit,
+                // so an ACK for the ORIGINAL transmission would be measured
+                // against the retransmit time, yielding a spuriously small RTT
+                // that drags SRTT/RTO down and triggers more spurious
+                // retransmits. Only unambiguous (never-retried) frames qualify.
+                if frame.retries == 0 {
+                    let rtt = now.duration_since(frame.last_sent);
+                    min_rtt = Some(min_rtt.map_or(rtt, |m| m.min(rtt)));
                 }
             }
         }
 
         self.sent_history.retain(|frame| !nonce_in_ranges(frame.nonce, ranges));
 
-        // Notify congestion controller
+        // Notify congestion controller. Feed an RTT sample only when we had at
+        // least one unambiguous ACK; otherwise update the window without
+        // polluting the RTT estimator.
         if acked_bytes > 0 {
-            self.cc.on_ack(acked_bytes, min_rtt);
+            match min_rtt {
+                Some(rtt) => self.cc.on_ack(acked_bytes, rtt),
+                None => self.cc.on_ack_no_rtt(acked_bytes),
+            }
         }
     }
 }
@@ -745,26 +900,6 @@ fn nonce_in_ranges(nonce: u64, ranges: &[(u64, u64)]) -> bool {
     ranges.iter().any(|(start, end)| nonce >= *start && nonce <= *end)
 }
 
-fn derive_split_keys(base_key: &[u8; 32], role: NoiseRole) -> ([u8; 32], [u8; 32]) {
-    let mut initiator_key = [0u8; 32];
-    let mut responder_key = [0u8; 32];
-
-    let mut h1 = Sha256::new();
-    h1.update(base_key);
-    h1.update(b"ostp-initiator");
-    initiator_key.copy_from_slice(&h1.finalize());
-
-    let mut h2 = Sha256::new();
-    h2.update(base_key);
-    h2.update(b"ostp-responder");
-    responder_key.copy_from_slice(&h2.finalize());
-
-    match role {
-        NoiseRole::Initiator => (initiator_key, responder_key),
-        NoiseRole::Responder => (responder_key, initiator_key),
-    }
-}
-
 #[cfg(test)]
 mod tests {
     use super::*;
@@ -996,4 +1131,154 @@ mod tests {
         let _ = client.on_event(OstpEvent::Tick).unwrap();
         let _ = server.on_event(OstpEvent::Tick).unwrap();
     }
+
+    /// A retry may only be charged to a frame that was actually retransmitted.
+    ///
+    /// The retransmit loop is budget-limited per tick. It used to bump
+    /// `retries` and reset `last_sent` for every due frame regardless of
+    /// whether the budget allowed it to actually send — so under heavy loss
+    /// (exactly when the budget runs out) frames racked up retries they never
+    /// received, and the zombie eviction dropped them after `max_retries + 2`
+    /// such rounds. That data was never delivered and never would be: the
+    /// stream stalls permanently while the session itself stays up.
+    #[test]
+    fn test_retransmit_budget_charges_retries_only_for_frames_actually_sent() {
+        let (mut client, _server) = do_handshake();
+
+        // Queue far more in-flight frames than a single tick's budget allows.
+        const FRAMES: usize = 40;
+        for i in 0..FRAMES {
+            let payload = Bytes::from(vec![i as u8; 200]);
+            client.on_event(OstpEvent::Outbound(1, payload)).unwrap();
+        }
+        assert_eq!(client.in_flight_count(), FRAMES);
+        assert_eq!(client.total_retries(), 0, "nothing retransmitted yet");
+
+        // Let every frame's RTO lapse so that on the next tick all FRAMES frames
+        // are due at once and the per-tick budget is guaranteed to run out. The
+        // effective RTO here is max(cc.rto(), config rto_ms) = 100ms at retries=0.
+        std::thread::sleep(Duration::from_millis(150));
+
+        let sent = count_datagrams(&client.on_event(OstpEvent::Tick).unwrap());
+
+        assert!(sent > 0, "expected some retransmits after the RTO lapsed");
+        assert!(
+            sent < FRAMES,
+            "budget should have capped this tick below the {FRAMES} due frames, got {sent}"
+        );
+        assert_eq!(
+            client.total_retries(),
+            sent,
+            "charged {} retries but only put {} frames on the wire — the \
+             difference is phantom retries that will silently evict live data",
+            client.total_retries(),
+            sent
+        );
+        assert_eq!(
+            client.in_flight_count(),
+            FRAMES,
+            "nothing was acked, so no frame may be evicted yet"
+        );
+    }
+
+    /// Count how many datagrams an action tree actually puts on the wire.
+    fn count_datagrams(action: &ProtocolAction) -> usize {
+        match action {
+            ProtocolAction::SendDatagram(_) => 1,
+            ProtocolAction::Multiple(list) => list.iter().map(count_datagrams).sum(),
+            _ => 0,
+        }
+    }
+
+    /// Count how many application payloads an action tree actually delivers.
+    fn delivered_payloads(action: &ProtocolAction) -> Vec<Bytes> {
+        match action {
+            ProtocolAction::DeliverApp(_, data) => vec![data.clone()],
+            ProtocolAction::Multiple(list) => list.iter().flat_map(delivered_payloads).collect(),
+            _ => Vec::new(),
+        }
+    }
+
+    /// Build `count` data frames on `client`, returning them without delivering
+    /// any — lets a test choose which ones to "lose" in transit.
+    fn make_data_frames(client: &mut ProtocolMachine, count: u8) -> Vec<Bytes> {
+        (0..count)
+            .map(|i| {
+                let payload = Bytes::from(vec![i; 32]);
+                match client.on_event(OstpEvent::Outbound(1, payload)).unwrap() {
+                    ProtocolAction::SendDatagram(d) => d,
+                    _ => panic!("expected SendDatagram for frame {i}"),
+                }
+            })
+            .collect()
+    }
+
+    /// The freeze this fixes: a frame is lost, the sender eventually stops
+    /// retransmitting it, and the receiver — which gates delivery on
+    /// `expected_recv_nonce` — waits for it forever. Every later frame piles up
+    /// undelivered while the transport itself stays healthy, so nothing upstream
+    /// notices. Recovery must eventually skip the hole and release the backlog.
+    #[test]
+    fn test_gap_recovery_releases_permanently_stalled_frames() {
+        let (mut client, mut server) = do_handshake();
+        let frames = make_data_frames(&mut client, 4);
+
+        // Frame 0 arrives in order and is delivered straight through.
+        let action = server.on_event(OstpEvent::Inbound(frames[0].clone())).unwrap();
+        assert_eq!(delivered_payloads(&action).len(), 1, "in-order frame should deliver");
+
+        // Frame 1 is lost. 2 and 3 arrive but must be held back — delivering them
+        // now would reorder the stream.
+        for idx in [2usize, 3] {
+            let action = server.on_event(OstpEvent::Inbound(frames[idx].clone())).unwrap();
+            assert!(
+                delivered_payloads(&action).is_empty(),
+                "frame {idx} must stay buffered behind the missing frame"
+            );
+        }
+
+        // Stand in for "the sender exhausted its retries and dropped frame 1":
+        // the sequence has not advanced for longer than the recovery timeout.
+        server.last_recv_advance = Instant::now() - GAP_RECOVERY_MAX - Duration::from_secs(1);
+
+        // The next inbound frame (a retransmitted duplicate, which is exactly what
+        // a real stalled session keeps receiving) must unblock the backlog.
+        let action = server.on_event(OstpEvent::Inbound(frames[0].clone())).unwrap();
+        let delivered = delivered_payloads(&action);
+        assert_eq!(
+            delivered.len(),
+            2,
+            "both buffered frames must be released once the gap is declared unrecoverable"
+        );
+        // ...and in order: frame 2 before frame 3.
+        assert_eq!(delivered[0][0], 2);
+        assert_eq!(delivered[1][0], 3);
+    }
+
+    /// Recovery must not be trigger-happy: a frame that is merely late still has
+    /// to be waited for, or we would discard data the sender is about to resend.
+    #[test]
+    fn test_gap_recovery_does_not_fire_before_timeout() {
+        let (mut client, mut server) = do_handshake();
+        let frames = make_data_frames(&mut client, 3);
+
+        server.on_event(OstpEvent::Inbound(frames[0].clone())).unwrap();
+        let action = server.on_event(OstpEvent::Inbound(frames[2].clone())).unwrap();
+        assert!(delivered_payloads(&action).is_empty());
+
+        // Well inside the timeout — the gap must still be respected.
+        let action = server.on_event(OstpEvent::Inbound(frames[0].clone())).unwrap();
+        assert!(
+            delivered_payloads(&action).is_empty(),
+            "must keep waiting while retransmission is still plausible"
+        );
+
+        // And once the genuinely-late frame shows up, normal in-order delivery
+        // resumes with nothing dropped.
+        let action = server.on_event(OstpEvent::Inbound(frames[1].clone())).unwrap();
+        let delivered = delivered_payloads(&action);
+        assert_eq!(delivered.len(), 2, "late frame plus the buffered one");
+        assert_eq!(delivered[0][0], 1);
+        assert_eq!(delivered[1][0], 2);
+    }
 }

ostp-core/src/resumption.rs

@@ -1,307 +0,0 @@
-//! 0-RTT Session Resumption for OSTP.
-//!
-//! When a client has previously connected to a server, it can cache
-//! a "session ticket" that allows it to send encrypted data in the
-//! very first packet — eliminating the handshake round-trip entirely.
-//!
-//! How it works:
-//! 1. After a successful handshake, the server issues a SessionTicket
-//!    containing enough state to resume the session.
-//! 2. The client stores the ticket locally (encrypted with the PSK).
-//! 3. On reconnection, the client sends a ResumptionRequest with the
-//!    ticket + early data in the first packet.
-//! 4. The server validates the ticket and immediately begins processing
-//!    data, achieving 0-RTT.
-//!
-//! Security considerations:
-//! - Tickets have a TTL (default 3600s) to limit replay window.
-//! - The server maintains a ticket nonce set to prevent replay.
-//! - Early data is idempotent by protocol design (relay CONNECT is safe
-//!   because duplicate CONNECTs to the same target are no-ops).
-
-use std::collections::HashSet;
-use std::time::{Duration, SystemTime, UNIX_EPOCH};
-
-use sha2::{Digest, Sha256};
-
-/// A session ticket that allows 0-RTT resumption.
-#[derive(Debug, Clone)]
-pub struct SessionTicket {
-    /// Unique ticket identifier (prevents replay)
-    pub ticket_id: [u8; 16],
-    /// Server session ID to resume
-    pub session_id: u32,
-    /// Derived cipher key for early data
-    pub cipher_key: [u8; 32],
-    /// Timestamp of issuance (seconds since epoch)
-    pub issued_at: u64,
-    /// Time-to-live in seconds
-    pub ttl: u64,
-}
-
-/// Maximum ticket age (1 hour default)
-const DEFAULT_TICKET_TTL: u64 = 3600;
-/// Maximum tickets in the anti-replay set
-const MAX_REPLAY_SET: usize = 10000;
-
-impl SessionTicket {
-    /// Create a new session ticket from the transport key material.
-    pub fn new(session_id: u32, transport_key: &[u8; 32], psk: &[u8; 32]) -> Self {
-        let now = SystemTime::now()
-            .duration_since(UNIX_EPOCH)
-            .unwrap_or_default()
-            .as_secs();
-
-        // Derive ticket ID from key material + timestamp
-        let mut hasher = Sha256::new();
-        hasher.update(transport_key);
-        hasher.update(now.to_be_bytes());
-        hasher.update(b"ostp-ticket-id");
-        let hash = hasher.finalize();
-        let mut ticket_id = [0u8; 16];
-        ticket_id.copy_from_slice(&hash[..16]);
-
-        // Derive cipher key for early data from PSK + ticket
-        let mut key_hasher = Sha256::new();
-        key_hasher.update(psk);
-        key_hasher.update(ticket_id);
-        key_hasher.update(b"ostp-early-data-key");
-        let cipher_key_hash = key_hasher.finalize();
-        let mut cipher_key = [0u8; 32];
-        cipher_key.copy_from_slice(&cipher_key_hash);
-
-        Self {
-            ticket_id,
-            session_id,
-            cipher_key,
-            issued_at: now,
-            ttl: DEFAULT_TICKET_TTL,
-        }
-    }
-
-    /// Check if the ticket has expired.
-    pub fn is_expired(&self) -> bool {
-        let now = SystemTime::now()
-            .duration_since(UNIX_EPOCH)
-            .unwrap_or_default()
-            .as_secs();
-        now > self.issued_at + self.ttl
-    }
-
-    /// Serialize the ticket to bytes for storage/transmission.
-    /// Wire format: [ticket_id:16][session_id:4][cipher_key:32][issued_at:8][ttl:8]
-    pub fn to_bytes(&self) -> Vec<u8> {
-        let mut out = Vec::with_capacity(68);
-        out.extend_from_slice(&self.ticket_id);
-        out.extend_from_slice(&self.session_id.to_be_bytes());
-        out.extend_from_slice(&self.cipher_key);
-        out.extend_from_slice(&self.issued_at.to_be_bytes());
-        out.extend_from_slice(&self.ttl.to_be_bytes());
-        out
-    }
-
-    /// Deserialize a ticket from bytes.
-    pub fn from_bytes(data: &[u8]) -> Option<Self> {
-        if data.len() < 68 {
-            return None;
-        }
-        let mut ticket_id = [0u8; 16];
-        ticket_id.copy_from_slice(&data[0..16]);
-
-        let session_id = u32::from_be_bytes(data[16..20].try_into().ok()?);
-
-        let mut cipher_key = [0u8; 32];
-        cipher_key.copy_from_slice(&data[20..52]);
-
-        let issued_at = u64::from_be_bytes(data[52..60].try_into().ok()?);
-        let ttl = u64::from_be_bytes(data[60..68].try_into().ok()?);
-
-        Some(Self {
-            ticket_id,
-            session_id,
-            cipher_key,
-            issued_at,
-            ttl,
-        })
-    }
-
-    /// Encrypt the ticket with a PSK for client-side storage.
-    /// Uses a simple XOR cipher with HMAC-SHA256 derived key.
-    pub fn encrypt(&self, psk: &[u8; 32]) -> Vec<u8> {
-        let raw = self.to_bytes();
-        let mut enc_key_hasher = Sha256::new();
-        enc_key_hasher.update(psk);
-        enc_key_hasher.update(b"ostp-ticket-encryption");
-        let enc_key = enc_key_hasher.finalize();
-
-        let mut encrypted = raw.clone();
-        for (i, byte) in encrypted.iter_mut().enumerate() {
-            *byte ^= enc_key[i % 32];
-        }
-        encrypted
-    }
-
-    /// Decrypt a ticket from encrypted bytes.
-    pub fn decrypt(encrypted: &[u8], psk: &[u8; 32]) -> Option<Self> {
-        let mut enc_key_hasher = Sha256::new();
-        enc_key_hasher.update(psk);
-        enc_key_hasher.update(b"ostp-ticket-encryption");
-        let enc_key = enc_key_hasher.finalize();
-
-        let mut decrypted = encrypted.to_vec();
-        for (i, byte) in decrypted.iter_mut().enumerate() {
-            *byte ^= enc_key[i % 32];
-        }
-        Self::from_bytes(&decrypted)
-    }
-}
-
-/// Server-side anti-replay guard for session tickets.
-#[allow(dead_code)]
-pub struct TicketValidator {
-    /// Set of consumed ticket IDs (prevents replay)
-    consumed: HashSet<[u8; 16]>,
-    /// PSK for ticket validation
-    psk: [u8; 32],
-    /// Maximum age for tickets
-    max_age: Duration,
-}
-
-impl TicketValidator {
-    pub fn new(psk: [u8; 32]) -> Self {
-        Self {
-            consumed: HashSet::new(),
-            psk,
-            max_age: Duration::from_secs(DEFAULT_TICKET_TTL),
-        }
-    }
-
-    /// Validate a ticket from the client. Returns the ticket if valid,
-    /// or None if expired, replayed, or invalid.
-    pub fn validate(&mut self, encrypted_ticket: &[u8]) -> Option<SessionTicket> {
-        let ticket = SessionTicket::decrypt(encrypted_ticket, &self.psk)?;
-
-        // Check expiry
-        if ticket.is_expired() {
-            tracing::debug!("0-RTT ticket rejected: expired");
-            return None;
-        }
-
-        // Check replay
-        if self.consumed.contains(&ticket.ticket_id) {
-            tracing::warn!("0-RTT ticket rejected: replay detected");
-            return None;
-        }
-
-        // Accept and mark as consumed
-        self.consumed.insert(ticket.ticket_id);
-
-        // Garbage collection: remove old entries when set grows too large
-        if self.consumed.len() > MAX_REPLAY_SET {
-            // Simple strategy: clear the entire set. This is safe because
-            // expired tickets would fail the expiry check anyway.
-            self.consumed.clear();
-            self.consumed.insert(ticket.ticket_id);
-            tracing::debug!("0-RTT replay set cleared (overflow)");
-        }
-
-        tracing::debug!("0-RTT ticket accepted: session_id={}", ticket.session_id);
-        Some(ticket)
-    }
-
-    /// Issue a new ticket for a completed session.
-    pub fn issue_ticket(&self, session_id: u32, transport_key: &[u8; 32]) -> SessionTicket {
-        SessionTicket::new(session_id, transport_key, &self.psk)
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-
-    #[test]
-    fn test_ticket_serialize_roundtrip() {
-        let psk = [42u8; 32];
-        let key = [1u8; 32];
-        let ticket = SessionTicket::new(12345, &key, &psk);
-
-        let bytes = ticket.to_bytes();
-        let restored = SessionTicket::from_bytes(&bytes).unwrap();
-
-        assert_eq!(ticket.ticket_id, restored.ticket_id);
-        assert_eq!(ticket.session_id, restored.session_id);
-        assert_eq!(ticket.cipher_key, restored.cipher_key);
-        assert_eq!(ticket.issued_at, restored.issued_at);
-    }
-
-    #[test]
-    fn test_ticket_encrypt_decrypt() {
-        let psk = [42u8; 32];
-        let key = [1u8; 32];
-        let ticket = SessionTicket::new(99, &key, &psk);
-
-        let encrypted = ticket.encrypt(&psk);
-        let decrypted = SessionTicket::decrypt(&encrypted, &psk).unwrap();
-
-        assert_eq!(ticket.ticket_id, decrypted.ticket_id);
-        assert_eq!(ticket.session_id, decrypted.session_id);
-    }
-
-    #[test]
-    fn test_ticket_wrong_psk_fails() {
-        let psk = [42u8; 32];
-        let wrong_psk = [99u8; 32];
-        let key = [1u8; 32];
-        let ticket = SessionTicket::new(1, &key, &psk);
-        let encrypted = ticket.encrypt(&psk);
-
-        // Decrypting with wrong PSK produces garbage, from_bytes should
-        // still return Some but ticket_id won't match
-        let decrypted = SessionTicket::decrypt(&encrypted, &wrong_psk);
-        // It may parse but the data will be wrong
-        if let Some(d) = decrypted {
-            assert_ne!(d.ticket_id, ticket.ticket_id);
-        }
-    }
-
-    #[test]
-    fn test_ticket_not_expired() {
-        let psk = [42u8; 32];
-        let key = [1u8; 32];
-        let ticket = SessionTicket::new(1, &key, &psk);
-        assert!(!ticket.is_expired());
-    }
-
-    #[test]
-    fn test_validator_replay_protection() {
-        let psk = [42u8; 32];
-        let key = [1u8; 32];
-        let mut validator = TicketValidator::new(psk);
-
-        let ticket = validator.issue_ticket(1, &key);
-        let encrypted = ticket.encrypt(&psk);
-
-        // First use should succeed
-        assert!(validator.validate(&encrypted).is_some());
-
-        // Replay should fail
-        assert!(validator.validate(&encrypted).is_none());
-    }
-
-    #[test]
-    fn test_validator_different_tickets() {
-        let psk = [42u8; 32];
-        let mut validator = TicketValidator::new(psk);
-
-        let ticket1 = validator.issue_ticket(1, &[1u8; 32]);
-        let ticket2 = validator.issue_ticket(2, &[2u8; 32]);
-
-        assert!(validator.validate(&ticket1.encrypt(&psk)).is_some());
-        assert!(validator.validate(&ticket2.encrypt(&psk)).is_some());
-    }
-
-    #[test]
-    fn test_truncated_ticket_fails() {
-        assert!(SessionTicket::from_bytes(&[0u8; 10]).is_none());
-    }
-}

ostp-flutter/android/app/build.gradle.kts

@@ -1,3 +1,6 @@
+import java.io.FileInputStream
+import java.util.Properties
+
 plugins {
     id("com.android.application")
     id("kotlin-android")
@@ -5,6 +8,37 @@ plugins {
     id("dev.flutter.flutter-gradle-plugin")
 }
 
+// ── Release signing material ────────────────────────────────────────────────
+// Supplied out-of-band and never committed: either an `android/key.properties`
+// file (local release builds) or OSTP_KEYSTORE_* environment variables (CI).
+//
+// This exists because the release build used to be signed with the DEBUG
+// keystore (the stock Flutter template TODO). Android identifies an app by
+// applicationId + signing key, and refuses to update across a key change. The
+// debug keystore is auto-generated per machine, and CI runners are ephemeral,
+// so every published build carried a different random key — which is why
+// updating on top of a previous install failed with "App not installed" /
+// "unable to parse the package" and only a full uninstall+reinstall worked.
+val keystoreProperties = Properties().apply {
+    val propsFile = rootProject.file("key.properties")
+    if (propsFile.exists()) {
+        FileInputStream(propsFile).use { load(it) }
+    }
+}
+
+// Blank counts as absent. GitHub Actions substitutes an EMPTY STRING (not an
+// unset variable) for a secret that doesn't exist, so `getenv(...) ?: fallback`
+// silently kept the empty value — the elvis operator only catches null. That is
+// how an unset ANDROID_KEY_PASSWORD ended up being used as the literal key
+// password instead of falling back to the store password, producing Gradle's
+// "Get Key failed: Given final block not properly padded".
+fun signingSetting(propKey: String, envKey: String): String? =
+    (keystoreProperties.getProperty(propKey) ?: System.getenv(envKey))
+        ?.takeIf { it.isNotBlank() }
+
+val releaseStorePath: String? = signingSetting("storeFile", "OSTP_KEYSTORE_PATH")
+val hasReleaseSigning: Boolean = !releaseStorePath.isNullOrBlank()
+
 android {
     namespace = "com.ospab.ostp_client"
     compileSdk = flutter.compileSdkVersion
@@ -28,13 +62,50 @@ android {
         targetSdk = flutter.targetSdkVersion
         versionCode = flutter.versionCode
         versionName = flutter.versionName
+        
+        ndk {
+            abiFilters += listOf("armeabi-v7a", "arm64-v8a", "x86_64")
+        }
+    }
+
+    signingConfigs {
+        create("release") {
+            if (hasReleaseSigning) {
+                val store = signingSetting("storePassword", "OSTP_KEYSTORE_PASSWORD")
+                storeFile = file(releaseStorePath!!)
+                storePassword = store
+                keyAlias = signingSetting("keyAlias", "OSTP_KEY_ALIAS")
+                // PKCS12 (the keytool default since Java 9, and what our upload
+                // keystore is) cannot hold a key password that differs from the
+                // store password — the format simply has no place to put one. So
+                // treat a missing key password as "same as the store password"
+                // instead of demanding a secret that, for this keystore, can only
+                // ever be a duplicate. An explicit value still wins, for the older
+                // JKS format where the two genuinely can differ.
+                keyPassword = signingSetting("keyPassword", "OSTP_KEY_PASSWORD") ?: store
+            }
+        }
     }
 
     buildTypes {
         release {
-            // TODO: Add your own signing config for the release build.
-            // Signing with the debug keys for now, so `flutter run --release` works.
-            signingConfig = signingConfigs.getByName("debug")
+            // Use the real upload key when one was supplied; otherwise fall back to
+            // the debug keystore so a plain local `flutter build apk --release`
+            // still works for development. Anything PUBLISHED must take the first
+            // branch — a debug-signed build cannot be updated over, and its key is
+            // machine-local, so it also can't be reproduced later.
+            if (hasReleaseSigning) {
+                signingConfig = signingConfigs.getByName("release")
+            } else {
+                logger.warn(
+                    "OSTP: no release keystore configured (android/key.properties or " +
+                    "OSTP_KEYSTORE_PATH) - falling back to the DEBUG keystore. This APK " +
+                    "is for local use only: users cannot update over it, and the key is " +
+                    "not reproducible on another machine."
+                )
+                signingConfig = signingConfigs.getByName("debug")
+            }
+            proguardFiles(getDefaultProguardFile("proguard-android-optimize.txt"), "proguard-rules.pro")
         }
     }
 }

ostp-flutter/android/app/proguard-rules.pro

@@ -0,0 +1,3 @@
+-keep class net.ostp.client.OstpClientSdk { *; }
+-keep class com.ospab.ostp_client.OstpVpnService { *; }
+-keep class com.ospab.ostp_client.MainActivity { *; }

ostp-flutter/android/app/src/main/AndroidManifest.xml

@@ -1,6 +1,7 @@
 <manifest xmlns:android="http://schemas.android.com/apk/res/android">
     <uses-permission android:name="android.permission.INTERNET"/>
     <uses-permission android:name="android.permission.ACCESS_NETWORK_STATE"/>
+    <uses-permission android:name="android.permission.CHANGE_NETWORK_STATE"/>
     <uses-permission android:name="android.permission.QUERY_ALL_PACKAGES"/>
     <uses-permission android:name="android.permission.FOREGROUND_SERVICE"/>
     <uses-permission android:name="android.permission.FOREGROUND_SERVICE_CONNECTED_DEVICE"/>
@@ -9,7 +10,8 @@
     <application
         android:label="ostp_client"
         android:name="${applicationName}"
-        android:icon="@mipmap/ic_launcher"
+        android:icon="@mipmap/launcher_icon"
+        android:roundIcon="@mipmap/launcher_icon_round"
         android:extractNativeLibs="true">
         <activity
             android:name=".MainActivity"
@@ -32,6 +34,9 @@
                 <action android:name="android.intent.action.MAIN"/>
                 <category android:name="android.intent.category.LAUNCHER"/>
             </intent-filter>
+            <intent-filter>
+                <action android:name="android.service.quicksettings.action.QS_TILE_PREFERENCES"/>
+            </intent-filter>
         </activity>
         <!-- Don't delete the meta-data below.
              This is used by the Flutter tool to generate GeneratedPluginRegistrant.java -->
@@ -42,6 +47,7 @@
         <service
             android:name=".OstpVpnService"
             android:permission="android.permission.BIND_VPN_SERVICE"
+            android:foregroundServiceType="connectedDevice"
             android:exported="false">
             <intent-filter>
                 <action android:name="android.net.VpnService"/>
@@ -51,7 +57,7 @@
         <!-- Quick Settings Tile -->
         <service
             android:name=".OstpTileService"
-            android:icon="@mipmap/ic_launcher"
+            android:icon="@mipmap/launcher_icon"
             android:label="OSTP VPN"
             android:permission="android.permission.BIND_QUICK_SETTINGS_TILE"
             android:exported="true">

ostp-flutter/android/app/src/main/kotlin/com/ospab/ostp_client/MainActivity.kt

@@ -92,28 +92,46 @@ class MainActivity : FlutterActivity() {
                         val metrics = net.ostp.client.OstpClientSdk.getMetrics()
                         result.success(metrics ?: "{}")
                     } catch (e: Throwable) {
+                        // Surfaced into the in-app log viewer (not just logcat) so a
+                        // broken traffic counter is diagnosable from a user's bug
+                        // report without adb access.
+                        android.util.Log.e("MainActivity", "getMetrics failed", e)
+                        try {
+                            net.ostp.client.OstpClientSdk.addLog("getMetrics failed: ${e.javaClass.simpleName}: ${e.message}")
+                        } catch (_: Throwable) {}
                         result.error("ERROR", e.message, null)
                     }
                 }
                 "getInstalledApps" -> {
-                    try {
-                        val pm = packageManager
-                        val apps = pm.getInstalledApplications(PackageManager.GET_META_DATA)
-                        val list = apps.map { app ->
-                            val isSystem = ((app.flags and ApplicationInfo.FLAG_SYSTEM) != 0) &&
-                                           (pm.getLaunchIntentForPackage(app.packageName) == null)
-                            val iconBase64 = getAppIconBase64(pm, app)
-                            mapOf(
-                                "name" to pm.getApplicationLabel(app).toString(),
-                                "package" to app.packageName,
-                                "isSystem" to isSystem,
-                                "icon" to (iconBase64 ?: "")
-                            )
+                    // MethodChannel handlers run on the main/UI thread by default.
+                    // Enumerating every installed package AND decoding+re-encoding
+                    // each one's icon to PNG/base64 is expensive (100+ apps is
+                    // common) — done inline here it blocked the main thread for
+                    // 10-15s, during which Flutter couldn't render ANY frame, not
+                    // even the "loading" spinner, so the screen just appeared to
+                    // hang before jumping straight to the fully-loaded list.
+                    // Do the work on a background thread; only the final
+                    // `result.success(...)` needs to hop back onto the UI thread.
+                    val pm = packageManager
+                    Thread {
+                        try {
+                            val apps = pm.getInstalledApplications(PackageManager.GET_META_DATA)
+                            val list = apps.map { app ->
+                                val isSystem = ((app.flags and ApplicationInfo.FLAG_SYSTEM) != 0) &&
+                                               (pm.getLaunchIntentForPackage(app.packageName) == null)
+                                val iconBase64 = getAppIconBase64(pm, app)
+                                mapOf(
+                                    "name" to pm.getApplicationLabel(app).toString(),
+                                    "package" to app.packageName,
+                                    "isSystem" to isSystem,
+                                    "icon" to (iconBase64 ?: "")
+                                )
+                            }
+                            runOnUiThread { result.success(list) }
+                        } catch (e: Exception) {
+                            runOnUiThread { result.error("ERROR", e.message, null) }
                         }
-                        result.success(list)
-                    } catch (e: Exception) {
-                        result.error("ERROR", e.message, null)
-                    }
+                    }.start()
                 }
                 else -> result.notImplemented()
             }
@@ -133,6 +151,6 @@ class MainActivity : FlutterActivity() {
         if (pendingConfigJson != null) {
             intent.putExtra("configJson", pendingConfigJson)
         }
-        startService(intent)
+        androidx.core.content.ContextCompat.startForegroundService(this, intent)
     }
 }

ostp-flutter/android/app/src/main/kotlin/com/ospab/ostp_client/OstpTileService.kt

@@ -34,6 +34,19 @@ class OstpTileService : TileService() {
             val configJson = prefs.getString("latest_config_json", null)
             
             if (configJson != null) {
+                // Check if VPN consent is needed
+                val vpnIntent = android.net.VpnService.prepare(this)
+                if (vpnIntent != null) {
+                    // Consent needed, launch app
+                    val appIntent = packageManager.getLaunchIntentForPackage(packageName)?.apply {
+                        addFlags(Intent.FLAG_ACTIVITY_NEW_TASK or Intent.FLAG_ACTIVITY_SINGLE_TOP)
+                    }
+                    if (appIntent != null) {
+                        startActivityAndCollapse(appIntent)
+                    }
+                    return
+                }
+
                 val startIntent = Intent(this, OstpVpnService::class.java).apply {
                     action = "START"
                     putExtra("configJson", configJson)

ostp-flutter/android/app/src/main/kotlin/com/ospab/ostp_client/OstpVpnService.kt

@@ -16,6 +16,7 @@ import java.io.IOException
 import androidx.annotation.Keep
 import androidx.core.app.NotificationCompat
 import androidx.core.app.NotificationManagerCompat
+import androidx.core.app.ServiceCompat
 
 @Keep
 class OstpVpnService : VpnService() {
@@ -43,6 +44,7 @@ class OstpVpnService : VpnService() {
 
     private var vpnInterface: ParcelFileDescriptor? = null
     private var wakeLock: PowerManager.WakeLock? = null
+    private var networkCallback: android.net.ConnectivityManager.NetworkCallback? = null
 
     override fun onCreate() {
         super.onCreate()
@@ -55,7 +57,7 @@ class OstpVpnService : VpnService() {
         if (action == "START") {
             val configJson = intent.getStringExtra("configJson") ?: return START_NOT_STICKY
             // Launch foreground immediately so Android doesn't kill us
-            startForeground(NOTIF_ID, buildNotification(connecting = true))
+            ServiceCompat.startForeground(this, NOTIF_ID, buildNotification(connecting = true), ServiceInfo.FOREGROUND_SERVICE_TYPE_CONNECTED_DEVICE)
             startVpn(configJson)
         } else if (action == "STOP") {
             stopVpn()
@@ -144,6 +146,41 @@ class OstpVpnService : VpnService() {
         }
     }
 
+    private fun registerNetworkCallback() {
+        if (networkCallback != null) return
+        try {
+            val cm = getSystemService(android.content.Context.CONNECTIVITY_SERVICE) as android.net.ConnectivityManager
+            networkCallback = object : android.net.ConnectivityManager.NetworkCallback() {
+                override fun onAvailable(network: android.net.Network) {
+                    super.onAvailable(network)
+                    OstpClientSdk.notifyNetworkChanged()
+                }
+                override fun onLost(network: android.net.Network) {
+                    super.onLost(network)
+                    OstpClientSdk.notifyNetworkChanged()
+                }
+            }
+            val request = android.net.NetworkRequest.Builder()
+                .addCapability(android.net.NetworkCapabilities.NET_CAPABILITY_INTERNET)
+                .build()
+            cm.registerNetworkCallback(request, networkCallback!!)
+        } catch (e: Throwable) {
+            Log.e("OstpVpnService", "Failed to register NetworkCallback", e)
+        }
+    }
+
+    private fun unregisterNetworkCallback() {
+        try {
+            if (networkCallback != null) {
+                val cm = getSystemService(android.content.Context.CONNECTIVITY_SERVICE) as android.net.ConnectivityManager
+                cm.unregisterNetworkCallback(networkCallback!!)
+                networkCallback = null
+            }
+        } catch (e: Throwable) {
+            Log.e("OstpVpnService", "Failed to unregister NetworkCallback", e)
+        }
+    }
+
     private fun startVpn(configJson: String) {
         if (vpnInterface != null) return
 
@@ -162,8 +199,13 @@ class OstpVpnService : VpnService() {
                 .addRoute("::", 0)
                 .addDnsServer(dnsServer)
                 .setMtu(Math.max(1280, json.optJSONObject("ostp")?.optInt("mtu", 1140) ?: 1140))
-                
+
+            // Always add fallback IPv4 DNS servers
+            try { builder.addDnsServer("1.1.1.1") } catch (e: Throwable) {}
             try { builder.addDnsServer("8.8.8.8") } catch (e: Throwable) {}
+            // NOTE: Do NOT add IPv6 DNS servers here — Android would send DNS
+            // queries over IPv6, but our smoltcp TUN stack processes them as
+            // IPv4 only, causing all DNS to silently fail on LTE (IPv6-only networks).
 
             if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.Q) {
                 builder.allowBypass()
@@ -230,8 +272,13 @@ class OstpVpnService : VpnService() {
 
         } catch (e: Throwable) {
             Log.e("OstpVpnService", "Error starting VPN", e)
+            android.os.Handler(android.os.Looper.getMainLooper()).post {
+                android.widget.Toast.makeText(applicationContext, "VPN Error: ${e.message}", android.widget.Toast.LENGTH_LONG).show()
+            }
             stopVpn()
         }
+        
+        registerNetworkCallback()
     }
 
     private fun stopVpn() {
@@ -248,6 +295,7 @@ class OstpVpnService : VpnService() {
 
         stopForeground(true)
         OstpTileService.requestListeningState(applicationContext)
+        unregisterNetworkCallback()
         stopSelf()
     }
 

ostp-flutter/android/app/src/main/kotlin/net/ostp/client/OstpClientSdk.kt

@@ -46,4 +46,8 @@ object OstpClientSdk {
     @Keep
     @JvmStatic
     external fun addLog(logMsg: String)
+
+    @Keep
+    @JvmStatic
+    external fun notifyNetworkChanged()
 }

ostp-flutter/android/app/src/main/res/mipmap-anydpi-v26/launcher_icon.xml

@@ -0,0 +1,5 @@
+<?xml version="1.0" encoding="utf-8"?>
+<adaptive-icon xmlns:android="http://schemas.android.com/apk/res/android">
+  <background android:drawable="@color/ic_launcher_background"/>
+  <foreground android:drawable="@drawable/ic_launcher_foreground"/>
+</adaptive-icon>