Merge pull request #208 from telemt/flow

Middle-End protocol hardening

.github/workflows/release.yml

@@ -3,8 +3,8 @@ name: Release
 on:
   push:
     tags:
-      - '[0-9]+.[0-9]+.[0-9]+'  # Matches tags like 3.0.0, 3.1.2, etc.
-  workflow_dispatch:  # Manual trigger from GitHub Actions UI
+      - '[0-9]+.[0-9]+.[0-9]+' # Matches tags like 3.0.0, 3.1.2, etc.
+  workflow_dispatch: # Manual trigger from GitHub Actions UI
 
 permissions:
   contents: read
@@ -25,10 +25,16 @@ jobs:
         include:
           - target: x86_64-unknown-linux-gnu
             artifact_name: telemt
-            asset_name: telemt-x86_64-linux
+            asset_name: telemt-x86_64-linux-gnu
           - target: aarch64-unknown-linux-gnu
             artifact_name: telemt
-            asset_name: telemt-aarch64-linux
+            asset_name: telemt-aarch64-linux-gnu
+          - target: x86_64-unknown-linux-musl
+            artifact_name: telemt
+            asset_name: telemt-x86_64-linux-musl
+          - target: aarch64-unknown-linux-musl
+            artifact_name: telemt
+            asset_name: telemt-aarch64-linux-musl
 
     steps:
       - name: Checkout repository
@@ -56,12 +62,13 @@ jobs:
           restore-keys: |
             ${{ runner.os }}-${{ matrix.target }}-cargo-
 
+      - name: Install cross
+        run: cargo install cross --git https://github.com/cross-rs/cross
+
       - name: Build Release
-        uses: actions-rs/cargo@ae10961054e4aa8bff448f48a500763b90d5c550 # v1.0.1
-        with:
-          use-cross: true
-          command: build
-          args: --release --target ${{ matrix.target }}
+        env:
+          RUSTFLAGS: ${{ contains(matrix.target, 'musl') && '-C target-feature=+crt-static' || '' }}
+        run: cross build --release --target ${{ matrix.target }}
 
       - name: Package binary
         run: |
@@ -77,6 +84,32 @@ jobs:
             target/${{ matrix.target }}/release/${{ matrix.asset_name }}.tar.gz
             target/${{ matrix.target }}/release/${{ matrix.asset_name }}.sha256
 
+  build-docker-image:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v3
+
+      - name: Set up QEMU
+        uses: docker/setup-qemu-action@v3
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v2
+
+      - name: Login to GitHub Container Registry
+        uses: docker/login-action@v2
+        with:
+          registry: ghcr.io
+          username: ${{ github.repository_owner }}
+          password: ${{ secrets.TOKEN_GH_DEPLOY }}
+
+      - name: Build and push
+        uses: docker/build-push-action@v6
+        with:
+          context: .
+          push: true
+          tags: ${{ github.ref }}
+
   release:
     name: Create Release
     needs: build
@@ -85,13 +118,42 @@ jobs:
       contents: write
 
     steps:
+      - name: Checkout repository
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          fetch-depth: 0
+          token: ${{ secrets.GITHUB_TOKEN }}
+
       - name: Download all artifacts
         uses: actions/download-artifact@fa0a91b85d4f404e444e00e005971372dc801d16 # v4.1.8
         with:
           path: artifacts
 
+      - name: Update version in Cargo.toml and Cargo.lock
+        run: |
+          # Extract version from tag (remove 'v' prefix if present)
+          VERSION="${GITHUB_REF#refs/tags/}"
+          VERSION="${VERSION#v}"
+
+          # Install cargo-edit for version bumping
+          cargo install cargo-edit
+
+          # Update Cargo.toml version
+          cargo set-version "$VERSION"
+
+          # Configure git
+          git config user.name "github-actions[bot]"
+          git config user.email "github-actions[bot]@users.noreply.github.com"
+
+          # Commit and push changes
+          #git add Cargo.toml Cargo.lock
+          #git commit -m "chore: bump version to $VERSION" || echo "No changes to commit"
+          #git push origin HEAD:main
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+
       - name: Create Release
-        uses: softprops/action-gh-release@c95fe1489396fe360a41fb53f90de6ddce8c4c8a # v2.2.1
+        uses: softprops/action-gh-release@v2
         with:
           files: artifacts/**/*
           generate_release_notes: true

AGENTS_SYSTEM_PROMT.md

@@ -1,6 +1,7 @@
 ## System Prompt — Production Rust Codebase: Modification and Architecture Guidelines
 
-You are a senior Rust systems engineer acting as a strict code reviewer and implementation partner. Your responses are precise, minimal, and architecturally sound. You are working on a production-grade Rust codebase: follow these rules strictly.
+You are a senior Rust Engineer and pricipal Rust Architect acting as a strict code reviewer and implementation partner. 
+Your responses are precise, minimal, and architecturally sound. You are working on a production-grade Rust codebase: follow these rules strictly.
 
 ---
 
@@ -32,6 +33,11 @@ The user can override this behavior with explicit commands:
 - `"Make minimal changes"` — no coordinated fixes, narrowest possible diff.
 - `"Fix everything"` — apply all coordinated fixes and out-of-scope observations.
 
+### Core Rule
+
+The codebase must never enter an invalid intermediate state.
+No response may leave the repository in a condition that requires follow-up fixes.
+
 ---
 
 ### 1. Comments and Documentation
@@ -131,16 +137,32 @@ You MUST:
 - Document non-obvious logic with comments that describe *why*, not *what*.
 - Limit changes strictly to the requested scope (plus coordinated fixes per Section 0).
 - Keep all existing symbol names unless renaming is explicitly requested.
-- Preserve global formatting as-is.
+- Preserve global formatting as-is
+- Result every modification in a self-contained, compilable, runnable state of the codebase
 
 You MUST NOT:
 
-- Use placeholders: no `// ... rest of code`, no `// implement here`, no `/* TODO */` stubs that replace existing working code. Write full, working implementation. If the implementation is unclear, ask first.
-- Refactor code outside the requested scope.
-- Make speculative improvements.
+- Use placeholders: no `// ... rest of code`, no `// implement here`, no `/* TODO */` stubs that replace existing working code. Write full, working implementation. If the implementation is unclear, ask first
+- Refactor code outside the requested scope
+- Make speculative improvements
+- Spawn multiple agents for EDITING
+- Produce partial changes
+- Introduce references to entities that are not yet implemented
+- Leave TODO placeholders in production paths
 
 Note: `todo!()` and `unimplemented!()` are allowed as idiomatic Rust markers for genuinely unfinished code paths.
 
+Every change must:
+   - compile,
+   - pass type checks,
+   - have no broken imports,
+   - preserve invariants,
+   - not rely on future patches.
+
+If the task requires multiple phases:
+   - either implement all required phases,
+   - or explicitly refuse and explain missing dependencies.
+
 ---
 
 ### 8. Decision Process for Complex Changes
@@ -160,6 +182,7 @@ When facing a non-trivial modification, follow this sequence:
 - When provided with partial code, assume the rest of the codebase exists and functions correctly unless stated otherwise.
 - Reference existing types, functions, and module structures by their actual names as shown in the provided code.
 - When the provided context is insufficient to make a safe change, request the missing context explicitly.
+- Spawn multiple agents for SEARCHING information, code, functions
 
 ---
 
@@ -167,14 +190,14 @@ When facing a non-trivial modification, follow this sequence:
 
 #### Language Policy
 
-- Code, comments, commit messages, documentation: **English**.
-- Reasoning and explanations in response text: **Russian**.
+- Code, comments, commit messages, documentation ONLY ON **English**!
+- Reasoning and explanations in response text on language from promt
 
 #### Response Structure
 
 Your response MUST consist of two sections:
 
-**Section 1: `## Reasoning` (in Russian)**
+**Section 1: `## Reasoning`**
 
 - What needs to be done and why.
 - Which files and modules are affected.
@@ -205,3 +228,183 @@ If the response exceeds the output limit:
 2. List the files that will be provided in subsequent parts.
 3. Wait for user confirmation before continuing.
 4. No single file may be split across parts.
+
+## 11. Anti-LLM Degeneration Safeguards (Principal-Paranoid, Visionary)
+
+This section exists to prevent common LLM failure modes: scope creep, semantic drift, cargo-cult refactors, performance regressions, contract breakage, and hidden behavior changes.
+
+### 11.1 Non-Negotiable Invariants
+
+- **No semantic drift:** Do not reinterpret requirements, rename concepts, or change meaning of existing terms.
+- **No “helpful refactors”:** Any refactor not explicitly requested is forbidden.
+- **No architectural drift:** Do not introduce new layers, patterns, abstractions, or “clean architecture” migrations unless requested.
+- **No dependency drift:** Do not add crates, features, or versions unless explicitly requested.
+- **No behavior drift:** If a change could alter runtime behavior, you MUST call it out explicitly in `## Reasoning` and justify it.
+
+### 11.2 Minimal Surface Area Rule
+
+- Touch the smallest number of files possible.
+- Prefer local changes over cross-cutting edits.
+- Do not “align style” across a file/module—only adjust the modified region.
+- Do not reorder items, imports, or code unless required for correctness.
+
+### 11.3 No Implicit Contract Changes
+
+Contracts include:
+- public APIs, trait bounds, visibility, error types, timeouts/retries, logging semantics, metrics semantics,
+- protocol formats, framing, padding, keepalive cadence, state machine transitions,
+- concurrency guarantees, cancellation behavior, backpressure behavior.
+
+Rule:
+- If you change a contract, you MUST update all dependents in the same patch AND document the contract delta explicitly.
+
+### 11.4 Hot-Path Preservation (Performance Paranoia)
+
+- Do not introduce extra allocations, cloning, or formatting in hot paths.
+- Do not add logging/metrics on hot paths unless requested.
+- Do not add new locks or broaden lock scope.
+- Prefer `&str` / slices / borrowed data where the codebase already does so.
+- Avoid `String` building for errors/logs if it changes current patterns.
+
+If you cannot prove performance neutrality, label it as risk in `## Reasoning`.
+
+### 11.5 Async / Concurrency Safety (Cancellation & Backpressure)
+
+- No blocking calls inside async contexts.
+- Preserve cancellation safety: do not introduce `await` between lock acquisition and critical invariants unless already present.
+- Preserve backpressure: do not replace bounded channels with unbounded, do not remove flow control.
+- Do not change task lifecycle semantics (spawn patterns, join handles, shutdown order) unless requested.
+- Do not introduce `tokio::spawn` / background tasks unless explicitly requested.
+
+### 11.6 Error Semantics Integrity
+
+- Do not replace structured errors with generic strings.
+- Do not widen/narrow error types or change error categories without explicit approval.
+- Avoid introducing panics in production paths (`unwrap`, `expect`) unless the codebase already treats that path as impossible and documented.
+
+### 11.7 “No New Abstractions” Default
+
+Default stance:
+- No new traits, generics, macros, builder patterns, type-level cleverness, or “frameworking”.
+- If abstraction is necessary, prefer the smallest possible local helper (private function) and justify it.
+
+### 11.8 Negative-Diff Protection
+
+Avoid “diff inflation” patterns:
+- mass edits,
+- moving code between files,
+- rewrapping long lines,
+- rearranging module order,
+- renaming for aesthetics.
+
+If a diff becomes large, STOP and ask before proceeding.
+
+### 11.9 Consistency with Existing Style (But Not Style Refactors)
+
+- Follow existing conventions of the touched module (naming, error style, return patterns).
+- Do not enforce global “best practices” that the codebase does not already use.
+
+### 11.10 Two-Phase Safety Gate (Plan → Patch)
+
+For non-trivial changes:
+1) Provide a micro-plan (1–5 bullets): what files, what functions, what invariants, what risks.
+2) Implement exactly that plan—no extra improvements.
+
+### 11.11 Pre-Response Checklist (Hard Gate)
+
+Before final output, verify internally:
+
+- No unresolved symbols / broken imports.
+- No partially updated call sites.
+- No new public surface changes unless requested.
+- No transitional states / TODO placeholders replacing working code.
+- Changes are atomic: the repository remains buildable and runnable.
+- Any behavior change is explicitly stated.
+
+If any check fails: fix it before responding.
+
+### 11.12 Truthfulness Policy (No Hallucinated Claims)
+
+- Do not claim “this compiles” or “tests pass” unless you actually verified with the available tooling/context.
+- If verification is not possible, state: “Not executed; reasoning-based consistency check only.”
+
+### 11.13 Visionary Guardrail: Preserve Optionality
+
+When multiple valid designs exist, prefer the one that:
+- minimally constrains future evolution,
+- preserves existing extension points,
+- avoids locking the project into a new paradigm,
+- keeps interfaces stable and implementation local.
+
+Default to reversible changes.
+
+### 11.14 Stop Conditions
+
+STOP and ask targeted questions if:
+- required context is missing,
+- a change would cross module boundaries,
+- a contract might change,
+- concurrency/protocol invariants are unclear,
+- the diff is growing beyond a minimal patch.
+
+No guessing.
+
+### 12. Invariant Preservation
+
+You MUST explicitly preserve:
+- Thread-safety guarantees (`Send` / `Sync` expectations).
+- Memory safety assumptions (no hidden `unsafe` expansions).
+- Lock ordering and deadlock invariants.
+- State machine correctness (no new invalid transitions).
+- Backward compatibility of serialized formats (if applicable).
+
+If a change touches concurrency, networking, protocol logic, or state machines,
+you MUST explain why existing invariants remain valid.
+
+### 13. Error Handling Policy
+
+- Do not replace structured errors with generic strings.
+- Preserve existing error propagation semantics.
+- Do not widen or narrow error types without approval.
+- Avoid introducing panics in production paths.
+- Prefer explicit error mapping over implicit conversions.
+
+### 14. Test Safety
+
+- Do not modify existing tests unless the task explicitly requires it.
+- Do not weaken assertions.
+- Preserve determinism in testable components.
+
+### 15. Security Constraints
+
+- Do not weaken cryptographic assumptions.
+- Do not modify key derivation logic without explicit request.
+- Do not change constant-time behavior.
+- Do not introduce logging of secrets.
+- Preserve TLS/MTProto protocol correctness.
+
+### 16. Logging Policy
+
+- Do not introduce excessive logging in hot paths.
+- Do not log sensitive data.
+- Preserve existing log levels and style.
+
+### 17. Pre-Response Verification Checklist
+
+Before producing the final answer, verify internally:
+
+- The change compiles conceptually.
+- No unresolved symbols exist.
+- All modified call sites are updated.
+- No accidental behavioral changes were introduced.
+- Architectural boundaries remain intact.
+
+### 18. Atomic Change Principle
+Every patch must be **atomic and production-safe**.
+* **Self-contained** — no dependency on future patches or unimplemented components.
+* **Build-safe** — the project must compile successfully after the change.
+* **Contract-consistent** — no partial interface or behavioral changes; all dependent code must be updated within the same patch.
+* **No transitional states** — no placeholders, incomplete refactors, or temporary inconsistencies.
+
+**Invariant:** After any single patch, the repository remains fully functional and buildable.
+

CONTRIBUTING.md

@@ -0,0 +1,14 @@
+# Pull Requests - Rules
+## General
+- ONLY signed and verified commits
+- ONLY from your name
+- DO NOT commit with `codex` or `claude` as author/commiter
+- PREFER `flow` branch for development, not `main`
+
+## AI
+We are not against modern tools, like AI, where you act as a principal or architect, but we consider it important:
+
+- you really understand what you're doing
+- you understand the relationships and dependencies of the components being modified
+- you understand the architecture of Telegram MTProto, MTProxy, Middle-End KDF at least generically
+- you DO NOT commit for the sake of commits, but to help the community, core-developers and ordinary users

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "telemt"
-version = "3.0.3"
+version = "3.0.9"
 edition = "2024"
 
 [dependencies]
@@ -20,15 +20,18 @@ sha1 = "0.10"
 md-5 = "0.10"
 hmac = "0.12"
 crc32fast = "1.4"
+crc32c = "0.6"
 zeroize = { version = "1.8", features = ["derive"] }
 
 # Network
 socket2 = { version = "0.5", features = ["all"] }
+nix = { version = "0.28", default-features = false, features = ["net"] }
 
 # Serialization
 serde = { version = "1.0", features = ["derive"] }
 serde_json = "1.0"
 toml = "0.8"
+x509-parser = "0.15"
 
 # Utils
 bytes = "1.9"
@@ -47,13 +50,19 @@ regex = "1.11"
 crossbeam-queue = "0.3"
 num-bigint = "0.4"
 num-traits = "0.2"
+anyhow = "1.0"
 
 # HTTP
 reqwest = { version = "0.12", features = ["rustls-tls"], default-features = false }
+notify = { version = "6", features = ["macos_fsevent"] }
+ipnetwork = "0.20"
 hyper = { version = "1", features = ["server", "http1"] }
 hyper-util = { version = "0.1", features = ["tokio", "server-auto"] }
 http-body-util = "0.1"
 httpdate = "1.0"
+tokio-rustls = { version = "0.26", default-features = false, features = ["tls12"] }
+rustls = { version = "0.23", default-features = false, features = ["std", "tls12", "ring"] }
+webpki-roots = "0.26"
 
 [dev-dependencies]
 tokio-test = "0.4"

LICENSING.md

@@ -0,0 +1,17 @@
+# LICENSING
+## Licenses for Versions
+| Version | License       |
+|---------|---------------|
+| 1.0     | NO LICNESE    |
+| 1.1     | NO LICENSE    |
+| 1.2     | NO LICENSE    |
+| 2.0     | NO LICENSE    |
+| 3.0     | TELEMT UL 1   |
+
+### License Types
+- **NO LICENSE** = ***ALL RIGHT RESERVED***
+- **TELEMT UL1** - work in progress license for source code of `telemt`, which encourages:
+  - fair use, 
+  - contributions, 
+  - distribution, 
+  - but prohibits NOT mentioning the authors

README.md

@@ -10,74 +10,40 @@
 
 ### 🇷🇺 RU
 
-15 февраля мы опубликовали `telemt 3` с поддержкой Middle-End Proxy, а значит:
+18 февраля мы опубликовали `telemt 3.0.3`, он имеет:
 
-- с функциональными медиа, в том числе с CDN/DC=203  
-- с Ad-tag — показывайте спонсорский канал и собирайте статистику через официального бота  
-- с новым подходом к безопасности и асинхронности  
-- с высокоточной диагностикой криптографии через `ME_DIAG`  
+- улучшенный механизм Middle-End Health Check
+- высокоскоростное восстановление инициализации Middle-End
+- меньше задержек на hot-path
+- более корректную работу в Dualstack, а именно - IPv6 Middle-End
+- аккуратное переподключение клиента без дрифта сессий между Middle-End
+- автоматическая деградация на Direct-DC при массовой (>2 ME-DC-групп) недоступности Middle-End
+- автодетект IP за NAT, при возможности - будет выполнен хендшейк с ME, при неудаче - автодеградация
+- единственный известный специальный DC=203 уже добавлен в код: медиа загружаются с CDN в Direct-DC режиме
 
-Для использования нужно:
+[Здесь вы можете найти релиз](https://github.com/telemt/telemt/releases/tag/3.0.3)
 
-1. Версия `telemt` ≥3.0.0
-2. Выполнение любого из наборов условий:
-   - публичный IP для исходящих соединений установлен на интерфейса инстанса с `telemt`
-     - ЛИБО
-   - вы используете NAT 1:1 + включили STUN-пробинг
-3. В конфиге, в секции `[general]` указать:
-```toml
-use_middle_proxy = true
-```
-
-Если условия из пункта 1 не выполняются:
-1. Выключите ME-режим:
-   - установите `use_middle_proxy = false`
-     - ЛИБО
-   - Middle-End Proxy будет выключен автоматически по таймауту, но это займёт больше времени при запуске
-2. В конфиге, добавьте в конец:
-```toml
-[dc_overrides]
-"203" = "91.105.192.100:443"
-```
-
-Если у вас есть компетенции в асинхронных сетевых приложениях, анализе трафика, реверс-инжиниринге или сетевых расследованиях — мы открыты к идеям и pull requests.
+Если у вас есть компетенции в асинхронных сетевых приложениях, анализе трафика, реверс-инжиниринге или сетевых расследованиях - мы открыты к идеям и pull requests!
 
 </td>
 <td width="50%" valign="top">
 
 ### 🇬🇧 EN
 
-On February 15, we released `telemt 3` with support for Middle-End Proxy, which means:
+On February 18, we released `telemt 3.0.3`. This version introduces:
 
-- functional media, including CDN/DC=203  
-- Ad-tag support – promote a sponsored channel and collect statistics via Telegram bot  
-- new approach to security and asynchronicity  
-- high-precision cryptography diagnostics via `ME_DIAG`  
+- improved Middle-End Health Check method  
+- high-speed recovery of Middle-End init
+- reduced latency on the hot path  
+- correct Dualstack support: proper handling of IPv6 Middle-End  
+- *clean* client reconnection without session "drift" between Middle-End
+- automatic degradation to Direct-DC mode in case of large-scale (>2 ME-DC groups) Middle-End unavailability  
+- automatic public IP detection behind NAT; first - Middle-End handshake is performed, otherwise automatic degradation is applied  
+- known special DC=203 is now handled natively: media is delivered from the CDN via Direct-DC mode  
 
-To use this feature, the following requirements must be met:
-1. `telemt` version ≥ 3.0.0
-2. One of the following conditions satisfied:
-   - the instance running `telemt` has a public IP address assigned to its network interface for outbound connections  
-     - OR
-   - you are using 1:1 NAT and have STUN probing enabled  
-3. In the config file, under the `[general]` section, specify:
-```toml
-use_middle_proxy = true
-````
+[Release is available here](https://github.com/telemt/telemt/releases/tag/3.0.3)
 
-If the conditions from step 1 are not satisfied:
-1. Disable Middle-End mode:
-   - set `use_middle_proxy = false`
-     - OR
-   - Middle-End Proxy will be disabled automatically after a timeout, but this will increase startup time
-
-2. In the config file, add the following at the end:
-```toml
-[dc_overrides]
-"203" = "91.105.192.100:443"
-```
-
-If you have expertise in asynchronous network applications, traffic analysis, reverse engineering, or network forensics — we welcome ideas, suggestions, and pull requests.
+If you have expertise in asynchronous network applications, traffic analysis, reverse engineering, or network forensics - we welcome ideas and pull requests!
 
 </td>
 </tr>
@@ -86,7 +52,9 @@ If you have expertise in asynchronous network applications, traffic analysis, re
 # Features
 💥 The configuration structure has changed since version 1.1.0.0. change it in your environment!
 
-⚓ Our implementation of **TLS-fronting** is one of the most deeply debugged, focused, advanced and *almost* **"behaviorally consistent to real"**:  we are confident we have it right - [see evidence on our validation and traces](#recognizability-for-dpi-and-crawler) 
+⚓ Our implementation of **TLS-fronting** is one of the most deeply debugged, focused, advanced and *almost* **"behaviorally consistent to real"**:  we are confident we have it right - [see evidence on our validation and traces](#recognizability-for-dpi-and-crawler)
+
+⚓ Our ***Middle-End Pool*** is fastest by design in standard scenarios, compared to other implementations of connecting to the Middle-End Proxy: non dramatically, but usual
 
 # GOTO
 - [Features](#features)
@@ -127,7 +95,7 @@ If you have expertise in asynchronous network applications, traffic analysis, re
 **This software is designed for Debian-based OS: in addition to Debian, these are Ubuntu, Mint, Kali, MX and many other Linux**
 1. Download release
 ```bash
-wget https://github.com/telemt/telemt/releases/latest/download/telemt
+wget -qO- "https://github.com/telemt/telemt/releases/latest/download/telemt-$(uname -m)-linux-$(ldd --version 2>&1 | grep -iq musl && echo musl || echo gnu).tar.gz" | tar -xz
 ```
 2. Move to Bin Folder
 ```bash
@@ -210,98 +178,21 @@ then Ctrl+X -> Y -> Enter to save
 ```toml
 # === General Settings ===
 [general]
-# prefer_ipv6 is deprecated; use [network].prefer
-prefer_ipv6 = false
-fast_mode = true
-use_middle_proxy = false
-# ad_tag = "..."
-
-[network]
-ipv4 = true
-ipv6 = true   # set false to disable, omit for auto
-prefer = 4    # 4 or 6
-multipath = false
+# ad_tag = "00000000000000000000000000000000"
 
 [general.modes]
 classic = false
 secure = false
 tls = true
 
-# === Server Binding ===
-[server]
-port = 443
-listen_addr_ipv4 = "0.0.0.0"
-listen_addr_ipv6 = "::"
-# metrics_port = 9090
-# metrics_whitelist = ["127.0.0.1", "::1"]
-
-# Listen on multiple interfaces/IPs (overrides listen_addr_*)
-[[server.listeners]]
-ip = "0.0.0.0"
-# announce = "my.hostname.tld" # Optional: hostname for tg:// links
-# OR
-# announce = "1.2.3.4" # Optional: Public IP for tg:// links
-
-[[server.listeners]]
-ip = "::"
-
-# Users to show in the startup log (tg:// links)
-[general.links]
-show = ["hello"] # Users to show in the startup log (tg:// links)
-# public_host = "proxy.example.com"  # Host (IP or domain) for tg:// links
-# public_port = 443                  # Port for tg:// links (default: server.port)
-
-# === Timeouts (in seconds) ===
-[timeouts]
-client_handshake = 15
-tg_connect = 10
-client_keepalive = 60
-client_ack = 300
-
 # === Anti-Censorship & Masking ===
 [censorship]
 tls_domain = "petrovich.ru"
-mask = true
-mask_port = 443
-# mask_host = "petrovich.ru" # Defaults to tls_domain if not set
-# mask_unix_sock = "/var/run/nginx.sock" # Unix socket (mutually exclusive with mask_host)
-fake_cert_len = 2048
-
-# === Access Control & Users ===
-# username "hello" is used for example
-[access]
-replay_check_len = 65536
-ignore_time_skew = false
 
 [access.users]
 # format: "username" = "32_hex_chars_secret"
 hello = "00000000000000000000000000000000"
 
-# [access.user_max_tcp_conns]
-# hello = 50
-
-# [access.user_data_quota]
-# hello = 1073741824 # 1 GB
-
-# === Upstreams & Routing ===
-# By default, direct connection is used, but you can add SOCKS proxy
-
-# Direct - Default
-[[upstreams]]
-type = "direct"
-enabled = true
-weight = 10
-
-# SOCKS5
-# [[upstreams]]
-# type = "socks5"
-# address = "127.0.0.1:9050"
-# enabled = false
-# weight = 1
-
-# === DC Address Overrides ===
-# [dc_overrides]
-# "203" = "91.105.192.100:443"
 ```
 ### Advanced
 #### Adtag

config.toml

@@ -1,29 +1,69 @@
 # === General Settings ===
 [general]
-# prefer_ipv6 is deprecated; use [network].prefer instead
-prefer_ipv6 = false
 fast_mode = true
 use_middle_proxy = true
-#ad_tag = "00000000000000000000000000000000"
-
-[network]
-# Enable/disable families; ipv6 = true/false/auto(None)
-ipv4 = true
-ipv6 = true
-# prefer = 4 or 6
-prefer = 4
-multipath = false
+# ad_tag = "00000000000000000000000000000000"
+# Path to proxy-secret binary (auto-downloaded if missing).
+proxy_secret_path = "proxy-secret"
+# disable_colors = false  # Disable colored output in logs (useful for files/systemd)
 
+# === Log Level ===
 # Log level: debug | verbose | normal | silent
 # Can be overridden with --silent or --log-level CLI flags
 # RUST_LOG env var takes absolute priority over all of these
 log_level = "normal"
 
+# === Middle Proxy - ME ===
+# Public IP override for ME KDF when behind NAT; leave unset to auto-detect.
+# middle_proxy_nat_ip = "203.0.113.10"
+# Enable STUN probing to discover public IP:port for ME.
+middle_proxy_nat_probe = true
+# Primary STUN server (host:port); defaults to Telegram STUN when empty.
+middle_proxy_nat_stun = "stun.l.google.com:19302"
+# Optional fallback STUN servers list.
+middle_proxy_nat_stun_servers = ["stun1.l.google.com:19302", "stun2.l.google.com:19302"]
+# Desired number of concurrent ME writers in pool.
+middle_proxy_pool_size = 16
+# Pre-initialized warm-standby ME connections kept idle.
+middle_proxy_warm_standby = 8
+# Ignore STUN/interface mismatch and keep ME enabled even if IP differs.
+stun_iface_mismatch_ignore = false
+# Keepalive padding frames - fl==4
+me_keepalive_enabled = true
+me_keepalive_interval_secs = 25           # Period between keepalives
+me_keepalive_jitter_secs = 5              # Jitter added to interval
+me_keepalive_payload_random = true        # Randomize 4-byte payload (vs zeros)
+# Stagger extra ME connections on warmup to de-phase lifecycles.
+me_warmup_stagger_enabled = true
+me_warmup_step_delay_ms = 500             # Base delay between extra connects
+me_warmup_step_jitter_ms = 300            # Jitter for warmup delay
+# Reconnect policy knobs.
+me_reconnect_max_concurrent_per_dc = 1    # Parallel reconnects per DC - EXPERIMENTAL! UNSTABLE!
+me_reconnect_backoff_base_ms = 500        # Backoff start
+me_reconnect_backoff_cap_ms = 30000       # Backoff cap
+me_reconnect_fast_retry_count = 11        # Quick retries before backoff
+
 [general.modes]
 classic = false
 secure = false
 tls = true
 
+[general.links]
+show = "*"
+# show = ["alice", "bob"] # Only show links for alice and bob
+# show = "*"              # Show links for all users
+# public_host = "proxy.example.com"  # Host (IP or domain) for tg:// links
+# public_port = 443                  # Port for tg:// links (default: server.port)
+
+# === Network Parameters ===
+[network]
+# Enable/disable families: true/false/auto(None)
+ipv4 = true
+ipv6 = false # UNSTABLE WITH ME
+# prefer = 4 or 6
+prefer = 4
+multipath = false # EXPERIMENTAL!
+
 # === Server Binding ===
 [server]
 port = 443
@@ -31,38 +71,39 @@ listen_addr_ipv4 = "0.0.0.0"
 listen_addr_ipv6 = "::"
 # listen_unix_sock = "/var/run/telemt.sock" # Unix socket
 # listen_unix_sock_perm = "0666" # Socket file permissions
+# proxy_protocol = false           # Enable if behind HAProxy/nginx with PROXY protocol
 # metrics_port = 9090
 # metrics_whitelist = ["127.0.0.1", "::1"]
 
-# Listen on multiple interfaces/IPs (overrides listen_addr_*)
+# Listen on multiple interfaces/IPs - IPv4
 [[server.listeners]]
 ip = "0.0.0.0"
-# announce_ip = "1.2.3.4" # Optional: Public IP for tg:// links
 
+# Listen on multiple interfaces/IPs - IPv6
 [[server.listeners]]
 ip = "::"
 
-# Users to show in the startup log (tg:// links)
-[general.links]
-show = ["hello"] # Users to show in the startup log (tg:// links)
-# public_host = "proxy.example.com"  # Host (IP or domain) for tg:// links
-# public_port = 443                  # Port for tg:// links (default: server.port)
-
 # === Timeouts (in seconds) ===
 [timeouts]
-client_handshake = 15
+client_handshake = 30
 tg_connect = 10
 client_keepalive = 60
 client_ack = 300
+# Quick ME reconnects for single-address DCs (count and per-attempt timeout, ms).
+me_one_retry = 12
+me_one_timeout_ms = 1200
 
 # === Anti-Censorship & Masking ===
 [censorship]
 tls_domain = "petrovich.ru"
+# tls_domains = ["example.com", "cdn.example.net"] # Additional domains for EE links
 mask = true
 mask_port = 443
 # mask_host = "petrovich.ru" # Defaults to tls_domain if not set
 # mask_unix_sock = "/var/run/nginx.sock" # Unix socket (mutually exclusive with mask_host)
 fake_cert_len = 2048
+# tls_emulation = false        # Fetch real cert lengths and emulate TLS records
+# tls_front_dir = "tlsfront"   # Cache directory for TLS emulation
 
 # === Access Control & Users ===
 [access]
@@ -83,11 +124,17 @@ hello = "00000000000000000000000000000000"
 # [access.user_data_quota]
 # hello = 1073741824 # 1 GB
 
+# [access.user_expirations]
+# format: username = "[year]-[month]-[day]T[hour]:[minute]:[second]Z" UTC
+# hello = "2027-01-01T00:00:00Z"
+
 # === Upstreams & Routing ===
 [[upstreams]]
 type = "direct"
 enabled = true
 weight = 10
+# interface = "192.168.1.100"           # Bind outgoing to specific IP or iface name
+# bind_addresses = ["192.168.1.100"]    # List for round-robin binding (family must match target)
 
 # [[upstreams]]
 # type = "socks5"

src/cli.rs

@@ -213,6 +213,7 @@ listen_addr_ipv6 = "::"
 
 [[server.listeners]]
 ip = "0.0.0.0"
+# reuse_allow = false # Set true only when intentionally running multiple telemt instances on same port
 
 [[server.listeners]]
 ip = "::"

src/config/defaults.rs

@@ -1,5 +1,6 @@
 use std::net::IpAddr;
 use std::collections::HashMap;
+use ipnetwork::IpNetwork;
 use serde::Deserialize;
 
 // Helper defaults kept private to the config module.
@@ -23,6 +24,10 @@ pub(crate) fn default_fake_cert_len() -> usize {
     2048
 }
 
+pub(crate) fn default_tls_front_dir() -> String {
+    "tlsfront".to_string()
+}
+
 pub(crate) fn default_replay_check_len() -> usize {
     65_536
 }
@@ -62,8 +67,11 @@ pub(crate) fn default_weight() -> u16 {
     1
 }
 
-pub(crate) fn default_metrics_whitelist() -> Vec<IpAddr> {
-    vec!["127.0.0.1".parse().unwrap(), "::1".parse().unwrap()]
+pub(crate) fn default_metrics_whitelist() -> Vec<IpNetwork> {
+    vec![
+        "127.0.0.1/32".parse().unwrap(),
+        "::1/128".parse().unwrap(),
+    ]
 }
 
 pub(crate) fn default_prefer_4() -> u8 {
@@ -74,6 +82,103 @@ pub(crate) fn default_unknown_dc_log_path() -> Option<String> {
     Some("unknown-dc.txt".to_string())
 }
 
+pub(crate) fn default_pool_size() -> usize {
+    2
+}
+
+pub(crate) fn default_keepalive_interval() -> u64 {
+    25
+}
+
+pub(crate) fn default_keepalive_jitter() -> u64 {
+    5
+}
+
+pub(crate) fn default_warmup_step_delay_ms() -> u64 {
+    500
+}
+
+pub(crate) fn default_warmup_step_jitter_ms() -> u64 {
+    300
+}
+
+pub(crate) fn default_reconnect_backoff_base_ms() -> u64 {
+    500
+}
+
+pub(crate) fn default_reconnect_backoff_cap_ms() -> u64 {
+    30_000
+}
+
+pub(crate) fn default_crypto_pending_buffer() -> usize {
+    256 * 1024
+}
+
+pub(crate) fn default_max_client_frame() -> usize {
+    16 * 1024 * 1024
+}
+
+pub(crate) fn default_tls_new_session_tickets() -> u8 {
+    0
+}
+
+pub(crate) fn default_server_hello_delay_min_ms() -> u64 {
+    0
+}
+
+pub(crate) fn default_server_hello_delay_max_ms() -> u64 {
+    0
+}
+
+pub(crate) fn default_alpn_enforce() -> bool {
+    true
+}
+
+pub(crate) fn default_stun_servers() -> Vec<String> {
+    vec![
+        "stun.l.google.com:19302".to_string(),
+        "stun1.l.google.com:19302".to_string(),
+        "stun2.l.google.com:19302".to_string(),
+        "stun.stunprotocol.org:3478".to_string(),
+        "stun.voip.eutelia.it:3478".to_string(),
+    ]
+}
+
+pub(crate) fn default_http_ip_detect_urls() -> Vec<String> {
+    vec![
+        "https://ifconfig.me/ip".to_string(),
+        "https://api.ipify.org".to_string(),
+    ]
+}
+
+pub(crate) fn default_cache_public_ip_path() -> String {
+    "cache/public_ip.txt".to_string()
+}
+
+pub(crate) fn default_proxy_secret_reload_secs() -> u64 {
+    12 * 60 * 60
+}
+
+pub(crate) fn default_proxy_config_reload_secs() -> u64 {
+    12 * 60 * 60
+}
+
+pub(crate) fn default_ntp_check() -> bool {
+    true
+}
+
+pub(crate) fn default_ntp_servers() -> Vec<String> {
+    vec!["pool.ntp.org".to_string()]
+}
+
+pub(crate) fn default_fast_mode_min_tls_record() -> usize {
+    0
+}
+
+pub(crate) fn default_degradation_min_unavailable_dc_groups() -> u8 {
+    2
+}
+
 // Custom deserializer helpers
 
 #[derive(Deserialize)]

src/config/hot_reload.rs

@@ -0,0 +1,421 @@
+//! Hot-reload: watches the config file via inotify (Linux) / FSEvents (macOS)
+//! / ReadDirectoryChangesW (Windows) using the `notify` crate.
+//! SIGHUP is also supported on Unix as an additional manual trigger.
+//!
+//! # What can be reloaded without restart
+//!
+//! | Section   | Field                         | Effect                            |
+//! |-----------|-------------------------------|-----------------------------------|
+//! | `general` | `log_level`                   | Filter updated via `log_level_tx` |
+//! | `general` | `ad_tag`                      | Passed on next connection         |
+//! | `general` | `middle_proxy_pool_size`      | Passed on next connection         |
+//! | `general` | `me_keepalive_*`              | Passed on next connection         |
+//! | `access`  | All user/quota fields         | Effective immediately             |
+//!
+//! Fields that require re-binding sockets (`server.port`, `censorship.*`,
+//! `network.*`, `use_middle_proxy`) are **not** applied; a warning is emitted.
+
+use std::net::IpAddr;
+use std::path::PathBuf;
+use std::sync::Arc;
+
+use notify::{EventKind, RecursiveMode, Watcher, recommended_watcher};
+use tokio::sync::{mpsc, watch};
+use tracing::{error, info, warn};
+
+use crate::config::LogLevel;
+use super::load::ProxyConfig;
+
+// ── Hot fields ────────────────────────────────────────────────────────────────
+
+/// Fields that are safe to swap without restarting listeners.
+#[derive(Debug, Clone, PartialEq)]
+pub struct HotFields {
+    pub log_level:               LogLevel,
+    pub ad_tag:                  Option<String>,
+    pub middle_proxy_pool_size:  usize,
+    pub me_keepalive_enabled:    bool,
+    pub me_keepalive_interval_secs: u64,
+    pub me_keepalive_jitter_secs:   u64,
+    pub me_keepalive_payload_random: bool,
+    pub access:                  crate::config::AccessConfig,
+}
+
+impl HotFields {
+    pub fn from_config(cfg: &ProxyConfig) -> Self {
+        Self {
+            log_level:               cfg.general.log_level.clone(),
+            ad_tag:                  cfg.general.ad_tag.clone(),
+            middle_proxy_pool_size:  cfg.general.middle_proxy_pool_size,
+            me_keepalive_enabled:    cfg.general.me_keepalive_enabled,
+            me_keepalive_interval_secs: cfg.general.me_keepalive_interval_secs,
+            me_keepalive_jitter_secs:   cfg.general.me_keepalive_jitter_secs,
+            me_keepalive_payload_random: cfg.general.me_keepalive_payload_random,
+            access:                  cfg.access.clone(),
+        }
+    }
+}
+
+// ── Helpers ───────────────────────────────────────────────────────────────────
+
+/// Warn if any non-hot fields changed (require restart).
+fn warn_non_hot_changes(old: &ProxyConfig, new: &ProxyConfig) {
+    if old.server.port != new.server.port {
+        warn!(
+            "config reload: server.port changed ({} → {}); restart required",
+            old.server.port, new.server.port
+        );
+    }
+    if old.censorship.tls_domain != new.censorship.tls_domain {
+        warn!(
+            "config reload: censorship.tls_domain changed ('{}' → '{}'); restart required",
+            old.censorship.tls_domain, new.censorship.tls_domain
+        );
+    }
+    if old.network.ipv4 != new.network.ipv4 || old.network.ipv6 != new.network.ipv6 {
+        warn!("config reload: network.ipv4/ipv6 changed; restart required");
+    }
+    if old.general.use_middle_proxy != new.general.use_middle_proxy {
+        warn!("config reload: use_middle_proxy changed; restart required");
+    }
+}
+
+/// Resolve the public host for link generation — mirrors the logic in main.rs.
+///
+/// Priority:
+/// 1. `[general.links] public_host` — explicit override in config
+/// 2. `detected_ip_v4` — from STUN/interface probe at startup
+/// 3. `detected_ip_v6` — fallback
+/// 4. `"UNKNOWN"` — warn the user to set `public_host`
+fn resolve_link_host(
+    cfg: &ProxyConfig,
+    detected_ip_v4: Option<IpAddr>,
+    detected_ip_v6: Option<IpAddr>,
+) -> String {
+    if let Some(ref h) = cfg.general.links.public_host {
+        return h.clone();
+    }
+    detected_ip_v4
+        .or(detected_ip_v6)
+        .map(|ip| ip.to_string())
+        .unwrap_or_else(|| {
+            warn!(
+                "config reload: could not determine public IP for proxy links. \
+                 Set [general.links] public_host in config."
+            );
+            "UNKNOWN".to_string()
+        })
+}
+
+/// Print TG proxy links for a single user — mirrors print_proxy_links() in main.rs.
+fn print_user_links(user: &str, secret: &str, host: &str, port: u16, cfg: &ProxyConfig) {
+    info!(target: "telemt::links", "--- New user: {} ---", user);
+    if cfg.general.modes.classic {
+        info!(
+            target: "telemt::links",
+            "  Classic: tg://proxy?server={}&port={}&secret={}",
+            host, port, secret
+        );
+    }
+    if cfg.general.modes.secure {
+        info!(
+            target: "telemt::links",
+            "  DD:      tg://proxy?server={}&port={}&secret=dd{}",
+            host, port, secret
+        );
+    }
+    if cfg.general.modes.tls {
+        let mut domains = vec![cfg.censorship.tls_domain.clone()];
+        for d in &cfg.censorship.tls_domains {
+            if !domains.contains(d) {
+                domains.push(d.clone());
+            }
+        }
+        for domain in &domains {
+            let domain_hex = hex::encode(domain.as_bytes());
+            info!(
+                target: "telemt::links",
+                "  EE-TLS:  tg://proxy?server={}&port={}&secret=ee{}{}",
+                host, port, secret, domain_hex
+            );
+        }
+    }
+    info!(target: "telemt::links", "--------------------");
+}
+
+/// Log all detected changes and emit TG links for new users.
+fn log_changes(
+    old_hot: &HotFields,
+    new_hot: &HotFields,
+    new_cfg: &ProxyConfig,
+    log_tx: &watch::Sender<LogLevel>,
+    detected_ip_v4: Option<IpAddr>,
+    detected_ip_v6: Option<IpAddr>,
+) {
+    if old_hot.log_level != new_hot.log_level {
+        info!(
+            "config reload: log_level: '{}' → '{}'",
+            old_hot.log_level, new_hot.log_level
+        );
+        log_tx.send(new_hot.log_level.clone()).ok();
+    }
+
+    if old_hot.ad_tag != new_hot.ad_tag {
+        info!(
+            "config reload: ad_tag: {} → {}",
+            old_hot.ad_tag.as_deref().unwrap_or("none"),
+            new_hot.ad_tag.as_deref().unwrap_or("none"),
+        );
+    }
+
+    if old_hot.middle_proxy_pool_size != new_hot.middle_proxy_pool_size {
+        info!(
+            "config reload: middle_proxy_pool_size: {} → {}",
+            old_hot.middle_proxy_pool_size, new_hot.middle_proxy_pool_size,
+        );
+    }
+
+    if old_hot.me_keepalive_enabled        != new_hot.me_keepalive_enabled
+    || old_hot.me_keepalive_interval_secs  != new_hot.me_keepalive_interval_secs
+    || old_hot.me_keepalive_jitter_secs    != new_hot.me_keepalive_jitter_secs
+    || old_hot.me_keepalive_payload_random != new_hot.me_keepalive_payload_random
+    {
+        info!(
+            "config reload: me_keepalive: enabled={} interval={}s jitter={}s random_payload={}",
+            new_hot.me_keepalive_enabled,
+            new_hot.me_keepalive_interval_secs,
+            new_hot.me_keepalive_jitter_secs,
+            new_hot.me_keepalive_payload_random,
+        );
+    }
+
+    if old_hot.access.users != new_hot.access.users {
+        let mut added: Vec<&String> = new_hot.access.users.keys()
+            .filter(|u| !old_hot.access.users.contains_key(*u))
+            .collect();
+        added.sort();
+
+        let mut removed: Vec<&String> = old_hot.access.users.keys()
+            .filter(|u| !new_hot.access.users.contains_key(*u))
+            .collect();
+        removed.sort();
+
+        let mut changed: Vec<&String> = new_hot.access.users.keys()
+            .filter(|u| {
+                old_hot.access.users.get(*u)
+                    .map(|s| s != &new_hot.access.users[*u])
+                    .unwrap_or(false)
+            })
+            .collect();
+        changed.sort();
+
+        if !added.is_empty() {
+            info!(
+                "config reload: users added: [{}]",
+                added.iter().map(|s| s.as_str()).collect::<Vec<_>>().join(", ")
+            );
+            let host = resolve_link_host(new_cfg, detected_ip_v4, detected_ip_v6);
+            let port = new_cfg.general.links.public_port.unwrap_or(new_cfg.server.port);
+            for user in &added {
+                if let Some(secret) = new_hot.access.users.get(*user) {
+                    print_user_links(user, secret, &host, port, new_cfg);
+                }
+            }
+        }
+        if !removed.is_empty() {
+            info!(
+                "config reload: users removed: [{}]",
+                removed.iter().map(|s| s.as_str()).collect::<Vec<_>>().join(", ")
+            );
+        }
+        if !changed.is_empty() {
+            info!(
+                "config reload: users secret changed: [{}]",
+                changed.iter().map(|s| s.as_str()).collect::<Vec<_>>().join(", ")
+            );
+        }
+    }
+
+    if old_hot.access.user_max_tcp_conns != new_hot.access.user_max_tcp_conns {
+        info!(
+            "config reload: user_max_tcp_conns updated ({} entries)",
+            new_hot.access.user_max_tcp_conns.len()
+        );
+    }
+    if old_hot.access.user_expirations != new_hot.access.user_expirations {
+        info!(
+            "config reload: user_expirations updated ({} entries)",
+            new_hot.access.user_expirations.len()
+        );
+    }
+    if old_hot.access.user_data_quota != new_hot.access.user_data_quota {
+        info!(
+            "config reload: user_data_quota updated ({} entries)",
+            new_hot.access.user_data_quota.len()
+        );
+    }
+    if old_hot.access.user_max_unique_ips != new_hot.access.user_max_unique_ips {
+        info!(
+            "config reload: user_max_unique_ips updated ({} entries)",
+            new_hot.access.user_max_unique_ips.len()
+        );
+    }
+}
+
+/// Load config, validate, diff against current, and broadcast if changed.
+fn reload_config(
+    config_path: &PathBuf,
+    config_tx: &watch::Sender<Arc<ProxyConfig>>,
+    log_tx: &watch::Sender<LogLevel>,
+    detected_ip_v4: Option<IpAddr>,
+    detected_ip_v6: Option<IpAddr>,
+) {
+    let new_cfg = match ProxyConfig::load(config_path) {
+        Ok(c) => c,
+        Err(e) => {
+            error!("config reload: failed to parse {:?}: {}", config_path, e);
+            return;
+        }
+    };
+
+    if let Err(e) = new_cfg.validate() {
+        error!("config reload: validation failed: {}; keeping old config", e);
+        return;
+    }
+
+    let old_cfg = config_tx.borrow().clone();
+    let old_hot = HotFields::from_config(&old_cfg);
+    let new_hot = HotFields::from_config(&new_cfg);
+
+    if old_hot == new_hot {
+        return;
+    }
+
+    warn_non_hot_changes(&old_cfg, &new_cfg);
+    log_changes(&old_hot, &new_hot, &new_cfg, log_tx, detected_ip_v4, detected_ip_v6);
+    config_tx.send(Arc::new(new_cfg)).ok();
+}
+
+// ── Public API ────────────────────────────────────────────────────────────────
+
+/// Spawn the hot-reload watcher task.
+///
+/// Uses `notify` (inotify on Linux) to detect file changes instantly.
+/// SIGHUP is also handled on Unix as an additional manual trigger.
+///
+/// `detected_ip_v4` / `detected_ip_v6` are the IPs discovered during the
+/// startup probe — used when generating proxy links for newly added users,
+/// matching the same logic as the startup output.
+pub fn spawn_config_watcher(
+    config_path: PathBuf,
+    initial: Arc<ProxyConfig>,
+    detected_ip_v4: Option<IpAddr>,
+    detected_ip_v6: Option<IpAddr>,
+) -> (watch::Receiver<Arc<ProxyConfig>>, watch::Receiver<LogLevel>) {
+    let initial_level = initial.general.log_level.clone();
+    let (config_tx, config_rx) = watch::channel(initial);
+    let (log_tx, log_rx)       = watch::channel(initial_level);
+
+    // Bridge: sync notify callbacks → async task via mpsc.
+    let (notify_tx, mut notify_rx) = mpsc::channel::<()>(4);
+
+    // Canonicalize so path matches what notify returns (absolute) in events.
+    let config_path = match config_path.canonicalize() {
+        Ok(p) => p,
+        Err(_) => config_path.to_path_buf(),
+    };
+
+    // Watch the parent directory rather than the file itself, because many
+    // editors (vim, nano) and systemd write via rename, which would cause
+    // inotify to lose track of the original inode.
+    let watch_dir = config_path
+        .parent()
+        .unwrap_or_else(|| std::path::Path::new("."))
+        .to_path_buf();
+
+    // ── inotify watcher (instant on local fs) ────────────────────────────
+    let config_file = config_path.clone();
+    let tx_inotify  = notify_tx.clone();
+    let inotify_ok = match recommended_watcher(move |res: notify::Result<notify::Event>| {
+        let Ok(event) = res else { return };
+        let is_our_file = event.paths.iter().any(|p| p == &config_file);
+        if !is_our_file { return; }
+        if matches!(event.kind, EventKind::Modify(_) | EventKind::Create(_) | EventKind::Remove(_)) {
+            let _ = tx_inotify.try_send(());
+        }
+    }) {
+        Ok(mut w) => match w.watch(&watch_dir, RecursiveMode::NonRecursive) {
+            Ok(()) => {
+                info!("config watcher: inotify active on {:?}", config_path);
+                Box::leak(Box::new(w));
+                true
+            }
+            Err(e) => { warn!("config watcher: inotify watch failed: {}", e); false }
+        },
+        Err(e) => { warn!("config watcher: inotify unavailable: {}", e); false }
+    };
+
+    // ── poll watcher (always active, fixes Docker bind mounts / NFS) ─────
+    // inotify does not receive events for files mounted from the host into
+    // a container. PollWatcher compares file contents every 3 s and fires
+    // on any change regardless of the underlying fs.
+    let config_file2 = config_path.clone();
+    let tx_poll      = notify_tx.clone();
+    match notify::poll::PollWatcher::new(
+        move |res: notify::Result<notify::Event>| {
+            let Ok(event) = res else { return };
+            let is_our_file = event.paths.iter().any(|p| p == &config_file2);
+            if !is_our_file { return; }
+            if matches!(event.kind, EventKind::Modify(_) | EventKind::Create(_) | EventKind::Remove(_)) {
+                let _ = tx_poll.try_send(());
+            }
+        },
+        notify::Config::default()
+            .with_poll_interval(std::time::Duration::from_secs(3))
+            .with_compare_contents(true),
+    ) {
+        Ok(mut w) => match w.watch(&config_path, RecursiveMode::NonRecursive) {
+            Ok(()) => {
+                if inotify_ok {
+                    info!("config watcher: poll watcher also active (Docker/NFS safe)");
+                } else {
+                    info!("config watcher: poll watcher active on {:?} (3s interval)", config_path);
+                }
+                Box::leak(Box::new(w));
+            }
+            Err(e) => warn!("config watcher: poll watch failed: {}", e),
+        },
+        Err(e) => warn!("config watcher: poll watcher unavailable: {}", e),
+    }
+
+    // ── event loop ───────────────────────────────────────────────────────
+    tokio::spawn(async move {
+        #[cfg(unix)]
+        let mut sighup = {
+            use tokio::signal::unix::{SignalKind, signal};
+            signal(SignalKind::hangup()).expect("Failed to register SIGHUP handler")
+        };
+
+        loop {
+            #[cfg(unix)]
+            tokio::select! {
+                msg = notify_rx.recv() => {
+                    if msg.is_none() { break; }
+                }
+                _ = sighup.recv() => {
+                    info!("SIGHUP received — reloading {:?}", config_path);
+                }
+            }
+            #[cfg(not(unix))]
+            if notify_rx.recv().await.is_none() { break; }
+
+            // Debounce: drain extra events that arrive within 50 ms.
+            tokio::time::sleep(std::time::Duration::from_millis(50)).await;
+            while notify_rx.try_recv().is_ok() {}
+
+            reload_config(&config_path, &config_tx, &log_tx, detected_ip_v4, detected_ip_v6);
+        }
+    });
+
+    (config_rx, log_rx)
+}

src/config/load.rs

@@ -11,6 +11,32 @@ use crate::error::{ProxyError, Result};
 use super::defaults::*;
 use super::types::*;
 
+fn preprocess_includes(content: &str, base_dir: &Path, depth: u8) -> Result<String> {
+    if depth > 10 {
+        return Err(ProxyError::Config("Include depth > 10".into()));
+    }
+    let mut output = String::with_capacity(content.len());
+    for line in content.lines() {
+        let trimmed = line.trim();
+        if let Some(rest) = trimmed.strip_prefix("include") {
+            let rest = rest.trim();
+            if let Some(rest) = rest.strip_prefix('=') {
+                let path_str = rest.trim().trim_matches('"');
+                let resolved = base_dir.join(path_str);
+                let included = std::fs::read_to_string(&resolved)
+                    .map_err(|e| ProxyError::Config(e.to_string()))?;
+                let included_dir = resolved.parent().unwrap_or(base_dir);
+                output.push_str(&preprocess_includes(&included, included_dir, depth + 1)?);
+                output.push('\n');
+                continue;
+            }
+        }
+        output.push_str(line);
+        output.push('\n');
+    }
+    Ok(output)
+}
+
 fn validate_network_cfg(net: &mut NetworkConfig) -> Result<()> {
     if !net.ipv4 && matches!(net.ipv6, Some(false)) {
         return Err(ProxyError::Config(
@@ -84,10 +110,12 @@ pub struct ProxyConfig {
 impl ProxyConfig {
     pub fn load<P: AsRef<Path>>(path: P) -> Result<Self> {
         let content =
-            std::fs::read_to_string(path).map_err(|e| ProxyError::Config(e.to_string()))?;
+            std::fs::read_to_string(&path).map_err(|e| ProxyError::Config(e.to_string()))?;
+        let base_dir = path.as_ref().parent().unwrap_or(Path::new("."));
+        let processed = preprocess_includes(&content, base_dir, 0)?;
 
         let mut config: ProxyConfig =
-            toml::from_str(&content).map_err(|e| ProxyError::Config(e.to_string()))?;
+            toml::from_str(&processed).map_err(|e| ProxyError::Config(e.to_string()))?;
 
         // Validate secrets.
         for (user, secret) in &config.access.users {
@@ -135,6 +163,21 @@ impl ProxyConfig {
             config.censorship.mask_host = Some(config.censorship.tls_domain.clone());
         }
 
+        // Merge primary + extra TLS domains, deduplicate (primary always first).
+        if !config.censorship.tls_domains.is_empty() {
+            let mut all = Vec::with_capacity(1 + config.censorship.tls_domains.len());
+            all.push(config.censorship.tls_domain.clone());
+            for d in std::mem::take(&mut config.censorship.tls_domains) {
+                if !d.is_empty() && !all.contains(&d) {
+                    all.push(d);
+                }
+            }
+            // keep primary as tls_domain; store remaining back to tls_domains
+            if all.len() > 1 {
+                config.censorship.tls_domains = all[1..].to_vec();
+            }
+        }
+
         // Migration: prefer_ipv6 -> network.prefer.
         if config.general.prefer_ipv6 {
             if config.network.prefer == 4 {
@@ -151,8 +194,14 @@ impl ProxyConfig {
 
         validate_network_cfg(&mut config.network)?;
 
-        // Random fake_cert_len.
-        config.censorship.fake_cert_len = rand::rng().gen_range(1024..4096);
+        if config.general.use_middle_proxy && config.network.ipv6 == Some(true) {
+            warn!("IPv6 with Middle Proxy is experimental and may cause KDF address mismatch; consider disabling IPv6 or ME");
+        }
+
+        // Random fake_cert_len only when default is in use.
+        if !config.censorship.tls_emulation && config.censorship.fake_cert_len == default_fake_cert_len() {
+            config.censorship.fake_cert_len = rand::rng().gen_range(1024..4096);
+        }
 
         // Resolve listen_tcp: explicit value wins, otherwise auto-detect.
         // If unix socket is set → TCP only when listen_addr_ipv4 or listeners are explicitly provided.
@@ -177,6 +226,8 @@ impl ProxyConfig {
                     ip: ipv4,
                     announce: None,
                     announce_ip: None,
+                    proxy_protocol: None,
+                    reuse_allow: false,
                 });
             }
             if let Some(ipv6_str) = &config.server.listen_addr_ipv6 {
@@ -185,6 +236,8 @@ impl ProxyConfig {
                         ip: ipv6,
                         announce: None,
                         announce_ip: None,
+                        proxy_protocol: None,
+                        reuse_allow: false,
                     });
                 }
             }
@@ -205,9 +258,11 @@ impl ProxyConfig {
         // Migration: Populate upstreams if empty (Default Direct).
         if config.upstreams.is_empty() {
             config.upstreams.push(UpstreamConfig {
-                upstream_type: UpstreamType::Direct { interface: None },
+                upstream_type: UpstreamType::Direct { interface: None, bind_addresses: None },
                 weight: 1,
                 enabled: true,
+                scopes: String::new(),
+                selected_scope: String::new(),
             });
         }
 

src/config/mod.rs

@@ -3,6 +3,7 @@
 pub(crate) mod defaults;
 mod types;
 mod load;
+pub mod hot_reload;
 
 pub use load::ProxyConfig;
 pub use types::*;

src/config/types.rs

@@ -1,4 +1,5 @@
 use chrono::{DateTime, Utc};
+use ipnetwork::IpNetwork;
 use serde::{Deserialize, Serialize};
 use std::collections::HashMap;
 use std::net::IpAddr;
@@ -73,8 +74,8 @@ pub struct ProxyModes {
 impl Default for ProxyModes {
     fn default() -> Self {
         Self {
-            classic: true,
-            secure: true,
+            classic: false,
+            secure: false,
             tls: true,
         }
     }
@@ -95,15 +96,35 @@ pub struct NetworkConfig {
 
     #[serde(default)]
     pub multipath: bool,
+
+    /// STUN servers list for public IP discovery.
+    #[serde(default = "default_stun_servers")]
+    pub stun_servers: Vec<String>,
+
+    /// Enable TCP STUN fallback when UDP is blocked.
+    #[serde(default)]
+    pub stun_tcp_fallback: bool,
+
+    /// HTTP-based public IP detection endpoints (fallback after STUN).
+    #[serde(default = "default_http_ip_detect_urls")]
+    pub http_ip_detect_urls: Vec<String>,
+
+    /// Cache file path for detected public IP.
+    #[serde(default = "default_cache_public_ip_path")]
+    pub cache_public_ip_path: String,
 }
 
 impl Default for NetworkConfig {
     fn default() -> Self {
         Self {
             ipv4: true,
-            ipv6: None,
+            ipv6: Some(false),
             prefer: 4,
             multipath: false,
+            stun_servers: default_stun_servers(),
+            stun_tcp_fallback: true,
+            http_ip_detect_urls: default_http_ip_detect_urls(),
+            cache_public_ip_path: default_cache_public_ip_path(),
         }
     }
 }
@@ -143,6 +164,71 @@ pub struct GeneralConfig {
     #[serde(default)]
     pub middle_proxy_nat_stun: Option<String>,
 
+    /// Optional list of STUN servers for NAT probing fallback.
+    #[serde(default)]
+    pub middle_proxy_nat_stun_servers: Vec<String>,
+
+    /// Desired size of active Middle-Proxy writer pool.
+    #[serde(default = "default_pool_size")]
+    pub middle_proxy_pool_size: usize,
+
+    /// Number of warm standby ME connections kept pre-initialized.
+    #[serde(default)]
+    pub middle_proxy_warm_standby: usize,
+
+    /// Enable ME keepalive padding frames.
+    #[serde(default = "default_true")]
+    pub me_keepalive_enabled: bool,
+
+    /// Keepalive interval in seconds.
+    #[serde(default = "default_keepalive_interval")]
+    pub me_keepalive_interval_secs: u64,
+
+    /// Keepalive jitter in seconds.
+    #[serde(default = "default_keepalive_jitter")]
+    pub me_keepalive_jitter_secs: u64,
+
+    /// Keepalive payload randomized (4 bytes); otherwise zeros.
+    #[serde(default = "default_true")]
+    pub me_keepalive_payload_random: bool,
+
+    /// Max pending ciphertext buffer per client writer (bytes).
+    /// Controls FakeTLS backpressure vs throughput.
+    #[serde(default = "default_crypto_pending_buffer")]
+    pub crypto_pending_buffer: usize,
+
+    /// Maximum allowed client MTProto frame size (bytes).
+    #[serde(default = "default_max_client_frame")]
+    pub max_client_frame: usize,
+
+    /// Enable staggered warmup of extra ME writers.
+    #[serde(default = "default_true")]
+    pub me_warmup_stagger_enabled: bool,
+
+    /// Base delay between warmup connections in ms.
+    #[serde(default = "default_warmup_step_delay_ms")]
+    pub me_warmup_step_delay_ms: u64,
+
+    /// Jitter for warmup delay in ms.
+    #[serde(default = "default_warmup_step_jitter_ms")]
+    pub me_warmup_step_jitter_ms: u64,
+
+    /// Max concurrent reconnect attempts per DC.
+    #[serde(default)]
+    pub me_reconnect_max_concurrent_per_dc: u32,
+
+    /// Base backoff in ms for reconnect.
+    #[serde(default = "default_reconnect_backoff_base_ms")]
+    pub me_reconnect_backoff_base_ms: u64,
+
+    /// Cap backoff in ms for reconnect.
+    #[serde(default = "default_reconnect_backoff_cap_ms")]
+    pub me_reconnect_backoff_cap_ms: u64,
+
+    /// Fast retry attempts before backoff.
+    #[serde(default)]
+    pub me_reconnect_fast_retry_count: u32,
+
     /// Ignore STUN/interface IP mismatch (keep using Middle Proxy even if NAT detected).
     #[serde(default)]
     pub stun_iface_mismatch_ignore: bool,
@@ -161,6 +247,34 @@ pub struct GeneralConfig {
     /// [general.links] — proxy link generation overrides.
     #[serde(default)]
     pub links: LinksConfig,
+
+    /// Minimum TLS record size when fast_mode coalescing is enabled (0 = disabled).
+    #[serde(default = "default_fast_mode_min_tls_record")]
+    pub fast_mode_min_tls_record: usize,
+
+    /// Automatically reload proxy-secret every N seconds.
+    #[serde(default = "default_proxy_secret_reload_secs")]
+    pub proxy_secret_auto_reload_secs: u64,
+
+    /// Automatically reload proxy-multi.conf every N seconds.
+    #[serde(default = "default_proxy_config_reload_secs")]
+    pub proxy_config_auto_reload_secs: u64,
+
+    /// Enable NTP drift check at startup.
+    #[serde(default = "default_ntp_check")]
+    pub ntp_check: bool,
+
+    /// NTP servers for drift check.
+    #[serde(default = "default_ntp_servers")]
+    pub ntp_servers: Vec<String>,
+
+    /// Enable auto-degradation from ME to Direct-DC.
+    #[serde(default = "default_true")]
+    pub auto_degradation_enabled: bool,
+
+    /// Minimum unavailable ME DC groups before degrading.
+    #[serde(default = "default_degradation_min_unavailable_dc_groups")]
+    pub degradation_min_unavailable_dc_groups: u8,
 }
 
 impl Default for GeneralConfig {
@@ -175,11 +289,34 @@ impl Default for GeneralConfig {
             middle_proxy_nat_ip: None,
             middle_proxy_nat_probe: false,
             middle_proxy_nat_stun: None,
+            middle_proxy_nat_stun_servers: Vec::new(),
+            middle_proxy_pool_size: default_pool_size(),
+            middle_proxy_warm_standby: 8,
+            me_keepalive_enabled: true,
+            me_keepalive_interval_secs: default_keepalive_interval(),
+            me_keepalive_jitter_secs: default_keepalive_jitter(),
+            me_keepalive_payload_random: true,
+            me_warmup_stagger_enabled: true,
+            me_warmup_step_delay_ms: default_warmup_step_delay_ms(),
+            me_warmup_step_jitter_ms: default_warmup_step_jitter_ms(),
+            me_reconnect_max_concurrent_per_dc: 4,
+            me_reconnect_backoff_base_ms: default_reconnect_backoff_base_ms(),
+            me_reconnect_backoff_cap_ms: default_reconnect_backoff_cap_ms(),
+            me_reconnect_fast_retry_count: 8,
             stun_iface_mismatch_ignore: false,
             unknown_dc_log_path: default_unknown_dc_log_path(),
             log_level: LogLevel::Normal,
             disable_colors: false,
             links: LinksConfig::default(),
+            crypto_pending_buffer: default_crypto_pending_buffer(),
+            max_client_frame: default_max_client_frame(),
+            fast_mode_min_tls_record: default_fast_mode_min_tls_record(),
+            proxy_secret_auto_reload_secs: default_proxy_secret_reload_secs(),
+            proxy_config_auto_reload_secs: default_proxy_config_reload_secs(),
+            ntp_check: default_ntp_check(),
+            ntp_servers: default_ntp_servers(),
+            auto_degradation_enabled: true,
+            degradation_min_unavailable_dc_groups: default_degradation_min_unavailable_dc_groups(),
         }
     }
 }
@@ -225,11 +362,16 @@ pub struct ServerConfig {
     #[serde(default)]
     pub listen_tcp: Option<bool>,
 
+    /// Accept HAProxy PROXY protocol headers on incoming connections.
+    /// When enabled, real client IPs are extracted from PROXY v1/v2 headers.
+    #[serde(default)]
+    pub proxy_protocol: bool,
+
     #[serde(default)]
     pub metrics_port: Option<u16>,
 
     #[serde(default = "default_metrics_whitelist")]
-    pub metrics_whitelist: Vec<IpAddr>,
+    pub metrics_whitelist: Vec<IpNetwork>,
 
     #[serde(default)]
     pub listeners: Vec<ListenerConfig>,
@@ -244,6 +386,7 @@ impl Default for ServerConfig {
             listen_unix_sock: None,
             listen_unix_sock_perm: None,
             listen_tcp: None,
+            proxy_protocol: false,
             metrics_port: None,
             metrics_whitelist: default_metrics_whitelist(),
             listeners: Vec::new(),
@@ -292,6 +435,10 @@ pub struct AntiCensorshipConfig {
     #[serde(default = "default_tls_domain")]
     pub tls_domain: String,
 
+    /// Additional TLS domains for generating multiple proxy links.
+    #[serde(default)]
+    pub tls_domains: Vec<String>,
+
     #[serde(default = "default_true")]
     pub mask: bool,
 
@@ -306,22 +453,53 @@ pub struct AntiCensorshipConfig {
 
     #[serde(default = "default_fake_cert_len")]
     pub fake_cert_len: usize,
+
+    /// Enable TLS certificate emulation using cached real certificates.
+    #[serde(default)]
+    pub tls_emulation: bool,
+
+    /// Directory to store TLS front cache (on disk).
+    #[serde(default = "default_tls_front_dir")]
+    pub tls_front_dir: String,
+
+    /// Minimum server_hello delay in milliseconds (anti-fingerprint).
+    #[serde(default = "default_server_hello_delay_min_ms")]
+    pub server_hello_delay_min_ms: u64,
+
+    /// Maximum server_hello delay in milliseconds.
+    #[serde(default = "default_server_hello_delay_max_ms")]
+    pub server_hello_delay_max_ms: u64,
+
+    /// Number of NewSessionTicket messages to emit post-handshake.
+    #[serde(default = "default_tls_new_session_tickets")]
+    pub tls_new_session_tickets: u8,
+
+    /// Enforce ALPN echo of client preference.
+    #[serde(default = "default_alpn_enforce")]
+    pub alpn_enforce: bool,
 }
 
 impl Default for AntiCensorshipConfig {
     fn default() -> Self {
         Self {
             tls_domain: default_tls_domain(),
+            tls_domains: Vec::new(),
             mask: true,
             mask_host: None,
             mask_port: default_mask_port(),
             mask_unix_sock: None,
             fake_cert_len: default_fake_cert_len(),
+            tls_emulation: false,
+            tls_front_dir: default_tls_front_dir(),
+            server_hello_delay_min_ms: default_server_hello_delay_min_ms(),
+            server_hello_delay_max_ms: default_server_hello_delay_max_ms(),
+            tls_new_session_tickets: default_tls_new_session_tickets(),
+            alpn_enforce: default_alpn_enforce(),
         }
     }
 }
 
-#[derive(Debug, Clone, Serialize, Deserialize)]
+#[derive(Debug, Clone, PartialEq, Serialize, Deserialize)]
 pub struct AccessConfig {
     #[serde(default)]
     pub users: HashMap<String, String>,
@@ -376,6 +554,8 @@ pub enum UpstreamType {
     Direct {
         #[serde(default)]
         interface: Option<String>,
+        #[serde(default)]
+        bind_addresses: Option<Vec<String>>,
     },
     Socks4 {
         address: String,
@@ -403,6 +583,10 @@ pub struct UpstreamConfig {
     pub weight: u16,
     #[serde(default = "default_true")]
     pub enabled: bool,
+    #[serde(default)]
+    pub scopes: String,
+    #[serde(skip)]
+    pub selected_scope: String,
 }
 
 #[derive(Debug, Clone, Serialize, Deserialize)]
@@ -416,6 +600,13 @@ pub struct ListenerConfig {
     /// Migrated to `announce` automatically if `announce` is not set.
     #[serde(default)]
     pub announce_ip: Option<IpAddr>,
+    /// Per-listener PROXY protocol override. When set, overrides global server.proxy_protocol.
+    #[serde(default)]
+    pub proxy_protocol: Option<bool>,
+    /// Allow multiple telemt instances to listen on the same IP:port (SO_REUSEPORT).
+    /// Default is false for safety.
+    #[serde(default)]
+    pub reuse_allow: bool,
 }
 
 // ============= ShowLink =============

src/crypto/hash.rs

@@ -55,6 +55,11 @@ pub fn crc32(data: &[u8]) -> u32 {
     crc32fast::hash(data)
 }
 
+/// CRC32C (Castagnoli)
+pub fn crc32c(data: &[u8]) -> u32 {
+    crc32c::crc32c(data)
+}
+
 /// Build the exact prekey buffer used by Telegram Middle Proxy KDF.
 ///
 /// Returned buffer layout (IPv4):

src/crypto/mod.rs

@@ -5,5 +5,8 @@ pub mod hash;
 pub mod random;
 
 pub use aes::{AesCtr, AesCbc};
-pub use hash::{sha256, sha256_hmac, sha1, md5, crc32, derive_middleproxy_keys, build_middleproxy_prekey};
+pub use hash::{
+    build_middleproxy_prekey, crc32, crc32c, derive_middleproxy_keys, md5, sha1, sha256,
+    sha256_hmac,
+};
 pub use random::SecureRandom;

src/crypto/random.rs

@@ -11,6 +11,9 @@ pub struct SecureRandom {
     inner: Mutex<SecureRandomInner>,
 }
 
+unsafe impl Send for SecureRandom {}
+unsafe impl Sync for SecureRandom {}
+
 struct SecureRandomInner {
     rng: StdRng,
     cipher: AesCtr,
@@ -211,4 +214,4 @@ mod tests {
         
         assert_ne!(shuffled, original);
     }
-}
+}

src/main.rs

@@ -3,6 +3,7 @@
 use std::net::SocketAddr;
 use std::sync::Arc;
 use std::time::Duration;
+use rand::Rng;
 use tokio::net::TcpListener;
 use tokio::signal;
 use tokio::sync::Semaphore;
@@ -23,9 +24,11 @@ mod proxy;
 mod stats;
 mod stream;
 mod transport;
+mod tls_front;
 mod util;
 
 use crate::config::{LogLevel, ProxyConfig};
+use crate::config::hot_reload::spawn_config_watcher;
 use crate::crypto::SecureRandom;
 use crate::ip_tracker::UserIpTracker;
 use crate::network::probe::{decide_network_capabilities, log_probe_result, run_probe};
@@ -35,7 +38,8 @@ use crate::stream::BufferPool;
 use crate::transport::middle_proxy::{
     MePool, fetch_proxy_config, run_me_ping, MePingFamily, MePingSample, format_sample_line,
 };
-use crate::transport::{ListenOptions, UpstreamManager, create_listener};
+use crate::transport::{ListenOptions, UpstreamManager, create_listener, find_listener_processes};
+use crate::tls_front::TlsFrontCache;
 
 fn parse_cli() -> (String, bool, Option<String>) {
     let mut config_path = "config.toml".to_string();
@@ -92,6 +96,10 @@ fn parse_cli() -> (String, bool, Option<String>) {
                 eprintln!("    --no-start             Don't start the service after install");
                 std::process::exit(0);
             }
+            "--version" | "-V" => {
+                println!("telemt {}", env!("CARGO_PKG_VERSION"));
+                std::process::exit(0);
+            }
             s if !s.starts_with('-') => {
                 config_path = s.to_string();
             }
@@ -106,34 +114,47 @@ fn parse_cli() -> (String, bool, Option<String>) {
 }
 
 fn print_proxy_links(host: &str, port: u16, config: &ProxyConfig) {
-    info!("--- Proxy Links ({}) ---", host);
+    info!(target: "telemt::links", "--- Proxy Links ({}) ---", host);
     for user_name in config.general.links.show.resolve_users(&config.access.users) {
         if let Some(secret) = config.access.users.get(user_name) {
-            info!("User: {}", user_name);
+            info!(target: "telemt::links", "User: {}", user_name);
             if config.general.modes.classic {
                 info!(
+                    target: "telemt::links",
                     "  Classic: tg://proxy?server={}&port={}&secret={}",
                     host, port, secret
                 );
             }
             if config.general.modes.secure {
                 info!(
+                    target: "telemt::links",
                     "  DD:      tg://proxy?server={}&port={}&secret=dd{}",
                     host, port, secret
                 );
             }
             if config.general.modes.tls {
-                let domain_hex = hex::encode(&config.censorship.tls_domain);
-                info!(
-                    "  EE-TLS:  tg://proxy?server={}&port={}&secret=ee{}{}",
-                    host, port, secret, domain_hex
-                );
+                let mut domains = Vec::with_capacity(1 + config.censorship.tls_domains.len());
+                domains.push(config.censorship.tls_domain.clone());
+                for d in &config.censorship.tls_domains {
+                    if !domains.contains(d) {
+                        domains.push(d.clone());
+                    }
+                }
+
+                for domain in domains {
+                    let domain_hex = hex::encode(&domain);
+                    info!(
+                        target: "telemt::links",
+                        "  EE-TLS:  tg://proxy?server={}&port={}&secret=ee{}{}",
+                        host, port, secret, domain_hex
+                    );
+                }
             }
         } else {
-            warn!("User '{}' in show_link not found", user_name);
+            warn!(target: "telemt::links", "User '{}' in show_link not found", user_name);
         }
     }
-    info!("------------------------");
+    info!(target: "telemt::links", "------------------------");
 }
 
 #[tokio::main]
@@ -192,6 +213,9 @@ async fn main() -> std::result::Result<(), Box<dyn std::error::Error>> {
         "Modes: classic={} secure={} tls={}",
         config.general.modes.classic, config.general.modes.secure, config.general.modes.tls
     );
+    if config.general.modes.classic {
+        warn!("Classic mode is vulnerable to DPI detection; enable only for legacy clients");
+    }
     info!("TLS domain: {}", config.censorship.tls_domain);
     if let Some(ref sock) = config.censorship.mask_unix_sock {
         info!("Mask: {} -> unix:{}", config.censorship.mask, sock);
@@ -317,6 +341,7 @@ match crate::transport::middle_proxy::fetch_proxy_secret(proxy_secret_path).awai
                     config.general.middle_proxy_nat_ip,
                     config.general.middle_proxy_nat_probe,
                     config.general.middle_proxy_nat_stun.clone(),
+                    config.general.middle_proxy_nat_stun_servers.clone(),
                     probe.detected_ipv6,
                     config.timeouts.me_one_retry,
                     config.timeouts.me_one_timeout_ms,
@@ -325,18 +350,32 @@ match crate::transport::middle_proxy::fetch_proxy_secret(proxy_secret_path).awai
                     cfg_v4.default_dc.or(cfg_v6.default_dc),
                     decision.clone(),
                     rng.clone(),
+                    stats.clone(),
+                    config.general.me_keepalive_enabled,
+                    config.general.me_keepalive_interval_secs,
+                    config.general.me_keepalive_jitter_secs,
+                    config.general.me_keepalive_payload_random,
+                    config.general.me_warmup_stagger_enabled,
+                    config.general.me_warmup_step_delay_ms,
+                    config.general.me_warmup_step_jitter_ms,
+                    config.general.me_reconnect_max_concurrent_per_dc,
+                    config.general.me_reconnect_backoff_base_ms,
+                    config.general.me_reconnect_backoff_cap_ms,
+                    config.general.me_reconnect_fast_retry_count,
                 );
 
-                match pool.init(2, &rng).await {
+                let pool_size = config.general.middle_proxy_pool_size.max(1);
+                match pool.init(pool_size, &rng).await {
                     Ok(()) => {
                         info!("Middle-End pool initialized successfully");
 
                         // Phase 4: Start health monitor
                         let pool_clone = pool.clone();
                         let rng_clone = rng.clone();
+                        let min_conns = pool_size;
                         tokio::spawn(async move {
                             crate::transport::middle_proxy::me_health_monitor(
-                                pool_clone, rng_clone, 2,
+                                pool_clone, rng_clone, min_conns,
                             )
                             .await;
                         });
@@ -403,6 +442,74 @@ match crate::transport::middle_proxy::fetch_proxy_secret(proxy_secret_path).awai
     let upstream_manager = Arc::new(UpstreamManager::new(config.upstreams.clone()));
     let buffer_pool = Arc::new(BufferPool::with_config(16 * 1024, 4096));
 
+    // TLS front cache (optional emulation)
+    let mut tls_domains = Vec::with_capacity(1 + config.censorship.tls_domains.len());
+    tls_domains.push(config.censorship.tls_domain.clone());
+    for d in &config.censorship.tls_domains {
+        if !tls_domains.contains(d) {
+            tls_domains.push(d.clone());
+        }
+    }
+
+    let tls_cache: Option<Arc<TlsFrontCache>> = if config.censorship.tls_emulation {
+        let cache = Arc::new(TlsFrontCache::new(
+            &tls_domains,
+            config.censorship.fake_cert_len,
+            &config.censorship.tls_front_dir,
+        ));
+
+        cache.load_from_disk().await;
+
+        let port = config.censorship.mask_port;
+        let mask_host = config.censorship.mask_host.clone()
+            .unwrap_or_else(|| config.censorship.tls_domain.clone());
+        // Initial synchronous fetch to warm cache before serving clients.
+        for domain in tls_domains.clone() {
+            match crate::tls_front::fetcher::fetch_real_tls(
+                &mask_host,
+                port,
+                &domain,
+                Duration::from_secs(5),
+                Some(upstream_manager.clone()),
+            )
+            .await
+            {
+                Ok(res) => cache.update_from_fetch(&domain, res).await,
+                Err(e) => warn!(domain = %domain, error = %e, "TLS emulation fetch failed"),
+            }
+        }
+
+        // Periodic refresh with jitter.
+        let cache_clone = cache.clone();
+        let domains = tls_domains.clone();
+        let upstream_for_task = upstream_manager.clone();
+        tokio::spawn(async move {
+            loop {
+                let base_secs = rand::rng().random_range(4 * 3600..=6 * 3600);
+                let jitter_secs = rand::rng().random_range(0..=7200);
+                tokio::time::sleep(Duration::from_secs(base_secs + jitter_secs)).await;
+                for domain in &domains {
+                    match crate::tls_front::fetcher::fetch_real_tls(
+                        &mask_host,
+                        port,
+                        domain,
+                        Duration::from_secs(5),
+                        Some(upstream_for_task.clone()),
+                    )
+                    .await
+                    {
+                        Ok(res) => cache_clone.update_from_fetch(domain, res).await,
+                        Err(e) => warn!(domain = %domain, error = %e, "TLS emulation refresh failed"),
+                    }
+                }
+            }
+        });
+
+        Some(cache)
+    } else {
+        None
+    };
+
     // Middle-End ping before DC connectivity
     if let Some(ref pool) = me_pool {
         let me_results = run_me_ping(pool, &rng).await;
@@ -582,6 +689,19 @@ match crate::transport::middle_proxy::fetch_proxy_secret(proxy_secret_path).awai
         detected_ip_v4, detected_ip_v6
     );
 
+    // ── Hot-reload watcher ────────────────────────────────────────────────
+    // Uses inotify to detect file changes instantly (SIGHUP also works).
+    // detected_ip_v4/v6 are passed so newly added users get correct TG links.
+    let (config_rx, mut log_level_rx): (
+        tokio::sync::watch::Receiver<Arc<ProxyConfig>>,
+        tokio::sync::watch::Receiver<LogLevel>,
+    ) = spawn_config_watcher(
+        std::path::PathBuf::from(&config_path),
+        config.clone(),
+        detected_ip_v4,
+        detected_ip_v6,
+    );
+
     let mut listeners = Vec::new();
 
     for listener_conf in &config.server.listeners {
@@ -595,6 +715,7 @@ match crate::transport::middle_proxy::fetch_proxy_secret(proxy_secret_path).awai
             continue;
         }
         let options = ListenOptions {
+            reuse_port: listener_conf.reuse_allow,
             ipv6_only: listener_conf.ip.is_ipv6(),
             ..Default::default()
         };
@@ -603,6 +724,8 @@ match crate::transport::middle_proxy::fetch_proxy_secret(proxy_secret_path).awai
             Ok(socket) => {
                 let listener = TcpListener::from_std(socket.into())?;
                 info!("Listening on {}", addr);
+                let listener_proxy_protocol =
+                    listener_conf.proxy_protocol.unwrap_or(config.server.proxy_protocol);
 
                 // Resolve the public host for link generation
                 let public_host = if let Some(ref announce) = listener_conf.announce {
@@ -628,10 +751,36 @@ match crate::transport::middle_proxy::fetch_proxy_secret(proxy_secret_path).awai
                     print_proxy_links(&public_host, link_port, &config);
                 }
 
-                listeners.push(listener);
+                listeners.push((listener, listener_proxy_protocol));
             }
             Err(e) => {
-                error!("Failed to bind to {}: {}", addr, e);
+                if e.kind() == std::io::ErrorKind::AddrInUse {
+                    let owners = find_listener_processes(addr);
+                    if owners.is_empty() {
+                        error!(
+                            %addr,
+                            "Failed to bind: address already in use (owner process unresolved)"
+                        );
+                    } else {
+                        for owner in owners {
+                            error!(
+                                %addr,
+                                pid = owner.pid,
+                                process = %owner.process,
+                                "Failed to bind: address already in use"
+                            );
+                        }
+                    }
+
+                    if !listener_conf.reuse_allow {
+                        error!(
+                            %addr,
+                            "reuse_allow=false; set [[server.listeners]].reuse_allow=true to allow multi-instance listening"
+                        );
+                    }
+                } else {
+                    error!("Failed to bind to {}: {}", addr, e);
+                }
             }
         }
     }
@@ -686,13 +835,14 @@ match crate::transport::middle_proxy::fetch_proxy_secret(proxy_secret_path).awai
 
         has_unix_listener = true;
 
-        let config = config.clone();
+        let mut config_rx_unix: tokio::sync::watch::Receiver<Arc<ProxyConfig>> = config_rx.clone();
         let stats = stats.clone();
         let upstream_manager = upstream_manager.clone();
         let replay_checker = replay_checker.clone();
         let buffer_pool = buffer_pool.clone();
         let rng = rng.clone();
         let me_pool = me_pool.clone();
+        let tls_cache = tls_cache.clone();
         let ip_tracker = ip_tracker.clone();
 
         tokio::spawn(async move {
@@ -704,20 +854,22 @@ match crate::transport::middle_proxy::fetch_proxy_secret(proxy_secret_path).awai
                         let conn_id = unix_conn_counter.fetch_add(1, std::sync::atomic::Ordering::Relaxed);
                         let fake_peer = SocketAddr::from(([127, 0, 0, 1], (conn_id % 65535) as u16));
 
-                        let config = config.clone();
+                        let config = config_rx_unix.borrow_and_update().clone();
                         let stats = stats.clone();
                         let upstream_manager = upstream_manager.clone();
                         let replay_checker = replay_checker.clone();
                         let buffer_pool = buffer_pool.clone();
                         let rng = rng.clone();
                         let me_pool = me_pool.clone();
+                        let tls_cache = tls_cache.clone();
                         let ip_tracker = ip_tracker.clone();
+                        let proxy_protocol_enabled = config.server.proxy_protocol;
 
                         tokio::spawn(async move {
                             if let Err(e) = crate::proxy::client::handle_client_stream(
                                 stream, fake_peer, config, stats,
                                 upstream_manager, replay_checker, buffer_pool, rng,
-                                me_pool, ip_tracker,
+                                me_pool, tls_cache, ip_tracker, proxy_protocol_enabled,
                             ).await {
                                 debug!(error = %e, "Unix socket connection error");
                             }
@@ -740,6 +892,8 @@ match crate::transport::middle_proxy::fetch_proxy_secret(proxy_secret_path).awai
     // Switch to user-configured log level after startup
     let runtime_filter = if has_rust_log {
         EnvFilter::from_default_env()
+    } else if matches!(effective_log_level, LogLevel::Silent) {
+        EnvFilter::new("warn,telemt::links=info")
     } else {
         EnvFilter::new(effective_log_level.to_filter_str())
     };
@@ -747,6 +901,20 @@ match crate::transport::middle_proxy::fetch_proxy_secret(proxy_secret_path).awai
         .reload(runtime_filter)
         .expect("Failed to switch log filter");
 
+    // Apply log_level changes from hot-reload to the tracing filter.
+    tokio::spawn(async move {
+        loop {
+            if log_level_rx.changed().await.is_err() {
+                break;
+            }
+            let level = log_level_rx.borrow_and_update().clone();
+            let new_filter = tracing_subscriber::EnvFilter::new(level.to_filter_str());
+            if let Err(e) = filter_handle.reload(new_filter) {
+                tracing::error!("config reload: failed to update log filter: {}", e);
+            }
+        }
+    });
+
     if let Some(port) = config.server.metrics_port {
         let stats = stats.clone();
         let whitelist = config.server.metrics_whitelist.clone();
@@ -755,28 +923,31 @@ match crate::transport::middle_proxy::fetch_proxy_secret(proxy_secret_path).awai
         });
     }
 
-    for listener in listeners {
-        let config = config.clone();
+    for (listener, listener_proxy_protocol) in listeners {
+        let mut config_rx: tokio::sync::watch::Receiver<Arc<ProxyConfig>> = config_rx.clone();
         let stats = stats.clone();
         let upstream_manager = upstream_manager.clone();
         let replay_checker = replay_checker.clone();
         let buffer_pool = buffer_pool.clone();
         let rng = rng.clone();
         let me_pool = me_pool.clone();
+        let tls_cache = tls_cache.clone();
         let ip_tracker = ip_tracker.clone();
 
         tokio::spawn(async move {
             loop {
                 match listener.accept().await {
                     Ok((stream, peer_addr)) => {
-                        let config = config.clone();
+                        let config = config_rx.borrow_and_update().clone();
                         let stats = stats.clone();
                         let upstream_manager = upstream_manager.clone();
                         let replay_checker = replay_checker.clone();
                         let buffer_pool = buffer_pool.clone();
                         let rng = rng.clone();
                         let me_pool = me_pool.clone();
+                        let tls_cache = tls_cache.clone();
                         let ip_tracker = ip_tracker.clone();
+                        let proxy_protocol_enabled = listener_proxy_protocol;
 
                         tokio::spawn(async move {
                             if let Err(e) = ClientHandler::new(
@@ -789,12 +960,47 @@ match crate::transport::middle_proxy::fetch_proxy_secret(proxy_secret_path).awai
                                 buffer_pool,
                                 rng,
                                 me_pool,
+                                tls_cache,
                                 ip_tracker,
+                                proxy_protocol_enabled,
                             )
                             .run()
                             .await
                             {
-                                debug!(peer = %peer_addr, error = %e, "Connection error");
+                                let peer_closed = matches!(
+                                    &e,
+                                    crate::error::ProxyError::Io(ioe)
+                                        if matches!(
+                                            ioe.kind(),
+                                            std::io::ErrorKind::ConnectionReset
+                                                | std::io::ErrorKind::ConnectionAborted
+                                                | std::io::ErrorKind::BrokenPipe
+                                                | std::io::ErrorKind::NotConnected
+                                        )
+                                ) || matches!(
+                                    &e,
+                                    crate::error::ProxyError::Stream(
+                                        crate::error::StreamError::Io(ioe)
+                                    )
+                                        if matches!(
+                                            ioe.kind(),
+                                            std::io::ErrorKind::ConnectionReset
+                                                | std::io::ErrorKind::ConnectionAborted
+                                                | std::io::ErrorKind::BrokenPipe
+                                                | std::io::ErrorKind::NotConnected
+                                        )
+                                );
+
+                                let me_closed = matches!(
+                                    &e,
+                                    crate::error::ProxyError::Proxy(msg) if msg == "ME connection lost"
+                                );
+
+                                match (peer_closed, me_closed) {
+                                    (true, _) => debug!(peer = %peer_addr, error = %e, "Connection closed by client"),
+                                    (_, true) => warn!(peer = %peer_addr, error = %e, "Connection closed: Middle-End dropped session"),
+                                    _ => warn!(peer = %peer_addr, error = %e, "Connection closed with error"),
+                                }
                             }
                         });
                     }

src/metrics.rs

@@ -1,18 +1,19 @@
 use std::convert::Infallible;
-use std::net::{IpAddr, SocketAddr};
+use std::net::SocketAddr;
 use std::sync::Arc;
 
-use http_body_util::Full;
+use http_body_util::{Full, BodyExt};
 use hyper::body::Bytes;
 use hyper::server::conn::http1;
 use hyper::service::service_fn;
 use hyper::{Request, Response, StatusCode};
+use ipnetwork::IpNetwork;
 use tokio::net::TcpListener;
 use tracing::{info, warn, debug};
 
 use crate::stats::Stats;
 
-pub async fn serve(port: u16, stats: Arc<Stats>, whitelist: Vec<IpAddr>) {
+pub async fn serve(port: u16, stats: Arc<Stats>, whitelist: Vec<IpNetwork>) {
     let addr = SocketAddr::from(([0, 0, 0, 0], port));
     let listener = match TcpListener::bind(addr).await {
         Ok(l) => l,
@@ -32,7 +33,7 @@ pub async fn serve(port: u16, stats: Arc<Stats>, whitelist: Vec<IpAddr>) {
             }
         };
 
-        if !whitelist.is_empty() && !whitelist.contains(&peer.ip()) {
+        if !whitelist.is_empty() && !whitelist.iter().any(|net| net.contains(peer.ip())) {
             debug!(peer = %peer, "Metrics request denied by whitelist");
             continue;
         }
@@ -53,7 +54,7 @@ pub async fn serve(port: u16, stats: Arc<Stats>, whitelist: Vec<IpAddr>) {
     }
 }
 
-fn handle(req: Request<hyper::body::Incoming>, stats: &Stats) -> Result<Response<Full<Bytes>>, Infallible> {
+fn handle<B>(req: Request<B>, stats: &Stats) -> Result<Response<Full<Bytes>>, Infallible> {
     if req.uri().path() != "/metrics" {
         let resp = Response::builder()
             .status(StatusCode::NOT_FOUND)
@@ -91,6 +92,54 @@ fn render_metrics(stats: &Stats) -> String {
     let _ = writeln!(out, "# TYPE telemt_handshake_timeouts_total counter");
     let _ = writeln!(out, "telemt_handshake_timeouts_total {}", stats.get_handshake_timeouts());
 
+    let _ = writeln!(out, "# HELP telemt_me_keepalive_sent_total ME keepalive frames sent");
+    let _ = writeln!(out, "# TYPE telemt_me_keepalive_sent_total counter");
+    let _ = writeln!(out, "telemt_me_keepalive_sent_total {}", stats.get_me_keepalive_sent());
+
+    let _ = writeln!(out, "# HELP telemt_me_keepalive_failed_total ME keepalive send failures");
+    let _ = writeln!(out, "# TYPE telemt_me_keepalive_failed_total counter");
+    let _ = writeln!(out, "telemt_me_keepalive_failed_total {}", stats.get_me_keepalive_failed());
+
+    let _ = writeln!(out, "# HELP telemt_me_keepalive_pong_total ME keepalive pong replies");
+    let _ = writeln!(out, "# TYPE telemt_me_keepalive_pong_total counter");
+    let _ = writeln!(out, "telemt_me_keepalive_pong_total {}", stats.get_me_keepalive_pong());
+
+    let _ = writeln!(out, "# HELP telemt_me_keepalive_timeout_total ME keepalive ping timeouts");
+    let _ = writeln!(out, "# TYPE telemt_me_keepalive_timeout_total counter");
+    let _ = writeln!(out, "telemt_me_keepalive_timeout_total {}", stats.get_me_keepalive_timeout());
+
+    let _ = writeln!(out, "# HELP telemt_me_reconnect_attempts_total ME reconnect attempts");
+    let _ = writeln!(out, "# TYPE telemt_me_reconnect_attempts_total counter");
+    let _ = writeln!(out, "telemt_me_reconnect_attempts_total {}", stats.get_me_reconnect_attempts());
+
+    let _ = writeln!(out, "# HELP telemt_me_reconnect_success_total ME reconnect successes");
+    let _ = writeln!(out, "# TYPE telemt_me_reconnect_success_total counter");
+    let _ = writeln!(out, "telemt_me_reconnect_success_total {}", stats.get_me_reconnect_success());
+
+    let _ = writeln!(out, "# HELP telemt_me_crc_mismatch_total ME CRC mismatches");
+    let _ = writeln!(out, "# TYPE telemt_me_crc_mismatch_total counter");
+    let _ = writeln!(out, "telemt_me_crc_mismatch_total {}", stats.get_me_crc_mismatch());
+
+    let _ = writeln!(out, "# HELP telemt_me_seq_mismatch_total ME sequence mismatches");
+    let _ = writeln!(out, "# TYPE telemt_me_seq_mismatch_total counter");
+    let _ = writeln!(out, "telemt_me_seq_mismatch_total {}", stats.get_me_seq_mismatch());
+
+    let _ = writeln!(out, "# HELP telemt_me_route_drop_no_conn_total ME route drops: no conn");
+    let _ = writeln!(out, "# TYPE telemt_me_route_drop_no_conn_total counter");
+    let _ = writeln!(out, "telemt_me_route_drop_no_conn_total {}", stats.get_me_route_drop_no_conn());
+
+    let _ = writeln!(out, "# HELP telemt_me_route_drop_channel_closed_total ME route drops: channel closed");
+    let _ = writeln!(out, "# TYPE telemt_me_route_drop_channel_closed_total counter");
+    let _ = writeln!(out, "telemt_me_route_drop_channel_closed_total {}", stats.get_me_route_drop_channel_closed());
+
+    let _ = writeln!(out, "# HELP telemt_me_route_drop_queue_full_total ME route drops: queue full");
+    let _ = writeln!(out, "# TYPE telemt_me_route_drop_queue_full_total counter");
+    let _ = writeln!(out, "telemt_me_route_drop_queue_full_total {}", stats.get_me_route_drop_queue_full());
+
+    let _ = writeln!(out, "# HELP telemt_secure_padding_invalid_total Invalid secure frame lengths");
+    let _ = writeln!(out, "# TYPE telemt_secure_padding_invalid_total counter");
+    let _ = writeln!(out, "telemt_secure_padding_invalid_total {}", stats.get_secure_padding_invalid());
+
     let _ = writeln!(out, "# HELP telemt_user_connections_total Per-user total connections");
     let _ = writeln!(out, "# TYPE telemt_user_connections_total counter");
     let _ = writeln!(out, "# HELP telemt_user_connections_current Per-user active connections");
@@ -177,21 +226,20 @@ mod tests {
         stats.increment_connects_all();
         stats.increment_connects_all();
 
-        let port = 19091u16;
-        let s = stats.clone();
-        tokio::spawn(async move {
-            serve(port, s, vec![]).await;
-        });
-        tokio::time::sleep(std::time::Duration::from_millis(50)).await;
+        let req = Request::builder()
+            .uri("/metrics")
+            .body(())
+            .unwrap();
+        let resp = handle(req, &stats).unwrap();
+        assert_eq!(resp.status(), StatusCode::OK);
+        let body = resp.into_body().collect().await.unwrap().to_bytes();
+        assert!(std::str::from_utf8(body.as_ref()).unwrap().contains("telemt_connections_total 3"));
 
-        let resp = reqwest::get(format!("http://127.0.0.1:{}/metrics", port))
-            .await.unwrap();
-        assert_eq!(resp.status(), 200);
-        let body = resp.text().await.unwrap();
-        assert!(body.contains("telemt_connections_total 3"));
-
-        let resp404 = reqwest::get(format!("http://127.0.0.1:{}/other", port))
-            .await.unwrap();
-        assert_eq!(resp404.status(), 404);
+        let req404 = Request::builder()
+            .uri("/other")
+            .body(())
+            .unwrap();
+        let resp404 = handle(req404, &stats).unwrap();
+        assert_eq!(resp404.status(), StatusCode::NOT_FOUND);
     }
 }

src/network/stun.rs

@@ -50,10 +50,17 @@ pub async fn stun_probe_family(stun_addr: &str, family: IpFamily) -> Result<Opti
 
     let target_addr = resolve_stun_addr(stun_addr, family).await?;
     if let Some(addr) = target_addr {
-        socket
-            .connect(addr)
-            .await
-            .map_err(|e| ProxyError::Proxy(format!("STUN connect failed: {e}")))?;
+        match socket.connect(addr).await {
+            Ok(()) => {}
+            Err(e) if family == IpFamily::V6 && matches!(
+                e.kind(),
+                std::io::ErrorKind::NetworkUnreachable
+                | std::io::ErrorKind::HostUnreachable
+                | std::io::ErrorKind::Unsupported
+                | std::io::ErrorKind::NetworkDown
+            ) => return Ok(None),
+            Err(e) => return Err(ProxyError::Proxy(format!("STUN connect failed: {e}"))),
+        }
     } else {
         return Ok(None);
     }

src/protocol/constants.rs

@@ -1,6 +1,8 @@
 //! Protocol constants and datacenter addresses
 
 use std::net::{IpAddr, Ipv4Addr, Ipv6Addr};
+
+use crate::crypto::SecureRandom;
 use std::sync::LazyLock;
 
 // ============= Telegram Datacenters =============
@@ -151,7 +153,32 @@ pub const TLS_RECORD_ALERT: u8 = 0x15;
 /// Maximum TLS record size
 pub const MAX_TLS_RECORD_SIZE: usize = 16384;
 /// Maximum TLS chunk size (with overhead)
-pub const MAX_TLS_CHUNK_SIZE: usize = 16384 + 24;
+/// RFC 8446 §5.2 allows up to 16384 + 256 bytes of ciphertext
+pub const MAX_TLS_CHUNK_SIZE: usize = 16384 + 256;
+
+/// Secure Intermediate payload is expected to be 4-byte aligned.
+pub fn is_valid_secure_payload_len(data_len: usize) -> bool {
+    data_len % 4 == 0
+}
+
+/// Compute Secure Intermediate payload length from wire length.
+/// Secure mode strips up to 3 random tail bytes by truncating to 4-byte boundary.
+pub fn secure_payload_len_from_wire_len(wire_len: usize) -> Option<usize> {
+    if wire_len < 4 {
+        return None;
+    }
+    Some(wire_len - (wire_len % 4))
+}
+
+/// Generate padding length for Secure Intermediate protocol.
+/// Data must be 4-byte aligned; padding is 1..=3 so total is never divisible by 4.
+pub fn secure_padding_len(data_len: usize, rng: &SecureRandom) -> usize {
+    debug_assert!(
+        is_valid_secure_payload_len(data_len),
+        "Secure payload must be 4-byte aligned, got {data_len}"
+    );
+    (rng.range(3) + 1) as usize
+}
 
 // ============= Timeouts =============
 
@@ -284,6 +311,10 @@ pub mod rpc_flags {
         pub const FLAG_ABRIDGED: u32      = 0x40000000;
         pub const FLAG_QUICKACK: u32      = 0x80000000;
     }
+
+    pub mod rpc_crypto_flags {
+        pub const USE_CRC32C: u32 = 0x800;
+    }
     
     pub const ME_CONNECT_TIMEOUT_SECS: u64 = 5;
     pub const ME_HANDSHAKE_TIMEOUT_SECS: u64 = 10;
@@ -319,4 +350,43 @@ mod tests {
         assert_eq!(TG_DATACENTERS_V4.len(), 5);
         assert_eq!(TG_DATACENTERS_V6.len(), 5);
     }
-}
+
+    #[test]
+    fn secure_padding_never_produces_aligned_total() {
+        let rng = SecureRandom::new();
+        for data_len in (0..1000).step_by(4) {
+            for _ in 0..100 {
+                let padding = secure_padding_len(data_len, &rng);
+                assert!(
+                    padding <= 3,
+                    "padding out of range: data_len={data_len}, padding={padding}"
+                );
+                assert_ne!(
+                    (data_len + padding) % 4,
+                    0,
+                    "invariant violated: data_len={data_len}, padding={padding}, total={}",
+                    data_len + padding
+                );
+            }
+        }
+    }
+
+    #[test]
+    fn secure_wire_len_roundtrip_for_aligned_payload() {
+        for payload_len in (4..4096).step_by(4) {
+            for padding in 0..=3usize {
+                let wire_len = payload_len + padding;
+                let recovered = secure_payload_len_from_wire_len(wire_len);
+                assert_eq!(recovered, Some(payload_len));
+            }
+        }
+    }
+
+    #[test]
+    fn secure_wire_len_rejects_too_short_frames() {
+        assert_eq!(secure_payload_len_from_wire_len(0), None);
+        assert_eq!(secure_payload_len_from_wire_len(1), None);
+        assert_eq!(secure_payload_len_from_wire_len(2), None);
+        assert_eq!(secure_payload_len_from_wire_len(3), None);
+    }
+}

src/protocol/tls.rs

@@ -32,6 +32,7 @@ pub const TIME_SKEW_MAX: i64 = 10 * 60;  // 10 minutes after
 mod extension_type {
     pub const KEY_SHARE: u16 = 0x0033;
     pub const SUPPORTED_VERSIONS: u16 = 0x002b;
+    pub const ALPN: u16 = 0x0010;
 }
 
 /// TLS Cipher Suites
@@ -62,6 +63,7 @@ pub struct TlsValidation {
 // ============= TLS Extension Builder =============
 
 /// Builder for TLS extensions with correct length calculation
+#[derive(Clone)]
 struct TlsExtensionBuilder {
     extensions: Vec<u8>,
 }
@@ -108,6 +110,27 @@ impl TlsExtensionBuilder {
         
         self
     }
+
+    /// Add ALPN extension with a single selected protocol.
+    fn add_alpn(&mut self, proto: &[u8]) -> &mut Self {
+        // Extension type: ALPN (0x0010)
+        self.extensions.extend_from_slice(&extension_type::ALPN.to_be_bytes());
+
+        // ALPN extension format:
+        // extension_data length (2 bytes)
+        //   protocols length (2 bytes)
+        //     protocol name length (1 byte)
+        //     protocol name bytes
+        let proto_len = proto.len() as u8;
+        let list_len: u16 = 1 + proto_len as u16;
+        let ext_len: u16 = 2 + list_len;
+
+        self.extensions.extend_from_slice(&ext_len.to_be_bytes());
+        self.extensions.extend_from_slice(&list_len.to_be_bytes());
+        self.extensions.push(proto_len);
+        self.extensions.extend_from_slice(proto);
+        self
+    }
     
     /// Build final extensions with length prefix
     fn build(self) -> Vec<u8> {
@@ -144,6 +167,8 @@ struct ServerHelloBuilder {
     compression: u8,
     /// Extensions
     extensions: TlsExtensionBuilder,
+    /// Selected ALPN protocol (if any)
+    alpn: Option<Vec<u8>>,
 }
 
 impl ServerHelloBuilder {
@@ -154,6 +179,7 @@ impl ServerHelloBuilder {
             cipher_suite: cipher_suite::TLS_AES_128_GCM_SHA256,
             compression: 0x00,
             extensions: TlsExtensionBuilder::new(),
+            alpn: None,
         }
     }
     
@@ -167,10 +193,19 @@ impl ServerHelloBuilder {
         self.extensions.add_supported_versions(0x0304);
         self
     }
+
+    fn with_alpn(mut self, proto: Option<Vec<u8>>) -> Self {
+        self.alpn = proto;
+        self
+    }
     
     /// Build ServerHello message (without record header)
     fn build_message(&self) -> Vec<u8> {
-        let extensions = self.extensions.extensions.clone();
+        let mut ext_builder = self.extensions.clone();
+        if let Some(ref alpn) = self.alpn {
+            ext_builder.add_alpn(alpn);
+        }
+        let extensions = ext_builder.extensions.clone();
         let extensions_len = extensions.len() as u16;
         
         // Calculate total length
@@ -350,13 +385,19 @@ pub fn build_server_hello(
     session_id: &[u8],
     fake_cert_len: usize,
     rng: &SecureRandom,
+    alpn: Option<Vec<u8>>,
+    new_session_tickets: u8,
 ) -> Vec<u8> {
+    const MIN_APP_DATA: usize = 64;
+    const MAX_APP_DATA: usize = 16640; // RFC 8446 §5.2 upper bound
+    let fake_cert_len = fake_cert_len.max(MIN_APP_DATA).min(MAX_APP_DATA);
     let x25519_key = gen_fake_x25519_key(rng);
     
     // Build ServerHello
     let server_hello = ServerHelloBuilder::new(session_id.to_vec())
         .with_x25519_key(&x25519_key)
         .with_tls13_version()
+        .with_alpn(alpn)
         .build_record();
     
     // Build Change Cipher Spec record
@@ -373,15 +414,35 @@ pub fn build_server_hello(
     app_data_record.push(TLS_RECORD_APPLICATION);
     app_data_record.extend_from_slice(&TLS_VERSION);
     app_data_record.extend_from_slice(&(fake_cert_len as u16).to_be_bytes());
+    // Fill ApplicationData with fully random bytes of desired length to avoid
+    // deterministic DPI fingerprints (fixed inner content type markers).
     app_data_record.extend_from_slice(&fake_cert);
     
+    // Build optional NewSessionTicket records (TLS 1.3 handshake messages are encrypted;
+    // here we mimic with opaque ApplicationData records of plausible size).
+    let mut tickets = Vec::new();
+    if new_session_tickets > 0 {
+        for _ in 0..new_session_tickets {
+            let ticket_len: usize = rng.range(48) + 48; // 48-95 bytes
+            let mut record = Vec::with_capacity(5 + ticket_len);
+            record.push(TLS_RECORD_APPLICATION);
+            record.extend_from_slice(&TLS_VERSION);
+            record.extend_from_slice(&(ticket_len as u16).to_be_bytes());
+            record.extend_from_slice(&rng.bytes(ticket_len));
+            tickets.push(record);
+        }
+    }
+
     // Combine all records
     let mut response = Vec::with_capacity(
-        server_hello.len() + change_cipher_spec.len() + app_data_record.len()
+        server_hello.len() + change_cipher_spec.len() + app_data_record.len() + tickets.iter().map(|r| r.len()).sum::<usize>()
     );
     response.extend_from_slice(&server_hello);
     response.extend_from_slice(&change_cipher_spec);
     response.extend_from_slice(&app_data_record);
+    for t in &tickets {
+        response.extend_from_slice(t);
+    }
     
     // Compute HMAC for the response
     let mut hmac_input = Vec::with_capacity(TLS_DIGEST_LEN + response.len());
@@ -397,6 +458,131 @@ pub fn build_server_hello(
     response
 }
 
+/// Extract SNI (server_name) from a TLS ClientHello.
+pub fn extract_sni_from_client_hello(handshake: &[u8]) -> Option<String> {
+    if handshake.len() < 43 || handshake[0] != TLS_RECORD_HANDSHAKE {
+        return None;
+    }
+
+    let mut pos = 5; // after record header
+    if handshake.get(pos).copied()? != 0x01 {
+        return None; // not ClientHello
+    }
+
+    // Handshake length bytes
+    pos += 4; // type + len (3)
+
+    // version (2) + random (32)
+    pos += 2 + 32;
+    if pos + 1 > handshake.len() {
+        return None;
+    }
+
+    let session_id_len = *handshake.get(pos)? as usize;
+    pos += 1 + session_id_len;
+    if pos + 2 > handshake.len() {
+        return None;
+    }
+
+    let cipher_suites_len = u16::from_be_bytes([handshake[pos], handshake[pos + 1]]) as usize;
+    pos += 2 + cipher_suites_len;
+    if pos + 1 > handshake.len() {
+        return None;
+    }
+
+    let comp_len = *handshake.get(pos)? as usize;
+    pos += 1 + comp_len;
+    if pos + 2 > handshake.len() {
+        return None;
+    }
+
+    let ext_len = u16::from_be_bytes([handshake[pos], handshake[pos + 1]]) as usize;
+    pos += 2;
+    let ext_end = pos + ext_len;
+    if ext_end > handshake.len() {
+        return None;
+    }
+
+    while pos + 4 <= ext_end {
+        let etype = u16::from_be_bytes([handshake[pos], handshake[pos + 1]]);
+        let elen = u16::from_be_bytes([handshake[pos + 2], handshake[pos + 3]]) as usize;
+        pos += 4;
+        if pos + elen > ext_end {
+            break;
+        }
+        if etype == 0x0000 && elen >= 5 {
+            // server_name extension
+            let list_len = u16::from_be_bytes([handshake[pos], handshake[pos + 1]]) as usize;
+            let mut sn_pos = pos + 2;
+            let sn_end = std::cmp::min(sn_pos + list_len, pos + elen);
+            while sn_pos + 3 <= sn_end {
+                let name_type = handshake[sn_pos];
+                let name_len = u16::from_be_bytes([handshake[sn_pos + 1], handshake[sn_pos + 2]]) as usize;
+                sn_pos += 3;
+                if sn_pos + name_len > sn_end {
+                    break;
+                }
+                if name_type == 0 && name_len > 0 {
+                    if let Ok(host) = std::str::from_utf8(&handshake[sn_pos..sn_pos + name_len]) {
+                        return Some(host.to_string());
+                    }
+                }
+                sn_pos += name_len;
+            }
+        }
+        pos += elen;
+    }
+
+    None
+}
+
+/// Extract ALPN protocol list from ClientHello, return in offered order.
+pub fn extract_alpn_from_client_hello(handshake: &[u8]) -> Vec<Vec<u8>> {
+    let mut pos = 5; // after record header
+    if handshake.get(pos) != Some(&0x01) {
+        return Vec::new();
+    }
+    pos += 4; // type + len
+    pos += 2 + 32; // version + random
+    if pos >= handshake.len() { return Vec::new(); }
+    let session_id_len = *handshake.get(pos).unwrap_or(&0) as usize;
+    pos += 1 + session_id_len;
+    if pos + 2 > handshake.len() { return Vec::new(); }
+    let cipher_len = u16::from_be_bytes([handshake[pos], handshake[pos+1]]) as usize;
+    pos += 2 + cipher_len;
+    if pos >= handshake.len() { return Vec::new(); }
+    let comp_len = *handshake.get(pos).unwrap_or(&0) as usize;
+    pos += 1 + comp_len;
+    if pos + 2 > handshake.len() { return Vec::new(); }
+    let ext_len = u16::from_be_bytes([handshake[pos], handshake[pos+1]]) as usize;
+    pos += 2;
+    let ext_end = pos + ext_len;
+    if ext_end > handshake.len() { return Vec::new(); }
+    let mut out = Vec::new();
+    while pos + 4 <= ext_end {
+        let etype = u16::from_be_bytes([handshake[pos], handshake[pos+1]]);
+        let elen = u16::from_be_bytes([handshake[pos+2], handshake[pos+3]]) as usize;
+        pos += 4;
+        if pos + elen > ext_end { break; }
+        if etype == extension_type::ALPN && elen >= 3 {
+            let list_len = u16::from_be_bytes([handshake[pos], handshake[pos+1]]) as usize;
+            let mut lp = pos + 2;
+            let list_end = (pos + 2).saturating_add(list_len).min(pos + elen);
+            while lp + 1 <= list_end {
+                let plen = handshake[lp] as usize;
+                lp += 1;
+                if lp + plen > list_end { break; }
+                out.push(handshake[lp..lp+plen].to_vec());
+                lp += plen;
+            }
+            break;
+        }
+        pos += elen;
+    }
+    out
+}
+
+
 /// Check if bytes look like a TLS ClientHello
 pub fn is_tls_handshake(first_bytes: &[u8]) -> bool {
     if first_bytes.len() < 3 {
@@ -575,7 +761,7 @@ mod tests {
         let session_id = vec![0xAA; 32];
         
         let rng = SecureRandom::new();
-        let response = build_server_hello(secret, &client_digest, &session_id, 2048, &rng);
+        let response = build_server_hello(secret, &client_digest, &session_id, 2048, &rng, None, 0);
         
         // Should have at least 3 records
         assert!(response.len() > 100);
@@ -608,8 +794,8 @@ mod tests {
         let session_id = vec![0xAA; 32];
         
         let rng = SecureRandom::new();
-        let response1 = build_server_hello(secret, &client_digest, &session_id, 1024, &rng);
-        let response2 = build_server_hello(secret, &client_digest, &session_id, 1024, &rng);
+        let response1 = build_server_hello(secret, &client_digest, &session_id, 1024, &rng, None, 0);
+        let response2 = build_server_hello(secret, &client_digest, &session_id, 1024, &rng, None, 0);
         
         // Digest position should have non-zero data
         let digest1 = &response1[TLS_DIGEST_POS..TLS_DIGEST_POS + TLS_DIGEST_LEN];
@@ -668,4 +854,101 @@ mod tests {
         // Should return None (no match) but not panic
         assert!(result.is_none());
     }
+
+    fn build_client_hello_with_exts(exts: Vec<(u16, Vec<u8>)>, host: &str) -> Vec<u8> {
+        let mut body = Vec::new();
+        body.extend_from_slice(&TLS_VERSION); // legacy version
+        body.extend_from_slice(&[0u8; 32]); // random
+        body.push(0); // session id len
+        body.extend_from_slice(&2u16.to_be_bytes()); // cipher suites len
+        body.extend_from_slice(&[0x13, 0x01]); // TLS_AES_128_GCM_SHA256
+        body.push(1); // compression len
+        body.push(0); // null compression
+
+        // Build SNI extension
+        let host_bytes = host.as_bytes();
+        let mut sni_ext = Vec::new();
+        sni_ext.extend_from_slice(&(host_bytes.len() as u16 + 3).to_be_bytes());
+        sni_ext.push(0);
+        sni_ext.extend_from_slice(&(host_bytes.len() as u16).to_be_bytes());
+        sni_ext.extend_from_slice(host_bytes);
+
+        let mut ext_blob = Vec::new();
+        for (typ, data) in exts {
+            ext_blob.extend_from_slice(&typ.to_be_bytes());
+            ext_blob.extend_from_slice(&(data.len() as u16).to_be_bytes());
+            ext_blob.extend_from_slice(&data);
+        }
+        // SNI last
+        ext_blob.extend_from_slice(&0x0000u16.to_be_bytes());
+        ext_blob.extend_from_slice(&(sni_ext.len() as u16).to_be_bytes());
+        ext_blob.extend_from_slice(&sni_ext);
+
+        body.extend_from_slice(&(ext_blob.len() as u16).to_be_bytes());
+        body.extend_from_slice(&ext_blob);
+
+        let mut handshake = Vec::new();
+        handshake.push(0x01); // ClientHello
+        let len_bytes = (body.len() as u32).to_be_bytes();
+        handshake.extend_from_slice(&len_bytes[1..4]);
+        handshake.extend_from_slice(&body);
+
+        let mut record = Vec::new();
+        record.push(TLS_RECORD_HANDSHAKE);
+        record.extend_from_slice(&[0x03, 0x01]);
+        record.extend_from_slice(&(handshake.len() as u16).to_be_bytes());
+        record.extend_from_slice(&handshake);
+        record
+    }
+
+    #[test]
+    fn test_extract_sni_with_grease_extension() {
+        // GREASE type 0x0a0a with zero length before SNI
+        let ch = build_client_hello_with_exts(vec![(0x0a0a, Vec::new())], "example.com");
+        let sni = extract_sni_from_client_hello(&ch);
+        assert_eq!(sni.as_deref(), Some("example.com"));
+    }
+
+    #[test]
+    fn test_extract_sni_tolerates_empty_unknown_extension() {
+        let ch = build_client_hello_with_exts(vec![(0x1234, Vec::new())], "test.local");
+        let sni = extract_sni_from_client_hello(&ch);
+        assert_eq!(sni.as_deref(), Some("test.local"));
+    }
+
+    #[test]
+    fn test_extract_alpn_single() {
+        let mut alpn_data = Vec::new();
+        // list length = 3 (1 length byte + "h2")
+        alpn_data.extend_from_slice(&3u16.to_be_bytes());
+        alpn_data.push(2);
+        alpn_data.extend_from_slice(b"h2");
+        let ch = build_client_hello_with_exts(vec![(0x0010, alpn_data)], "alpn.test");
+        let alpn = extract_alpn_from_client_hello(&ch);
+        let alpn_str: Vec<String> = alpn
+            .iter()
+            .map(|p| std::str::from_utf8(p).unwrap().to_string())
+            .collect();
+        assert_eq!(alpn_str, vec!["h2"]);
+    }
+
+    #[test]
+    fn test_extract_alpn_multiple() {
+        let mut alpn_data = Vec::new();
+        // list length = 11 (sum of per-proto lengths including length bytes)
+        alpn_data.extend_from_slice(&11u16.to_be_bytes());
+        alpn_data.push(2);
+        alpn_data.extend_from_slice(b"h2");
+        alpn_data.push(4);
+        alpn_data.extend_from_slice(b"spdy");
+        alpn_data.push(2);
+        alpn_data.extend_from_slice(b"h3");
+        let ch = build_client_hello_with_exts(vec![(0x0010, alpn_data)], "alpn.test");
+        let alpn = extract_alpn_from_client_hello(&ch);
+        let alpn_str: Vec<String> = alpn
+            .iter()
+            .map(|p| std::str::from_utf8(p).unwrap().to_string())
+            .collect();
+        assert_eq!(alpn_str, vec!["h2", "spdy", "h3"]);
+    }
 }

src/proxy/client.rs

@@ -30,7 +30,9 @@ use crate::protocol::tls;
 use crate::stats::{ReplayChecker, Stats};
 use crate::stream::{BufferPool, CryptoReader, CryptoWriter};
 use crate::transport::middle_proxy::MePool;
-use crate::transport::{UpstreamManager, configure_client_socket};
+use crate::transport::{UpstreamManager, configure_client_socket, parse_proxy_protocol};
+use crate::transport::socket::normalize_ip;
+use crate::tls_front::TlsFrontCache;
 
 use crate::proxy::direct_relay::handle_via_direct;
 use crate::proxy::handshake::{HandshakeSuccess, handle_mtproto_handshake, handle_tls_handshake};
@@ -47,13 +49,36 @@ pub async fn handle_client_stream<S>(
     buffer_pool: Arc<BufferPool>,
     rng: Arc<SecureRandom>,
     me_pool: Option<Arc<MePool>>,
+    tls_cache: Option<Arc<TlsFrontCache>>,
     ip_tracker: Arc<UserIpTracker>,
+    proxy_protocol_enabled: bool,
 ) -> Result<()>
 where
     S: AsyncRead + AsyncWrite + Unpin + Send + 'static,
 {
     stats.increment_connects_all();
-    debug!(peer = %peer, "New connection (generic stream)");
+    let mut real_peer = normalize_ip(peer);
+
+    if proxy_protocol_enabled {
+        match parse_proxy_protocol(&mut stream, peer).await {
+            Ok(info) => {
+                debug!(
+                    peer = %peer,
+                    client = %info.src_addr,
+                    version = info.version,
+                    "PROXY protocol header parsed"
+                );
+                real_peer = normalize_ip(info.src_addr);
+            }
+            Err(e) => {
+                stats.increment_connects_bad();
+                warn!(peer = %peer, error = %e, "Invalid PROXY protocol header");
+                return Err(e);
+            }
+        }
+    }
+
+    debug!(peer = %real_peer, "New connection (generic stream)");
 
     let handshake_timeout = Duration::from_secs(config.timeouts.client_handshake);
     let stats_for_timeout = stats.clone();
@@ -69,13 +94,13 @@ where
         stream.read_exact(&mut first_bytes).await?;
 
         let is_tls = tls::is_tls_handshake(&first_bytes[..3]);
-        debug!(peer = %peer, is_tls = is_tls, "Handshake type detected");
+        debug!(peer = %real_peer, is_tls = is_tls, "Handshake type detected");
 
         if is_tls {
             let tls_len = u16::from_be_bytes([first_bytes[3], first_bytes[4]]) as usize;
 
             if tls_len < 512 {
-                debug!(peer = %peer, tls_len = tls_len, "TLS handshake too short");
+                debug!(peer = %real_peer, tls_len = tls_len, "TLS handshake too short");
                 stats.increment_connects_bad();
                 let (reader, writer) = tokio::io::split(stream);
                 handle_bad_client(reader, writer, &first_bytes, &config).await;
@@ -89,8 +114,8 @@ where
             let (read_half, write_half) = tokio::io::split(stream);
 
             let (mut tls_reader, tls_writer, _tls_user) = match handle_tls_handshake(
-                &handshake, read_half, write_half, peer,
-                &config, &replay_checker, &rng,
+                &handshake, read_half, write_half, real_peer,
+                &config, &replay_checker, &rng, tls_cache.clone(),
             ).await {
                 HandshakeResult::Success(result) => result,
                 HandshakeResult::BadClient { reader, writer } => {
@@ -107,7 +132,7 @@ where
                 .map_err(|_| ProxyError::InvalidHandshake("Short MTProto handshake".into()))?;
 
             let (crypto_reader, crypto_writer, success) = match handle_mtproto_handshake(
-                &mtproto_handshake, tls_reader, tls_writer, peer,
+                &mtproto_handshake, tls_reader, tls_writer, real_peer,
                 &config, &replay_checker, true,
             ).await {
                 HandshakeResult::Success(result) => result,
@@ -123,12 +148,12 @@ where
                 RunningClientHandler::handle_authenticated_static(
                     crypto_reader, crypto_writer, success,
                     upstream_manager, stats, config, buffer_pool, rng, me_pool,
-                    local_addr, peer, ip_tracker.clone(),
+                    local_addr, real_peer, ip_tracker.clone(),
                 ),
             )))
         } else {
             if !config.general.modes.classic && !config.general.modes.secure {
-                debug!(peer = %peer, "Non-TLS modes disabled");
+                debug!(peer = %real_peer, "Non-TLS modes disabled");
                 stats.increment_connects_bad();
                 let (reader, writer) = tokio::io::split(stream);
                 handle_bad_client(reader, writer, &first_bytes, &config).await;
@@ -142,7 +167,7 @@ where
             let (read_half, write_half) = tokio::io::split(stream);
 
             let (crypto_reader, crypto_writer, success) = match handle_mtproto_handshake(
-                &handshake, read_half, write_half, peer,
+                &handshake, read_half, write_half, real_peer,
                 &config, &replay_checker, false,
             ).await {
                 HandshakeResult::Success(result) => result,
@@ -166,7 +191,7 @@ where
                     rng,
                     me_pool,
                     local_addr,
-                    peer,
+                    real_peer,
                     ip_tracker.clone(),
                 )
             )))
@@ -203,7 +228,9 @@ pub struct RunningClientHandler {
     buffer_pool: Arc<BufferPool>,
     rng: Arc<SecureRandom>,
     me_pool: Option<Arc<MePool>>,
+    tls_cache: Option<Arc<TlsFrontCache>>,
     ip_tracker: Arc<UserIpTracker>,
+    proxy_protocol_enabled: bool,
 }
 
 impl ClientHandler {
@@ -217,7 +244,9 @@ impl ClientHandler {
         buffer_pool: Arc<BufferPool>,
         rng: Arc<SecureRandom>,
         me_pool: Option<Arc<MePool>>,
+        tls_cache: Option<Arc<TlsFrontCache>>,
         ip_tracker: Arc<UserIpTracker>,
+        proxy_protocol_enabled: bool,
     ) -> RunningClientHandler {
         RunningClientHandler {
             stream,
@@ -229,7 +258,9 @@ impl ClientHandler {
             buffer_pool,
             rng,
             me_pool,
+            tls_cache,
             ip_tracker,
+            proxy_protocol_enabled,
         }
     }
 }
@@ -238,6 +269,7 @@ impl RunningClientHandler {
     pub async fn run(mut self) -> Result<()> {
         self.stats.increment_connects_all();
 
+        self.peer = normalize_ip(self.peer);
         let peer = self.peer;
         let ip_tracker = self.ip_tracker.clone();
         debug!(peer = %peer, "New connection");
@@ -275,6 +307,25 @@ impl RunningClientHandler {
     }
 
     async fn do_handshake(mut self) -> Result<HandshakeOutcome> {
+        if self.proxy_protocol_enabled {
+            match parse_proxy_protocol(&mut self.stream, self.peer).await {
+                Ok(info) => {
+                    debug!(
+                        peer = %self.peer,
+                        client = %info.src_addr,
+                        version = info.version,
+                        "PROXY protocol header parsed"
+                    );
+                    self.peer = normalize_ip(info.src_addr);
+                }
+                Err(e) => {
+                    self.stats.increment_connects_bad();
+                    warn!(peer = %self.peer, error = %e, "Invalid PROXY protocol header");
+                    return Err(e);
+                }
+            }
+        }
+
         let mut first_bytes = [0u8; 5];
         self.stream.read_exact(&mut first_bytes).await?;
 
@@ -327,6 +378,7 @@ impl RunningClientHandler {
             &config,
             &replay_checker,
             &self.rng,
+            self.tls_cache.clone(),
         )
         .await
         {

src/proxy/direct_relay.rs

@@ -45,7 +45,7 @@ where
     );
 
     let tg_stream = upstream_manager
-        .connect(dc_addr, Some(success.dc_idx))
+        .connect(dc_addr, Some(success.dc_idx), user.strip_prefix("scope_").filter(|s| !s.is_empty()))
         .await?;
 
     debug!(peer = %success.peer, dc_addr = %dc_addr, "Connected, performing TG handshake");
@@ -178,8 +178,9 @@ async fn do_tg_handshake_static(
 
     let (read_half, write_half) = stream.into_split();
 
+    let max_pending = config.general.crypto_pending_buffer;
     Ok((
         CryptoReader::new(read_half, tg_decryptor),
-        CryptoWriter::new(write_half, tg_encryptor),
+        CryptoWriter::new(write_half, tg_encryptor, max_pending),
     ))
 }

src/proxy/handshake.rs

@@ -1,17 +1,20 @@
 //! MTProto Handshake
 
 use std::net::SocketAddr;
+use std::sync::Arc;
 use tokio::io::{AsyncRead, AsyncWrite, AsyncWriteExt};
 use tracing::{debug, warn, trace, info};
 use zeroize::Zeroize;
 
 use crate::crypto::{sha256, AesCtr, SecureRandom};
+use rand::Rng;
 use crate::protocol::constants::*;
 use crate::protocol::tls;
 use crate::stream::{FakeTlsReader, FakeTlsWriter, CryptoReader, CryptoWriter};
 use crate::error::{ProxyError, HandshakeResult};
 use crate::stats::ReplayChecker;
 use crate::config::ProxyConfig;
+use crate::tls_front::{TlsFrontCache, emulator};
 
 /// Result of successful handshake
 ///
@@ -55,6 +58,7 @@ pub async fn handle_tls_handshake<R, W>(
     config: &ProxyConfig,
     replay_checker: &ReplayChecker,
     rng: &SecureRandom,
+    tls_cache: Option<Arc<TlsFrontCache>>,
 ) -> HandshakeResult<(FakeTlsReader<R>, FakeTlsWriter<W>, String), R, W>
 where
     R: AsyncRead + Unpin,
@@ -102,13 +106,72 @@ where
         None => return HandshakeResult::BadClient { reader, writer },
     };
 
-    let response = tls::build_server_hello(
-        secret,
-        &validation.digest,
-        &validation.session_id,
-        config.censorship.fake_cert_len,
-        rng,
-    );
+    let cached = if config.censorship.tls_emulation {
+        if let Some(cache) = tls_cache.as_ref() {
+            if let Some(sni) = tls::extract_sni_from_client_hello(handshake) {
+                Some(cache.get(&sni).await)
+            } else {
+                Some(cache.get(&config.censorship.tls_domain).await)
+            }
+        } else {
+            None
+        }
+    } else {
+        None
+    };
+
+    let alpn_list = if config.censorship.alpn_enforce {
+        tls::extract_alpn_from_client_hello(handshake)
+    } else {
+        Vec::new()
+    };
+    let selected_alpn = if config.censorship.alpn_enforce {
+        if alpn_list.iter().any(|p| p == b"h2") {
+            Some(b"h2".to_vec())
+        } else if alpn_list.iter().any(|p| p == b"http/1.1") {
+            Some(b"http/1.1".to_vec())
+        } else {
+            None
+        }
+    } else {
+        None
+    };
+
+    let response = if let Some(cached_entry) = cached {
+        emulator::build_emulated_server_hello(
+            secret,
+            &validation.digest,
+            &validation.session_id,
+            &cached_entry,
+            rng,
+            selected_alpn.clone(),
+            config.censorship.tls_new_session_tickets,
+        )
+    } else {
+        tls::build_server_hello(
+            secret,
+            &validation.digest,
+            &validation.session_id,
+            config.censorship.fake_cert_len,
+            rng,
+            selected_alpn.clone(),
+            config.censorship.tls_new_session_tickets,
+        )
+    };
+
+    // Optional anti-fingerprint delay before sending ServerHello.
+    if config.censorship.server_hello_delay_max_ms > 0 {
+        let min = config.censorship.server_hello_delay_min_ms;
+        let max = config.censorship.server_hello_delay_max_ms.max(min);
+        let delay_ms = if max == min {
+            max
+        } else {
+            rand::rng().random_range(min..=max)
+        };
+        if delay_ms > 0 {
+            tokio::time::sleep(std::time::Duration::from_millis(delay_ms)).await;
+        }
+    }
 
     debug!(peer = %peer, response_len = response.len(), "Sending TLS ServerHello");
 
@@ -190,7 +253,11 @@ where
 
         let mode_ok = match proto_tag {
             ProtoTag::Secure => {
-                if is_tls { config.general.modes.tls } else { config.general.modes.secure }
+                if is_tls {
+                    config.general.modes.tls || config.general.modes.secure
+                } else {
+                    config.general.modes.secure || config.general.modes.tls
+                }
             }
             ProtoTag::Intermediate | ProtoTag::Abridged => config.general.modes.classic,
         };
@@ -237,9 +304,10 @@ where
             "MTProto handshake successful"
         );
 
+        let max_pending = config.general.crypto_pending_buffer;
         return HandshakeResult::Success((
             CryptoReader::new(reader, decryptor),
-            CryptoWriter::new(writer, encryptor),
+            CryptoWriter::new(writer, encryptor, max_pending),
             success,
         ));
     }

src/proxy/masking.rs

@@ -1,7 +1,7 @@
 //! Masking - forward unrecognized traffic to mask host
 
-use std::time::Duration;
 use std::str;
+use std::time::Duration;
 use tokio::net::TcpStream;
 #[cfg(unix)]
 use tokio::net::UnixStream;
@@ -11,9 +11,9 @@ use tracing::debug;
 use crate::config::ProxyConfig;
 
 const MASK_TIMEOUT: Duration = Duration::from_secs(5);
-    /// Maximum duration for the entire masking relay.
-    /// Limits resource consumption from slow-loris attacks and port scanners.
-    const MASK_RELAY_TIMEOUT: Duration = Duration::from_secs(60);
+/// Maximum duration for the entire masking relay.
+/// Limits resource consumption from slow-loris attacks and port scanners.
+const MASK_RELAY_TIMEOUT: Duration = Duration::from_secs(60);
 const MASK_BUFFER_SIZE: usize = 8192;
 
 /// Detect client type based on initial data
@@ -78,7 +78,9 @@ where
         match connect_result {
             Ok(Ok(stream)) => {
                 let (mask_read, mask_write) = stream.into_split();
-                relay_to_mask(reader, writer, mask_read, mask_write, initial_data).await;
+                if timeout(MASK_RELAY_TIMEOUT, relay_to_mask(reader, writer, mask_read, mask_write, initial_data)).await.is_err() {
+                    debug!("Mask relay timed out (unix socket)");
+                }
             }
             Ok(Err(e)) => {
                 debug!(error = %e, "Failed to connect to mask unix socket");
@@ -110,7 +112,9 @@ where
     match connect_result {
         Ok(Ok(stream)) => {
             let (mask_read, mask_write) = stream.into_split();
-            relay_to_mask(reader, writer, mask_read, mask_write, initial_data).await;
+            if timeout(MASK_RELAY_TIMEOUT, relay_to_mask(reader, writer, mask_read, mask_write, initial_data)).await.is_err() {
+                debug!("Mask relay timed out");
+            }
         }
         Ok(Err(e)) => {
             debug!(error = %e, "Failed to connect to mask host");

src/proxy/middle_relay.rs

@@ -2,24 +2,30 @@ use std::net::SocketAddr;
 use std::sync::Arc;
 
 use tokio::io::{AsyncRead, AsyncReadExt, AsyncWrite, AsyncWriteExt};
-use tracing::{debug, info, trace};
+use tokio::sync::{mpsc, oneshot};
+use tracing::{debug, info, trace, warn};
 
 use crate::config::ProxyConfig;
 use crate::crypto::SecureRandom;
 use crate::error::{ProxyError, Result};
-use crate::protocol::constants::*;
+use crate::protocol::constants::{*, secure_padding_len};
 use crate::proxy::handshake::HandshakeSuccess;
 use crate::stats::Stats;
 use crate::stream::{BufferPool, CryptoReader, CryptoWriter};
 use crate::transport::middle_proxy::{MePool, MeResponse, proto_flags_for_tag};
 
+enum C2MeCommand {
+    Data { payload: Vec<u8>, flags: u32 },
+    Close,
+}
+
 pub(crate) async fn handle_via_middle_proxy<R, W>(
     mut crypto_reader: CryptoReader<R>,
-    mut crypto_writer: CryptoWriter<W>,
+    crypto_writer: CryptoWriter<W>,
     success: HandshakeSuccess,
     me_pool: Arc<MePool>,
     stats: Arc<Stats>,
-    _config: Arc<ProxyConfig>,
+    config: Arc<ProxyConfig>,
     _buffer_pool: Arc<BufferPool>,
     local_addr: SocketAddr,
     rng: Arc<SecureRandom>,
@@ -41,7 +47,7 @@ where
         "Routing via Middle-End"
     );
 
-    let (conn_id, mut me_rx) = me_pool.registry().register().await;
+    let (conn_id, me_rx) = me_pool.registry().register().await;
 
     stats.increment_user_connects(&user);
     stats.increment_user_curr_connects(&user);
@@ -56,59 +62,169 @@ where
 
     let translated_local_addr = me_pool.translate_our_addr(local_addr);
 
-    let result: Result<()> = loop {
-        tokio::select! {
-            client_frame = read_client_payload(&mut crypto_reader, proto_tag) => {
-                match client_frame {
-                    Ok(Some((payload, quickack))) => {
-                        trace!(conn_id, bytes = payload.len(), "C->ME frame");
-                        stats.add_user_octets_from(&user, payload.len() as u64);
-                        let mut flags = proto_flags;
-                        if quickack {
-                            flags |= RPC_FLAG_QUICKACK;
-                        }
-                        if payload.len() >= 8 && payload[..8].iter().all(|b| *b == 0) {
-                            flags |= RPC_FLAG_NOT_ENCRYPTED;
-                        }
-                        me_pool.send_proxy_req(
-                            conn_id,
-                            success.dc_idx,
-                            peer,
-                            translated_local_addr,
-                            &payload,
-                            flags,
-                        ).await?;
-                    }
-                    Ok(None) => {
-                        debug!(conn_id, "Client EOF");
-                        let _ = me_pool.send_close(conn_id).await;
-                        break Ok(());
-                    }
-                    Err(e) => break Err(e),
+    let frame_limit = config.general.max_client_frame;
+
+    let (c2me_tx, mut c2me_rx) = mpsc::channel::<C2MeCommand>(1024);
+    let me_pool_c2me = me_pool.clone();
+    let c2me_sender = tokio::spawn(async move {
+        while let Some(cmd) = c2me_rx.recv().await {
+            match cmd {
+                C2MeCommand::Data { payload, flags } => {
+                    me_pool_c2me.send_proxy_req(
+                        conn_id,
+                        success.dc_idx,
+                        peer,
+                        translated_local_addr,
+                        &payload,
+                        flags,
+                    ).await?;
                 }
-            }
-            me_msg = me_rx.recv() => {
-                match me_msg {
-                    Some(MeResponse::Data { flags, data }) => {
-                        trace!(conn_id, bytes = data.len(), flags, "ME->C data");
-                        stats.add_user_octets_to(&user, data.len() as u64);
-                        write_client_payload(&mut crypto_writer, proto_tag, flags, &data, rng.as_ref()).await?;
-                    }
-                    Some(MeResponse::Ack(confirm)) => {
-                        trace!(conn_id, confirm, "ME->C quickack");
-                        write_client_ack(&mut crypto_writer, proto_tag, confirm).await?;
-                    }
-                    Some(MeResponse::Close) => {
-                        debug!(conn_id, "ME sent close");
-                        break Ok(());
-                    }
-                    None => {
-                        debug!(conn_id, "ME channel closed");
-                        break Err(ProxyError::Proxy("ME connection lost".into()));
-                    }
+                C2MeCommand::Close => {
+                    let _ = me_pool_c2me.send_close(conn_id).await;
+                    return Ok(());
                 }
             }
         }
+        Ok(())
+    });
+
+    let (stop_tx, mut stop_rx) = oneshot::channel::<()>();
+    let mut me_rx_task = me_rx;
+    let stats_clone = stats.clone();
+    let rng_clone = rng.clone();
+    let user_clone = user.clone();
+    let me_writer = tokio::spawn(async move {
+        let mut writer = crypto_writer;
+        loop {
+            tokio::select! {
+                msg = me_rx_task.recv() => {
+                    match msg {
+                        Some(MeResponse::Data { flags, data }) => {
+                            trace!(conn_id, bytes = data.len(), flags, "ME->C data");
+                            stats_clone.add_user_octets_to(&user_clone, data.len() as u64);
+                            write_client_payload(&mut writer, proto_tag, flags, &data, rng_clone.as_ref()).await?;
+
+                            // Drain all immediately queued ME responses and flush once.
+                            while let Ok(next) = me_rx_task.try_recv() {
+                                match next {
+                                    MeResponse::Data { flags, data } => {
+                                        trace!(conn_id, bytes = data.len(), flags, "ME->C data (batched)");
+                                        stats_clone.add_user_octets_to(&user_clone, data.len() as u64);
+                                        write_client_payload(
+                                            &mut writer,
+                                            proto_tag,
+                                            flags,
+                                            &data,
+                                            rng_clone.as_ref(),
+                                        ).await?;
+                                    }
+                                    MeResponse::Ack(confirm) => {
+                                        trace!(conn_id, confirm, "ME->C quickack (batched)");
+                                        write_client_ack(&mut writer, proto_tag, confirm).await?;
+                                    }
+                                    MeResponse::Close => {
+                                        debug!(conn_id, "ME sent close (batched)");
+                                        let _ = writer.flush().await;
+                                        return Ok(());
+                                    }
+                                }
+                            }
+
+                            writer.flush().await.map_err(ProxyError::Io)?;
+                        }
+                        Some(MeResponse::Ack(confirm)) => {
+                            trace!(conn_id, confirm, "ME->C quickack");
+                            write_client_ack(&mut writer, proto_tag, confirm).await?;
+                        }
+                        Some(MeResponse::Close) => {
+                            debug!(conn_id, "ME sent close");
+                            let _ = writer.flush().await;
+                            return Ok(());
+                        }
+                        None => {
+                            debug!(conn_id, "ME channel closed");
+                            return Err(ProxyError::Proxy("ME connection lost".into()));
+                        }
+                    }
+                }
+                _ = &mut stop_rx => {
+                    debug!(conn_id, "ME writer stop signal");
+                    return Ok(());
+                }
+            }
+        }
+    });
+
+    let mut main_result: Result<()> = Ok(());
+    let mut client_closed = false;
+    let mut frame_counter: u64 = 0;
+    loop {
+        match read_client_payload(
+            &mut crypto_reader,
+            proto_tag,
+            frame_limit,
+            &user,
+            &mut frame_counter,
+            &stats,
+        ).await {
+            Ok(Some((payload, quickack))) => {
+                trace!(conn_id, bytes = payload.len(), "C->ME frame");
+                stats.add_user_octets_from(&user, payload.len() as u64);
+                let mut flags = proto_flags;
+                if quickack {
+                    flags |= RPC_FLAG_QUICKACK;
+                }
+                if payload.len() >= 8 && payload[..8].iter().all(|b| *b == 0) {
+                    flags |= RPC_FLAG_NOT_ENCRYPTED;
+                }
+                // Keep client read loop lightweight: route heavy ME send path via a dedicated task.
+                if c2me_tx
+                    .send(C2MeCommand::Data { payload, flags })
+                    .await
+                    .is_err()
+                {
+                    main_result = Err(ProxyError::Proxy("ME sender channel closed".into()));
+                    break;
+                }
+            }
+            Ok(None) => {
+                debug!(conn_id, "Client EOF");
+                client_closed = true;
+                let _ = c2me_tx.send(C2MeCommand::Close).await;
+                break;
+            }
+            Err(e) => {
+                main_result = Err(e);
+                break;
+            }
+        }
+    }
+
+    drop(c2me_tx);
+    let c2me_result = c2me_sender
+        .await
+        .unwrap_or_else(|e| Err(ProxyError::Proxy(format!("ME sender join error: {e}"))));
+
+    let _ = stop_tx.send(());
+    let mut writer_result = me_writer
+        .await
+        .unwrap_or_else(|e| Err(ProxyError::Proxy(format!("ME writer join error: {e}"))));
+
+    // When client closes, but ME channel stopped as unregistered - it isnt error
+    if client_closed {
+        if matches!(
+            writer_result,
+            Err(ProxyError::Proxy(ref msg)) if msg == "ME connection lost"
+        ) {
+            writer_result = Ok(());
+        }
+    }
+
+    let result = match (main_result, c2me_result, writer_result) {
+        (Ok(()), Ok(()), Ok(())) => Ok(()),
+        (Err(e), _, _) => Err(e),
+        (_, Err(e), _) => Err(e),
+        (_, _, Err(e)) => Err(e),
     };
 
     debug!(user = %user, conn_id, "ME relay cleanup");
@@ -120,66 +236,125 @@ where
 async fn read_client_payload<R>(
     client_reader: &mut CryptoReader<R>,
     proto_tag: ProtoTag,
+    max_frame: usize,
+    user: &str,
+    frame_counter: &mut u64,
+    stats: &Stats,
 ) -> Result<Option<(Vec<u8>, bool)>>
 where
     R: AsyncRead + Unpin + Send + 'static,
 {
-    let (len, quickack) = match proto_tag {
-        ProtoTag::Abridged => {
-            let mut first = [0u8; 1];
-            match client_reader.read_exact(&mut first).await {
-                Ok(_) => {}
-                Err(e) if e.kind() == std::io::ErrorKind::UnexpectedEof => return Ok(None),
-                Err(e) => return Err(ProxyError::Io(e)),
+    loop {
+        let (len, quickack, raw_len_bytes) = match proto_tag {
+            ProtoTag::Abridged => {
+                let mut first = [0u8; 1];
+                match client_reader.read_exact(&mut first).await {
+                    Ok(_) => {}
+                    Err(e) if e.kind() == std::io::ErrorKind::UnexpectedEof => return Ok(None),
+                    Err(e) => return Err(ProxyError::Io(e)),
+                }
+
+                let quickack = (first[0] & 0x80) != 0;
+                let len_words = if (first[0] & 0x7f) == 0x7f {
+                    let mut ext = [0u8; 3];
+                    client_reader
+                        .read_exact(&mut ext)
+                        .await
+                        .map_err(ProxyError::Io)?;
+                    u32::from_le_bytes([ext[0], ext[1], ext[2], 0]) as usize
+                } else {
+                    (first[0] & 0x7f) as usize
+                };
+
+                let len = len_words
+                    .checked_mul(4)
+                    .ok_or_else(|| ProxyError::Proxy("Abridged frame length overflow".into()))?;
+                (len, quickack, None)
             }
-
-            let quickack = (first[0] & 0x80) != 0;
-            let len_words = if (first[0] & 0x7f) == 0x7f {
-                let mut ext = [0u8; 3];
-                client_reader
-                    .read_exact(&mut ext)
-                    .await
-                    .map_err(ProxyError::Io)?;
-                u32::from_le_bytes([ext[0], ext[1], ext[2], 0]) as usize
-            } else {
-                (first[0] & 0x7f) as usize
-            };
-
-            let len = len_words
-                .checked_mul(4)
-                .ok_or_else(|| ProxyError::Proxy("Abridged frame length overflow".into()))?;
-            (len, quickack)
-        }
-        ProtoTag::Intermediate | ProtoTag::Secure => {
-            let mut len_buf = [0u8; 4];
-            match client_reader.read_exact(&mut len_buf).await {
-                Ok(_) => {}
-                Err(e) if e.kind() == std::io::ErrorKind::UnexpectedEof => return Ok(None),
-                Err(e) => return Err(ProxyError::Io(e)),
+            ProtoTag::Intermediate | ProtoTag::Secure => {
+                let mut len_buf = [0u8; 4];
+                match client_reader.read_exact(&mut len_buf).await {
+                    Ok(_) => {}
+                    Err(e) if e.kind() == std::io::ErrorKind::UnexpectedEof => return Ok(None),
+                    Err(e) => return Err(ProxyError::Io(e)),
+                }
+                let quickack = (len_buf[3] & 0x80) != 0;
+                (
+                    (u32::from_le_bytes(len_buf) & 0x7fff_ffff) as usize,
+                    quickack,
+                    Some(len_buf),
+                )
             }
-            let quickack = (len_buf[3] & 0x80) != 0;
-            ((u32::from_le_bytes(len_buf) & 0x7fff_ffff) as usize, quickack)
+        };
+
+        if len == 0 {
+            continue;
         }
-    };
-
-    if len > 16 * 1024 * 1024 {
-        return Err(ProxyError::Proxy(format!("Frame too large: {len}")));
-    }
-
-    let mut payload = vec![0u8; len];
-    client_reader
-        .read_exact(&mut payload)
-        .await
-        .map_err(ProxyError::Io)?;
-
-    // Secure Intermediate: remove random padding (last len%4 bytes)
-    if proto_tag == ProtoTag::Secure {
-        let rem = len % 4;
-        if rem != 0 && payload.len() >= rem {
-            payload.truncate(len - rem);
+        if len < 4 && proto_tag != ProtoTag::Abridged {
+            warn!(
+                user = %user,
+                len,
+                proto = ?proto_tag,
+                "Frame too small — corrupt or probe"
+            );
+            return Err(ProxyError::Proxy(format!("Frame too small: {len}")));
         }
+
+        if len > max_frame {
+            let len_buf = raw_len_bytes.unwrap_or((len as u32).to_le_bytes());
+            let looks_like_tls = raw_len_bytes
+                .map(|b| b[0] == 0x16 && b[1] == 0x03)
+                .unwrap_or(false);
+            let looks_like_http = raw_len_bytes
+                .map(|b| matches!(b[0], b'G' | b'P' | b'H' | b'C' | b'D'))
+                .unwrap_or(false);
+            warn!(
+                user = %user,
+                raw_len = len,
+                raw_len_hex = format_args!("0x{:08x}", len),
+                raw_bytes = format_args!(
+                    "{:02x} {:02x} {:02x} {:02x}",
+                    len_buf[0], len_buf[1], len_buf[2], len_buf[3]
+                ),
+                proto = ?proto_tag,
+                tls_like = looks_like_tls,
+                http_like = looks_like_http,
+                frames_ok = *frame_counter,
+                "Frame too large — crypto desync forensics"
+            );
+            return Err(ProxyError::Proxy(format!(
+                "Frame too large: {len} (max {max_frame}), frames_ok={}",
+                *frame_counter
+            )));
+        }
+
+        let secure_payload_len = if proto_tag == ProtoTag::Secure {
+            match secure_payload_len_from_wire_len(len) {
+                Some(payload_len) => payload_len,
+                None => {
+                    stats.increment_secure_padding_invalid();
+                    return Err(ProxyError::Proxy(format!(
+                        "Invalid secure frame length: {len}"
+                    )));
+                }
+            }
+        } else {
+            len
+        };
+
+        let mut payload = vec![0u8; len];
+        client_reader
+            .read_exact(&mut payload)
+            .await
+            .map_err(ProxyError::Io)?;
+
+        // Secure Intermediate: strip validated trailing padding bytes.
+        if proto_tag == ProtoTag::Secure {
+            payload.truncate(secure_payload_len);
+        }
+        *frame_counter += 1;
+        return Ok(Some((payload, quickack)));
     }
-    Ok(Some((payload, quickack)))
 }
 
 async fn write_client_payload<W>(
@@ -209,8 +384,11 @@ where
                 if quickack {
                     first |= 0x80;
                 }
+                let mut frame_buf = Vec::with_capacity(1 + data.len());
+                frame_buf.push(first);
+                frame_buf.extend_from_slice(data);
                 client_writer
-                    .write_all(&[first])
+                    .write_all(&frame_buf)
                     .await
                     .map_err(ProxyError::Io)?;
             } else if len_words < (1 << 24) {
@@ -219,8 +397,11 @@ where
                     first |= 0x80;
                 }
                 let lw = (len_words as u32).to_le_bytes();
+                let mut frame_buf = Vec::with_capacity(4 + data.len());
+                frame_buf.extend_from_slice(&[first, lw[0], lw[1], lw[2]]);
+                frame_buf.extend_from_slice(data);
                 client_writer
-                    .write_all(&[first, lw[0], lw[1], lw[2]])
+                    .write_all(&frame_buf)
                     .await
                     .map_err(ProxyError::Io)?;
             } else {
@@ -229,47 +410,37 @@ where
                     data.len()
                 )));
             }
-
-            client_writer
-                .write_all(data)
-                .await
-                .map_err(ProxyError::Io)?;
         }
         ProtoTag::Intermediate | ProtoTag::Secure => {
             let padding_len = if proto_tag == ProtoTag::Secure {
-                (rng.bytes(1)[0] % 4) as usize
+                if !is_valid_secure_payload_len(data.len()) {
+                    return Err(ProxyError::Proxy(format!(
+                        "Secure payload must be 4-byte aligned, got {}",
+                        data.len()
+                    )));
+                }
+                secure_padding_len(data.len(), rng)
             } else {
                 0
             };
-            let mut len = (data.len() + padding_len) as u32;
+            let mut len_val = (data.len() + padding_len) as u32;
             if quickack {
-                len |= 0x8000_0000;
+                len_val |= 0x8000_0000;
             }
-            client_writer
-                .write_all(&len.to_le_bytes())
-                .await
-                .map_err(ProxyError::Io)?;
-            client_writer
-                .write_all(data)
-                .await
-                .map_err(ProxyError::Io)?;
+            let total = 4 + data.len() + padding_len;
+            let mut frame_buf = Vec::with_capacity(total);
+            frame_buf.extend_from_slice(&len_val.to_le_bytes());
+            frame_buf.extend_from_slice(data);
             if padding_len > 0 {
-                let pad = rng.bytes(padding_len);
-                client_writer
-                    .write_all(&pad)
-                    .await
-                    .map_err(ProxyError::Io)?;
+                frame_buf.extend_from_slice(&rng.bytes(padding_len));
             }
+            client_writer
+                .write_all(&frame_buf)
+                .await
+                .map_err(ProxyError::Io)?;
         }
     }
 
-    // Avoid unconditional per-frame flush (throughput killer on large downloads).
-    // Flush only when low-latency ack semantics are requested or when
-    // CryptoWriter has buffered pending ciphertext that must be drained.
-    if quickack || client_writer.has_pending() {
-        client_writer.flush().await.map_err(ProxyError::Io)?;
-    }
-
     Ok(())
 }
 

src/stats/mod.rs

@@ -19,6 +19,18 @@ pub struct Stats {
     connects_all: AtomicU64,
     connects_bad: AtomicU64,
     handshake_timeouts: AtomicU64,
+    me_keepalive_sent: AtomicU64,
+    me_keepalive_failed: AtomicU64,
+    me_keepalive_pong: AtomicU64,
+    me_keepalive_timeout: AtomicU64,
+    me_reconnect_attempts: AtomicU64,
+    me_reconnect_success: AtomicU64,
+    me_crc_mismatch: AtomicU64,
+    me_seq_mismatch: AtomicU64,
+    me_route_drop_no_conn: AtomicU64,
+    me_route_drop_channel_closed: AtomicU64,
+    me_route_drop_queue_full: AtomicU64,
+    secure_padding_invalid: AtomicU64,
     user_stats: DashMap<String, UserStats>,
     start_time: parking_lot::RwLock<Option<Instant>>,
 }
@@ -43,8 +55,47 @@ impl Stats {
     pub fn increment_connects_all(&self) { self.connects_all.fetch_add(1, Ordering::Relaxed); }
     pub fn increment_connects_bad(&self) { self.connects_bad.fetch_add(1, Ordering::Relaxed); }
     pub fn increment_handshake_timeouts(&self) { self.handshake_timeouts.fetch_add(1, Ordering::Relaxed); }
+    pub fn increment_me_keepalive_sent(&self) { self.me_keepalive_sent.fetch_add(1, Ordering::Relaxed); }
+    pub fn increment_me_keepalive_failed(&self) { self.me_keepalive_failed.fetch_add(1, Ordering::Relaxed); }
+    pub fn increment_me_keepalive_pong(&self) { self.me_keepalive_pong.fetch_add(1, Ordering::Relaxed); }
+    pub fn increment_me_keepalive_timeout(&self) { self.me_keepalive_timeout.fetch_add(1, Ordering::Relaxed); }
+    pub fn increment_me_keepalive_timeout_by(&self, value: u64) {
+        self.me_keepalive_timeout.fetch_add(value, Ordering::Relaxed);
+    }
+    pub fn increment_me_reconnect_attempt(&self) { self.me_reconnect_attempts.fetch_add(1, Ordering::Relaxed); }
+    pub fn increment_me_reconnect_success(&self) { self.me_reconnect_success.fetch_add(1, Ordering::Relaxed); }
+    pub fn increment_me_crc_mismatch(&self) { self.me_crc_mismatch.fetch_add(1, Ordering::Relaxed); }
+    pub fn increment_me_seq_mismatch(&self) { self.me_seq_mismatch.fetch_add(1, Ordering::Relaxed); }
+    pub fn increment_me_route_drop_no_conn(&self) { self.me_route_drop_no_conn.fetch_add(1, Ordering::Relaxed); }
+    pub fn increment_me_route_drop_channel_closed(&self) {
+        self.me_route_drop_channel_closed.fetch_add(1, Ordering::Relaxed);
+    }
+    pub fn increment_me_route_drop_queue_full(&self) {
+        self.me_route_drop_queue_full.fetch_add(1, Ordering::Relaxed);
+    }
+    pub fn increment_secure_padding_invalid(&self) {
+        self.secure_padding_invalid.fetch_add(1, Ordering::Relaxed);
+    }
     pub fn get_connects_all(&self) -> u64 { self.connects_all.load(Ordering::Relaxed) }
     pub fn get_connects_bad(&self) -> u64 { self.connects_bad.load(Ordering::Relaxed) }
+    pub fn get_me_keepalive_sent(&self) -> u64 { self.me_keepalive_sent.load(Ordering::Relaxed) }
+    pub fn get_me_keepalive_failed(&self) -> u64 { self.me_keepalive_failed.load(Ordering::Relaxed) }
+    pub fn get_me_keepalive_pong(&self) -> u64 { self.me_keepalive_pong.load(Ordering::Relaxed) }
+    pub fn get_me_keepalive_timeout(&self) -> u64 { self.me_keepalive_timeout.load(Ordering::Relaxed) }
+    pub fn get_me_reconnect_attempts(&self) -> u64 { self.me_reconnect_attempts.load(Ordering::Relaxed) }
+    pub fn get_me_reconnect_success(&self) -> u64 { self.me_reconnect_success.load(Ordering::Relaxed) }
+    pub fn get_me_crc_mismatch(&self) -> u64 { self.me_crc_mismatch.load(Ordering::Relaxed) }
+    pub fn get_me_seq_mismatch(&self) -> u64 { self.me_seq_mismatch.load(Ordering::Relaxed) }
+    pub fn get_me_route_drop_no_conn(&self) -> u64 { self.me_route_drop_no_conn.load(Ordering::Relaxed) }
+    pub fn get_me_route_drop_channel_closed(&self) -> u64 {
+        self.me_route_drop_channel_closed.load(Ordering::Relaxed)
+    }
+    pub fn get_me_route_drop_queue_full(&self) -> u64 {
+        self.me_route_drop_queue_full.load(Ordering::Relaxed)
+    }
+    pub fn get_secure_padding_invalid(&self) -> u64 {
+        self.secure_padding_invalid.load(Ordering::Relaxed)
+    }
     
     pub fn increment_user_connects(&self, user: &str) {
         self.user_stats.entry(user.to_string()).or_default()

src/stream/crypto_stream.rs

@@ -34,7 +34,7 @@
 //! └────────────────────────────────────────┘
 //!
 //! Backpressure
-//! - pending ciphertext buffer is bounded (MAX_PENDING_WRITE)
+//! - pending ciphertext buffer is bounded (configurable per connection)
 //! - pending is full and upstream is pending 
 //!   -> poll_write returns Poll::Pending
 //!   -> do not accept any plaintext
@@ -62,10 +62,9 @@ use super::state::{StreamState, YieldBuffer};
 
 // ============= Constants =============
 
-/// Maximum size for pending ciphertext buffer (bounded backpressure).
-/// Reduced to 64KB to prevent bufferbloat on mobile networks.
-/// 512KB was causing high latency on 3G/LTE connections.
-const MAX_PENDING_WRITE: usize = 64 * 1024;
+/// Default size for pending ciphertext buffer (bounded backpressure).
+/// Actual limit is supplied at runtime from configuration.
+const DEFAULT_MAX_PENDING_WRITE: usize = 64 * 1024;
 
 /// Default read buffer capacity (reader mostly decrypts in-place into caller buffer).
 const DEFAULT_READ_CAPACITY: usize = 16 * 1024;
@@ -427,15 +426,22 @@ pub struct CryptoWriter<W> {
     encryptor: AesCtr,
     state: CryptoWriterState,
     scratch: BytesMut,
+    max_pending_write: usize,
 }
 
 impl<W> CryptoWriter<W> {
-    pub fn new(upstream: W, encryptor: AesCtr) -> Self {
+    pub fn new(upstream: W, encryptor: AesCtr, max_pending_write: usize) -> Self {
+        let max_pending = if max_pending_write == 0 {
+            DEFAULT_MAX_PENDING_WRITE
+        } else {
+            max_pending_write
+        };
         Self {
             upstream,
             encryptor,
             state: CryptoWriterState::Idle,
             scratch: BytesMut::with_capacity(16 * 1024),
+            max_pending_write: max_pending.max(4 * 1024),
         }
     }
 
@@ -484,10 +490,10 @@ impl<W> CryptoWriter<W> {
     }
 
     /// Ensure we are in Flushing state and return mutable pending buffer.
-    fn ensure_pending<'a>(state: &'a mut CryptoWriterState) -> &'a mut PendingCiphertext {
+    fn ensure_pending<'a>(state: &'a mut CryptoWriterState, max_pending: usize) -> &'a mut PendingCiphertext {
         if matches!(state, CryptoWriterState::Idle) {
             *state = CryptoWriterState::Flushing {
-                pending: PendingCiphertext::new(MAX_PENDING_WRITE),
+                pending: PendingCiphertext::new(max_pending),
             };
         }
 
@@ -498,14 +504,14 @@ impl<W> CryptoWriter<W> {
     }
 
     /// Select how many plaintext bytes can be accepted in buffering path
-    fn select_to_accept_for_buffering(state: &CryptoWriterState, buf_len: usize) -> usize {
+    fn select_to_accept_for_buffering(state: &CryptoWriterState, buf_len: usize, max_pending: usize) -> usize {
         if buf_len == 0 {
             return 0;
         }
 
         match state {
             CryptoWriterState::Flushing { pending } => buf_len.min(pending.remaining_capacity()),
-            CryptoWriterState::Idle => buf_len.min(MAX_PENDING_WRITE),
+            CryptoWriterState::Idle => buf_len.min(max_pending),
             CryptoWriterState::Poisoned { .. } => 0,
         }
     }
@@ -603,7 +609,7 @@ impl<W: AsyncWrite + Unpin> AsyncWrite for CryptoWriter<W> {
                 Poll::Pending => {
                     // Upstream blocked. Apply ideal backpressure
                     let to_accept =
-                        Self::select_to_accept_for_buffering(&this.state, buf.len());
+                        Self::select_to_accept_for_buffering(&this.state, buf.len(), this.max_pending_write);
 
                     if to_accept == 0 {
                         trace!(
@@ -618,7 +624,7 @@ impl<W: AsyncWrite + Unpin> AsyncWrite for CryptoWriter<W> {
 
                     // Disjoint borrows
                     let encryptor = &mut this.encryptor;
-                    let pending = Self::ensure_pending(&mut this.state);
+                    let pending = Self::ensure_pending(&mut this.state, this.max_pending_write);
 
                     if let Err(e) = pending.push_encrypted(encryptor, plaintext) {
                         if e.kind() == ErrorKind::WouldBlock {
@@ -635,7 +641,7 @@ impl<W: AsyncWrite + Unpin> AsyncWrite for CryptoWriter<W> {
         // 2) Fast path: pending empty -> write-through
         debug_assert!(matches!(this.state, CryptoWriterState::Idle));
 
-        let to_accept = buf.len().min(MAX_PENDING_WRITE);
+        let to_accept = buf.len().min(this.max_pending_write);
         let plaintext = &buf[..to_accept];
 
         Self::encrypt_into_scratch(&mut this.encryptor, &mut this.scratch, plaintext);
@@ -645,7 +651,7 @@ impl<W: AsyncWrite + Unpin> AsyncWrite for CryptoWriter<W> {
                 // Upstream blocked: buffer FULL ciphertext for accepted bytes.
                 let ciphertext = std::mem::take(&mut this.scratch);
 
-                let pending = Self::ensure_pending(&mut this.state);
+                let pending = Self::ensure_pending(&mut this.state, this.max_pending_write);
                 pending.replace_with(ciphertext);
 
                 Poll::Ready(Ok(to_accept))
@@ -672,7 +678,7 @@ impl<W: AsyncWrite + Unpin> AsyncWrite for CryptoWriter<W> {
                 let remainder = this.scratch.split_off(n);
                 this.scratch.clear();
 
-                let pending = Self::ensure_pending(&mut this.state);
+                let pending = Self::ensure_pending(&mut this.state, this.max_pending_write);
                 pending.replace_with(remainder);
 
                 Poll::Ready(Ok(to_accept))
@@ -767,4 +773,4 @@ impl<S: AsyncWrite + Unpin> AsyncWrite for PassthroughStream<S> {
     fn poll_shutdown(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Result<()>> {
         Pin::new(&mut self.inner).poll_shutdown(cx)
     }
-}
+}

src/stream/frame_codec.rs

@@ -8,7 +8,9 @@ use std::io::{self, Error, ErrorKind};
 use std::sync::Arc;
 use tokio_util::codec::{Decoder, Encoder};
 
-use crate::protocol::constants::ProtoTag;
+use crate::protocol::constants::{
+    ProtoTag, is_valid_secure_payload_len, secure_padding_len, secure_payload_len_from_wire_len,
+};
 use crate::crypto::SecureRandom;
 use super::frame::{Frame, FrameMeta, FrameCodec as FrameCodecTrait};
 
@@ -274,13 +276,13 @@ fn decode_secure(src: &mut BytesMut, max_size: usize) -> io::Result<Option<Frame
         return Ok(None);
     }
     
-    // Calculate padding (indicated by length not divisible by 4)
-    let padding_len = len % 4;
-    let data_len = if padding_len != 0 {
-        len - padding_len
-    } else {
-        len
-    };
+    let data_len = secure_payload_len_from_wire_len(len).ok_or_else(|| {
+        Error::new(
+            ErrorKind::InvalidData,
+            format!("invalid secure frame length: {len}"),
+        )
+    })?;
+    let padding_len = len - data_len;
     
     meta.padding_len = padding_len as u8;
     
@@ -303,14 +305,15 @@ fn encode_secure(frame: &Frame, dst: &mut BytesMut, rng: &SecureRandom) -> io::R
         return Ok(());
     }
     
-    // Generate padding to make length not divisible by 4
-    let padding_len = if data.len() % 4 == 0 {
-        // Add 1-3 bytes to make it non-aligned
-        (rng.range(3) + 1) as usize
-    } else {
-        // Already non-aligned, can add 0-3
-        rng.range(4) as usize
-    };
+    if !is_valid_secure_payload_len(data.len()) {
+        return Err(Error::new(
+            ErrorKind::InvalidData,
+            format!("secure payload must be 4-byte aligned, got {}", data.len()),
+        ));
+    }
+
+    // Generate padding that keeps total length non-divisible by 4.
+    let padding_len = secure_padding_len(data.len(), rng);
     
     let total_len = data.len() + padding_len;
     dst.reserve(4 + total_len);
@@ -625,4 +628,4 @@ mod tests {
         let result = codec.decode(&mut buf);
         assert!(result.is_err());
     }
-}
+}

src/stream/frame_stream.rs

@@ -232,11 +232,13 @@ impl<R: AsyncRead + Unpin> SecureIntermediateFrameReader<R> {
         let mut data = vec![0u8; len];
         self.upstream.read_exact(&mut data).await?;
         
-        // Strip padding (not aligned to 4)
-        if len % 4 != 0 {
-            let actual_len = len - (len % 4);
-            data.truncate(actual_len);
-        }
+        let payload_len = secure_payload_len_from_wire_len(len).ok_or_else(|| {
+            Error::new(
+                ErrorKind::InvalidData,
+                format!("Invalid secure frame length: {len}"),
+            )
+        })?;
+        data.truncate(payload_len);
         
         Ok((Bytes::from(data), meta))
     }
@@ -267,8 +269,15 @@ impl<W: AsyncWrite + Unpin> SecureIntermediateFrameWriter<W> {
             return Ok(());
         }
         
-        // Add random padding (0-3 bytes)
-        let padding_len = self.rng.range(4);
+        if !is_valid_secure_payload_len(data.len()) {
+            return Err(Error::new(
+                ErrorKind::InvalidData,
+                format!("Secure payload must be 4-byte aligned, got {}", data.len()),
+            ));
+        }
+
+        // Add padding so total length is never divisible by 4 (MTProto Secure)
+        let padding_len = secure_padding_len(data.len(), &self.rng);
         let padding = self.rng.bytes(padding_len);
         
         let total_len = data.len() + padding_len;
@@ -550,9 +559,7 @@ mod tests {
         writer.flush().await.unwrap();
         
         let (received, _meta) = reader.read_frame().await.unwrap();
-        // Received should have padding stripped to align to 4
-        let expected_len = (data.len() / 4) * 4;
-        assert_eq!(received.len(), expected_len);
+        assert_eq!(received.len(), data.len());
     }
     
     #[tokio::test]
@@ -585,4 +592,4 @@ mod tests {
         let (received, _) = reader.read_frame().await.unwrap();
         assert_eq!(&received[..], &data[..]);
     }
-}
+}

src/stream/mod.rs

@@ -40,4 +40,4 @@ pub use frame_stream::{
     SecureIntermediateFrameReader, SecureIntermediateFrameWriter,
     MtprotoFrameReader, MtprotoFrameWriter,
     FrameReaderKind, FrameWriterKind,
-};
+};

src/stream/tls_stream.rs

@@ -25,7 +25,8 @@
 //! - However, the on-the-wire record length can exceed 16384 because TLS 1.3
 //!   uses AEAD and can include tag/overhead/padding.
 //! - Telegram FakeTLS clients (notably iOS) may send Application Data records
-//!   with length up to 16384 + 24 bytes. We accept that as MAX_TLS_CHUNK_SIZE.
+//!   with length up to 16384 + 256 bytes (RFC 8446 §5.2). We accept that as
+//!   MAX_TLS_CHUNK_SIZE.
 //!
 //! If you reject those (e.g. validate length <= 16384), you will see errors like:
 //!   "TLS record too large: 16408 bytes"
@@ -52,9 +53,8 @@ use super::state::{StreamState, HeaderBuffer, YieldBuffer, WriteBuffer};
 const TLS_HEADER_SIZE: usize = 5;
 
 /// Maximum TLS fragment size we emit for Application Data.
-/// Real TLS 1.3 ciphertexts often add ~16-24 bytes AEAD overhead, so to mimic
-/// on-the-wire record sizes we allow up to 16384 + 24 bytes of plaintext.
-const MAX_TLS_PAYLOAD: usize = 16384 + 24;
+/// Real TLS 1.3 allows up to 16384 + 256 bytes of ciphertext (incl. tag).
+const MAX_TLS_PAYLOAD: usize = 16384 + 256;
 
 /// Maximum pending write buffer for one record remainder.
 /// Note: we never queue unlimited amount of data here; state holds at most one record.
@@ -91,7 +91,7 @@ impl TlsRecordHeader {
     /// - We accept TLS 1.0 header version for ClientHello-like records (0x03 0x01),
     ///   and TLS 1.2/1.3 style version bytes for the rest (we use TLS_VERSION = 0x03 0x03).
     /// - For Application Data, Telegram FakeTLS may send payload length up to
-    ///   MAX_TLS_CHUNK_SIZE (16384 + 24).
+    ///   MAX_TLS_CHUNK_SIZE (16384 + 256).
     /// - For other record types we keep stricter bounds to avoid memory abuse.
     fn validate(&self) -> Result<()> {
         // Version: accept TLS 1.0 header (ClientHello quirk) and TLS_VERSION (0x0303).
@@ -105,7 +105,7 @@ impl TlsRecordHeader {
         let len = self.length as usize;
 
         // Length checks depend on record type.
-        // Telegram FakeTLS: ApplicationData length may be 16384 + 24.
+        // Telegram FakeTLS: ApplicationData length may be 16384 + 256.
         match self.record_type {
             TLS_RECORD_APPLICATION => {
                 if len > MAX_TLS_CHUNK_SIZE {
@@ -755,9 +755,6 @@ impl<W: AsyncWrite + Unpin> AsyncWrite for FakeTlsWriter<W> {
                     payload_size: chunk_size,
                 };
 
-                // Wake to retry flushing soon.
-                cx.waker().wake_by_ref();
-
                 Poll::Ready(Ok(chunk_size))
             }
         }

src/tls_front/cache.rs

@@ -0,0 +1,163 @@
+use std::collections::HashMap;
+use std::path::{Path, PathBuf};
+use std::sync::Arc;
+use std::time::{SystemTime, Duration};
+
+use tokio::sync::RwLock;
+use tokio::time::sleep;
+use tracing::{debug, warn, info};
+
+use crate::tls_front::types::{CachedTlsData, ParsedServerHello, TlsFetchResult};
+
+/// Lightweight in-memory + optional on-disk cache for TLS fronting data.
+#[derive(Debug)]
+pub struct TlsFrontCache {
+    memory: RwLock<HashMap<String, Arc<CachedTlsData>>>,
+    default: Arc<CachedTlsData>,
+    disk_path: PathBuf,
+}
+
+impl TlsFrontCache {
+    pub fn new(domains: &[String], default_len: usize, disk_path: impl AsRef<Path>) -> Self {
+        let default_template = ParsedServerHello {
+            version: [0x03, 0x03],
+            random: [0u8; 32],
+            session_id: Vec::new(),
+            cipher_suite: [0x13, 0x01],
+            compression: 0,
+            extensions: Vec::new(),
+        };
+
+        let default = Arc::new(CachedTlsData {
+            server_hello_template: default_template,
+            cert_info: None,
+            app_data_records_sizes: vec![default_len],
+            total_app_data_len: default_len,
+            fetched_at: SystemTime::now(),
+            domain: "default".to_string(),
+        });
+
+        let mut map = HashMap::new();
+        for d in domains {
+            map.insert(d.clone(), default.clone());
+        }
+
+        Self {
+            memory: RwLock::new(map),
+            default,
+            disk_path: disk_path.as_ref().to_path_buf(),
+        }
+    }
+
+    pub async fn get(&self, sni: &str) -> Arc<CachedTlsData> {
+        let guard = self.memory.read().await;
+        guard.get(sni).cloned().unwrap_or_else(|| self.default.clone())
+    }
+
+    pub async fn set(&self, domain: &str, data: CachedTlsData) {
+        let mut guard = self.memory.write().await;
+        guard.insert(domain.to_string(), Arc::new(data));
+    }
+
+    pub async fn load_from_disk(&self) {
+        let path = self.disk_path.clone();
+        if tokio::fs::create_dir_all(&path).await.is_err() {
+            return;
+        }
+        let mut loaded = 0usize;
+        if let Ok(mut dir) = tokio::fs::read_dir(&path).await {
+            while let Ok(Some(entry)) = dir.next_entry().await {
+                if let Ok(name) = entry.file_name().into_string() {
+                    if !name.ends_with(".json") {
+                        continue;
+                    }
+                    if let Ok(data) = tokio::fs::read(entry.path()).await {
+                        if let Ok(mut cached) = serde_json::from_slice::<CachedTlsData>(&data) {
+                            if cached.domain.is_empty()
+                                || cached.domain.len() > 255
+                                || !cached.domain.chars().all(|c| c.is_ascii_alphanumeric() || c == '.' || c == '-')
+                            {
+                                warn!(file = %name, "Skipping TLS cache entry with invalid domain");
+                                continue;
+                            }
+                            // fetched_at is skipped during deserialization; approximate with file mtime if available.
+                            if let Ok(meta) = entry.metadata().await {
+                                if let Ok(modified) = meta.modified() {
+                                    cached.fetched_at = modified;
+                                }
+                            }
+                            // Drop entries older than 72h
+                            if let Ok(age) = cached.fetched_at.elapsed() {
+                                if age > Duration::from_secs(72 * 3600) {
+                                    warn!(domain = %cached.domain, "Skipping stale TLS cache entry (>72h)");
+                                    continue;
+                                }
+                            }
+                            let domain = cached.domain.clone();
+                            self.set(&domain, cached).await;
+                            loaded += 1;
+                        }
+                    }
+                }
+            }
+        }
+        if loaded > 0 {
+            info!(count = loaded, "Loaded TLS cache entries from disk");
+        }
+    }
+
+    async fn persist(&self, domain: &str, data: &CachedTlsData) {
+        if tokio::fs::create_dir_all(&self.disk_path).await.is_err() {
+            return;
+        }
+        let fname = format!("{}.json", domain.replace(['/', '\\'], "_"));
+        let path = self.disk_path.join(fname);
+        if let Ok(json) = serde_json::to_vec_pretty(data) {
+            // best-effort write
+            let _ = tokio::fs::write(path, json).await;
+        }
+    }
+
+    /// Spawn background updater that periodically refreshes cached domains using provided fetcher.
+    pub fn spawn_updater<F>(
+        self: Arc<Self>,
+        domains: Vec<String>,
+        interval: Duration,
+        fetcher: F,
+    ) where
+        F: Fn(String) -> tokio::task::JoinHandle<()> + Send + Sync + 'static,
+    {
+        tokio::spawn(async move {
+            loop {
+                for domain in &domains {
+                    fetcher(domain.clone()).await;
+                }
+                sleep(interval).await;
+            }
+        });
+    }
+
+    /// Replace cached entry from a fetch result.
+    pub async fn update_from_fetch(&self, domain: &str, fetched: TlsFetchResult) {
+        let data = CachedTlsData {
+            server_hello_template: fetched.server_hello_parsed,
+            cert_info: fetched.cert_info,
+            app_data_records_sizes: fetched.app_data_records_sizes.clone(),
+            total_app_data_len: fetched.total_app_data_len,
+            fetched_at: SystemTime::now(),
+            domain: domain.to_string(),
+        };
+
+        self.set(domain, data.clone()).await;
+        self.persist(domain, &data).await;
+        debug!(domain = %domain, len = fetched.total_app_data_len, "TLS cache updated");
+    }
+
+    pub fn default_entry(&self) -> Arc<CachedTlsData> {
+        self.default.clone()
+    }
+
+    pub fn disk_path(&self) -> &Path {
+        &self.disk_path
+    }
+}

src/tls_front/emulator.rs

@@ -0,0 +1,160 @@
+use crate::crypto::{sha256_hmac, SecureRandom};
+use crate::protocol::constants::{
+    TLS_RECORD_APPLICATION, TLS_RECORD_CHANGE_CIPHER, TLS_RECORD_HANDSHAKE, TLS_VERSION,
+};
+use crate::protocol::tls::{TLS_DIGEST_LEN, TLS_DIGEST_POS, gen_fake_x25519_key};
+use crate::tls_front::types::CachedTlsData;
+
+const MIN_APP_DATA: usize = 64;
+const MAX_APP_DATA: usize = 16640; // RFC 8446 §5.2 allows up to 2^14 + 256
+
+fn jitter_and_clamp_sizes(sizes: &[usize], rng: &SecureRandom) -> Vec<usize> {
+    sizes
+        .iter()
+        .map(|&size| {
+            let base = size.max(MIN_APP_DATA).min(MAX_APP_DATA);
+            let jitter_range = ((base as f64) * 0.03).round() as i64;
+            if jitter_range == 0 {
+                return base;
+            }
+            let mut rand_bytes = [0u8; 2];
+            rand_bytes.copy_from_slice(&rng.bytes(2));
+            let span = 2 * jitter_range + 1;
+            let delta = (u16::from_le_bytes(rand_bytes) as i64 % span) - jitter_range;
+            let adjusted = (base as i64 + delta).clamp(MIN_APP_DATA as i64, MAX_APP_DATA as i64);
+            adjusted as usize
+        })
+        .collect()
+}
+
+/// Build a ServerHello + CCS + ApplicationData sequence using cached TLS metadata.
+pub fn build_emulated_server_hello(
+    secret: &[u8],
+    client_digest: &[u8; TLS_DIGEST_LEN],
+    session_id: &[u8],
+    cached: &CachedTlsData,
+    rng: &SecureRandom,
+    alpn: Option<Vec<u8>>,
+    new_session_tickets: u8,
+) -> Vec<u8> {
+    // --- ServerHello ---
+    let mut extensions = Vec::new();
+    // KeyShare (x25519)
+    let key = gen_fake_x25519_key(rng);
+    extensions.extend_from_slice(&0x0033u16.to_be_bytes()); // key_share
+    extensions.extend_from_slice(&(2 + 2 + 32u16).to_be_bytes()); // len
+    extensions.extend_from_slice(&0x001du16.to_be_bytes()); // X25519
+    extensions.extend_from_slice(&(32u16).to_be_bytes());
+    extensions.extend_from_slice(&key);
+    // supported_versions (TLS1.3)
+    extensions.extend_from_slice(&0x002bu16.to_be_bytes());
+    extensions.extend_from_slice(&(2u16).to_be_bytes());
+    extensions.extend_from_slice(&0x0304u16.to_be_bytes());
+    if let Some(alpn_proto) = &alpn {
+        extensions.extend_from_slice(&0x0010u16.to_be_bytes());
+        let list_len: u16 = 1 + alpn_proto.len() as u16;
+        let ext_len: u16 = 2 + list_len;
+        extensions.extend_from_slice(&ext_len.to_be_bytes());
+        extensions.extend_from_slice(&list_len.to_be_bytes());
+        extensions.push(alpn_proto.len() as u8);
+        extensions.extend_from_slice(alpn_proto);
+    }
+
+    let extensions_len = extensions.len() as u16;
+
+    let body_len = 2 + // version
+        32 + // random
+        1 + session_id.len() + // session id
+        2 + // cipher
+        1 + // compression
+        2 + extensions.len(); // extensions
+
+    let mut message = Vec::with_capacity(4 + body_len);
+    message.push(0x02); // ServerHello
+    let len_bytes = (body_len as u32).to_be_bytes();
+    message.extend_from_slice(&len_bytes[1..4]);
+    message.extend_from_slice(&cached.server_hello_template.version); // 0x0303
+    message.extend_from_slice(&[0u8; 32]); // random placeholder
+    message.push(session_id.len() as u8);
+    message.extend_from_slice(session_id);
+    let cipher = if cached.server_hello_template.cipher_suite == [0, 0] {
+        [0x13, 0x01]
+    } else {
+        cached.server_hello_template.cipher_suite
+    };
+    message.extend_from_slice(&cipher);
+    message.push(cached.server_hello_template.compression);
+    message.extend_from_slice(&extensions_len.to_be_bytes());
+    message.extend_from_slice(&extensions);
+
+    let mut server_hello = Vec::with_capacity(5 + message.len());
+    server_hello.push(TLS_RECORD_HANDSHAKE);
+    server_hello.extend_from_slice(&TLS_VERSION);
+    server_hello.extend_from_slice(&(message.len() as u16).to_be_bytes());
+    server_hello.extend_from_slice(&message);
+
+    // --- ChangeCipherSpec ---
+    let change_cipher_spec = [
+        TLS_RECORD_CHANGE_CIPHER,
+        TLS_VERSION[0],
+        TLS_VERSION[1],
+        0x00,
+        0x01,
+        0x01,
+    ];
+
+    // --- ApplicationData (fake encrypted records) ---
+    // Use the same number and sizes of ApplicationData records as the cached server.
+    let mut sizes = cached.app_data_records_sizes.clone();
+    if sizes.is_empty() {
+        sizes.push(cached.total_app_data_len.max(1024));
+    }
+    let sizes = jitter_and_clamp_sizes(&sizes, rng);
+
+    let mut app_data = Vec::new();
+    for size in sizes {
+        let mut rec = Vec::with_capacity(5 + size);
+        rec.push(TLS_RECORD_APPLICATION);
+        rec.extend_from_slice(&TLS_VERSION);
+        rec.extend_from_slice(&(size as u16).to_be_bytes());
+        if size > 17 {
+            let body_len = size - 17;
+            rec.extend_from_slice(&rng.bytes(body_len));
+            rec.push(0x16); // inner content type marker (handshake)
+            rec.extend_from_slice(&rng.bytes(16)); // AEAD-like tag
+        } else {
+            rec.extend_from_slice(&rng.bytes(size));
+        }
+        app_data.extend_from_slice(&rec);
+    }
+
+    // --- Combine ---
+    // Optional NewSessionTicket mimic records (opaque ApplicationData for fingerprint).
+    let mut tickets = Vec::new();
+    if new_session_tickets > 0 {
+        for _ in 0..new_session_tickets {
+            let ticket_len: usize = rng.range(48) + 48;
+            let mut rec = Vec::with_capacity(5 + ticket_len);
+            rec.push(TLS_RECORD_APPLICATION);
+            rec.extend_from_slice(&TLS_VERSION);
+            rec.extend_from_slice(&(ticket_len as u16).to_be_bytes());
+            rec.extend_from_slice(&rng.bytes(ticket_len));
+            tickets.extend_from_slice(&rec);
+        }
+    }
+
+    let mut response = Vec::with_capacity(server_hello.len() + change_cipher_spec.len() + app_data.len() + tickets.len());
+    response.extend_from_slice(&server_hello);
+    response.extend_from_slice(&change_cipher_spec);
+    response.extend_from_slice(&app_data);
+    response.extend_from_slice(&tickets);
+
+    // --- HMAC ---
+    let mut hmac_input = Vec::with_capacity(TLS_DIGEST_LEN + response.len());
+    hmac_input.extend_from_slice(client_digest);
+    hmac_input.extend_from_slice(&response);
+    let digest = sha256_hmac(secret, &hmac_input);
+    response[TLS_DIGEST_POS..TLS_DIGEST_POS + TLS_DIGEST_LEN].copy_from_slice(&digest);
+
+    response
+}

src/tls_front/fetcher.rs

@@ -0,0 +1,465 @@
+use std::sync::Arc;
+use std::time::Duration;
+
+use anyhow::{Context, Result, anyhow};
+use tokio::io::{AsyncReadExt, AsyncWriteExt};
+use tokio::net::TcpStream;
+use tokio::time::timeout;
+use tokio_rustls::client::TlsStream;
+use tokio_rustls::TlsConnector;
+use tracing::{debug, warn};
+
+use rustls::client::danger::{HandshakeSignatureValid, ServerCertVerified, ServerCertVerifier};
+use rustls::client::ClientConfig;
+use rustls::pki_types::{CertificateDer, ServerName, UnixTime};
+use rustls::{DigitallySignedStruct, Error as RustlsError};
+
+use x509_parser::prelude::FromDer;
+use x509_parser::certificate::X509Certificate;
+
+use crate::crypto::SecureRandom;
+use crate::protocol::constants::{TLS_RECORD_APPLICATION, TLS_RECORD_HANDSHAKE};
+use crate::tls_front::types::{ParsedServerHello, TlsExtension, TlsFetchResult, ParsedCertificateInfo};
+
+/// No-op verifier: accept any certificate (we only need lengths and metadata).
+#[derive(Debug)]
+struct NoVerify;
+
+impl ServerCertVerifier for NoVerify {
+    fn verify_server_cert(
+        &self,
+        _end_entity: &CertificateDer<'_>,
+        _intermediates: &[CertificateDer<'_>],
+        _server_name: &ServerName<'_>,
+        _ocsp: &[u8],
+        _now: UnixTime,
+    ) -> Result<ServerCertVerified, RustlsError> {
+        Ok(ServerCertVerified::assertion())
+    }
+
+    fn verify_tls12_signature(
+        &self,
+        _message: &[u8],
+        _cert: &CertificateDer<'_>,
+        _dss: &DigitallySignedStruct,
+    ) -> Result<HandshakeSignatureValid, RustlsError> {
+        Ok(HandshakeSignatureValid::assertion())
+    }
+
+    fn verify_tls13_signature(
+        &self,
+        _message: &[u8],
+        _cert: &CertificateDer<'_>,
+        _dss: &DigitallySignedStruct,
+    ) -> Result<HandshakeSignatureValid, RustlsError> {
+        Ok(HandshakeSignatureValid::assertion())
+    }
+
+    fn supported_verify_schemes(&self) -> Vec<rustls::SignatureScheme> {
+        use rustls::SignatureScheme::*;
+        vec![
+            RSA_PKCS1_SHA256,
+            RSA_PSS_SHA256,
+            ECDSA_NISTP256_SHA256,
+            ECDSA_NISTP384_SHA384,
+        ]
+    }
+}
+
+fn build_client_config() -> Arc<ClientConfig> {
+    let root = rustls::RootCertStore::empty();
+
+    let provider = rustls::crypto::ring::default_provider();
+    let mut config = ClientConfig::builder_with_provider(Arc::new(provider))
+        .with_protocol_versions(&[&rustls::version::TLS13, &rustls::version::TLS12])
+        .expect("protocol versions")
+        .with_root_certificates(root)
+        .with_no_client_auth();
+
+    config
+        .dangerous()
+        .set_certificate_verifier(Arc::new(NoVerify));
+
+    Arc::new(config)
+}
+
+fn build_client_hello(sni: &str, rng: &SecureRandom) -> Vec<u8> {
+    // === ClientHello body ===
+    let mut body = Vec::new();
+
+    // Legacy version (TLS 1.0) as in real ClientHello headers
+    body.extend_from_slice(&[0x03, 0x03]);
+
+    // Random
+    body.extend_from_slice(&rng.bytes(32));
+
+    // Session ID: empty
+    body.push(0);
+
+    // Cipher suites (common minimal set, TLS1.3 + a few 1.2 fallbacks)
+    let cipher_suites: [u8; 10] = [
+        0x13, 0x01, // TLS_AES_128_GCM_SHA256
+        0x13, 0x02, // TLS_AES_256_GCM_SHA384
+        0x13, 0x03, // TLS_CHACHA20_POLY1305_SHA256
+        0x00, 0x2f, // TLS_RSA_WITH_AES_128_CBC_SHA (legacy)
+        0x00, 0xff, // RENEGOTIATION_INFO_SCSV
+    ];
+    body.extend_from_slice(&(cipher_suites.len() as u16).to_be_bytes());
+    body.extend_from_slice(&cipher_suites);
+
+    // Compression methods: null only
+    body.push(1);
+    body.push(0);
+
+    // === Extensions ===
+    let mut exts = Vec::new();
+
+    // server_name (SNI)
+    let sni_bytes = sni.as_bytes();
+    let mut sni_ext = Vec::with_capacity(5 + sni_bytes.len());
+    sni_ext.extend_from_slice(&(sni_bytes.len() as u16 + 3).to_be_bytes());
+    sni_ext.push(0); // host_name
+    sni_ext.extend_from_slice(&(sni_bytes.len() as u16).to_be_bytes());
+    sni_ext.extend_from_slice(sni_bytes);
+    exts.extend_from_slice(&0x0000u16.to_be_bytes());
+    exts.extend_from_slice(&(sni_ext.len() as u16).to_be_bytes());
+    exts.extend_from_slice(&sni_ext);
+
+    // supported_groups
+    let groups: [u16; 2] = [0x001d, 0x0017]; // x25519, secp256r1
+    exts.extend_from_slice(&0x000au16.to_be_bytes());
+    exts.extend_from_slice(&((2 + groups.len() * 2) as u16).to_be_bytes());
+    exts.extend_from_slice(&(groups.len() as u16 * 2).to_be_bytes());
+    for g in groups { exts.extend_from_slice(&g.to_be_bytes()); }
+
+    // signature_algorithms
+    let sig_algs: [u16; 4] = [0x0804, 0x0805, 0x0403, 0x0503]; // rsa_pss_rsae_sha256/384, ecdsa_secp256r1_sha256, rsa_pkcs1_sha256
+    exts.extend_from_slice(&0x000du16.to_be_bytes());
+    exts.extend_from_slice(&((2 + sig_algs.len() * 2) as u16).to_be_bytes());
+    exts.extend_from_slice(&(sig_algs.len() as u16 * 2).to_be_bytes());
+    for a in sig_algs { exts.extend_from_slice(&a.to_be_bytes()); }
+
+    // supported_versions (TLS1.3 + TLS1.2)
+    let versions: [u16; 2] = [0x0304, 0x0303];
+    exts.extend_from_slice(&0x002bu16.to_be_bytes());
+    exts.extend_from_slice(&((1 + versions.len() * 2) as u16).to_be_bytes());
+    exts.push((versions.len() * 2) as u8);
+    for v in versions { exts.extend_from_slice(&v.to_be_bytes()); }
+
+    // key_share (x25519)
+    let key = gen_key_share(rng);
+    let mut keyshare = Vec::with_capacity(4 + key.len());
+    keyshare.extend_from_slice(&0x001du16.to_be_bytes()); // group
+    keyshare.extend_from_slice(&(key.len() as u16).to_be_bytes());
+    keyshare.extend_from_slice(&key);
+    exts.extend_from_slice(&0x0033u16.to_be_bytes());
+    exts.extend_from_slice(&((2 + keyshare.len()) as u16).to_be_bytes());
+    exts.extend_from_slice(&(keyshare.len() as u16).to_be_bytes());
+    exts.extend_from_slice(&keyshare);
+
+    // ALPN (http/1.1)
+    let alpn_proto = b"http/1.1";
+    exts.extend_from_slice(&0x0010u16.to_be_bytes());
+    exts.extend_from_slice(&((2 + 1 + alpn_proto.len()) as u16).to_be_bytes());
+    exts.extend_from_slice(&((1 + alpn_proto.len()) as u16).to_be_bytes());
+    exts.push(alpn_proto.len() as u8);
+    exts.extend_from_slice(alpn_proto);
+
+    // padding to reduce recognizability and keep length ~500 bytes
+    const TARGET_EXT_LEN: usize = 180;
+    if exts.len() < TARGET_EXT_LEN {
+        let remaining = TARGET_EXT_LEN - exts.len();
+        if remaining > 4 {
+            let pad_len = remaining - 4; // minus type+len
+            exts.extend_from_slice(&0x0015u16.to_be_bytes()); // padding extension
+            exts.extend_from_slice(&(pad_len as u16).to_be_bytes());
+            exts.resize(exts.len() + pad_len, 0);
+        }
+    }
+
+    // Extensions length prefix
+    body.extend_from_slice(&(exts.len() as u16).to_be_bytes());
+    body.extend_from_slice(&exts);
+
+    // === Handshake wrapper ===
+    let mut handshake = Vec::new();
+    handshake.push(0x01); // ClientHello
+    let len_bytes = (body.len() as u32).to_be_bytes();
+    handshake.extend_from_slice(&len_bytes[1..4]);
+    handshake.extend_from_slice(&body);
+
+    // === Record ===
+    let mut record = Vec::new();
+    record.push(TLS_RECORD_HANDSHAKE);
+    record.extend_from_slice(&[0x03, 0x01]); // legacy record version
+    record.extend_from_slice(&(handshake.len() as u16).to_be_bytes());
+    record.extend_from_slice(&handshake);
+
+    record
+}
+
+fn gen_key_share(rng: &SecureRandom) -> [u8; 32] {
+    let mut key = [0u8; 32];
+    key.copy_from_slice(&rng.bytes(32));
+    key
+}
+
+async fn read_tls_record(stream: &mut TcpStream) -> Result<(u8, Vec<u8>)> {
+    let mut header = [0u8; 5];
+    stream.read_exact(&mut header).await?;
+    let len = u16::from_be_bytes([header[3], header[4]]) as usize;
+    let mut body = vec![0u8; len];
+    stream.read_exact(&mut body).await?;
+    Ok((header[0], body))
+}
+
+fn parse_server_hello(body: &[u8]) -> Option<ParsedServerHello> {
+    if body.len() < 4 || body[0] != 0x02 {
+        return None;
+    }
+
+    let msg_len = u32::from_be_bytes([0, body[1], body[2], body[3]]) as usize;
+    if msg_len + 4 > body.len() {
+        return None;
+    }
+
+    let mut pos = 4;
+    let version = [*body.get(pos)?, *body.get(pos + 1)?];
+    pos += 2;
+
+    let mut random = [0u8; 32];
+    random.copy_from_slice(body.get(pos..pos + 32)?);
+    pos += 32;
+
+    let session_len = *body.get(pos)? as usize;
+    pos += 1;
+    let session_id = body.get(pos..pos + session_len)?.to_vec();
+    pos += session_len;
+
+    let cipher_suite = [*body.get(pos)?, *body.get(pos + 1)?];
+    pos += 2;
+
+    let compression = *body.get(pos)?;
+    pos += 1;
+
+    let ext_len = u16::from_be_bytes([*body.get(pos)?, *body.get(pos + 1)?]) as usize;
+    pos += 2;
+    let ext_end = pos.checked_add(ext_len)?;
+    if ext_end > body.len() {
+        return None;
+    }
+
+    let mut extensions = Vec::new();
+    while pos + 4 <= ext_end {
+        let etype = u16::from_be_bytes([body[pos], body[pos + 1]]);
+        let elen = u16::from_be_bytes([body[pos + 2], body[pos + 3]]) as usize;
+        pos += 4;
+        let data = body.get(pos..pos + elen)?.to_vec();
+        pos += elen;
+        extensions.push(TlsExtension { ext_type: etype, data });
+    }
+
+    Some(ParsedServerHello {
+        version,
+        random,
+        session_id,
+        cipher_suite,
+        compression,
+        extensions,
+    })
+}
+
+fn parse_cert_info(certs: &[CertificateDer<'static>]) -> Option<ParsedCertificateInfo> {
+    let first = certs.first()?;
+    let (_rem, cert) = X509Certificate::from_der(first.as_ref()).ok()?;
+
+    let not_before = Some(cert.validity().not_before.to_datetime().unix_timestamp());
+    let not_after = Some(cert.validity().not_after.to_datetime().unix_timestamp());
+
+    let issuer_cn = cert
+        .issuer()
+        .iter_common_name()
+        .next()
+        .and_then(|cn| cn.as_str().ok())
+        .map(|s| s.to_string());
+
+    let subject_cn = cert
+        .subject()
+        .iter_common_name()
+        .next()
+        .and_then(|cn| cn.as_str().ok())
+        .map(|s| s.to_string());
+
+    let san_names = cert
+        .subject_alternative_name()
+        .ok()
+        .flatten()
+        .map(|san| {
+            san.value
+                .general_names
+                .iter()
+                .filter_map(|gn| match gn {
+                    x509_parser::extensions::GeneralName::DNSName(n) => Some(n.to_string()),
+                    _ => None,
+                })
+                .collect::<Vec<_>>()
+        })
+        .unwrap_or_default();
+
+    Some(ParsedCertificateInfo {
+        not_after_unix: not_after,
+        not_before_unix: not_before,
+        issuer_cn,
+        subject_cn,
+        san_names,
+    })
+}
+
+async fn fetch_via_raw_tls(
+    host: &str,
+    port: u16,
+    sni: &str,
+    connect_timeout: Duration,
+) -> Result<TlsFetchResult> {
+    let addr = format!("{host}:{port}");
+    let mut stream = timeout(connect_timeout, TcpStream::connect(addr)).await??;
+
+    let rng = SecureRandom::new();
+    let client_hello = build_client_hello(sni, &rng);
+    timeout(connect_timeout, async {
+        stream.write_all(&client_hello).await?;
+        stream.flush().await?;
+        Ok::<(), std::io::Error>(())
+    })
+    .await??;
+
+    let mut records = Vec::new();
+    // Read up to 4 records: ServerHello, CCS, and up to two ApplicationData.
+    for _ in 0..4 {
+        match timeout(connect_timeout, read_tls_record(&mut stream)).await {
+            Ok(Ok(rec)) => records.push(rec),
+            Ok(Err(e)) => return Err(e.into()),
+            Err(_) => break,
+        }
+        if records.len() >= 3 && records.iter().any(|(t, _)| *t == TLS_RECORD_APPLICATION) {
+            break;
+        }
+    }
+
+    let mut app_sizes = Vec::new();
+    let mut server_hello = None;
+    for (t, body) in &records {
+        if *t == TLS_RECORD_HANDSHAKE && server_hello.is_none() {
+            server_hello = parse_server_hello(body);
+        } else if *t == TLS_RECORD_APPLICATION {
+            app_sizes.push(body.len());
+        }
+    }
+
+    let parsed = server_hello.ok_or_else(|| anyhow!("ServerHello not received"))?;
+    let total_app_data_len = app_sizes.iter().sum::<usize>().max(1024);
+
+    Ok(TlsFetchResult {
+        server_hello_parsed: parsed,
+        app_data_records_sizes: if app_sizes.is_empty() {
+            vec![total_app_data_len]
+        } else {
+            app_sizes
+        },
+        total_app_data_len,
+        cert_info: None,
+    })
+}
+
+/// Fetch real TLS metadata for the given SNI: negotiated cipher and cert lengths.
+pub async fn fetch_real_tls(
+    host: &str,
+    port: u16,
+    sni: &str,
+    connect_timeout: Duration,
+    upstream: Option<std::sync::Arc<crate::transport::UpstreamManager>>,
+) -> Result<TlsFetchResult> {
+    // Preferred path: raw TLS probe for accurate record sizing
+    match fetch_via_raw_tls(host, port, sni, connect_timeout).await {
+        Ok(res) => return Ok(res),
+        Err(e) => {
+            warn!(sni = %sni, error = %e, "Raw TLS fetch failed, falling back to rustls");
+        }
+    }
+
+    // Fallback: rustls handshake to at least get certificate sizes
+    let stream = if let Some(manager) = upstream {
+        // Resolve host to SocketAddr
+        if let Ok(mut addrs) = tokio::net::lookup_host((host, port)).await {
+            if let Some(addr) = addrs.find(|a| a.is_ipv4()) {
+                match manager.connect(addr, None, None).await {
+                    Ok(s) => s,
+                    Err(e) => {
+                        warn!(sni = %sni, error = %e, "Upstream connect failed, using direct connect");
+                        timeout(connect_timeout, TcpStream::connect((host, port))).await??
+                    }
+                }
+            } else {
+                timeout(connect_timeout, TcpStream::connect((host, port))).await??
+            }
+        } else {
+            timeout(connect_timeout, TcpStream::connect((host, port))).await??
+        }
+    } else {
+        timeout(connect_timeout, TcpStream::connect((host, port))).await??
+    };
+
+    let config = build_client_config();
+    let connector = TlsConnector::from(config);
+
+    let server_name = ServerName::try_from(sni.to_owned())
+        .or_else(|_| ServerName::try_from(host.to_owned()))
+        .map_err(|_| RustlsError::General("invalid SNI".into()))?;
+
+    let tls_stream: TlsStream<TcpStream> = connector.connect(server_name, stream).await?;
+
+    // Extract negotiated parameters and certificates
+    let (_io, session) = tls_stream.get_ref();
+    let cipher_suite = session
+        .negotiated_cipher_suite()
+        .map(|s| u16::from(s.suite()).to_be_bytes())
+        .unwrap_or([0x13, 0x01]);
+
+    let certs: Vec<CertificateDer<'static>> = session
+        .peer_certificates()
+        .map(|slice| slice.to_vec())
+        .unwrap_or_default();
+
+    let total_cert_len: usize = certs.iter().map(|c| c.len()).sum::<usize>().max(1024);
+    let cert_info = parse_cert_info(&certs);
+
+    // Heuristic: split across two records if large to mimic real servers a bit.
+    let app_data_records_sizes = if total_cert_len > 3000 {
+        vec![total_cert_len / 2, total_cert_len - total_cert_len / 2]
+    } else {
+        vec![total_cert_len]
+    };
+
+    let parsed = ParsedServerHello {
+        version: [0x03, 0x03],
+        random: [0u8; 32],
+        session_id: Vec::new(),
+        cipher_suite,
+        compression: 0,
+        extensions: Vec::new(),
+    };
+
+    debug!(
+        sni = %sni,
+        len = total_cert_len,
+        cipher = format!("0x{:04x}", u16::from_be_bytes(cipher_suite)),
+        "Fetched TLS metadata via rustls"
+    );
+
+    Ok(TlsFetchResult {
+        server_hello_parsed: parsed,
+        app_data_records_sizes: app_data_records_sizes.clone(),
+        total_app_data_len: app_data_records_sizes.iter().sum(),
+        cert_info,
+    })
+}

src/tls_front/mod.rs

@@ -0,0 +1,7 @@
+pub mod types;
+pub mod cache;
+pub mod fetcher;
+pub mod emulator;
+
+pub use cache::TlsFrontCache;
+pub use types::{CachedTlsData, TlsFetchResult};

src/tls_front/types.rs

@@ -0,0 +1,55 @@
+use std::time::SystemTime;
+use serde::{Serialize, Deserialize};
+
+/// Parsed representation of an unencrypted TLS ServerHello.
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct ParsedServerHello {
+    pub version: [u8; 2],
+    pub random: [u8; 32],
+    pub session_id: Vec<u8>,
+    pub cipher_suite: [u8; 2],
+    pub compression: u8,
+    pub extensions: Vec<TlsExtension>,
+}
+
+/// Generic TLS extension container.
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct TlsExtension {
+    pub ext_type: u16,
+    pub data: Vec<u8>,
+}
+
+/// Basic certificate metadata (optional, informative).
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct ParsedCertificateInfo {
+    pub not_after_unix: Option<i64>,
+    pub not_before_unix: Option<i64>,
+    pub issuer_cn: Option<String>,
+    pub subject_cn: Option<String>,
+    pub san_names: Vec<String>,
+}
+
+/// Cached data per SNI used by the emulator.
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct CachedTlsData {
+    pub server_hello_template: ParsedServerHello,
+    pub cert_info: Option<ParsedCertificateInfo>,
+    pub app_data_records_sizes: Vec<usize>,
+    pub total_app_data_len: usize,
+    #[serde(default = "now_system_time", skip_serializing, skip_deserializing)]
+    pub fetched_at: SystemTime,
+    pub domain: String,
+}
+
+fn now_system_time() -> SystemTime {
+    SystemTime::now()
+}
+
+/// Result of attempting to fetch real TLS artifacts.
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct TlsFetchResult {
+    pub server_hello_parsed: ParsedServerHello,
+    pub app_data_records_sizes: Vec<usize>,
+    pub total_app_data_len: usize,
+    pub cert_info: Option<ParsedCertificateInfo>,
+}

src/transport/middle_proxy/codec.rs

@@ -1,16 +1,53 @@
 use tokio::io::{AsyncReadExt, AsyncWriteExt};
 
-use crate::crypto::{AesCbc, crc32};
+use crate::crypto::{AesCbc, crc32, crc32c};
 use crate::error::{ProxyError, Result};
 use crate::protocol::constants::*;
 
-pub(crate) fn build_rpc_frame(seq_no: i32, payload: &[u8]) -> Vec<u8> {
+/// Commands sent to dedicated writer tasks to avoid mutex contention on TCP writes.
+pub(crate) enum WriterCommand {
+    Data(Vec<u8>),
+    DataAndFlush(Vec<u8>),
+    Close,
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub(crate) enum RpcChecksumMode {
+    Crc32,
+    Crc32c,
+}
+
+impl RpcChecksumMode {
+    pub(crate) fn from_handshake_flags(flags: u32) -> Self {
+        if (flags & rpc_crypto_flags::USE_CRC32C) != 0 {
+            Self::Crc32c
+        } else {
+            Self::Crc32
+        }
+    }
+
+    pub(crate) fn advertised_flags(self) -> u32 {
+        match self {
+            Self::Crc32 => 0,
+            Self::Crc32c => rpc_crypto_flags::USE_CRC32C,
+        }
+    }
+}
+
+pub(crate) fn rpc_crc(mode: RpcChecksumMode, data: &[u8]) -> u32 {
+    match mode {
+        RpcChecksumMode::Crc32 => crc32(data),
+        RpcChecksumMode::Crc32c => crc32c(data),
+    }
+}
+
+pub(crate) fn build_rpc_frame(seq_no: i32, payload: &[u8], crc_mode: RpcChecksumMode) -> Vec<u8> {
     let total_len = (4 + 4 + payload.len() + 4) as u32;
     let mut frame = Vec::with_capacity(total_len as usize);
     frame.extend_from_slice(&total_len.to_le_bytes());
     frame.extend_from_slice(&seq_no.to_le_bytes());
     frame.extend_from_slice(payload);
-    let c = crc32(&frame);
+    let c = rpc_crc(crc_mode, &frame);
     frame.extend_from_slice(&c.to_le_bytes());
     frame
 }
@@ -37,7 +74,7 @@ pub(crate) async fn read_rpc_frame_plaintext(
 
     let crc_offset = total_len - 4;
     let expected_crc = u32::from_le_bytes(full[crc_offset..crc_offset + 4].try_into().unwrap());
-    let actual_crc = crc32(&full[..crc_offset]);
+    let actual_crc = rpc_crc(RpcChecksumMode::Crc32, &full[..crc_offset]);
     if expected_crc != actual_crc {
         return Err(ProxyError::InvalidHandshake(format!(
             "CRC mismatch: 0x{expected_crc:08x} vs 0x{actual_crc:08x}"
@@ -87,24 +124,52 @@ pub(crate) fn build_handshake_payload(
     our_port: u16,
     peer_ip: [u8; 4],
     peer_port: u16,
+    flags: u32,
 ) -> [u8; 32] {
     let mut p = [0u8; 32];
     p[0..4].copy_from_slice(&RPC_HANDSHAKE_U32.to_le_bytes());
+    p[4..8].copy_from_slice(&flags.to_le_bytes());
 
-    // Keep C memory layout compatibility for PID IPv4 bytes.
+    // process_id sender_pid
     p[8..12].copy_from_slice(&our_ip);
     p[12..14].copy_from_slice(&our_port.to_le_bytes());
-    let pid = (std::process::id() & 0xffff) as u16;
-    p[14..16].copy_from_slice(&pid.to_le_bytes());
+    p[14..16].copy_from_slice(&process_pid16().to_le_bytes());
+    p[16..20].copy_from_slice(&process_utime().to_le_bytes());
+
+    // process_id peer_pid
+    p[20..24].copy_from_slice(&peer_ip);
+    p[24..26].copy_from_slice(&peer_port.to_le_bytes());
+    p[26..28].copy_from_slice(&0u16.to_le_bytes());
+    p[28..32].copy_from_slice(&0u32.to_le_bytes());
+    p
+}
+
+pub(crate) fn parse_handshake_flags(payload: &[u8]) -> Result<u32> {
+    if payload.len() != 32 {
+        return Err(ProxyError::InvalidHandshake(format!(
+            "Bad handshake payload len: {}",
+            payload.len()
+        )));
+    }
+    let hs_type = u32::from_le_bytes(payload[0..4].try_into().unwrap());
+    if hs_type != RPC_HANDSHAKE_U32 {
+        return Err(ProxyError::InvalidHandshake(format!(
+            "Expected HANDSHAKE 0x{RPC_HANDSHAKE_U32:08x}, got 0x{hs_type:08x}"
+        )));
+    }
+    Ok(u32::from_le_bytes(payload[4..8].try_into().unwrap()))
+}
+
+fn process_pid16() -> u16 {
+    (std::process::id() & 0xffff) as u16
+}
+
+fn process_utime() -> u32 {
     let utime = std::time::SystemTime::now()
         .duration_since(std::time::UNIX_EPOCH)
         .unwrap_or_default()
         .as_secs() as u32;
-    p[16..20].copy_from_slice(&utime.to_le_bytes());
-
-    p[20..24].copy_from_slice(&peer_ip);
-    p[24..26].copy_from_slice(&peer_port.to_le_bytes());
-    p
+    utime
 }
 
 pub(crate) fn cbc_encrypt_padded(
@@ -152,11 +217,12 @@ pub(crate) struct RpcWriter {
     pub(crate) key: [u8; 32],
     pub(crate) iv: [u8; 16],
     pub(crate) seq_no: i32,
+    pub(crate) crc_mode: RpcChecksumMode,
 }
 
 impl RpcWriter {
     pub(crate) async fn send(&mut self, payload: &[u8]) -> Result<()> {
-        let frame = build_rpc_frame(self.seq_no, payload);
+        let frame = build_rpc_frame(self.seq_no, payload, self.crc_mode);
         self.seq_no += 1;
 
         let pad = (16 - (frame.len() % 16)) % 16;

src/transport/middle_proxy/config_updater.rs

@@ -13,6 +13,24 @@ use super::secret::download_proxy_secret;
 use crate::crypto::SecureRandom;
 use std::time::SystemTime;
 
+async fn retry_fetch(url: &str) -> Option<ProxyConfigData> {
+    let delays = [1u64, 5, 15];
+    for (i, d) in delays.iter().enumerate() {
+        match fetch_proxy_config(url).await {
+            Ok(cfg) => return Some(cfg),
+            Err(e) => {
+                if i == delays.len() - 1 {
+                    warn!(error = %e, url, "fetch_proxy_config failed");
+                } else {
+                    debug!(error = %e, url, "fetch_proxy_config retrying");
+                    tokio::time::sleep(Duration::from_secs(*d)).await;
+                }
+            }
+        }
+    }
+    None
+}
+
 #[derive(Debug, Clone, Default)]
 pub struct ProxyConfigData {
     pub map: HashMap<i32, Vec<(IpAddr, u16)>>,
@@ -118,7 +136,8 @@ pub async fn me_config_updater(pool: Arc<MePool>, rng: Arc<SecureRandom>, interv
         tick.tick().await;
 
         // Update proxy config v4
-        if let Ok(cfg) = fetch_proxy_config("https://core.telegram.org/getProxyConfig").await {
+        let cfg_v4 = retry_fetch("https://core.telegram.org/getProxyConfig").await;
+        if let Some(cfg) = cfg_v4 {
             let changed = pool.update_proxy_maps(cfg.map.clone(), None).await;
             if let Some(dc) = cfg.default_dc {
                 pool.default_dc.store(dc, std::sync::atomic::Ordering::Relaxed);
@@ -129,14 +148,20 @@ pub async fn me_config_updater(pool: Arc<MePool>, rng: Arc<SecureRandom>, interv
             } else {
                 debug!("ME config v4 unchanged");
             }
-        } else {
-            warn!("getProxyConfig update failed");
         }
 
         // Update proxy config v6 (optional)
-        if let Ok(cfg_v6) = fetch_proxy_config("https://core.telegram.org/getProxyConfigV6").await {
-            let _ = pool.update_proxy_maps(HashMap::new(), Some(cfg_v6.map)).await;
+        let cfg_v6 = retry_fetch("https://core.telegram.org/getProxyConfigV6").await;
+        if let Some(cfg_v6) = cfg_v6 {
+            let changed = pool.update_proxy_maps(HashMap::new(), Some(cfg_v6.map)).await;
+            if changed {
+                info!("ME config updated (v6), reconciling connections");
+                pool.reconcile_connections(&rng).await;
+            } else {
+                debug!("ME config v6 unchanged");
+            }
         }
+        pool.reset_stun_state();
 
         // Update proxy-secret
         match download_proxy_secret().await {

src/transport/middle_proxy/handshake.rs

@@ -18,13 +18,14 @@ use crate::crypto::{SecureRandom, build_middleproxy_prekey, derive_middleproxy_k
 use crate::error::{ProxyError, Result};
 use crate::network::IpFamily;
 use crate::protocol::constants::{
-    ME_CONNECT_TIMEOUT_SECS, ME_HANDSHAKE_TIMEOUT_SECS, RPC_CRYPTO_AES_U32, RPC_HANDSHAKE_ERROR_U32,
-    RPC_HANDSHAKE_U32, RPC_PING_U32, RPC_PONG_U32, RPC_NONCE_U32,
+    ME_CONNECT_TIMEOUT_SECS, ME_HANDSHAKE_TIMEOUT_SECS, RPC_CRYPTO_AES_U32,
+    RPC_HANDSHAKE_ERROR_U32, rpc_crypto_flags,
 };
 
 use super::codec::{
-    build_handshake_payload, build_nonce_payload, build_rpc_frame, cbc_decrypt_inplace,
-    cbc_encrypt_padded, parse_nonce_payload, read_rpc_frame_plaintext,
+    RpcChecksumMode, build_handshake_payload, build_nonce_payload, build_rpc_frame,
+    cbc_decrypt_inplace, cbc_encrypt_padded, parse_handshake_flags, parse_nonce_payload,
+    read_rpc_frame_plaintext, rpc_crc,
 };
 use super::wire::{extract_ip_material, IpMaterial};
 use super::MePool;
@@ -37,6 +38,7 @@ pub(crate) struct HandshakeOutput {
     pub read_iv: [u8; 16],
     pub write_key: [u8; 32],
     pub write_iv: [u8; 16],
+    pub crc_mode: RpcChecksumMode,
     pub handshake_ms: f64,
 }
 
@@ -146,7 +148,7 @@ impl MePool {
 
         let ks = self.key_selector().await;
         let nonce_payload = build_nonce_payload(ks, crypto_ts, &my_nonce);
-        let nonce_frame = build_rpc_frame(-2, &nonce_payload);
+        let nonce_frame = build_rpc_frame(-2, &nonce_payload, RpcChecksumMode::Crc32);
         let dump = hex_dump(&nonce_frame[..nonce_frame.len().min(44)]);
         debug!(
             key_selector = format_args!("0x{ks:08x}"),
@@ -284,8 +286,15 @@ impl MePool {
             srv_v6_opt.as_ref(),
         );
 
-        let hs_payload = build_handshake_payload(hs_our_ip, local_addr.port(), hs_peer_ip, peer_addr.port());
-        let hs_frame = build_rpc_frame(-1, &hs_payload);
+        let requested_crc_mode = RpcChecksumMode::Crc32c;
+        let hs_payload = build_handshake_payload(
+            hs_our_ip,
+            local_addr.port(),
+            hs_peer_ip,
+            peer_addr.port(),
+            requested_crc_mode.advertised_flags(),
+        );
+        let hs_frame = build_rpc_frame(-1, &hs_payload, RpcChecksumMode::Crc32);
         if diag_level >= 1 {
             info!(
                 write_key = %hex_dump(&wk),
@@ -314,7 +323,7 @@ impl MePool {
             );
         }
 
-        let (encrypted_hs, mut write_iv) = cbc_encrypt_padded(&wk, &wi, &hs_frame)?;
+        let (encrypted_hs, write_iv) = cbc_encrypt_padded(&wk, &wi, &hs_frame)?;
         if diag_level >= 1 {
             info!(
                 hs_cipher = %hex_dump(&encrypted_hs),
@@ -328,6 +337,7 @@ impl MePool {
         let mut enc_buf = BytesMut::with_capacity(256);
         let mut dec_buf = BytesMut::with_capacity(256);
         let mut read_iv = ri;
+        let mut negotiated_crc_mode = RpcChecksumMode::Crc32;
         let mut handshake_ok = false;
 
         while Instant::now() < deadline && !handshake_ok {
@@ -375,17 +385,23 @@ impl MePool {
                 let frame = dec_buf.split_to(fl);
                 let pe = fl - 4;
                 let ec = u32::from_le_bytes(frame[pe..pe + 4].try_into().unwrap());
-                let ac = crate::crypto::crc32(&frame[..pe]);
+                let ac = rpc_crc(RpcChecksumMode::Crc32, &frame[..pe]);
                 if ec != ac {
                     return Err(ProxyError::InvalidHandshake(format!(
                         "HS CRC mismatch: 0x{ec:08x} vs 0x{ac:08x}"
                     )));
                 }
 
-                let hs_type = u32::from_le_bytes(frame[8..12].try_into().unwrap());
+                let hs_payload = &frame[8..pe];
+                if hs_payload.len() < 4 {
+                    return Err(ProxyError::InvalidHandshake(
+                        "Handshake payload too short".to_string(),
+                    ));
+                }
+                let hs_type = u32::from_le_bytes(hs_payload[0..4].try_into().unwrap());
                 if hs_type == RPC_HANDSHAKE_ERROR_U32 {
-                    let err_code = if frame.len() >= 16 {
-                        i32::from_le_bytes(frame[12..16].try_into().unwrap())
+                    let err_code = if hs_payload.len() >= 8 {
+                        i32::from_le_bytes(hs_payload[4..8].try_into().unwrap())
                     } else {
                         -1
                     };
@@ -393,11 +409,21 @@ impl MePool {
                         "ME rejected handshake (error={err_code})"
                     )));
                 }
-                if hs_type != RPC_HANDSHAKE_U32 {
+                let hs_flags = parse_handshake_flags(hs_payload)?;
+                if hs_flags & 0xff != 0 {
                     return Err(ProxyError::InvalidHandshake(format!(
-                        "Expected HANDSHAKE 0x{RPC_HANDSHAKE_U32:08x}, got 0x{hs_type:08x}"
+                        "Unsupported handshake flags: 0x{hs_flags:08x}"
                     )));
                 }
+                negotiated_crc_mode = if (hs_flags & requested_crc_mode.advertised_flags()) != 0 {
+                    RpcChecksumMode::from_handshake_flags(hs_flags)
+                } else if (hs_flags & rpc_crypto_flags::USE_CRC32C) != 0 {
+                    return Err(ProxyError::InvalidHandshake(format!(
+                        "Peer negotiated unsupported CRC flags: 0x{hs_flags:08x}"
+                    )));
+                } else {
+                    RpcChecksumMode::Crc32
+                };
 
                 handshake_ok = true;
                 break;
@@ -418,6 +444,7 @@ impl MePool {
             read_iv,
             write_key: wk,
             write_iv,
+            crc_mode: negotiated_crc_mode,
             handshake_ms,
         })
     }

src/transport/middle_proxy/health.rs

@@ -5,25 +5,30 @@ use std::time::{Duration, Instant};
 
 use tracing::{debug, info, warn};
 use rand::seq::SliceRandom;
+use rand::Rng;
 
 use crate::crypto::SecureRandom;
 use crate::network::IpFamily;
 
 use super::MePool;
 
+const HEALTH_INTERVAL_SECS: u64 = 1;
+const JITTER_FRAC_NUM: u64 = 2; // jitter up to 50% of backoff
+const MAX_CONCURRENT_PER_DC_DEFAULT: usize = 1;
+
 pub async fn me_health_monitor(pool: Arc<MePool>, rng: Arc<SecureRandom>, _min_connections: usize) {
     let mut backoff: HashMap<(i32, IpFamily), u64> = HashMap::new();
-    let mut last_attempt: HashMap<(i32, IpFamily), Instant> = HashMap::new();
-    let mut inflight_single: HashSet<(i32, IpFamily)> = HashSet::new();
+    let mut next_attempt: HashMap<(i32, IpFamily), Instant> = HashMap::new();
+    let mut inflight: HashMap<(i32, IpFamily), usize> = HashMap::new();
     loop {
-        tokio::time::sleep(Duration::from_secs(30)).await;
+        tokio::time::sleep(Duration::from_secs(HEALTH_INTERVAL_SECS)).await;
         check_family(
             IpFamily::V4,
             &pool,
             &rng,
             &mut backoff,
-            &mut last_attempt,
-            &mut inflight_single,
+            &mut next_attempt,
+            &mut inflight,
         )
         .await;
         check_family(
@@ -31,8 +36,8 @@ pub async fn me_health_monitor(pool: Arc<MePool>, rng: Arc<SecureRandom>, _min_c
             &pool,
             &rng,
             &mut backoff,
-            &mut last_attempt,
-            &mut inflight_single,
+            &mut next_attempt,
+            &mut inflight,
         )
         .await;
     }
@@ -43,8 +48,8 @@ async fn check_family(
     pool: &Arc<MePool>,
     rng: &Arc<SecureRandom>,
     backoff: &mut HashMap<(i32, IpFamily), u64>,
-    last_attempt: &mut HashMap<(i32, IpFamily), Instant>,
-    inflight_single: &mut HashSet<(i32, IpFamily)>,
+    next_attempt: &mut HashMap<(i32, IpFamily), Instant>,
+    inflight: &mut HashMap<(i32, IpFamily), usize>,
 ) {
     let enabled = match family {
         IpFamily::V4 => pool.decision.ipv4_me,
@@ -80,95 +85,60 @@ async fn check_family(
     for (dc, dc_addrs) in entries {
         let has_coverage = dc_addrs.iter().any(|a| writer_addrs.contains(a));
         if has_coverage {
-            inflight_single.remove(&(dc, family));
-            continue;
-        }
-        let key = (dc, family);
-        let delay = *backoff.get(&key).unwrap_or(&30);
-        let now = Instant::now();
-        if let Some(last) = last_attempt.get(&key) {
-            if now.duration_since(*last).as_secs() < delay {
-                continue;
-            }
-        }
-        if dc_addrs.len() == 1 {
-            // Single ME address: fast retries then slower background retries.
-            if inflight_single.contains(&key) {
-                continue;
-            }
-            inflight_single.insert(key);
-            let addr = dc_addrs[0];
-            let dc_id = dc;
-            let pool_clone = pool.clone();
-            let rng_clone = rng.clone();
-            let timeout = pool.me_one_timeout;
-            let quick_attempts = pool.me_one_retry.max(1);
-            tokio::spawn(async move {
-                let mut success = false;
-                for _ in 0..quick_attempts {
-                    let res = tokio::time::timeout(timeout, pool_clone.connect_one(addr, rng_clone.as_ref())).await;
-                    match res {
-                        Ok(Ok(())) => {
-                            info!(%addr, dc = %dc_id, ?family, "ME reconnected for DC coverage");
-                            success = true;
-                            break;
-                        }
-                        Ok(Err(e)) => debug!(%addr, dc = %dc_id, error = %e, ?family, "ME reconnect failed"),
-                        Err(_) => debug!(%addr, dc = %dc_id, ?family, "ME reconnect timed out"),
-                    }
-                    tokio::time::sleep(Duration::from_millis(1000)).await;
-                }
-                if success {
-                    return;
-                }
-                let timeout_ms = timeout.as_millis();
-                warn!(
-                    dc = %dc_id,
-                    ?family,
-                    attempts = quick_attempts,
-                    timeout_ms,
-                    "DC={} has no ME coverage: {} tries * {} ms... retry in 5 seconds...",
-                    dc_id,
-                    quick_attempts,
-                    timeout_ms
-                );
-                loop {
-                    tokio::time::sleep(Duration::from_secs(5)).await;
-                    let res = tokio::time::timeout(timeout, pool_clone.connect_one(addr, rng_clone.as_ref())).await;
-                    match res {
-                        Ok(Ok(())) => {
-                            info!(%addr, dc = %dc_id, ?family, "ME reconnected for DC coverage");
-                            break;
-                        }
-                        Ok(Err(e)) => debug!(%addr, dc = %dc_id, error = %e, ?family, "ME reconnect failed"),
-                        Err(_) => debug!(%addr, dc = %dc_id, ?family, "ME reconnect timed out"),
-                    }
-                }
-                // will drop inflight flag in outer loop when coverage detected
-            });
             continue;
         }
 
-        warn!(dc = %dc, delay, ?family, "DC has no ME coverage, reconnecting...");
-        let mut shuffled = dc_addrs.clone();
-        shuffled.shuffle(&mut rand::rng());
-        let mut reconnected = false;
-        for addr in shuffled {
-            match pool.connect_one(addr, rng.as_ref()).await {
-                Ok(()) => {
-                    info!(%addr, dc = %dc, ?family, "ME reconnected for DC coverage");
-                    backoff.insert(key, 30);
-                    last_attempt.insert(key, now);
-                    reconnected = true;
-                    break;
-                }
-                Err(e) => debug!(%addr, dc = %dc, error = %e, ?family, "ME reconnect failed"),
+        let key = (dc, family);
+        let now = Instant::now();
+        if let Some(ts) = next_attempt.get(&key) {
+            if now < *ts {
+                continue;
             }
         }
-        if !reconnected {
-            let next = (*backoff.get(&key).unwrap_or(&30)).saturating_mul(2).min(300);
-            backoff.insert(key, next);
-            last_attempt.insert(key, now);
+
+        let max_concurrent = pool.me_reconnect_max_concurrent_per_dc.max(1) as usize;
+        if *inflight.get(&key).unwrap_or(&0) >= max_concurrent {
+            return;
+        }
+        *inflight.entry(key).or_insert(0) += 1;
+
+        let mut shuffled = dc_addrs.clone();
+        shuffled.shuffle(&mut rand::rng());
+        let mut success = false;
+        for addr in shuffled {
+            let res = tokio::time::timeout(pool.me_one_timeout, pool.connect_one(addr, rng.as_ref())).await;
+            match res {
+                Ok(Ok(())) => {
+                    info!(%addr, dc = %dc, ?family, "ME reconnected for DC coverage");
+                    pool.stats.increment_me_reconnect_success();
+                    backoff.insert(key, pool.me_reconnect_backoff_base.as_millis() as u64);
+                    let jitter = pool.me_reconnect_backoff_base.as_millis() as u64 / JITTER_FRAC_NUM;
+                    let wait = pool.me_reconnect_backoff_base
+                        + Duration::from_millis(rand::rng().random_range(0..=jitter.max(1)));
+                    next_attempt.insert(key, now + wait);
+                    success = true;
+                    break;
+                }
+                Ok(Err(e)) => {
+                    pool.stats.increment_me_reconnect_attempt();
+                    debug!(%addr, dc = %dc, error = %e, ?family, "ME reconnect failed")
+                }
+                Err(_) => debug!(%addr, dc = %dc, ?family, "ME reconnect timed out"),
+            }
+        }
+        if !success {
+            pool.stats.increment_me_reconnect_attempt();
+            let curr = *backoff.get(&key).unwrap_or(&(pool.me_reconnect_backoff_base.as_millis() as u64));
+            let next_ms = (curr.saturating_mul(2)).min(pool.me_reconnect_backoff_cap.as_millis() as u64);
+            backoff.insert(key, next_ms);
+            let jitter = next_ms / JITTER_FRAC_NUM;
+            let wait = Duration::from_millis(next_ms)
+                + Duration::from_millis(rand::rng().random_range(0..=jitter.max(1)));
+            next_attempt.insert(key, now + wait);
+            warn!(dc = %dc, backoff_ms = next_ms, ?family, "DC has no ME coverage, scheduled reconnect");
+        }
+        if let Some(v) = inflight.get_mut(&key) {
+            *v = v.saturating_sub(1);
         }
     }
 }

src/transport/middle_proxy/pool.rs

@@ -1,14 +1,14 @@
 use std::collections::HashMap;
 use std::net::{IpAddr, Ipv6Addr, SocketAddr};
 use std::sync::Arc;
-use std::sync::atomic::{AtomicBool, AtomicI32, AtomicU64, Ordering};
+use std::sync::atomic::{AtomicBool, AtomicI32, AtomicU64, AtomicUsize, Ordering};
 use bytes::BytesMut;
 use rand::Rng;
 use rand::seq::SliceRandom;
-use tokio::sync::{Mutex, RwLock};
+use tokio::sync::{Mutex, RwLock, mpsc, Notify};
 use tokio_util::sync::CancellationToken;
 use tracing::{debug, info, warn};
-use std::time::Duration;
+use std::time::{Duration, Instant};
 
 use crate::crypto::SecureRandom;
 use crate::error::{ProxyError, Result};
@@ -17,11 +17,9 @@ use crate::network::IpFamily;
 use crate::protocol::constants::*;
 
 use super::ConnRegistry;
-use super::registry::{BoundConn, ConnMeta};
-use super::codec::RpcWriter;
+use super::registry::BoundConn;
+use super::codec::{RpcWriter, WriterCommand};
 use super::reader::reader_loop;
-use super::MeResponse;
-
 const ME_ACTIVE_PING_SECS: u64 = 25;
 const ME_ACTIVE_PING_JITTER_SECS: i64 = 5;
 
@@ -29,7 +27,7 @@ const ME_ACTIVE_PING_JITTER_SECS: i64 = 5;
 pub struct MeWriter {
     pub id: u64,
     pub addr: SocketAddr,
-    pub writer: Arc<Mutex<RpcWriter>>,
+    pub tx: mpsc::Sender<WriterCommand>,
     pub cancel: CancellationToken,
     pub degraded: Arc<AtomicBool>,
     pub draining: Arc<AtomicBool>,
@@ -47,11 +45,24 @@ pub struct MePool {
     pub(super) nat_ip_detected: Arc<RwLock<Option<IpAddr>>>,
     pub(super) nat_probe: bool,
     pub(super) nat_stun: Option<String>,
+    pub(super) nat_stun_servers: Vec<String>,
     pub(super) detected_ipv6: Option<Ipv6Addr>,
     pub(super) nat_probe_attempts: std::sync::atomic::AtomicU8,
     pub(super) nat_probe_disabled: std::sync::atomic::AtomicBool,
+    pub(super) stun_backoff_until: Arc<RwLock<Option<Instant>>>,
     pub(super) me_one_retry: u8,
     pub(super) me_one_timeout: Duration,
+    pub(super) me_keepalive_enabled: bool,
+    pub(super) me_keepalive_interval: Duration,
+    pub(super) me_keepalive_jitter: Duration,
+    pub(super) me_keepalive_payload_random: bool,
+    pub(super) me_warmup_stagger_enabled: bool,
+    pub(super) me_warmup_step_delay: Duration,
+    pub(super) me_warmup_step_jitter: Duration,
+    pub(super) me_reconnect_max_concurrent_per_dc: u32,
+    pub(super) me_reconnect_backoff_base: Duration,
+    pub(super) me_reconnect_backoff_cap: Duration,
+    pub(super) me_reconnect_fast_retry_count: u32,
     pub(super) proxy_map_v4: Arc<RwLock<HashMap<i32, Vec<(IpAddr, u16)>>>>,
     pub(super) proxy_map_v6: Arc<RwLock<HashMap<i32, Vec<(IpAddr, u16)>>>>,
     pub(super) default_dc: AtomicI32,
@@ -59,6 +70,9 @@ pub struct MePool {
     pub(super) ping_tracker: Arc<Mutex<HashMap<i64, (std::time::Instant, u64)>>>,
     pub(super) rtt_stats: Arc<Mutex<HashMap<u64, (f64, f64)>>>,
     pub(super) nat_reflection_cache: Arc<Mutex<NatReflectionCache>>,
+    pub(super) writer_available: Arc<Notify>,
+    pub(super) conn_count: AtomicUsize,
+    pub(super) stats: Arc<crate::stats::Stats>,
     pool_size: usize,
 }
 
@@ -75,6 +89,7 @@ impl MePool {
         nat_ip: Option<IpAddr>,
         nat_probe: bool,
         nat_stun: Option<String>,
+        nat_stun_servers: Vec<String>,
         detected_ipv6: Option<Ipv6Addr>,
         me_one_retry: u8,
         me_one_timeout_ms: u64,
@@ -83,6 +98,18 @@ impl MePool {
         default_dc: Option<i32>,
         decision: NetworkDecision,
         rng: Arc<SecureRandom>,
+        stats: Arc<crate::stats::Stats>,
+        me_keepalive_enabled: bool,
+        me_keepalive_interval_secs: u64,
+        me_keepalive_jitter_secs: u64,
+        me_keepalive_payload_random: bool,
+        me_warmup_stagger_enabled: bool,
+        me_warmup_step_delay_ms: u64,
+        me_warmup_step_jitter_ms: u64,
+        me_reconnect_max_concurrent_per_dc: u32,
+        me_reconnect_backoff_base_ms: u64,
+        me_reconnect_backoff_cap_ms: u64,
+        me_reconnect_fast_retry_count: u32,
     ) -> Arc<Self> {
         Arc::new(Self {
             registry: Arc::new(ConnRegistry::new()),
@@ -96,11 +123,25 @@ impl MePool {
             nat_ip_detected: Arc::new(RwLock::new(None)),
             nat_probe,
             nat_stun,
+            nat_stun_servers,
             detected_ipv6,
             nat_probe_attempts: std::sync::atomic::AtomicU8::new(0),
             nat_probe_disabled: std::sync::atomic::AtomicBool::new(false),
+            stun_backoff_until: Arc::new(RwLock::new(None)),
             me_one_retry,
             me_one_timeout: Duration::from_millis(me_one_timeout_ms),
+            stats,
+            me_keepalive_enabled,
+            me_keepalive_interval: Duration::from_secs(me_keepalive_interval_secs),
+            me_keepalive_jitter: Duration::from_secs(me_keepalive_jitter_secs),
+            me_keepalive_payload_random,
+            me_warmup_stagger_enabled,
+            me_warmup_step_delay: Duration::from_millis(me_warmup_step_delay_ms),
+            me_warmup_step_jitter: Duration::from_millis(me_warmup_step_jitter_ms),
+            me_reconnect_max_concurrent_per_dc,
+            me_reconnect_backoff_base: Duration::from_millis(me_reconnect_backoff_base_ms),
+            me_reconnect_backoff_cap: Duration::from_millis(me_reconnect_backoff_cap_ms),
+            me_reconnect_fast_retry_count,
             pool_size: 2,
             proxy_map_v4: Arc::new(RwLock::new(proxy_map_v4)),
             proxy_map_v6: Arc::new(RwLock::new(proxy_map_v6)),
@@ -109,6 +150,8 @@ impl MePool {
             ping_tracker: Arc::new(Mutex::new(HashMap::new())),
             rtt_stats: Arc::new(Mutex::new(HashMap::new())),
             nat_reflection_cache: Arc::new(Mutex::new(NatReflectionCache::default())),
+            writer_available: Arc::new(Notify::new()),
+            conn_count: AtomicUsize::new(0),
         })
     }
 
@@ -116,6 +159,11 @@ impl MePool {
         self.proxy_tag.is_some()
     }
 
+    pub fn reset_stun_state(&self) {
+        self.nat_probe_attempts.store(0, Ordering::Relaxed);
+        self.nat_probe_disabled.store(false, Ordering::Relaxed);
+    }
+
     pub fn translate_our_addr(&self, addr: SocketAddr) -> SocketAddr {
         let ip = self.translate_ip_for_nat(addr.ip());
         SocketAddr::new(ip, addr.port())
@@ -132,7 +180,11 @@ impl MePool {
     pub async fn reconcile_connections(self: &Arc<Self>, rng: &SecureRandom) {
         use std::collections::HashSet;
         let writers = self.writers.read().await;
-        let current: HashSet<SocketAddr> = writers.iter().map(|w| w.addr).collect();
+        let current: HashSet<SocketAddr> = writers
+            .iter()
+            .filter(|w| !w.draining.load(Ordering::Relaxed))
+            .map(|w| w.addr)
+            .collect();
         drop(writers);
 
         for family in self.family_order() {
@@ -175,12 +227,36 @@ impl MePool {
             let mut guard = self.proxy_map_v6.write().await;
             if !v6.is_empty() && *guard != v6 {
                 *guard = v6;
+                changed = true;
+            }
+        }
+        // Ensure negative DC entries mirror positives when absent (Telegram convention).
+        {
+            let mut guard = self.proxy_map_v4.write().await;
+            let keys: Vec<i32> = guard.keys().cloned().collect();
+            for k in keys.iter().cloned().filter(|k| *k > 0) {
+                if !guard.contains_key(&-k) {
+                    if let Some(addrs) = guard.get(&k).cloned() {
+                        guard.insert(-k, addrs);
+                    }
+                }
+            }
+        }
+        {
+            let mut guard = self.proxy_map_v6.write().await;
+            let keys: Vec<i32> = guard.keys().cloned().collect();
+            for k in keys.iter().cloned().filter(|k| *k > 0) {
+                if !guard.contains_key(&-k) {
+                    if let Some(addrs) = guard.get(&k).cloned() {
+                        guard.insert(-k, addrs);
+                    }
+                }
             }
         }
         changed
     }
 
-    pub async fn update_secret(&self, new_secret: Vec<u8>) -> bool {
+    pub async fn update_secret(self: &Arc<Self>, new_secret: Vec<u8>) -> bool {
         if new_secret.len() < 32 {
             warn!(len = new_secret.len(), "proxy-secret update ignored (too short)");
             return false;
@@ -195,10 +271,14 @@ impl MePool {
         false
     }
 
-    pub async fn reconnect_all(&self) {
-        // Graceful: do not drop all at once. New connections will use updated secret.
-        // Existing writers remain until health monitor replaces them.
-        // No-op here to avoid total outage.
+    pub async fn reconnect_all(self: &Arc<Self>) {
+        let ws = self.writers.read().await.clone();
+        for w in ws {
+            if let Ok(()) = self.connect_one(w.addr, self.rng.as_ref()).await {
+                self.mark_writer_draining(w.id).await;
+                tokio::time::sleep(Duration::from_secs(2)).await;
+            }
+        }
     }
 
     pub(super) async fn key_selector(&self) -> u32 {
@@ -277,7 +357,23 @@ impl MePool {
                 return Err(ProxyError::Proxy("Too many ME DC init failures, falling back to direct".into()));
             }
 
-            // Additional connections up to pool_size total (round-robin across DCs)
+        // Additional connections up to pool_size total (round-robin across DCs), staggered to de-phase lifecycles.
+        if self.me_warmup_stagger_enabled {
+            for (dc, addrs) in dc_addrs.iter() {
+                for (ip, port) in addrs {
+                    if self.connection_count() >= pool_size {
+                        break;
+                    }
+                    let addr = SocketAddr::new(*ip, *port);
+                    let jitter = rand::rng().random_range(0..=self.me_warmup_step_jitter.as_millis() as u64);
+                    let delay_ms = self.me_warmup_step_delay.as_millis() as u64 + jitter;
+                    tokio::time::sleep(Duration::from_millis(delay_ms)).await;
+                    if let Err(e) = self.connect_one(addr, rng.as_ref()).await {
+                        debug!(%addr, dc = %dc, error = %e, "Extra ME connect failed (staggered)");
+                    }
+                }
+            }
+        } else {
             for (dc, addrs) in dc_addrs.iter() {
                 for (ip, port) in addrs {
                     if self.connection_count() >= pool_size {
@@ -292,6 +388,7 @@ impl MePool {
                     break;
                 }
             }
+        }
 
             if !self.decision.effective_multipath && self.connection_count() > 0 {
                 break;
@@ -317,50 +414,93 @@ impl MePool {
         let cancel = CancellationToken::new();
         let degraded = Arc::new(AtomicBool::new(false));
         let draining = Arc::new(AtomicBool::new(false));
-        let rpc_w = Arc::new(Mutex::new(RpcWriter {
+        let (tx, mut rx) = mpsc::channel::<WriterCommand>(4096);
+        let tx_for_keepalive = tx.clone();
+        let mut rpc_writer = RpcWriter {
             writer: hs.wr,
             key: hs.write_key,
             iv: hs.write_iv,
             seq_no: 0,
-        }));
+            crc_mode: hs.crc_mode,
+        };
+        let cancel_wr = cancel.clone();
+        tokio::spawn(async move {
+            loop {
+                tokio::select! {
+                    cmd = rx.recv() => {
+                        match cmd {
+                            Some(WriterCommand::Data(payload)) => {
+                                if rpc_writer.send(&payload).await.is_err() { break; }
+                            }
+                            Some(WriterCommand::DataAndFlush(payload)) => {
+                                if rpc_writer.send_and_flush(&payload).await.is_err() { break; }
+                            }
+                            Some(WriterCommand::Close) | None => break,
+                        }
+                    }
+                    _ = cancel_wr.cancelled() => break,
+                }
+            }
+        });
         let writer = MeWriter {
             id: writer_id,
             addr,
-            writer: rpc_w.clone(),
+            tx: tx.clone(),
             cancel: cancel.clone(),
             degraded: degraded.clone(),
             draining: draining.clone(),
         };
         self.writers.write().await.push(writer.clone());
+        self.conn_count.fetch_add(1, Ordering::Relaxed);
+        self.writer_available.notify_waiters();
 
         let reg = self.registry.clone();
         let writers_arc = self.writers_arc();
         let ping_tracker = self.ping_tracker.clone();
+        let ping_tracker_reader = ping_tracker.clone();
         let rtt_stats = self.rtt_stats.clone();
+        let stats_reader = self.stats.clone();
+        let stats_ping = self.stats.clone();
+        let stats_keepalive = self.stats.clone();
         let pool = Arc::downgrade(self);
         let cancel_ping = cancel.clone();
-        let rpc_w_ping = rpc_w.clone();
+        let tx_ping = tx.clone();
         let ping_tracker_ping = ping_tracker.clone();
+        let cleanup_done = Arc::new(AtomicBool::new(false));
+        let cleanup_for_reader = cleanup_done.clone();
+        let cleanup_for_ping = cleanup_done.clone();
+        let keepalive_enabled = self.me_keepalive_enabled;
+        let keepalive_interval = self.me_keepalive_interval;
+        let keepalive_jitter = self.me_keepalive_jitter;
+        let cancel_reader_token = cancel.clone();
+        let cancel_ping_token = cancel_ping.clone();
+        let cancel_keepalive_token = cancel.clone();
 
         tokio::spawn(async move {
-            let cancel_reader = cancel.clone();
             let res = reader_loop(
                 hs.rd,
                 hs.read_key,
                 hs.read_iv,
+                hs.crc_mode,
                 reg.clone(),
                 BytesMut::new(),
                 BytesMut::new(),
-                rpc_w.clone(),
-                ping_tracker.clone(),
+                tx.clone(),
+                ping_tracker_reader,
                 rtt_stats.clone(),
+                stats_reader,
                 writer_id,
                 degraded.clone(),
-                cancel_reader.clone(),
+                cancel_reader_token.clone(),
             )
             .await;
             if let Some(pool) = pool.upgrade() {
-                pool.remove_writer_and_close_clients(writer_id).await;
+                if cleanup_for_reader
+                    .compare_exchange(false, true, Ordering::AcqRel, Ordering::Relaxed)
+                    .is_ok()
+                {
+                    pool.remove_writer_and_close_clients(writer_id).await;
+                }
             }
             if let Err(e) = res {
                 warn!(error = %e, "ME reader ended");
@@ -378,7 +518,7 @@ impl MePool {
                     .random_range(-ME_ACTIVE_PING_JITTER_SECS..=ME_ACTIVE_PING_JITTER_SECS);
                 let wait = (ME_ACTIVE_PING_SECS as i64 + jitter).max(5) as u64;
                 tokio::select! {
-                    _ = cancel_ping.cancelled() => {
+                    _ = cancel_ping_token.cancelled() => {
                         break;
                     }
                     _ = tokio::time::sleep(Duration::from_secs(wait)) => {}
@@ -389,20 +529,71 @@ impl MePool {
                 p.extend_from_slice(&sent_id.to_le_bytes());
                 {
                     let mut tracker = ping_tracker_ping.lock().await;
+                    let before = tracker.len();
+                    tracker.retain(|_, (ts, _)| ts.elapsed() < Duration::from_secs(120));
+                    let expired = before.saturating_sub(tracker.len());
+                    if expired > 0 {
+                        stats_ping.increment_me_keepalive_timeout_by(expired as u64);
+                    }
                     tracker.insert(sent_id, (std::time::Instant::now(), writer_id));
                 }
                 ping_id = ping_id.wrapping_add(1);
-                if let Err(e) = rpc_w_ping.lock().await.send_and_flush(&p).await {
-                    debug!(error = %e, "Active ME ping failed, removing dead writer");
+                if tx_ping.send(WriterCommand::DataAndFlush(p)).await.is_err() {
+                    debug!("Active ME ping failed, removing dead writer");
                     cancel_ping.cancel();
                     if let Some(pool) = pool_ping.upgrade() {
-                        pool.remove_writer_and_close_clients(writer_id).await;
+                        if cleanup_for_ping
+                            .compare_exchange(false, true, Ordering::AcqRel, Ordering::Relaxed)
+                            .is_ok()
+                        {
+                            pool.remove_writer_and_close_clients(writer_id).await;
+                        }
                     }
                     break;
                 }
             }
         });
 
+        if keepalive_enabled {
+            let tx_keepalive = tx_for_keepalive;
+            let cancel_keepalive = cancel_keepalive_token;
+            let ping_tracker_keepalive = ping_tracker.clone();
+            tokio::spawn(async move {
+                // Per-writer jittered start to avoid phase sync.
+                let jitter_cap_ms = keepalive_interval.as_millis() / 2;
+                let effective_jitter_ms = keepalive_jitter.as_millis().min(jitter_cap_ms).max(1);
+                let initial_jitter_ms = rand::rng().random_range(0..=effective_jitter_ms as u64);
+                tokio::time::sleep(Duration::from_millis(initial_jitter_ms)).await;
+                let mut ping_id: i64 = rand::random::<i64>();
+                loop {
+                    tokio::select! {
+                        _ = cancel_keepalive.cancelled() => break,
+                        _ = tokio::time::sleep(keepalive_interval + Duration::from_millis(rand::rng().random_range(0..=effective_jitter_ms as u64))) => {}
+                    }
+                    let sent_id = ping_id;
+                    ping_id = ping_id.wrapping_add(1);
+                    let mut p = Vec::with_capacity(12);
+                    p.extend_from_slice(&RPC_PING_U32.to_le_bytes());
+                    p.extend_from_slice(&sent_id.to_le_bytes());
+                    {
+                        let mut tracker = ping_tracker_keepalive.lock().await;
+                        let before = tracker.len();
+                        tracker.retain(|_, (ts, _)| ts.elapsed() < Duration::from_secs(120));
+                        let expired = before.saturating_sub(tracker.len());
+                        if expired > 0 {
+                            stats_keepalive.increment_me_keepalive_timeout_by(expired as u64);
+                        }
+                        tracker.insert(sent_id, (std::time::Instant::now(), writer_id));
+                    }
+                    stats_keepalive.increment_me_keepalive_sent();
+                    if tx_keepalive.send(WriterCommand::DataAndFlush(p)).await.is_err() {
+                        stats_keepalive.increment_me_keepalive_failed();
+                        break;
+                    }
+                }
+            });
+        }
+
         Ok(())
     }
 
@@ -430,7 +621,7 @@ impl MePool {
         false
     }
 
-    pub(crate) async fn remove_writer_and_close_clients(&self, writer_id: u64) {
+    pub(crate) async fn remove_writer_and_close_clients(self: &Arc<Self>, writer_id: u64) {
         let conns = self.remove_writer_only(writer_id).await;
         for bound in conns {
             let _ = self.registry.route(bound.conn_id, super::MeResponse::Close).await;
@@ -444,8 +635,11 @@ impl MePool {
             if let Some(pos) = ws.iter().position(|w| w.id == writer_id) {
                 let w = ws.remove(pos);
                 w.cancel.cancel();
+                let _ = w.tx.send(WriterCommand::Close).await;
+                self.conn_count.fetch_sub(1, Ordering::Relaxed);
             }
         }
+        self.rtt_stats.lock().await.remove(&writer_id);
         self.registry.writer_lost(writer_id).await
     }
 
@@ -459,8 +653,14 @@ impl MePool {
 
         let pool = Arc::downgrade(self);
         tokio::spawn(async move {
+            let deadline = Instant::now() + Duration::from_secs(300);
             loop {
                 if let Some(p) = pool.upgrade() {
+                    if Instant::now() >= deadline {
+                        warn!(writer_id, "Drain timeout, force-closing");
+                        let _ = p.remove_writer_and_close_clients(writer_id).await;
+                        break;
+                    }
                     if p.registry.is_writer_empty(writer_id).await {
                         let _ = p.remove_writer_only(writer_id).await;
                         break;

src/transport/middle_proxy/pool_nat.rs

@@ -98,16 +98,18 @@ impl MePool {
         family: IpFamily,
     ) -> Option<std::net::SocketAddr> {
         const STUN_CACHE_TTL: Duration = Duration::from_secs(600);
-        // If STUN probing was disabled after attempts, reuse cached (even stale) or skip.
-        if self.nat_probe_disabled.load(std::sync::atomic::Ordering::Relaxed) {
-            if let Ok(cache) = self.nat_reflection_cache.try_lock() {
-                let slot = match family {
-                    IpFamily::V4 => cache.v4,
-                    IpFamily::V6 => cache.v6,
-                };
-                return slot.map(|(_, addr)| addr);
+        // Backoff window
+        if let Some(until) = *self.stun_backoff_until.read().await {
+            if Instant::now() < until {
+                if let Ok(cache) = self.nat_reflection_cache.try_lock() {
+                    let slot = match family {
+                        IpFamily::V4 => cache.v4,
+                        IpFamily::V6 => cache.v6,
+                    };
+                    return slot.map(|(_, addr)| addr);
+                }
+                return None;
             }
-            return None;
         }
 
         if let Ok(mut cache) = self.nat_reflection_cache.try_lock() {
@@ -123,48 +125,42 @@ impl MePool {
         }
 
         let attempt = self.nat_probe_attempts.fetch_add(1, std::sync::atomic::Ordering::Relaxed);
-        if attempt >= 2 {
-            self.nat_probe_disabled.store(true, std::sync::atomic::Ordering::Relaxed);
-            return None;
-        }
+        let servers = if !self.nat_stun_servers.is_empty() {
+            self.nat_stun_servers.clone()
+        } else if let Some(s) = &self.nat_stun {
+            vec![s.clone()]
+        } else {
+            vec!["stun.l.google.com:19302".to_string()]
+        };
 
-        let stun_addr = self
-            .nat_stun
-            .clone()
-            .unwrap_or_else(|| "stun.l.google.com:19302".to_string());
-        match stun_probe_dual(&stun_addr).await {
-            Ok(res) => {
-                let picked: Option<StunProbeResult> = match family {
-                    IpFamily::V4 => res.v4,
-                    IpFamily::V6 => res.v6,
-                };
-                if let Some(result) = picked {
-                    info!(local = %result.local_addr, reflected = %result.reflected_addr, family = ?family, "NAT probe: reflected address");
-                    if let Ok(mut cache) = self.nat_reflection_cache.try_lock() {
-                        let slot = match family {
-                            IpFamily::V4 => &mut cache.v4,
-                            IpFamily::V6 => &mut cache.v6,
-                        };
-                        *slot = Some((Instant::now(), result.reflected_addr));
+        for stun_addr in servers {
+            match stun_probe_dual(&stun_addr).await {
+                Ok(res) => {
+                    let picked: Option<StunProbeResult> = match family {
+                        IpFamily::V4 => res.v4,
+                        IpFamily::V6 => res.v6,
+                    };
+                    if let Some(result) = picked {
+                        info!(local = %result.local_addr, reflected = %result.reflected_addr, family = ?family, stun = %stun_addr, "NAT probe: reflected address");
+                        self.nat_probe_attempts.store(0, std::sync::atomic::Ordering::Relaxed);
+                        if let Ok(mut cache) = self.nat_reflection_cache.try_lock() {
+                            let slot = match family {
+                                IpFamily::V4 => &mut cache.v4,
+                                IpFamily::V6 => &mut cache.v6,
+                            };
+                            *slot = Some((Instant::now(), result.reflected_addr));
+                        }
+                        return Some(result.reflected_addr);
                     }
-                    Some(result.reflected_addr)
-                } else {
-                    None
                 }
-            }
-            Err(e) => {
-                let attempts = attempt + 1;
-                if attempts <= 2 {
-                    warn!(error = %e, attempt = attempts, "NAT probe failed");
-                } else {
-                    debug!(error = %e, attempt = attempts, "NAT probe suppressed after max attempts");
+                Err(e) => {
+                    warn!(error = %e, stun = %stun_addr, attempt = attempt + 1, "NAT probe failed, trying next server");
                 }
-                if attempts >= 2 {
-                    self.nat_probe_disabled.store(true, std::sync::atomic::Ordering::Relaxed);
-                }
-                None
             }
         }
+        let backoff = Duration::from_secs(60 * 2u64.pow((attempt as u32).min(6)));
+        *self.stun_backoff_until.write().await = Some(Instant::now() + backoff);
+        None
     }
 }
 

src/transport/middle_proxy/reader.rs

@@ -6,27 +6,31 @@ use std::time::Instant;
 use bytes::{Bytes, BytesMut};
 use tokio::io::AsyncReadExt;
 use tokio::net::TcpStream;
-use tokio::sync::Mutex;
+use tokio::sync::{Mutex, mpsc};
 use tokio_util::sync::CancellationToken;
 use tracing::{debug, trace, warn};
 
-use crate::crypto::{AesCbc, crc32};
+use crate::crypto::AesCbc;
 use crate::error::{ProxyError, Result};
 use crate::protocol::constants::*;
+use crate::stats::Stats;
 
-use super::codec::RpcWriter;
+use super::codec::{RpcChecksumMode, WriterCommand, rpc_crc};
+use super::registry::RouteResult;
 use super::{ConnRegistry, MeResponse};
 
 pub(crate) async fn reader_loop(
     mut rd: tokio::io::ReadHalf<TcpStream>,
     dk: [u8; 32],
     mut div: [u8; 16],
+    crc_mode: RpcChecksumMode,
     reg: Arc<ConnRegistry>,
     enc_leftover: BytesMut,
     mut dec: BytesMut,
-    writer: Arc<Mutex<RpcWriter>>,
+    tx: mpsc::Sender<WriterCommand>,
     ping_tracker: Arc<Mutex<HashMap<i64, (Instant, u64)>>>,
     rtt_stats: Arc<Mutex<HashMap<u64, (f64, f64)>>>,
+    stats: Arc<Stats>,
     _writer_id: u64,
     degraded: Arc<AtomicBool>,
     cancel: CancellationToken,
@@ -78,18 +82,28 @@ pub(crate) async fn reader_loop(
             let frame = dec.split_to(fl);
             let pe = fl - 4;
             let ec = u32::from_le_bytes(frame[pe..pe + 4].try_into().unwrap());
-            if crc32(&frame[..pe]) != ec {
-                warn!("CRC mismatch in data frame");
-                continue;
+            let actual_crc = rpc_crc(crc_mode, &frame[..pe]);
+            if actual_crc != ec {
+                stats.increment_me_crc_mismatch();
+                warn!(
+                    frame_len = fl,
+                    expected_crc = format_args!("0x{ec:08x}"),
+                    actual_crc = format_args!("0x{actual_crc:08x}"),
+                    "CRC mismatch — CBC crypto desync, aborting ME connection"
+                );
+                return Err(ProxyError::Proxy("CRC mismatch (crypto desync)".into()));
             }
 
             let seq_no = i32::from_le_bytes(frame[4..8].try_into().unwrap());
             if seq_no != expected_seq {
+                stats.increment_me_seq_mismatch();
                 warn!(seq_no, expected = expected_seq, "ME RPC seq mismatch");
-                expected_seq = seq_no.wrapping_add(1);
-            } else {
-                expected_seq = expected_seq.wrapping_add(1);
+                return Err(ProxyError::SeqNoMismatch {
+                    expected: expected_seq,
+                    got: seq_no,
+                });
             }
+            expected_seq = expected_seq.wrapping_add(1);
 
             let payload = &frame[8..pe];
             if payload.len() < 4 {
@@ -106,9 +120,15 @@ pub(crate) async fn reader_loop(
                 trace!(cid, flags, len = data.len(), "RPC_PROXY_ANS");
 
                 let routed = reg.route(cid, MeResponse::Data { flags, data }).await;
-                if !routed {
+                if !matches!(routed, RouteResult::Routed) {
+                    match routed {
+                        RouteResult::NoConn => stats.increment_me_route_drop_no_conn(),
+                        RouteResult::ChannelClosed => stats.increment_me_route_drop_channel_closed(),
+                        RouteResult::QueueFull => stats.increment_me_route_drop_queue_full(),
+                        RouteResult::Routed => {}
+                    }
                     reg.unregister(cid).await;
-                    send_close_conn(&writer, cid).await;
+                    send_close_conn(&tx, cid).await;
                 }
             } else if pt == RPC_SIMPLE_ACK_U32 && body.len() >= 12 {
                 let cid = u64::from_le_bytes(body[0..8].try_into().unwrap());
@@ -116,9 +136,15 @@ pub(crate) async fn reader_loop(
                 trace!(cid, cfm, "RPC_SIMPLE_ACK");
 
                 let routed = reg.route(cid, MeResponse::Ack(cfm)).await;
-                if !routed {
+                if !matches!(routed, RouteResult::Routed) {
+                    match routed {
+                        RouteResult::NoConn => stats.increment_me_route_drop_no_conn(),
+                        RouteResult::ChannelClosed => stats.increment_me_route_drop_channel_closed(),
+                        RouteResult::QueueFull => stats.increment_me_route_drop_queue_full(),
+                        RouteResult::Routed => {}
+                    }
                     reg.unregister(cid).await;
-                    send_close_conn(&writer, cid).await;
+                    send_close_conn(&tx, cid).await;
                 }
             } else if pt == RPC_CLOSE_EXT_U32 && body.len() >= 8 {
                 let cid = u64::from_le_bytes(body[0..8].try_into().unwrap());
@@ -136,12 +162,13 @@ pub(crate) async fn reader_loop(
                 let mut pong = Vec::with_capacity(12);
                 pong.extend_from_slice(&RPC_PONG_U32.to_le_bytes());
                 pong.extend_from_slice(&ping_id.to_le_bytes());
-                if let Err(e) = writer.lock().await.send_and_flush(&pong).await {
-                    warn!(error = %e, "PONG send failed");
+                if tx.send(WriterCommand::DataAndFlush(pong)).await.is_err() {
+                    warn!("PONG send failed");
                     break;
                 }
             } else if pt == RPC_PONG_U32 && body.len() >= 8 {
                 let ping_id = i64::from_le_bytes(body[0..8].try_into().unwrap());
+                stats.increment_me_keepalive_pong();
                 if let Some((sent, wid)) = {
                     let mut guard = ping_tracker.lock().await;
                     guard.remove(&ping_id)
@@ -171,12 +198,10 @@ pub(crate) async fn reader_loop(
     }
 }
 
-async fn send_close_conn(writer: &Arc<Mutex<RpcWriter>>, conn_id: u64) {
+async fn send_close_conn(tx: &mpsc::Sender<WriterCommand>, conn_id: u64) {
     let mut p = Vec::with_capacity(12);
     p.extend_from_slice(&RPC_CLOSE_CONN_U32.to_le_bytes());
     p.extend_from_slice(&conn_id.to_le_bytes());
 
-    if let Err(e) = writer.lock().await.send_and_flush(&p).await {
-        debug!(conn_id, error = %e, "Failed to send RPC_CLOSE_CONN");
-    }
+    let _ = tx.send(WriterCommand::DataAndFlush(p)).await;
 }

src/transport/middle_proxy/registry.rs

@@ -1,13 +1,23 @@
 use std::collections::{HashMap, HashSet};
 use std::net::SocketAddr;
 use std::sync::atomic::{AtomicU64, Ordering};
-use std::sync::Arc;
 
-use tokio::sync::{mpsc, Mutex, RwLock};
+use tokio::sync::{mpsc, RwLock};
+use tokio::sync::mpsc::error::TrySendError;
 
-use super::codec::RpcWriter;
+use super::codec::WriterCommand;
 use super::MeResponse;
 
+const ROUTE_CHANNEL_CAPACITY: usize = 4096;
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum RouteResult {
+    Routed,
+    NoConn,
+    ChannelClosed,
+    QueueFull,
+}
+
 #[derive(Clone)]
 pub struct ConnMeta {
     pub target_dc: i16,
@@ -25,12 +35,12 @@ pub struct BoundConn {
 #[derive(Clone)]
 pub struct ConnWriter {
     pub writer_id: u64,
-    pub writer: Arc<Mutex<RpcWriter>>,
+    pub tx: mpsc::Sender<WriterCommand>,
 }
 
 struct RegistryInner {
     map: HashMap<u64, mpsc::Sender<MeResponse>>,
-    writers: HashMap<u64, Arc<Mutex<RpcWriter>>>,
+    writers: HashMap<u64, mpsc::Sender<WriterCommand>>,
     writer_for_conn: HashMap<u64, u64>,
     conns_for_writer: HashMap<u64, HashSet<u64>>,
     meta: HashMap<u64, ConnMeta>,
@@ -64,7 +74,7 @@ impl ConnRegistry {
 
     pub async fn register(&self) -> (u64, mpsc::Receiver<MeResponse>) {
         let id = self.next_id.fetch_add(1, Ordering::Relaxed);
-        let (tx, rx) = mpsc::channel(1024);
+        let (tx, rx) = mpsc::channel(ROUTE_CHANNEL_CAPACITY);
         self.inner.write().await.map.insert(id, tx);
         (id, rx)
     }
@@ -83,12 +93,16 @@ impl ConnRegistry {
         None
     }
 
-    pub async fn route(&self, id: u64, resp: MeResponse) -> bool {
+    pub async fn route(&self, id: u64, resp: MeResponse) -> RouteResult {
         let inner = self.inner.read().await;
         if let Some(tx) = inner.map.get(&id) {
-            tx.try_send(resp).is_ok()
+            match tx.try_send(resp) {
+                Ok(()) => RouteResult::Routed,
+                Err(TrySendError::Closed(_)) => RouteResult::ChannelClosed,
+                Err(TrySendError::Full(_)) => RouteResult::QueueFull,
+            }
         } else {
-            false
+            RouteResult::NoConn
         }
     }
 
@@ -96,13 +110,13 @@ impl ConnRegistry {
         &self,
         conn_id: u64,
         writer_id: u64,
-        writer: Arc<Mutex<RpcWriter>>,
+        tx: mpsc::Sender<WriterCommand>,
         meta: ConnMeta,
     ) {
         let mut inner = self.inner.write().await;
         inner.meta.entry(conn_id).or_insert(meta);
         inner.writer_for_conn.insert(conn_id, writer_id);
-        inner.writers.entry(writer_id).or_insert_with(|| writer.clone());
+        inner.writers.entry(writer_id).or_insert_with(|| tx.clone());
         inner
             .conns_for_writer
             .entry(writer_id)
@@ -114,7 +128,7 @@ impl ConnRegistry {
         let inner = self.inner.read().await;
         let writer_id = inner.writer_for_conn.get(&conn_id).cloned()?;
         let writer = inner.writers.get(&writer_id).cloned()?;
-        Some(ConnWriter { writer_id, writer })
+        Some(ConnWriter { writer_id, tx: writer })
     }
 
     pub async fn writer_lost(&self, writer_id: u64) -> Vec<BoundConn> {

src/transport/middle_proxy/rotation.rs

@@ -31,8 +31,17 @@ pub async fn me_rotation_task(pool: Arc<MePool>, rng: Arc<SecureRandom>, interva
         info!(addr = %w.addr, writer_id = w.id, "Rotating ME connection");
         match pool.connect_one(w.addr, rng.as_ref()).await {
             Ok(()) => {
-                // Mark old writer for graceful drain; removal happens when sessions finish.
-                pool.mark_writer_draining(w.id).await;
+                tokio::time::sleep(Duration::from_secs(2)).await;
+                let ws = pool.writers.read().await;
+                let new_alive = ws.iter().any(|nw|
+                    nw.id != w.id && nw.addr == w.addr && !nw.degraded.load(Ordering::Relaxed) && !nw.draining.load(Ordering::Relaxed)
+                );
+                drop(ws);
+                if new_alive {
+                    pool.mark_writer_draining(w.id).await;
+                } else {
+                    warn!(addr = %w.addr, writer_id = w.id, "New writer died, keeping old");
+                }
             }
             Err(e) => {
                 warn!(addr = %w.addr, writer_id = w.id, error = %e, "ME rotation connect failed");

src/transport/middle_proxy/send.rs

@@ -10,6 +10,7 @@ use crate::network::IpFamily;
 use crate::protocol::constants::RPC_CLOSE_EXT_U32;
 
 use super::MePool;
+use super::codec::WriterCommand;
 use super::wire::build_proxy_req_payload;
 use rand::seq::SliceRandom;
 use super::registry::ConnMeta;
@@ -43,18 +44,15 @@ impl MePool {
         loop {
             if let Some(current) = self.registry.get_writer(conn_id).await {
                 let send_res = {
-                    if let Ok(mut guard) = current.writer.try_lock() {
-                        let r = guard.send(&payload).await;
-                        drop(guard);
-                        r
-                    } else {
-                        current.writer.lock().await.send(&payload).await
-                    }
+                    current
+                        .tx
+                        .send(WriterCommand::Data(payload.clone()))
+                        .await
                 };
                 match send_res {
                     Ok(()) => return Ok(()),
-                    Err(e) => {
-                        warn!(error = %e, writer_id = current.writer_id, "ME write failed");
+                    Err(_) => {
+                        warn!(writer_id = current.writer_id, "ME writer channel closed");
                         self.remove_writer_and_close_clients(current.writer_id).await;
                         continue;
                     }
@@ -64,7 +62,26 @@ impl MePool {
             let mut writers_snapshot = {
                 let ws = self.writers.read().await;
                 if ws.is_empty() {
-                    return Err(ProxyError::Proxy("All ME connections dead".into()));
+                    drop(ws);
+                    for family in self.family_order() {
+                        let map = match family {
+                            IpFamily::V4 => self.proxy_map_v4.read().await.clone(),
+                            IpFamily::V6 => self.proxy_map_v6.read().await.clone(),
+                        };
+                        for (_dc, addrs) in map.iter() {
+                            for (ip, port) in addrs {
+                                let addr = SocketAddr::new(*ip, *port);
+                                if self.connect_one(addr, self.rng.as_ref()).await.is_ok() {
+                                    self.writer_available.notify_waiters();
+                                    break;
+                                }
+                            }
+                        }
+                    }
+                    if tokio::time::timeout(Duration::from_secs(3), self.writer_available.notified()).await.is_err() {
+                        return Err(ProxyError::Proxy("All ME connections dead (waited 3s)".into()));
+                    }
+                    continue;
                 }
                 ws.clone()
             };
@@ -96,9 +113,10 @@ impl MePool {
                         writers_snapshot = ws2.clone();
                         drop(ws2);
                         candidate_indices = self.candidate_indices_for_dc(&writers_snapshot, target_dc).await;
-                        break;
+                        if !candidate_indices.is_empty() {
+                            break;
+                        }
                     }
-                    drop(map_guard);
                 }
                 if candidate_indices.is_empty() {
                     return Err(ProxyError::Proxy("No ME writers available for target DC".into()));
@@ -120,22 +138,15 @@ impl MePool {
                 if w.draining.load(Ordering::Relaxed) {
                     continue;
                 }
-                if let Ok(mut guard) = w.writer.try_lock() {
-                    let send_res = guard.send(&payload).await;
-                    drop(guard);
-                    match send_res {
-                        Ok(()) => {
-                            self.registry
-                                .bind_writer(conn_id, w.id, w.writer.clone(), meta.clone())
-                                .await;
-                            return Ok(());
-                        }
-                        Err(e) => {
-                            warn!(error = %e, writer_id = w.id, "ME write failed");
-                            self.remove_writer_and_close_clients(w.id).await;
-                            continue;
-                        }
-                    }
+                if w.tx.send(WriterCommand::Data(payload.clone())).await.is_ok() {
+                    self.registry
+                        .bind_writer(conn_id, w.id, w.tx.clone(), meta.clone())
+                        .await;
+                    return Ok(());
+                } else {
+                    warn!(writer_id = w.id, "ME writer channel closed");
+                    self.remove_writer_and_close_clients(w.id).await;
+                    continue;
                 }
             }
 
@@ -143,15 +154,15 @@ impl MePool {
             if w.draining.load(Ordering::Relaxed) {
                 continue;
             }
-            match w.writer.lock().await.send(&payload).await {
+            match w.tx.send(WriterCommand::Data(payload.clone())).await {
                 Ok(()) => {
                     self.registry
-                        .bind_writer(conn_id, w.id, w.writer.clone(), meta.clone())
+                        .bind_writer(conn_id, w.id, w.tx.clone(), meta.clone())
                         .await;
                     return Ok(());
                 }
-                Err(e) => {
-                    warn!(error = %e, writer_id = w.id, "ME write failed (blocking)");
+                Err(_) => {
+                    warn!(writer_id = w.id, "ME writer channel closed (blocking)");
                     self.remove_writer_and_close_clients(w.id).await;
                 }
             }
@@ -163,8 +174,8 @@ impl MePool {
             let mut p = Vec::with_capacity(12);
             p.extend_from_slice(&RPC_CLOSE_EXT_U32.to_le_bytes());
             p.extend_from_slice(&conn_id.to_le_bytes());
-            if let Err(e) = w.writer.lock().await.send_and_flush(&p).await {
-                debug!(error = %e, "ME close write failed");
+            if w.tx.send(WriterCommand::DataAndFlush(p)).await.is_err() {
+                debug!("ME close write failed");
                 self.remove_writer_and_close_clients(w.writer_id).await;
             }
         } else {
@@ -176,7 +187,7 @@ impl MePool {
     }
 
     pub fn connection_count(&self) -> usize {
-        self.writers.try_read().map(|w| w.len()).unwrap_or(0)
+        self.conn_count.load(Ordering::Relaxed)
     }
     
     pub(super) async fn candidate_indices_for_dc(

src/transport/proxy_protocol.rs

@@ -283,6 +283,58 @@ impl Default for ProxyProtocolV1Builder {
     }
 }
 
+/// Builder for PROXY protocol v2 header
+pub struct ProxyProtocolV2Builder {
+    src: Option<SocketAddr>,
+    dst: Option<SocketAddr>,
+}
+
+impl ProxyProtocolV2Builder {
+    pub fn new() -> Self {
+        Self { src: None, dst: None }
+    }
+
+    pub fn with_addrs(mut self, src: SocketAddr, dst: SocketAddr) -> Self {
+        self.src = Some(src);
+        self.dst = Some(dst);
+        self
+    }
+
+    pub fn build(&self) -> Vec<u8> {
+        let mut header = Vec::new();
+        header.extend_from_slice(PROXY_V2_SIGNATURE);
+        // version 2, PROXY command
+        header.push(0x21);
+
+        match (self.src, self.dst) {
+            (Some(SocketAddr::V4(src)), Some(SocketAddr::V4(dst))) => {
+                header.push(0x11); // INET + STREAM
+                header.extend_from_slice(&(12u16).to_be_bytes());
+                header.extend_from_slice(&src.ip().octets());
+                header.extend_from_slice(&dst.ip().octets());
+                header.extend_from_slice(&src.port().to_be_bytes());
+                header.extend_from_slice(&dst.port().to_be_bytes());
+            }
+            (Some(SocketAddr::V6(src)), Some(SocketAddr::V6(dst))) => {
+                header.push(0x21); // INET6 + STREAM
+                header.extend_from_slice(&(36u16).to_be_bytes());
+                header.extend_from_slice(&src.ip().octets());
+                header.extend_from_slice(&dst.ip().octets());
+                header.extend_from_slice(&src.port().to_be_bytes());
+                header.extend_from_slice(&dst.port().to_be_bytes());
+            }
+            _ => {
+                // LOCAL/UNSPEC: no address information
+                header[12] = 0x20; // version 2, LOCAL command
+                header.push(0x00);
+                header.extend_from_slice(&0u16.to_be_bytes());
+            }
+        }
+
+        header
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;
@@ -378,4 +430,4 @@ mod tests {
         let header = ProxyProtocolV1Builder::new().build();
         assert_eq!(header, b"PROXY UNKNOWN\r\n");
     }
-}
+}

src/transport/socket.rs

@@ -1,5 +1,7 @@
 //! TCP Socket Configuration
 
+use std::collections::HashSet;
+use std::fs;
 use std::io::Result;
 use std::net::{SocketAddr, IpAddr};
 use std::time::Duration;
@@ -122,6 +124,38 @@ pub fn get_local_addr(stream: &TcpStream) -> Option<SocketAddr> {
     stream.local_addr().ok()
 }
 
+/// Resolve primary IP address of a network interface by name.
+/// Returns the first address matching the requested family (IPv4/IPv6).
+#[cfg(unix)]
+pub fn resolve_interface_ip(name: &str, want_ipv6: bool) -> Option<IpAddr> {
+    use nix::ifaddrs::getifaddrs;
+
+    if let Ok(addrs) = getifaddrs() {
+        for iface in addrs {
+            if iface.interface_name == name {
+                if let Some(address) = iface.address {
+                    if let Some(v4) = address.as_sockaddr_in() {
+                        if !want_ipv6 {
+                            return Some(IpAddr::V4(v4.ip()));
+                        }
+                    } else if let Some(v6) = address.as_sockaddr_in6() {
+                        if want_ipv6 {
+                            return Some(IpAddr::V6(v6.ip().clone()));
+                        }
+                    }
+                }
+            }
+        }
+    }
+    None
+}
+
+/// Stub for non-Unix platforms: interface name resolution unsupported.
+#[cfg(not(unix))]
+pub fn resolve_interface_ip(_name: &str, _want_ipv6: bool) -> Option<IpAddr> {
+    None
+}
+
 /// Get peer address of a socket
 pub fn get_peer_addr(stream: &TcpStream) -> Option<SocketAddr> {
     stream.peer_addr().ok()
@@ -202,6 +236,133 @@ pub fn create_listener(addr: SocketAddr, options: &ListenOptions) -> Result<Sock
     Ok(socket)
 }
 
+/// Best-effort process list for listeners occupying the same local TCP port.
+#[derive(Debug, Clone)]
+pub struct ListenerProcessInfo {
+    pub pid: u32,
+    pub process: String,
+}
+
+/// Find processes currently listening on the local TCP port of `addr`.
+/// Returns an empty list when unsupported or when no owners can be resolved.
+pub fn find_listener_processes(addr: SocketAddr) -> Vec<ListenerProcessInfo> {
+    #[cfg(target_os = "linux")]
+    {
+        find_listener_processes_linux(addr)
+    }
+    #[cfg(not(target_os = "linux"))]
+    {
+        let _ = addr;
+        Vec::new()
+    }
+}
+
+#[cfg(target_os = "linux")]
+fn find_listener_processes_linux(addr: SocketAddr) -> Vec<ListenerProcessInfo> {
+    let inodes = listening_inodes_for_port(addr);
+    if inodes.is_empty() {
+        return Vec::new();
+    }
+
+    let mut out = Vec::new();
+
+    let proc_entries = match fs::read_dir("/proc") {
+        Ok(entries) => entries,
+        Err(_) => return out,
+    };
+
+    for entry in proc_entries.flatten() {
+        let pid = match entry.file_name().to_string_lossy().parse::<u32>() {
+            Ok(pid) => pid,
+            Err(_) => continue,
+        };
+
+        let fd_dir = entry.path().join("fd");
+        let fd_entries = match fs::read_dir(fd_dir) {
+            Ok(entries) => entries,
+            Err(_) => continue,
+        };
+
+        let mut matched = false;
+        for fd in fd_entries.flatten() {
+            let link_target = match fs::read_link(fd.path()) {
+                Ok(link) => link,
+                Err(_) => continue,
+            };
+
+            let link_str = link_target.to_string_lossy();
+            let Some(rest) = link_str.strip_prefix("socket:[") else {
+                continue;
+            };
+            let Some(inode_str) = rest.strip_suffix(']') else {
+                continue;
+            };
+            let Ok(inode) = inode_str.parse::<u64>() else {
+                continue;
+            };
+
+            if inodes.contains(&inode) {
+                matched = true;
+                break;
+            }
+        }
+
+        if matched {
+            let process = fs::read_to_string(entry.path().join("comm"))
+                .ok()
+                .map(|s| s.trim().to_string())
+                .filter(|s| !s.is_empty())
+                .unwrap_or_else(|| "unknown".to_string());
+            out.push(ListenerProcessInfo { pid, process });
+        }
+    }
+
+    out.sort_by_key(|p| p.pid);
+    out.dedup_by_key(|p| p.pid);
+    out
+}
+
+#[cfg(target_os = "linux")]
+fn listening_inodes_for_port(addr: SocketAddr) -> HashSet<u64> {
+    let path = match addr {
+        SocketAddr::V4(_) => "/proc/net/tcp",
+        SocketAddr::V6(_) => "/proc/net/tcp6",
+    };
+
+    let mut inodes = HashSet::new();
+    let Ok(data) = fs::read_to_string(path) else {
+        return inodes;
+    };
+
+    for line in data.lines().skip(1) {
+        let cols: Vec<&str> = line.split_whitespace().collect();
+        if cols.len() < 10 {
+            continue;
+        }
+
+        // LISTEN state in /proc/net/tcp*
+        if cols[3] != "0A" {
+            continue;
+        }
+
+        let Some(port_hex) = cols[1].split(':').nth(1) else {
+            continue;
+        };
+        let Ok(port) = u16::from_str_radix(port_hex, 16) else {
+            continue;
+        };
+        if port != addr.port() {
+            continue;
+        }
+
+        if let Ok(inode) = cols[9].parse::<u64>() {
+            inodes.insert(inode);
+        }
+    }
+
+    inodes
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;

src/transport/upstream.rs

@@ -5,6 +5,7 @@
 use std::collections::HashMap;
 use std::net::{SocketAddr, IpAddr};
 use std::sync::Arc;
+use std::sync::atomic::{AtomicUsize, Ordering};
 use std::time::Duration;
 use tokio::net::TcpStream;
 use tokio::sync::RwLock;
@@ -15,7 +16,7 @@ use tracing::{debug, warn, info, trace};
 use crate::config::{UpstreamConfig, UpstreamType};
 use crate::error::{Result, ProxyError};
 use crate::protocol::constants::{TG_DATACENTERS_V4, TG_DATACENTERS_V6, TG_DATACENTER_PORT};
-use crate::transport::socket::create_outgoing_socket_bound;
+use crate::transport::socket::{create_outgoing_socket_bound, resolve_interface_ip};
 use crate::transport::socks::{connect_socks4, connect_socks5};
 
 /// Number of Telegram datacenters
@@ -23,6 +24,8 @@ const NUM_DCS: usize = 5;
 
 /// Timeout for individual DC ping attempt
 const DC_PING_TIMEOUT_SECS: u64 = 5;
+/// Timeout for direct TG DC TCP connect readiness.
+const DIRECT_CONNECT_TIMEOUT_SECS: u64 = 10;
 
 // ============= RTT Tracking =============
 
@@ -84,6 +87,8 @@ struct UpstreamState {
     dc_latency: [LatencyEma; NUM_DCS],
     /// Per-DC IP version preference (learned from connectivity tests)
     dc_ip_pref: [IpPreference; NUM_DCS],
+    /// Round-robin counter for bind_addresses selection
+    bind_rr: Arc<AtomicUsize>,
 }
 
 impl UpstreamState {
@@ -95,6 +100,7 @@ impl UpstreamState {
             last_check: std::time::Instant::now(),
             dc_latency: [LatencyEma::new(0.3); NUM_DCS],
             dc_ip_pref: [IpPreference::Unknown; NUM_DCS],
+            bind_rr: Arc::new(AtomicUsize::new(0)),
         }
     }
 
@@ -166,21 +172,85 @@ impl UpstreamManager {
         }
     }
 
+    fn resolve_bind_address(
+        interface: &Option<String>,
+        bind_addresses: &Option<Vec<String>>,
+        target: SocketAddr,
+        rr: Option<&AtomicUsize>,
+    ) -> Option<IpAddr> {
+        let want_ipv6 = target.is_ipv6();
+
+        if let Some(addrs) = bind_addresses {
+            let candidates: Vec<IpAddr> = addrs
+                .iter()
+                .filter_map(|s| s.parse::<IpAddr>().ok())
+                .filter(|ip| ip.is_ipv6() == want_ipv6)
+                .collect();
+
+            if !candidates.is_empty() {
+                if let Some(counter) = rr {
+                    let idx = counter.fetch_add(1, Ordering::Relaxed) % candidates.len();
+                    return Some(candidates[idx]);
+                }
+                return candidates.first().copied();
+            }
+        }
+
+        if let Some(iface) = interface {
+            if let Ok(ip) = iface.parse::<IpAddr>() {
+                if ip.is_ipv6() == want_ipv6 {
+                    return Some(ip);
+                }
+            } else {
+                #[cfg(unix)]
+                if let Some(ip) = resolve_interface_ip(iface, want_ipv6) {
+                    return Some(ip);
+                }
+            }
+        }
+
+        None
+    }
+
     /// Select upstream using latency-weighted random selection.
-    async fn select_upstream(&self, dc_idx: Option<i16>) -> Option<usize> {
+    async fn select_upstream(&self, dc_idx: Option<i16>, scope: Option<&str>) -> Option<usize> {
         let upstreams = self.upstreams.read().await;
         if upstreams.is_empty() {
             return None;
         }
-
-        let healthy: Vec<usize> = upstreams.iter()
+        // Scope filter:
+        //   If scope is set: only scoped and matched items
+        //   If scope is not set: only unscoped items
+        let filtered_upstreams : Vec<usize> = upstreams.iter()
             .enumerate()
-            .filter(|(_, u)| u.healthy)
+            .filter(|(_, u)| {
+                scope.map_or(
+                    u.config.scopes.is_empty(),
+                    |req_scope| {
+                        u.config.scopes
+                            .split(',')
+                            .map(str::trim)
+                            .any(|s| s == req_scope)
+                    }
+                )
+            })
             .map(|(i, _)| i)
             .collect();
 
+        // Healthy filter
+        let healthy: Vec<usize> = filtered_upstreams.iter()
+            .filter(|&&i| upstreams[i].healthy)
+            .copied()
+            .collect();
+
+        if filtered_upstreams.is_empty() {
+            warn!(scope = scope, "No upstreams available! Using first (direct?)");
+            return None;
+        }
+
         if healthy.is_empty() {
-            return Some(rand::rng().gen_range(0..upstreams.len()));
+            warn!(scope = scope, "No healthy upstreams available! Using random.");
+            return Some(filtered_upstreams[rand::rng().gen_range(0..filtered_upstreams.len())]);
         }
 
         if healthy.len() == 1 {
@@ -222,18 +292,28 @@ impl UpstreamManager {
     }
 
     /// Connect to target through a selected upstream.
-    pub async fn connect(&self, target: SocketAddr, dc_idx: Option<i16>) -> Result<TcpStream> {
-        let idx = self.select_upstream(dc_idx).await
+    pub async fn connect(&self, target: SocketAddr, dc_idx: Option<i16>, scope: Option<&str>) -> Result<TcpStream> {
+        let idx = self.select_upstream(dc_idx, scope).await
             .ok_or_else(|| ProxyError::Config("No upstreams available".to_string()))?;
 
-        let upstream = {
+        let mut upstream = {
             let guard = self.upstreams.read().await;
             guard[idx].config.clone()
         };
 
+        // Set scope for configuration copy
+        if let Some(s) = scope {
+            upstream.selected_scope = s.to_string();
+        }
+
         let start = Instant::now();
 
-        match self.connect_via_upstream(&upstream, target).await {
+        let bind_rr = {
+            let guard = self.upstreams.read().await;
+            guard.get(idx).map(|u| u.bind_rr.clone())
+        };
+
+        match self.connect_via_upstream(&upstream, target, bind_rr).await {
             Ok(stream) => {
                 let rtt_ms = start.elapsed().as_secs_f64() * 1000.0;
                 let mut guard = self.upstreams.write().await;
@@ -265,13 +345,27 @@ impl UpstreamManager {
         }
     }
 
-    async fn connect_via_upstream(&self, config: &UpstreamConfig, target: SocketAddr) -> Result<TcpStream> {
+    async fn connect_via_upstream(
+        &self,
+        config: &UpstreamConfig,
+        target: SocketAddr,
+        bind_rr: Option<Arc<AtomicUsize>>,
+    ) -> Result<TcpStream> {
         match &config.upstream_type {
-            UpstreamType::Direct { interface } => {
-                let bind_ip = interface.as_ref()
-                    .and_then(|s| s.parse::<IpAddr>().ok());
+            UpstreamType::Direct { interface, bind_addresses } => {
+                let bind_ip = Self::resolve_bind_address(
+                    interface,
+                    bind_addresses,
+                    target,
+                    bind_rr.as_deref(),
+                );
 
                 let socket = create_outgoing_socket_bound(target, bind_ip)?;
+                if let Some(ip) = bind_ip {
+                    debug!(bind = %ip, target = %target, "Bound outgoing socket");
+                } else if interface.is_some() || bind_addresses.is_some() {
+                    debug!(target = %target, "No matching bind address for target family");
+                }
 
                 socket.set_nonblocking(true)?;
                 match socket.connect(&target.into()) {
@@ -283,7 +377,16 @@ impl UpstreamManager {
                 let std_stream: std::net::TcpStream = socket.into();
                 let stream = TcpStream::from_std(std_stream)?;
 
-                stream.writable().await?;
+                let connect_timeout = Duration::from_secs(DIRECT_CONNECT_TIMEOUT_SECS);
+                match tokio::time::timeout(connect_timeout, stream.writable()).await {
+                    Ok(Ok(())) => {}
+                    Ok(Err(e)) => return Err(ProxyError::Io(e)),
+                    Err(_) => {
+                        return Err(ProxyError::ConnectionTimeout {
+                            addr: target.to_string(),
+                        });
+                    }
+                }
                 if let Some(e) = stream.take_error()? {
                     return Err(ProxyError::Io(e));
                 }
@@ -291,57 +394,97 @@ impl UpstreamManager {
                 Ok(stream)
             },
             UpstreamType::Socks4 { address, interface, user_id } => {
-                let proxy_addr: SocketAddr = address.parse()
-                    .map_err(|_| ProxyError::Config("Invalid SOCKS4 address".to_string()))?;
+                // Try to parse as SocketAddr first (IP:port), otherwise treat as hostname:port
+                let mut stream = if let Ok(proxy_addr) = address.parse::<SocketAddr>() {
+                    // IP:port format - use socket with optional interface binding
+                    let bind_ip = Self::resolve_bind_address(
+                        interface,
+                        &None,
+                        proxy_addr,
+                        bind_rr.as_deref(),
+                    );
 
-                let bind_ip = interface.as_ref()
-                    .and_then(|s| s.parse::<IpAddr>().ok());
+                    let socket = create_outgoing_socket_bound(proxy_addr, bind_ip)?;
 
-                let socket = create_outgoing_socket_bound(proxy_addr, bind_ip)?;
+                    socket.set_nonblocking(true)?;
+                    match socket.connect(&proxy_addr.into()) {
+                        Ok(()) => {},
+                        Err(err) if err.raw_os_error() == Some(libc::EINPROGRESS) || err.kind() == std::io::ErrorKind::WouldBlock => {},
+                        Err(err) => return Err(ProxyError::Io(err)),
+                    }
 
-                socket.set_nonblocking(true)?;
-                match socket.connect(&proxy_addr.into()) {
-                    Ok(()) => {},
-                    Err(err) if err.raw_os_error() == Some(libc::EINPROGRESS) || err.kind() == std::io::ErrorKind::WouldBlock => {},
-                    Err(err) => return Err(ProxyError::Io(err)),
-                }
+                    let std_stream: std::net::TcpStream = socket.into();
+                    let stream = TcpStream::from_std(std_stream)?;
 
-                let std_stream: std::net::TcpStream = socket.into();
-                let mut stream = TcpStream::from_std(std_stream)?;
+                    stream.writable().await?;
+                    if let Some(e) = stream.take_error()? {
+                        return Err(ProxyError::Io(e));
+                    }
+                    stream
+                } else {
+                    // Hostname:port format - use tokio DNS resolution
+                    // Note: interface binding is not supported for hostnames
+                    if interface.is_some() {
+                        warn!("SOCKS4 interface binding is not supported for hostname addresses, ignoring");
+                    }
+                    TcpStream::connect(address).await
+                        .map_err(ProxyError::Io)?
+                };
 
-                stream.writable().await?;
-                if let Some(e) = stream.take_error()? {
-                    return Err(ProxyError::Io(e));
-                }
+                // replace socks user_id with config.selected_scope, if set
+                let scope: Option<&str> = Some(config.selected_scope.as_str())
+                    .filter(|s| !s.is_empty());
+                let _user_id: Option<&str> = scope.or(user_id.as_deref());
 
-                connect_socks4(&mut stream, target, user_id.as_deref()).await?;
+                connect_socks4(&mut stream, target, _user_id).await?;
                 Ok(stream)
             },
             UpstreamType::Socks5 { address, interface, username, password } => {
-                let proxy_addr: SocketAddr = address.parse()
-                    .map_err(|_| ProxyError::Config("Invalid SOCKS5 address".to_string()))?;
+                // Try to parse as SocketAddr first (IP:port), otherwise treat as hostname:port
+                let mut stream = if let Ok(proxy_addr) = address.parse::<SocketAddr>() {
+                    // IP:port format - use socket with optional interface binding
+                    let bind_ip = Self::resolve_bind_address(
+                        interface,
+                        &None,
+                        proxy_addr,
+                        bind_rr.as_deref(),
+                    );
 
-                let bind_ip = interface.as_ref()
-                    .and_then(|s| s.parse::<IpAddr>().ok());
+                    let socket = create_outgoing_socket_bound(proxy_addr, bind_ip)?;
 
-                let socket = create_outgoing_socket_bound(proxy_addr, bind_ip)?;
+                    socket.set_nonblocking(true)?;
+                    match socket.connect(&proxy_addr.into()) {
+                        Ok(()) => {},
+                        Err(err) if err.raw_os_error() == Some(libc::EINPROGRESS) || err.kind() == std::io::ErrorKind::WouldBlock => {},
+                        Err(err) => return Err(ProxyError::Io(err)),
+                    }
 
-                socket.set_nonblocking(true)?;
-                match socket.connect(&proxy_addr.into()) {
-                    Ok(()) => {},
-                    Err(err) if err.raw_os_error() == Some(libc::EINPROGRESS) || err.kind() == std::io::ErrorKind::WouldBlock => {},
-                    Err(err) => return Err(ProxyError::Io(err)),
-                }
+                    let std_stream: std::net::TcpStream = socket.into();
+                    let stream = TcpStream::from_std(std_stream)?;
 
-                let std_stream: std::net::TcpStream = socket.into();
-                let mut stream = TcpStream::from_std(std_stream)?;
+                    stream.writable().await?;
+                    if let Some(e) = stream.take_error()? {
+                        return Err(ProxyError::Io(e));
+                    }
+                    stream
+                } else {
+                    // Hostname:port format - use tokio DNS resolution
+                    // Note: interface binding is not supported for hostnames
+                    if interface.is_some() {
+                        warn!("SOCKS5 interface binding is not supported for hostname addresses, ignoring");
+                    }
+                    TcpStream::connect(address).await
+                        .map_err(ProxyError::Io)?
+                };
 
-                stream.writable().await?;
-                if let Some(e) = stream.take_error()? {
-                    return Err(ProxyError::Io(e));
-                }
+                debug!(config = ?config, "Socks5 connection");
+                // replace socks user:pass with config.selected_scope, if set
+                let scope: Option<&str> = Some(config.selected_scope.as_str())
+                    .filter(|s| !s.is_empty());
+                let _username: Option<&str> = scope.or(username.as_deref());
+                let _password: Option<&str> = scope.or(password.as_deref());
 
-                connect_socks5(&mut stream, target, username.as_deref(), password.as_deref()).await?;
+                connect_socks5(&mut stream, target, _username, _password).await?;
                 Ok(stream)
             },
         }
@@ -358,18 +501,18 @@ impl UpstreamManager {
         ipv4_enabled: bool,
         ipv6_enabled: bool,
     ) -> Vec<StartupPingResult> {
-        let upstreams: Vec<(usize, UpstreamConfig)> = {
+        let upstreams: Vec<(usize, UpstreamConfig, Arc<AtomicUsize>)> = {
             let guard = self.upstreams.read().await;
             guard.iter().enumerate()
-                .map(|(i, u)| (i, u.config.clone()))
+                .map(|(i, u)| (i, u.config.clone(), u.bind_rr.clone()))
                 .collect()
         };
 
         let mut all_results = Vec::new();
 
-        for (upstream_idx, upstream_config) in &upstreams {
+        for (upstream_idx, upstream_config, bind_rr) in &upstreams {
             let upstream_name = match &upstream_config.upstream_type {
-                UpstreamType::Direct { interface } => {
+                UpstreamType::Direct { interface, .. } => {
                     format!("direct{}", interface.as_ref().map(|i| format!(" ({})", i)).unwrap_or_default())
                 }
                 UpstreamType::Socks4 { address, .. } => format!("socks4://{}", address),
@@ -384,7 +527,7 @@ impl UpstreamManager {
 
                     let result = tokio::time::timeout(
                         Duration::from_secs(DC_PING_TIMEOUT_SECS),
-                        self.ping_single_dc(&upstream_config, addr_v6)
+                        self.ping_single_dc(&upstream_config, Some(bind_rr.clone()), addr_v6)
                     ).await;
 
                     let ping_result = match result {
@@ -435,7 +578,7 @@ impl UpstreamManager {
 
                     let result = tokio::time::timeout(
                         Duration::from_secs(DC_PING_TIMEOUT_SECS),
-                        self.ping_single_dc(&upstream_config, addr_v4)
+                        self.ping_single_dc(&upstream_config, Some(bind_rr.clone()), addr_v4)
                     ).await;
 
                     let ping_result = match result {
@@ -498,7 +641,7 @@ impl UpstreamManager {
                             }
                             let result = tokio::time::timeout(
                                 Duration::from_secs(DC_PING_TIMEOUT_SECS),
-                                self.ping_single_dc(&upstream_config, addr)
+                                self.ping_single_dc(&upstream_config, Some(bind_rr.clone()), addr)
                             ).await;
 
                             let ping_result = match result {
@@ -567,9 +710,14 @@ impl UpstreamManager {
         all_results
     }
 
-    async fn ping_single_dc(&self, config: &UpstreamConfig, target: SocketAddr) -> Result<f64> {
+    async fn ping_single_dc(
+        &self,
+        config: &UpstreamConfig,
+        bind_rr: Option<Arc<AtomicUsize>>,
+        target: SocketAddr,
+    ) -> Result<f64> {
         let start = Instant::now();
-        let _stream = self.connect_via_upstream(config, target).await?;
+        let _stream = self.connect_via_upstream(config, target, bind_rr).await?;
         Ok(start.elapsed().as_secs_f64() * 1000.0)
     }
 
@@ -609,15 +757,16 @@ impl UpstreamManager {
             let count = self.upstreams.read().await.len();
 
             for i in 0..count {
-                let config = {
+                let (config, bind_rr) = {
                     let guard = self.upstreams.read().await;
-                    guard[i].config.clone()
+                    let u = &guard[i];
+                    (u.config.clone(), u.bind_rr.clone())
                 };
 
                 let start = Instant::now();
                 let result = tokio::time::timeout(
                     Duration::from_secs(10),
-                    self.connect_via_upstream(&config, dc_addr)
+                    self.connect_via_upstream(&config, dc_addr, Some(bind_rr.clone()))
                 ).await;
 
                 match result {
@@ -646,7 +795,7 @@ impl UpstreamManager {
                             let start2 = Instant::now();
                             let result2 = tokio::time::timeout(
                                 Duration::from_secs(10),
-                                self.connect_via_upstream(&config, fallback_addr)
+                                self.connect_via_upstream(&config, fallback_addr, Some(bind_rr.clone()))
                             ).await;
 
                             let mut guard = self.upstreams.write().await;

telemt.service

@@ -7,6 +7,7 @@ Type=simple
 WorkingDirectory=/bin
 ExecStart=/bin/telemt /etc/telemt.toml
 Restart=on-failure
+LimitNOFILE=65536
 
 [Install]
 WantedBy=multi-user.target