New Relay on Tokio Copy Bidirectional

1.2.0.2 Special DC support

.github/workflows/rust.yml

@@ -10,10 +10,15 @@ env:
   CARGO_TERM_COLOR: always
 
 jobs:
-  build-and-test:
-    name: Build & Test
+  build:
+    name: Build
     runs-on: ubuntu-latest
 
+    permissions:
+      contents: read
+      actions: write
+      checks: write
+
     steps:
     - name: Checkout repository
       uses: actions/checkout@v4

Cargo.toml

@@ -1,15 +1,14 @@
 [package]
 name = "telemt"
-version = "1.0.0"
-edition = "2021"
-rust-version = "1.75"
+version = "1.2.0"
+edition = "2024"
 
 [dependencies]
 # C
 libc = "0.2"
 
 # Async runtime
-tokio = { version = "1.35", features = ["full", "tracing"] }
+tokio = { version = "1.42", features = ["full", "tracing"] }
 tokio-util = { version = "0.7", features = ["codec"] }
 
 # Crypto
@@ -20,41 +19,41 @@ sha2 = "0.10"
 sha1 = "0.10"
 md-5 = "0.10"
 hmac = "0.12"
-crc32fast = "1.3"
+crc32fast = "1.4"
+zeroize = { version = "1.8", features = ["derive"] }
 
 # Network
 socket2 = { version = "0.5", features = ["all"] }
-rustls = "0.22"
 
-# Serial
+# Serialization
 serde = { version = "1.0", features = ["derive"] }
 serde_json = "1.0"
 toml = "0.8"
 
 # Utils
-bytes = "1.5"
-thiserror = "1.0"
+bytes = "1.9"
+thiserror = "2.0"
 tracing = "0.1"
 tracing-subscriber = { version = "0.3", features = ["env-filter"] }
 parking_lot = "0.12"
 dashmap = "5.5"
 lru = "0.12"
-rand = "0.8"
+rand = "0.9"
 chrono = { version = "0.4", features = ["serde"] }
 hex = "0.4"
-base64 = "0.21"
+base64 = "0.22"
 url = "2.5"
-regex = "1.10"
-once_cell = "1.19"
+regex = "1.11"
 crossbeam-queue = "0.3"
 
 # HTTP
-reqwest = { version = "0.11", features = ["rustls-tls"], default-features = false }
+reqwest = { version = "0.12", features = ["rustls-tls"], default-features = false }
 
 [dev-dependencies]
 tokio-test = "0.4"
 criterion = "0.5"
 proptest = "1.4"
+futures = "0.3"
 
 [[bench]]
 name = "crypto_bench"

README.md

@@ -2,16 +2,49 @@
 
 **Telemt** is a fast, secure, and feature-rich server written in Rust: it fully implements the official Telegram proxy algo and adds many production-ready improvements such as connection pooling, replay protection, detailed statistics, masking from "prying" eyes
 
+## Emergency
+**Важное сообщение для пользователей из России**
+
+Мы работаем над проектом с Нового года и сейчас готовим новый релиз - 1.2
+
+В нём имплементируется поддержка Middle Proxy Protocol - основного терминатора для Ad Tag:
+работа над ним идёт с 6 ферваля, а уже 10 февраля произошли "громкие события"...
+
+Если у вас есть компетенции в асинхронных сетевых приложениях - мы открыты к предложениям и pull requests
+
+**Important message for users from Russia**
+
+We've been working on the project since December 30 and are currently preparing a new release – 1.2
+
+It implements support for the Middle Proxy Protocol – the primary point for the Ad Tag:
+development on it started on February 6th, and by February 10th, "big activity" in Russia had already "taken place"...
+
+If you have expertise in asynchronous network applications – we are open to ideas and pull requests!
+
+# 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) 
+
 # GOTO
 - [Features](#features)
 - [Quick Start Guide](#quick-start-guide)
-  - [Build](#build)
 - [How to use?](#how-to-use)
   - [Systemd Method](#telemt-via-systemd)
+- [Configuration](#configuration)
+  - [Minimal Configuration](#minimal-configuration-for-first-start)
+  - [Advanced](#advanced)
+    - [Adtag](#adtag)
+    - [Listening and Announce IPs](#listening-and-announce-ips)
+    - [Upstream Manager](#upstream-manager)
+      - [IP](#bind-on-ip)
+      - [SOCKS](#socks45-as-upstream)
 - [FAQ](#faq)
+  - [Recognizability for DPI + crawler](#recognizability-for-dpi-and-crawler)
   - [Telegram Calls](#telegram-calls-via-mtproxy)
   - [DPI](#how-does-dpi-see-mtproxy-tls)
   - [Whitelist on Network Level](#whitelist-on-ip)
+- [Build](#build)
 - [Why Rust?](#why-rust)
 
 ## Features
@@ -27,25 +60,27 @@
 - Extensive logging via `trace` and `debug` with `RUST_LOG` method
 
 ## Quick Start Guide
-
-### Build
+**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
-# Cloning repo
-git clone https://github.com/telemt/telemt 
-# Changing Directory to telemt
-cd telemt
-# Starting Release Build
-cargo build --release
-# Move to /bin
-mv ./target/release/telemt /bin
-# Make executable
-chmod +x /bin/telemt
-# Lets go!
-telemt config.toml
+wget https://github.com/telemt/telemt/releases/latest/download/telemt
 ```
+2. Move to Bin Folder
+```bash
+mv telemt /bin
+```
+4. Make Executable
+```bash
+chmod +x /bin/telemt
+```
+5. Go to [How to use?](#how-to-use) section for for further steps
 
 ## How to use?
 ### Telemt via Systemd
+**This instruction "assume" that you:**
+- logged in as root or executed `su -` / `sudo su`
+- you already have an assembled and executable `telemt` in /bin folder as a result of the [Quick Start Guide](#quick-start-guide) or [Build](#build)
+
 **0. Check port and generate secrets**
 
 The port you have selected for use should be MISSING from the list, when:
@@ -57,6 +92,14 @@ Generate 16 bytes/32 characters HEX with OpenSSL or another way:
 ```bash
 openssl rand -hex 16
 ```
+OR
+```bash
+xxd -l 16 -p /dev/urandom
+```
+OR
+```bash
+python3 -c 'import os; print(os.urandom(16).hex())'
+```
 
 **1. Place your config to /etc/telemt.toml**
 
@@ -64,28 +107,8 @@ Open nano
 ```bash
 nano /etc/telemt.toml
 ```
-```bash
-port = 443                              # Listening port
+paste your config from [Configuration](#configuration) section
 
-[users]
-hello = "00000000000000000000000000000000" # Replace the secret with one generated before
-
-[modes]
-classic = false                         # Plain obfuscated mode
-secure = false                          # dd-prefix mode
-tls = true                              # Fake TLS - ee-prefix
-
-tls_domain = "petrovich.ru"             # Domain for ee-secret and masking
-mask = true                             # Enable masking of bad traffic
-mask_host = "petrovich.ru"              # Optional override for mask destination
-mask_port = 443                         # Port for masking
-
-prefer_ipv6 = false                     # Try IPv6 DCs first if true
-fast_mode = true                        # Use "fast" obfuscation variant
-
-client_keepalive = 600                  # Seconds
-client_ack_timeout = 300                # Seconds
-```
 then Ctrl+X -> Y -> Enter to save
 
 **2. Create service on /etc/systemd/system/telemt.service**
@@ -117,9 +140,224 @@ then Ctrl+X -> Y -> Enter to save
 
 **5.** In Shell type `systemctl enable telemt` - then telemt will start with system startup, after the network is up
 
+## Configuration
+### Minimal Configuration for First Start
+```toml
+# === UI ===
+# Users to show in the startup log (tg:// links)
+show_link = ["hello"]
+
+# === General Settings ===
+[general]
+prefer_ipv6 = false
+fast_mode = true
+use_middle_proxy = false
+# ad_tag = "..."
+
+[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_ip = "1.2.3.4" # Optional: Public IP for tg:// links
+
+[[server.listeners]]
+ip = "::"
+
+# === 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
+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
+```
+### Advanced
+#### Adtag
+To use channel advertising and usage statistics from Telegram, get Adtag from [@mtproxybot](https://t.me/mtproxybot), add this parameter to section `[General]`
+```toml
+ad_tag = "00000000000000000000000000000000" # Replace zeros to your adtag from @mtproxybot
+```
+#### Listening and Announce IPs
+To specify listening address and/or address in links, add to section `[[server.listeners]]` of config.toml:
+```toml
+[[server.listeners]]
+ip = "0.0.0.0"          # 0.0.0.0 = all IPs; your IP = specific listening
+announce_ip = "1.2.3.4" # IP in links; comment with # if not used
+```
+#### Upstream Manager
+To specify upstream, add to section `[[upstreams]]` of config.toml:
+##### Bind on IP
+```toml
+[[upstreams]]
+type = "direct"
+weight = 1
+enabled = true
+interface = "192.168.1.100" # Change to your outgoing IP
+```
+##### SOCKS4/5 as Upstream
+- Without Auth:
+```toml
+[[upstreams]]
+type = "socks5"            # Specify SOCKS4 or SOCKS5
+address = "1.2.3.4:1234"   # SOCKS-server Address
+weight = 1                 # Set Weight for Scenarios
+enabled = true
+```
+
+- With Auth:
+```toml
+[[upstreams]]
+type = "socks5"            # Specify SOCKS4 or SOCKS5
+address = "1.2.3.4:1234"   # SOCKS-server Address
+username = "user"          # Username for Auth on SOCKS-server
+password = "pass"          # Password for Auth on SOCKS-server
+weight = 1                 # Set Weight for Scenarios
+enabled = true
+```
+
 ## FAQ
+### Recognizability for DPI and crawler
+Since version 1.1.0.0, we have debugged masking perfectly: for all clients without "presenting" a key, 
+we transparently direct traffic to the target host!
+
+- We consider this a breakthrough aspect, which has no stable analogues today
+- Based on this: if `telemt` configured correctly, **TLS mode is completely identical to real-life handshake + communication** with a specified host
+- Here is our evidence:
+    - 212.220.88.77 - "dummy" host, running `telemt`
+    - `petrovich.ru` - `tls` + `masking` host, in HEX: `706574726f766963682e7275`
+    - **No MITM + No Fake Certificates/Crypto** = pure transparent *TCP Splice* to "best" upstream: MTProxy or tls/mask-host:
+      - DPI see legitimate HTTPS to `tls_host`, including *valid chain-of-trust* and entropy
+      - Crawlers completely satisfied receiving responses from `mask_host`
+  #### Client WITH secret-key accesses the MTProxy resource:
+  
+  <img width="360" height="439" alt="telemt" src="https://github.com/user-attachments/assets/39352afb-4a11-4ecc-9d91-9e8cfb20607d" />
+  
+  #### Client WITHOUT secret-key gets transparent access to the specified resource:
+    - with trusted certificate
+    - with original handshake
+    - with full request-response way
+    - with low-latency overhead
+```bash
+root@debian:~/telemt# curl -v -I --resolve petrovich.ru:443:212.220.88.77 https://petrovich.ru/
+* Added petrovich.ru:443:212.220.88.77 to DNS cache
+* Hostname petrovich.ru was found in DNS cache
+*   Trying 212.220.88.77:443...
+* Connected to petrovich.ru (212.220.88.77) port 443 (#0)
+* ALPN: offers h2,http/1.1
+* TLSv1.3 (OUT), TLS handshake, Client hello (1):
+*  CAfile: /etc/ssl/certs/ca-certificates.crt
+*  CApath: /etc/ssl/certs
+* TLSv1.3 (IN), TLS handshake, Server hello (2):
+* TLSv1.3 (IN), TLS handshake, Encrypted Extensions (8):
+* TLSv1.3 (IN), TLS handshake, Certificate (11):
+* TLSv1.3 (IN), TLS handshake, CERT verify (15):
+* TLSv1.3 (IN), TLS handshake, Finished (20):
+* TLSv1.3 (OUT), TLS change cipher, Change cipher spec (1):
+* TLSv1.3 (OUT), TLS handshake, Finished (20):
+* SSL connection using TLSv1.3 / TLS_AES_256_GCM_SHA384
+* ALPN: server did not agree on a protocol. Uses default.
+* Server certificate:
+*  subject: C=RU; ST=Saint Petersburg; L=Saint Petersburg; O=STD Petrovich; CN=*.petrovich.ru
+*  start date: Jan 28 11:21:01 2025 GMT
+*  expire date: Mar  1 11:21:00 2026 GMT
+*  subjectAltName: host "petrovich.ru" matched cert's "petrovich.ru"
+*  issuer: C=BE; O=GlobalSign nv-sa; CN=GlobalSign RSA OV SSL CA 2018
+*  SSL certificate verify ok.
+* using HTTP/1.x
+> HEAD / HTTP/1.1
+> Host: petrovich.ru
+> User-Agent: curl/7.88.1
+> Accept: */*
+> 
+* TLSv1.3 (IN), TLS handshake, Newsession Ticket (4):
+* TLSv1.3 (IN), TLS handshake, Newsession Ticket (4):
+* old SSL session ID is stale, removing
+< HTTP/1.1 200 OK
+HTTP/1.1 200 OK
+< Server: Variti/0.9.3a
+Server: Variti/0.9.3a
+< Date: Thu, 01 Jan 2026 00:0000 GMT
+Date: Thu, 01 Jan 2026 00:0000 GMT
+< Access-Control-Allow-Origin: *
+Access-Control-Allow-Origin: *
+< Content-Type: text/html
+Content-Type: text/html
+< Cache-Control: no-store
+Cache-Control: no-store
+< Expires: Thu, 01 Jan 2026 00:0000 GMT
+Expires: Thu, 01 Jan 2026 00:0000 GMT
+< Pragma: no-cache
+Pragma: no-cache
+< Set-Cookie: ipp_uid=XXXXX/XXXXX/XXXXX==; Expires=Tue, 31 Dec 2040 23:59:59 GMT; Domain=.petrovich.ru; Path=/
+Set-Cookie: ipp_uid=XXXXX/XXXXX/XXXXX==; Expires=Tue, 31 Dec 2040 23:59:59 GMT; Domain=.petrovich.ru; Path=/
+< Content-Type: text/html
+Content-Type: text/html
+< Content-Length: 31253
+Content-Length: 31253
+< Connection: keep-alive
+Connection: keep-alive
+< Keep-Alive: timeout=60
+Keep-Alive: timeout=60
+
+< 
+* Connection #0 to host petrovich.ru left intact
+
+```
+- We challenged ourselves, we kept trying and we didn't only *beat the air*: now, we have something to show you
+  - Do not just take our word for it? - This is great and we respect that: you can build your own `telemt` or download a build and check it right now
 ### Telegram Calls via MTProxy
-- Telegram architecture does **NOT allow calls via MTProxy**, but only via SOCKS5, which cannot be obfuscated
+- Telegram architecture **does NOT allow calls via MTProxy**, but only via SOCKS5, which cannot be obfuscated
 ### How does DPI see MTProxy TLS?
 - DPI sees MTProxy in Fake TLS (ee) mode as TLS 1.3
 - the SNI you specify sends both the client and the server;
@@ -127,11 +365,34 @@ then Ctrl+X -> Y -> Enter to save
 - high entropy, which is normal for AES-encrypted traffic;
 ### Whitelist on IP
 - MTProxy cannot work when there is: 
-  - no IP connectivity to the target host
+  - no IP connectivity to the target host: Russian Whitelist on Mobile Networks - "Белый список"
   - OR all TCP traffic is blocked
-  - OR all TLS traffic is blocked,
+  - OR high entropy/encrypted traffic is blocked: content filters at universities and critical infrastructure
+  - OR all TLS traffic is blocked
+  - OR specified port is blocked: use 443 to make it "like real"
+  - OR provided SNI is blocked: use "officially approved"/innocuous name
 - like most protocols on the Internet; 
-- this situation is observed in China behind the Great Chinese Firewall and in Russia on mobile networks
+- these situations are observed:
+  - in China behind the Great Firewall
+  - in Russia on mobile networks, less in wired networks
+  - in Iran during "activity"
+
+
+## Build
+```bash
+# Cloning repo
+git clone https://github.com/telemt/telemt 
+# Changing Directory to telemt
+cd telemt
+# Starting Release Build
+cargo build --release
+# Move to /bin
+mv ./target/release/telemt /bin
+# Make executable
+chmod +x /bin/telemt
+# Lets go!
+telemt config.toml
+```
 
 ## Why Rust?
 - Long-running reliability and idempotent behavior
@@ -140,6 +401,10 @@ then Ctrl+X -> Y -> Enter to save
 - Memory safety and reduced attack surface
 - Tokio's asynchronous architecture
 
+## Issues
+- ✅ [SOCKS5 as Upstream](https://github.com/telemt/telemt/issues/1) -> added Upstream Management
+- ✅ [iOS - Media Upload Hanging-in-Loop](https://github.com/telemt/telemt/issues/2)
+
 ## Roadmap
 - Public IP in links
 - Config Reload-on-fly

config.toml

@@ -1,13 +1,79 @@
-port = 443
+# === UI ===
+# Users to show in the startup log (tg:// links)
+show_link = ["hello"]
 
-[users]
-user1 = "00000000000000000000000000000000"
+# === General Settings ===
+[general]
+prefer_ipv6 = false
+fast_mode = true
+use_middle_proxy = true
+ad_tag = "00000000000000000000000000000000"
 
-[modes]
-classic = true
-secure = true
+# 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"
+
+[general.modes]
+classic = false
+secure = false
 tls = true
 
-tls_domain = "www.github.com"
-fast_mode = true
-prefer_ipv6 = false
+# === 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_ip = "1.2.3.4" # Optional: Public IP for tg:// links
+
+[[server.listeners]]
+ip = "::"
+
+# === Timeouts (in seconds) ===
+[timeouts]
+client_handshake = 15
+tg_connect = 10
+client_keepalive = 60
+client_ack = 300
+
+# === Anti-Censorship & Masking ===
+[censorship]
+tls_domain = "google.ru"
+mask = true
+mask_port = 443
+# mask_host = "petrovich.ru" # Defaults to tls_domain if not set
+fake_cert_len = 2048
+
+# === Access Control & Users ===
+[access]
+replay_check_len = 65536
+replay_window_secs = 1800
+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 ===
+[[upstreams]]
+type = "direct"
+enabled = true
+weight = 10
+
+# [[upstreams]]
+# type = "socks5"
+# address = "127.0.0.1:9050"
+# enabled = false
+# weight = 1

src/cli.rs

@@ -0,0 +1,300 @@
+//! CLI commands: --init (fire-and-forget setup)
+
+use std::fs;
+use std::path::{Path, PathBuf};
+use std::process::Command;
+use rand::Rng;
+
+/// Options for the init command
+pub struct InitOptions {
+    pub port: u16,
+    pub domain: String,
+    pub secret: Option<String>,
+    pub username: String,
+    pub config_dir: PathBuf,
+    pub no_start: bool,
+}
+
+impl Default for InitOptions {
+    fn default() -> Self {
+        Self {
+            port: 443,
+            domain: "www.google.com".to_string(),
+            secret: None,
+            username: "user".to_string(),
+            config_dir: PathBuf::from("/etc/telemt"),
+            no_start: false,
+        }
+    }
+}
+
+/// Parse --init subcommand options from CLI args.
+///
+/// Returns `Some(InitOptions)` if `--init` was found, `None` otherwise.
+pub fn parse_init_args(args: &[String]) -> Option<InitOptions> {
+    if !args.iter().any(|a| a == "--init") {
+        return None;
+    }
+    
+    let mut opts = InitOptions::default();
+    let mut i = 0;
+    
+    while i < args.len() {
+        match args[i].as_str() {
+            "--port" => {
+                i += 1;
+                if i < args.len() {
+                    opts.port = args[i].parse().unwrap_or(443);
+                }
+            }
+            "--domain" => {
+                i += 1;
+                if i < args.len() {
+                    opts.domain = args[i].clone();
+                }
+            }
+            "--secret" => {
+                i += 1;
+                if i < args.len() {
+                    opts.secret = Some(args[i].clone());
+                }
+            }
+            "--user" => {
+                i += 1;
+                if i < args.len() {
+                    opts.username = args[i].clone();
+                }
+            }
+            "--config-dir" => {
+                i += 1;
+                if i < args.len() {
+                    opts.config_dir = PathBuf::from(&args[i]);
+                }
+            }
+            "--no-start" => {
+                opts.no_start = true;
+            }
+            _ => {}
+        }
+        i += 1;
+    }
+    
+    Some(opts)
+}
+
+/// Run the fire-and-forget setup.
+pub fn run_init(opts: InitOptions) -> Result<(), Box<dyn std::error::Error>> {
+    eprintln!("[telemt] Fire-and-forget setup");
+    eprintln!();
+    
+    // 1. Generate or validate secret
+    let secret = match opts.secret {
+        Some(s) => {
+            if s.len() != 32 || !s.chars().all(|c| c.is_ascii_hexdigit()) {
+                eprintln!("[error] Secret must be exactly 32 hex characters");
+                std::process::exit(1);
+            }
+            s
+        }
+        None => generate_secret(),
+    };
+    
+    eprintln!("[+] Secret: {}", secret);
+    eprintln!("[+] User:   {}", opts.username);
+    eprintln!("[+] Port:   {}", opts.port);
+    eprintln!("[+] Domain: {}", opts.domain);
+    
+    // 2. Create config directory
+    fs::create_dir_all(&opts.config_dir)?;
+    let config_path = opts.config_dir.join("config.toml");
+    
+    // 3. Write config
+    let config_content = generate_config(&opts.username, &secret, opts.port, &opts.domain);
+    fs::write(&config_path, &config_content)?;
+    eprintln!("[+] Config written to {}", config_path.display());
+    
+    // 4. Write systemd unit
+    let exe_path = std::env::current_exe()
+        .unwrap_or_else(|_| PathBuf::from("/usr/local/bin/telemt"));
+    
+    let unit_path = Path::new("/etc/systemd/system/telemt.service");
+    let unit_content = generate_systemd_unit(&exe_path, &config_path);
+    
+    match fs::write(unit_path, &unit_content) {
+        Ok(()) => {
+            eprintln!("[+] Systemd unit written to {}", unit_path.display());
+        }
+        Err(e) => {
+            eprintln!("[!] Cannot write systemd unit (run as root?): {}", e);
+            eprintln!("[!] Manual unit file content:");
+            eprintln!("{}", unit_content);
+            
+            // Still print links and config
+            print_links(&opts.username, &secret, opts.port, &opts.domain);
+            return Ok(());
+        }
+    }
+    
+    // 5. Reload systemd
+    run_cmd("systemctl", &["daemon-reload"]);
+    
+    // 6. Enable service
+    run_cmd("systemctl", &["enable", "telemt.service"]);
+    eprintln!("[+] Service enabled");
+    
+    // 7. Start service (unless --no-start)
+    if !opts.no_start {
+        run_cmd("systemctl", &["start", "telemt.service"]);
+        eprintln!("[+] Service started");
+        
+        // Brief delay then check status
+        std::thread::sleep(std::time::Duration::from_secs(1));
+        let status = Command::new("systemctl")
+            .args(["is-active", "telemt.service"])
+            .output();
+        
+        match status {
+            Ok(out) if out.status.success() => {
+                eprintln!("[+] Service is running");
+            }
+            _ => {
+                eprintln!("[!] Service may not have started correctly");
+                eprintln!("[!] Check: journalctl -u telemt.service -n 20");
+            }
+        }
+    } else {
+        eprintln!("[+] Service not started (--no-start)");
+        eprintln!("[+] Start manually: systemctl start telemt.service");
+    }
+    
+    eprintln!();
+    
+    // 8. Print links
+    print_links(&opts.username, &secret, opts.port, &opts.domain);
+    
+    Ok(())
+}
+
+fn generate_secret() -> String {
+    let mut rng = rand::rng();
+    let bytes: Vec<u8> = (0..16).map(|_| rng.random::<u8>()).collect();
+    hex::encode(bytes)
+}
+
+fn generate_config(username: &str, secret: &str, port: u16, domain: &str) -> String {
+    format!(
+r#"# Telemt MTProxy — auto-generated config
+# Re-run `telemt --init` to regenerate
+
+show_link = ["{username}"]
+
+[general]
+prefer_ipv6 = false
+fast_mode = true
+use_middle_proxy = false
+log_level = "normal"
+
+[general.modes]
+classic = false
+secure = false
+tls = true
+
+[server]
+port = {port}
+listen_addr_ipv4 = "0.0.0.0"
+listen_addr_ipv6 = "::"
+
+[[server.listeners]]
+ip = "0.0.0.0"
+
+[[server.listeners]]
+ip = "::"
+
+[timeouts]
+client_handshake = 15
+tg_connect = 10
+client_keepalive = 60
+client_ack = 300
+
+[censorship]
+tls_domain = "{domain}"
+mask = true
+mask_port = 443
+fake_cert_len = 2048
+
+[access]
+replay_check_len = 65536
+replay_window_secs = 1800
+ignore_time_skew = false
+
+[access.users]
+{username} = "{secret}"
+
+[[upstreams]]
+type = "direct"
+enabled = true
+weight = 10
+"#,
+        username = username,
+        secret = secret,
+        port = port,
+        domain = domain,
+    )
+}
+
+fn generate_systemd_unit(exe_path: &Path, config_path: &Path) -> String {
+    format!(
+r#"[Unit]
+Description=Telemt MTProxy
+Documentation=https://github.com/nicepkg/telemt
+After=network-online.target
+Wants=network-online.target
+
+[Service]
+Type=simple
+ExecStart={exe} {config}
+Restart=always
+RestartSec=5
+LimitNOFILE=65535
+# Security hardening
+NoNewPrivileges=true
+ProtectSystem=strict
+ProtectHome=true
+ReadWritePaths=/etc/telemt
+PrivateTmp=true
+
+[Install]
+WantedBy=multi-user.target
+"#,
+        exe = exe_path.display(),
+        config = config_path.display(),
+    )
+}
+
+fn run_cmd(cmd: &str, args: &[&str]) {
+    match Command::new(cmd).args(args).output() {
+        Ok(output) => {
+            if !output.status.success() {
+                let stderr = String::from_utf8_lossy(&output.stderr);
+                eprintln!("[!] {} {} failed: {}", cmd, args.join(" "), stderr.trim());
+            }
+        }
+        Err(e) => {
+            eprintln!("[!] Failed to run {} {}: {}", cmd, args.join(" "), e);
+        }
+    }
+}
+
+fn print_links(username: &str, secret: &str, port: u16, domain: &str) {
+    let domain_hex = hex::encode(domain);
+    
+    println!("=== Proxy Links ===");
+    println!("[{}]", username);
+    println!("  EE-TLS:  tg://proxy?server=YOUR_SERVER_IP&port={}&secret=ee{}{}", 
+        port, secret, domain_hex);
+    println!();
+    println!("Replace YOUR_SERVER_IP with your server's public IP.");
+    println!("The proxy will auto-detect and display the correct link on startup.");
+    println!("Check: journalctl -u telemt.service | head -30");
+    println!("===================");
+}

src/config/mod.rs

@@ -1,12 +1,89 @@
 //! Configuration
 
 use std::collections::HashMap;
-use std::net::IpAddr;
+use std::net::{IpAddr, SocketAddr};
 use std::path::Path;
 use chrono::{DateTime, Utc};
 use serde::{Deserialize, Serialize};
 use crate::error::{ProxyError, Result};
 
+// ============= Helper Defaults =============
+
+fn default_true() -> bool { true }
+fn default_port() -> u16 { 443 }
+fn default_tls_domain() -> String { "www.google.com".to_string() }
+fn default_mask_port() -> u16 { 443 }
+fn default_replay_check_len() -> usize { 65536 }
+fn default_replay_window_secs() -> u64 { 1800 }
+fn default_handshake_timeout() -> u64 { 15 }
+fn default_connect_timeout() -> u64 { 10 }
+fn default_keepalive() -> u64 { 60 }
+fn default_ack_timeout() -> u64 { 300 }
+fn default_listen_addr() -> String { "0.0.0.0".to_string() }
+fn default_fake_cert_len() -> usize { 2048 }
+fn default_weight() -> u16 { 1 }
+fn default_metrics_whitelist() -> Vec<IpAddr> {
+    vec![
+        "127.0.0.1".parse().unwrap(),
+        "::1".parse().unwrap(),
+    ]
+}
+
+// ============= Log Level =============
+
+/// Logging verbosity level
+#[derive(Debug, Clone, PartialEq, Serialize, Deserialize, Default)]
+#[serde(rename_all = "lowercase")]
+pub enum LogLevel {
+    /// All messages including trace (trace + debug + info + warn + error)
+    Debug,
+    /// Detailed operational logs (debug + info + warn + error)
+    Verbose,
+    /// Standard operational logs (info + warn + error)
+    #[default]
+    Normal,
+    /// Minimal output: only warnings and errors (warn + error).
+    /// Startup messages (config, DC connectivity, proxy links) are always shown
+    /// via info! before the filter is applied.
+    Silent,
+}
+
+impl LogLevel {
+    /// Convert to tracing EnvFilter directive string
+    pub fn to_filter_str(&self) -> &'static str {
+        match self {
+            LogLevel::Debug => "trace",
+            LogLevel::Verbose => "debug",
+            LogLevel::Normal => "info",
+            LogLevel::Silent => "warn",
+        }
+    }
+
+    /// Parse from a loose string (CLI argument)
+    pub fn from_str_loose(s: &str) -> Self {
+        match s.to_lowercase().as_str() {
+            "debug" | "trace" => LogLevel::Debug,
+            "verbose" => LogLevel::Verbose,
+            "normal" | "info" => LogLevel::Normal,
+            "silent" | "quiet" | "error" | "warn" => LogLevel::Silent,
+            _ => LogLevel::Normal,
+        }
+    }
+}
+
+impl std::fmt::Display for LogLevel {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            LogLevel::Debug => write!(f, "debug"),
+            LogLevel::Verbose => write!(f, "verbose"),
+            LogLevel::Normal => write!(f, "normal"),
+            LogLevel::Silent => write!(f, "silent"),
+        }
+    }
+}
+
+// ============= Sub-Configs =============
+
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct ProxyModes {
     #[serde(default)]
@@ -17,8 +94,6 @@ pub struct ProxyModes {
     pub tls: bool,
 }
 
-fn default_true() -> bool { true }
-
 impl Default for ProxyModes {
     fn default() -> Self {
         Self { classic: true, secure: true, tls: true }
@@ -26,31 +101,10 @@ impl Default for ProxyModes {
 }
 
 #[derive(Debug, Clone, Serialize, Deserialize)]
-pub struct ProxyConfig {
-    #[serde(default = "default_port")]
-    pub port: u16,
-    
-    #[serde(default)]
-    pub users: HashMap<String, String>,
-    
-    #[serde(default)]
-    pub ad_tag: Option<String>,
-    
+pub struct GeneralConfig {
     #[serde(default)]
     pub modes: ProxyModes,
     
-    #[serde(default = "default_tls_domain")]
-    pub tls_domain: String,
-    
-    #[serde(default = "default_true")]
-    pub mask: bool,
-    
-    #[serde(default)]
-    pub mask_host: Option<String>,
-    
-    #[serde(default = "default_mask_port")]
-    pub mask_port: u16,
-    
     #[serde(default)]
     pub prefer_ipv6: bool,
     
@@ -61,31 +115,29 @@ pub struct ProxyConfig {
     pub use_middle_proxy: bool,
 
     #[serde(default)]
-    pub user_max_tcp_conns: HashMap<String, usize>,
+    pub ad_tag: Option<String>,
     
     #[serde(default)]
-    pub user_expirations: HashMap<String, DateTime<Utc>>,
+    pub log_level: LogLevel,
+}
 
-    #[serde(default)]
-    pub user_data_quota: HashMap<String, u64>,
+impl Default for GeneralConfig {
+    fn default() -> Self {
+        Self {
+            modes: ProxyModes::default(),
+            prefer_ipv6: false,
+            fast_mode: true,
+            use_middle_proxy: false,
+            ad_tag: None,
+            log_level: LogLevel::Normal,
+        }
+    }
+}
 
-    #[serde(default = "default_replay_check_len")]
-    pub replay_check_len: usize,
-    
-    #[serde(default)]
-    pub ignore_time_skew: bool,
-    
-    #[serde(default = "default_handshake_timeout")]
-    pub client_handshake_timeout: u64,
-    
-    #[serde(default = "default_connect_timeout")]
-    pub tg_connect_timeout: u64,
-    
-    #[serde(default = "default_keepalive")]
-    pub client_keepalive: u64,
-    
-    #[serde(default = "default_ack_timeout")]
-    pub client_ack_timeout: u64,
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct ServerConfig {
+    #[serde(default = "default_port")]
+    pub port: u16,
 
     #[serde(default = "default_listen_addr")]
     pub listen_addr_ipv4: String,
@@ -102,64 +154,205 @@ pub struct ProxyConfig {
     #[serde(default = "default_metrics_whitelist")]
     pub metrics_whitelist: Vec<IpAddr>,
 
-    #[serde(default = "default_fake_cert_len")]
-    pub fake_cert_len: usize,
+    #[serde(default)]
+    pub listeners: Vec<ListenerConfig>,
 }
 
-fn default_port() -> u16 { 443 }
-fn default_tls_domain() -> String { "www.google.com".to_string() }
-fn default_mask_port() -> u16 { 443 }
-fn default_replay_check_len() -> usize { 65536 }
-fn default_handshake_timeout() -> u64 { 10 }
-fn default_connect_timeout() -> u64 { 10 }
-fn default_keepalive() -> u64 { 600 }
-fn default_ack_timeout() -> u64 { 300 }
-fn default_listen_addr() -> String { "0.0.0.0".to_string() }
-fn default_fake_cert_len() -> usize { 2048 }
-
-fn default_metrics_whitelist() -> Vec<IpAddr> {
-    vec![
-        "127.0.0.1".parse().unwrap(),
-        "::1".parse().unwrap(),
-    ]
-}
-
-impl Default for ProxyConfig {
+impl Default for ServerConfig {
     fn default() -> Self {
-        let mut users = HashMap::new();
-        users.insert("default".to_string(), "00000000000000000000000000000000".to_string());
-        
         Self {
             port: default_port(),
-            users,
-            ad_tag: None,
-            modes: ProxyModes::default(),
-            tls_domain: default_tls_domain(),
-            mask: true,
-            mask_host: None,
-            mask_port: default_mask_port(),
-            prefer_ipv6: false,
-            fast_mode: true,
-            use_middle_proxy: false,
-            user_max_tcp_conns: HashMap::new(),
-            user_expirations: HashMap::new(),
-            user_data_quota: HashMap::new(),
-            replay_check_len: default_replay_check_len(),
-            ignore_time_skew: false,
-            client_handshake_timeout: default_handshake_timeout(),
-            tg_connect_timeout: default_connect_timeout(),
-            client_keepalive: default_keepalive(),
-            client_ack_timeout: default_ack_timeout(),
             listen_addr_ipv4: default_listen_addr(),
             listen_addr_ipv6: Some("::".to_string()),
             listen_unix_sock: None,
             metrics_port: None,
             metrics_whitelist: default_metrics_whitelist(),
+            listeners: Vec::new(),
+        }
+    }
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct TimeoutsConfig {
+    #[serde(default = "default_handshake_timeout")]
+    pub client_handshake: u64,
+    
+    #[serde(default = "default_connect_timeout")]
+    pub tg_connect: u64,
+    
+    #[serde(default = "default_keepalive")]
+    pub client_keepalive: u64,
+    
+    #[serde(default = "default_ack_timeout")]
+    pub client_ack: u64,
+}
+
+impl Default for TimeoutsConfig {
+    fn default() -> Self {
+        Self {
+            client_handshake: default_handshake_timeout(),
+            tg_connect: default_connect_timeout(),
+            client_keepalive: default_keepalive(),
+            client_ack: default_ack_timeout(),
+        }
+    }
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct AntiCensorshipConfig {
+    #[serde(default = "default_tls_domain")]
+    pub tls_domain: String,
+    
+    #[serde(default = "default_true")]
+    pub mask: bool,
+    
+    #[serde(default)]
+    pub mask_host: Option<String>,
+    
+    #[serde(default = "default_mask_port")]
+    pub mask_port: u16,
+
+    #[serde(default = "default_fake_cert_len")]
+    pub fake_cert_len: usize,
+}
+
+impl Default for AntiCensorshipConfig {
+    fn default() -> Self {
+        Self {
+            tls_domain: default_tls_domain(),
+            mask: true,
+            mask_host: None,
+            mask_port: default_mask_port(),
             fake_cert_len: default_fake_cert_len(),
         }
     }
 }
 
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct AccessConfig {
+    #[serde(default)]
+    pub users: HashMap<String, String>,
+
+    #[serde(default)]
+    pub user_max_tcp_conns: HashMap<String, usize>,
+    
+    #[serde(default)]
+    pub user_expirations: HashMap<String, DateTime<Utc>>,
+    
+    #[serde(default)]
+    pub user_data_quota: HashMap<String, u64>,
+
+    #[serde(default = "default_replay_check_len")]
+    pub replay_check_len: usize,
+    
+    #[serde(default = "default_replay_window_secs")]
+    pub replay_window_secs: u64,
+    
+    #[serde(default)]
+    pub ignore_time_skew: bool,
+}
+
+impl Default for AccessConfig {
+    fn default() -> Self {
+        let mut users = HashMap::new();
+        users.insert("default".to_string(), "00000000000000000000000000000000".to_string());
+        Self {
+            users,
+            user_max_tcp_conns: HashMap::new(),
+            user_expirations: HashMap::new(),
+            user_data_quota: HashMap::new(),
+            replay_check_len: default_replay_check_len(),
+            replay_window_secs: default_replay_window_secs(),
+            ignore_time_skew: false,
+        }
+    }
+}
+
+// ============= Aux Structures =============
+
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
+#[serde(tag = "type", rename_all = "lowercase")]
+pub enum UpstreamType {
+    Direct {
+        #[serde(default)]
+        interface: Option<String>,
+    },
+    Socks4 {
+        address: String,
+        #[serde(default)]
+        interface: Option<String>,
+        #[serde(default)]
+        user_id: Option<String>,
+    },
+    Socks5 {
+        address: String,
+        #[serde(default)]
+        interface: Option<String>,
+        #[serde(default)]
+        username: Option<String>,
+        #[serde(default)]
+        password: Option<String>,
+    },
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct UpstreamConfig {
+    #[serde(flatten)]
+    pub upstream_type: UpstreamType,
+    #[serde(default = "default_weight")]
+    pub weight: u16,
+    #[serde(default = "default_true")]
+    pub enabled: bool,
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct ListenerConfig {
+    pub ip: IpAddr,
+    #[serde(default)]
+    pub announce_ip: Option<IpAddr>,
+}
+
+// ============= Main Config =============
+
+#[derive(Debug, Clone, Serialize, Deserialize, Default)]
+pub struct ProxyConfig {
+    #[serde(default)]
+    pub general: GeneralConfig,
+
+    #[serde(default)]
+    pub server: ServerConfig,
+
+    #[serde(default)]
+    pub timeouts: TimeoutsConfig,
+
+    #[serde(default)]
+    pub censorship: AntiCensorshipConfig,
+
+    #[serde(default)]
+    pub access: AccessConfig,
+
+    #[serde(default)]
+    pub upstreams: Vec<UpstreamConfig>,
+
+    #[serde(default)]
+    pub show_link: Vec<String>,
+
+    /// DC address overrides for non-standard DCs (CDN, media, test, etc.)
+    /// Keys are DC indices as strings, values are "ip:port" addresses.
+    /// Matches the C implementation's `proxy_for <dc_id> <ip>:<port>` config directive.
+    /// Example in config.toml:
+    ///   [dc_overrides]
+    ///   "203" = "149.154.175.100:443"
+    #[serde(default)]
+    pub dc_overrides: HashMap<String, String>,
+
+    /// Default DC index (1-5) for unmapped non-standard DCs.
+    /// Matches the C implementation's `default <dc_id>` config directive.
+    /// If not set, defaults to 2 (matching Telegram's official `default 2;` in proxy-multi.conf).
+    #[serde(default)]
+    pub default_dc: Option<u8>,
+}
+
 impl ProxyConfig {
     pub fn load<P: AsRef<Path>>(path: P) -> Result<Self> {
         let content = std::fs::read_to_string(path)
@@ -169,7 +362,7 @@ impl ProxyConfig {
             .map_err(|e| ProxyError::Config(e.to_string()))?;
         
         // Validate secrets
-        for (user, secret) in &config.users {
+        for (user, secret) in &config.access.users {
             if !secret.chars().all(|c| c.is_ascii_hexdigit()) || secret.len() != 32 {
                 return Err(ProxyError::InvalidSecret {
                     user: user.clone(),
@@ -178,50 +371,65 @@ impl ProxyConfig {
             }
         }
         
-        // Default mask_host
-        if config.mask_host.is_none() {
-            config.mask_host = Some(config.tls_domain.clone());
+        // Validate tls_domain
+        if config.censorship.tls_domain.is_empty() {
+            return Err(ProxyError::Config("tls_domain cannot be empty".to_string()));
+        }
+        
+        // Default mask_host to tls_domain if not set
+        if config.censorship.mask_host.is_none() {
+            config.censorship.mask_host = Some(config.censorship.tls_domain.clone());
         }
         
         // Random fake_cert_len
         use rand::Rng;
-        config.fake_cert_len = rand::thread_rng().gen_range(1024..4096);
+        config.censorship.fake_cert_len = rand::rng().gen_range(1024..4096);
+        
+        // Migration: Populate listeners if empty
+        if config.server.listeners.is_empty() {
+            if let Ok(ipv4) = config.server.listen_addr_ipv4.parse::<IpAddr>() {
+                config.server.listeners.push(ListenerConfig {
+                    ip: ipv4,
+                    announce_ip: None,
+                });
+            }
+            if let Some(ipv6_str) = &config.server.listen_addr_ipv6 {
+                 if let Ok(ipv6) = ipv6_str.parse::<IpAddr>() {
+                    config.server.listeners.push(ListenerConfig {
+                        ip: ipv6,
+                        announce_ip: None,
+                    });
+                }
+            }
+        }
+
+        // Migration: Populate upstreams if empty (Default Direct)
+        if config.upstreams.is_empty() {
+             config.upstreams.push(UpstreamConfig {
+                upstream_type: UpstreamType::Direct { interface: None },
+                weight: 1,
+                enabled: true,
+            });
+        }
         
         Ok(config)
     }
     
     pub fn validate(&self) -> Result<()> {
-        if self.users.is_empty() {
+        if self.access.users.is_empty() {
             return Err(ProxyError::Config("No users configured".to_string()));
         }
         
-        if !self.modes.classic && !self.modes.secure && !self.modes.tls {
+        if !self.general.modes.classic && !self.general.modes.secure && !self.general.modes.tls {
             return Err(ProxyError::Config("No modes enabled".to_string()));
         }
         
+        if self.censorship.tls_domain.contains(' ') || self.censorship.tls_domain.contains('/') {
+            return Err(ProxyError::Config(
+                format!("Invalid tls_domain: '{}'. Must be a valid domain name", self.censorship.tls_domain)
+            ));
+        }
+        
         Ok(())
     }
 }
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-    
-    #[test]
-    fn test_default_config() {
-        let config = ProxyConfig::default();
-        assert_eq!(config.port, 443);
-        assert!(config.modes.tls);
-        assert_eq!(config.client_keepalive, 600);
-        assert_eq!(config.client_ack_timeout, 300);
-    }
-    
-    #[test]
-    fn test_config_validate() {
-        let mut config = ProxyConfig::default();
-        assert!(config.validate().is_ok());
-        
-        config.users.clear();
-        assert!(config.validate().is_err());
-    }
-}

src/crypto/aes.rs

@@ -1,9 +1,19 @@
 //! AES encryption implementations
 //!
 //! Provides AES-256-CTR and AES-256-CBC modes for MTProto encryption.
+//!
+//! ## Zeroize policy
+//!
+//! - `AesCbc` stores raw key/IV bytes and zeroizes them on drop.
+//! - `AesCtr` wraps an opaque `Aes256Ctr` cipher from the `ctr` crate.
+//!   The expanded key schedule lives inside that type and cannot be
+//!   zeroized from outside. Callers that hold raw key material (e.g.
+//!   `HandshakeSuccess`, `ObfuscationParams`) are responsible for
+//!   zeroizing their own copies.
 
 use aes::Aes256;
 use ctr::{Ctr128BE, cipher::{KeyIvInit, StreamCipher}};
+use zeroize::Zeroize;
 use crate::error::{ProxyError, Result};
 
 type Aes256Ctr = Ctr128BE<Aes256>;
@@ -12,7 +22,12 @@ type Aes256Ctr = Ctr128BE<Aes256>;
 
 /// AES-256-CTR encryptor/decryptor
 /// 
-/// CTR mode is symmetric - encryption and decryption are the same operation.
+/// CTR mode is symmetric — encryption and decryption are the same operation.
+///
+/// **Zeroize note:** The inner `Aes256Ctr` cipher state (expanded key schedule
+/// + counter) is opaque and cannot be zeroized. If you need to protect key
+/// material, zeroize the `[u8; 32]` key and `u128` IV at the call site
+/// before dropping them.
 pub struct AesCtr {
     cipher: Aes256Ctr,
 }
@@ -62,14 +77,23 @@ impl AesCtr {
 
 /// AES-256-CBC cipher with proper chaining
 ///
-/// Unlike CTR mode, CBC is NOT symmetric - encryption and decryption
+/// Unlike CTR mode, CBC is NOT symmetric — encryption and decryption
 /// are different operations. This implementation handles CBC chaining
 /// correctly across multiple blocks.
+///
+/// Key and IV are zeroized on drop.
 pub struct AesCbc {
     key: [u8; 32],
     iv: [u8; 16],
 }
 
+impl Drop for AesCbc {
+    fn drop(&mut self) {
+        self.key.zeroize();
+        self.iv.zeroize();
+    }
+}
+
 impl AesCbc {
     /// AES block size
     const BLOCK_SIZE: usize = 16;
@@ -141,17 +165,9 @@ impl AesCbc {
         
         for chunk in data.chunks(Self::BLOCK_SIZE) {
             let plaintext: [u8; 16] = chunk.try_into().unwrap();
-            
-            // XOR plaintext with previous ciphertext (or IV for first block)
             let xored = Self::xor_blocks(&plaintext, &prev_ciphertext);
-            
-            // Encrypt the XORed block
             let ciphertext = self.encrypt_block(&xored, &key_schedule);
-            
-            // Save for next iteration
             prev_ciphertext = ciphertext;
-            
-            // Append to result
             result.extend_from_slice(&ciphertext);
         }
         
@@ -180,17 +196,9 @@ impl AesCbc {
         
         for chunk in data.chunks(Self::BLOCK_SIZE) {
             let ciphertext: [u8; 16] = chunk.try_into().unwrap();
-            
-            // Decrypt the block
             let decrypted = self.decrypt_block(&ciphertext, &key_schedule);
-            
-            // XOR with previous ciphertext (or IV for first block)
             let plaintext = Self::xor_blocks(&decrypted, &prev_ciphertext);
-            
-            // Save current ciphertext for next iteration
             prev_ciphertext = ciphertext;
-            
-            // Append to result
             result.extend_from_slice(&plaintext);
         }
         
@@ -217,16 +225,13 @@ impl AesCbc {
         for i in (0..data.len()).step_by(Self::BLOCK_SIZE) {
             let block = &mut data[i..i + Self::BLOCK_SIZE];
             
-            // XOR with previous ciphertext
             for j in 0..Self::BLOCK_SIZE {
                 block[j] ^= prev_ciphertext[j];
             }
             
-            // Encrypt in-place
             let block_array: &mut [u8; 16] = block.try_into().unwrap();
             *block_array = self.encrypt_block(block_array, &key_schedule);
             
-            // Save for next iteration
             prev_ciphertext = *block_array;
         }
         
@@ -248,26 +253,20 @@ impl AesCbc {
         use aes::cipher::KeyInit;
         let key_schedule = aes::Aes256::new((&self.key).into());
         
-        // For in-place decryption, we need to save ciphertext blocks
-        // before we overwrite them
         let mut prev_ciphertext = self.iv;
         
         for i in (0..data.len()).step_by(Self::BLOCK_SIZE) {
             let block = &mut data[i..i + Self::BLOCK_SIZE];
             
-            // Save current ciphertext before modifying
             let current_ciphertext: [u8; 16] = block.try_into().unwrap();
             
-            // Decrypt in-place
             let block_array: &mut [u8; 16] = block.try_into().unwrap();
             *block_array = self.decrypt_block(block_array, &key_schedule);
             
-            // XOR with previous ciphertext
             for j in 0..Self::BLOCK_SIZE {
                 block[j] ^= prev_ciphertext[j];
             }
             
-            // Save for next iteration
             prev_ciphertext = current_ciphertext;
         }
         
@@ -347,10 +346,8 @@ mod tests {
         let mut cipher = AesCtr::new(&key, iv);
         cipher.apply(&mut data);
         
-        // Encrypted should be different
         assert_ne!(&data[..], original);
         
-        // Decrypt with fresh cipher
         let mut cipher = AesCtr::new(&key, iv);
         cipher.apply(&mut data);
         
@@ -364,7 +361,7 @@ mod tests {
         let key = [0u8; 32];
         let iv = [0u8; 16];
         
-        let original = [0u8; 32]; // 2 blocks
+        let original = [0u8; 32];
         
         let cipher = AesCbc::new(key, iv);
         let encrypted = cipher.encrypt(&original).unwrap();
@@ -375,31 +372,25 @@ mod tests {
     
     #[test]
     fn test_aes_cbc_chaining_works() {
-        // This is the key test - verify CBC chaining is correct
         let key = [0x42u8; 32];
         let iv = [0x00u8; 16];
         
-        // Two IDENTICAL plaintext blocks
         let plaintext = [0xAAu8; 32];
         
         let cipher = AesCbc::new(key, iv);
         let ciphertext = cipher.encrypt(&plaintext).unwrap();
         
-        // With proper CBC, identical plaintext blocks produce DIFFERENT ciphertext
         let block1 = &ciphertext[0..16];
         let block2 = &ciphertext[16..32];
         
         assert_ne!(
             block1, block2,
-            "CBC chaining broken: identical plaintext blocks produced identical ciphertext. \
-             This indicates ECB mode, not CBC!"
+            "CBC chaining broken: identical plaintext blocks produced identical ciphertext"
         );
     }
     
     #[test]
     fn test_aes_cbc_known_vector() {
-        // Test with known NIST test vector
-        // AES-256-CBC with zero key and zero IV
         let key = [0u8; 32];
         let iv = [0u8; 16];
         let plaintext = [0u8; 16];
@@ -407,11 +398,9 @@ mod tests {
         let cipher = AesCbc::new(key, iv);
         let ciphertext = cipher.encrypt(&plaintext).unwrap();
         
-        // Decrypt and verify roundtrip
         let decrypted = cipher.decrypt(&ciphertext).unwrap();
         assert_eq!(plaintext.as_slice(), decrypted.as_slice());
         
-        // Ciphertext should not be all zeros
         assert_ne!(ciphertext.as_slice(), plaintext.as_slice());
     }
     
@@ -420,7 +409,6 @@ mod tests {
         let key = [0x12u8; 32];
         let iv = [0x34u8; 16];
         
-        // 5 blocks = 80 bytes
         let plaintext: Vec<u8> = (0..80).collect();
         
         let cipher = AesCbc::new(key, iv);
@@ -435,7 +423,7 @@ mod tests {
         let key = [0x12u8; 32];
         let iv = [0x34u8; 16];
         
-        let original = [0x56u8; 48]; // 3 blocks
+        let original = [0x56u8; 48];
         let mut buffer = original;
         
         let cipher = AesCbc::new(key, iv);
@@ -462,41 +450,33 @@ mod tests {
     fn test_aes_cbc_unaligned_error() {
         let cipher = AesCbc::new([0u8; 32], [0u8; 16]);
         
-        // 15 bytes - not aligned to block size
         let result = cipher.encrypt(&[0u8; 15]);
         assert!(result.is_err());
         
-        // 17 bytes - not aligned
         let result = cipher.encrypt(&[0u8; 17]);
         assert!(result.is_err());
     }
     
     #[test]
     fn test_aes_cbc_avalanche_effect() {
-        // Changing one bit in plaintext should change entire ciphertext block
-        // and all subsequent blocks (due to chaining)
         let key = [0xAB; 32];
         let iv = [0xCD; 16];
         
-        let mut plaintext1 = [0u8; 32];
+        let plaintext1 = [0u8; 32];
         let mut plaintext2 = [0u8; 32];
-        plaintext2[0] = 0x01; // Single bit difference in first block
+        plaintext2[0] = 0x01;
         
         let cipher = AesCbc::new(key, iv);
         
         let ciphertext1 = cipher.encrypt(&plaintext1).unwrap();
         let ciphertext2 = cipher.encrypt(&plaintext2).unwrap();
         
-        // First blocks should be different
         assert_ne!(&ciphertext1[0..16], &ciphertext2[0..16]);
-        
-        // Second blocks should ALSO be different (chaining effect)
         assert_ne!(&ciphertext1[16..32], &ciphertext2[16..32]);
     }
     
     #[test]
     fn test_aes_cbc_iv_matters() {
-        // Same plaintext with different IVs should produce different ciphertext
         let key = [0x55; 32];
         let plaintext = [0x77u8; 16];
         
@@ -511,7 +491,6 @@ mod tests {
     
     #[test]
     fn test_aes_cbc_deterministic() {
-        // Same key, IV, plaintext should always produce same ciphertext
         let key = [0x99; 32];
         let iv = [0x88; 16];
         let plaintext = [0x77u8; 32];
@@ -524,6 +503,23 @@ mod tests {
         assert_eq!(ciphertext1, ciphertext2);
     }
     
+    // ============= Zeroize Tests =============
+    
+    #[test]
+    fn test_aes_cbc_zeroize_on_drop() {
+        let key = [0xAA; 32];
+        let iv = [0xBB; 16];
+        
+        let cipher = AesCbc::new(key, iv);
+        // Verify key/iv are set
+        assert_eq!(cipher.key, [0xAA; 32]);
+        assert_eq!(cipher.iv, [0xBB; 16]);
+        
+        drop(cipher);
+        // After drop, key/iv are zeroized (can't observe directly,
+        // but the Drop impl runs without panic)
+    }
+    
     // ============= Error Handling Tests =============
     
     #[test]

src/crypto/hash.rs

@@ -1,3 +1,16 @@
+//! Cryptographic hash functions
+//!
+//! ## Protocol-required algorithms
+//!
+//! This module exposes MD5 and SHA-1 alongside SHA-256. These weaker
+//! hash functions are **required by the Telegram Middle Proxy protocol**
+//! (`derive_middleproxy_keys`) and cannot be replaced without breaking
+//! compatibility. They are NOT used for any security-sensitive purpose
+//! outside of that specific key derivation scheme mandated by Telegram.
+//!
+//! Static analysis tools (CodeQL, cargo-audit) may flag them — the
+//! usages are intentional and protocol-mandated.
+
 use hmac::{Hmac, Mac};
 use sha2::Sha256;
 use md5::Md5;
@@ -21,14 +34,16 @@ pub fn sha256_hmac(key: &[u8], data: &[u8]) -> [u8; 32] {
     mac.finalize().into_bytes().into()
 }
 
-/// SHA-1
+/// SHA-1 — **protocol-required** by Telegram Middle Proxy key derivation.
+/// Not used for general-purpose hashing.
 pub fn sha1(data: &[u8]) -> [u8; 20] {
     let mut hasher = Sha1::new();
     hasher.update(data);
     hasher.finalize().into()
 }
 
-/// MD5
+/// MD5 — **protocol-required** by Telegram Middle Proxy key derivation.
+/// Not used for general-purpose hashing.
 pub fn md5(data: &[u8]) -> [u8; 16] {
     let mut hasher = Md5::new();
     hasher.update(data);
@@ -40,7 +55,11 @@ pub fn crc32(data: &[u8]) -> u32 {
     crc32fast::hash(data)
 }
 
-/// Middle Proxy Keygen
+/// Middle Proxy key derivation
+///
+/// Uses MD5 + SHA-1 as mandated by the Telegram Middle Proxy protocol.
+/// These algorithms are NOT replaceable here — changing them would break
+/// interoperability with Telegram's middle proxy infrastructure.
 pub fn derive_middleproxy_keys(
     nonce_srv: &[u8; 16],
     nonce_clt: &[u8; 16],

src/crypto/mod.rs

@@ -6,4 +6,4 @@ pub mod random;
 
 pub use aes::{AesCtr, AesCbc};
 pub use hash::{sha256, sha256_hmac, sha1, md5, crc32};
-pub use random::{SecureRandom, SECURE_RANDOM};
+pub use random::SecureRandom;

src/crypto/random.rs

@@ -3,11 +3,8 @@
 use rand::{Rng, RngCore, SeedableRng};
 use rand::rngs::StdRng;
 use parking_lot::Mutex;
+use zeroize::Zeroize;
 use crate::crypto::AesCtr;
-use once_cell::sync::Lazy;
-
-/// Global secure random instance
-pub static SECURE_RANDOM: Lazy<SecureRandom> = Lazy::new(SecureRandom::new);
 
 /// Cryptographically secure PRNG with AES-CTR
 pub struct SecureRandom {
@@ -20,18 +17,30 @@ struct SecureRandomInner {
     buffer: Vec<u8>,
 }
 
+impl Drop for SecureRandomInner {
+    fn drop(&mut self) {
+        self.buffer.zeroize();
+    }
+}
+
 impl SecureRandom {
     pub fn new() -> Self {
-        let mut rng = StdRng::from_entropy();
+        let mut seed_source = rand::rng();
+        let mut rng = StdRng::from_rng(&mut seed_source);
         
         let mut key = [0u8; 32];
         rng.fill_bytes(&mut key);
-        let iv: u128 = rng.gen();
+        let iv: u128 = rng.random();
+        
+        let cipher = AesCtr::new(&key, iv);
+        
+        // Zeroize local key copy — cipher already consumed it
+        key.zeroize();
         
         Self {
             inner: Mutex::new(SecureRandomInner {
                 rng,
-                cipher: AesCtr::new(&key, iv),
+                cipher,
                 buffer: Vec::with_capacity(1024),
             }),
         }
@@ -78,7 +87,6 @@ impl SecureRandom {
             result |= (b as u64) << (i * 8);
         }
         
-        // Mask extra bits
         if k < 64 {
             result &= (1u64 << k) - 1;
         }
@@ -107,13 +115,13 @@ impl SecureRandom {
     /// Generate random u32
     pub fn u32(&self) -> u32 {
         let mut inner = self.inner.lock();
-        inner.rng.gen()
+        inner.rng.random()
     }
     
     /// Generate random u64
     pub fn u64(&self) -> u64 {
         let mut inner = self.inner.lock();
-        inner.rng.gen()
+        inner.rng.random()
     }
 }
 
@@ -162,12 +170,10 @@ mod tests {
     fn test_bits() {
         let rng = SecureRandom::new();
         
-        // Single bit should be 0 or 1
         for _ in 0..100 {
             assert!(rng.bits(1) <= 1);
         }
         
-        // 8 bits should be 0-255
         for _ in 0..100 {
             assert!(rng.bits(8) <= 255);
         }
@@ -185,10 +191,8 @@ mod tests {
             }
         }
         
-        // Should have seen all items
         assert_eq!(seen.len(), 5);
         
-        // Empty slice should return None
         let empty: Vec<i32> = vec![];
         assert!(rng.choose(&empty).is_none());
     }
@@ -201,12 +205,10 @@ mod tests {
         let mut shuffled = original.clone();
         rng.shuffle(&mut shuffled);
         
-        // Should contain same elements
         let mut sorted = shuffled.clone();
         sorted.sort();
         assert_eq!(sorted, original);
         
-        // Should be different order (with very high probability)
         assert_ne!(shuffled, original);
     }
 }

src/error.rs

@@ -118,16 +118,13 @@ pub trait Recoverable {
 impl Recoverable for StreamError {
     fn is_recoverable(&self) -> bool {
         match self {
-            // Partial operations can be retried
             Self::PartialRead { .. } | Self::PartialWrite { .. } => true,
-            // I/O errors depend on kind
             Self::Io(e) => matches!(
                 e.kind(),
                 std::io::ErrorKind::WouldBlock 
                 | std::io::ErrorKind::Interrupted
                 | std::io::ErrorKind::TimedOut
             ),
-            // These are not recoverable
             Self::Poisoned { .. } 
             | Self::BufferOverflow { .. }
             | Self::InvalidFrame { .. }
@@ -137,13 +134,9 @@ impl Recoverable for StreamError {
     
     fn can_continue(&self) -> bool {
         match self {
-            // Poisoned stream cannot be used
             Self::Poisoned { .. } => false,
-            // EOF means stream is done
             Self::UnexpectedEof => false,
-            // Buffer overflow is fatal
             Self::BufferOverflow { .. } => false,
-            // Others might allow continuation
             _ => true,
         }
     }
@@ -235,6 +228,9 @@ pub enum ProxyError {
     #[error("Invalid proxy protocol header")]
     InvalidProxyProtocol,
     
+    #[error("Proxy error: {0}")]
+    Proxy(String),
+    
     // ============= Config Errors =============
     
     #[error("Config error: {0}")]
@@ -294,16 +290,16 @@ pub type StreamResult<T> = std::result::Result<T, StreamError>;
 
 /// Result with optional bad client handling
 #[derive(Debug)]
-pub enum HandshakeResult<T> {
+pub enum HandshakeResult<T, R, W> {
     /// Handshake succeeded
     Success(T),
-    /// Client failed validation, needs masking
-    BadClient,
+    /// Client failed validation, needs masking. Returns ownership of streams.
+    BadClient { reader: R, writer: W },
     /// Error occurred
     Error(ProxyError),
 }
 
-impl<T> HandshakeResult<T> {
+impl<T, R, W> HandshakeResult<T, R, W> {
     /// Check if successful
     pub fn is_success(&self) -> bool {
         matches!(self, HandshakeResult::Success(_))
@@ -311,49 +307,32 @@ impl<T> HandshakeResult<T> {
     
     /// Check if bad client
     pub fn is_bad_client(&self) -> bool {
-        matches!(self, HandshakeResult::BadClient)
-    }
-    
-    /// Convert to Result, treating BadClient as error
-    pub fn into_result(self) -> Result<T> {
-        match self {
-            HandshakeResult::Success(v) => Ok(v),
-            HandshakeResult::BadClient => Err(ProxyError::InvalidHandshake("Bad client".into())),
-            HandshakeResult::Error(e) => Err(e),
-        }
+        matches!(self, HandshakeResult::BadClient { .. })
     }
     
     /// Map the success value
-    pub fn map<U, F: FnOnce(T) -> U>(self, f: F) -> HandshakeResult<U> {
+    pub fn map<U, F: FnOnce(T) -> U>(self, f: F) -> HandshakeResult<U, R, W> {
         match self {
             HandshakeResult::Success(v) => HandshakeResult::Success(f(v)),
-            HandshakeResult::BadClient => HandshakeResult::BadClient,
+            HandshakeResult::BadClient { reader, writer } => HandshakeResult::BadClient { reader, writer },
             HandshakeResult::Error(e) => HandshakeResult::Error(e),
         }
     }
-    
-    /// Convert success to Option
-    pub fn ok(self) -> Option<T> {
-        match self {
-            HandshakeResult::Success(v) => Some(v),
-            _ => None,
-        }
-    }
 }
 
-impl<T> From<ProxyError> for HandshakeResult<T> {
+impl<T, R, W> From<ProxyError> for HandshakeResult<T, R, W> {
     fn from(err: ProxyError) -> Self {
         HandshakeResult::Error(err)
     }
 }
 
-impl<T> From<std::io::Error> for HandshakeResult<T> {
+impl<T, R, W> From<std::io::Error> for HandshakeResult<T, R, W> {
     fn from(err: std::io::Error) -> Self {
         HandshakeResult::Error(ProxyError::Io(err))
     }
 }
 
-impl<T> From<StreamError> for HandshakeResult<T> {
+impl<T, R, W> From<StreamError> for HandshakeResult<T, R, W> {
     fn from(err: StreamError) -> Self {
         HandshakeResult::Error(ProxyError::Stream(err))
     }
@@ -397,18 +376,18 @@ mod tests {
     
     #[test]
     fn test_handshake_result() {
-        let success: HandshakeResult<i32> = HandshakeResult::Success(42);
+        let success: HandshakeResult<i32, (), ()> = HandshakeResult::Success(42);
         assert!(success.is_success());
         assert!(!success.is_bad_client());
         
-        let bad: HandshakeResult<i32> = HandshakeResult::BadClient;
+        let bad: HandshakeResult<i32, (), ()> = HandshakeResult::BadClient { reader: (), writer: () };
         assert!(!bad.is_success());
         assert!(bad.is_bad_client());
     }
     
     #[test]
     fn test_handshake_result_map() {
-        let success: HandshakeResult<i32> = HandshakeResult::Success(42);
+        let success: HandshakeResult<i32, (), ()> = HandshakeResult::Success(42);
         let mapped = success.map(|x| x * 2);
         
         match mapped {

src/main.rs

@@ -1,158 +1,306 @@
 //! Telemt - MTProxy on Rust
 
-use std::sync::Arc;
 use std::net::SocketAddr;
+use std::sync::Arc;
+use std::time::Duration;
 use tokio::net::TcpListener;
 use tokio::signal;
-use tracing::{info, error, Level};
-use tracing_subscriber::{FmtSubscriber, EnvFilter};
+use tokio::sync::Semaphore;
+use tracing::{info, error, warn, debug};
+use tracing_subscriber::{fmt, EnvFilter, reload, prelude::*};
 
-mod error;
+mod cli;
+mod config;
 mod crypto;
+mod error;
 mod protocol;
+mod proxy;
+mod stats;
 mod stream;
 mod transport;
-mod proxy;
-mod config;
-mod stats;
 mod util;
 
-use config::ProxyConfig;
-use stats::{Stats, ReplayChecker};
-use transport::ConnectionPool;
-use proxy::ClientHandler;
+use crate::config::{ProxyConfig, LogLevel};
+use crate::proxy::ClientHandler;
+use crate::stats::{Stats, ReplayChecker};
+use crate::crypto::SecureRandom;
+use crate::transport::{create_listener, ListenOptions, UpstreamManager};
+use crate::util::ip::detect_ip;
+use crate::stream::BufferPool;
 
-#[tokio::main]
-async fn main() -> std::result::Result<(), Box<dyn std::error::Error>> {
-    // Initialize logging with env filter
-    // Use RUST_LOG=debug or RUST_LOG=trace for more details
-    let filter = EnvFilter::try_from_default_env()
-        .unwrap_or_else(|_| EnvFilter::new("info"));
+fn parse_cli() -> (String, bool, Option<String>) {
+    let mut config_path = "config.toml".to_string();
+    let mut silent = false;
+    let mut log_level: Option<String> = None;
     
-    let subscriber = FmtSubscriber::builder()
-        .with_env_filter(filter)
-        .with_target(true)
-        .with_thread_ids(false)
-        .with_file(false)
-        .with_line_number(false)
-        .finish();
+    let args: Vec<String> = std::env::args().skip(1).collect();
     
-    tracing::subscriber::set_global_default(subscriber)?;
-    
-    // Load configuration
-    let config_path = std::env::args()
-        .nth(1)
-        .unwrap_or_else(|| "config.toml".to_string());
-    
-    info!("Loading configuration from {}", config_path);
-    
-    let config = ProxyConfig::load(&config_path).unwrap_or_else(|e| {
-        error!("Failed to load config: {}", e);
-        info!("Using default configuration");
-        ProxyConfig::default()
-    });
-    
-    if let Err(e) = config.validate() {
-        error!("Invalid configuration: {}", e);
+    // Check for --init first (handled before tokio)
+    if let Some(init_opts) = cli::parse_init_args(&args) {
+        if let Err(e) = cli::run_init(init_opts) {
+            eprintln!("[telemt] Init failed: {}", e);
             std::process::exit(1);
         }
-    
-    let config = Arc::new(config);
-    
-    info!("Starting MTProto Proxy on port {}", config.port);
-    info!("Fast mode: {}", config.fast_mode);
-    info!("Modes: classic={}, secure={}, tls={}", 
-        config.modes.classic, config.modes.secure, config.modes.tls);
-    
-    // Initialize components
-    let stats = Arc::new(Stats::new());
-    let replay_checker = Arc::new(ReplayChecker::new(config.replay_check_len));
-    let pool = Arc::new(ConnectionPool::new());
-    
-    // Create handler
-    let handler = Arc::new(ClientHandler::new(
-        Arc::clone(&config),
-        Arc::clone(&stats),
-        Arc::clone(&replay_checker),
-        Arc::clone(&pool),
-    ));
-    
-    // Start listener
-    let addr: SocketAddr = format!("{}:{}", config.listen_addr_ipv4, config.port)
-        .parse()?;
-    
-    let listener = TcpListener::bind(addr).await?;
-    info!("Listening on {}", addr);
-    
-    // Print proxy links
-    print_proxy_links(&config);
-    
-    info!("Use RUST_LOG=debug or RUST_LOG=trace for more detailed logging");
-    
-    // Main accept loop
-    let accept_loop = async {
-        loop {
-            match listener.accept().await {
-                Ok((stream, peer)) => {
-                    let handler = Arc::clone(&handler);
-                    tokio::spawn(async move {
-                        handler.handle(stream, peer).await;
-                    });
+        std::process::exit(0);
     }
+    
+    let mut i = 0;
+    while i < args.len() {
+        match args[i].as_str() {
+            "--silent" | "-s" => { silent = true; }
+            "--log-level" => {
+                i += 1;
+                if i < args.len() { log_level = Some(args[i].clone()); }
+            }
+            s if s.starts_with("--log-level=") => {
+                log_level = Some(s.trim_start_matches("--log-level=").to_string());
+            }
+            "--help" | "-h" => {
+                eprintln!("Usage: telemt [config.toml] [OPTIONS]");
+                eprintln!();
+                eprintln!("Options:");
+                eprintln!("  --silent, -s            Suppress info logs");
+                eprintln!("  --log-level <LEVEL>     debug|verbose|normal|silent");
+                eprintln!("  --help, -h              Show this help");
+                eprintln!();
+                eprintln!("Setup (fire-and-forget):");
+                eprintln!("  --init                  Generate config, install systemd service, start");
+                eprintln!("    --port <PORT>          Listen port (default: 443)");
+                eprintln!("    --domain <DOMAIN>      TLS domain for masking (default: www.google.com)");
+                eprintln!("    --secret <HEX>         32-char hex secret (auto-generated if omitted)");
+                eprintln!("    --user <NAME>          Username (default: user)");
+                eprintln!("    --config-dir <DIR>     Config directory (default: /etc/telemt)");
+                eprintln!("    --no-start             Don't start the service after install");
+                std::process::exit(0);
+            }
+            s if !s.starts_with('-') => { config_path = s.to_string(); }
+            other => { eprintln!("Unknown option: {}", other); }
+        }
+        i += 1;
+    }
+    
+    (config_path, silent, log_level)
+}
+
+#[tokio::main]
+async fn main() -> Result<(), Box<dyn std::error::Error>> {
+    let (config_path, cli_silent, cli_log_level) = parse_cli();
+
+    let config = match ProxyConfig::load(&config_path) {
+        Ok(c) => c,
         Err(e) => {
-                    error!("Accept error: {}", e);
-                }
+            if std::path::Path::new(&config_path).exists() {
+                eprintln!("[telemt] Error: {}", e);
+                std::process::exit(1);
+            } else {
+                let default = ProxyConfig::default();
+                std::fs::write(&config_path, toml::to_string_pretty(&default).unwrap()).unwrap();
+                eprintln!("[telemt] Created default config at {}", config_path);
+                default
             }
         }
     };
     
-    // Graceful shutdown
-    tokio::select! {
-        _ = accept_loop => {}
-        _ = signal::ctrl_c() => {
-            info!("Shutting down...");
+    if let Err(e) = config.validate() {
+        eprintln!("[telemt] Invalid config: {}", e);
+        std::process::exit(1);
+    }
+
+    let has_rust_log = std::env::var("RUST_LOG").is_ok();
+    let effective_log_level = if cli_silent {
+        LogLevel::Silent
+    } else if let Some(ref s) = cli_log_level {
+        LogLevel::from_str_loose(s)
+    } else {
+        config.general.log_level.clone()
+    };
+
+    // Start with INFO so startup messages are always visible,
+    // then switch to user-configured level after startup
+    let (filter_layer, filter_handle) = reload::Layer::new(EnvFilter::new("info"));
+    tracing_subscriber::registry()
+        .with(filter_layer)
+        .with(fmt::Layer::default())
+        .init();
+
+    info!("Telemt MTProxy v{}", env!("CARGO_PKG_VERSION"));
+    info!("Log level: {}", effective_log_level);
+    info!("Modes: classic={} secure={} tls={}",
+        config.general.modes.classic,
+        config.general.modes.secure,
+        config.general.modes.tls);
+    info!("TLS domain: {}", config.censorship.tls_domain);
+    info!("Mask: {} -> {}:{}",
+        config.censorship.mask,
+        config.censorship.mask_host.as_deref().unwrap_or(&config.censorship.tls_domain),
+        config.censorship.mask_port);
+
+    if config.censorship.tls_domain == "www.google.com" {
+        warn!("Using default tls_domain. Consider setting a custom domain.");
+    }
+    
+    let prefer_ipv6 = config.general.prefer_ipv6;
+    let config = Arc::new(config);
+    let stats = Arc::new(Stats::new());
+    let rng = Arc::new(SecureRandom::new());
+    
+    let replay_checker = Arc::new(ReplayChecker::new(
+        config.access.replay_check_len,
+        Duration::from_secs(config.access.replay_window_secs),
+    ));
+    
+    let upstream_manager = Arc::new(UpstreamManager::new(config.upstreams.clone()));
+    let buffer_pool = Arc::new(BufferPool::with_config(16 * 1024, 4096));
+    
+        // Connection concurrency limit — prevents OOM under SYN flood / connection storm.
+        // 10000 is generous; each connection uses ~64KB (2x 16KB relay buffers + overhead).
+        // 10000 connections ≈ 640MB peak memory.
+        let max_connections = Arc::new(Semaphore::new(10_000));
+    
+    // Startup DC ping
+    info!("=== Telegram DC Connectivity ===");
+    let ping_results = upstream_manager.ping_all_dcs(prefer_ipv6).await;
+    for upstream_result in &ping_results {
+        info!("  via {}", upstream_result.upstream_name);
+        for dc in &upstream_result.results {
+            match (&dc.rtt_ms, &dc.error) {
+                (Some(rtt), _) => {
+                    info!("    DC{} ({:>21}):  {:.0}ms", dc.dc_idx, dc.dc_addr, rtt);
+                }
+                (None, Some(err)) => {
+                    info!("    DC{} ({:>21}):  FAIL ({})", dc.dc_idx, dc.dc_addr, err);
+                }
+                _ => {
+                    info!("    DC{} ({:>21}):  FAIL", dc.dc_idx, dc.dc_addr);
+                }
+            }
+        }
+    }
+    info!("================================");
+    
+    // Background tasks
+    let um_clone = upstream_manager.clone();
+    tokio::spawn(async move { um_clone.run_health_checks(prefer_ipv6).await; });
+    
+    let rc_clone = replay_checker.clone();
+    tokio::spawn(async move { rc_clone.run_periodic_cleanup().await; });
+
+    let detected_ip = detect_ip().await;
+    debug!("Detected IPs: v4={:?} v6={:?}", detected_ip.ipv4, detected_ip.ipv6);
+
+    let mut listeners = Vec::new();
+    
+    for listener_conf in &config.server.listeners {
+        let addr = SocketAddr::new(listener_conf.ip, config.server.port);
+        let options = ListenOptions {
+            ipv6_only: listener_conf.ip.is_ipv6(),
+            ..Default::default()
+        };
+        
+        match create_listener(addr, &options) {
+            Ok(socket) => {
+                let listener = TcpListener::from_std(socket.into())?;
+                info!("Listening on {}", addr);
+                
+                let public_ip = if let Some(ip) = listener_conf.announce_ip {
+                    ip
+                } else if listener_conf.ip.is_unspecified() {
+                    if listener_conf.ip.is_ipv4() {
+                        detected_ip.ipv4.unwrap_or(listener_conf.ip)
+                    } else {
+                        detected_ip.ipv6.unwrap_or(listener_conf.ip)
+                    }
+                } else {
+                    listener_conf.ip
+                };
+
+                if !config.show_link.is_empty() {
+                    info!("--- Proxy Links ({}) ---", public_ip);
+                    for user_name in &config.show_link {
+                        if let Some(secret) = config.access.users.get(user_name) {
+                            info!("User: {}", user_name);
+                            if config.general.modes.classic {
+                                info!("  Classic: tg://proxy?server={}&port={}&secret={}",
+                                    public_ip, config.server.port, secret);
+                            }
+                            if config.general.modes.secure {
+                                info!("  DD:      tg://proxy?server={}&port={}&secret=dd{}",
+                                    public_ip, config.server.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{}{}",
+                                    public_ip, config.server.port, secret, domain_hex);
+                            }
+                        } else {
+                            warn!("User '{}' in show_link not found", user_name);
+                        }
+                    }
+                    info!("------------------------");
+                }
+                
+                listeners.push(listener);
+            },
+            Err(e) => {
+                error!("Failed to bind to {}: {}", addr, e);
+            }
         }
     }
     
-    // Cleanup
-    pool.close_all().await;
+    if listeners.is_empty() {
+        error!("No listeners. Exiting.");
+        std::process::exit(1);
+    }
+
+    // Switch to user-configured log level after startup
+    let runtime_filter = if has_rust_log {
+        EnvFilter::from_default_env()
+    } else {
+        EnvFilter::new(effective_log_level.to_filter_str())
+    };
+    filter_handle.reload(runtime_filter).expect("Failed to switch log filter");
+
+    for listener in listeners {
+        let config = config.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();
+        
+        tokio::spawn(async move {
+            loop {
+                match listener.accept().await {
+                    Ok((stream, peer_addr)) => {
+                        let config = config.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();
+                        
+                        tokio::spawn(async move {
+                            if let Err(e) = ClientHandler::new(
+                                stream, peer_addr, config, stats,
+                                upstream_manager, replay_checker, buffer_pool, rng
+                            ).run().await {
+                                debug!(peer = %peer_addr, error = %e, "Connection error");
+                            }
+                        });
+                    }
+                    Err(e) => {
+                        error!("Accept error: {}", e);
+                        tokio::time::sleep(Duration::from_millis(100)).await;
+                    }
+                }
+            }
+        });
+    }
+
+    match signal::ctrl_c().await {
+        Ok(()) => info!("Shutting down..."),
+        Err(e) => error!("Signal error: {}", e),
+    }
 
-    info!("Goodbye!");
     Ok(())
 }
-
-fn print_proxy_links(config: &ProxyConfig) {
-    println!("\n=== Proxy Links ===\n");
-    
-    for (user, secret) in &config.users {
-        if config.modes.tls {
-            let tls_secret = format!(
-                "ee{}{}",
-                secret,
-                hex::encode(config.tls_domain.as_bytes())
-            );
-            println!(
-                "{} (TLS): tg://proxy?server=IP&port={}&secret={}",
-                user, config.port, tls_secret
-            );
-        }
-        
-        if config.modes.secure {
-            println!(
-                "{} (Secure): tg://proxy?server=IP&port={}&secret=dd{}",
-                user, config.port, secret
-            );
-        }
-        
-        if config.modes.classic {
-            println!(
-                "{} (Classic): tg://proxy?server=IP&port={}&secret={}",
-                user, config.port, secret
-            );
-        }
-        
-        println!();
-    }
-    
-    println!("===================\n");
-}

src/protocol/constants.rs

@@ -1,13 +1,13 @@
 //! Protocol constants and datacenter addresses
 
 use std::net::{IpAddr, Ipv4Addr, Ipv6Addr};
-use once_cell::sync::Lazy;
+use std::sync::LazyLock;
 
 // ============= Telegram Datacenters =============
 
 pub const TG_DATACENTER_PORT: u16 = 443;
 
-pub static TG_DATACENTERS_V4: Lazy<Vec<IpAddr>> = Lazy::new(|| {
+pub static TG_DATACENTERS_V4: LazyLock<Vec<IpAddr>> = LazyLock::new(|| {
     vec![
         IpAddr::V4(Ipv4Addr::new(149, 154, 175, 50)),
         IpAddr::V4(Ipv4Addr::new(149, 154, 167, 51)),
@@ -17,7 +17,7 @@ pub static TG_DATACENTERS_V4: Lazy<Vec<IpAddr>> = Lazy::new(|| {
     ]
 });
 
-pub static TG_DATACENTERS_V6: Lazy<Vec<IpAddr>> = Lazy::new(|| {
+pub static TG_DATACENTERS_V6: LazyLock<Vec<IpAddr>> = LazyLock::new(|| {
     vec![
         IpAddr::V6("2001:b28:f23d:f001::a".parse().unwrap()),
         IpAddr::V6("2001:67c:04e8:f002::a".parse().unwrap()),
@@ -29,8 +29,8 @@ pub static TG_DATACENTERS_V6: Lazy<Vec<IpAddr>> = Lazy::new(|| {
 
 // ============= Middle Proxies (for advertising) =============
 
-pub static TG_MIDDLE_PROXIES_V4: Lazy<std::collections::HashMap<i32, Vec<(IpAddr, u16)>>> = 
-    Lazy::new(|| {
+pub static TG_MIDDLE_PROXIES_V4: LazyLock<std::collections::HashMap<i32, Vec<(IpAddr, u16)>>> = 
+    LazyLock::new(|| {
         let mut m = std::collections::HashMap::new();
         m.insert(1, vec![(IpAddr::V4(Ipv4Addr::new(149, 154, 175, 50)), 8888)]);
         m.insert(-1, vec![(IpAddr::V4(Ipv4Addr::new(149, 154, 175, 50)), 8888)]);
@@ -45,8 +45,8 @@ pub static TG_MIDDLE_PROXIES_V4: Lazy<std::collections::HashMap<i32, Vec<(IpAddr
         m
     });
 
-pub static TG_MIDDLE_PROXIES_V6: Lazy<std::collections::HashMap<i32, Vec<(IpAddr, u16)>>> = 
-    Lazy::new(|| {
+pub static TG_MIDDLE_PROXIES_V6: LazyLock<std::collections::HashMap<i32, Vec<(IpAddr, u16)>>> = 
+    LazyLock::new(|| {
         let mut m = std::collections::HashMap::new();
         m.insert(1, vec![(IpAddr::V6("2001:b28:f23d:f001::d".parse().unwrap()), 8888)]);
         m.insert(-1, vec![(IpAddr::V6("2001:b28:f23d:f001::d".parse().unwrap()), 8888)]);
@@ -167,7 +167,8 @@ pub const DEFAULT_ACK_TIMEOUT_SECS: u64 = 300;
 // ============= Buffer Sizes =============
 
 /// Default buffer size
-pub const DEFAULT_BUFFER_SIZE: usize = 65536;
+pub const DEFAULT_BUFFER_SIZE: usize = 16384;
+
 /// Small buffer size for bad client handling
 pub const SMALL_BUFFER_SIZE: usize = 8192;
 

src/protocol/obfuscation.rs

@@ -1,10 +1,13 @@
 //! MTProto Obfuscation
 
+use zeroize::Zeroize;
 use crate::crypto::{sha256, AesCtr};
 use crate::error::Result;
 use super::constants::*;
 
 /// Obfuscation parameters from handshake
+///
+/// Key material is zeroized on drop.
 #[derive(Debug, Clone)]
 pub struct ObfuscationParams {
     /// Key for decrypting client -> proxy traffic
@@ -21,25 +24,31 @@ pub struct ObfuscationParams {
     pub dc_idx: i16,
 }
 
+impl Drop for ObfuscationParams {
+    fn drop(&mut self) {
+        self.decrypt_key.zeroize();
+        self.decrypt_iv.zeroize();
+        self.encrypt_key.zeroize();
+        self.encrypt_iv.zeroize();
+    }
+}
+
 impl ObfuscationParams {
     /// Parse obfuscation parameters from handshake bytes
     /// Returns None if handshake doesn't match any user secret
     pub fn from_handshake(
         handshake: &[u8; HANDSHAKE_LEN],
-        secrets: &[(String, Vec<u8>)], // (username, secret_bytes)
+        secrets: &[(String, Vec<u8>)],
     ) -> Option<(Self, String)> {
-        // Extract prekey and IV for decryption
         let dec_prekey_iv = &handshake[SKIP_LEN..SKIP_LEN + PREKEY_LEN + IV_LEN];
         let dec_prekey = &dec_prekey_iv[..PREKEY_LEN];
         let dec_iv_bytes = &dec_prekey_iv[PREKEY_LEN..];
         
-        // Reversed for encryption direction
         let enc_prekey_iv: Vec<u8> = dec_prekey_iv.iter().rev().copied().collect();
         let enc_prekey = &enc_prekey_iv[..PREKEY_LEN];
         let enc_iv_bytes = &enc_prekey_iv[PREKEY_LEN..];
         
         for (username, secret) in secrets {
-            // Derive decryption key
             let mut dec_key_input = Vec::with_capacity(PREKEY_LEN + secret.len());
             dec_key_input.extend_from_slice(dec_prekey);
             dec_key_input.extend_from_slice(secret);
@@ -47,26 +56,22 @@ impl ObfuscationParams {
             
             let decrypt_iv = u128::from_be_bytes(dec_iv_bytes.try_into().unwrap());
             
-            // Create decryptor and decrypt handshake
             let mut decryptor = AesCtr::new(&decrypt_key, decrypt_iv);
             let decrypted = decryptor.decrypt(handshake);
             
-            // Check protocol tag
             let tag_bytes: [u8; 4] = decrypted[PROTO_TAG_POS..PROTO_TAG_POS + 4]
                 .try_into()
                 .unwrap();
             
             let proto_tag = match ProtoTag::from_bytes(tag_bytes) {
                 Some(tag) => tag,
-                None => continue, // Try next secret
+                None => continue,
             };
             
-            // Extract DC index
             let dc_idx = i16::from_le_bytes(
                 decrypted[DC_IDX_POS..DC_IDX_POS + 2].try_into().unwrap()
             );
             
-            // Derive encryption key
             let mut enc_key_input = Vec::with_capacity(PREKEY_LEN + secret.len());
             enc_key_input.extend_from_slice(enc_prekey);
             enc_key_input.extend_from_slice(secret);
@@ -123,18 +128,15 @@ pub fn generate_nonce<R: FnMut(usize) -> Vec<u8>>(mut random_bytes: R) -> [u8; H
 
 /// Check if nonce is valid (not matching reserved patterns)
 pub fn is_valid_nonce(nonce: &[u8; HANDSHAKE_LEN]) -> bool {
-    // Check first byte
     if RESERVED_NONCE_FIRST_BYTES.contains(&nonce[0]) {
         return false;
     }
     
-    // Check first 4 bytes
     let first_four: [u8; 4] = nonce[..4].try_into().unwrap();
     if RESERVED_NONCE_BEGINNINGS.contains(&first_four) {
         return false;
     }
     
-    // Check bytes 4-7
     let continue_four: [u8; 4] = nonce[4..8].try_into().unwrap();
     if RESERVED_NONCE_CONTINUES.contains(&continue_four) {
         return false;
@@ -147,12 +149,10 @@ pub fn is_valid_nonce(nonce: &[u8; HANDSHAKE_LEN]) -> bool {
 pub fn prepare_tg_nonce(
     nonce: &mut [u8; HANDSHAKE_LEN],
     proto_tag: ProtoTag,
-    enc_key_iv: Option<&[u8]>, // For fast mode
+    enc_key_iv: Option<&[u8]>,
 ) {
-    // Set protocol tag
     nonce[PROTO_TAG_POS..PROTO_TAG_POS + 4].copy_from_slice(&proto_tag.to_bytes());
     
-    // For fast mode, copy the reversed enc_key_iv
     if let Some(key_iv) = enc_key_iv {
         let reversed: Vec<u8> = key_iv.iter().rev().copied().collect();
         nonce[SKIP_LEN..SKIP_LEN + KEY_LEN + IV_LEN].copy_from_slice(&reversed);
@@ -161,14 +161,12 @@ pub fn prepare_tg_nonce(
 
 /// Encrypt the outgoing nonce for Telegram
 pub fn encrypt_nonce(nonce: &[u8; HANDSHAKE_LEN]) -> Vec<u8> {
-    // Derive encryption key from the nonce itself
     let key_iv = &nonce[SKIP_LEN..SKIP_LEN + KEY_LEN + IV_LEN];
     let enc_key = sha256(key_iv);
     let enc_iv = u128::from_be_bytes(key_iv[..IV_LEN].try_into().unwrap());
     
     let mut encryptor = AesCtr::new(&enc_key, enc_iv);
     
-    // Only encrypt from PROTO_TAG_POS onwards
     let mut result = nonce.to_vec();
     let encrypted_part = encryptor.encrypt(&nonce[PROTO_TAG_POS..]);
     result[PROTO_TAG_POS..].copy_from_slice(&encrypted_part);
@@ -182,22 +180,18 @@ mod tests {
     
     #[test]
     fn test_is_valid_nonce() {
-        // Valid nonce
         let mut valid = [0x42u8; HANDSHAKE_LEN];
         valid[4..8].copy_from_slice(&[1, 2, 3, 4]);
         assert!(is_valid_nonce(&valid));
         
-        // Invalid: starts with 0xef
         let mut invalid = [0x00u8; HANDSHAKE_LEN];
         invalid[0] = 0xef;
         assert!(!is_valid_nonce(&invalid));
         
-        // Invalid: starts with HEAD
         let mut invalid = [0x00u8; HANDSHAKE_LEN];
         invalid[..4].copy_from_slice(b"HEAD");
         assert!(!is_valid_nonce(&invalid));
         
-        // Invalid: bytes 4-7 are zeros
         let mut invalid = [0x42u8; HANDSHAKE_LEN];
         invalid[4..8].copy_from_slice(&[0, 0, 0, 0]);
         assert!(!is_valid_nonce(&invalid));

src/protocol/tls.rs

@@ -4,7 +4,7 @@
 //! for domain fronting. The handshake looks like valid TLS 1.3 but
 //! actually carries MTProto authentication data.
 
-use crate::crypto::{sha256_hmac, random::SECURE_RANDOM};
+use crate::crypto::{sha256_hmac, SecureRandom};
 use crate::error::{ProxyError, Result};
 use super::constants::*;
 use std::time::{SystemTime, UNIX_EPOCH};
@@ -315,8 +315,8 @@ pub fn validate_tls_handshake(
 ///
 /// This generates random bytes that look like a valid X25519 public key.
 /// Since we're not doing real TLS, the actual cryptographic properties don't matter.
-pub fn gen_fake_x25519_key() -> [u8; 32] {
-    let bytes = SECURE_RANDOM.bytes(32);
+pub fn gen_fake_x25519_key(rng: &SecureRandom) -> [u8; 32] {
+    let bytes = rng.bytes(32);
     bytes.try_into().unwrap()
 }
 
@@ -333,8 +333,9 @@ pub fn build_server_hello(
     client_digest: &[u8; TLS_DIGEST_LEN],
     session_id: &[u8],
     fake_cert_len: usize,
+    rng: &SecureRandom,
 ) -> Vec<u8> {
-    let x25519_key = gen_fake_x25519_key();
+    let x25519_key = gen_fake_x25519_key(rng);
     
     // Build ServerHello
     let server_hello = ServerHelloBuilder::new(session_id.to_vec())
@@ -351,7 +352,7 @@ pub fn build_server_hello(
     ];
     
     // Build fake certificate (Application Data record)
-    let fake_cert = SECURE_RANDOM.bytes(fake_cert_len);
+    let fake_cert = rng.bytes(fake_cert_len);
     let mut app_data_record = Vec::with_capacity(5 + fake_cert_len);
     app_data_record.push(TLS_RECORD_APPLICATION);
     app_data_record.extend_from_slice(&TLS_VERSION);
@@ -489,8 +490,9 @@ mod tests {
     
     #[test]
     fn test_gen_fake_x25519_key() {
-        let key1 = gen_fake_x25519_key();
-        let key2 = gen_fake_x25519_key();
+        let rng = SecureRandom::new();
+        let key1 = gen_fake_x25519_key(&rng);
+        let key2 = gen_fake_x25519_key(&rng);
         
         assert_eq!(key1.len(), 32);
         assert_eq!(key2.len(), 32);
@@ -545,7 +547,8 @@ mod tests {
         let client_digest = [0x42u8; 32];
         let session_id = vec![0xAA; 32];
         
-        let response = build_server_hello(secret, &client_digest, &session_id, 2048);
+        let rng = SecureRandom::new();
+        let response = build_server_hello(secret, &client_digest, &session_id, 2048, &rng);
         
         // Should have at least 3 records
         assert!(response.len() > 100);
@@ -577,8 +580,9 @@ mod tests {
         let client_digest = [0x42u8; 32];
         let session_id = vec![0xAA; 32];
         
-        let response1 = build_server_hello(secret, &client_digest, &session_id, 1024);
-        let response2 = build_server_hello(secret, &client_digest, &session_id, 1024);
+        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);
         
         // Digest position should have non-zero data
         let digest1 = &response1[TLS_DIGEST_POS..TLS_DIGEST_POS + TLS_DIGEST_LEN];

src/proxy/client.rs

@@ -13,103 +13,104 @@ use crate::error::{ProxyError, Result, HandshakeResult};
 use crate::protocol::constants::*;
 use crate::protocol::tls;
 use crate::stats::{Stats, ReplayChecker};
-use crate::transport::{ConnectionPool, configure_client_socket};
-use crate::stream::{CryptoReader, CryptoWriter, FakeTlsReader, FakeTlsWriter};
-use crate::crypto::AesCtr;
+use crate::transport::{configure_client_socket, UpstreamManager};
+use crate::stream::{CryptoReader, CryptoWriter, FakeTlsReader, FakeTlsWriter, BufferPool};
+use crate::crypto::{AesCtr, SecureRandom};
 
-use super::handshake::{
+use crate::proxy::handshake::{
     handle_tls_handshake, handle_mtproto_handshake, 
     HandshakeSuccess, generate_tg_nonce, encrypt_tg_nonce,
 };
-use super::relay::relay_bidirectional;
-use super::masking::handle_bad_client;
+use crate::proxy::relay::relay_bidirectional;
+use crate::proxy::masking::handle_bad_client;
 
-/// Client connection handler
-pub struct ClientHandler {
+pub struct ClientHandler;
+
+pub struct RunningClientHandler {
+    stream: TcpStream,
+    peer: SocketAddr,
     config: Arc<ProxyConfig>,
     stats: Arc<Stats>,
     replay_checker: Arc<ReplayChecker>,
-    pool: Arc<ConnectionPool>,
+    upstream_manager: Arc<UpstreamManager>,
+    buffer_pool: Arc<BufferPool>,
+    rng: Arc<SecureRandom>,
 }
 
 impl ClientHandler {
-    /// Create new client handler
     pub fn new(
+        stream: TcpStream,
+        peer: SocketAddr,
         config: Arc<ProxyConfig>,
         stats: Arc<Stats>,
+        upstream_manager: Arc<UpstreamManager>,
         replay_checker: Arc<ReplayChecker>,
-        pool: Arc<ConnectionPool>,
-    ) -> Self {
-        Self {
-            config,
-            stats,
-            replay_checker,
-            pool,
+        buffer_pool: Arc<BufferPool>,
+        rng: Arc<SecureRandom>,
+    ) -> RunningClientHandler {
+        RunningClientHandler {
+            stream, peer, config, stats, replay_checker,
+            upstream_manager, buffer_pool, rng,
+        }
     }
 }
 
-    /// Handle a client connection
-    pub async fn handle(&self, stream: TcpStream, peer: SocketAddr) {
+impl RunningClientHandler {
+    pub async fn run(mut self) -> Result<()> {
         self.stats.increment_connects_all();
         
+        let peer = self.peer;
         debug!(peer = %peer, "New connection");
         
-        // Configure socket
         if let Err(e) = configure_client_socket(
-            &stream,
-            self.config.client_keepalive,
-            self.config.client_ack_timeout,
+            &self.stream,
+            self.config.timeouts.client_keepalive,
+            self.config.timeouts.client_ack,
         ) {
             debug!(peer = %peer, error = %e, "Failed to configure client socket");
         }
         
-        // Perform handshake with timeout
-        let handshake_timeout = Duration::from_secs(self.config.client_handshake_timeout);
+        let handshake_timeout = Duration::from_secs(self.config.timeouts.client_handshake);
+        let stats = self.stats.clone();
         
-        let result = timeout(
-            handshake_timeout,
-            self.do_handshake(stream, peer)
-        ).await;
+        let result = timeout(handshake_timeout, self.do_handshake()).await;
         
         match result {
             Ok(Ok(())) => {
                 debug!(peer = %peer, "Connection handled successfully");
+                Ok(())
             }
             Ok(Err(e)) => {
                 debug!(peer = %peer, error = %e, "Handshake failed");
+                Err(e)
             }
             Err(_) => {
-                self.stats.increment_handshake_timeouts();
+                stats.increment_handshake_timeouts();
                 debug!(peer = %peer, "Handshake timeout");
+                Err(ProxyError::TgHandshakeTimeout)
             }
         }
     }
     
-    /// Perform handshake and relay
-    async fn do_handshake(&self, mut stream: TcpStream, peer: SocketAddr) -> Result<()> {
-        // Read first bytes to determine handshake type
+    async fn do_handshake(mut self) -> Result<()> {
         let mut first_bytes = [0u8; 5];
-        stream.read_exact(&mut first_bytes).await?;
+        self.stream.read_exact(&mut first_bytes).await?;
         
         let is_tls = tls::is_tls_handshake(&first_bytes[..3]);
+        let peer = self.peer;
         
-        debug!(peer = %peer, is_tls = is_tls, first_bytes = %hex::encode(&first_bytes), "Handshake type detected");
+        debug!(peer = %peer, is_tls = is_tls, "Handshake type detected");
         
         if is_tls {
-            self.handle_tls_client(stream, peer, first_bytes).await
+            self.handle_tls_client(first_bytes).await
         } else {
-            self.handle_direct_client(stream, peer, first_bytes).await
+            self.handle_direct_client(first_bytes).await
         }
     }
     
-    /// Handle TLS-wrapped client
-    async fn handle_tls_client(
-        &self,
-        mut stream: TcpStream,
-        peer: SocketAddr,
-        first_bytes: [u8; 5],
-    ) -> Result<()> {
-        // Read TLS handshake length
+    async fn handle_tls_client(mut self, first_bytes: [u8; 5]) -> Result<()> {
+        let peer = self.peer;
+        
         let tls_len = u16::from_be_bytes([first_bytes[3], first_bytes[4]]) as usize;
         
         debug!(peer = %peer, tls_len = tls_len, "Reading TLS handshake");
@@ -117,113 +118,111 @@ impl ClientHandler {
         if tls_len < 512 {
             debug!(peer = %peer, tls_len = tls_len, "TLS handshake too short");
             self.stats.increment_connects_bad();
-            handle_bad_client(stream, &first_bytes, &self.config).await;
+            let (reader, writer) = self.stream.into_split();
+            handle_bad_client(reader, writer, &first_bytes, &self.config).await;
             return Ok(());
         }
         
-        // Read full TLS handshake
         let mut handshake = vec![0u8; 5 + tls_len];
         handshake[..5].copy_from_slice(&first_bytes);
-        stream.read_exact(&mut handshake[5..]).await?;
+        self.stream.read_exact(&mut handshake[5..]).await?;
         
-        // Split stream for reading/writing
-        let (read_half, write_half) = stream.into_split();
+        let config = self.config.clone();
+        let replay_checker = self.replay_checker.clone();
+        let stats = self.stats.clone();
+        let buffer_pool = self.buffer_pool.clone();
+        
+        let (read_half, write_half) = self.stream.into_split();
         
-        // Handle TLS handshake
         let (mut tls_reader, tls_writer, _tls_user) = match handle_tls_handshake(
-            &handshake,
-            read_half,
-            write_half,
-            peer,
-            &self.config,
-            &self.replay_checker,
+            &handshake, read_half, write_half, peer,
+            &config, &replay_checker, &self.rng,
         ).await {
             HandshakeResult::Success(result) => result,
-            HandshakeResult::BadClient => {
-                self.stats.increment_connects_bad();
+            HandshakeResult::BadClient { reader, writer } => {
+                stats.increment_connects_bad();
+                handle_bad_client(reader, writer, &handshake, &config).await;
                 return Ok(());
             }
             HandshakeResult::Error(e) => return Err(e),
         };
         
-        // Read MTProto handshake through TLS
         debug!(peer = %peer, "Reading MTProto handshake through TLS");
         let mtproto_data = tls_reader.read_exact(HANDSHAKE_LEN).await?;
         let mtproto_handshake: [u8; HANDSHAKE_LEN] = mtproto_data[..].try_into()
             .map_err(|_| ProxyError::InvalidHandshake("Short MTProto handshake".into()))?;
         
-        // Handle MTProto handshake
         let (crypto_reader, crypto_writer, success) = match handle_mtproto_handshake(
-            &mtproto_handshake,
-            tls_reader,
-            tls_writer,
-            peer,
-            &self.config,
-            &self.replay_checker,
-            true,
+            &mtproto_handshake, tls_reader, tls_writer, peer,
+            &config, &replay_checker, true,
         ).await {
             HandshakeResult::Success(result) => result,
-            HandshakeResult::BadClient => {
-                self.stats.increment_connects_bad();
+            HandshakeResult::BadClient { reader: _, writer: _ } => {
+                stats.increment_connects_bad();
+                debug!(peer = %peer, "Valid TLS but invalid MTProto handshake");
                 return Ok(());
             }
             HandshakeResult::Error(e) => return Err(e),
         };
         
-        // Handle authenticated client
-        self.handle_authenticated_inner(crypto_reader, crypto_writer, success).await
+        Self::handle_authenticated_static(
+            crypto_reader, crypto_writer, success,
+            self.upstream_manager, self.stats, self.config,
+            buffer_pool, self.rng,
+        ).await
     }
     
-    /// Handle direct (non-TLS) client
-    async fn handle_direct_client(
-        &self,
-        mut stream: TcpStream,
-        peer: SocketAddr,
-        first_bytes: [u8; 5],
-    ) -> Result<()> {
-        // Check if non-TLS modes are enabled
-        if !self.config.modes.classic && !self.config.modes.secure {
+    async fn handle_direct_client(mut self, first_bytes: [u8; 5]) -> Result<()> {
+        let peer = self.peer;
+        
+        if !self.config.general.modes.classic && !self.config.general.modes.secure {
             debug!(peer = %peer, "Non-TLS modes disabled");
             self.stats.increment_connects_bad();
-            handle_bad_client(stream, &first_bytes, &self.config).await;
+            let (reader, writer) = self.stream.into_split();
+            handle_bad_client(reader, writer, &first_bytes, &self.config).await;
             return Ok(());
         }
         
-        // Read rest of handshake
         let mut handshake = [0u8; HANDSHAKE_LEN];
         handshake[..5].copy_from_slice(&first_bytes);
-        stream.read_exact(&mut handshake[5..]).await?;
+        self.stream.read_exact(&mut handshake[5..]).await?;
         
-        // Split stream
-        let (read_half, write_half) = stream.into_split();
+        let config = self.config.clone();
+        let replay_checker = self.replay_checker.clone();
+        let stats = self.stats.clone();
+        let buffer_pool = self.buffer_pool.clone();
+        
+        let (read_half, write_half) = self.stream.into_split();
         
-        // Handle MTProto handshake
         let (crypto_reader, crypto_writer, success) = match handle_mtproto_handshake(
-            &handshake,
-            read_half,
-            write_half,
-            peer,
-            &self.config,
-            &self.replay_checker,
-            false,
+            &handshake, read_half, write_half, peer,
+            &config, &replay_checker, false,
         ).await {
             HandshakeResult::Success(result) => result,
-            HandshakeResult::BadClient => {
-                self.stats.increment_connects_bad();
+            HandshakeResult::BadClient { reader, writer } => {
+                stats.increment_connects_bad();
+                handle_bad_client(reader, writer, &handshake, &config).await;
                 return Ok(());
             }
             HandshakeResult::Error(e) => return Err(e),
         };
         
-        self.handle_authenticated_inner(crypto_reader, crypto_writer, success).await
+        Self::handle_authenticated_static(
+            crypto_reader, crypto_writer, success,
+            self.upstream_manager, self.stats, self.config,
+            buffer_pool, self.rng,
+        ).await
     }
     
-    /// Handle authenticated client - connect to Telegram and relay
-    async fn handle_authenticated_inner<R, W>(
-        &self,
+    async fn handle_authenticated_static<R, W>(
         client_reader: CryptoReader<R>,
         client_writer: CryptoWriter<W>,
         success: HandshakeSuccess,
+        upstream_manager: Arc<UpstreamManager>,
+        stats: Arc<Stats>,
+        config: Arc<ProxyConfig>,
+        buffer_pool: Arc<BufferPool>,
+        rng: Arc<SecureRandom>,
     ) -> Result<()>
     where
         R: AsyncRead + Unpin + Send + 'static,
@@ -231,14 +230,12 @@ impl ClientHandler {
     {
         let user = &success.user;
         
-        // Check user limits
-        if let Err(e) = self.check_user_limits(user) {
+        if let Err(e) = Self::check_user_limits_static(user, &config, &stats) {
             warn!(user = %user, error = %e, "User limit exceeded");
             return Err(e);
         }
         
-        // Get datacenter address
-        let dc_addr = self.get_dc_addr(success.dc_idx)?;
+        let dc_addr = Self::get_dc_addr_static(success.dc_idx, &config)?;
         
         info!(
             user = %user,
@@ -246,69 +243,54 @@ impl ClientHandler {
             dc = success.dc_idx,
             dc_addr = %dc_addr,
             proto = ?success.proto_tag,
-            fast_mode = self.config.fast_mode,
             "Connecting to Telegram"
         );
         
-        // Connect to Telegram
-        let tg_stream = self.pool.get(dc_addr).await?;
+        // Pass dc_idx for latency-based upstream selection
+        let tg_stream = upstream_manager.connect(dc_addr, Some(success.dc_idx)).await?;
         
-        debug!(peer = %success.peer, dc_addr = %dc_addr, "Connected to Telegram, performing handshake");
+        debug!(peer = %success.peer, dc_addr = %dc_addr, "Connected, performing TG handshake");
         
-        // Perform Telegram handshake and get crypto streams
-        let (tg_reader, tg_writer) = self.do_tg_handshake(
-            tg_stream, 
-            &success,
+        let (tg_reader, tg_writer) = Self::do_tg_handshake_static(
+            tg_stream, &success, &config, rng.as_ref(),
         ).await?;
         
-        debug!(peer = %success.peer, "Telegram handshake complete, starting relay");
+        debug!(peer = %success.peer, "TG handshake complete, starting relay");
         
-        // Update stats
-        self.stats.increment_user_connects(user);
-        self.stats.increment_user_curr_connects(user);
+        stats.increment_user_connects(user);
+        stats.increment_user_curr_connects(user);
         
-        // Relay traffic - передаём Arc::clone(&self.stats)
         let relay_result = relay_bidirectional(
-            client_reader,
-            client_writer,
-            tg_reader,
-            tg_writer,
-            user,
-            Arc::clone(&self.stats),
+            client_reader, client_writer,
+            tg_reader, tg_writer,
+            user, Arc::clone(&stats), buffer_pool,
         ).await;
         
-        // Update stats
-        self.stats.decrement_user_curr_connects(user);
+        stats.decrement_user_curr_connects(user);
         
         match &relay_result {
-            Ok(()) => debug!(user = %user, peer = %success.peer, "Relay completed normally"),
-            Err(e) => debug!(user = %user, peer = %success.peer, error = %e, "Relay ended with error"),
+            Ok(()) => debug!(user = %user, "Relay completed"),
+            Err(e) => debug!(user = %user, error = %e, "Relay ended with error"),
         }
         
         relay_result
     }
     
-    /// Check user limits (expiration, connection count, data quota)
-    fn check_user_limits(&self, user: &str) -> Result<()> {
-        // Check expiration
-        if let Some(expiration) = self.config.user_expirations.get(user) {
+    fn check_user_limits_static(user: &str, config: &ProxyConfig, stats: &Stats) -> Result<()> {
+        if let Some(expiration) = config.access.user_expirations.get(user) {
             if chrono::Utc::now() > *expiration {
                 return Err(ProxyError::UserExpired { user: user.to_string() });
             }
         }
         
-        // Check connection limit
-        if let Some(limit) = self.config.user_max_tcp_conns.get(user) {
-            let current = self.stats.get_user_curr_connects(user);
-            if current >= *limit as u64 {
+        if let Some(limit) = config.access.user_max_tcp_conns.get(user) {
+            if stats.get_user_curr_connects(user) >= *limit as u64 {
                 return Err(ProxyError::ConnectionLimitExceeded { user: user.to_string() });
             }
         }
         
-        // Check data quota
-        if let Some(quota) = self.config.user_data_quota.get(user) {
-            let used = self.stats.get_user_total_octets(user);
-            if used >= *quota {
+        if let Some(quota) = config.access.user_data_quota.get(user) {
+            if stats.get_user_total_octets(user) >= *quota {
                 return Err(ProxyError::DataQuotaExceeded { user: user.to_string() });
             }
         }
@@ -316,63 +298,105 @@ impl ClientHandler {
         Ok(())
     }
     
-    /// Get datacenter address by index
-    fn get_dc_addr(&self, dc_idx: i16) -> Result<SocketAddr> {
-        let idx = (dc_idx.abs() - 1) as usize;
-        
-        let datacenters = if self.config.prefer_ipv6 {
+    /// Resolve DC index to a target address.
+    ///
+    /// Matches the C implementation's behavior exactly:
+    ///
+    /// 1. Look up DC in known clusters (standard DCs ±1..±5)
+    /// 2. If not found and `force=1` → fall back to `default_cluster`
+    ///
+    /// In the C code:
+    /// - `proxy-multi.conf` is downloaded from Telegram, contains only DC ±1..±5
+    /// - `default 2;` directive sets the default cluster
+    /// - `mf_cluster_lookup(CurConf, target_dc, 1)` returns default_cluster
+    ///   for any unknown DC (like CDN DC 203)
+    ///
+    /// So DC 203, DC 101, DC -300, etc. all route to the default DC (2).
+    /// There is NO modular arithmetic in the C implementation.
+    fn get_dc_addr_static(dc_idx: i16, config: &ProxyConfig) -> Result<SocketAddr> {
+        let datacenters = if config.general.prefer_ipv6 {
             &*TG_DATACENTERS_V6
         } else {
             &*TG_DATACENTERS_V4
         };
     
-        datacenters.get(idx)
-            .map(|ip| SocketAddr::new(*ip, TG_DATACENTER_PORT))
-            .ok_or_else(|| ProxyError::InvalidHandshake(
-                format!("Invalid DC index: {}", dc_idx)
-            ))
+        let num_dcs = datacenters.len(); // 5
+    
+        // === Step 1: Check dc_overrides (like C's `proxy_for <dc> <ip>:<port>`) ===
+        let dc_key = dc_idx.to_string();
+        if let Some(addr_str) = config.dc_overrides.get(&dc_key) {
+            match addr_str.parse::<SocketAddr>() {
+                Ok(addr) => {
+                    debug!(dc_idx = dc_idx, addr = %addr, "Using DC override from config");
+                    return Ok(addr);
+                }
+                Err(_) => {
+                    warn!(dc_idx = dc_idx, addr_str = %addr_str,
+                        "Invalid DC override address in config, ignoring");
+                }
+            }
         }
     
-    /// Perform handshake with Telegram server
-    /// Returns crypto reader and writer for TG connection
-    async fn do_tg_handshake(
-        &self,
-        mut stream: TcpStream,
-        success: &HandshakeSuccess,
-    ) -> Result<(CryptoReader<tokio::net::tcp::OwnedReadHalf>, CryptoWriter<tokio::net::tcp::OwnedWriteHalf>)> {
-        // Generate nonce with keys for TG
-        let (nonce, tg_enc_key, tg_enc_iv, tg_dec_key, tg_dec_iv) = generate_tg_nonce(
-            success.proto_tag,
-            &success.dec_key,  // Client's dec key
-            success.dec_iv,
-            self.config.fast_mode,
+        // === Step 2: Standard DCs ±1..±5 — direct lookup ===
+        let abs_dc = dc_idx.unsigned_abs() as usize;
+        if abs_dc >= 1 && abs_dc <= num_dcs {
+            return Ok(SocketAddr::new(datacenters[abs_dc - 1], TG_DATACENTER_PORT));
+        }
+    
+        // === Step 3: Unknown DC — fall back to default_cluster ===
+        // Exactly like C's `mf_cluster_lookup(CurConf, target_dc, force=1)`
+        // which returns `MC->default_cluster` when the DC is not found.
+        // Telegram's proxy-multi.conf uses `default 2;`
+        let default_dc = config.default_dc.unwrap_or(2) as usize;
+        let fallback_idx = if default_dc >= 1 && default_dc <= num_dcs {
+            default_dc - 1
+        } else {
+            1 // DC 2 (index 1) — matches Telegram's `default 2;`
+        };
+    
+        info!(
+            original_dc = dc_idx,
+            fallback_dc = (fallback_idx + 1) as u16,
+            fallback_addr = %datacenters[fallback_idx],
+            "Special DC ---> default_cluster"
+        );
+    
+        Ok(SocketAddr::new(datacenters[fallback_idx], TG_DATACENTER_PORT))
+    }
+    
+    async fn do_tg_handshake_static(
+        mut stream: TcpStream,
+        success: &HandshakeSuccess,
+        config: &ProxyConfig,
+        rng: &SecureRandom,
+    ) -> Result<(CryptoReader<tokio::net::tcp::OwnedReadHalf>, CryptoWriter<tokio::net::tcp::OwnedWriteHalf>)> {
+        let (nonce, tg_enc_key, tg_enc_iv, tg_dec_key, tg_dec_iv) = generate_tg_nonce(
+            success.proto_tag,
+            &success.dec_key,
+            success.dec_iv,
+            rng,
+            config.general.fast_mode,
         );
         
-        // Encrypt nonce
         let encrypted_nonce = encrypt_tg_nonce(&nonce);
         
         debug!(
             peer = %success.peer,
             nonce_head = %hex::encode(&nonce[..16]),
-            encrypted_head = %hex::encode(&encrypted_nonce[..16]),
             "Sending nonce to Telegram"
         );
         
-        // Send to Telegram
         stream.write_all(&encrypted_nonce).await?;
         stream.flush().await?;
         
-        debug!(peer = %success.peer, "Nonce sent to Telegram");
-        
-        // Split stream and wrap with crypto
         let (read_half, write_half) = stream.into_split();
         
         let decryptor = AesCtr::new(&tg_dec_key, tg_dec_iv);
         let encryptor = AesCtr::new(&tg_enc_key, tg_enc_iv);
         
-        let tg_reader = CryptoReader::new(read_half, decryptor);
-        let tg_writer = CryptoWriter::new(write_half, encryptor);
-        
-        Ok((tg_reader, tg_writer))
+        Ok((
+            CryptoReader::new(read_half, decryptor),
+            CryptoWriter::new(write_half, encryptor),
+        ))
     }
 }

src/proxy/handshake.rs

@@ -1,11 +1,11 @@
-//! MTProto Handshake Magics
+//! MTProto Handshake
 
 use std::net::SocketAddr;
 use tokio::io::{AsyncRead, AsyncWrite, AsyncWriteExt};
 use tracing::{debug, warn, trace, info};
+use zeroize::Zeroize;
 
-use crate::crypto::{sha256, AesCtr};
-use crate::crypto::random::SECURE_RANDOM;
+use crate::crypto::{sha256, AesCtr, SecureRandom};
 use crate::protocol::constants::*;
 use crate::protocol::tls;
 use crate::stream::{FakeTlsReader, FakeTlsWriter, CryptoReader, CryptoWriter};
@@ -14,6 +14,9 @@ use crate::stats::ReplayChecker;
 use crate::config::ProxyConfig;
 
 /// Result of successful handshake
+///
+/// Key material (`dec_key`, `dec_iv`, `enc_key`, `enc_iv`) is
+/// zeroized on drop.
 #[derive(Debug, Clone)]
 pub struct HandshakeSuccess {
     /// Authenticated user name
@@ -34,6 +37,15 @@ pub struct HandshakeSuccess {
     pub is_tls: bool,
 }
 
+impl Drop for HandshakeSuccess {
+    fn drop(&mut self) {
+        self.dec_key.zeroize();
+        self.dec_iv.zeroize();
+        self.enc_key.zeroize();
+        self.enc_iv.zeroize();
+    }
+}
+
 /// Handle fake TLS handshake
 pub async fn handle_tls_handshake<R, W>(
     handshake: &[u8],
@@ -42,76 +54,74 @@ pub async fn handle_tls_handshake<R, W>(
     peer: SocketAddr,
     config: &ProxyConfig,
     replay_checker: &ReplayChecker,
-) -> HandshakeResult<(FakeTlsReader<R>, FakeTlsWriter<W>, String)>
+    rng: &SecureRandom,
+) -> HandshakeResult<(FakeTlsReader<R>, FakeTlsWriter<W>, String), R, W>
 where
     R: AsyncRead + Unpin,
     W: AsyncWrite + Unpin,
 {
     debug!(peer = %peer, handshake_len = handshake.len(), "Processing TLS handshake");
     
-    // Check minimum length
     if handshake.len() < tls::TLS_DIGEST_POS + tls::TLS_DIGEST_LEN + 1 {
         debug!(peer = %peer, "TLS handshake too short");
-        return HandshakeResult::BadClient;
+        return HandshakeResult::BadClient { reader, writer };
     }
     
-    // Extract digest for replay check
     let digest = &handshake[tls::TLS_DIGEST_POS..tls::TLS_DIGEST_POS + tls::TLS_DIGEST_LEN];
     let digest_half = &digest[..tls::TLS_DIGEST_HALF_LEN];
     
-    // Check for replay
     if replay_checker.check_tls_digest(digest_half) {
-        warn!(peer = %peer, "TLS replay attack detected");
-        return HandshakeResult::BadClient;
+        warn!(peer = %peer, "TLS replay attack detected (duplicate digest)");
+        return HandshakeResult::BadClient { reader, writer };
     }
     
-    // Build secrets list
-    let secrets: Vec<(String, Vec<u8>)> = config.users.iter()
+    let secrets: Vec<(String, Vec<u8>)> = config.access.users.iter()
         .filter_map(|(name, hex)| {
             hex::decode(hex).ok().map(|bytes| (name.clone(), bytes))
         })
         .collect();
     
-    debug!(peer = %peer, num_users = secrets.len(), "Validating TLS handshake against users");
-    
-    // Validate handshake
     let validation = match tls::validate_tls_handshake(
         handshake,
         &secrets,
-        config.ignore_time_skew,
+        config.access.ignore_time_skew,
     ) {
         Some(v) => v,
         None => {
-            debug!(peer = %peer, "TLS handshake validation failed - no matching user");
-            return HandshakeResult::BadClient;
+            debug!(
+                peer = %peer, 
+                ignore_time_skew = config.access.ignore_time_skew,
+                "TLS handshake validation failed - no matching user or time skew"
+            );
+            return HandshakeResult::BadClient { reader, writer };
         }
     };
     
-    // Get secret for response
     let secret = match secrets.iter().find(|(name, _)| *name == validation.user) {
         Some((_, s)) => s,
-        None => return HandshakeResult::BadClient,
+        None => return HandshakeResult::BadClient { reader, writer },
     };
     
-    // Build and send response
     let response = tls::build_server_hello(
         secret,
         &validation.digest,
         &validation.session_id,
-        config.fake_cert_len,
+        config.censorship.fake_cert_len,
+        rng,
     );
     
     debug!(peer = %peer, response_len = response.len(), "Sending TLS ServerHello");
     
     if let Err(e) = writer.write_all(&response).await {
+        warn!(peer = %peer, error = %e, "Failed to write TLS ServerHello");
         return HandshakeResult::Error(ProxyError::Io(e));
     }
     
     if let Err(e) = writer.flush().await {
+        warn!(peer = %peer, error = %e, "Failed to flush TLS ServerHello");
         return HandshakeResult::Error(ProxyError::Io(e));
     }
     
-    // Record for replay protection
     replay_checker.add_tls_digest(digest_half);
     
     info!(
@@ -136,39 +146,28 @@ pub async fn handle_mtproto_handshake<R, W>(
     config: &ProxyConfig,
     replay_checker: &ReplayChecker,
     is_tls: bool,
-) -> HandshakeResult<(CryptoReader<R>, CryptoWriter<W>, HandshakeSuccess)>
+) -> HandshakeResult<(CryptoReader<R>, CryptoWriter<W>, HandshakeSuccess), R, W>
 where
     R: AsyncRead + Unpin + Send,
     W: AsyncWrite + Unpin + Send,
 {
     trace!(peer = %peer, handshake = ?hex::encode(handshake), "MTProto handshake bytes");
     
-    // Extract prekey and IV
     let dec_prekey_iv = &handshake[SKIP_LEN..SKIP_LEN + PREKEY_LEN + IV_LEN];
     
-    debug!(
-        peer = %peer,
-        dec_prekey_iv = %hex::encode(dec_prekey_iv),
-        "Extracted prekey+IV from handshake"
-    );
-    
-    // Check for replay
     if replay_checker.check_handshake(dec_prekey_iv) {
         warn!(peer = %peer, "MTProto replay attack detected");
-        return HandshakeResult::BadClient;
+        return HandshakeResult::BadClient { reader, writer };
     }
     
-    // Reversed for encryption direction
     let enc_prekey_iv: Vec<u8> = dec_prekey_iv.iter().rev().copied().collect();
     
-    // Try each user's secret
-    for (user, secret_hex) in &config.users {
+    for (user, secret_hex) in &config.access.users {
         let secret = match hex::decode(secret_hex) {
             Ok(s) => s,
             Err(_) => continue,
         };
         
-        // Derive decryption key
         let dec_prekey = &dec_prekey_iv[..PREKEY_LEN];
         let dec_iv_bytes = &dec_prekey_iv[PREKEY_LEN..];
         
@@ -179,38 +178,23 @@ where
         
         let dec_iv = u128::from_be_bytes(dec_iv_bytes.try_into().unwrap());
         
-        // Decrypt handshake to check protocol tag
         let mut decryptor = AesCtr::new(&dec_key, dec_iv);
         let decrypted = decryptor.decrypt(handshake);
         
-        trace!(
-            peer = %peer,
-            user = %user,
-            decrypted_tail = %hex::encode(&decrypted[PROTO_TAG_POS..]),
-            "Decrypted handshake tail"
-        );
-        
-        // Check protocol tag
         let tag_bytes: [u8; 4] = decrypted[PROTO_TAG_POS..PROTO_TAG_POS + 4]
             .try_into()
             .unwrap();
         
         let proto_tag = match ProtoTag::from_bytes(tag_bytes) {
             Some(tag) => tag,
-            None => {
-                trace!(peer = %peer, user = %user, tag = %hex::encode(tag_bytes), "Invalid proto tag");
-                continue;
-            }
+            None => continue,
         };
         
-        debug!(peer = %peer, user = %user, proto = ?proto_tag, "Found valid proto tag");
-        
-        // Check if mode is enabled
         let mode_ok = match proto_tag {
             ProtoTag::Secure => {
-                if is_tls { config.modes.tls } else { config.modes.secure }
+                if is_tls { config.general.modes.tls } else { config.general.modes.secure }
             }
-            ProtoTag::Intermediate | ProtoTag::Abridged => config.modes.classic,
+            ProtoTag::Intermediate | ProtoTag::Abridged => config.general.modes.classic,
         };
         
         if !mode_ok {
@@ -218,12 +202,10 @@ where
             continue;
         }
         
-        // Extract DC index
         let dc_idx = i16::from_le_bytes(
             decrypted[DC_IDX_POS..DC_IDX_POS + 2].try_into().unwrap()
         );
         
-        // Derive encryption key
         let enc_prekey = &enc_prekey_iv[..PREKEY_LEN];
         let enc_iv_bytes = &enc_prekey_iv[PREKEY_LEN..];
         
@@ -234,10 +216,8 @@ where
         
         let enc_iv = u128::from_be_bytes(enc_iv_bytes.try_into().unwrap());
         
-        // Record for replay protection
         replay_checker.add_handshake(dec_prekey_iv);
         
-        // Create new cipher instances
         let decryptor = AesCtr::new(&dec_key, dec_iv);
         let encryptor = AesCtr::new(&enc_key, enc_iv);
         
@@ -270,56 +250,37 @@ where
     }
     
     debug!(peer = %peer, "MTProto handshake: no matching user found");
-    HandshakeResult::BadClient
+    HandshakeResult::BadClient { reader, writer }
 }
 
 /// Generate nonce for Telegram connection
-/// 
-/// In FAST MODE: we use the same keys for TG as for client, but reversed.
-/// This means: client's enc_key becomes TG's dec_key and vice versa.
 pub fn generate_tg_nonce(
     proto_tag: ProtoTag, 
     client_dec_key: &[u8; 32],
     client_dec_iv: u128,
+    rng: &SecureRandom,
     fast_mode: bool,
 ) -> ([u8; HANDSHAKE_LEN], [u8; 32], u128, [u8; 32], u128) {
     loop {
-        let bytes = SECURE_RANDOM.bytes(HANDSHAKE_LEN);
+        let bytes = rng.bytes(HANDSHAKE_LEN);
         let mut nonce: [u8; HANDSHAKE_LEN] = bytes.try_into().unwrap();
         
-        // Check reserved patterns
-        if RESERVED_NONCE_FIRST_BYTES.contains(&nonce[0]) {
-            continue;
-        }
+        if RESERVED_NONCE_FIRST_BYTES.contains(&nonce[0]) { continue; }
         
         let first_four: [u8; 4] = nonce[..4].try_into().unwrap();
-        if RESERVED_NONCE_BEGINNINGS.contains(&first_four) {
-            continue;
-        }
+        if RESERVED_NONCE_BEGINNINGS.contains(&first_four) { continue; }
         
         let continue_four: [u8; 4] = nonce[4..8].try_into().unwrap();
-        if RESERVED_NONCE_CONTINUES.contains(&continue_four) {
-            continue;
-        }
+        if RESERVED_NONCE_CONTINUES.contains(&continue_four) { continue; }
         
-        // Set protocol tag
         nonce[PROTO_TAG_POS..PROTO_TAG_POS + 4].copy_from_slice(&proto_tag.to_bytes());
         
-        // Fast mode: copy client's dec_key+iv (this becomes TG's enc direction)
-        // In fast mode, we make TG use the same keys as client but swapped:
-        // - What we decrypt FROM TG = what we encrypt TO client (so no re-encryption needed)
-        // - What we encrypt TO TG = what we decrypt FROM client
         if fast_mode {
-            // Put client's dec_key + dec_iv into nonce[8:56]
-            // This will be used by TG for encryption TO us
             nonce[SKIP_LEN..SKIP_LEN + KEY_LEN].copy_from_slice(client_dec_key);
             nonce[SKIP_LEN + KEY_LEN..SKIP_LEN + KEY_LEN + IV_LEN]
                 .copy_from_slice(&client_dec_iv.to_be_bytes());
         }
         
-        // Now compute what keys WE will use for TG connection
-        // enc_key_iv = nonce[8:56] (for encrypting TO TG)
-        // dec_key_iv = nonce[8:56] reversed (for decrypting FROM TG)
         let enc_key_iv = &nonce[SKIP_LEN..SKIP_LEN + KEY_LEN + IV_LEN];
         let dec_key_iv: Vec<u8> = enc_key_iv.iter().rev().copied().collect();
         
@@ -329,44 +290,22 @@ pub fn generate_tg_nonce(
         let tg_dec_key: [u8; 32] = dec_key_iv[..KEY_LEN].try_into().unwrap();
         let tg_dec_iv = u128::from_be_bytes(dec_key_iv[KEY_LEN..].try_into().unwrap());
         
-        debug!(
-            fast_mode = fast_mode,
-            tg_enc_key = %hex::encode(&tg_enc_key[..8]),
-            tg_dec_key = %hex::encode(&tg_dec_key[..8]),
-            "Generated TG nonce"
-        );
-        
         return (nonce, tg_enc_key, tg_enc_iv, tg_dec_key, tg_dec_iv);
     }
 }
 
 /// Encrypt nonce for sending to Telegram
-/// 
-/// Only the part from PROTO_TAG_POS onwards is encrypted.
-/// The encryption key is derived from enc_key_iv in the nonce itself.
 pub fn encrypt_tg_nonce(nonce: &[u8; HANDSHAKE_LEN]) -> Vec<u8> {
-    // enc_key_iv is at nonce[8:56]
     let enc_key_iv = &nonce[SKIP_LEN..SKIP_LEN + KEY_LEN + IV_LEN];
-    
-    // Key for encrypting is just the first 32 bytes of enc_key_iv
     let key: [u8; 32] = enc_key_iv[..KEY_LEN].try_into().unwrap();
     let iv = u128::from_be_bytes(enc_key_iv[KEY_LEN..].try_into().unwrap());
     
     let mut encryptor = AesCtr::new(&key, iv);
-    
-    // Encrypt the entire nonce first, then take only the encrypted tail
     let encrypted_full = encryptor.encrypt(nonce);
     
-    // Result: unencrypted head + encrypted tail
     let mut result = nonce[..PROTO_TAG_POS].to_vec();
     result.extend_from_slice(&encrypted_full[PROTO_TAG_POS..]);
     
-    trace!(
-        original = %hex::encode(&nonce[PROTO_TAG_POS..]),
-        encrypted = %hex::encode(&result[PROTO_TAG_POS..]),
-        "Encrypted nonce tail"
-    );
-    
     result
 }
 
@@ -379,13 +318,12 @@ mod tests {
         let client_dec_key = [0x42u8; 32];
         let client_dec_iv = 12345u128;
         
-        let (nonce, tg_enc_key, tg_enc_iv, tg_dec_key, tg_dec_iv) = 
-            generate_tg_nonce(ProtoTag::Secure, &client_dec_key, client_dec_iv, false);
+        let rng = SecureRandom::new();
+        let (nonce, _tg_enc_key, _tg_enc_iv, _tg_dec_key, _tg_dec_iv) = 
+            generate_tg_nonce(ProtoTag::Secure, &client_dec_key, client_dec_iv, &rng, false);
         
-        // Check length
         assert_eq!(nonce.len(), HANDSHAKE_LEN);
         
-        // Check proto tag is set
         let tag_bytes: [u8; 4] = nonce[PROTO_TAG_POS..PROTO_TAG_POS + 4].try_into().unwrap();
         assert_eq!(ProtoTag::from_bytes(tag_bytes), Some(ProtoTag::Secure));
     }
@@ -395,17 +333,35 @@ mod tests {
         let client_dec_key = [0x42u8; 32];
         let client_dec_iv = 12345u128;
         
+        let rng = SecureRandom::new();
         let (nonce, _, _, _, _) = 
-            generate_tg_nonce(ProtoTag::Secure, &client_dec_key, client_dec_iv, false);
+            generate_tg_nonce(ProtoTag::Secure, &client_dec_key, client_dec_iv, &rng, false);
         
         let encrypted = encrypt_tg_nonce(&nonce);
         
         assert_eq!(encrypted.len(), HANDSHAKE_LEN);
-        
-        // First PROTO_TAG_POS bytes should be unchanged
         assert_eq!(&encrypted[..PROTO_TAG_POS], &nonce[..PROTO_TAG_POS]);
-        
-        // Rest should be different (encrypted)
         assert_ne!(&encrypted[PROTO_TAG_POS..], &nonce[PROTO_TAG_POS..]);
     }
+    
+    #[test]
+    fn test_handshake_success_zeroize_on_drop() {
+        let success = HandshakeSuccess {
+            user: "test".to_string(),
+            dc_idx: 2,
+            proto_tag: ProtoTag::Secure,
+            dec_key: [0xAA; 32],
+            dec_iv: 0xBBBBBBBB,
+            enc_key: [0xCC; 32],
+            enc_iv: 0xDDDDDDDD,
+            peer: "127.0.0.1:1234".parse().unwrap(),
+            is_tls: true,
+        };
+        
+        assert_eq!(success.dec_key, [0xAA; 32]);
+        assert_eq!(success.enc_key, [0xCC; 32]);
+        
+        drop(success);
+        // Drop impl zeroizes key material without panic
+    }
 }

src/proxy/masking.rs

@@ -1,35 +1,76 @@
 //! Masking - forward unrecognized traffic to mask host
 
 use std::time::Duration;
+use std::str;
 use tokio::net::TcpStream;
-use tokio::io::{AsyncReadExt, AsyncWriteExt};
+use tokio::io::{AsyncRead, AsyncWrite, AsyncReadExt, AsyncWriteExt};
 use tokio::time::timeout;
 use tracing::debug;
 use crate::config::ProxyConfig;
-use crate::transport::set_linger_zero;
 
 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);
 const MASK_BUFFER_SIZE: usize = 8192;
 
+/// Detect client type based on initial data
+fn detect_client_type(data: &[u8]) -> &'static str {
+    // Check for HTTP request
+    if data.len() > 4 {
+        if data.starts_with(b"GET ") || data.starts_with(b"POST") ||
+           data.starts_with(b"HEAD") || data.starts_with(b"PUT ") ||
+           data.starts_with(b"DELETE") || data.starts_with(b"OPTIONS") {
+            return "HTTP";
+        }
+    }
+    
+    // Check for TLS ClientHello (0x16 = handshake, 0x03 0x01-0x03 = TLS version)
+    if data.len() > 3 && data[0] == 0x16 && data[1] == 0x03 {
+        return "TLS-scanner";
+    }
+    
+    // Check for SSH
+    if data.starts_with(b"SSH-") {
+        return "SSH";
+    }
+    
+    // Port scanner (very short data)
+    if data.len() < 10 {
+        return "port-scanner";
+    }
+    
+    "unknown"
+}
+
 /// Handle a bad client by forwarding to mask host
-pub async fn handle_bad_client(
-    mut client: TcpStream,
+pub async fn handle_bad_client<R, W>(
+    mut reader: R,
+    mut writer: W,
     initial_data: &[u8],
     config: &ProxyConfig,
-) {
-    if !config.mask {
+) 
+where 
+    R: AsyncRead + Unpin + Send + 'static,
+    W: AsyncWrite + Unpin + Send + 'static,
+{
+    if !config.censorship.mask {
         // Masking disabled, just consume data
-        consume_client_data(client).await;
+        consume_client_data(reader).await;
         return;
     }
     
-    let mask_host = config.mask_host.as_deref()
-        .unwrap_or(&config.tls_domain);
-    let mask_port = config.mask_port;
+    let client_type = detect_client_type(initial_data);
+    
+    let mask_host = config.censorship.mask_host.as_deref()
+        .unwrap_or(&config.censorship.tls_domain);
+    let mask_port = config.censorship.mask_port;
     
     debug!(
+        client_type = client_type,
         host = %mask_host,
         port = mask_port,
+        data_len = initial_data.len(),
         "Forwarding bad client to mask host"
     );
     
@@ -40,33 +81,32 @@ pub async fn handle_bad_client(
         TcpStream::connect(&mask_addr)
     ).await;
     
-    let mut mask_stream = match connect_result {
+    let mask_stream = match connect_result {
         Ok(Ok(s)) => s,
         Ok(Err(e)) => {
             debug!(error = %e, "Failed to connect to mask host");
-            consume_client_data(client).await;
+            consume_client_data(reader).await;
             return;
         }
         Err(_) => {
             debug!("Timeout connecting to mask host");
-            consume_client_data(client).await;
+            consume_client_data(reader).await;
             return;
         }
     };
     
+    let (mut mask_read, mut mask_write) = mask_stream.into_split();
+    
     // Send initial data to mask host
-    if mask_stream.write_all(initial_data).await.is_err() {
+    if mask_write.write_all(initial_data).await.is_err() {
         return;
     }
     
     // Relay traffic
-    let (mut client_read, mut client_write) = client.into_split();
-    let (mut mask_read, mut mask_write) = mask_stream.into_split();
-    
     let c2m = tokio::spawn(async move {
         let mut buf = vec![0u8; MASK_BUFFER_SIZE];
         loop {
-            match client_read.read(&mut buf).await {
+            match reader.read(&mut buf).await {
                 Ok(0) | Err(_) => {
                     let _ = mask_write.shutdown().await;
                     break;
@@ -85,11 +125,11 @@ pub async fn handle_bad_client(
         loop {
             match mask_read.read(&mut buf).await {
                 Ok(0) | Err(_) => {
-                    let _ = client_write.shutdown().await;
+                    let _ = writer.shutdown().await;
                     break;
                 }
                 Ok(n) => {
-                    if client_write.write_all(&buf[..n]).await.is_err() {
+                    if writer.write_all(&buf[..n]).await.is_err() {
                         break;
                     }
                 }
@@ -105,9 +145,9 @@ pub async fn handle_bad_client(
 }
 
 /// Just consume all data from client without responding
-async fn consume_client_data(mut client: TcpStream) {
+async fn consume_client_data<R: AsyncRead + Unpin>(mut reader: R) {
     let mut buf = vec![0u8; MASK_BUFFER_SIZE];
-    while let Ok(n) = client.read(&mut buf).await {
+    while let Ok(n) = reader.read(&mut buf).await {
         if n == 0 {
             break;
         }

src/proxy/relay.rs

@@ -1,22 +1,320 @@
-//! Bidirectional Relay
+//! Bidirectional Relay — poll-based, no head-of-line blocking
+//!
+//! ## What changed and why
+//!
+//! Previous implementation used a single-task `select! { biased; ... }` loop
+//! where each branch called `write_all()`. This caused head-of-line blocking:
+//! while `write_all()` waited for a slow writer (e.g. client on 3G downloading
+//! media), the entire loop was blocked — the other direction couldn't make progress.
+//!
+//! Symptoms observed in production:
+//! - Media loading at ~8 KB/s despite fast server connection
+//! - Stop-and-go pattern with 50–500ms gaps between chunks
+//! - `biased` select starving S→C direction
+//! - Some users unable to load media at all
+//!
+//! ## New architecture
+//!
+//! Uses `tokio::io::copy_bidirectional` which polls both directions concurrently
+//! in a single task via non-blocking `poll_read` / `poll_write` calls:
+//!
+//! Old (select! + write_all — BLOCKING):
+//!
+//!   loop {
+//!       select! {
+//!           biased;
+//!           data = client.read()  => { server.write_all(data).await; }  ← BLOCKS here
+//!           data = server.read()  => { client.write_all(data).await; }  ← can't run
+//!       }
+//!   }
+//!
+//! New (copy_bidirectional — CONCURRENT):
+//!
+//!   poll(cx) {
+//!       // Both directions polled in the same poll cycle
+//!       C→S: poll_read(client) → poll_write(server)   // non-blocking
+//!       S→C: poll_read(server) → poll_write(client)   // non-blocking
+//!       // If one writer is Pending, the other direction still progresses
+//!   }
+//!
+//! Benefits:
+//! - No head-of-line blocking: slow client download doesn't block uploads
+//! - No biased starvation: fair polling of both directions
+//! - Proper flush: `copy_bidirectional` calls `poll_flush` when reader stalls,
+//!   so CryptoWriter's pending ciphertext is always drained (fixes "stuck at 95%")
+//! - No deadlock risk: old write_all could deadlock when both TCP buffers filled;
+//!   poll-based approach lets TCP flow control work correctly
+//!
+//! Stats tracking:
+//! - `StatsIo` wraps client side, intercepts `poll_read` / `poll_write`
+//! - `poll_read` on client = C→S (client sending) → `octets_from`, `msgs_from`
+//! - `poll_write` on client = S→C (to client)     → `octets_to`, `msgs_to`
+//! - `SharedCounters` (atomics) let the watchdog read stats without locking
 
+use std::io;
+use std::pin::Pin;
 use std::sync::Arc;
-use tokio::io::{AsyncRead, AsyncWrite, AsyncReadExt, AsyncWriteExt};
+use std::sync::atomic::{AtomicU64, Ordering};
+use std::task::{Context, Poll};
+use std::time::Duration;
+use tokio::io::{AsyncRead, AsyncWrite, AsyncWriteExt, ReadBuf, copy_bidirectional};
+use tokio::time::Instant;
 use tracing::{debug, trace, warn};
 use crate::error::Result;
 use crate::stats::Stats;
-use std::sync::atomic::{AtomicU64, Ordering};
+use crate::stream::BufferPool;
 
-const BUFFER_SIZE: usize = 65536;
+// ============= Constants =============
 
-/// Relay data bidirectionally between client and server
+/// Activity timeout for iOS compatibility.
+///
+/// iOS keeps Telegram connections alive in background for up to 30 minutes.
+/// Closing earlier causes unnecessary reconnects and handshake overhead.
+const ACTIVITY_TIMEOUT: Duration = Duration::from_secs(1800);
+
+/// Watchdog check interval — also used for periodic rate logging.
+///
+/// 10 seconds gives responsive timeout detection (±10s accuracy)
+/// without measurable overhead from atomic reads.
+const WATCHDOG_INTERVAL: Duration = Duration::from_secs(10);
+
+// ============= CombinedStream =============
+
+/// Combines separate read and write halves into a single bidirectional stream.
+///
+/// `copy_bidirectional` requires `AsyncRead + AsyncWrite` on each side,
+/// but the handshake layer produces split reader/writer pairs
+/// (e.g. `CryptoReader<FakeTlsReader<OwnedReadHalf>>` + `CryptoWriter<...>`).
+///
+/// This wrapper reunifies them with zero overhead — each trait method
+/// delegates directly to the corresponding half. No buffering, no copies.
+///
+/// Safety: `poll_read` only touches `reader`, `poll_write` only touches `writer`,
+/// so there's no aliasing even though both are called on the same `&mut self`.
+struct CombinedStream<R, W> {
+    reader: R,
+    writer: W,
+}
+
+impl<R, W> CombinedStream<R, W> {
+    fn new(reader: R, writer: W) -> Self {
+        Self { reader, writer }
+    }
+}
+
+impl<R: AsyncRead + Unpin, W: Unpin> AsyncRead for CombinedStream<R, W> {
+    #[inline]
+    fn poll_read(
+        self: Pin<&mut Self>,
+        cx: &mut Context<'_>,
+        buf: &mut ReadBuf<'_>,
+    ) -> Poll<io::Result<()>> {
+        Pin::new(&mut self.get_mut().reader).poll_read(cx, buf)
+    }
+}
+
+impl<R: Unpin, W: AsyncWrite + Unpin> AsyncWrite for CombinedStream<R, W> {
+    #[inline]
+    fn poll_write(
+        self: Pin<&mut Self>,
+        cx: &mut Context<'_>,
+        buf: &[u8],
+    ) -> Poll<io::Result<usize>> {
+        Pin::new(&mut self.get_mut().writer).poll_write(cx, buf)
+    }
+
+    #[inline]
+    fn poll_flush(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<io::Result<()>> {
+        Pin::new(&mut self.get_mut().writer).poll_flush(cx)
+    }
+
+    #[inline]
+    fn poll_shutdown(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<io::Result<()>> {
+        Pin::new(&mut self.get_mut().writer).poll_shutdown(cx)
+    }
+}
+
+// ============= SharedCounters =============
+
+/// Atomic counters shared between the relay (via StatsIo) and the watchdog task.
+///
+/// Using `Relaxed` ordering is sufficient because:
+/// - Counters are monotonically increasing (no ABA problem)
+/// - Slight staleness in watchdog reads is harmless (±10s check interval anyway)
+/// - No ordering dependencies between different counters
+struct SharedCounters {
+    /// Bytes read from client (C→S direction)
+    c2s_bytes: AtomicU64,
+    /// Bytes written to client (S→C direction)
+    s2c_bytes: AtomicU64,
+    /// Number of poll_read completions (≈ C→S chunks)
+    c2s_ops: AtomicU64,
+    /// Number of poll_write completions (≈ S→C chunks)
+    s2c_ops: AtomicU64,
+    /// Milliseconds since relay epoch of last I/O activity
+    last_activity_ms: AtomicU64,
+}
+
+impl SharedCounters {
+    fn new() -> Self {
+        Self {
+            c2s_bytes: AtomicU64::new(0),
+            s2c_bytes: AtomicU64::new(0),
+            c2s_ops: AtomicU64::new(0),
+            s2c_ops: AtomicU64::new(0),
+            last_activity_ms: AtomicU64::new(0),
+        }
+    }
+
+    /// Record activity at this instant.
+    #[inline]
+    fn touch(&self, now: Instant, epoch: Instant) {
+        let ms = now.duration_since(epoch).as_millis() as u64;
+        self.last_activity_ms.store(ms, Ordering::Relaxed);
+    }
+
+    /// How long since last recorded activity.
+    fn idle_duration(&self, now: Instant, epoch: Instant) -> Duration {
+        let last_ms = self.last_activity_ms.load(Ordering::Relaxed);
+        let now_ms = now.duration_since(epoch).as_millis() as u64;
+        Duration::from_millis(now_ms.saturating_sub(last_ms))
+    }
+}
+
+// ============= StatsIo =============
+
+/// Transparent I/O wrapper that tracks per-user statistics and activity.
+///
+/// Wraps the **client** side of the relay. Direction mapping:
+///
+/// | poll method  | direction | stats updated                        |
+/// |-------------|-----------|--------------------------------------|
+/// | `poll_read`  | C→S       | `octets_from`, `msgs_from`, counters |
+/// | `poll_write` | S→C       | `octets_to`, `msgs_to`, counters     |
+///
+/// Both update the shared activity timestamp for the watchdog.
+///
+/// Note on message counts: the original code counted one `read()`/`write_all()`
+/// as one "message". Here we count `poll_read`/`poll_write` completions instead.
+/// Byte counts are identical; op counts may differ slightly due to different
+/// internal buffering in `copy_bidirectional`. This is fine for monitoring.
+struct StatsIo<S> {
+    inner: S,
+    counters: Arc<SharedCounters>,
+    stats: Arc<Stats>,
+    user: String,
+    epoch: Instant,
+}
+
+impl<S> StatsIo<S> {
+    fn new(
+        inner: S,
+        counters: Arc<SharedCounters>,
+        stats: Arc<Stats>,
+        user: String,
+        epoch: Instant,
+    ) -> Self {
+        // Mark initial activity so the watchdog doesn't fire before data flows
+        counters.touch(Instant::now(), epoch);
+        Self { inner, counters, stats, user, epoch }
+    }
+}
+
+impl<S: AsyncRead + Unpin> AsyncRead for StatsIo<S> {
+    fn poll_read(
+        self: Pin<&mut Self>,
+        cx: &mut Context<'_>,
+        buf: &mut ReadBuf<'_>,
+    ) -> Poll<io::Result<()>> {
+        let this = self.get_mut();
+        let before = buf.filled().len();
+
+        match Pin::new(&mut this.inner).poll_read(cx, buf) {
+            Poll::Ready(Ok(())) => {
+                let n = buf.filled().len() - before;
+                if n > 0 {
+                    // C→S: client sent data
+                    this.counters.c2s_bytes.fetch_add(n as u64, Ordering::Relaxed);
+                    this.counters.c2s_ops.fetch_add(1, Ordering::Relaxed);
+                    this.counters.touch(Instant::now(), this.epoch);
+
+                    this.stats.add_user_octets_from(&this.user, n as u64);
+                    this.stats.increment_user_msgs_from(&this.user);
+
+                    trace!(user = %this.user, bytes = n, "C->S");
+                }
+                Poll::Ready(Ok(()))
+            }
+            other => other,
+        }
+    }
+}
+
+impl<S: AsyncWrite + Unpin> AsyncWrite for StatsIo<S> {
+    fn poll_write(
+        self: Pin<&mut Self>,
+        cx: &mut Context<'_>,
+        buf: &[u8],
+    ) -> Poll<io::Result<usize>> {
+        let this = self.get_mut();
+
+        match Pin::new(&mut this.inner).poll_write(cx, buf) {
+            Poll::Ready(Ok(n)) => {
+                if n > 0 {
+                    // S→C: data written to client
+                    this.counters.s2c_bytes.fetch_add(n as u64, Ordering::Relaxed);
+                    this.counters.s2c_ops.fetch_add(1, Ordering::Relaxed);
+                    this.counters.touch(Instant::now(), this.epoch);
+
+                    this.stats.add_user_octets_to(&this.user, n as u64);
+                    this.stats.increment_user_msgs_to(&this.user);
+
+                    trace!(user = %this.user, bytes = n, "S->C");
+                }
+                Poll::Ready(Ok(n))
+            }
+            other => other,
+        }
+    }
+
+    #[inline]
+    fn poll_flush(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<io::Result<()>> {
+        Pin::new(&mut self.get_mut().inner).poll_flush(cx)
+    }
+
+    #[inline]
+    fn poll_shutdown(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<io::Result<()>> {
+        Pin::new(&mut self.get_mut().inner).poll_shutdown(cx)
+    }
+}
+
+// ============= Relay =============
+
+/// Relay data bidirectionally between client and server.
+///
+/// Uses `tokio::io::copy_bidirectional` for concurrent, non-blocking data transfer.
+///
+/// ## API compatibility
+///
+/// Signature is identical to the previous implementation. The `_buffer_pool`
+/// parameter is retained for call-site compatibility — `copy_bidirectional`
+/// manages its own internal buffers (8 KB per direction).
+///
+/// ## Guarantees preserved
+///
+/// - Activity timeout: 30 minutes of inactivity → clean shutdown
+/// - Per-user stats: bytes and ops counted per direction
+/// - Periodic rate logging: every 10 seconds when active
+/// - Clean shutdown: both write sides are shut down on exit
+/// - Error propagation: I/O errors are returned as `ProxyError::Io`
 pub async fn relay_bidirectional<CR, CW, SR, SW>(
-    mut client_reader: CR,
-    mut client_writer: CW,
-    mut server_reader: SR,
-    mut server_writer: SW,
+    client_reader: CR,
+    client_writer: CW,
+    server_reader: SR,
+    server_writer: SW,
     user: &str,
     stats: Arc<Stats>,
+    _buffer_pool: Arc<BufferPool>,
 ) -> Result<()>
 where
     CR: AsyncRead + Unpin + Send + 'static,
@@ -24,139 +322,145 @@ where
     SR: AsyncRead + Unpin + Send + 'static,
     SW: AsyncWrite + Unpin + Send + 'static,
 {
-    let user_c2s = user.to_string();
-    let user_s2c = user.to_string();
+    let epoch = Instant::now();
+    let counters = Arc::new(SharedCounters::new());
+    let user_owned = user.to_string();
 
-    // Используем Arc::clone вместо stats.clone()
-    let stats_c2s = Arc::clone(&stats);
-    let stats_s2c = Arc::clone(&stats);
+    // ── Combine split halves into bidirectional streams ──────────────
+    let client_combined = CombinedStream::new(client_reader, client_writer);
+    let mut server = CombinedStream::new(server_reader, server_writer);
 
-    let c2s_bytes = Arc::new(AtomicU64::new(0));
-    let s2c_bytes = Arc::new(AtomicU64::new(0));
-    let c2s_bytes_clone = Arc::clone(&c2s_bytes);
-    let s2c_bytes_clone = Arc::clone(&s2c_bytes);
+    // Wrap client with stats/activity tracking
+    let mut client = StatsIo::new(
+        client_combined,
+        Arc::clone(&counters),
+        Arc::clone(&stats),
+        user_owned.clone(),
+        epoch,
+    );
 
-    // Client -> Server task
-    let c2s = tokio::spawn(async move {
-        let mut buf = vec![0u8; BUFFER_SIZE];
-        let mut total_bytes = 0u64;
-        let mut msg_count = 0u64;
+    // ── Watchdog: activity timeout + periodic rate logging ──────────
+    let wd_counters = Arc::clone(&counters);
+    let wd_user = user_owned.clone();
+
+    let watchdog = async {
+        let mut prev_c2s: u64 = 0;
+        let mut prev_s2c: u64 = 0;
 
         loop {
-            match client_reader.read(&mut buf).await {
-                Ok(0) => {
+            tokio::time::sleep(WATCHDOG_INTERVAL).await;
+
+            let now = Instant::now();
+            let idle = wd_counters.idle_duration(now, epoch);
+
+            // ── Activity timeout ────────────────────────────────────
+            if idle >= ACTIVITY_TIMEOUT {
+                let c2s = wd_counters.c2s_bytes.load(Ordering::Relaxed);
+                let s2c = wd_counters.s2c_bytes.load(Ordering::Relaxed);
+                warn!(
+                    user = %wd_user,
+                    c2s_bytes = c2s,
+                    s2c_bytes = s2c,
+                    idle_secs = idle.as_secs(),
+                    "Activity timeout"
+                );
+                return; // Causes select! to cancel copy_bidirectional
+            }
+
+            // ── Periodic rate logging ───────────────────────────────
+            let c2s = wd_counters.c2s_bytes.load(Ordering::Relaxed);
+            let s2c = wd_counters.s2c_bytes.load(Ordering::Relaxed);
+            let c2s_delta = c2s - prev_c2s;
+            let s2c_delta = s2c - prev_s2c;
+
+            if c2s_delta > 0 || s2c_delta > 0 {
+                let secs = WATCHDOG_INTERVAL.as_secs_f64();
                 debug!(
-                        user = %user_c2s, 
-                        total_bytes = total_bytes,
-                        msgs = msg_count,
-                        "Client closed connection (C->S)"
+                    user = %wd_user,
+                    c2s_kbps = (c2s_delta as f64 / secs / 1024.0) as u64,
+                    s2c_kbps = (s2c_delta as f64 / secs / 1024.0) as u64,
+                    c2s_total = c2s,
+                    s2c_total = s2c,
+                    "Relay active"
                 );
-                    let _ = server_writer.shutdown().await;
-                    break;
             }
-                Ok(n) => {
-                    total_bytes += n as u64;
-                    msg_count += 1;
-                    c2s_bytes_clone.store(total_bytes, Ordering::Relaxed);
 
-                    stats_c2s.add_user_octets_from(&user_c2s, n as u64);
-                    stats_c2s.increment_user_msgs_from(&user_c2s);
+            prev_c2s = c2s;
+            prev_s2c = s2c;
+        }
+    };
 
-                    trace!(
-                        user = %user_c2s,
-                        bytes = n,
-                        total = total_bytes,
-                        data_preview = %hex::encode(&buf[..n.min(32)]),
-                        "C->S data"
-                    );
+    // ── Run bidirectional copy + watchdog concurrently ───────────────
+    //
+    // copy_bidirectional polls both directions in the same poll() call:
+    //   C→S: poll_read(client/StatsIo) → poll_write(server)
+    //   S→C: poll_read(server)         → poll_write(client/StatsIo)
+    //
+    // When one direction's writer returns Pending, the other direction
+    // continues — no head-of-line blocking.
+    //
+    // When the watchdog fires, select! drops the copy future,
+    // releasing the &mut borrows on client and server.
+    let copy_result = tokio::select! {
+        result = copy_bidirectional(&mut client, &mut server) => Some(result),
+        _ = watchdog => None, // Activity timeout — cancel relay
+    };
 
-                    if let Err(e) = server_writer.write_all(&buf[..n]).await {
-                        debug!(user = %user_c2s, error = %e, "Failed to write to server");
-                        break;
-                    }
-                    if let Err(e) = server_writer.flush().await {
-                        debug!(user = %user_c2s, error = %e, "Failed to flush to server");
-                        break;
-                    }
-                }
-                Err(e) => {
-                    debug!(user = %user_c2s, error = %e, total_bytes = total_bytes, "Client read error");
-                    break;
-                }
-            }
-        }
-    });
+    // ── Clean shutdown ──────────────────────────────────────────────
+    // After select!, the losing future is dropped, borrows released.
+    // Shut down both write sides for clean TCP FIN.
+    let _ = client.shutdown().await;
+    let _ = server.shutdown().await;
 
-    // Server -> Client task
-    let s2c = tokio::spawn(async move {
-        let mut buf = vec![0u8; BUFFER_SIZE];
-        let mut total_bytes = 0u64;
-        let mut msg_count = 0u64;
+    // ── Final logging ───────────────────────────────────────────────
+    let c2s_ops = counters.c2s_ops.load(Ordering::Relaxed);
+    let s2c_ops = counters.s2c_ops.load(Ordering::Relaxed);
+    let duration = epoch.elapsed();
 
-        loop {
-            match server_reader.read(&mut buf).await {
-                Ok(0) => {
+    match copy_result {
+        Some(Ok((c2s, s2c))) => {
+            // Normal completion — one side closed the connection
             debug!(
-                        user = %user_s2c,
-                        total_bytes = total_bytes,
-                        msgs = msg_count,
-                        "Server closed connection (S->C)"
-                    );
-                    let _ = client_writer.shutdown().await;
-                    break;
-                }
-                Ok(n) => {
-                    total_bytes += n as u64;
-                    msg_count += 1;
-                    s2c_bytes_clone.store(total_bytes, Ordering::Relaxed);
-                    
-                    stats_s2c.add_user_octets_to(&user_s2c, n as u64);
-                    stats_s2c.increment_user_msgs_to(&user_s2c);
-                    
-                    trace!(
-                        user = %user_s2c,
-                        bytes = n,
-                        total = total_bytes,
-                        data_preview = %hex::encode(&buf[..n.min(32)]),
-                        "S->C data"
-                    );
-                    
-                    if let Err(e) = client_writer.write_all(&buf[..n]).await {
-                        debug!(user = %user_s2c, error = %e, "Failed to write to client");
-                        break;
-                    }
-                    if let Err(e) = client_writer.flush().await {
-                        debug!(user = %user_s2c, error = %e, "Failed to flush to client");
-                        break;
-                    }
-                }
-                Err(e) => {
-                    debug!(user = %user_s2c, error = %e, total_bytes = total_bytes, "Server read error");
-                    break;
-                }
-            }
-        }
-    });
-    
-    // Wait for either direction to complete
-    tokio::select! {
-        result = c2s => {
-            if let Err(e) = result {
-                warn!(error = %e, "C->S task panicked");
-            }
-        }
-        result = s2c => {
-            if let Err(e) = result {
-                warn!(error = %e, "S->C task panicked");
-            }
-        }
-    }
-    
-    debug!(
-        c2s_bytes = c2s_bytes.load(Ordering::Relaxed),
-        s2c_bytes = s2c_bytes.load(Ordering::Relaxed),
+                user = %user_owned,
+                c2s_bytes = c2s,
+                s2c_bytes = s2c,
+                c2s_msgs = c2s_ops,
+                s2c_msgs = s2c_ops,
+                duration_secs = duration.as_secs(),
                 "Relay finished"
             );
-    
             Ok(())
         }
+        Some(Err(e)) => {
+            // I/O error in one of the directions
+            let c2s = counters.c2s_bytes.load(Ordering::Relaxed);
+            let s2c = counters.s2c_bytes.load(Ordering::Relaxed);
+            debug!(
+                user = %user_owned,
+                c2s_bytes = c2s,
+                s2c_bytes = s2c,
+                c2s_msgs = c2s_ops,
+                s2c_msgs = s2c_ops,
+                duration_secs = duration.as_secs(),
+                error = %e,
+                "Relay error"
+            );
+            Err(e.into())
+        }
+        None => {
+            // Activity timeout (watchdog fired)
+            let c2s = counters.c2s_bytes.load(Ordering::Relaxed);
+            let s2c = counters.s2c_bytes.load(Ordering::Relaxed);
+            debug!(
+                user = %user_owned,
+                c2s_bytes = c2s,
+                s2c_bytes = s2c,
+                c2s_msgs = c2s_ops,
+                s2c_msgs = s2c_ops,
+                duration_secs = duration.as_secs(),
+                "Relay finished (activity timeout)"
+            );
+            Ok(())
+        }
+    }
+}

src/stats/mod.rs

@@ -1,29 +1,28 @@
-//! Statistics
+//! Statistics and replay protection
 
 use std::sync::atomic::{AtomicU64, Ordering};
 use std::sync::Arc;
-use std::time::Instant;
+use std::time::{Instant, Duration};
 use dashmap::DashMap;
-use parking_lot::RwLock;
+use parking_lot::Mutex;
 use lru::LruCache;
 use std::num::NonZeroUsize;
+use std::hash::{Hash, Hasher};
+use std::collections::hash_map::DefaultHasher;
+use std::collections::VecDeque;
+use tracing::debug;
+
+// ============= Stats =============
 
-/// Thread-safe statistics
 #[derive(Default)]
 pub struct Stats {
-    // Global counters
     connects_all: AtomicU64,
     connects_bad: AtomicU64,
     handshake_timeouts: AtomicU64,
-    
-    // Per-user stats
     user_stats: DashMap<String, UserStats>,
-    
-    // Start time
-    start_time: RwLock<Option<Instant>>,
+    start_time: parking_lot::RwLock<Option<Instant>>,
 }
 
-/// Per-user statistics
 #[derive(Default)]
 pub struct UserStats {
     pub connects: AtomicU64,
@@ -41,42 +40,20 @@ impl Stats {
         stats
     }
     
-    // Global stats
-    pub fn increment_connects_all(&self) {
-        self.connects_all.fetch_add(1, Ordering::Relaxed);
-    }
+    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 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 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 get_connects_all(&self) -> u64 {
-        self.connects_all.load(Ordering::Relaxed)
-    }
-    
-    pub fn get_connects_bad(&self) -> u64 {
-        self.connects_bad.load(Ordering::Relaxed)
-    }
-    
-    // User stats
     pub fn increment_user_connects(&self, user: &str) {
-        self.user_stats
-            .entry(user.to_string())
-            .or_default()
-            .connects
-            .fetch_add(1, Ordering::Relaxed);
+        self.user_stats.entry(user.to_string()).or_default()
+            .connects.fetch_add(1, Ordering::Relaxed);
     }
     
     pub fn increment_user_curr_connects(&self, user: &str) {
-        self.user_stats
-            .entry(user.to_string())
-            .or_default()
-            .curr_connects
-            .fetch_add(1, Ordering::Relaxed);
+        self.user_stats.entry(user.to_string()).or_default()
+            .curr_connects.fetch_add(1, Ordering::Relaxed);
     }
     
     pub fn decrement_user_curr_connects(&self, user: &str) {
@@ -86,47 +63,33 @@ impl Stats {
     }
     
     pub fn get_user_curr_connects(&self, user: &str) -> u64 {
-        self.user_stats
-            .get(user)
+        self.user_stats.get(user)
             .map(|s| s.curr_connects.load(Ordering::Relaxed))
             .unwrap_or(0)
     }
     
     pub fn add_user_octets_from(&self, user: &str, bytes: u64) {
-        self.user_stats
-            .entry(user.to_string())
-            .or_default()
-            .octets_from_client
-            .fetch_add(bytes, Ordering::Relaxed);
+        self.user_stats.entry(user.to_string()).or_default()
+            .octets_from_client.fetch_add(bytes, Ordering::Relaxed);
     }
     
     pub fn add_user_octets_to(&self, user: &str, bytes: u64) {
-        self.user_stats
-            .entry(user.to_string())
-            .or_default()
-            .octets_to_client
-            .fetch_add(bytes, Ordering::Relaxed);
+        self.user_stats.entry(user.to_string()).or_default()
+            .octets_to_client.fetch_add(bytes, Ordering::Relaxed);
     }
     
     pub fn increment_user_msgs_from(&self, user: &str) {
-        self.user_stats
-            .entry(user.to_string())
-            .or_default()
-            .msgs_from_client
-            .fetch_add(1, Ordering::Relaxed);
+        self.user_stats.entry(user.to_string()).or_default()
+            .msgs_from_client.fetch_add(1, Ordering::Relaxed);
     }
     
     pub fn increment_user_msgs_to(&self, user: &str) {
-        self.user_stats
-            .entry(user.to_string())
-            .or_default()
-            .msgs_to_client
-            .fetch_add(1, Ordering::Relaxed);
+        self.user_stats.entry(user.to_string()).or_default()
+            .msgs_to_client.fetch_add(1, Ordering::Relaxed);
     }
     
     pub fn get_user_total_octets(&self, user: &str) -> u64 {
-        self.user_stats
-            .get(user)
+        self.user_stats.get(user)
             .map(|s| {
                 s.octets_from_client.load(Ordering::Relaxed) +
                 s.octets_to_client.load(Ordering::Relaxed)
@@ -141,37 +104,209 @@ impl Stats {
     }
 }
 
-// Arc<Stats> Hightech Stats :D
+// ============= Replay Checker =============
 
-/// Replay attack checker using LRU cache
 pub struct ReplayChecker {
-    handshakes: RwLock<LruCache<Vec<u8>, ()>>,
-    tls_digests: RwLock<LruCache<Vec<u8>, ()>>,
+    shards: Vec<Mutex<ReplayShard>>,
+    shard_mask: usize,
+    window: Duration,
+    checks: AtomicU64,
+    hits: AtomicU64,
+    additions: AtomicU64,
+    cleanups: AtomicU64,
+}
+
+struct ReplayEntry {
+    seen_at: Instant,
+    seq: u64,
+}
+
+struct ReplayShard {
+    cache: LruCache<Box<[u8]>, ReplayEntry>,
+    queue: VecDeque<(Instant, Box<[u8]>, u64)>,
+    seq_counter: u64,
+}
+
+impl ReplayShard {
+    fn new(cap: NonZeroUsize) -> Self {
+        Self {
+            cache: LruCache::new(cap),
+            queue: VecDeque::with_capacity(cap.get()),
+            seq_counter: 0,
+        }
+    }
+    
+    fn next_seq(&mut self) -> u64 {
+        self.seq_counter += 1;
+        self.seq_counter
+    }
+
+    fn cleanup(&mut self, now: Instant, window: Duration) {
+        if window.is_zero() {
+            return;
+        }
+        let cutoff = now.checked_sub(window).unwrap_or(now);
+        
+        while let Some((ts, _, _)) = self.queue.front() {
+            if *ts >= cutoff {
+                break;
+            }
+            let (_, key, queue_seq) = self.queue.pop_front().unwrap();
+            
+            // Use key.as_ref() to get &[u8] — avoids Borrow<Q> ambiguity
+            // between Borrow<[u8]> and Borrow<Box<[u8]>>
+            if let Some(entry) = self.cache.peek(key.as_ref()) {
+                if entry.seq == queue_seq {
+                    self.cache.pop(key.as_ref());
+                }
+            }
+        }
+    }
+    
+    fn check(&mut self, key: &[u8], now: Instant, window: Duration) -> bool {
+        self.cleanup(now, window);
+        // key is &[u8], resolves Q=[u8] via Box<[u8]>: Borrow<[u8]>
+        self.cache.get(key).is_some()
+    }
+    
+    fn add(&mut self, key: &[u8], now: Instant, window: Duration) {
+        self.cleanup(now, window);
+        
+        let seq = self.next_seq();
+        let boxed_key: Box<[u8]> = key.into();
+        
+        self.cache.put(boxed_key.clone(), ReplayEntry { seen_at: now, seq });
+        self.queue.push_back((now, boxed_key, seq));
+    }
+    
+    fn len(&self) -> usize {
+        self.cache.len()
+    }
 }
 
 impl ReplayChecker {
-    pub fn new(capacity: usize) -> Self {
-        let cap = NonZeroUsize::new(capacity.max(1)).unwrap();
+    pub fn new(total_capacity: usize, window: Duration) -> Self {
+        let num_shards = 64;
+        let shard_capacity = (total_capacity / num_shards).max(1);
+        let cap = NonZeroUsize::new(shard_capacity).unwrap();
+
+        let mut shards = Vec::with_capacity(num_shards);
+        for _ in 0..num_shards {
+            shards.push(Mutex::new(ReplayShard::new(cap)));
+        }
+
         Self {
-            handshakes: RwLock::new(LruCache::new(cap)),
-            tls_digests: RwLock::new(LruCache::new(cap)),
+            shards,
+            shard_mask: num_shards - 1,
+            window,
+            checks: AtomicU64::new(0),
+            hits: AtomicU64::new(0),
+            additions: AtomicU64::new(0),
+            cleanups: AtomicU64::new(0),
         }
     }
 
-    pub fn check_handshake(&self, data: &[u8]) -> bool {
-        self.handshakes.read().contains(&data.to_vec())
+    fn get_shard_idx(&self, key: &[u8]) -> usize {
+        let mut hasher = DefaultHasher::new();
+        key.hash(&mut hasher);
+        (hasher.finish() as usize) & self.shard_mask
     }
 
-    pub fn add_handshake(&self, data: &[u8]) {
-        self.handshakes.write().put(data.to_vec(), ());
+    fn check(&self, data: &[u8]) -> bool {
+        self.checks.fetch_add(1, Ordering::Relaxed);
+        let idx = self.get_shard_idx(data);
+        let mut shard = self.shards[idx].lock();
+        let found = shard.check(data, Instant::now(), self.window);
+        if found {
+            self.hits.fetch_add(1, Ordering::Relaxed);
+        }
+        found
     }
 
-    pub fn check_tls_digest(&self, data: &[u8]) -> bool {
-        self.tls_digests.read().contains(&data.to_vec())
+    fn add(&self, data: &[u8]) {
+        self.additions.fetch_add(1, Ordering::Relaxed);
+        let idx = self.get_shard_idx(data);
+        let mut shard = self.shards[idx].lock();
+        shard.add(data, Instant::now(), self.window);
     }
 
-    pub fn add_tls_digest(&self, data: &[u8]) {
-        self.tls_digests.write().put(data.to_vec(), ());
+    pub fn check_handshake(&self, data: &[u8]) -> bool { self.check(data) }
+    pub fn add_handshake(&self, data: &[u8]) { self.add(data) }
+    pub fn check_tls_digest(&self, data: &[u8]) -> bool { self.check(data) }
+    pub fn add_tls_digest(&self, data: &[u8]) { self.add(data) }
+    
+    pub fn stats(&self) -> ReplayStats {
+        let mut total_entries = 0;
+        let mut total_queue_len = 0;
+        for shard in &self.shards {
+            let s = shard.lock();
+            total_entries += s.cache.len();
+            total_queue_len += s.queue.len();
+        }
+        
+        ReplayStats {
+            total_entries,
+            total_queue_len,
+            total_checks: self.checks.load(Ordering::Relaxed),
+            total_hits: self.hits.load(Ordering::Relaxed),
+            total_additions: self.additions.load(Ordering::Relaxed),
+            total_cleanups: self.cleanups.load(Ordering::Relaxed),
+            num_shards: self.shards.len(),
+            window_secs: self.window.as_secs(),
+        }
+    }
+    
+    pub async fn run_periodic_cleanup(&self) {
+        let interval = if self.window.as_secs() > 60 {
+            Duration::from_secs(30)
+        } else {
+            Duration::from_secs(self.window.as_secs().max(1) / 2)
+        };
+        
+        loop {
+            tokio::time::sleep(interval).await;
+            
+            let now = Instant::now();
+            let mut cleaned = 0usize;
+            
+            for shard_mutex in &self.shards {
+                let mut shard = shard_mutex.lock();
+                let before = shard.len();
+                shard.cleanup(now, self.window);
+                let after = shard.len();
+                cleaned += before.saturating_sub(after);
+            }
+            
+            self.cleanups.fetch_add(1, Ordering::Relaxed);
+            
+            if cleaned > 0 {
+                debug!(cleaned = cleaned, "Replay checker: periodic cleanup");
+            }
+        }
+    }
+}
+
+#[derive(Debug, Clone)]
+pub struct ReplayStats {
+    pub total_entries: usize,
+    pub total_queue_len: usize,
+    pub total_checks: u64,
+    pub total_hits: u64,
+    pub total_additions: u64,
+    pub total_cleanups: u64,
+    pub num_shards: usize,
+    pub window_secs: u64,
+}
+
+impl ReplayStats {
+    pub fn hit_rate(&self) -> f64 {
+        if self.total_checks == 0 { 0.0 }
+        else { (self.total_hits as f64 / self.total_checks as f64) * 100.0 }
+    }
+    
+    pub fn ghost_ratio(&self) -> f64 {
+        if self.total_entries == 0 { 0.0 }
+        else { self.total_queue_len as f64 / self.total_entries as f64 }
     }
 }
 
@@ -182,42 +317,60 @@ mod tests {
     #[test]
     fn test_stats_shared_counters() {
         let stats = Arc::new(Stats::new());
-        
-        // Симулируем использование из разных "задач"
-        let stats1 = Arc::clone(&stats);
-        let stats2 = Arc::clone(&stats);
-        
-        stats1.increment_connects_all();
-        stats2.increment_connects_all();
-        stats1.increment_connects_all();
-        
-        // Все инкременты должны быть видны
+        stats.increment_connects_all();
+        stats.increment_connects_all();
+        stats.increment_connects_all();
         assert_eq!(stats.get_connects_all(), 3);
     }
     
     #[test]
-    fn test_user_stats_shared() {
-        let stats = Arc::new(Stats::new());
-        
-        let stats1 = Arc::clone(&stats);
-        let stats2 = Arc::clone(&stats);
-        
-        stats1.add_user_octets_from("user1", 100);
-        stats2.add_user_octets_from("user1", 200);
-        stats1.add_user_octets_to("user1", 50);
-        
-        assert_eq!(stats.get_user_total_octets("user1"), 350);
+    fn test_replay_checker_basic() {
+        let checker = ReplayChecker::new(100, Duration::from_secs(60));
+        assert!(!checker.check_handshake(b"test1"));
+        checker.add_handshake(b"test1");
+        assert!(checker.check_handshake(b"test1"));
+        assert!(!checker.check_handshake(b"test2"));
     }
     
     #[test]
-    fn test_concurrent_user_connects() {
-        let stats = Arc::new(Stats::new());
+    fn test_replay_checker_duplicate_add() {
+        let checker = ReplayChecker::new(100, Duration::from_secs(60));
+        checker.add_handshake(b"dup");
+        checker.add_handshake(b"dup");
+        assert!(checker.check_handshake(b"dup"));
+    }
     
-        stats.increment_user_curr_connects("user1");
-        stats.increment_user_curr_connects("user1");
-        assert_eq!(stats.get_user_curr_connects("user1"), 2);
+    #[test]
+    fn test_replay_checker_expiration() {
+        let checker = ReplayChecker::new(100, Duration::from_millis(50));
+        checker.add_handshake(b"expire");
+        assert!(checker.check_handshake(b"expire"));
+        std::thread::sleep(Duration::from_millis(100));
+        assert!(!checker.check_handshake(b"expire"));
+    }
     
-        stats.decrement_user_curr_connects("user1");
-        assert_eq!(stats.get_user_curr_connects("user1"), 1);
+    #[test]
+    fn test_replay_checker_stats() {
+        let checker = ReplayChecker::new(100, Duration::from_secs(60));
+        checker.add_handshake(b"k1");
+        checker.add_handshake(b"k2");
+        checker.check_handshake(b"k1");
+        checker.check_handshake(b"k3");
+        let stats = checker.stats();
+        assert_eq!(stats.total_additions, 2);
+        assert_eq!(stats.total_checks, 2);
+        assert_eq!(stats.total_hits, 1);
+    }
+    
+    #[test]
+    fn test_replay_checker_many_keys() {
+        let checker = ReplayChecker::new(1000, Duration::from_secs(60));
+        for i in 0..500u32 {
+            checker.add(&i.to_le_bytes());
+        }
+        for i in 0..500u32 {
+            assert!(checker.check(&i.to_le_bytes()));
+        }
+        assert_eq!(checker.stats().total_entries, 500);
     }
 }

src/stream/buffer_pool.rs

@@ -11,8 +11,9 @@ use std::sync::Arc;
 
 // ============= Configuration =============
 
-/// Default buffer size (64KB - good for MTProto)
-pub const DEFAULT_BUFFER_SIZE: usize = 64 * 1024;
+/// Default buffer size
+/// CHANGED: Reduced from 64KB to 16KB to match TLS record size and prevent bufferbloat.
+pub const DEFAULT_BUFFER_SIZE: usize = 16 * 1024;
 
 /// Default maximum number of pooled buffers
 pub const DEFAULT_MAX_BUFFERS: usize = 1024;

src/stream/frame.rs

@@ -5,8 +5,10 @@
 
 use bytes::{Bytes, BytesMut};
 use std::io::Result;
+use std::sync::Arc;
 
 use crate::protocol::constants::ProtoTag;
+use crate::crypto::SecureRandom;
 
 // ============= Frame Types =============
 
@@ -147,11 +149,11 @@ pub trait FrameCodec: Send + Sync {
 // ============= Codec Factory =============
 
 /// Create a frame codec for the given protocol tag
-pub fn create_codec(proto_tag: ProtoTag) -> Box<dyn FrameCodec> {
+pub fn create_codec(proto_tag: ProtoTag, rng: Arc<SecureRandom>) -> Box<dyn FrameCodec> {
     match proto_tag {
-        ProtoTag::Abridged => Box::new(super::frame_codec::AbridgedCodec::new()),
-        ProtoTag::Intermediate => Box::new(super::frame_codec::IntermediateCodec::new()),
-        ProtoTag::Secure => Box::new(super::frame_codec::SecureCodec::new()),
+        ProtoTag::Abridged => Box::new(crate::stream::frame_codec::AbridgedCodec::new()),
+        ProtoTag::Intermediate => Box::new(crate::stream::frame_codec::IntermediateCodec::new()),
+        ProtoTag::Secure => Box::new(crate::stream::frame_codec::SecureCodec::new(rng)),
     }
 }
 

src/stream/frame_codec.rs

@@ -5,9 +5,11 @@
 
 use bytes::{Bytes, BytesMut, BufMut};
 use std::io::{self, Error, ErrorKind};
+use std::sync::Arc;
 use tokio_util::codec::{Decoder, Encoder};
 
 use crate::protocol::constants::ProtoTag;
+use crate::crypto::SecureRandom;
 use super::frame::{Frame, FrameMeta, FrameCodec as FrameCodecTrait};
 
 // ============= Unified Codec =============
@@ -21,14 +23,17 @@ pub struct FrameCodec {
     proto_tag: ProtoTag,
     /// Maximum allowed frame size
     max_frame_size: usize,
+    /// RNG for secure padding
+    rng: Arc<SecureRandom>,
 }
 
 impl FrameCodec {
     /// Create a new codec for the given protocol
-    pub fn new(proto_tag: ProtoTag) -> Self {
+    pub fn new(proto_tag: ProtoTag, rng: Arc<SecureRandom>) -> Self {
         Self {
             proto_tag,
             max_frame_size: 16 * 1024 * 1024, // 16MB default
+            rng,
         }
     }
     
@@ -64,7 +69,7 @@ impl Encoder<Frame> for FrameCodec {
         match self.proto_tag {
             ProtoTag::Abridged => encode_abridged(&frame, dst),
             ProtoTag::Intermediate => encode_intermediate(&frame, dst),
-            ProtoTag::Secure => encode_secure(&frame, dst),
+            ProtoTag::Secure => encode_secure(&frame, dst, &self.rng),
         }
     }
 }
@@ -288,9 +293,7 @@ fn decode_secure(src: &mut BytesMut, max_size: usize) -> io::Result<Option<Frame
     Ok(Some(Frame::with_meta(data, meta)))
 }
 
-fn encode_secure(frame: &Frame, dst: &mut BytesMut) -> io::Result<()> {
-    use crate::crypto::random::SECURE_RANDOM;
-    
+fn encode_secure(frame: &Frame, dst: &mut BytesMut, rng: &SecureRandom) -> io::Result<()> {
     let data = &frame.data;
     
     // Simple ACK: just send data
@@ -303,10 +306,10 @@ fn encode_secure(frame: &Frame, dst: &mut BytesMut) -> io::Result<()> {
     // 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
-        (SECURE_RANDOM.range(3) + 1) as usize
+        (rng.range(3) + 1) as usize
     } else {
         // Already non-aligned, can add 0-3
-        SECURE_RANDOM.range(4) as usize
+        rng.range(4) as usize
     };
     
     let total_len = data.len() + padding_len;
@@ -321,7 +324,7 @@ fn encode_secure(frame: &Frame, dst: &mut BytesMut) -> io::Result<()> {
     dst.extend_from_slice(data);
     
     if padding_len > 0 {
-        let padding = SECURE_RANDOM.bytes(padding_len);
+        let padding = rng.bytes(padding_len);
         dst.extend_from_slice(&padding);
     }
     
@@ -445,19 +448,21 @@ impl FrameCodecTrait for IntermediateCodec {
 /// Secure Intermediate protocol codec
 pub struct SecureCodec {
     max_frame_size: usize,
+    rng: Arc<SecureRandom>,
 }
 
 impl SecureCodec {
-    pub fn new() -> Self {
+    pub fn new(rng: Arc<SecureRandom>) -> Self {
         Self {
             max_frame_size: 16 * 1024 * 1024,
+            rng,
         }
     }
 }
 
 impl Default for SecureCodec {
     fn default() -> Self {
-        Self::new()
+        Self::new(Arc::new(SecureRandom::new()))
     }
 }
 
@@ -474,7 +479,7 @@ impl Encoder<Frame> for SecureCodec {
     type Error = io::Error;
     
     fn encode(&mut self, frame: Frame, dst: &mut BytesMut) -> Result<(), Self::Error> {
-        encode_secure(&frame, dst)
+        encode_secure(&frame, dst, &self.rng)
     }
 }
 
@@ -485,7 +490,7 @@ impl FrameCodecTrait for SecureCodec {
     
     fn encode(&self, frame: &Frame, dst: &mut BytesMut) -> io::Result<usize> {
         let before = dst.len();
-        encode_secure(frame, dst)?;
+        encode_secure(frame, dst, &self.rng)?;
         Ok(dst.len() - before)
     }
     
@@ -506,6 +511,8 @@ mod tests {
     use tokio_util::codec::{FramedRead, FramedWrite};
     use tokio::io::duplex;
     use futures::{SinkExt, StreamExt};
+    use crate::crypto::SecureRandom;
+    use std::sync::Arc;
     
     #[tokio::test]
     async fn test_framed_abridged() {
@@ -541,8 +548,8 @@ mod tests {
     async fn test_framed_secure() {
         let (client, server) = duplex(4096);
         
-        let mut writer = FramedWrite::new(client, SecureCodec::new());
-        let mut reader = FramedRead::new(server, SecureCodec::new());
+        let mut writer = FramedWrite::new(client, SecureCodec::new(Arc::new(SecureRandom::new())));
+        let mut reader = FramedRead::new(server, SecureCodec::new(Arc::new(SecureRandom::new())));
         
         let original = Bytes::from_static(&[1, 2, 3, 4, 5, 6, 7, 8]);
         let frame = Frame::new(original.clone());
@@ -557,8 +564,8 @@ mod tests {
         for proto_tag in [ProtoTag::Abridged, ProtoTag::Intermediate, ProtoTag::Secure] {
             let (client, server) = duplex(4096);
             
-            let mut writer = FramedWrite::new(client, FrameCodec::new(proto_tag));
-            let mut reader = FramedRead::new(server, FrameCodec::new(proto_tag));
+            let mut writer = FramedWrite::new(client, FrameCodec::new(proto_tag, Arc::new(SecureRandom::new())));
+            let mut reader = FramedRead::new(server, FrameCodec::new(proto_tag, Arc::new(SecureRandom::new())));
             
             // Use 4-byte aligned data for abridged compatibility
             let original = Bytes::from_static(&[1, 2, 3, 4, 5, 6, 7, 8]);
@@ -607,7 +614,7 @@ mod tests {
     
     #[test]
     fn test_frame_too_large() {
-        let mut codec = FrameCodec::new(ProtoTag::Intermediate)
+        let mut codec = FrameCodec::new(ProtoTag::Intermediate, Arc::new(SecureRandom::new()))
             .with_max_frame_size(100);
         
         // Create a "frame" that claims to be very large

src/stream/frame_stream.rs

@@ -4,8 +4,8 @@ use bytes::{Bytes, BytesMut};
 use std::io::{Error, ErrorKind, Result};
 use tokio::io::{AsyncRead, AsyncWrite, AsyncReadExt, AsyncWriteExt};
 use crate::protocol::constants::*;
-use crate::crypto::crc32;
-use crate::crypto::random::SECURE_RANDOM;
+use crate::crypto::{crc32, SecureRandom};
+use std::sync::Arc;
 use super::traits::{FrameMeta, LayeredStream};
 
 // ============= Abridged (Compact) Frame =============
@@ -251,11 +251,12 @@ impl<R> LayeredStream<R> for SecureIntermediateFrameReader<R> {
 /// Writer for secure intermediate MTProto framing
 pub struct SecureIntermediateFrameWriter<W> {
     upstream: W,
+    rng: Arc<SecureRandom>,
 }
 
 impl<W> SecureIntermediateFrameWriter<W> {
-    pub fn new(upstream: W) -> Self {
-        Self { upstream }
+    pub fn new(upstream: W, rng: Arc<SecureRandom>) -> Self {
+        Self { upstream, rng }
     }
 }
 
@@ -267,8 +268,8 @@ impl<W: AsyncWrite + Unpin> SecureIntermediateFrameWriter<W> {
         }
         
         // Add random padding (0-3 bytes)
-        let padding_len = SECURE_RANDOM.range(4);
-        let padding = SECURE_RANDOM.bytes(padding_len);
+        let padding_len = self.rng.range(4);
+        let padding = self.rng.bytes(padding_len);
         
         let total_len = data.len() + padding_len;
         let len_bytes = (total_len as u32).to_le_bytes();
@@ -454,11 +455,11 @@ pub enum FrameWriterKind<W> {
 }
 
 impl<W: AsyncWrite + Unpin> FrameWriterKind<W> {
-    pub fn new(upstream: W, proto_tag: ProtoTag) -> Self {
+    pub fn new(upstream: W, proto_tag: ProtoTag, rng: Arc<SecureRandom>) -> Self {
         match proto_tag {
             ProtoTag::Abridged => FrameWriterKind::Abridged(AbridgedFrameWriter::new(upstream)),
             ProtoTag::Intermediate => FrameWriterKind::Intermediate(IntermediateFrameWriter::new(upstream)),
-            ProtoTag::Secure => FrameWriterKind::SecureIntermediate(SecureIntermediateFrameWriter::new(upstream)),
+            ProtoTag::Secure => FrameWriterKind::SecureIntermediate(SecureIntermediateFrameWriter::new(upstream, rng)),
         }
     }
     
@@ -483,6 +484,8 @@ impl<W: AsyncWrite + Unpin> FrameWriterKind<W> {
 mod tests {
     use super::*;
     use tokio::io::duplex;
+    use std::sync::Arc;
+    use crate::crypto::SecureRandom;
     
     #[tokio::test]
     async fn test_abridged_roundtrip() {
@@ -539,7 +542,7 @@ mod tests {
     async fn test_secure_intermediate_padding() {
         let (client, server) = duplex(1024);
         
-        let mut writer = SecureIntermediateFrameWriter::new(client);
+        let mut writer = SecureIntermediateFrameWriter::new(client, Arc::new(SecureRandom::new()));
         let mut reader = SecureIntermediateFrameReader::new(server);
         
         let data = vec![1u8, 2, 3, 4, 5, 6, 7, 8];
@@ -572,7 +575,7 @@ mod tests {
     async fn test_frame_reader_kind() {
         let (client, server) = duplex(1024);
         
-        let mut writer = FrameWriterKind::new(client, ProtoTag::Intermediate);
+        let mut writer = FrameWriterKind::new(client, ProtoTag::Intermediate, Arc::new(SecureRandom::new()));
         let mut reader = FrameReaderKind::new(server, ProtoTag::Intermediate);
         
         let data = vec![1u8, 2, 3, 4];

src/stream/tls_stream.rs

@@ -1,17 +1,36 @@
 //! Fake TLS 1.3 stream wrappers
 //!
-//! This module provides stateful async stream wrappers that handle
-//! TLS record framing with proper partial read/write handling.
+//! This module provides stateful async stream wrappers that handle TLS record
+//! framing with proper partial read/write handling.
 //!
-//! These are "fake" TLS streams - they wrap data in valid TLS 1.3
-//! Application Data records but don't perform actual TLS encryption.
-//! The actual encryption is handled by the crypto layer underneath.
+//! These are "fake" TLS streams:
+//! - We wrap raw bytes into syntactically valid TLS 1.3 records (Application Data).
+//! - We DO NOT perform real TLS handshake/encryption.
+//! - Real crypto for MTProto is handled by the crypto layer underneath.
+//!
+//! Why do we need this?
+//! Telegram MTProto proxy "FakeTLS" mode uses a TLS-looking outer layer for
+//! domain fronting / traffic camouflage. iOS Telegram clients are known to
+//! produce slightly different TLS record sizing patterns than Android/Desktop,
+//! including records that exceed 16384 payload bytes by a small overhead.
 //!
 //! Key design principles:
 //! - Explicit state machines for all async operations
 //! - Never lose data on partial reads
 //! - Atomic TLS record formation for writes
 //! - Proper handling of all TLS record types
+//!
+//! Important nuance (Telegram FakeTLS):
+//! - The TLS spec limits "plaintext fragments" to 2^14 (16384) bytes.
+//! - 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.
+//!
+//! If you reject those (e.g. validate length <= 16384), you will see errors like:
+//!   "TLS record too large: 16408 bytes"
+//! and uploads from iOS will break (media/file sending), while small traffic
+//! may still work.
 
 use bytes::{Bytes, BytesMut, BufMut};
 use std::io::{self, Error, ErrorKind, Result};
@@ -20,25 +39,29 @@ use std::task::{Context, Poll};
 use tokio::io::{AsyncRead, AsyncWrite, AsyncReadExt, AsyncWriteExt, ReadBuf};
 
 use crate::protocol::constants::{
-    TLS_VERSION, TLS_RECORD_APPLICATION, TLS_RECORD_CHANGE_CIPHER,
-    TLS_RECORD_HANDSHAKE, TLS_RECORD_ALERT, MAX_TLS_RECORD_SIZE,
+    TLS_VERSION,
+    TLS_RECORD_APPLICATION, TLS_RECORD_CHANGE_CIPHER,
+    TLS_RECORD_HANDSHAKE, TLS_RECORD_ALERT,
+    MAX_TLS_CHUNK_SIZE,
 };
 use super::state::{StreamState, HeaderBuffer, YieldBuffer, WriteBuffer};
 
 // ============= Constants =============
 
-/// TLS record header size
+/// TLS record header size (type + version + length)
 const TLS_HEADER_SIZE: usize = 5;
 
-/// Maximum TLS record payload size (16KB as per TLS spec)
+/// Maximum TLS fragment size per spec (plaintext fragment).
+/// We use this for *outgoing* chunking, because we build plain ApplicationData records.
 const MAX_TLS_PAYLOAD: usize = 16384;
 
-/// Maximum pending write buffer
+/// Maximum pending write buffer for one record remainder.
+/// Note: we never queue unlimited amount of data here; state holds at most one record.
 const MAX_PENDING_WRITE: usize = 64 * 1024;
 
 // ============= TLS Record Types =============
 
-/// Parsed TLS record header
+/// Parsed TLS record header (5 bytes)
 #[derive(Debug, Clone, Copy)]
 struct TlsRecordHeader {
     /// Record type (0x17 = Application Data, 0x14 = Change Cipher, etc.)
@@ -50,22 +73,27 @@ struct TlsRecordHeader {
 }
 
 impl TlsRecordHeader {
-    /// Parse header from 5 bytes
+    /// Parse header from exactly 5 bytes.
+    ///
+    /// This currently never returns None, but is kept as Option to allow future
+    /// stricter parsing rules without changing callers.
     fn parse(header: &[u8; 5]) -> Option<Self> {
         let record_type = header[0];
         let version = [header[1], header[2]];
         let length = u16::from_be_bytes([header[3], header[4]]);
-        
-        Some(Self {
-            record_type,
-            version,
-            length,
-        })
+        Some(Self { record_type, version, length })
     }
 
-    /// Validate the header
+    /// Validate the header.
+    ///
+    /// Nuances:
+    /// - 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).
+    /// - For other record types we keep stricter bounds to avoid memory abuse.
     fn validate(&self) -> Result<()> {
-        // Check version (accept TLS 1.0 for ClientHello, TLS 1.2/1.3 for others)
+        // Version: accept TLS 1.0 header (ClientHello quirk) and TLS_VERSION (0x0303).
         if self.version != [0x03, 0x01] && self.version != TLS_VERSION {
             return Err(Error::new(
                 ErrorKind::InvalidData,
@@ -73,27 +101,36 @@ impl TlsRecordHeader {
             ));
         }
 
-        // Check length
-        if self.length as usize > MAX_TLS_RECORD_SIZE {
+        let len = self.length as usize;
+
+        // Length checks depend on record type.
+        // Telegram FakeTLS: ApplicationData length may be 16384 + 24.
+        match self.record_type {
+            TLS_RECORD_APPLICATION => {
+                if len > MAX_TLS_CHUNK_SIZE {
                     return Err(Error::new(
                         ErrorKind::InvalidData,
-                format!("TLS record too large: {} bytes", self.length),
+                        format!("TLS record too large: {} bytes (max {})", len, MAX_TLS_CHUNK_SIZE),
                     ));
                 }
+            }
+
+            // ChangeCipherSpec/Alert/Handshake should never be that large for our usage
+            // (post-handshake we don't expect Handshake at all).
+            // Keep strict to reduce attack surface.
+            _ => {
+                if len > MAX_TLS_PAYLOAD {
+                    return Err(Error::new(
+                        ErrorKind::InvalidData,
+                        format!("TLS control record too large: {} bytes (max {})", len, MAX_TLS_PAYLOAD),
+                    ));
+                }
+            }
+        }
 
         Ok(())
     }
 
-    /// Check if this is an application data record
-    fn is_application_data(&self) -> bool {
-        self.record_type == TLS_RECORD_APPLICATION
-    }
-    
-    /// Check if this is a change cipher spec record (should be skipped)
-    fn is_change_cipher_spec(&self) -> bool {
-        self.record_type == TLS_RECORD_CHANGE_CIPHER
-    }
-    
     /// Build header bytes
     fn to_bytes(&self) -> [u8; 5] {
         [
@@ -120,25 +157,20 @@ enum TlsReaderState {
         header: HeaderBuffer<TLS_HEADER_SIZE>,
     },
 
-    /// Reading the TLS record body
+    /// Reading the TLS record body (payload)
     ReadingBody {
-        /// Parsed record type
         record_type: u8,
-        /// Total body length
         length: usize,
-        /// Buffer for body data
         buffer: BytesMut,
     },
 
-    /// Have decrypted data ready to yield to caller
+    /// Have buffered data ready to yield to caller
     Yielding {
-        /// Buffer containing data to yield
         buffer: YieldBuffer,
     },
 
     /// Stream encountered an error and cannot be used
     Poisoned {
-        /// The error that caused poisoning
         error: Option<io::Error>,
     },
 }
@@ -165,12 +197,13 @@ impl StreamState for TlsReaderState {
 
 // ============= FakeTlsReader =============
 
-/// Reader that unwraps TLS 1.3 records with proper state machine
+/// Reader that unwraps TLS records (FakeTLS).
 ///
-/// This reader handles partial reads correctly by maintaining internal state
-/// and never losing any data that has been read from upstream.
+/// This wrapper is responsible ONLY for TLS record framing and skipping
+/// non-data records (like CCS). It does not decrypt TLS: payload bytes are passed
+/// as-is to upper layers (crypto stream).
 ///
-/// # State Machine
+/// State machine overview:
 ///
 /// ┌──────────┐                    ┌───────────────┐
 /// │   Idle   │ -----------------> │ ReadingHeader │
@@ -178,103 +211,69 @@ impl StreamState for TlsReaderState {
 ///      ▲                                  │
 ///      │                           header complete
 ///      │                                  │
-///      │                                  │
+///      │                                  ▼
 ///      │                          ┌───────────────┐
 ///      │        skip record       │  ReadingBody  │
 ///      │ <-------- (CCS) -------- │               │
 ///      │                          └───────┬───────┘
 ///      │                                  │
 ///      │                              body complete
-///      │      drained                     │
-///      │ <-----------------┐              │
-///      │                   │      ┌───────────────┐
-///      │                   └----- │   Yielding    │
+///      │                                  ▼
+///      │                          ┌───────────────┐
+///      │                          │   Yielding    │
 ///      │                          └───────────────┘
 ///      │
 ///      │    errors / w any state
-///      │
+///      ▼
 /// ┌───────────────────────────────────────────────┐
 /// │                    Poisoned                   │
 /// └───────────────────────────────────────────────┘
 ///
+/// NOTE: We must correctly handle partial reads from upstream:
+/// - do not assume header arrives in one poll
+/// - do not assume body arrives in one poll
+/// - never lose already-read bytes
 pub struct FakeTlsReader<R> {
-    /// Upstream reader
     upstream: R,
-    /// Current state
     state: TlsReaderState,
 }
 
 impl<R> FakeTlsReader<R> {
-    /// Create new fake TLS reader
     pub fn new(upstream: R) -> Self {
-        Self {
-            upstream,
-            state: TlsReaderState::Idle,
-        }
+        Self { upstream, state: TlsReaderState::Idle }
     }
 
-    /// Get reference to upstream
-    pub fn get_ref(&self) -> &R {
-        &self.upstream
-    }
+    pub fn get_ref(&self) -> &R { &self.upstream }
+    pub fn get_mut(&mut self) -> &mut R { &mut self.upstream }
+    pub fn into_inner(self) -> R { self.upstream }
 
-    /// Get mutable reference to upstream
-    pub fn get_mut(&mut self) -> &mut R {
-        &mut self.upstream
-    }
+    pub fn is_poisoned(&self) -> bool { self.state.is_poisoned() }
+    pub fn state_name(&self) -> &'static str { self.state.state_name() }
 
-    /// Consume and return upstream
-    pub fn into_inner(self) -> R {
-        self.upstream
-    }
-    
-    /// Check if stream is in poisoned state
-    pub fn is_poisoned(&self) -> bool {
-        self.state.is_poisoned()
-    }
-    
-    /// Get current state name (for debugging)
-    pub fn state_name(&self) -> &'static str {
-        self.state.state_name()
-    }
-    
-    /// Transition to poisoned state
     fn poison(&mut self, error: io::Error) {
         self.state = TlsReaderState::Poisoned { error: Some(error) };
     }
 
-    /// Take error from poisoned state
     fn take_poison_error(&mut self) -> io::Error {
         match &mut self.state {
-            TlsReaderState::Poisoned { error } => {
-                error.take().unwrap_or_else(|| {
+            TlsReaderState::Poisoned { error } => error.take().unwrap_or_else(|| {
                 io::Error::new(ErrorKind::Other, "stream previously poisoned")
-                })
-            }
+            }),
             _ => io::Error::new(ErrorKind::Other, "stream not poisoned"),
         }
     }
 }
 
-/// Result of polling for header completion
 enum HeaderPollResult {
-    /// Need more data
     Pending,
-    /// EOF at record boundary (clean close)
     Eof,
-    /// Header complete, parsed successfully
     Complete(TlsRecordHeader),
-    /// Error occurred
     Error(io::Error),
 }
 
-/// Result of polling for body completion
 enum BodyPollResult {
-    /// Need more data
     Pending,
-    /// Body complete
     Complete(Bytes),
-    /// Error occurred
     Error(io::Error),
 }
 
@@ -291,7 +290,7 @@ impl<R: AsyncRead + Unpin> AsyncRead for FakeTlsReader<R> {
             let state = std::mem::replace(&mut this.state, TlsReaderState::Idle);
 
             match state {
-                // Poisoned state - return error
+                // Poisoned state: always return the stored error
                 TlsReaderState::Poisoned { error } => {
                     this.state = TlsReaderState::Poisoned { error: None };
                     let err = error.unwrap_or_else(|| {
@@ -300,20 +299,18 @@ impl<R: AsyncRead + Unpin> AsyncRead for FakeTlsReader<R> {
                     return Poll::Ready(Err(err));
                 }
 
-                // Have buffered data to yield
+                // Yield buffered plaintext to caller
                 TlsReaderState::Yielding { mut buffer } => {
                     if buf.remaining() == 0 {
                         this.state = TlsReaderState::Yielding { buffer };
                         return Poll::Ready(Ok(()));
                     }
 
-                    // Copy as much as possible to output
                     let to_copy = buffer.remaining().min(buf.remaining());
                     let dst = buf.initialize_unfilled_to(to_copy);
                     let copied = buffer.copy_to(dst);
                     buf.advance(copied);
 
-                    // If buffer is drained, transition to Idle
                     if buffer.is_empty() {
                         this.state = TlsReaderState::Idle;
                     } else {
@@ -323,23 +320,21 @@ impl<R: AsyncRead + Unpin> AsyncRead for FakeTlsReader<R> {
                     return Poll::Ready(Ok(()));
                 }
 
-                // Ready to read a new TLS record
+                // Start reading new record
                 TlsReaderState::Idle => {
                     if buf.remaining() == 0 {
                         this.state = TlsReaderState::Idle;
                         return Poll::Ready(Ok(()));
                     }
 
-                    // Start reading header
                     this.state = TlsReaderState::ReadingHeader {
                         header: HeaderBuffer::new(),
                     };
-                    // Continue to ReadingHeader
+                    // loop continues and will handle ReadingHeader
                 }
 
-                // Reading TLS record header
+                // Read TLS header (5 bytes)
                 TlsReaderState::ReadingHeader { mut header } => {
-                    // Poll to fill header
                     let result = poll_read_header(&mut this.upstream, cx, &mut header);
 
                     match result {
@@ -348,6 +343,7 @@ impl<R: AsyncRead + Unpin> AsyncRead for FakeTlsReader<R> {
                             return Poll::Pending;
                         }
                         HeaderPollResult::Eof => {
+                            // Clean EOF at record boundary
                             this.state = TlsReaderState::Idle;
                             return Poll::Ready(Ok(()));
                         }
@@ -356,15 +352,12 @@ impl<R: AsyncRead + Unpin> AsyncRead for FakeTlsReader<R> {
                             return Poll::Ready(Err(e));
                         }
                         HeaderPollResult::Complete(parsed) => {
-                            // Validate header
                             if let Err(e) = parsed.validate() {
                                 this.poison(Error::new(e.kind(), e.to_string()));
                                 return Poll::Ready(Err(e));
                             }
 
                             let length = parsed.length as usize;
-                            
-                            // Transition to reading body
                             this.state = TlsReaderState::ReadingBody {
                                 record_type: parsed.record_type,
                                 length,
@@ -374,7 +367,7 @@ impl<R: AsyncRead + Unpin> AsyncRead for FakeTlsReader<R> {
                     }
                 }
 
-                // Reading TLS record body
+                // Read TLS payload
                 TlsReaderState::ReadingBody { record_type, length, mut buffer } => {
                     let result = poll_read_body(&mut this.upstream, cx, &mut buffer, length);
 
@@ -388,15 +381,15 @@ impl<R: AsyncRead + Unpin> AsyncRead for FakeTlsReader<R> {
                             return Poll::Ready(Err(e));
                         }
                         BodyPollResult::Complete(data) => {
-                            // Handle different record types
                             match record_type {
                                 TLS_RECORD_CHANGE_CIPHER => {
-                                    // Skip Change Cipher Spec, read next record
+                                    // CCS is expected in some clients, ignore it.
                                     this.state = TlsReaderState::Idle;
                                     continue;
                                 }
+
                                 TLS_RECORD_APPLICATION => {
-                                    // Application data - yield to caller
+                                    // This is what we actually want.
                                     if data.is_empty() {
                                         this.state = TlsReaderState::Idle;
                                         continue;
@@ -405,25 +398,26 @@ impl<R: AsyncRead + Unpin> AsyncRead for FakeTlsReader<R> {
                                     this.state = TlsReaderState::Yielding {
                                         buffer: YieldBuffer::new(data),
                                     };
-                                    // Continue to yield
+                                    // loop continues and will yield immediately
                                 }
+
                                 TLS_RECORD_ALERT => {
-                                    // TLS Alert - treat as EOF
+                                    // Treat TLS alert as EOF-like termination.
                                     this.state = TlsReaderState::Idle;
                                     return Poll::Ready(Ok(()));
                                 }
+
                                 TLS_RECORD_HANDSHAKE => {
-                                    let err = Error::new(
-                                        ErrorKind::InvalidData,
-                                        "unexpected TLS handshake record"
-                                    );
+                                    // After FakeTLS handshake is done, we do not expect any Handshake records.
+                                    let err = Error::new(ErrorKind::InvalidData, "unexpected TLS handshake record");
                                     this.poison(Error::new(err.kind(), err.to_string()));
                                     return Poll::Ready(Err(err));
                                 }
+
                                 _ => {
                                     let err = Error::new(
                                         ErrorKind::InvalidData,
-                                        format!("unknown TLS record type: 0x{:02x}", record_type)
+                                        format!("unknown TLS record type: 0x{:02x}", record_type),
                                     );
                                     this.poison(Error::new(err.kind(), err.to_string()));
                                     return Poll::Ready(Err(err));
@@ -459,8 +453,10 @@ fn poll_read_header<R: AsyncRead + Unpin>(
                     } else {
                         return HeaderPollResult::Error(Error::new(
                             ErrorKind::UnexpectedEof,
-                            format!("unexpected EOF in TLS header (got {} of 5 bytes)", 
-                                header.as_slice().len())
+                            format!(
+                                "unexpected EOF in TLS header (got {} of 5 bytes)",
+                                header.as_slice().len()
+                            ),
                         ));
                     }
                 }
@@ -469,14 +465,10 @@ fn poll_read_header<R: AsyncRead + Unpin>(
         }
     }
 
-    // Parse header
     let header_bytes = *header.as_array();
     match TlsRecordHeader::parse(&header_bytes) {
         Some(h) => HeaderPollResult::Complete(h),
-        None => HeaderPollResult::Error(Error::new(
-            ErrorKind::InvalidData,
-            "failed to parse TLS header"
-        )),
+        None => HeaderPollResult::Error(Error::new(ErrorKind::InvalidData, "failed to parse TLS header")),
     }
 }
 
@@ -487,10 +479,12 @@ fn poll_read_body<R: AsyncRead + Unpin>(
     buffer: &mut BytesMut,
     target_len: usize,
 ) -> BodyPollResult {
+    // NOTE: This implementation uses a temporary Vec to avoid tricky borrow/lifetime
+    // issues with BytesMut spare capacity and ReadBuf across polls.
+    // It's safe and correct; optimization is possible if needed.
     while buffer.len() < target_len {
         let remaining = target_len - buffer.len();
 
-        // Read into a temporary buffer
         let mut temp = vec![0u8; remaining.min(8192)];
         let mut read_buf = ReadBuf::new(&mut temp);
 
@@ -502,8 +496,11 @@ fn poll_read_body<R: AsyncRead + Unpin>(
                 if n == 0 {
                     return BodyPollResult::Error(Error::new(
                         ErrorKind::UnexpectedEof,
-                        format!("unexpected EOF in TLS body (got {} of {} bytes)",
-                            buffer.len(), target_len)
+                        format!(
+                            "unexpected EOF in TLS body (got {} of {} bytes)",
+                            buffer.len(),
+                            target_len
+                        ),
                     ));
                 }
                 buffer.extend_from_slice(&temp[..n]);
@@ -515,10 +512,9 @@ fn poll_read_body<R: AsyncRead + Unpin>(
 }
 
 impl<R: AsyncRead + Unpin> FakeTlsReader<R> {
-    /// Read exactly n bytes through TLS layer
+    /// Read exactly n bytes through TLS layer.
     ///
-    /// This is a convenience method that accumulates data across
-    /// multiple TLS records until exactly n bytes are available.
+    /// This accumulates data across multiple TLS ApplicationData records.
     pub async fn read_exact(&mut self, n: usize) -> Result<Bytes> {
         if self.is_poisoned() {
             return Err(self.take_poison_error());
@@ -533,7 +529,7 @@ impl<R: AsyncRead + Unpin> FakeTlsReader<R> {
             if read == 0 {
                 return Err(Error::new(
                     ErrorKind::UnexpectedEof,
-                    format!("expected {} bytes, got {}", n, result.len())
+                    format!("expected {} bytes, got {}", n, result.len()),
                 ));
             }
 
@@ -546,23 +542,19 @@ impl<R: AsyncRead + Unpin> FakeTlsReader<R> {
 
 // ============= FakeTlsWriter State =============
 
-/// State machine states for FakeTlsWriter
 #[derive(Debug)]
 enum TlsWriterState {
     /// Ready to accept new data
     Idle,
 
-    /// Writing a complete TLS record
+    /// Writing a complete TLS record (header + body), possibly partially
     WritingRecord {
-        /// Complete record (header + body) to write
         record: WriteBuffer,
-        /// Original payload size (for return value calculation)
         payload_size: usize,
     },
 
     /// Stream encountered an error and cannot be used
     Poisoned {
-        /// The error that caused poisoning
         error: Option<io::Error>,
     },
 }
@@ -587,94 +579,46 @@ impl StreamState for TlsWriterState {
 
 // ============= FakeTlsWriter =============
 
-/// Writer that wraps data in TLS 1.3 records with proper state machine
+/// Writer that wraps bytes into TLS 1.3 Application Data records.
 ///
-/// This writer handles partial writes correctly by:
-/// - Building complete TLS records before writing
-/// - Maintaining internal state for partial record writes
-/// - Never splitting a record mid-write to upstream
-///
-/// # State Machine
-///
-/// ┌──────────┐     write      ┌─────────────────┐
-/// │   Idle   │ -------------> │  WritingRecord  │
-/// │          │ <------------- │                 │
-/// └──────────┘    complete    └─────────────────┘
-///      │                              │
-///      │          < errors >          │ 
-///      │                              │
-/// ┌─────────────────────────────────────────────┐
-/// │                Poisoned                     │
-/// └─────────────────────────────────────────────┘
-///
-/// # Record Formation
-///
-/// Data is chunked into records of at most MAX_TLS_PAYLOAD bytes.
-/// Each record has a 5-byte header prepended.
+/// We chunk outgoing data into records of <= 16384 payload bytes (MAX_TLS_PAYLOAD).
+/// We do not try to mimic AEAD overhead on the wire; Telegram clients accept it.
+/// If you want to be more camouflage-accurate later, you could add optional padding
+/// to produce records sized closer to MAX_TLS_CHUNK_SIZE.
 pub struct FakeTlsWriter<W> {
-    /// Upstream writer
     upstream: W,
-    /// Current state
     state: TlsWriterState,
 }
 
 impl<W> FakeTlsWriter<W> {
-    /// Create new fake TLS writer
     pub fn new(upstream: W) -> Self {
-        Self {
-            upstream,
-            state: TlsWriterState::Idle,
-        }
+        Self { upstream, state: TlsWriterState::Idle }
     }
 
-    /// Get reference to upstream
-    pub fn get_ref(&self) -> &W {
-        &self.upstream
-    }
+    pub fn get_ref(&self) -> &W { &self.upstream }
+    pub fn get_mut(&mut self) -> &mut W { &mut self.upstream }
+    pub fn into_inner(self) -> W { self.upstream }
 
-    /// Get mutable reference to upstream
-    pub fn get_mut(&mut self) -> &mut W {
-        &mut self.upstream
-    }
+    pub fn is_poisoned(&self) -> bool { self.state.is_poisoned() }
+    pub fn state_name(&self) -> &'static str { self.state.state_name() }
 
-    /// Consume and return upstream
-    pub fn into_inner(self) -> W {
-        self.upstream
-    }
-    
-    /// Check if stream is in poisoned state
-    pub fn is_poisoned(&self) -> bool {
-        self.state.is_poisoned()
-    }
-    
-    /// Get current state name (for debugging)
-    pub fn state_name(&self) -> &'static str {
-        self.state.state_name()
-    }
-    
-    /// Check if there's a pending record to write
     pub fn has_pending(&self) -> bool {
         matches!(&self.state, TlsWriterState::WritingRecord { record, .. } if !record.is_empty())
     }
 
-    /// Transition to poisoned state
     fn poison(&mut self, error: io::Error) {
         self.state = TlsWriterState::Poisoned { error: Some(error) };
     }
 
-    /// Take error from poisoned state
     fn take_poison_error(&mut self) -> io::Error {
         match &mut self.state {
-            TlsWriterState::Poisoned { error } => {
-                error.take().unwrap_or_else(|| {
+            TlsWriterState::Poisoned { error } => error.take().unwrap_or_else(|| {
                 io::Error::new(ErrorKind::Other, "stream previously poisoned")
-                })
-            }
+            }),
             _ => io::Error::new(ErrorKind::Other, "stream not poisoned"),
         }
     }
 
-    /// Build a TLS Application Data record
     fn build_record(data: &[u8]) -> BytesMut {
         let header = TlsRecordHeader {
             record_type: TLS_RECORD_APPLICATION,
@@ -689,18 +633,13 @@ impl<W> FakeTlsWriter<W> {
     }
 }
 
-/// Result of flushing pending record
 enum FlushResult {
-    /// All data flushed, returns payload size
     Complete(usize),
-    /// Need to wait for upstream
     Pending,
-    /// Error occurred
     Error(io::Error),
 }
 
 impl<W: AsyncWrite + Unpin> FakeTlsWriter<W> {
-    /// Try to flush pending record to upstream (standalone logic)
     fn poll_flush_record_inner(
         upstream: &mut W,
         cx: &mut Context<'_>,
@@ -710,19 +649,14 @@ impl<W: AsyncWrite + Unpin> FakeTlsWriter<W> {
             let data = record.pending();
             match Pin::new(&mut *upstream).poll_write(cx, data) {
                 Poll::Pending => return FlushResult::Pending,
-                
                 Poll::Ready(Err(e)) => return FlushResult::Error(e),
-                
                 Poll::Ready(Ok(0)) => {
                     return FlushResult::Error(Error::new(
                         ErrorKind::WriteZero,
-                        "upstream returned 0 bytes written"
+                        "upstream returned 0 bytes written",
                     ));
                 }
-                
-                Poll::Ready(Ok(n)) => {
-                    record.advance(n);
-                }
+                Poll::Ready(Ok(n)) => record.advance(n),
             }
         }
 
@@ -738,7 +672,7 @@ impl<W: AsyncWrite + Unpin> AsyncWrite for FakeTlsWriter<W> {
     ) -> Poll<Result<usize>> {
         let this = self.get_mut();
 
-        // Take ownership of state
+        // Take ownership of state to avoid borrow conflicts.
         let state = std::mem::replace(&mut this.state, TlsWriterState::Idle);
 
         match state {
@@ -751,7 +685,7 @@ impl<W: AsyncWrite + Unpin> AsyncWrite for FakeTlsWriter<W> {
             }
 
             TlsWriterState::WritingRecord { mut record, payload_size } => {
-                // Continue flushing existing record
+                // Finish writing previous record before accepting new bytes.
                 match Self::poll_flush_record_inner(&mut this.upstream, cx, &mut record) {
                     FlushResult::Pending => {
                         this.state = TlsWriterState::WritingRecord { record, payload_size };
@@ -763,7 +697,7 @@ impl<W: AsyncWrite + Unpin> AsyncWrite for FakeTlsWriter<W> {
                     }
                     FlushResult::Complete(_) => {
                         this.state = TlsWriterState::Idle;
-                        // Fall through to handle new write
+                        // continue to accept new buf below
                     }
                 }
             }
@@ -782,19 +716,18 @@ impl<W: AsyncWrite + Unpin> AsyncWrite for FakeTlsWriter<W> {
         let chunk_size = buf.len().min(MAX_TLS_PAYLOAD);
         let chunk = &buf[..chunk_size];
 
-        // Build the complete record
+        // Build the complete record (header + payload)
         let record_data = Self::build_record(chunk);
 
-        // Try to write directly first
         match Pin::new(&mut this.upstream).poll_write(cx, &record_data) {
             Poll::Ready(Ok(n)) if n == record_data.len() => {
-                // Complete record written
                 Poll::Ready(Ok(chunk_size))
             }
 
             Poll::Ready(Ok(n)) => {
-                // Partial write - buffer the rest
+                // Partial write of the record: store remainder.
                 let mut write_buffer = WriteBuffer::with_max_size(MAX_PENDING_WRITE);
+                // record_data length is <= 16389, fits MAX_PENDING_WRITE
                 let _ = write_buffer.extend(&record_data[n..]);
 
                 this.state = TlsWriterState::WritingRecord {
@@ -802,7 +735,7 @@ impl<W: AsyncWrite + Unpin> AsyncWrite for FakeTlsWriter<W> {
                     payload_size: chunk_size,
                 };
 
-                // We've accepted chunk_size bytes from caller
+                // We have accepted chunk_size bytes from caller.
                 Poll::Ready(Ok(chunk_size))
             }
 
@@ -812,7 +745,7 @@ impl<W: AsyncWrite + Unpin> AsyncWrite for FakeTlsWriter<W> {
             }
 
             Poll::Pending => {
-                // Buffer the entire record
+                // Buffer entire record and report success for this chunk.
                 let mut write_buffer = WriteBuffer::with_max_size(MAX_PENDING_WRITE);
                 let _ = write_buffer.extend(&record_data);
 
@@ -821,10 +754,9 @@ impl<W: AsyncWrite + Unpin> AsyncWrite for FakeTlsWriter<W> {
                     payload_size: chunk_size,
                 };
 
-                // Wake to try again
+                // Wake to retry flushing soon.
                 cx.waker().wake_by_ref();
 
-                // We've accepted chunk_size bytes from caller
                 Poll::Ready(Ok(chunk_size))
             }
         }
@@ -833,7 +765,6 @@ impl<W: AsyncWrite + Unpin> AsyncWrite for FakeTlsWriter<W> {
     fn poll_flush(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Result<()>> {
         let this = self.get_mut();
 
-        // Take ownership of state
         let state = std::mem::replace(&mut this.state, TlsWriterState::Idle);
 
         match state {
@@ -866,48 +797,33 @@ impl<W: AsyncWrite + Unpin> AsyncWrite for FakeTlsWriter<W> {
             }
         }
 
-        // Flush upstream
         Pin::new(&mut this.upstream).poll_flush(cx)
     }
 
     fn poll_shutdown(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Result<()>> {
         let this = self.get_mut();
 
-        // Take ownership of state
         let state = std::mem::replace(&mut this.state, TlsWriterState::Idle);
 
         match state {
-            TlsWriterState::WritingRecord { mut record, payload_size } => {
-                // Try to flush pending (best effort)
-                match Self::poll_flush_record_inner(&mut this.upstream, cx, &mut record) {
-                    FlushResult::Pending => {
-                        // Can't complete flush, continue with shutdown anyway
+            TlsWriterState::WritingRecord { mut record, payload_size: _ } => {
+                // Best-effort flush (do not block shutdown forever).
+                let _ = Self::poll_flush_record_inner(&mut this.upstream, cx, &mut record);
                 this.state = TlsWriterState::Idle;
             }
-                    FlushResult::Error(_) => {
-                        // Ignore errors during shutdown
-                        this.state = TlsWriterState::Idle;
-                    }
-                    FlushResult::Complete(_) => {
-                        this.state = TlsWriterState::Idle;
-                    }
-                }
-            }
             _ => {
                 this.state = TlsWriterState::Idle;
             }
         }
 
-        // Shutdown upstream
         Pin::new(&mut this.upstream).poll_shutdown(cx)
     }
 }
 
 impl<W: AsyncWrite + Unpin> FakeTlsWriter<W> {
-    /// Write all data wrapped in TLS records (async method)
+    /// Write all data wrapped in TLS records.
     ///
-    /// This convenience method handles chunking large data into
-    /// multiple TLS records automatically.
+    /// Convenience method that chunks into <= 16384 records.
     pub async fn write_all_tls(&mut self, data: &[u8]) -> Result<()> {
         let mut written = 0;
         while written < data.len() {

src/transport/mod.rs

@@ -3,7 +3,11 @@
 pub mod pool;
 pub mod proxy_protocol;
 pub mod socket;
+pub mod socks;
+pub mod upstream;
 
 pub use pool::ConnectionPool;
 pub use proxy_protocol::{ProxyProtocolInfo, parse_proxy_protocol};
 pub use socket::*;
+pub use socks::*;
+pub use upstream::{UpstreamManager, StartupPingResult, DcPingResult};

src/transport/socket.rs

@@ -1,7 +1,7 @@
 //! TCP Socket Configuration
 
 use std::io::Result;
-use std::net::SocketAddr;
+use std::net::{SocketAddr, IpAddr};
 use std::time::Duration;
 use tokio::net::TcpStream;
 use socket2::{Socket, TcpKeepalive, Domain, Type, Protocol};
@@ -30,20 +30,13 @@ pub fn configure_tcp_socket(
         socket.set_tcp_keepalive(&keepalive)?;
     }
     
-    // Set buffer sizes
-    set_buffer_sizes(&socket, 65536, 65536)?;
+    // CHANGED: Removed manual buffer size setting (was 256KB).
+    // Allowing the OS kernel to handle TCP window scaling (Autotuning) is critical
+    // for mobile clients to avoid bufferbloat and stalled connections during uploads.
     
     Ok(())
 }
 
-/// Set socket buffer sizes
-fn set_buffer_sizes(socket: &socket2::SockRef, recv: usize, send: usize) -> Result<()> {
-    // These may fail on some systems, so we ignore errors
-    let _ = socket.set_recv_buffer_size(recv);
-    let _ = socket.set_send_buffer_size(send);
-    Ok(())
-}
-
 /// Configure socket for accepting client connections
 pub fn configure_client_socket(
     stream: &TcpStream,
@@ -65,6 +58,8 @@ pub fn configure_client_socket(
     socket.set_tcp_keepalive(&keepalive)?;
     
     // Set TCP user timeout (Linux only)
+    // NOTE: iOS does not support TCP_USER_TIMEOUT - application-level timeout 
+    // is implemented in relay_bidirectional instead
     #[cfg(target_os = "linux")]
     {
         use std::os::unix::io::AsRawFd;
@@ -93,6 +88,11 @@ pub fn set_linger_zero(stream: &TcpStream) -> Result<()> {
 
 /// Create a new TCP socket for outgoing connections
 pub fn create_outgoing_socket(addr: SocketAddr) -> Result<Socket> {
+    create_outgoing_socket_bound(addr, None)
+}
+
+/// Create a new TCP socket for outgoing connections, optionally bound to a specific interface
+pub fn create_outgoing_socket_bound(addr: SocketAddr, bind_addr: Option<IpAddr>) -> Result<Socket> {
     let domain = if addr.is_ipv4() {
         Domain::IPV4
     } else {
@@ -107,9 +107,16 @@ pub fn create_outgoing_socket(addr: SocketAddr) -> Result<Socket> {
     // Disable Nagle
     socket.set_nodelay(true)?;
 
+    if let Some(bind_ip) = bind_addr {
+        let bind_sock_addr = SocketAddr::new(bind_ip, 0);
+        socket.bind(&bind_sock_addr.into())?;
+        debug!("Bound outgoing socket to {}", bind_ip);
+    }
+    
     Ok(socket)
 }
 
+
 /// Get local address of a socket
 pub fn get_local_addr(stream: &TcpStream) -> Option<SocketAddr> {
     stream.local_addr().ok()

src/transport/socks.rs

@@ -0,0 +1,145 @@
+//! SOCKS4/5 Client Implementation
+
+use std::net::{IpAddr, SocketAddr};
+use tokio::io::{AsyncRead, AsyncReadExt, AsyncWrite, AsyncWriteExt};
+use tokio::net::TcpStream;
+use crate::error::{ProxyError, Result};
+
+pub async fn connect_socks4(
+    stream: &mut TcpStream,
+    target: SocketAddr,
+    user_id: Option<&str>,
+) -> Result<()> {
+    let ip = match target.ip() {
+        IpAddr::V4(ip) => ip,
+        IpAddr::V6(_) => return Err(ProxyError::Proxy("SOCKS4 does not support IPv6".to_string())),
+    };
+    
+    let port = target.port();
+    let user = user_id.unwrap_or("").as_bytes();
+    
+    // VN (4) | CD (1) | DSTPORT (2) | DSTIP (4) | USERID (variable) | NULL (1)
+    let mut buf = Vec::with_capacity(9 + user.len());
+    buf.push(4); // VN
+    buf.push(1); // CD (CONNECT)
+    buf.extend_from_slice(&port.to_be_bytes());
+    buf.extend_from_slice(&ip.octets());
+    buf.extend_from_slice(user);
+    buf.push(0); // NULL
+    
+    stream.write_all(&buf).await.map_err(|e| ProxyError::Io(e))?;
+    
+    // Response: VN (1) | CD (1) | DSTPORT (2) | DSTIP (4)
+    let mut resp = [0u8; 8];
+    stream.read_exact(&mut resp).await.map_err(|e| ProxyError::Io(e))?;
+    
+    if resp[1] != 90 {
+        return Err(ProxyError::Proxy(format!("SOCKS4 request rejected: code {}", resp[1])));
+    }
+    
+    Ok(())
+}
+
+pub async fn connect_socks5(
+    stream: &mut TcpStream,
+    target: SocketAddr,
+    username: Option<&str>,
+    password: Option<&str>,
+) -> Result<()> {
+    // 1. Auth negotiation
+    // VER (1) | NMETHODS (1) | METHODS (variable)
+    let mut methods = vec![0u8]; // No auth
+    if username.is_some() {
+        methods.push(2u8); // Username/Password
+    }
+    
+    let mut buf = vec![5u8, methods.len() as u8];
+    buf.extend_from_slice(&methods);
+    
+    stream.write_all(&buf).await.map_err(|e| ProxyError::Io(e))?;
+    
+    let mut resp = [0u8; 2];
+    stream.read_exact(&mut resp).await.map_err(|e| ProxyError::Io(e))?;
+    
+    if resp[0] != 5 {
+        return Err(ProxyError::Proxy("Invalid SOCKS5 version".to_string()));
+    }
+    
+    match resp[1] {
+        0 => {}, // No auth
+        2 => {
+            // Username/Password auth
+            if let (Some(u), Some(p)) = (username, password) {
+                let u_bytes = u.as_bytes();
+                let p_bytes = p.as_bytes();
+                
+                let mut auth_buf = Vec::with_capacity(3 + u_bytes.len() + p_bytes.len());
+                auth_buf.push(1); // VER
+                auth_buf.push(u_bytes.len() as u8);
+                auth_buf.extend_from_slice(u_bytes);
+                auth_buf.push(p_bytes.len() as u8);
+                auth_buf.extend_from_slice(p_bytes);
+                
+                stream.write_all(&auth_buf).await.map_err(|e| ProxyError::Io(e))?;
+                
+                let mut auth_resp = [0u8; 2];
+                stream.read_exact(&mut auth_resp).await.map_err(|e| ProxyError::Io(e))?;
+                
+                if auth_resp[1] != 0 {
+                    return Err(ProxyError::Proxy("SOCKS5 authentication failed".to_string()));
+                }
+            } else {
+                return Err(ProxyError::Proxy("SOCKS5 server requires authentication".to_string()));
+            }
+        },
+        _ => return Err(ProxyError::Proxy("Unsupported SOCKS5 auth method".to_string())),
+    }
+    
+    // 2. Connection request
+    // VER (1) | CMD (1) | RSV (1) | ATYP (1) | DST.ADDR (variable) | DST.PORT (2)
+    let mut req = vec![5u8, 1u8, 0u8]; // CONNECT
+    
+    match target {
+        SocketAddr::V4(v4) => {
+            req.push(1u8); // IPv4
+            req.extend_from_slice(&v4.ip().octets());
+        },
+        SocketAddr::V6(v6) => {
+            req.push(4u8); // IPv6
+            req.extend_from_slice(&v6.ip().octets());
+        },
+    }
+    
+    req.extend_from_slice(&target.port().to_be_bytes());
+    
+    stream.write_all(&req).await.map_err(|e| ProxyError::Io(e))?;
+    
+    // Response
+    let mut head = [0u8; 4];
+    stream.read_exact(&mut head).await.map_err(|e| ProxyError::Io(e))?;
+    
+    if head[1] != 0 {
+        return Err(ProxyError::Proxy(format!("SOCKS5 request failed: code {}", head[1])));
+    }
+    
+    // Skip address part of response
+    match head[3] {
+        1 => { // IPv4
+            let mut addr = [0u8; 4 + 2];
+            stream.read_exact(&mut addr).await.map_err(|e| ProxyError::Io(e))?;
+        },
+        3 => { // Domain
+            let mut len = [0u8; 1];
+            stream.read_exact(&mut len).await.map_err(|e| ProxyError::Io(e))?;
+            let mut addr = vec![0u8; len[0] as usize + 2];
+            stream.read_exact(&mut addr).await.map_err(|e| ProxyError::Io(e))?;
+        },
+        4 => { // IPv6
+            let mut addr = [0u8; 16 + 2];
+            stream.read_exact(&mut addr).await.map_err(|e| ProxyError::Io(e))?;
+        },
+        _ => return Err(ProxyError::Proxy("Invalid address type in SOCKS5 response".to_string())),
+    }
+    
+    Ok(())
+}

src/transport/upstream.rs

@@ -0,0 +1,488 @@
+//! Upstream Management with per-DC latency-weighted selection
+
+use std::net::{SocketAddr, IpAddr};
+use std::sync::Arc;
+use std::time::Duration;
+use tokio::net::TcpStream;
+use tokio::sync::RwLock;
+use tokio::time::Instant;
+use rand::Rng;
+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::socks::{connect_socks4, connect_socks5};
+
+/// Number of Telegram datacenters
+const NUM_DCS: usize = 5;
+
+// ============= RTT Tracking =============
+
+#[derive(Debug, Clone, Copy)]
+struct LatencyEma {
+    value_ms: Option<f64>,
+    alpha: f64,
+}
+
+impl LatencyEma {
+    const fn new(alpha: f64) -> Self {
+        Self { value_ms: None, alpha }
+    }
+    
+    fn update(&mut self, sample_ms: f64) {
+        self.value_ms = Some(match self.value_ms {
+            None => sample_ms,
+            Some(prev) => prev * (1.0 - self.alpha) + sample_ms * self.alpha,
+        });
+    }
+    
+    fn get(&self) -> Option<f64> {
+        self.value_ms
+    }
+}
+
+// ============= Upstream State =============
+
+#[derive(Debug)]
+struct UpstreamState {
+    config: UpstreamConfig,
+    healthy: bool,
+    fails: u32,
+    last_check: std::time::Instant,
+    /// Per-DC latency EMA (index 0 = DC1, index 4 = DC5)
+    dc_latency: [LatencyEma; NUM_DCS],
+}
+
+impl UpstreamState {
+    fn new(config: UpstreamConfig) -> Self {
+        Self {
+            config,
+            healthy: true,
+            fails: 0,
+            last_check: std::time::Instant::now(),
+            dc_latency: [LatencyEma::new(0.3); NUM_DCS],
+        }
+    }
+    
+    /// Map DC index to latency array slot (0..NUM_DCS).
+    ///
+    /// Matches the C implementation's `mf_cluster_lookup` behavior:
+    /// - Standard DCs ±1..±5 → direct mapping to array index 0..4
+    /// - Unknown DCs (CDN, media, etc.) → default DC slot (index 1 = DC 2)
+    ///   This matches Telegram's `default 2;` in proxy-multi.conf.
+    /// - There is NO modular arithmetic in the C implementation.
+    fn dc_array_idx(dc_idx: i16) -> Option<usize> {
+        let abs_dc = dc_idx.unsigned_abs() as usize;
+        if abs_dc == 0 {
+            return None;
+        }
+        if abs_dc >= 1 && abs_dc <= NUM_DCS {
+            Some(abs_dc - 1)
+        } else {
+            // Unknown DC → default cluster (DC 2, index 1)
+            // Same as C: mf_cluster_lookup returns default_cluster
+            Some(1)
+        }
+    }
+    
+    /// Get latency for a specific DC, falling back to average across all known DCs
+    fn effective_latency(&self, dc_idx: Option<i16>) -> Option<f64> {
+        // Try DC-specific latency first
+        if let Some(di) = dc_idx.and_then(Self::dc_array_idx) {
+            if let Some(ms) = self.dc_latency[di].get() {
+                return Some(ms);
+            }
+        }
+        
+        // Fallback: average of all known DC latencies
+        let (sum, count) = self.dc_latency.iter()
+            .filter_map(|l| l.get())
+            .fold((0.0, 0u32), |(s, c), v| (s + v, c + 1));
+        
+        if count > 0 { Some(sum / count as f64) } else { None }
+    }
+}
+
+/// Result of a single DC ping
+#[derive(Debug, Clone)]
+pub struct DcPingResult {
+    pub dc_idx: usize,
+    pub dc_addr: SocketAddr,
+    pub rtt_ms: Option<f64>,
+    pub error: Option<String>,
+}
+
+/// Result of startup ping for one upstream
+#[derive(Debug, Clone)]
+pub struct StartupPingResult {
+    pub results: Vec<DcPingResult>,
+    pub upstream_name: String,
+}
+
+// ============= Upstream Manager =============
+
+#[derive(Clone)]
+pub struct UpstreamManager {
+    upstreams: Arc<RwLock<Vec<UpstreamState>>>,
+}
+
+impl UpstreamManager {
+    pub fn new(configs: Vec<UpstreamConfig>) -> Self {
+        let states = configs.into_iter()
+            .filter(|c| c.enabled)
+            .map(UpstreamState::new)
+            .collect();
+            
+        Self {
+            upstreams: Arc::new(RwLock::new(states)),
+        }
+    }
+    
+    /// Select upstream using latency-weighted random selection.
+    ///
+    /// `effective_weight = config_weight × latency_factor`
+    ///
+    /// where `latency_factor = 1000 / latency_ms` if latency is known,
+    /// or `1.0` if no latency data is available.
+    ///
+    /// This means a 50ms upstream gets factor 20, a 200ms upstream gets
+    /// factor 5 — the faster route is 4× more likely to be chosen
+    /// (all else being equal).
+    async fn select_upstream(&self, dc_idx: Option<i16>) -> Option<usize> {
+        let upstreams = self.upstreams.read().await;
+        if upstreams.is_empty() {
+            return None;
+        }
+
+        let healthy: Vec<usize> = upstreams.iter()
+            .enumerate()
+            .filter(|(_, u)| u.healthy)
+            .map(|(i, _)| i)
+            .collect();
+            
+        if healthy.is_empty() {
+            // All unhealthy — pick any
+            return Some(rand::rng().gen_range(0..upstreams.len()));
+        }
+        
+        if healthy.len() == 1 {
+            return Some(healthy[0]);
+        }
+        
+        // Calculate latency-weighted scores
+        let weights: Vec<(usize, f64)> = healthy.iter().map(|&i| {
+            let base = upstreams[i].config.weight as f64;
+            let latency_factor = upstreams[i].effective_latency(dc_idx)
+                .map(|ms| if ms > 1.0 { 1000.0 / ms } else { 1000.0 })
+                .unwrap_or(1.0);
+            
+            (i, base * latency_factor)
+        }).collect();
+        
+        let total: f64 = weights.iter().map(|(_, w)| w).sum();
+        
+        if total <= 0.0 {
+            return Some(healthy[rand::rng().gen_range(0..healthy.len())]);
+        }
+        
+        let mut choice: f64 = rand::rng().gen_range(0.0..total);
+        
+        for &(idx, weight) in &weights {
+            if choice < weight {
+                trace!(
+                    upstream = idx,
+                    dc = ?dc_idx,
+                    weight = format!("{:.2}", weight),
+                    total = format!("{:.2}", total),
+                    "Upstream selected"
+                );
+                return Some(idx);
+            }
+            choice -= weight;
+        }
+        
+        Some(healthy[0])
+    }
+    
+    /// Connect to target through a selected upstream.
+    ///
+    /// `dc_idx` is used for latency-based upstream selection and RTT tracking.
+    /// Pass `None` if DC index is unknown.
+    pub async fn connect(&self, target: SocketAddr, dc_idx: Option<i16>) -> Result<TcpStream> {
+        let idx = self.select_upstream(dc_idx).await
+            .ok_or_else(|| ProxyError::Config("No upstreams available".to_string()))?;
+            
+        let upstream = {
+            let guard = self.upstreams.read().await;
+            guard[idx].config.clone()
+        };
+        
+        let start = Instant::now();
+        
+        match self.connect_via_upstream(&upstream, target).await {
+            Ok(stream) => {
+                let rtt_ms = start.elapsed().as_secs_f64() * 1000.0;
+                let mut guard = self.upstreams.write().await;
+                if let Some(u) = guard.get_mut(idx) {
+                    if !u.healthy {
+                        debug!(rtt_ms = format!("{:.1}", rtt_ms), "Upstream recovered");
+                    }
+                    u.healthy = true;
+                    u.fails = 0;
+                    
+                    // Store per-DC latency
+                    if let Some(di) = dc_idx.and_then(UpstreamState::dc_array_idx) {
+                        u.dc_latency[di].update(rtt_ms);
+                    }
+                }
+                Ok(stream)
+            },
+            Err(e) => {
+                let mut guard = self.upstreams.write().await;
+                if let Some(u) = guard.get_mut(idx) {
+                    u.fails += 1;
+                    warn!(fails = u.fails, "Upstream failed: {}", e);
+                    if u.fails > 3 {
+                        u.healthy = false;
+                        warn!("Upstream marked unhealthy");
+                    }
+                }
+                Err(e)
+            }
+        }
+    }
+    
+    async fn connect_via_upstream(&self, config: &UpstreamConfig, target: SocketAddr) -> Result<TcpStream> {
+        match &config.upstream_type {
+            UpstreamType::Direct { interface } => {
+                let bind_ip = interface.as_ref()
+                    .and_then(|s| s.parse::<IpAddr>().ok());
+                
+                let socket = create_outgoing_socket_bound(target, bind_ip)?;
+                
+                socket.set_nonblocking(true)?;
+                match socket.connect(&target.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)?;
+                
+                stream.writable().await?;
+                if let Some(e) = stream.take_error()? {
+                    return Err(ProxyError::Io(e));
+                }
+                
+                Ok(stream)
+            },
+            UpstreamType::Socks4 { address, interface, user_id } => {
+                let proxy_addr: SocketAddr = address.parse()
+                    .map_err(|_| ProxyError::Config("Invalid SOCKS4 address".to_string()))?;
+                    
+                let bind_ip = interface.as_ref()
+                    .and_then(|s| s.parse::<IpAddr>().ok());
+                
+                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)),
+                }
+                
+                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));
+                }
+                
+                connect_socks4(&mut stream, target, user_id.as_deref()).await?;
+                Ok(stream)
+            },
+            UpstreamType::Socks5 { address, interface, username, password } => {
+                let proxy_addr: SocketAddr = address.parse()
+                    .map_err(|_| ProxyError::Config("Invalid SOCKS5 address".to_string()))?;
+                    
+                let bind_ip = interface.as_ref()
+                    .and_then(|s| s.parse::<IpAddr>().ok());
+                
+                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)),
+                }
+                
+                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));
+                }
+                
+                connect_socks5(&mut stream, target, username.as_deref(), password.as_deref()).await?;
+                Ok(stream)
+            },
+        }
+    }
+    
+    // ============= Startup Ping =============
+    
+    /// Ping all Telegram DCs through all upstreams.
+    pub async fn ping_all_dcs(&self, prefer_ipv6: bool) -> Vec<StartupPingResult> {
+        let upstreams: Vec<(usize, UpstreamConfig)> = {
+            let guard = self.upstreams.read().await;
+            guard.iter().enumerate()
+                .map(|(i, u)| (i, u.config.clone()))
+                .collect()
+        };
+        
+        let datacenters = if prefer_ipv6 { &*TG_DATACENTERS_V6 } else { &*TG_DATACENTERS_V4 };
+        
+        let mut all_results = Vec::new();
+        
+        for (upstream_idx, upstream_config) in &upstreams {
+            let upstream_name = match &upstream_config.upstream_type {
+                UpstreamType::Direct { interface } => {
+                    format!("direct{}", interface.as_ref().map(|i| format!(" ({})", i)).unwrap_or_default())
+                }
+                UpstreamType::Socks4 { address, .. } => format!("socks4://{}", address),
+                UpstreamType::Socks5 { address, .. } => format!("socks5://{}", address),
+            };
+            
+            let mut dc_results = Vec::new();
+            
+            for (dc_zero_idx, dc_ip) in datacenters.iter().enumerate() {
+                let dc_addr = SocketAddr::new(*dc_ip, TG_DATACENTER_PORT);
+                
+                let ping_result = tokio::time::timeout(
+                    Duration::from_secs(5),
+                    self.ping_single_dc(upstream_config, dc_addr)
+                ).await;
+                
+                let result = match ping_result {
+                    Ok(Ok(rtt_ms)) => {
+                        // Store per-DC latency
+                        let mut guard = self.upstreams.write().await;
+                        if let Some(u) = guard.get_mut(*upstream_idx) {
+                            u.dc_latency[dc_zero_idx].update(rtt_ms);
+                        }
+                        DcPingResult {
+                            dc_idx: dc_zero_idx + 1,
+                            dc_addr,
+                            rtt_ms: Some(rtt_ms),
+                            error: None,
+                        }
+                    }
+                    Ok(Err(e)) => DcPingResult {
+                        dc_idx: dc_zero_idx + 1,
+                        dc_addr,
+                        rtt_ms: None,
+                        error: Some(e.to_string()),
+                    },
+                    Err(_) => DcPingResult {
+                        dc_idx: dc_zero_idx + 1,
+                        dc_addr,
+                        rtt_ms: None,
+                        error: Some("timeout (5s)".to_string()),
+                    },
+                };
+                
+                dc_results.push(result);
+            }
+            
+            all_results.push(StartupPingResult {
+                results: dc_results,
+                upstream_name,
+            });
+        }
+        
+        all_results
+    }
+    
+    async fn ping_single_dc(&self, config: &UpstreamConfig, target: SocketAddr) -> Result<f64> {
+        let start = Instant::now();
+        let _stream = self.connect_via_upstream(config, target).await?;
+        Ok(start.elapsed().as_secs_f64() * 1000.0)
+    }
+    
+    // ============= Health Checks =============
+    
+    /// Background health check: rotates through DCs, 30s interval.
+    pub async fn run_health_checks(&self, prefer_ipv6: bool) {
+        let datacenters = if prefer_ipv6 { &*TG_DATACENTERS_V6 } else { &*TG_DATACENTERS_V4 };
+        let mut dc_rotation = 0usize;
+        
+        loop {
+            tokio::time::sleep(Duration::from_secs(30)).await;
+            
+            let dc_zero_idx = dc_rotation % datacenters.len();
+            dc_rotation += 1;
+            
+            let check_target = SocketAddr::new(datacenters[dc_zero_idx], TG_DATACENTER_PORT);
+            
+            let count = self.upstreams.read().await.len();
+            for i in 0..count {
+                let config = {
+                    let guard = self.upstreams.read().await;
+                    guard[i].config.clone()
+                };
+                
+                let start = Instant::now();
+                let result = tokio::time::timeout(
+                    Duration::from_secs(10),
+                    self.connect_via_upstream(&config, check_target)
+                ).await;
+                
+                let mut guard = self.upstreams.write().await;
+                let u = &mut guard[i];
+                
+                match result {
+                    Ok(Ok(_stream)) => {
+                        let rtt_ms = start.elapsed().as_secs_f64() * 1000.0;
+                        u.dc_latency[dc_zero_idx].update(rtt_ms);
+                        
+                        if !u.healthy {
+                            info!(
+                                rtt = format!("{:.0}ms", rtt_ms),
+                                dc = dc_zero_idx + 1,
+                                "Upstream recovered"
+                            );
+                        }
+                        u.healthy = true;
+                        u.fails = 0;
+                    }
+                    Ok(Err(e)) => {
+                        u.fails += 1;
+                        debug!(dc = dc_zero_idx + 1, fails = u.fails,
+                            "Health check failed: {}", e);
+                        if u.fails > 3 {
+                            u.healthy = false;
+                            warn!("Upstream unhealthy (fails)");
+                        }
+                    }
+                    Err(_) => {
+                        u.fails += 1;
+                        debug!(dc = dc_zero_idx + 1, fails = u.fails,
+                            "Health check timeout");
+                        if u.fails > 3 {
+                            u.healthy = false;
+                            warn!("Upstream unhealthy (timeout)");
+                        }
+                    }
+                }
+                u.last_check = std::time::Instant::now();
+            }
+        }
+    }
+}

src/util/ip.rs

@@ -1,6 +1,6 @@
 //! IP Addr Detect
 
-use std::net::IpAddr;
+use std::net::{IpAddr, SocketAddr, UdpSocket};
 use std::time::Duration;
 use tracing::{debug, warn};
 
@@ -40,31 +40,77 @@ const IPV6_URLS: &[&str] = &[
     "http://api6.ipify.org/",
 ];
 
+/// Detect local IP address by connecting to a public DNS
+/// This does not actually send any packets
+fn get_local_ip(target: &str) -> Option<IpAddr> {
+    let socket = UdpSocket::bind("0.0.0.0:0").ok()?;
+    socket.connect(target).ok()?;
+    socket.local_addr().ok().map(|addr| addr.ip())
+}
+
+fn get_local_ipv6(target: &str) -> Option<IpAddr> {
+    let socket = UdpSocket::bind("[::]:0").ok()?;
+    socket.connect(target).ok()?;
+    socket.local_addr().ok().map(|addr| addr.ip())
+}
+
 /// Detect public IP addresses
 pub async fn detect_ip() -> IpInfo {
     let mut info = IpInfo::default();
 
-    // Detect IPv4
+    // Try to get local interface IP first (default gateway interface)
+    // We connect to Google DNS to find out which interface is used for routing
+    if let Some(ip) = get_local_ip("8.8.8.8:80") {
+        if ip.is_ipv4() && !ip.is_loopback() {
+             info.ipv4 = Some(ip);
+             debug!(ip = %ip, "Detected local IPv4 address via routing");
+        }
+    }
+
+    if let Some(ip) = get_local_ipv6("[2001:4860:4860::8888]:80") {
+        if ip.is_ipv6() && !ip.is_loopback() {
+            info.ipv6 = Some(ip);
+            debug!(ip = %ip, "Detected local IPv6 address via routing");
+        }
+    }
+    
+    // If local detection failed or returned private IP (and we want public), 
+    // or just as a fallback/verification, we might want to check external services.
+    // However, the requirement is: "if IP for listening is not set... it should be IP from interface... 
+    // if impossible - request external resources".
+    
+    // So if we found a local IP, we might be good. But often servers are behind NAT.
+    // If the local IP is private, we probably want the public IP for the tg:// link.
+    // Let's check if the detected IPs are private.
+    
+    let need_external_v4 = info.ipv4.map_or(true, |ip| is_private_ip(ip));
+    let need_external_v6 = info.ipv6.map_or(true, |ip| is_private_ip(ip));
+
+    if need_external_v4 {
+        debug!("Local IPv4 is private or missing, checking external services...");
         for url in IPV4_URLS {
             if let Some(ip) = fetch_ip(url).await {
                 if ip.is_ipv4() {
                     info.ipv4 = Some(ip);
-                debug!(ip = %ip, "Detected IPv4 address");
+                    debug!(ip = %ip, "Detected public IPv4 address");
                     break;
                 }
             }
         }
+    }
     
-    // Detect IPv6
+    if need_external_v6 {
+        debug!("Local IPv6 is private or missing, checking external services...");
         for url in IPV6_URLS {
             if let Some(ip) = fetch_ip(url).await {
                 if ip.is_ipv6() {
                     info.ipv6 = Some(ip);
-                debug!(ip = %ip, "Detected IPv6 address");
+                    debug!(ip = %ip, "Detected public IPv6 address");
                     break;
                 }
             }
         }
+    }
     
     if !info.has_any() {
         warn!("Failed to detect public IP address");
@@ -73,6 +119,17 @@ pub async fn detect_ip() -> IpInfo {
     info
 }
 
+fn is_private_ip(ip: IpAddr) -> bool {
+    match ip {
+        IpAddr::V4(ipv4) => {
+            ipv4.is_private() || ipv4.is_loopback() || ipv4.is_link_local()
+        }
+        IpAddr::V6(ipv6) => {
+            ipv6.is_loopback() || (ipv6.segments()[0] & 0xfe00) == 0xfc00 // Unique Local
+        }
+    }
+}
+
 /// Fetch IP from URL
 async fn fetch_ip(url: &str) -> Option<IpAddr> {
     let client = reqwest::Client::builder()