Merge branch 'master' into stable

This reverts commit dd1d0a6262.

Dockerfile

@@ -1,4 +1,4 @@
-FROM alpine:3.6
+FROM alpine:3.8
 
 RUN adduser tgproxy -u 10000 -D
 

README.md

@@ -24,3 +24,4 @@ The proxy can be launched:
 - with a custom config: `python3 mtprotoproxy.py [configfile]`
 - several times, clients will be automaticaly balanced between instances
 - using *PyPy* interprteter
+- with runtime statistics exported for [Prometheus](https://prometheus.io/): using [prometheus](https://github.com/alexbers/mtprotoproxy/tree/prometheus) branch

config.py

@@ -8,3 +8,7 @@ USERS = {
 
 # Tag for advertising, obtainable from @MTProxybot
 # AD_TAG = "3c09c680b76ee91a4c25ad51f742267d"
+
+# Uncommenting this do make a proxy harder to detect
+# But it can be incompatible with old clients
+# SECURE_ONLY = True

mtprotoproxy.py

@@ -144,15 +144,32 @@ USERS = config["USERS"]
 AD_TAG = bytes.fromhex(config.get("AD_TAG", ""))
 
 # load advanced settings
+# if IPv6 avaliable, use it by default
 PREFER_IPV6 = config.get("PREFER_IPV6", socket.has_ipv6)
 # disables tg->client trafic reencryption, faster but less secure
 FAST_MODE = config.get("FAST_MODE", True)
+# doesn't allow to connect in not-secure mode
+SECURE_ONLY = config.get("SECURE_ONLY", False)
+# delay in seconds between stats printing
 STATS_PRINT_PERIOD = config.get("STATS_PRINT_PERIOD", 600)
-PROXY_INFO_UPDATE_PERIOD = config.get("PROXY_INFO_UPDATE_PERIOD", 60*60*24)
-READ_BUF_SIZE = config.get("READ_BUF_SIZE", 16384)
-WRITE_BUF_SIZE = config.get("WRITE_BUF_SIZE", 65536)
-CLIENT_KEEPALIVE = config.get("CLIENT_KEEPALIVE", 60*30)
+# delay in seconds between middle proxy info updates
+PROXY_INFO_UPDATE_PERIOD = config.get("PROXY_INFO_UPDATE_PERIOD", 24*60*60)
+# max socket buffer size to the client direction, the more the faster, but more RAM hungry
+TO_CLT_BUFSIZE = config.get("TO_CLT_BUFSIZE", 16384)
+# max socket buffer size to the telegram servers direction
+TO_TG_BUFSIZE = config.get("TO_TG_BUFSIZE", 65536)
+# keepalive period for clients in secs
+CLIENT_KEEPALIVE = config.get("CLIENT_KEEPALIVE", 10*60)
+# drop client after this timeout if the handshake fail
 CLIENT_HANDSHAKE_TIMEOUT = config.get("CLIENT_HANDSHAKE_TIMEOUT", 10)
+# if client doesn't confirm data for this number of seconds, it is dropped
+CLIENT_ACK_TIMEOUT = config.get("CLIENT_ACK_TIMEOUT", 5*60)
+# telegram servers connect timeout in seconds
+TG_CONNECT_TIMEOUT = config.get("TG_CONNECT_TIMEOUT", 10)
+# listen address for IPv4
+LISTEN_ADDR_IPV4 = config.get("LISTEN_ADDR_IPV4", "0.0.0.0")
+# listen address for IPv6
+LISTEN_ADDR_IPV6 = config.get("LISTEN_ADDR_IPV6", "::")
 
 TG_DATACENTER_PORT = 443
 
@@ -203,6 +220,7 @@ DC_IDX_POS = 60
 
 PROTO_TAG_ABRIDGED = b"\xef\xef\xef\xef"
 PROTO_TAG_INTERMEDIATE = b"\xee\xee\xee\xee"
+PROTO_TAG_SECURE = b"\xdd\xdd\xdd\xdd"
 
 CBC_PADDING = 16
 PADDING_FILLER = b"\x04\x00\x00\x00"
@@ -222,14 +240,14 @@ def init_stats():
     stats = {user: collections.Counter() for user in USERS}
 
 
-def update_stats(user, connects=0, curr_connects=0, octets=0):
+def update_stats(user, connects=0, curr_connects=0, octets=0, msgs=0):
     global stats
 
     if user not in stats:
         stats[user] = collections.Counter()
 
     stats[user].update(connects=connects, curr_connects=curr_connects,
-                       octets=octets)
+                       octets=octets, msgs=msgs)
 
 
 class LayeredStreamReaderBase:
@@ -428,7 +446,6 @@ class MTProtoIntermediateFrameStreamReader(LayeredStreamReaderBase):
             msg_len -= 0x80000000
 
         data = await self.upstream.readexactly(msg_len)
-
         return data, extra
 
 
@@ -440,6 +457,38 @@ class MTProtoIntermediateFrameStreamWriter(LayeredStreamWriterBase):
             return self.upstream.write(int.to_bytes(len(data), 4, 'little') + data)
 
 
+class MTProtoSecureIntermediateFrameStreamReader(LayeredStreamReaderBase):
+    async def read(self, buf_size):
+        msg_len_bytes = await self.upstream.readexactly(4)
+        msg_len = int.from_bytes(msg_len_bytes, "little")
+
+        extra = {}
+        if msg_len > 0x80000000:
+            extra["QUICKACK_FLAG"] = True
+            msg_len -= 0x80000000
+
+        data = await self.upstream.readexactly(msg_len)
+
+        if msg_len % 4 != 0:
+            cut_border = msg_len - (msg_len % 4)
+            data = data[:cut_border]
+
+        return data, extra
+
+
+class MTProtoSecureIntermediateFrameStreamWriter(LayeredStreamWriterBase):
+    def write(self, data, extra={}):
+        MAX_PADDING_LEN = 4
+        if extra.get("SIMPLE_ACK"):
+            # TODO: make this unpredictable
+            return self.upstream.write(data)
+        else:
+            padding_len = random.randrange(MAX_PADDING_LEN)
+            padding = bytearray([random.randrange(256) for i in range(padding_len)])
+            padded_data_len_bytes = int.to_bytes(len(data) + padding_len, 4, 'little')
+            return self.upstream.write(padded_data_len_bytes + data + padding)
+
+
 class ProxyReqStreamReader(LayeredStreamReaderBase):
     async def read(self, msg):
         RPC_PROXY_ANS = b"\x0d\xda\x03\x44"
@@ -498,6 +547,7 @@ class ProxyReqStreamWriter(LayeredStreamWriterBase):
         FLAG_HAS_AD_TAG = 0x8
         FLAG_MAGIC = 0x1000
         FLAG_EXTMODE2 = 0x20000
+        FLAG_PAD = 0x8000000
         FLAG_INTERMEDIATE = 0x20000000
         FLAG_ABRIDGED = 0x40000000
         FLAG_QUICKACK = 0x80000000
@@ -512,6 +562,8 @@ class ProxyReqStreamWriter(LayeredStreamWriterBase):
             flags |= FLAG_ABRIDGED
         elif self.proto_tag == PROTO_TAG_INTERMEDIATE:
             flags |= FLAG_INTERMEDIATE
+        elif self.proto_tag == PROTO_TAG_SECURE:
+            flags |= FLAG_INTERMEDIATE | FLAG_PAD
 
         if extra.get("QUICKACK_FLAG"):
             flags |= FLAG_QUICKACK
@@ -548,7 +600,10 @@ async def handle_handshake(reader, writer):
         decrypted = decryptor.decrypt(handshake)
 
         proto_tag = decrypted[PROTO_TAG_POS:PROTO_TAG_POS+4]
-        if proto_tag not in (PROTO_TAG_ABRIDGED, PROTO_TAG_INTERMEDIATE):
+        if proto_tag not in (PROTO_TAG_ABRIDGED, PROTO_TAG_INTERMEDIATE, PROTO_TAG_SECURE):
+            continue
+
+        if SECURE_ONLY and proto_tag != PROTO_TAG_SECURE:
             continue
 
         dc_idx = int.from_bytes(decrypted[DC_IDX_POS:DC_IDX_POS+2], "little", signed=True)
@@ -557,13 +612,49 @@ async def handle_handshake(reader, writer):
         writer = CryptoWrappedStreamWriter(writer, encryptor)
         return reader, writer, proto_tag, user, dc_idx, enc_key + enc_iv
 
-    while await reader.read(READ_BUF_SIZE):
+    EMPTY_READ_BUF_SIZE = 4096
+    while await reader.read(EMPTY_READ_BUF_SIZE):
         # just consume all the data
         pass
 
     return False
 
 
+def set_keepalive(sock, interval=40, attempts=5):
+    sock.setsockopt(socket.SOL_SOCKET, socket.SO_KEEPALIVE, 1)
+    if hasattr(socket, "TCP_KEEPIDLE"):
+        sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_KEEPIDLE, interval)
+    if hasattr(socket, "TCP_KEEPINTVL"):
+        sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_KEEPINTVL, interval)
+    if hasattr(socket, "TCP_KEEPCNT"):
+        sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_KEEPCNT, attempts)
+
+
+def set_ack_timeout(sock, timeout):
+    if hasattr(socket, "TCP_USER_TIMEOUT"):
+        sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_USER_TIMEOUT, timeout*1000)
+
+
+def set_bufsizes(sock, recv_buf, send_buf):
+    sock.setsockopt(socket.SOL_SOCKET, socket.SO_RCVBUF, recv_buf)
+    sock.setsockopt(socket.SOL_SOCKET, socket.SO_SNDBUF, send_buf)
+
+
+async def open_connection_tryer(addr, port, limit, timeout, max_attempts=3):
+    for attempt in range(max_attempts-1):
+        try:
+            task = asyncio.open_connection(addr, port, limit=limit)
+            reader_tgt, writer_tgt = await asyncio.wait_for(task, timeout=timeout)
+            return reader_tgt, writer_tgt
+        except (OSError, asyncio.TimeoutError):
+            continue
+
+    # the last attempt
+    task = asyncio.open_connection(addr, port, limit=limit)
+    reader_tgt, writer_tgt = await asyncio.wait_for(task, timeout=timeout)
+    return reader_tgt, writer_tgt
+
+
 async def do_direct_handshake(proto_tag, dc_idx, dec_key_and_iv=None):
     RESERVED_NONCE_FIRST_CHARS = [b"\xef"]
     RESERVED_NONCE_BEGININGS = [b"\x48\x45\x41\x44", b"\x50\x4F\x53\x54",
@@ -582,15 +673,18 @@ async def do_direct_handshake(proto_tag, dc_idx, dec_key_and_iv=None):
         dc = TG_DATACENTERS_V4[dc_idx]
 
     try:
-        reader_tgt, writer_tgt = await asyncio.open_connection(dc, TG_DATACENTER_PORT,
-                                                               limit=READ_BUF_SIZE)
+        reader_tgt, writer_tgt = await open_connection_tryer(
+            dc, TG_DATACENTER_PORT, limit=TO_CLT_BUFSIZE, timeout=TG_CONNECT_TIMEOUT)
     except ConnectionRefusedError as E:
         print_err("Got connection refused while trying to connect to", dc, TG_DATACENTER_PORT)
         return False
-    except OSError as E:
+    except (OSError, asyncio.TimeoutError) as E:
         print_err("Unable to connect to", dc, TG_DATACENTER_PORT)
         return False
 
+    set_keepalive(writer_tgt.get_extra_info("socket"))
+    set_bufsizes(writer_tgt.get_extra_info("socket"), TO_CLT_BUFSIZE, TO_TG_BUFSIZE)
+
     while True:
         rnd = bytearray([random.randrange(0, 256) for i in range(HANDSHAKE_LEN)])
         if rnd[:1] in RESERVED_NONCE_FIRST_CHARS:
@@ -653,21 +747,6 @@ def get_middleproxy_aes_key_and_iv(nonce_srv, nonce_clt, clt_ts, srv_ip, clt_por
     return key, iv
 
 
-def set_keepalive(sock, interval=40, attempts=5):
-    sock.setsockopt(socket.SOL_SOCKET, socket.SO_KEEPALIVE, 1)
-    if hasattr(socket, "TCP_KEEPIDLE"):
-        sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_KEEPIDLE, interval)
-    if hasattr(socket, "TCP_KEEPINTVL"):
-        sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_KEEPINTVL, interval)
-    if hasattr(socket, "TCP_KEEPCNT"):
-        sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_KEEPCNT, attempts)
-
-
-def set_bufsizes(sock, recv_buf=READ_BUF_SIZE, send_buf=WRITE_BUF_SIZE):
-    sock.setsockopt(socket.SOL_SOCKET, socket.SO_RCVBUF, recv_buf)
-    sock.setsockopt(socket.SOL_SOCKET, socket.SO_SNDBUF, send_buf)
-
-
 async def do_middleproxy_handshake(proto_tag, dc_idx, cl_ip, cl_port):
     START_SEQ_NO = -2
     NONCE_LEN = 16
@@ -695,16 +774,18 @@ async def do_middleproxy_handshake(proto_tag, dc_idx, cl_ip, cl_port):
         addr, port = random.choice(TG_MIDDLE_PROXIES_V4[dc_idx])
 
     try:
-        reader_tgt, writer_tgt = await asyncio.open_connection(addr, port, limit=READ_BUF_SIZE)
-        set_keepalive(writer_tgt.get_extra_info("socket"))
-        set_bufsizes(writer_tgt.get_extra_info("socket"))
+        reader_tgt, writer_tgt = await open_connection_tryer(addr, port, limit=TO_CLT_BUFSIZE,
+                                                             timeout=TG_CONNECT_TIMEOUT)
     except ConnectionRefusedError as E:
         print_err("Got connection refused while trying to connect to", addr, port)
         return False
-    except OSError as E:
+    except (OSError, asyncio.TimeoutError) as E:
         print_err("Unable to connect to", addr, port)
         return False
 
+    set_keepalive(writer_tgt.get_extra_info("socket"))
+    set_bufsizes(writer_tgt.get_extra_info("socket"), TO_CLT_BUFSIZE, TO_TG_BUFSIZE)
+
     writer_tgt = MTProtoFrameStreamWriter(writer_tgt, START_SEQ_NO)
 
     key_selector = PROXY_SECRET[:4]
@@ -719,7 +800,7 @@ async def do_middleproxy_handshake(proto_tag, dc_idx, cl_ip, cl_port):
 
     old_reader = reader_tgt
     reader_tgt = MTProtoFrameStreamReader(reader_tgt, START_SEQ_NO)
-    ans = await reader_tgt.read(READ_BUF_SIZE)
+    ans = await reader_tgt.read(TO_CLT_BUFSIZE)
 
     if len(ans) != RPC_NONCE_ANS_LEN:
         return False
@@ -800,8 +881,9 @@ async def do_middleproxy_handshake(proto_tag, dc_idx, cl_ip, cl_port):
 
 
 async def handle_client(reader_clt, writer_clt):
-    set_keepalive(writer_clt.get_extra_info("socket"), CLIENT_KEEPALIVE)
-    set_bufsizes(writer_clt.get_extra_info("socket"))
+    set_keepalive(writer_clt.get_extra_info("socket"), CLIENT_KEEPALIVE, attempts=3)
+    set_ack_timeout(writer_clt.get_extra_info("socket"), CLIENT_ACK_TIMEOUT)
+    set_bufsizes(writer_clt.get_extra_info("socket"), TO_TG_BUFSIZE, TO_CLT_BUFSIZE)
 
     try:
         clt_data = await asyncio.wait_for(handle_handshake(reader_clt, writer_clt),
@@ -849,13 +931,16 @@ async def handle_client(reader_clt, writer_clt):
         elif proto_tag == PROTO_TAG_INTERMEDIATE:
             reader_clt = MTProtoIntermediateFrameStreamReader(reader_clt)
             writer_clt = MTProtoIntermediateFrameStreamWriter(writer_clt)
+        elif proto_tag == PROTO_TAG_SECURE:
+            reader_clt = MTProtoSecureIntermediateFrameStreamReader(reader_clt)
+            writer_clt = MTProtoSecureIntermediateFrameStreamWriter(writer_clt)
         else:
             return
 
-    async def connect_reader_to_writer(rd, wr, user):
+    async def connect_reader_to_writer(rd, wr, user, rd_buf_size):
         try:
             while True:
-                data = await rd.read(READ_BUF_SIZE)
+                data = await rd.read(rd_buf_size)
                 if isinstance(data, tuple):
                     data, extra = data
                 else:
@@ -866,15 +951,17 @@ async def handle_client(reader_clt, writer_clt):
                     await wr.drain()
                     return
                 else:
-                    update_stats(user, octets=len(data))
+                    update_stats(user, octets=len(data), msgs=1)
                     wr.write(data, extra)
                     await wr.drain()
         except (OSError, asyncio.streams.IncompleteReadError) as e:
             # print_err(e)
             pass
 
-    task_tg_to_clt = asyncio.ensure_future(connect_reader_to_writer(reader_tg, writer_clt, user))
-    task_clt_to_tg = asyncio.ensure_future(connect_reader_to_writer(reader_clt, writer_tg, user))
+    tg_to_clt = connect_reader_to_writer(reader_tg, writer_clt, user, TO_CLT_BUFSIZE)
+    clt_to_tg = connect_reader_to_writer(reader_clt, writer_tg, user, TO_TG_BUFSIZE)
+    task_tg_to_clt = asyncio.ensure_future(tg_to_clt)
+    task_clt_to_tg = asyncio.ensure_future(clt_to_tg)
 
     update_stats(user, curr_connects=1)
     await asyncio.wait([task_tg_to_clt, task_clt_to_tg], return_when=asyncio.FIRST_COMPLETED)
@@ -902,9 +989,9 @@ async def stats_printer():
 
         print("Stats for", time.strftime("%d.%m.%Y %H:%M:%S"))
         for user, stat in stats.items():
-            print("%s: %d connects (%d current), %.2f MB" % (
+            print("%s: %d connects (%d current), %.2f MB, %d msgs" % (
                 user, stat["connects"], stat["curr_connects"],
-                stat["octets"] / 1000000))
+                stat["octets"] / 1000000, stat["msgs"]))
         print(flush=True)
 
 
@@ -994,7 +1081,7 @@ def init_ip_info():
     TIMEOUT = 5
 
     try:
-        with urllib.request.urlopen('https://v4.ifconfig.co/ip', timeout=TIMEOUT) as f:
+        with urllib.request.urlopen('http://ipv4.myexternalip.com/raw', timeout=TIMEOUT) as f:
             if f.status != 200:
                 raise Exception("Invalid status code")
             my_ip_info["ipv4"] = f.read().decode().strip()
@@ -1003,7 +1090,7 @@ def init_ip_info():
 
     if PREFER_IPV6:
         try:
-            with urllib.request.urlopen('https://v6.ifconfig.co/ip', timeout=TIMEOUT) as f:
+            with urllib.request.urlopen('http://ipv6.myexternalip.com/raw', timeout=TIMEOUT) as f:
                 if f.status != 200:
                     raise Exception("Invalid status code")
                 my_ip_info["ipv6"] = f.read().decode().strip()
@@ -1028,7 +1115,12 @@ def print_tg_info():
 
     for user, secret in sorted(USERS.items(), key=lambda x: x[0]):
         for ip in ip_addrs:
-            params = {"server": ip, "port": PORT, "secret": secret}
+            if not SECURE_ONLY:
+                params = {"server": ip, "port": PORT, "secret": secret}
+                params_encodeded = urllib.parse.urlencode(params, safe=':')
+                print("{}: tg://proxy?{}".format(user, params_encodeded), flush=True)
+
+            params = {"server": ip, "port": PORT, "secret": "dd" + secret}
             params_encodeded = urllib.parse.urlencode(params, safe=':')
             print("{}: tg://proxy?{}".format(user, params_encodeded), flush=True)
 
@@ -1039,15 +1131,24 @@ def loop_exception_handler(loop, context):
     if exception:
         if isinstance(exception, TimeoutError):
             if transport:
-                print_err("Timeout, killing transport")
                 transport.abort()
                 return
         if isinstance(exception, OSError):
             IGNORE_ERRNO = {
-                10038  # operation on non-socket on Windows, likely because fd == -1
+                10038,  # operation on non-socket on Windows, likely because fd == -1
+                121,    # the semaphore timeout period has expired on Windows
+            }
+
+            FORCE_CLOSE_ERRNO = {
+                113,    # no route to host
+
             }
             if exception.errno in IGNORE_ERRNO:
                 return
+            elif exception.errno in FORCE_CLOSE_ERRNO:
+                if transport:
+                    transport.abort()
+                    return
 
     loop.default_exception_handler(context)
 
@@ -1071,13 +1172,13 @@ def main():
 
     reuse_port = hasattr(socket, "SO_REUSEPORT")
 
-    task_v4 = asyncio.start_server(handle_client_wrapper, '0.0.0.0', PORT,
-                                   limit=READ_BUF_SIZE, reuse_port=reuse_port, loop=loop)
+    task_v4 = asyncio.start_server(handle_client_wrapper, LISTEN_ADDR_IPV4, PORT,
+                                   limit=TO_TG_BUFSIZE, reuse_port=reuse_port, loop=loop)
     server_v4 = loop.run_until_complete(task_v4)
 
     if socket.has_ipv6:
-        task_v6 = asyncio.start_server(handle_client_wrapper, '::', PORT,
-                                       limit=READ_BUF_SIZE, reuse_port=reuse_port, loop=loop)
+        task_v6 = asyncio.start_server(handle_client_wrapper, LISTEN_ADDR_IPV6, PORT,
+                                       limit=TO_TG_BUFSIZE, reuse_port=reuse_port, loop=loop)
         server_v6 = loop.run_until_complete(task_v6)
 
     try: