bpo-37228: Fix loop.create_datagram_endpoint()'s usage of SO_REUSEADDR (GH-17311) (GH-17570)

ned-deily · aeros · web-flow · commit 95157c6a281c · 2019-12-10T23:49:23.000-05:00
(cherry picked from commit ab513a3) Co-authored-by: Kyle Stanley <aeros167@gmail.com>
diff --git a/Doc/library/asyncio-eventloop.rst b/Doc/library/asyncio-eventloop.rst
@@ -429,6 +429,21 @@ Opening network connections
                         reuse_address=None, reuse_port=None, \
                         allow_broadcast=None, sock=None)
 
+   .. note::
+      The parameter *reuse_address* is no longer supported, as using
+      :py:data:`~sockets.SO_REUSEADDR` poses a significant security concern for
+      UDP. Explicitly passing ``reuse_address=True`` will raise an exception.
+
+      When multiple processes with differing UIDs assign sockets to an
+      indentical UDP socket address with ``SO_REUSEADDR``, incoming packets can
+      become randomly distributed among the sockets.
+
+      For supported platforms, *reuse_port* can be used as a replacement for
+      similar functionality. With *reuse_port*,
+      :py:data:`~sockets.SO_REUSEPORT` is used instead, which specifically
+      prevents processes with differing UIDs from assigning sockets to the same
+      socket address.
+
    Create a datagram connection.
 
    The socket family can be either :py:data:`~socket.AF_INET`,
@@ -457,11 +472,6 @@ Opening network connections
      resolution. If given, these should all be integers from the
      corresponding :mod:`socket` module constants.
 
-   * *reuse_address* tells the kernel to reuse a local socket in
-     ``TIME_WAIT`` state, without waiting for its natural timeout to
-     expire. If not specified will automatically be set to ``True`` on
-     Unix.
-
    * *reuse_port* tells the kernel to allow this endpoint to be bound to the
      same port as other existing endpoints are bound to, so long as they all
      set this flag when being created. This option is not supported on Windows
@@ -485,6 +495,10 @@ Opening network connections
       The *family*, *proto*, *flags*, *reuse_address*, *reuse_port,
       *allow_broadcast*, and *sock* parameters were added.
 
+   .. versionchanged:: 3.7.6
+      The *reuse_address* parameter is no longer supported due to security
+      concerns.
+
 .. coroutinemethod:: loop.create_unix_connection(protocol_factory, \
                         path=None, \*, ssl=None, sock=None, \
                         server_hostname=None, ssl_handshake_timeout=None)
diff --git a/Lib/asyncio/base_events.py b/Lib/asyncio/base_events.py
@@ -61,6 +61,10 @@
 # Maximum timeout passed to select to avoid OS limitations
 MAXIMUM_SELECT_TIMEOUT = 24 * 3600
 
+# Used for deprecation and removal of `loop.create_datagram_endpoint()`'s
+# *reuse_address* parameter
+_unset = object()
+
 
 def _format_handle(handle):
     cb = handle._callback
@@ -1138,7 +1142,7 @@ async def start_tls(self, transport, protocol, sslcontext, *,
     async def create_datagram_endpoint(self, protocol_factory,
                                        local_addr=None, remote_addr=None, *,
                                        family=0, proto=0, flags=0,
-                                       reuse_address=None, reuse_port=None,
+                                       reuse_address=_unset, reuse_port=None,
                                        allow_broadcast=None, sock=None):
         """Create datagram connection."""
         if sock is not None:
@@ -1147,7 +1151,7 @@ async def create_datagram_endpoint(self, protocol_factory,
                     f'A UDP Socket was expected, got {sock!r}')
             if (local_addr or remote_addr or
                     family or proto or flags or
-                    reuse_address or reuse_port or allow_broadcast):
+                    reuse_port or allow_broadcast):
                 # show the problematic kwargs in exception msg
                 opts = dict(local_addr=local_addr, remote_addr=remote_addr,
                             family=family, proto=proto, flags=flags,
@@ -1201,8 +1205,18 @@ async def create_datagram_endpoint(self, protocol_factory,
 
             exceptions = []
 
-            if reuse_address is None:
-                reuse_address = os.name == 'posix' and sys.platform != 'cygwin'
+            # bpo-37228
+            if reuse_address is not _unset:
+                if reuse_address:
+                    raise ValueError("Passing `reuse_address=True` is no "
+                                     "longer supported, as the usage of "
+                                     "SO_REUSEPORT in UDP poses a significant "
+                                     "security concern.")
+                else:
+                    warnings.warn("The *reuse_address* parameter has been "
+                                  "deprecated as of 3.7.6 and is scheduled "
+                                  "for removal in 3.11.", DeprecationWarning,
+                                  stacklevel=2)
 
             for ((family, proto),
                  (local_address, remote_address)) in addr_pairs_info:
@@ -1211,9 +1225,6 @@ async def create_datagram_endpoint(self, protocol_factory,
                 try:
                     sock = socket.socket(
                         family=family, type=socket.SOCK_DGRAM, proto=proto)
-                    if reuse_address:
-                        sock.setsockopt(
-                            socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
                     if reuse_port:
                         _set_reuseport(sock)
                     if allow_broadcast:
diff --git a/Lib/test/test_asyncio/test_base_events.py b/Lib/test/test_asyncio/test_base_events.py
@@ -1708,10 +1708,6 @@ class FakeSock:
             MyDatagramProto, flags=1, sock=FakeSock())
         self.assertRaises(ValueError, self.loop.run_until_complete, fut)
 
-        fut = self.loop.create_datagram_endpoint(
-            MyDatagramProto, reuse_address=True, sock=FakeSock())
-        self.assertRaises(ValueError, self.loop.run_until_complete, fut)
-
         fut = self.loop.create_datagram_endpoint(
             MyDatagramProto, reuse_port=True, sock=FakeSock())
         self.assertRaises(ValueError, self.loop.run_until_complete, fut)
@@ -1722,7 +1718,6 @@ class FakeSock:
 
     def test_create_datagram_endpoint_sockopts(self):
         # Socket options should not be applied unless asked for.
-        # SO_REUSEADDR defaults to on for UNIX.
         # SO_REUSEPORT is not available on all platforms.
 
         coro = self.loop.create_datagram_endpoint(
@@ -1731,18 +1726,8 @@ def test_create_datagram_endpoint_sockopts(self):
         transport, protocol = self.loop.run_until_complete(coro)
         sock = transport.get_extra_info('socket')
 
-        reuse_address_default_on = (
-            os.name == 'posix' and sys.platform != 'cygwin')
         reuseport_supported = hasattr(socket, 'SO_REUSEPORT')
 
-        if reuse_address_default_on:
-            self.assertTrue(
-                sock.getsockopt(
-                    socket.SOL_SOCKET, socket.SO_REUSEADDR))
-        else:
-            self.assertFalse(
-                sock.getsockopt(
-                    socket.SOL_SOCKET, socket.SO_REUSEADDR))
         if reuseport_supported:
             self.assertFalse(
                 sock.getsockopt(
@@ -1758,13 +1743,12 @@ def test_create_datagram_endpoint_sockopts(self):
         coro = self.loop.create_datagram_endpoint(
             lambda: MyDatagramProto(create_future=True, loop=self.loop),
             local_addr=('127.0.0.1', 0),
-            reuse_address=True,
             reuse_port=reuseport_supported,
             allow_broadcast=True)
         transport, protocol = self.loop.run_until_complete(coro)
         sock = transport.get_extra_info('socket')
 
-        self.assertTrue(
+        self.assertFalse(
             sock.getsockopt(
                 socket.SOL_SOCKET, socket.SO_REUSEADDR))
         if reuseport_supported:
@@ -1779,6 +1763,29 @@ def test_create_datagram_endpoint_sockopts(self):
         self.loop.run_until_complete(protocol.done)
         self.assertEqual('CLOSED', protocol.state)
 
+    def test_create_datagram_endpoint_reuse_address_error(self):
+        # bpo-37228: Ensure that explicit passing of `reuse_address=True`
+        # raises an error, as it is not safe to use SO_REUSEADDR when using UDP
+
+        coro = self.loop.create_datagram_endpoint(
+            lambda: MyDatagramProto(create_future=True, loop=self.loop),
+            local_addr=('127.0.0.1', 0),
+            reuse_address=True)
+
+        with self.assertRaises(ValueError):
+            self.loop.run_until_complete(coro)
+
+    def test_create_datagram_endpoint_reuse_address_warning(self):
+        # bpo-37228: Deprecate *reuse_address* parameter
+
+        coro = self.loop.create_datagram_endpoint(
+            lambda: MyDatagramProto(create_future=True, loop=self.loop),
+            local_addr=('127.0.0.1', 0),
+            reuse_address=False)
+
+        with self.assertWarns(DeprecationWarning):
+            self.loop.run_until_complete(coro)
+
     @patch_socket
     def test_create_datagram_endpoint_nosoreuseport(self, m_socket):
         del m_socket.SO_REUSEPORT
diff --git a/Misc/NEWS.d/next/Security/2019-11-21-21-36-54.bpo-37228.yBZnFG.rst b/Misc/NEWS.d/next/Security/2019-11-21-21-36-54.bpo-37228.yBZnFG.rst
@@ -0,0 +1,6 @@
+Due to significant security concerns, the *reuse_address* parameter of
+:meth:`asyncio.loop.create_datagram_endpoint` is no longer supported. This is
+because of the behavior of ``SO_REUSEADDR`` in UDP. For more details, see the
+documentation for ``loop.create_datagram_endpoint()``.
+(Contributed by Kyle Stanley, Antoine Pitrou, and Yury Selivanov in
+:issue:`37228`.)
