Room Complexity Client Implementation (#5783)

changelog.d/5783.feature

@@ -0,0 +1 @@
+Synapse can now be configured to not join remote rooms of a given "complexity" (currently, state events) over federation. This option can be used to prevent adverse performance on resource-constrained homeservers.

docs/sample_config.yaml

@@ -278,6 +278,23 @@ listeners:
 # Used by phonehome stats to group together related servers.
 #server_context: context
 
+# Resource-constrained Homeserver Settings
+#
+# If limit_remote_rooms.enabled is True, the room complexity will be
+# checked before a user joins a new remote room. If it is above
+# limit_remote_rooms.complexity, it will disallow joining or
+# instantly leave.
+#
+# limit_remote_rooms.complexity_error can be set to customise the text
+# displayed to the user when a room above the complexity threshold has
+# its join cancelled.
+#
+# Uncomment the below lines to enable:
+#limit_remote_rooms:
+#  enabled: True
+#  complexity: 1.0
+#  complexity_error: "This room is too complex."
+
 # Whether to require a user to be in the room to add an alias to it.
 # Defaults to 'true'.
 #

synapse/config/server.py

@@ -18,6 +18,7 @@
 import logging
 import os.path
 
+import attr
 from netaddr import IPSet
 
 from synapse.api.room_versions import KNOWN_ROOM_VERSIONS
@@ -38,6 +39,12 @@ DEFAULT_BIND_ADDRESSES = ["::", "0.0.0.0"]
 
 DEFAULT_ROOM_VERSION = "4"
 
+ROOM_COMPLEXITY_TOO_GREAT = (
+    "Your homeserver is unable to join rooms this large or complex. "
+    "Please speak to your server administrator, or upgrade your instance "
+    "to join this room."
+)
+
 
 class ServerConfig(Config):
     def read_config(self, config, **kwargs):
@@ -247,6 +254,23 @@ class ServerConfig(Config):
 
         self.gc_thresholds = read_gc_thresholds(config.get("gc_thresholds", None))
 
+        @attr.s
+        class LimitRemoteRoomsConfig(object):
+            enabled = attr.ib(
+                validator=attr.validators.instance_of(bool), default=False
+            )
+            complexity = attr.ib(
+                validator=attr.validators.instance_of((int, float)), default=1.0
+            )
+            complexity_error = attr.ib(
+                validator=attr.validators.instance_of(str),
+                default=ROOM_COMPLEXITY_TOO_GREAT,
+            )
+
+        self.limit_remote_rooms = LimitRemoteRoomsConfig(
+            **config.get("limit_remote_rooms", {})
+        )
+
         bind_port = config.get("bind_port")
         if bind_port:
             if config.get("no_tls", False):
@@ -617,6 +641,23 @@ class ServerConfig(Config):
         # Used by phonehome stats to group together related servers.
         #server_context: context
 
+        # Resource-constrained Homeserver Settings
+        #
+        # If limit_remote_rooms.enabled is True, the room complexity will be
+        # checked before a user joins a new remote room. If it is above
+        # limit_remote_rooms.complexity, it will disallow joining or
+        # instantly leave.
+        #
+        # limit_remote_rooms.complexity_error can be set to customise the text
+        # displayed to the user when a room above the complexity threshold has
+        # its join cancelled.
+        #
+        # Uncomment the below lines to enable:
+        #limit_remote_rooms:
+        #  enabled: True
+        #  complexity: 1.0
+        #  complexity_error: "This room is too complex."
+
         # Whether to require a user to be in the room to add an alias to it.
         # Defaults to 'true'.
         #

synapse/federation/federation_client.py

@@ -993,3 +993,39 @@ class FederationClient(FederationBase):
                 )
 
         raise RuntimeError("Failed to send to any server.")
+
+    @defer.inlineCallbacks
+    def get_room_complexity(self, destination, room_id):
+        """
+        Fetch the complexity of a remote room from another server.
+
+        Args:
+            destination (str): The remote server
+            room_id (str): The room ID to ask about.
+
+        Returns:
+            Deferred[dict] or Deferred[None]: Dict contains the complexity
+            metric versions, while None means we could not fetch the complexity.
+        """
+        try:
+            complexity = yield self.transport_layer.get_room_complexity(
+                destination=destination, room_id=room_id
+            )
+            defer.returnValue(complexity)
+        except CodeMessageException as e:
+            # We didn't manage to get it -- probably a 404. We are okay if other
+            # servers don't give it to us.
+            logger.debug(
+                "Failed to fetch room complexity via %s for %s, got a %d",
+                destination,
+                room_id,
+                e.code,
+            )
+        except Exception:
+            logger.exception(
+                "Failed to fetch room complexity via %s for %s", destination, room_id
+            )
+
+        # If we don't manage to find it, return None. It's not an error if a
+        # server doesn't give it to us.
+        defer.returnValue(None)

synapse/federation/transport/client.py

@@ -21,7 +21,11 @@ from six.moves import urllib
 from twisted.internet import defer
 
 from synapse.api.constants import Membership
-from synapse.api.urls import FEDERATION_V1_PREFIX, FEDERATION_V2_PREFIX
+from synapse.api.urls import (
+    FEDERATION_UNSTABLE_PREFIX,
+    FEDERATION_V1_PREFIX,
+    FEDERATION_V2_PREFIX,
+)
 from synapse.logging.utils import log_function
 
 logger = logging.getLogger(__name__)
@@ -935,6 +939,23 @@ class TransportLayerClient(object):
             destination=destination, path=path, data=content, ignore_backoff=True
         )
 
+    def get_room_complexity(self, destination, room_id):
+        """
+        Args:
+            destination (str): The remote server
+            room_id (str): The room ID to ask about.
+        """
+        path = _create_path(FEDERATION_UNSTABLE_PREFIX, "/rooms/%s/complexity", room_id)
+
+        return self.client.get_json(destination=destination, path=path)
+
+
+def _create_path(federation_prefix, path, *args):
+    """
+    Ensures that all args are url encoded.
+    """
+    return federation_prefix + path % tuple(urllib.parse.quote(arg, "") for arg in args)
+
 
 def _create_v1_path(path, *args):
     """Creates a path against V1 federation API from the path template and
@@ -951,9 +972,7 @@ def _create_v1_path(path, *args):
     Returns:
         str
     """
-    return FEDERATION_V1_PREFIX + path % tuple(
-        urllib.parse.quote(arg, "") for arg in args
-    )
+    return _create_path(FEDERATION_V1_PREFIX, path, *args)
 
 
 def _create_v2_path(path, *args):
@@ -971,6 +990,4 @@ def _create_v2_path(path, *args):
     Returns:
         str
     """
-    return FEDERATION_V2_PREFIX + path % tuple(
-        urllib.parse.quote(arg, "") for arg in args
-    )
+    return _create_path(FEDERATION_V2_PREFIX, path, *args)

synapse/handlers/federation.py

@@ -2796,3 +2796,28 @@ class FederationHandler(BaseHandler):
             )
         else:
             return user_joined_room(self.distributor, user, room_id)
+
+    @defer.inlineCallbacks
+    def get_room_complexity(self, remote_room_hosts, room_id):
+        """
+        Fetch the complexity of a remote room over federation.
+
+        Args:
+            remote_room_hosts (list[str]): The remote servers to ask.
+            room_id (str): The room ID to ask about.
+
+        Returns:
+            Deferred[dict] or Deferred[None]: Dict contains the complexity
+            metric versions, while None means we could not fetch the complexity.
+        """
+
+        for host in remote_room_hosts:
+            res = yield self.federation_client.get_room_complexity(host, room_id)
+
+            # We got a result, return it.
+            if res:
+                defer.returnValue(res)
+
+        # We fell off the bottom, couldn't get the complexity from anyone. Oh
+        # well.
+        defer.returnValue(None)

synapse/handlers/room_member.py

@@ -26,8 +26,7 @@ from unpaddedbase64 import decode_base64
 
 from twisted.internet import defer
 
-import synapse.server
-import synapse.types
+from synapse import types
 from synapse.api.constants import EventTypes, Membership
 from synapse.api.errors import AuthError, Codes, HttpResponseException, SynapseError
 from synapse.types import RoomID, UserID
@@ -543,7 +542,7 @@ class RoomMemberHandler(object):
             ), "Sender (%s) must be same as requester (%s)" % (sender, requester.user)
             assert self.hs.is_mine(sender), "Sender must be our own: %s" % (sender,)
         else:
-            requester = synapse.types.create_requester(target_user)
+            requester = types.create_requester(target_user)
 
         prev_event = yield self.event_creation_handler.deduplicate_state_event(
             event, context
@@ -945,6 +944,47 @@ class RoomMemberMasterHandler(RoomMemberHandler):
         self.distributor.declare("user_joined_room")
         self.distributor.declare("user_left_room")
 
+    @defer.inlineCallbacks
+    def _is_remote_room_too_complex(self, room_id, remote_room_hosts):
+        """
+        Check if complexity of a remote room is too great.
+
+        Args:
+            room_id (str)
+            remote_room_hosts (list[str])
+
+        Returns: bool of whether the complexity is too great, or None
+            if unable to be fetched
+        """
+        max_complexity = self.hs.config.limit_remote_rooms.complexity
+        complexity = yield self.federation_handler.get_room_complexity(
+            remote_room_hosts, room_id
+        )
+
+        if complexity:
+            if complexity["v1"] > max_complexity:
+                return True
+            return False
+        return None
+
+    @defer.inlineCallbacks
+    def _is_local_room_too_complex(self, room_id):
+        """
+        Check if the complexity of a local room is too great.
+
+        Args:
+            room_id (str)
+
+        Returns: bool
+        """
+        max_complexity = self.hs.config.limit_remote_rooms.complexity
+        complexity = yield self.store.get_room_complexity(room_id)
+
+        if complexity["v1"] > max_complexity:
+            return True
+
+        return False
+
     @defer.inlineCallbacks
     def _remote_join(self, requester, remote_room_hosts, room_id, user, content):
         """Implements RoomMemberHandler._remote_join
@@ -952,7 +992,6 @@ class RoomMemberMasterHandler(RoomMemberHandler):
         # filter ourselves out of remote_room_hosts: do_invite_join ignores it
         # and if it is the only entry we'd like to return a 404 rather than a
         # 500.
-
         remote_room_hosts = [
             host for host in remote_room_hosts if host != self.hs.hostname
         ]
@@ -960,6 +999,18 @@ class RoomMemberMasterHandler(RoomMemberHandler):
         if len(remote_room_hosts) == 0:
             raise SynapseError(404, "No known servers")
 
+        if self.hs.config.limit_remote_rooms.enabled:
+            # Fetch the room complexity
+            too_complex = yield self._is_remote_room_too_complex(
+                room_id, remote_room_hosts
+            )
+            if too_complex is True:
+                raise SynapseError(
+                    code=400,
+                    msg=self.hs.config.limit_remote_rooms.complexity_error,
+                    errcode=Codes.RESOURCE_LIMIT_EXCEEDED,
+                )
+
         # We don't do an auth check if we are doing an invite
         # join dance for now, since we're kinda implicitly checking
         # that we are allowed to join when we decide whether or not we
@@ -969,6 +1020,31 @@ class RoomMemberMasterHandler(RoomMemberHandler):
         )
         yield self._user_joined_room(user, room_id)
 
+        # Check the room we just joined wasn't too large, if we didn't fetch the
+        # complexity of it before.
+        if self.hs.config.limit_remote_rooms.enabled:
+            if too_complex is False:
+                # We checked, and we're under the limit.
+                return
+
+            # Check again, but with the local state events
+            too_complex = yield self._is_local_room_too_complex(room_id)
+
+            if too_complex is False:
+                # We're under the limit.
+                return
+
+            # The room is too large. Leave.
+            requester = types.create_requester(user, None, False, None)
+            yield self.update_membership(
+                requester=requester, target=user, room_id=room_id, action="leave"
+            )
+            raise SynapseError(
+                code=400,
+                msg=self.hs.config.limit_remote_rooms.complexity_error,
+                errcode=Codes.RESOURCE_LIMIT_EXCEEDED,
+            )
+
     @defer.inlineCallbacks
     def _remote_reject_invite(self, requester, remote_room_hosts, room_id, target):
         """Implements RoomMemberHandler._remote_reject_invite

tests/federation/test_complexity.py

@@ -13,12 +13,16 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+from mock import Mock
+
 from twisted.internet import defer
 
+from synapse.api.errors import Codes, SynapseError
 from synapse.config.ratelimiting import FederationRateLimitConfig
 from synapse.federation.transport import server
 from synapse.rest import admin
 from synapse.rest.client.v1 import login, room
+from synapse.types import UserID
 from synapse.util.ratelimitutils import FederationRateLimiter
 
 from tests import unittest
@@ -33,9 +37,8 @@ class RoomComplexityTests(unittest.HomeserverTestCase):
     ]
 
     def default_config(self, name="test"):
-        config = super(RoomComplexityTests, self).default_config(name=name)
-        config["limit_large_remote_room_joins"] = True
-        config["limit_large_remote_room_complexity"] = 0.05
+        config = super().default_config(name=name)
+        config["limit_remote_rooms"] = {"enabled": True, "complexity": 0.05}
         return config
 
     def prepare(self, reactor, clock, homeserver):
@@ -88,3 +91,71 @@ class RoomComplexityTests(unittest.HomeserverTestCase):
         self.assertEquals(200, channel.code)
         complexity = channel.json_body["v1"]
         self.assertEqual(complexity, 1.23)
+
+    def test_join_too_large(self):
+
+        u1 = self.register_user("u1", "pass")
+
+        handler = self.hs.get_room_member_handler()
+        fed_transport = self.hs.get_federation_transport_client()
+
+        # Mock out some things, because we don't want to test the whole join
+        fed_transport.client.get_json = Mock(return_value=defer.succeed({"v1": 9999}))
+        handler.federation_handler.do_invite_join = Mock(return_value=defer.succeed(1))
+
+        d = handler._remote_join(
+            None,
+            ["otherserver.example"],
+            "roomid",
+            UserID.from_string(u1),
+            {"membership": "join"},
+        )
+
+        self.pump()
+
+        # The request failed with a SynapseError saying the resource limit was
+        # exceeded.
+        f = self.get_failure(d, SynapseError)
+        self.assertEqual(f.value.code, 400, f.value)
+        self.assertEqual(f.value.errcode, Codes.RESOURCE_LIMIT_EXCEEDED)
+
+    def test_join_too_large_once_joined(self):
+
+        u1 = self.register_user("u1", "pass")
+        u1_token = self.login("u1", "pass")
+
+        # Ok, this might seem a bit weird -- I want to test that we actually
+        # leave the room, but I don't want to simulate two servers. So, we make
+        # a local room, which we say we're joining remotely, even if there's no
+        # remote, because we mock that out. Then, we'll leave the (actually
+        # local) room, which will be propagated over federation in a real
+        # scenario.
+        room_1 = self.helper.create_room_as(u1, tok=u1_token)
+
+        handler = self.hs.get_room_member_handler()
+        fed_transport = self.hs.get_federation_transport_client()
+
+        # Mock out some things, because we don't want to test the whole join
+        fed_transport.client.get_json = Mock(return_value=defer.succeed(None))
+        handler.federation_handler.do_invite_join = Mock(return_value=defer.succeed(1))
+
+        # Artificially raise the complexity
+        self.hs.get_datastore().get_current_state_event_counts = lambda x: defer.succeed(
+            600
+        )
+
+        d = handler._remote_join(
+            None,
+            ["otherserver.example"],
+            room_1,
+            UserID.from_string(u1),
+            {"membership": "join"},
+        )
+
+        self.pump()
+
+        # The request failed with a SynapseError saying the resource limit was
+        # exceeded.
+        f = self.get_failure(d, SynapseError)
+        self.assertEqual(f.value.code, 400)
+        self.assertEqual(f.value.errcode, Codes.RESOURCE_LIMIT_EXCEEDED)