SS: Reset connection if token is unrecognized (#17529)

This triggers the client to start a new sliding sync connection. If we
don't do this and the client asks for the full range of rooms, we end up
sending down all rooms and their state from scratch (which can be very
slow)

This causes things like
https://github.com/element-hq/element-x-ios/issues/3115 after we restart
the server

---------

Co-authored-by: Eric Eastwood <eric.eastwood@beta.gouv.fr>

2 files changed, 45 insertions, 0 deletions

diff --git a/synapse/api/errors.py b/synapse/api/errors.py
index dd4a1ae706..99fc7eab54 100644
--- a/synapse/api/errors.py
+++ b/synapse/api/errors.py
@@ -128,6 +128,10 @@ class Codes(str, Enum):
     # MSC2677
     DUPLICATE_ANNOTATION = "M_DUPLICATE_ANNOTATION"
 
+    # MSC3575 we are telling the client they need to reset their sliding sync
+    # connection.
+    UNKNOWN_POS = "M_UNKNOWN_POS"
+
 
 class CodeMessageException(RuntimeError):
     """An exception with integer code, a message string attributes and optional headers.
@@ -847,3 +851,17 @@ class PartialStateConflictError(SynapseError):
             msg=PartialStateConflictError.message(),
             errcode=Codes.UNKNOWN,
         )
+
+
+class SlidingSyncUnknownPosition(SynapseError):
+    """An error that Synapse can return to signal to the client to expire their
+    sliding sync connection (i.e. send a new request without a `?since=`
+    param).
+    """
+
+    def __init__(self) -> None:
+        super().__init__(
+            HTTPStatus.BAD_REQUEST,
+            msg="Unknown position",
+            errcode=Codes.UNKNOWN_POS,
+        )
diff --git a/synapse/handlers/sliding_sync.py b/synapse/handlers/sliding_sync.py
index 8467766518..1936471345 100644
--- a/synapse/handlers/sliding_sync.py
+++ b/synapse/handlers/sliding_sync.py
@@ -47,6 +47,7 @@ from synapse.api.constants import (
     EventTypes,
     Membership,
 )
+from synapse.api.errors import SlidingSyncUnknownPosition
 from synapse.events import EventBase, StrippedStateEvent
 from synapse.events.utils import parse_stripped_state_event, strip_event
 from synapse.handlers.relations import BundledAggregations
@@ -491,6 +492,22 @@ class SlidingSyncHandler:
             # See https://github.com/matrix-org/matrix-doc/issues/1144
             raise NotImplementedError()
 
+        if from_token:
+            # Check that we recognize the connection position, if not tell the
+            # clients that they need to start again.
+            #
+            # If we don't do this and the client asks for the full range of
+            # rooms, we end up sending down all rooms and their state from
+            # scratch (which can be very slow). By expiring the connection we
+            # allow the client a chance to do an initial request with a smaller
+            # range of rooms to get them some results sooner but will end up
+            # taking the same amount of time (more with round-trips and
+            # re-processing) in the end to get everything again.
+            if not await self.connection_store.is_valid_token(
+                sync_config, from_token.connection_position
+            ):
+                raise SlidingSyncUnknownPosition()
+
         await self.connection_store.mark_token_seen(
             sync_config=sync_config,
             from_token=from_token,
@@ -2821,6 +2838,16 @@ class SlidingSyncConnectionStore:
         attr.Factory(dict)
     )
 
+    async def is_valid_token(
+        self, sync_config: SlidingSyncConfig, connection_token: int
+    ) -> bool:
+        """Return whether the connection token is valid/recognized"""
+        if connection_token == 0:
+            return True
+
+        conn_key = self._get_connection_key(sync_config)
+        return connection_token in self._connections.get(conn_key, {})
+
     async def have_sent_room(
         self, sync_config: SlidingSyncConfig, connection_token: int, room_id: str
     ) -> HaveSentRoom: