Add support for MSC2716 marker events (#10498)

* Make historical messages available to federated servers

Part of MSC2716: https://github.com/matrix-org/matrix-doc/pull/2716

Follow-up to https://github.com/matrix-org/synapse/pull/9247

* Debug message not available on federation

* Add base starting insertion point when no chunk ID is provided

* Fix messages from multiple senders in historical chunk

Follow-up to https://github.com/matrix-org/synapse/pull/9247

Part of MSC2716: https://github.com/matrix-org/matrix-doc/pull/2716

---

Previously, Synapse would throw a 403,
`Cannot force another user to join.`,
because we were trying to use `?user_id` from a single virtual user
which did not match with messages from other users in the chunk.

* Remove debug lines

* Messing with selecting insertion event extremeties

* Move db schema change to new version

* Add more better comments

* Make a fake requester with just what we need

See https://github.com/matrix-org/synapse/pull/10276#discussion_r660999080

* Store insertion events in table

* Make base insertion event float off on its own

See https://github.com/matrix-org/synapse/pull/10250#issuecomment-875711889

Conflicts:
	synapse/rest/client/v1/room.py

* Validate that the app service can actually control the given user

See https://github.com/matrix-org/synapse/pull/10276#issuecomment-876316455

Conflicts:
	synapse/rest/client/v1/room.py

* Add some better comments on what we're trying to check for

* Continue debugging

* Share validation logic

* Add inserted historical messages to /backfill response

* Remove debug sql queries

* Some marker event implemntation trials

* Clean up PR

* Rename insertion_event_id to just event_id

* Add some better sql comments

* More accurate description

* Add changelog

* Make it clear what MSC the change is part of

* Add more detail on which insertion event came through

* Address review and improve sql queries

* Only use event_id as unique constraint

* Fix test case where insertion event is already in the normal DAG

* Remove debug changes

* Add support for MSC2716 marker events

* Process markers when we receive it over federation

* WIP: make hs2 backfill historical messages after marker event

* hs2 to better ask for insertion event extremity

But running into the `sqlite3.IntegrityError: NOT NULL constraint failed: event_to_state_groups.state_group`
error

* Add insertion_event_extremities table

* Switch to chunk events so we can auth via power_levels

Previously, we were using `content.chunk_id` to connect one
chunk to another. But these events can be from any `sender`
and we can't tell who should be able to send historical events.
We know we only want the application service to do it but these
events have the sender of a real historical message, not the
application service user ID as the sender. Other federated homeservers
also have no indicator which senders are an application service on
the originating homeserver.

So we want to auth all of the MSC2716 events via power_levels
and have them be sent by the application service with proper
PL levels in the room.

* Switch to chunk events for federation

* Add unstable room version to support new historical PL

* Messy: Fix undefined state_group for federated historical events

```
2021-07-13 02:27:57,810 - synapse.handlers.federation - 1248 - ERROR - GET-4 - Failed to backfill from hs1 because NOT NULL constraint failed: event_to_state_groups.state_group
Traceback (most recent call last):
  File "/usr/local/lib/python3.8/site-packages/synapse/handlers/federation.py", line 1216, in try_backfill
    await self.backfill(
  File "/usr/local/lib/python3.8/site-packages/synapse/handlers/federation.py", line 1035, in backfill
    await self._auth_and_persist_event(dest, event, context, backfilled=True)
  File "/usr/local/lib/python3.8/site-packages/synapse/handlers/federation.py", line 2222, in _auth_and_persist_event
    await self._run_push_actions_and_persist_event(event, context, backfilled)
  File "/usr/local/lib/python3.8/site-packages/synapse/handlers/federation.py", line 2244, in _run_push_actions_and_persist_event
    await self.persist_events_and_notify(
  File "/usr/local/lib/python3.8/site-packages/synapse/handlers/federation.py", line 3290, in persist_events_and_notify
    events, max_stream_token = await self.storage.persistence.persist_events(
  File "/usr/local/lib/python3.8/site-packages/synapse/logging/opentracing.py", line 774, in _trace_inner
    return await func(*args, **kwargs)
  File "/usr/local/lib/python3.8/site-packages/synapse/storage/persist_events.py", line 320, in persist_events
    ret_vals = await yieldable_gather_results(enqueue, partitioned.items())
  File "/usr/local/lib/python3.8/site-packages/synapse/storage/persist_events.py", line 237, in handle_queue_loop
    ret = await self._per_item_callback(
  File "/usr/local/lib/python3.8/site-packages/synapse/storage/persist_events.py", line 577, in _persist_event_batch
    await self.persist_events_store._persist_events_and_state_updates(
  File "/usr/local/lib/python3.8/site-packages/synapse/storage/databases/main/events.py", line 176, in _persist_events_and_state_updates
    await self.db_pool.runInteraction(
  File "/usr/local/lib/python3.8/site-packages/synapse/storage/database.py", line 681, in runInteraction
    result = await self.runWithConnection(
  File "/usr/local/lib/python3.8/site-packages/synapse/storage/database.py", line 770, in runWithConnection
    return await make_deferred_yieldable(
  File "/usr/local/lib/python3.8/site-packages/twisted/python/threadpool.py", line 238, in inContext
    result = inContext.theWork()  # type: ignore[attr-defined]
  File "/usr/local/lib/python3.8/site-packages/twisted/python/threadpool.py", line 254, in <lambda>
    inContext.theWork = lambda: context.call(  # type: ignore[attr-defined]
  File "/usr/local/lib/python3.8/site-packages/twisted/python/context.py", line 118, in callWithContext
    return self.currentContext().callWithContext(ctx, func, *args, **kw)
  File "/usr/local/lib/python3.8/site-packages/twisted/python/context.py", line 83, in callWithContext
    return func(*args, **kw)
  File "/usr/local/lib/python3.8/site-packages/twisted/enterprise/adbapi.py", line 293, in _runWithConnection
    compat.reraise(excValue, excTraceback)
  File "/usr/local/lib/python3.8/site-packages/twisted/python/deprecate.py", line 298, in deprecatedFunction
    return function(*args, **kwargs)
  File "/usr/local/lib/python3.8/site-packages/twisted/python/compat.py", line 403, in reraise
    raise exception.with_traceback(traceback)
  File "/usr/local/lib/python3.8/site-packages/twisted/enterprise/adbapi.py", line 284, in _runWithConnection
    result = func(conn, *args, **kw)
  File "/usr/local/lib/python3.8/site-packages/synapse/storage/database.py", line 765, in inner_func
    return func(db_conn, *args, **kwargs)
  File "/usr/local/lib/python3.8/site-packages/synapse/storage/database.py", line 549, in new_transaction
    r = func(cursor, *args, **kwargs)
  File "/usr/local/lib/python3.8/site-packages/synapse/logging/utils.py", line 69, in wrapped
    return f(*args, **kwargs)
  File "/usr/local/lib/python3.8/site-packages/synapse/storage/databases/main/events.py", line 385, in _persist_events_txn
    self._store_event_state_mappings_txn(txn, events_and_contexts)
  File "/usr/local/lib/python3.8/site-packages/synapse/storage/databases/main/events.py", line 2065, in _store_event_state_mappings_txn
    self.db_pool.simple_insert_many_txn(
  File "/usr/local/lib/python3.8/site-packages/synapse/storage/database.py", line 923, in simple_insert_many_txn
    txn.execute_batch(sql, vals)
  File "/usr/local/lib/python3.8/site-packages/synapse/storage/database.py", line 280, in execute_batch
    self.executemany(sql, args)
  File "/usr/local/lib/python3.8/site-packages/synapse/storage/database.py", line 300, in executemany
    self._do_execute(self.txn.executemany, sql, *args)
  File "/usr/local/lib/python3.8/site-packages/synapse/storage/database.py", line 330, in _do_execute
    return func(sql, *args)
sqlite3.IntegrityError: NOT NULL constraint failed: event_to_state_groups.state_group
```

* Revert "Messy: Fix undefined state_group for federated historical events"

This reverts commit 187ab28611546321e02770944c86f30ee2bc742a.

* Fix federated events being rejected for no state_groups

Add fix from https://github.com/matrix-org/synapse/pull/10439
until it merges.

* Adapting to experimental room version

* Some log cleanup

* Add better comments around extremity fetching code and why

* Rename to be more accurate to what the function returns

* Add changelog

* Ignore rejected events

* Use simplified upsert

* Add Erik's explanation of extra event checks

See https://github.com/matrix-org/synapse/pull/10498#discussion_r680880332

* Clarify that the depth is not directly correlated to the backwards extremity that we return

See https://github.com/matrix-org/synapse/pull/10498#discussion_r681725404

* lock only matters for sqlite

See https://github.com/matrix-org/synapse/pull/10498#discussion_r681728061

* Move new SQL changes to its own delta file

* Clean up upsert docstring

* Bump database schema version (62)

2 files changed, 118 insertions, 20 deletions

diff --git a/synapse/storage/databases/main/event_federation.py b/synapse/storage/databases/main/event_federation.py
index 44018c1c31..bddf5ef192 100644
--- a/synapse/storage/databases/main/event_federation.py
+++ b/synapse/storage/databases/main/event_federation.py
@@ -671,27 +671,97 @@ class EventFederationWorkerStore(EventsWorkerStore, SignatureWorkerStore, SQLBas
         # Return all events where not all sets can reach them.
         return {eid for eid, n in event_to_missing_sets.items() if n}
 
-    async def get_oldest_events_with_depth_in_room(self, room_id):
+    async def get_oldest_event_ids_with_depth_in_room(self, room_id) -> Dict[str, int]:
+        """Gets the oldest events(backwards extremities) in the room along with the
+        aproximate depth.
+
+        We use this function so that we can compare and see if someones current
+        depth at their current scrollback is within pagination range of the
+        event extremeties. If the current depth is close to the depth of given
+        oldest event, we can trigger a backfill.
+
+        Args:
+            room_id: Room where we want to find the oldest events
+
+        Returns:
+            Map from event_id to depth
+        """
+
+        def get_oldest_event_ids_with_depth_in_room_txn(txn, room_id):
+            # Assemble a dictionary with event_id -> depth for the oldest events
+            # we know of in the room. Backwards extremeties are the oldest
+            # events we know of in the room but we only know of them because
+            # some other event referenced them by prev_event and aren't peristed
+            # in our database yet (meaning we don't know their depth
+            # specifically). So we need to look for the aproximate depth from
+            # the events connected to the current backwards extremeties.
+            sql = """
+                SELECT b.event_id, MAX(e.depth) FROM events as e
+                /**
+                 * Get the edge connections from the event_edges table
+                 * so we can see whether this event's prev_events points
+                 * to a backward extremity in the next join.
+                 */
+                INNER JOIN event_edges as g
+                ON g.event_id = e.event_id
+                /**
+                 * We find the "oldest" events in the room by looking for
+                 * events connected to backwards extremeties (oldest events
+                 * in the room that we know of so far).
+                 */
+                INNER JOIN event_backward_extremities as b
+                ON g.prev_event_id = b.event_id
+                WHERE b.room_id = ? AND g.is_state is ?
+                GROUP BY b.event_id
+            """
+
+            txn.execute(sql, (room_id, False))
+
+            return dict(txn)
+
         return await self.db_pool.runInteraction(
-            "get_oldest_events_with_depth_in_room",
-            self.get_oldest_events_with_depth_in_room_txn,
+            "get_oldest_event_ids_with_depth_in_room",
+            get_oldest_event_ids_with_depth_in_room_txn,
             room_id,
         )
 
-    def get_oldest_events_with_depth_in_room_txn(self, txn, room_id):
-        sql = (
-            "SELECT b.event_id, MAX(e.depth) FROM events as e"
-            " INNER JOIN event_edges as g"
-            " ON g.event_id = e.event_id"
-            " INNER JOIN event_backward_extremities as b"
-            " ON g.prev_event_id = b.event_id"
-            " WHERE b.room_id = ? AND g.is_state is ?"
-            " GROUP BY b.event_id"
-        )
+    async def get_insertion_event_backwards_extremities_in_room(
+        self, room_id
+    ) -> Dict[str, int]:
+        """Get the insertion events we know about that we haven't backfilled yet.
 
-        txn.execute(sql, (room_id, False))
+        We use this function so that we can compare and see if someones current
+        depth at their current scrollback is within pagination range of the
+        insertion event. If the current depth is close to the depth of given
+        insertion event, we can trigger a backfill.
 
-        return dict(txn)
+        Args:
+            room_id: Room where we want to find the oldest events
+
+        Returns:
+            Map from event_id to depth
+        """
+
+        def get_insertion_event_backwards_extremities_in_room_txn(txn, room_id):
+            sql = """
+                SELECT b.event_id, MAX(e.depth) FROM insertion_events as i
+                /* We only want insertion events that are also marked as backwards extremities */
+                INNER JOIN insertion_event_extremities as b USING (event_id)
+                /* Get the depth of the insertion event from the events table */
+                INNER JOIN events AS e USING (event_id)
+                WHERE b.room_id = ?
+                GROUP BY b.event_id
+            """
+
+            txn.execute(sql, (room_id,))
+
+            return dict(txn)
+
+        return await self.db_pool.runInteraction(
+            "get_insertion_event_backwards_extremities_in_room",
+            get_insertion_event_backwards_extremities_in_room_txn,
+            room_id,
+        )
 
     async def get_max_depth_of(self, event_ids: List[str]) -> Tuple[str, int]:
         """Returns the event ID and depth for the event that has the max depth from a set of event IDs
@@ -1041,7 +1111,6 @@ class EventFederationWorkerStore(EventsWorkerStore, SignatureWorkerStore, SQLBas
                     if row[1] not in event_results:
                         queue.put((-row[0], row[1]))
 
-            # Navigate up the DAG by prev_event
             txn.execute(query, (event_id, False, limit - len(event_results)))
             prev_event_id_results = txn.fetchall()
             logger.debug(
@@ -1136,6 +1205,19 @@ class EventFederationWorkerStore(EventsWorkerStore, SignatureWorkerStore, SQLBas
             _delete_old_forward_extrem_cache_txn,
         )
 
+    async def insert_insertion_extremity(self, event_id: str, room_id: str) -> None:
+        await self.db_pool.simple_upsert(
+            table="insertion_event_extremities",
+            keyvalues={"event_id": event_id},
+            values={
+                "event_id": event_id,
+                "room_id": room_id,
+            },
+            insertion_values={},
+            desc="insert_insertion_extremity",
+            lock=False,
+        )
+
     async def insert_received_event_to_staging(
         self, origin: str, event: EventBase
     ) -> None:
diff --git a/synapse/storage/databases/main/events.py b/synapse/storage/databases/main/events.py
index 86baf397fb..40b53274fb 100644
--- a/synapse/storage/databases/main/events.py
+++ b/synapse/storage/databases/main/events.py
@@ -1845,6 +1845,18 @@ class PersistEventsStore:
             },
         )
 
+        # When we receive an event with a `chunk_id` referencing the
+        # `next_chunk_id` of the insertion event, we can remove it from the
+        # `insertion_event_extremities` table.
+        sql = """
+            DELETE FROM insertion_event_extremities WHERE event_id IN (
+                SELECT event_id FROM insertion_events
+                WHERE next_chunk_id = ?
+            )
+        """
+
+        txn.execute(sql, (chunk_id,))
+
     def _handle_redaction(self, txn, redacted_event_id):
         """Handles receiving a redaction and checking whether we need to remove
         any redacted relations from the database.
@@ -2101,15 +2113,17 @@ class PersistEventsStore:
 
         Forward extremities are handled when we first start persisting the events.
         """
+        # From the events passed in, add all of the prev events as backwards extremities.
+        # Ignore any events that are already backwards extrems or outliers.
         query = (
             "INSERT INTO event_backward_extremities (event_id, room_id)"
             " SELECT ?, ? WHERE NOT EXISTS ("
-            " SELECT 1 FROM event_backward_extremities"
-            " WHERE event_id = ? AND room_id = ?"
+            "   SELECT 1 FROM event_backward_extremities"
+            "   WHERE event_id = ? AND room_id = ?"
             " )"
             " AND NOT EXISTS ("
-            " SELECT 1 FROM events WHERE event_id = ? AND room_id = ? "
-            " AND outlier = ?"
+            "   SELECT 1 FROM events WHERE event_id = ? AND room_id = ? "
+            "   AND outlier = ?"
             " )"
         )
 
@@ -2123,6 +2137,8 @@ class PersistEventsStore:
             ],
         )
 
+        # Delete all these events that we've already fetched and now know that their
+        # prev events are the new backwards extremeties.
         query = (
             "DELETE FROM event_backward_extremities"
             " WHERE event_id = ? AND room_id = ?"