Track when the pulled event signature fails (#13815)

Because we're doing the recording in `_check_sigs_and_hash_for_pulled_events_and_fetch` (previously named `_check_sigs_and_hash_and_fetch`), this means we will track signature failures for `backfill`, `get_room_state`, `get_event_auth`, and `get_missing_events` (all pulled event scenarios). And we also record signature failures from `get_pdu`.

Part of https://github.com/matrix-org/synapse/issues/13700

Part of https://github.com/matrix-org/synapse/issues/13676 and https://github.com/matrix-org/synapse/issues/13356

This PR will be especially important for https://github.com/matrix-org/synapse/pull/13816 so we can avoid the costly `_get_state_ids_after_missing_prev_event` down the line when `/messages` calls backfill.

1 files changed, 40 insertions, 10 deletions

diff --git a/synapse/federation/federation_client.py b/synapse/federation/federation_client.py
index 464672a3da..4dca711cd2 100644
--- a/synapse/federation/federation_client.py
+++ b/synapse/federation/federation_client.py
@@ -278,7 +278,7 @@ class FederationClient(FederationBase):
         pdus = [event_from_pdu_json(p, room_version) for p in transaction_data_pdus]
 
         # Check signatures and hash of pdus, removing any from the list that fail checks
-        pdus[:] = await self._check_sigs_and_hash_and_fetch(
+        pdus[:] = await self._check_sigs_and_hash_for_pulled_events_and_fetch(
             dest, pdus, room_version=room_version
         )
 
@@ -328,7 +328,17 @@ class FederationClient(FederationBase):
 
             # Check signatures are correct.
             try:
-                signed_pdu = await self._check_sigs_and_hash(room_version, pdu)
+
+                async def _record_failure_callback(
+                    event: EventBase, cause: str
+                ) -> None:
+                    await self.store.record_event_failed_pull_attempt(
+                        event.room_id, event.event_id, cause
+                    )
+
+                signed_pdu = await self._check_sigs_and_hash(
+                    room_version, pdu, _record_failure_callback
+                )
             except InvalidEventSignatureError as e:
                 errmsg = f"event id {pdu.event_id}: {e}"
                 logger.warning("%s", errmsg)
@@ -547,24 +557,28 @@ class FederationClient(FederationBase):
             len(auth_event_map),
         )
 
-        valid_auth_events = await self._check_sigs_and_hash_and_fetch(
+        valid_auth_events = await self._check_sigs_and_hash_for_pulled_events_and_fetch(
             destination, auth_event_map.values(), room_version
         )
 
-        valid_state_events = await self._check_sigs_and_hash_and_fetch(
-            destination, state_event_map.values(), room_version
+        valid_state_events = (
+            await self._check_sigs_and_hash_for_pulled_events_and_fetch(
+                destination, state_event_map.values(), room_version
+            )
         )
 
         return valid_state_events, valid_auth_events
 
     @trace
-    async def _check_sigs_and_hash_and_fetch(
+    async def _check_sigs_and_hash_for_pulled_events_and_fetch(
         self,
         origin: str,
         pdus: Collection[EventBase],
         room_version: RoomVersion,
     ) -> List[EventBase]:
-        """Checks the signatures and hashes of a list of events.
+        """
+        Checks the signatures and hashes of a list of pulled events we got from
+        federation and records any signature failures as failed pull attempts.
 
         If a PDU fails its signature check then we check if we have it in
         the database, and if not then request it from the sender's server (if that
@@ -597,11 +611,17 @@ class FederationClient(FederationBase):
 
         valid_pdus: List[EventBase] = []
 
+        async def _record_failure_callback(event: EventBase, cause: str) -> None:
+            await self.store.record_event_failed_pull_attempt(
+                event.room_id, event.event_id, cause
+            )
+
         async def _execute(pdu: EventBase) -> None:
             valid_pdu = await self._check_sigs_and_hash_and_fetch_one(
                 pdu=pdu,
                 origin=origin,
                 room_version=room_version,
+                record_failure_callback=_record_failure_callback,
             )
 
             if valid_pdu:
@@ -618,6 +638,9 @@ class FederationClient(FederationBase):
         pdu: EventBase,
         origin: str,
         room_version: RoomVersion,
+        record_failure_callback: Optional[
+            Callable[[EventBase, str], Awaitable[None]]
+        ] = None,
     ) -> Optional[EventBase]:
         """Takes a PDU and checks its signatures and hashes.
 
@@ -634,6 +657,11 @@ class FederationClient(FederationBase):
             origin
             pdu
             room_version
+            record_failure_callback: A callback to run whenever the given event
+                fails signature or hash checks. This includes exceptions
+                that would be normally be thrown/raised but also things like
+                checking for event tampering where we just return the redacted
+                event.
 
         Returns:
             The PDU (possibly redacted) if it has valid signatures and hashes.
@@ -641,7 +669,9 @@ class FederationClient(FederationBase):
         """
 
         try:
-            return await self._check_sigs_and_hash(room_version, pdu)
+            return await self._check_sigs_and_hash(
+                room_version, pdu, record_failure_callback
+            )
         except InvalidEventSignatureError as e:
             logger.warning(
                 "Signature on retrieved event %s was invalid (%s). "
@@ -694,7 +724,7 @@ class FederationClient(FederationBase):
 
         auth_chain = [event_from_pdu_json(p, room_version) for p in res["auth_chain"]]
 
-        signed_auth = await self._check_sigs_and_hash_and_fetch(
+        signed_auth = await self._check_sigs_and_hash_for_pulled_events_and_fetch(
             destination, auth_chain, room_version=room_version
         )
 
@@ -1401,7 +1431,7 @@ class FederationClient(FederationBase):
                 event_from_pdu_json(e, room_version) for e in content.get("events", [])
             ]
 
-            signed_events = await self._check_sigs_and_hash_and_fetch(
+            signed_events = await self._check_sigs_and_hash_for_pulled_events_and_fetch(
                 destination, events, room_version=room_version
             )
         except HttpResponseException as e: