Back off from more events, don't retry auth events

src/service/rooms/event_handler/mod.rs

@@ -119,7 +119,15 @@ impl Service {
             .ok_or_else(|| Error::bad_database("Failed to find first pdu in db."))?;
 
         let (incoming_pdu, val) = self
-            .handle_outlier_pdu(origin, &create_event, event_id, room_id, value, pub_key_map)
+            .handle_outlier_pdu(
+                origin,
+                &create_event,
+                event_id,
+                room_id,
+                value,
+                false,
+                pub_key_map,
+            )
             .await?;
         self.check_room_id(room_id, &incoming_pdu)?;
 
@@ -276,6 +284,7 @@ impl Service {
         event_id: &'a EventId,
         room_id: &'a RoomId,
         mut value: BTreeMap<String, CanonicalJsonValue>,
+        auth_events_known: bool,
         pub_key_map: &'a RwLock<BTreeMap<String, BTreeMap<String, Base64>>>,
     ) -> AsyncRecursiveType<'a, Result<(Arc<PduEvent>, BTreeMap<String, CanonicalJsonValue>)>> {
         Box::pin(async move {
@@ -343,23 +352,25 @@ impl Service {
 
             self.check_room_id(room_id, &incoming_pdu)?;
 
-            // 4. fetch any missing auth events doing all checks listed here starting at 1. These are not timeline events
+            if !auth_events_known {
-            // 5. Reject "due to auth events" if can't get all the auth events or some of the auth events are also rejected "due to auth events"
+                // 4. fetch any missing auth events doing all checks listed here starting at 1. These are not timeline events
-            // NOTE: Step 5 is not applied anymore because it failed too often
+                // 5. Reject "due to auth events" if can't get all the auth events or some of the auth events are also rejected "due to auth events"
-            debug!(event_id = ?incoming_pdu.event_id, "Fetching auth events");
+                // NOTE: Step 5 is not applied anymore because it failed too often
-            self.fetch_and_handle_outliers(
+                debug!(event_id = ?incoming_pdu.event_id, "Fetching auth events");
-                origin,
+                self.fetch_and_handle_outliers(
-                &incoming_pdu
+                    origin,
-                    .auth_events
+                    &incoming_pdu
-                    .iter()
+                        .auth_events
-                    .map(|x| Arc::from(&**x))
+                        .iter()
-                    .collect::<Vec<_>>(),
+                        .map(|x| Arc::from(&**x))
-                create_event,
+                        .collect::<Vec<_>>(),
-                room_id,
+                    create_event,
-                room_version_id,
+                    room_id,
-                pub_key_map,
+                    room_version_id,
-            )
+                    pub_key_map,
-            .await;
+                )
+                .await;
+            }
 
             // 6. Reject "due to auth events" if the event doesn't pass auth based on the auth events
             debug!(
@@ -1015,26 +1026,6 @@ impl Service {
 
             let mut pdus = vec![];
             for id in events {
-                if let Some((time, tries)) = services()
-                    .globals
-                    .bad_event_ratelimiter
-                    .read()
-                    .unwrap()
-                    .get(&**id)
-                {
-                    // Exponential backoff
-                    let mut min_elapsed_duration =
-                        Duration::from_secs(5 * 60) * (*tries) * (*tries);
-                    if min_elapsed_duration > Duration::from_secs(60 * 60 * 24) {
-                        min_elapsed_duration = Duration::from_secs(60 * 60 * 24);
-                    }
-
-                    if time.elapsed() < min_elapsed_duration {
-                        info!("Backing off from {}", id);
-                        continue;
-                    }
-                }
-
                 // a. Look in the main timeline (pduid_pdu tree)
                 // b. Look at outlier pdu tree
                 // (get_pdu_json checks both)
@@ -1052,6 +1043,26 @@ impl Service {
                 let mut events_all = HashSet::new();
                 let mut i = 0;
                 while let Some(next_id) = todo_auth_events.pop() {
+                    if let Some((time, tries)) = services()
+                        .globals
+                        .bad_event_ratelimiter
+                        .read()
+                        .unwrap()
+                        .get(&*next_id)
+                    {
+                        // Exponential backoff
+                        let mut min_elapsed_duration =
+                            Duration::from_secs(5 * 60) * (*tries) * (*tries);
+                        if min_elapsed_duration > Duration::from_secs(60 * 60 * 24) {
+                            min_elapsed_duration = Duration::from_secs(60 * 60 * 24);
+                        }
+
+                        if time.elapsed() < min_elapsed_duration {
+                            info!("Backing off from {}", next_id);
+                            continue;
+                        }
+                    }
+
                     if events_all.contains(&next_id) {
                         continue;
                     }
@@ -1062,7 +1073,7 @@ impl Service {
                     }
 
                     if let Ok(Some(_)) = services().rooms.timeline.get_pdu(&next_id) {
-                        trace!("Found {} in db", id);
+                        trace!("Found {} in db", next_id);
                         continue;
                     }
 
@@ -1121,6 +1132,26 @@ impl Service {
                 }
 
                 for (next_id, value) in events_in_reverse_order.iter().rev() {
+                    if let Some((time, tries)) = services()
+                        .globals
+                        .bad_event_ratelimiter
+                        .read()
+                        .unwrap()
+                        .get(&**next_id)
+                    {
+                        // Exponential backoff
+                        let mut min_elapsed_duration =
+                            Duration::from_secs(5 * 60) * (*tries) * (*tries);
+                        if min_elapsed_duration > Duration::from_secs(60 * 60 * 24) {
+                            min_elapsed_duration = Duration::from_secs(60 * 60 * 24);
+                        }
+
+                        if time.elapsed() < min_elapsed_duration {
+                            info!("Backing off from {}", next_id);
+                            continue;
+                        }
+                    }
+
                     match self
                         .handle_outlier_pdu(
                             origin,
@@ -1128,6 +1159,7 @@ impl Service {
                             next_id,
                             room_id,
                             value.clone(),
+                            true,
                             pub_key_map,
                         )
                         .await