Add MSC4040 matrix-fed service lookups

changelog.d/16137.feature

@@ -0,0 +1 @@
+Support resolving homeservers using `matrix-fed` DNS SRV records from [MSC4040](https://github.com/matrix-org/matrix-spec-proposals/pull/4040).

scripts-dev/federation_client.py

@@ -329,15 +329,26 @@ def _lookup(server_name: str) -> Tuple[str, int, str]:
                 raise ValueError("Invalid host:port '%s'" % (server_name,))
             return out[0], port, out[0]
 
+        # Look up SRV for Matrix 1.8 `matrix-fed` service first
         try:
-            srv = srvlookup.lookup("matrix", "tcp", server_name)[0]
+            srv = srvlookup.lookup("matrix-fed", "tcp", server_name)[0]
             print(
-                f"SRV lookup on _matrix._tcp.{server_name} gave {srv}",
+                f"SRV lookup on _matrix-fed._tcp.{server_name} gave {srv}",
                 file=sys.stderr,
             )
             return srv.host, srv.port, server_name
         except Exception:
-            return server_name, 8448, server_name
+            # Fall back to deprecated `matrix` service
+            try:
+                srv = srvlookup.lookup("matrix", "tcp", server_name)[0]
+                print(
+                    f"SRV lookup on _matrix._tcp.{server_name} gave {srv}",
+                    file=sys.stderr,
+                )
+                return srv.host, srv.port, server_name
+            except Exception:
+                # Fall even further back to just port 8448
+                return server_name, 8448, server_name
 
     @staticmethod
     def _get_well_known(server_name: str) -> Optional[str]:

synapse/http/federation/matrix_federation_agent.py

@@ -399,8 +399,11 @@ async def _resolve_server(self) -> List[Server]:
         if port or _is_ip_literal(host):
             return [Server(host, port or 8448)]
 
+        # Check _matrix-fed._tcp SRV record.
         logger.debug("Looking up SRV record for %s", host.decode(errors="replace"))
-        server_list = await self._srv_resolver.resolve_service(b"_matrix._tcp." + host)
+        server_list = await self._srv_resolver.resolve_service(
+            b"_matrix-fed._tcp." + host
+        )
 
         if server_list:
             logger.debug(
@@ -410,6 +413,20 @@ async def _resolve_server(self) -> List[Server]:
             )
             return server_list
 
+        # No _matrix-fed._tcp SRV record, fallback to legacy _matrix._tcp SRV record.
+        logger.debug(
+            "Looking up deprecated SRV record for %s", host.decode(errors="replace")
+        )
+        server_list = await self._srv_resolver.resolve_service(b"_matrix._tcp." + host)
+
+        if server_list:
+            logger.debug(
+                "Got %s from deprecated SRV lookup for %s",
+                ", ".join(map(str, server_list)),
+                host.decode(errors="replace"),
+            )
+            return server_list
+
         # No SRV records, so we fallback to host and 8448
         logger.debug("No SRV records for %s", host.decode(errors="replace"))
         return [Server(host, 8448)]