Prevent overload of incomming s2s connections

Three changes were introduced:

1) ejabberd_s2s_in now uses p1_fsm instead of gen_fsm.  And uses
   the {max_queue, N} option to kill the process if its input
   queue grows too much.
2) If a ejabberd_s2s_in process is overload and killed, the server
   that originated that connection is not allowed to connect back
   to us for X seconds (set to 60seconds on the source)
3) The list of blocked (both statically and dynamically by the above
   method) host is now also checked for hosts authenticating by
   starttls+sasl. Previusly it was only used during dialback.

2 files changed, 62 insertions, 6 deletions

diff --git a/src/ejabberd_s2s.erl b/src/ejabberd_s2s.erl
index a3ccb7d74..84fe92401 100644
--- a/src/ejabberd_s2s.erl
+++ b/src/ejabberd_s2s.erl
@@ -42,7 +42,11 @@
 	 allow_host/2,
 	 incoming_s2s_number/0,
 	 outgoing_s2s_number/0,
-	 migrate/1
+	 migrate/1,
+	 clean_temporarily_blocked_table/0,
+	 list_temporarily_blocked_hosts/0,
+	 external_host_overloaded/1,
+	 is_temporarly_blocked/1
 	]).
 
 %% gen_server callbacks
@@ -65,9 +69,14 @@
 -define(PREFIXED_NS,
         [{?NS_XMPP, ?NS_XMPP_pfx}, {?NS_DIALBACK, ?NS_DIALBACK_pfx}]).
 
+-define(S2S_OVERLOAD_BLOCK_PERIOD, 60).
+%% once a server is temporarly blocked, it stay blocked for 60 seconds
+
 -record(s2s, {fromto, pid, key}).
 -record(state, {}).
 
+-record(temporarily_blocked, {host, timestamp}).
+
 %%====================================================================
 %% API
 %%====================================================================
@@ -98,6 +107,31 @@ route(From, To, Packet) ->
             ok
     end.
 
+clean_temporarily_blocked_table() ->
+	mnesia:clear_table(temporarily_blocked).
+list_temporarily_blocked_hosts() ->
+	ets:tab2list(temporarily_blocked).
+
+external_host_overloaded(Host) ->
+	?INFO_MSG("Disabling connections from ~s for ~s seconds", [Host, ?S2S_OVERLOAD_BLOCK_PERIOD]),
+	mnesia:transaction( fun() ->
+		mnesia:write(#temporarily_blocked{host = Host, timestamp = now()})
+	end).
+
+is_temporarly_blocked(Host) ->
+	case mnesia:dirty_read(temporarily_blocked, Host) of
+		[] -> false;
+		[#temporarily_blocked{timestamp = T}=Entry] ->
+			case timer:now_diff(now(), T) of
+				N when N > ?S2S_OVERLOAD_BLOCK_PERIOD * 1000 * 1000 ->
+					mnesia:dirty_delete_object(Entry),
+					false;
+				_ ->
+					true
+			end
+	end.
+
+
 remove_connection(FromTo, Pid, Key) ->
     case catch mnesia:dirty_match_object(s2s, #s2s{fromto = FromTo,
 						   pid = Pid,
@@ -248,6 +282,7 @@ init([]) ->
     mnesia:add_table_copy(s2s, node(), ram_copies),
     ejabberd_hooks:add(node_hash_update, ?MODULE, migrate, 100),
     ejabberd_commands:register_commands(commands()),
+    mnesia:create_table(temporarily_blocked, [{ram_copies, [node()]}, {attributes, record_info(fields, temporarily_blocked)}]),
     {ok, #state{}}.
 
 %%--------------------------------------------------------------------
@@ -576,6 +611,8 @@ update_tables() ->
 
 %% Check if host is in blacklist or white list
 allow_host(MyServer, S2SHost) ->
+	allow_host2(MyServer, S2SHost) andalso (not is_temporarly_blocked(S2SHost)).
+allow_host2(MyServer, S2SHost) ->
     Hosts = ?MYHOSTS,
     case lists:dropwhile(
 	   fun(ParentDomain) ->
diff --git a/src/ejabberd_s2s_in.erl b/src/ejabberd_s2s_in.erl
index 367bce502..a0cf1a732 100644
--- a/src/ejabberd_s2s_in.erl
+++ b/src/ejabberd_s2s_in.erl
@@ -27,7 +27,7 @@
 -module(ejabberd_s2s_in).
 -author('alexey@process-one.net').
 
--behaviour(gen_fsm).
+-behaviour(p1_fsm).
 
 %% External exports
 -export([start/2,
@@ -80,10 +80,12 @@
 -define(FSMOPTS, []).
 -endif.
 
+-define(FSMLIMITS, [{max_queue, 2000}]). %% if queue grows more than this, we shutdown this connection.
+
 %% Module start with or without supervisor:
 -ifdef(NO_TRANSIENT_SUPERVISORS).
--define(SUPERVISOR_START, gen_fsm:start(ejabberd_s2s_in, [SockData, Opts],
-					?FSMOPTS)).
+-define(SUPERVISOR_START, p1_fsm:start(ejabberd_s2s_in, [SockData, Opts],
+					?FSMOPTS ++ ?FSMLIMITS)).
 -else.
 -define(SUPERVISOR_START, supervisor:start_child(ejabberd_s2s_in_sup,
 						 [SockData, Opts])).
@@ -102,7 +104,7 @@ start(SockData, Opts) ->
     ?SUPERVISOR_START.
 
 start_link(SockData, Opts) ->
-    gen_fsm:start_link(ejabberd_s2s_in, [SockData, Opts], ?FSMOPTS).
+    p1_fsm:start_link(ejabberd_s2s_in, [SockData, Opts], ?FSMOPTS ++ ?FSMLIMITS).
 
 socket_type() ->
     xml_stream.
@@ -323,7 +325,8 @@ wait_for_feature_request({xmlstreamelement, El}, StateData) ->
 			    false
 		    end,
 		    if
-			AuthRes ->
+		    	AllowRemoteHost = ejabberd_s2s:allow_host("", AuthDomain),
+			AuthRes andalso AllowRemoteHost ->
 			    (StateData#state.sockmod):reset_stream(
 			      StateData#state.socket),
 			    send_element(StateData,
@@ -622,9 +625,25 @@ handle_info(_, StateName, StateData) ->
 %%----------------------------------------------------------------------
 terminate(Reason, _StateName, StateData) ->
     ?DEBUG("terminated: ~p", [Reason]),
+    case Reason of
+	    {process_limit, _} ->
+		    [ejabberd_s2s:external_host_overloaded(Host) || Host <- get_external_hosts(StateData)];
+	    _ ->
+		    ok
+    end,
     (StateData#state.sockmod):close(StateData#state.socket),
     ok.
 
+get_external_hosts(StateData) ->
+    case StateData#state.authenticated of
+	    true ->
+		[StateData#state.auth_domain];
+	    false ->
+		Connections = StateData#state.connections,
+		[D || {{D, _}, established} <- dict:to_list(Connections)]
+    end.
+
+
 %%%----------------------------------------------------------------------
 %%% Internal functions
 %%%----------------------------------------------------------------------