Merge pull request #9834 from zmstone/0123-fix-idle_timeout-infinity

fix(emqx_connection): crash when idle_timeout is set to infinity

apps/emqx/i18n/emqx_schema_i18n.conf

@@ -656,8 +656,15 @@ mqtt 下所有的配置作为全局的默认值存在，它可以被 <code>zone<
 
     mqtt_idle_timeout {
         desc {
-            en: """After the TCP connection is established, if the MQTT CONNECT packet from the client is not received within the time specified by <code>idle_timeout</code>, the connection will be disconnected."""
-            zh: """TCP 连接建立后，如果在 <code>idle_timeout</code> 指定的时间内未收到客户端的 MQTT CONNECT 报文，则连接将被断开。"""
+            en: """After the TCP connection is established, if the MQTT CONNECT packet from the client is
+not received within the time specified by <code>idle_timeout</code>, the connection will be disconnected.
+After the CONNECT packet has been accepted by EMQX, if the connection idles for this long time,
+then the Erlang process is put to hibernation to save OS resources. Note: long <code>idle_timeout</code>
+interval may impose risk at the system if large number of malicious clients only establish connections
+but do not send any data."""
+            zh: """TCP 连接建立后，如果在 <code>idle_timeout</code> 指定的时间内未收到客户端的 MQTT CONNECT 报文，则连接将被断开。
+如果连接在 CONNECT 报文被 EMQX 接受之后空闲超过该时长，那么服务这个连接的 Erlang 进程会进入休眠以节省系统资源。
+注意，该配置值如果设置过大的情况下，如果大量恶意客户端只连接，但不发任何数据，可能会导致系统资源被恶意消耗。"""
         }
         label: {
             en: """Idle Timeout"""

apps/emqx/src/emqx.app.src

@@ -3,7 +3,7 @@
     {id, "emqx"},
     {description, "EMQX Core"},
     % strict semver, bump manually!
-    {vsn, "5.0.15"},
+    {vsn, "5.0.16"},
     {modules, []},
     {registered, []},
     {applications, [

apps/emqx/src/emqx_connection.erl

@@ -403,14 +403,19 @@ exit_on_sock_error(Reason) ->
 recvloop(
     Parent,
     State = #state{
-        idle_timeout = IdleTimeout,
+        idle_timeout = IdleTimeout0,
         zone = Zone
     }
 ) ->
+    IdleTimeout =
+        case IdleTimeout0 of
+            infinity -> infinity;
+            _ -> IdleTimeout0 + 100
+        end,
     receive
         Msg ->
             handle_recv(Msg, Parent, State)
-    after IdleTimeout + 100 ->
+    after IdleTimeout ->
         case emqx_olp:backoff_hibernation(Zone) of
             true ->
                 recvloop(Parent, State);

apps/emqx/test/emqx_client_SUITE.erl

@@ -24,6 +24,7 @@
 -include_lib("emqx/include/emqx_mqtt.hrl").
 -include_lib("eunit/include/eunit.hrl").
 -include_lib("common_test/include/ct.hrl").
+-include_lib("snabbkaffe/include/snabbkaffe.hrl").
 
 -define(TOPICS, [
     <<"TopicA">>,
@@ -43,6 +44,8 @@
     <<"TopicA/#">>
 ]).
 
+-define(WAIT(EXPR, ATTEMPTS), ?retry(1000, ATTEMPTS, EXPR)).
+
 all() ->
     [
         {group, mqttv3},
@@ -85,6 +88,12 @@ init_per_suite(Config) ->
 end_per_suite(_Config) ->
     emqx_common_test_helpers:stop_apps([]).
 
+init_per_testcase(_Case, Config) ->
+    Config.
+
+end_per_testcase(_Case, _Config) ->
+    emqx_config:put_zone_conf(default, [mqtt, idle_timeout], 15000).
+
 %%--------------------------------------------------------------------
 %% Test cases for MQTT v3
 %%--------------------------------------------------------------------
@@ -101,16 +110,35 @@ t_basic_v4(_Config) ->
 
 t_cm(_) ->
     emqx_config:put_zone_conf(default, [mqtt, idle_timeout], 1000),
-    ClientId = <<"myclient">>,
+    ClientId = atom_to_binary(?FUNCTION_NAME),
     {ok, C} = emqtt:start_link([{clientid, ClientId}]),
     {ok, _} = emqtt:connect(C),
-    ct:sleep(500),
-    #{clientinfo := #{clientid := ClientId}} = emqx_cm:get_chan_info(ClientId),
+    ?WAIT(#{clientinfo := #{clientid := ClientId}} = emqx_cm:get_chan_info(ClientId), 2),
     emqtt:subscribe(C, <<"mytopic">>, 0),
-    ct:sleep(1200),
-    Stats = emqx_cm:get_chan_stats(ClientId),
-    ?assertEqual(1, proplists:get_value(subscriptions_cnt, Stats)),
-    emqx_config:put_zone_conf(default, [mqtt, idle_timeout], 15000).
+    ?WAIT(
+        begin
+            Stats = emqx_cm:get_chan_stats(ClientId),
+            ?assertEqual(1, proplists:get_value(subscriptions_cnt, Stats))
+        end,
+        2
+    ),
+    ok.
+
+t_idle_timeout_infinity(_) ->
+    emqx_config:put_zone_conf(default, [mqtt, idle_timeout], infinity),
+    ClientId = atom_to_binary(?FUNCTION_NAME),
+    {ok, C} = emqtt:start_link([{clientid, ClientId}]),
+    {ok, _} = emqtt:connect(C),
+    ?WAIT(#{clientinfo := #{clientid := ClientId}} = emqx_cm:get_chan_info(ClientId), 2),
+    emqtt:subscribe(C, <<"mytopic">>, 0),
+    ?WAIT(
+        begin
+            Stats = emqx_cm:get_chan_stats(ClientId),
+            ?assertEqual(1, proplists:get_value(subscriptions_cnt, Stats))
+        end,
+        2
+    ),
+    ok.
 
 t_cm_registry(_) ->
     Children = supervisor:which_children(emqx_cm_sup),

changes/v5.0.16/fix-9834.en.md

@@ -0,0 +1 @@
+Allow `mqtt.idle_timeout` to be set to `infinity`

changes/v5.0.16/fix-9834.zh.md

@@ -0,0 +1 @@
+允许配置项 `mqtt.idle_timeout` 设置成 `infinity`