iot_cloud/src/transport/ssl_channel.erl
2026-05-10 23:44:35 +08:00

370 lines
16 KiB
Erlang

%%%-------------------------------------------------------------------
%%% @author licheng5
%%% @copyright (C) 2021, <COMPANY>
%%% @doc
%%%
%%% @end
%%% Created : 11. 1月 2021 上午12:17
%%%-------------------------------------------------------------------
-module(ssl_channel).
-author("licheng5").
-behaviour(ranch_protocol).
-define(INFLIGHT_TIMEOUT, 60000).
-define(SSL_IDLE_TIMEOUT, 120000).
%% API
-export([pub/4, container_call/4, cancel_command_call/2]).
-export([start_link/3, stop/2]).
%% gen_server callbacks
-export([init/3, handle_call/3, handle_cast/2, handle_info/2, code_change/3, terminate/2]).
-record(state, {
transport,
socket,
uuid :: undefined | binary(),
is_authed = false :: boolean(),
%% 用户进程id
host_pid = undefined,
%% iot 发起的 command 与 command_response 的对应关系
inflight = #{},
idle_timer_ref :: undefined | reference()
}).
-record(inflight_command, {
receiver_pid :: undefined | pid(),
timer_ref :: reference()
}).
-type request_ref() :: binary().
%% 向通道中写入消息
-spec pub(Pid :: pid(), Topic :: binary(), Qos :: integer(), Content :: binary()) -> ok.
pub(Pid, Topic, Qos, Content) when is_pid(Pid), is_binary(Topic), is_integer(Qos), is_binary(Content) ->
gen_server:cast(Pid, {pub, Topic, Qos, Content}).
-spec container_call(Pid :: pid(), ReceiverPid :: pid(), Ref :: request_ref(), Request :: map()) -> ok.
container_call(Pid, ReceiverPid, Ref, Request) when is_pid(Pid), is_pid(ReceiverPid), is_binary(Ref), is_map(Request) ->
gen_server:cast(Pid, {command_call, ReceiverPid, Ref, {container, Request}}),
ok.
-spec cancel_command_call(Pid :: pid(), Ref :: request_ref()) -> ok.
cancel_command_call(Pid, Ref) when is_pid(Pid), is_binary(Ref) ->
gen_server:call(Pid, {cancel_command_call, Ref}).
%% 关闭方法
-spec stop(Pid :: undefined | pid(), Reason :: any()) -> ok.
stop(undefined, _Reason) ->
ok;
stop(Pid, Reason) when is_pid(Pid) ->
gen_server:stop(Pid, Reason, 5000).
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% 逻辑处理方法
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
start_link(Ref, Transport, Opts) ->
{ok, proc_lib:spawn_link(?MODULE, init, [Ref, Transport, Opts])}.
init(Ref, Transport, _Opts = []) ->
ok = iot_log:set_metadata(),
{ok, Socket} = ranch:handshake(Ref),
logger:debug("[ssl_channel] get a new connection: ~p", [Socket]),
Transport:setopts(Socket, [binary, {active, true}, {packet, 4}]),
IdleTimerRef = start_idle_timer(),
gen_server:enter_loop(?MODULE, [], #state{transport = Transport, socket = Socket, idle_timer_ref = IdleTimerRef}).
handle_call({cancel_command_call, Ref}, _From, State = #state{inflight = Inflight}) ->
case maps:take(Ref, Inflight) of
{#inflight_command{timer_ref = TimerRef}, NInflight} ->
erlang:cancel_timer(TimerRef),
{reply, ok, State#state{inflight = NInflight}};
error ->
{reply, ok, State}
end;
handle_call(_Request, _From, State) ->
{reply, ok, State}.
%% 发送消息, 基于pub/sub机制
handle_cast({pub, Topic, Qos, Content}, State = #state{transport = Transport, socket = Socket}) ->
Packet = term_to_binary({<<"message">>, {<<"pub">>, #{<<"topic">> => Topic, <<"qos">> => Qos, <<"content">> => Content}}}),
case Transport:send(Socket, Packet) of
ok ->
{noreply, State};
{error, Reason} ->
logger:warning("[ssl_channel] send pub failed, reason: ~p", [Reason]),
{stop, {send_failed, Reason}, State}
end;
%% iot 请求 efka 时使用 command/command_response 语义。
handle_cast({command_call, ReceiverPid, Ref, Body}, State = #state{transport = Transport, socket = Socket, inflight = Inflight}) ->
Packet = term_to_binary({<<"command">>, Ref, encode_command_body(Body)}),
case Transport:send(Socket, Packet) of
ok ->
TimerRef = erlang:start_timer(?INFLIGHT_TIMEOUT, self(), {command_timeout, Ref}),
CommandInfo = #inflight_command{receiver_pid = ReceiverPid, timer_ref = TimerRef},
{noreply, State#state{inflight = maps:put(Ref, CommandInfo, Inflight)}};
{error, Reason} ->
logger:warning("[ssl_channel] send command failed, ref: ~p, reason: ~p", [Ref, Reason]),
deliver_command_error(ReceiverPid, Ref, {send_failed, Reason}),
{stop, {send_failed, Reason}, State}
end.
handle_info({timeout, TimerRef, {command_timeout, Ref}}, State = #state{inflight = Inflight}) ->
case maps:get(Ref, Inflight, undefined) of
#inflight_command{receiver_pid = ReceiverPid, timer_ref = TimerRef} ->
logger:warning("[ws_channel] command timeout, ref: ~p", [Ref]),
reply_command_timeout(ReceiverPid, Ref),
{noreply, State#state{inflight = maps:remove(Ref, Inflight)}};
_ ->
{noreply, State}
end;
handle_info({timeout, TimerRef, ssl_idle_timeout}, State = #state{idle_timer_ref = TimerRef}) ->
logger:notice("[ssl_channel] ssl channel idle timeout"),
{stop, ssl_idle_timeout, State#state{idle_timer_ref = undefined}};
handle_info({timeout, _TimerRef, ssl_idle_timeout}, State) ->
{noreply, State};
%% 关闭当前通道
handle_info({stop, Reason}, State) ->
{stop, Reason, State};
%% 主机进程挂掉的时候,需要关闭掉链接
handle_info({'DOWN', _, process, HostPid, Reason}, State = #state{uuid = UUID, host_pid = HostPid}) ->
logger:debug("[ws_channel] uuid: ~p, channel will close because host exited with reason: ~p", [UUID, Reason]),
{stop, Reason, State};
handle_info({ssl, Socket, PacketBin}, State = #state{socket = Socket}) when is_binary(PacketBin) ->
State1 = reset_idle_timer(State),
try binary_to_term(PacketBin, [safe]) of
{<<"request">>, Ref, Body} ->
handle_request_frame(Ref, Body, State1);
{<<"message">>, Body} ->
handle_message_frame(Body, State1);
{<<"command_response">>, Ref, Response} ->
handle_command_response_frame(Ref, Response, State1);
Other ->
logger:warning("[ssl_channel] unsupported packet: ~p", [Other]),
{stop, bad_packet, State1}
catch error:Error ->
logger:warning("[ssl_channel] binary_to_term get error: ~p, packet_size: ~p", [Error, byte_size(PacketBin)]),
{stop, bad_packet, State1}
end;
handle_info({ssl_closed, Socket}, State = #state{socket = Socket}) ->
logger:notice("[ssl_channel] ssl socket closed"),
{stop, normal, State};
handle_info({ssl_error, Socket, Reason}, State = #state{socket = Socket}) ->
logger:notice("[ssl_channel] ssl socket error: ~p", [Reason]),
{stop, normal, State};
handle_info({ssl_passive, Socket}, State = #state{transport = Transport, socket = Socket}) ->
ok = Transport:setopts(Socket, [{active, true}]),
{noreply, State};
handle_info(Info, State) ->
logger:warning("[ssl_channel] get a unknown message: ~p, state: ~p", [Info, State]),
{noreply, State}.
terminate(Reason, #state{inflight = Inflight, transport = Transport, socket = Socket, idle_timer_ref = IdleTimerRef}) ->
cancel_timer(IdleTimerRef),
maps:foreach(fun(Ref, CommandInfo) -> reply_command_closed(Ref, CommandInfo, Reason) end, Inflight),
Transport:close(Socket),
logger:warning("[ssl_channel] stop with reason: ~p", [Reason]),
ok.
code_change(_OldVsn, State, _Extra) ->
{ok, State}.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%% helper methods
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-spec handle_request_frame(request_ref(), tuple(), #state{}) -> {noreply, #state{}} | {stop, term(), #state{}}.
handle_request_frame(Ref, {<<"auth_request">>, #{<<"uuid">> := UUID}}, State = #state{is_authed = true}) ->
logger:warning("[ws_channel] repeated auth request, ref: ~p, uuid: ~p, close channel", [Ref, UUID]),
{stop, repeated_auth, State};
handle_request_frame(Ref, {<<"auth_request">>, #{<<"uuid">> := UUID, <<"token">> := Token, <<"timestamp">> := Timestamp}}, State = #state{transport = Transport, socket = Socket}) ->
maybe
ok ?= auth(Token, UUID, Timestamp),
{ok, HostPid} ?= iot_host:lookup_pid(UUID),
ok ?= iot_host:attach_channel(HostPid, self()),
erlang:monitor(process, HostPid),
ok = send_reply_frame(Transport, Socket, Ref, {<<"auth_response">>, <<"ok">>}),
logger:debug("[ws_channel] auth uuid: ~p", [UUID]),
{noreply, State#state{uuid = UUID, is_authed = true, host_pid = HostPid}}
else
{error, Reason} ->
logger:warning("[ws_channel] uuid: ~p, auth failed with reason: ~p", [UUID, Reason]),
case send_reply_frame(Transport, Socket, Ref, {<<"auth_response">>, {<<"error">>, {<<"failed">>, Reason}}}) of
ok ->
{stop, Reason, State};
{error, SendReason} ->
logger:warning("[ws_channel] uuid: ~p, send auth failed response failed: ~p", [UUID, SendReason]),
{stop, {send_failed, SendReason}, State}
end
end;
handle_request_frame(Ref, {<<"container">>, ContainerCommand}, State) ->
logger:warning("[ws_channel] unsupported request message type: container, ref: ~p, command: ~p", [Ref, ContainerCommand]),
{stop, normal, State};
handle_request_frame(Ref, Body, State) ->
logger:warning("[ws_channel] unsupported request body, ref: ~p, body: ~p", [Ref, Body]),
{stop, normal, State}.
-spec handle_message_frame(term(), #state{}) ->
{noreply, #state{}} | {stop, term(), #state{}}.
handle_message_frame(<<"ping">>, State = #state{transport = Transport, socket = Socket}) ->
Packet = term_to_binary({<<"message">>, <<"pong">>}),
case Transport:send(Socket, Packet) of
ok ->
{noreply, State};
{error, Reason} ->
logger:warning("[ssl_channel] send pong failed, reason: ~p", [Reason]),
{stop, {send_failed, Reason}, State}
end;
handle_message_frame({<<"data">>, #{<<"route_key">> := RouteKey, <<"metric">> := Metric}}, State = #state{host_pid = HostPid}) when is_pid(HostPid) ->
iot_host:handle(HostPid, {data, RouteKey, Metric}),
{noreply, State};
handle_message_frame({<<"task_event">>, Event}, State = #state{uuid = UUID, host_pid = HostPid}) when is_binary(UUID), is_pid(HostPid) ->
handle_event_stream_frame(UUID, Event),
{noreply, State};
handle_message_frame(Body, State) ->
logger:warning("[ssl_channel] unsupported message body: ~p", [Body]),
{noreply, State}.
-spec handle_event_stream_frame(binary(), map()) -> ok.
handle_event_stream_frame(UUID, #{<<"task_id">> := TaskId, <<"type">> := <<"close">>, <<"stream">> := Reason}) ->
logger:debug("[ssl_channel] get uuid: ~p, task_id: ~p, close with reason: ~p", [UUID, TaskId, Reason]),
iot_container_task_sup:close(UUID, TaskId, Reason);
handle_event_stream_frame(UUID, #{<<"task_id">> := TaskId, <<"type">> := Type, <<"stream">> := Stream}) ->
logger:debug("[ssl_channel] get uuid: ~p, task_id: ~p, type: ~ts, stream: ~ts", [UUID, TaskId, Type, Stream]),
iot_container_task_sup:stream(UUID, TaskId, Type, Stream);
handle_event_stream_frame(UUID, Event) ->
logger:warning("[ssl_channel] invalid task_event, uuid: ~p, event: ~p", [UUID, Event]),
ok.
-spec handle_command_response_frame(request_ref(), tuple(), #state{}) ->
{noreply, #state{}}.
handle_command_response_frame(Ref, Reply, State = #state{inflight = Inflight}) when is_binary(Ref) ->
case maps:take(Ref, Inflight) of
error ->
{noreply, State};
{#inflight_command{receiver_pid = ReceiverPid, timer_ref = TimerRef}, NInflight} ->
erlang:cancel_timer(TimerRef),
deliver_command_response(ReceiverPid, Ref, Reply),
{noreply, State#state{inflight = NInflight}}
end;
handle_command_response_frame(Ref, Reply, State) ->
logger:warning("[ws_channel] unexpected command_response frame, ref: ~p, reply: ~p", [Ref, Reply]),
{noreply, State}.
-spec send_reply_frame(module(), any(), request_ref(), tuple()) -> ok | {error, term()}.
send_reply_frame(Transport, Socket, Ref, Reply) ->
Packet = term_to_binary({<<"response">>, Ref, Reply}),
Transport:send(Socket, Packet).
-spec decode_command_response({binary(), term()} | tuple()) ->
ok | {ok, term()} | {error, term()}.
decode_command_response({<<"container">>, <<"ok">>}) ->
ok;
decode_command_response({<<"container">>, {<<"ok">>, Result}}) ->
{ok, Result};
decode_command_response({<<"container">>, {<<"error">>, Reason}}) ->
{error, Reason};
decode_command_response(_Reply) ->
{error, invalid_response}.
-spec deliver_command_response(undefined | pid(), request_ref(), tuple()) -> ok.
deliver_command_response(undefined, _Ref, _Reply) ->
ok;
deliver_command_response(ReceiverPid, Ref, Reply) when is_pid(ReceiverPid) ->
case is_process_alive(ReceiverPid) of
true ->
ReceiverPid ! {command_reply, Ref, decode_command_response(Reply)};
false ->
logger:warning("[ws_channel] get command_response: ~p, ref: ~p, but receiver_pid is deaded", [Reply, Ref])
end.
-spec deliver_command_error(undefined | pid(), request_ref(), term()) -> ok.
deliver_command_error(ReceiverPid, Ref, Reason) when is_pid(ReceiverPid) ->
ReceiverPid ! {command_reply, Ref, {error, Reason}},
ok;
deliver_command_error(_ReceiverPid, _Ref, _Reason) ->
ok.
-spec reply_command_timeout(undefined | pid(), request_ref()) -> ok.
reply_command_timeout(ReceiverPid, Ref) when is_pid(ReceiverPid) ->
ReceiverPid ! {command_reply, Ref, {error, timeout}},
ok;
reply_command_timeout(_ReceiverPid, _Ref) ->
ok.
-spec reply_command_closed(request_ref(), #inflight_command{}, term()) -> ok.
reply_command_closed(Ref, #inflight_command{receiver_pid = ReceiverPid}, Reason) when is_pid(ReceiverPid) ->
ReceiverPid ! {command_reply, Ref, {error, {channel_closed, Reason}}},
ok;
reply_command_closed(_Ref, _CommandInfo, _Reason) ->
ok.
%% 检测token是否是合法值
-spec auth(Token :: binary(), UUID :: binary(), Timestamp :: integer()) ->
ok | {error, Reason :: binary()}.
auth(Token, UUID, Timestamp) when is_binary(Token), is_binary(UUID), is_integer(Timestamp) ->
%% 1分钟内有效
Now = iot_util:current_time(),
case Timestamp =< Now andalso Now - Timestamp =< 60 of
true ->
case efka_client_store:auth(UUID, Token) of
true ->
ok;
false ->
{error, <<"invalid token">>}
end;
false ->
{error, <<"invalid timestamp">>}
end.
-spec encode_command_body({container, map()}) -> {binary(), map()}.
encode_command_body({container, Request}) ->
{<<"container">>, safe_term(Request)}.
-spec safe_term(term()) -> term().
safe_term(true) ->
true;
safe_term(false) ->
false;
safe_term(undefined) ->
undefined;
safe_term(Value) when is_atom(Value) ->
atom_to_binary(Value, utf8);
safe_term(Value) when is_map(Value) ->
maps:from_list([{safe_term(K), safe_term(V)} || {K, V} <- maps:to_list(Value)]);
safe_term(Value) when is_list(Value) ->
[safe_term(Item) || Item <- Value];
safe_term(Value) when is_tuple(Value) ->
list_to_tuple([safe_term(Item) || Item <- tuple_to_list(Value)]);
safe_term(Value) ->
Value.
-spec start_idle_timer() -> reference().
start_idle_timer() ->
erlang:start_timer(?SSL_IDLE_TIMEOUT, self(), ssl_idle_timeout).
-spec reset_idle_timer(#state{}) -> #state{}.
reset_idle_timer(State = #state{idle_timer_ref = TimerRef}) ->
cancel_timer(TimerRef),
State#state{idle_timer_ref = start_idle_timer()}.
-spec cancel_timer(undefined | reference()) -> ok.
cancel_timer(undefined) ->
ok;
cancel_timer(TimerRef) ->
_ = erlang:cancel_timer(TimerRef),
ok.