## ranch_acceptor.erl ##
启动一个loop/3循环，当接收到客户端的socket之后，把socket的控制进程改为连接监控进程ranch\_conns\_sup，连接监控进程中有对应的一些消息处理，接着调用ranch\_conns\_sup:start\_protocol/2发送{?MODULE, start\_protocol, self(), Socket},ranch\_conns\_sup进程自身对该消息进行处理，详情看ranch\_conns\_sup.erl的介绍，至此，ranch的监听端口的工作都已经准备完毕，（发现还有部分忽略了，需要实现ranch_protocol行为才能处理客户端消息的）现在就差客户端的连接进来了。
```
-spec loop(inet:socket(), module(), pid()) -> no_return().
loop(LSocket, Transport, ConnsSup) ->
_ = case Transport:accept(LSocket, infinity) of
{ok, CSocket} ->
case Transport:controlling_process(CSocket, ConnsSup) of
ok ->
%% This call will not return until process has been started
%% AND we are below the maximum number of connections.
ranch_conns_sup:start_protocol(ConnsSup, CSocket);
{error, _} ->
Transport:close(CSocket)
end;
%% Reduce the accept rate if we run out of file descriptors.
%% We can't accept anymore anyway, so we might as well wait
%% a little for the situation to resolve itself.
{error, emfile} ->
error_logger:warning_msg("Ranch acceptor reducing accept rate: out of file descriptors~n"),
receive after 100 -> ok end;
%% We want to crash if the listening socket got closed.
{error, Reason} when Reason =/= closed ->
ok
end,
flush(),
?MODULE:loop(LSocket, Transport, ConnsSup).
```