How handle only one client simultaneously in erlang gen_tcp? - tcp

Question

I have TCP server that listen Ip:Port.
listen(Ip, Port) ->
Opts = [
binary,
{active, false},
{packet, 0},
{reuseaddr, true},
{ip, Ip}
],
case gen_tcp:listen(Port, Opts) of
{ok, ListenSock} ->
?MODULE:loop_accept(ListenSock);
{error, Reason} ->
exit(Reason)
end.
loop_accept(ListenSock) ->
{ok, Sock} = gen_tcp:accept(ListenSock),
?MODULE:loop(Sock),
?MODULE:loop_accept(ListenSock).
loop(Sock) ->
case gen_tcp:recv(Sock, 0) of
{ok, Data} ->
gen_tcp:send(Sock, [<<"Response: ">>, Data]),
?MODULE:loop(Sock);
{error, Reason} ->
ok
end.
Task: when one client connected on Ip:Port (for example telnet Ip Port), another client trying connection must be dropped. In other words, exclusive usage of Ip:Port.
Questions:
How it's implement on Erlang using gen_tcp module?
It is possible resolve by options of gen_tcp:listen?
How to programmaticaly drop trying connection in Erlang?
P.S. I am new in erlang.

Answer 1

First, you can't recv() like that when you specify {packet, 0}. Read this answer about gen_tcp.
The server could:
Pid = spawn(?MODULE, loop, [Sock])
Monitor the process in #1:
Ref = monitor(process, Pid)
But to prevent a race condition, you should perform #1 and #2 in one step:
{Pid, Ref} = spawn_monitor(?MODULE, loop [Sock])
After gen_tcp:accept(ListenSock) executes, do:
gen_tcp:close(ListenSock)
Detect when the client terminates and therefore it's time to start listening for a new client:
receive {'DOWN', Ref, process, Pid, _Reason} ->
listen(Ip, Port)
Or, if the client will not terminate after it is done sending data, then you can detect when the client closes the socket in loop():
case gen_tcp:recv(Sock, 0) of
{ok, Data} ->
gen_tcp:send(Sock, [<<"Response: ">>, Data]),
?MODULE:loop(Sock);
{error, closed} ->
listen(Ip, Port);
{error, Reason} ->
ok
end
=====
backlog socket option (e.g. {backlog, 0}):
The backlog option sets an OS socket configuration parameter. From man listen:
The backlog parameter defines the maximum length for the queue of
pending connections. If a connection request arrives with the queue
full, the client may receive an error with an indication of
ECONNREFUSED. Alternatively, if the underlying protocol supports
retransmission, the request may be ignored so that retries may
succeed.
And, a good read is this thread at Perl Monks: TCP server: How to reject connections when busy? Some snippets about the backlog configuration:
So, it looks like connection request are just ignored (as TCP does
support retransmission)
...when the queue is full, then the system simply stops answering SYN
packets, which triggers their resending. As result, the peer gets no
indication that your server is busy. It simply keeps waiting for
connection to be established.