Assume a legacy Linux application listening on a UNIX domain socket /tmp/foo.
In addition to communicating with this legacy application over the UNIX domain socket mechanism I want to be able to connect to it via a TCP-connection on port say 1234.
What is the easiest way to bind to TCP port 1234 and then redirect all incoming connections to the UNIX domain socket /tmp/foo?
Turns out socat can be used to achieve this:
socat TCP-LISTEN:1234,reuseaddr,fork UNIX-CLIENT:/tmp/foo
And with a bit of added security:
socat TCP-LISTEN:1234,bind=127.0.0.1,reuseaddr,fork,su=nobody,range=127.0.0.0/8 UNIX-CLIENT:/tmp/foo
These examples have been tested and work as expected.
stunnel around it for more layers of complexity!
su=nobody part
-H tcp://127.0.0.1:4321 (note that it allows access to the docker daemon for anyone that can open a connection to 127.0.0.1. No uid/gid checks or anything, but that is also true for the netcat/socat approach)
Easiest? Probably Netcat (aka nc):
nc -l 1234 | nc -U /tmp/foo
The first command listens on port 1234 for incoming connections, and pipes the resulting data to the second command. The second connects to the Unix domain socket /tmp/foo, and writes its input to that socket. Note that this will only accept a single connection, and exit as soon as that connection is dropped. If you want to keep listening for more connections, use the -k option:
nc -lk 1234 | nc -U /tmp/foo
You can test that this is working by setting up a listener for that socket in one terminal:
nc -lUk /tmp/foo
And writing to it in another:
nc localhost 1234
socat, as recommended by knorv, is more capable, but more complicated to use.
nc -U option? I don't have a Quantal machine to test on, but on Precise nc has the -U option, and what I said in my comment works fine. Have you tried exactly what I wrote? If you aren't finding that nc works for you, I would recommend trying socat; it can do a lot more than nc, with much finer grained control of how it works, but it is a bit more complicated to figure out how to use.nc -l 1234 | nc -U /tmp/foo gives nc: invalid option -- 'U' and nc -h for help. socat works fine :)netcat-traditional and netcat-openbsd. netcat-openbsd is the one I have installed, and seems to have more features than netcat-traditional. You can install them both at once, with nc and netcat symlinked to one or the other.
socat which will forward the response back to us.You should be able to bind to TCP 1234, get a socket fd for /tmp/foo and use the select call to 'listen' for data on both 1234, and /tmp/foo. Any data written to 1234, you rewrite to /tmp/foo and vice-versa.
You now act as a proxy and transfer data back and forth.
And here is a web-page which might help: http://osr507doc.sco.com/en/netguide/dusockC.io_multiplexing.html
In additons to @knorv's answer: with xinetd it can work like a daemon
# cat /etc/xined.d/mysrv
service mysrv
{
disable = no
type = UNLISTED
socket_type = stream
protocol = tcp
wait = no
server = /usr/bin/socat
server_args = STDIN UNIX-CLIENT:/tmp/mysocket.sock
bind = 127.0.0.1
port = 1234
}
Not tried it : but it looks like 'lighttpd' can do this for you:
