gen_udp(3)
π¦ gen_udp β the Erlang Module Definition, bestie edition
okay so you found gen_udp(3), the Erlang module page for talking to UDP sockets, and honestly? it's giving "tiny module, huge lore." let's get into it π
β¨ NAME β¨
gen_udp β your interface to UDP sockets. that's it, that's the vibe. π«Ά
πΈ DESCRIPTION πΈ
This module provides functions for communicating over UDP protocol sockets. like, it's the way you do UDP in Erlang, no cap.
NOTE: Functions that create sockets can take an optional option; {inet_backend, Backend} that, if specified, has to be the first option. π (yes, first. it's the main character, it goes first.)
This selects the implementation backend towards the platform's socket API. This is a temporary option that will be ignored in a future release (so don't get attached, bestie π). The default is Backend = inet, which selects the traditional inet_drv.c driver. The other choice is Backend = socket, which selects the new socket module and its NIF implementation. The system default can be changed when the node is started with the application kernel's configuration variable inet_backend. For gen_udp with inet_backend = socket we have tried to be as "compatible" as possible, which has sometimes been impossible π. Here is a list of cases when the behaviour of inet-backend inet (default) and socket are different:
- The option
read_packetsis currently ignored. - Windows require sockets (
domain = inet | inet6) to be bound. Currently all sockets created on Windows withinet_backend = socketwill be bound. If the user does not provide an address,gen_udpwill try to 'figure out' an address itself.
so yeah. two backends, one of them is the default old reliable, the other one is new and does its best. noted. π
π DATA TYPES π
ip6_membership()
ip6_membership() ::
{MultiAddress :: inet:ip6_address(), IfIndex :: integer()}.
ip6_multicast_if()
ip6_multicast_if() :: integer().
IPv6 this multicast interface index (an integer). yeah it's just a lil int π§
ip_membership()
ip_membership() ::
{MultiAddress :: inet:ip4_address(),
Interface :: inet:ip4_address()} |
{MultiAddress :: inet:ip4_address(),
Address :: inet:ip4_address(),
IfIndex :: integer()}.
IP multicast membership.
The 3-tuple form isn't supported on all platforms. 'ifindex' defaults to zero (0) on platforms that supports the 3-tuple variant. so if your platform is being picky about the 3-tuple, that's why, not you. π
ip_multicast_if()
ip_multicast_if() :: inet:ip4_address().
membership()
membership() :: ip_membership() | ip6_membership().
multicast_if()
multicast_if() :: ip_multicast_if() | ip6_multicast_if().
open_option()
open_option() ::
{ip, inet:socket_address()} |
{fd, non_neg_integer()} |
{ifaddr,
socket:sockaddr_in() |
socket:sockaddr_in6() |
inet:socket_address()} |
inet:address_family() |
{port, inet:port_number()} |
{netns, file:filename_all()} |
{bind_to_device, binary()} |
option().
option()
option() ::
{active, true | false | once | -32768..32767} |
{add_membership, membership()} |
{broadcast, boolean()} |
{buffer, non_neg_integer()} |
{debug, boolean()} |
{deliver, port | term} |
{dontroute, boolean()} |
{drop_membership, membership()} |
{exclusiveaddruse, boolean()} |
{header, non_neg_integer()} |
{high_msgq_watermark, pos_integer()} |
{low_msgq_watermark, pos_integer()} |
{mode, list | binary} |
list | binary |
{multicast_if, multicast_if()} |
{multicast_loop, boolean()} |
{multicast_ttl, non_neg_integer()} |
{priority, non_neg_integer()} |
{raw,
Protocol :: non_neg_integer(),
OptionNum :: non_neg_integer(),
ValueBin :: binary()} |
{read_packets, non_neg_integer()} |
{recbuf, non_neg_integer()} |
{reuseaddr, boolean()} |
{reuseport, boolean()} |
{reuseport_lb, boolean()} |
{sndbuf, non_neg_integer()} |
{tos, non_neg_integer()} |
{tclass, non_neg_integer()} |
{ttl, non_neg_integer()} |
{recvtos, boolean()} |
{recvtclass, boolean()} |
{recvttl, boolean()} |
{ipv6_v6only, boolean()}.
look at all those options. a whole buffet. π
option_name()
option_name() ::
active | broadcast | buffer | debug | deliver | dontroute |
exclusiveaddruse | header | high_msgq_watermark |
low_msgq_watermark | mode | multicast_if | multicast_loop |
multicast_ttl | priority |
{raw,
Protocol :: non_neg_integer(),
OptionNum :: non_neg_integer(),
ValueSpec ::
(ValueSize :: non_neg_integer()) | (ValueBin ::
binary())} |
read_packets | recbuf | reuseaddr | reuseport | reuseport_lb |
sndbuf | tos | tclass | ttl | recvtos | recvtclass | recvttl |
pktoptions | ipv6_v6only.
socket()
socket() :: inet:socket().
A socket as returned by open/1,2. that's your lil handle, your ticket to the UDP party ποΈ
π FUNCTIONS π
close/1 β let it go ποΈ
close(Socket) -> ok when Socket :: socket().
Closes a UDP socket. bye bestie, thanks for the datagrams π
connect/2 β lock in π
connect(Socket, SockAddr) -> ok | {error, Reason}
when
Socket :: socket(),
SockAddr ::
socket:sockaddr_in() | socket:sockaddr_in6(),
Reason :: inet:posix().
Connect a UDP socket.
Connecting a UDP socket only means storing the specified (destination) socket address, as specified by SockAddr, so that the system knows where to send data. it's not like a TCP handshake, it's just... remembering. lowkey a notes app for one address. π
When the socket is "connected" it is not necessary to specify the destination address when sending a datagram. That is; send/2 may be used. slay, less typing.
It also means that the socket will only receive data from the connected address. Other messages are discarded on arrival by the OS protocol stack. so it's exclusive now, we love a committed socket. π
connect/3 β same energy, separate args β¨
connect(Socket, Address, Port) -> ok | {error, Reason}
when
Socket :: socket(),
Address :: inet:socket_address() | inet:hostname(),
Port :: inet:port_number(),
Reason :: inet:posix().
Connect a UDP socket.
See connect/2. yeah, same thing, same vibe. π
With this function the destination is specified with separate Address and Port arguments where Address may be an IP address or a host name. pick your fighter, either way it works. π
controlling_process/2 β passing the aux cord π§
controlling_process(Socket, Pid) -> ok | {error, Reason}
when
Socket :: socket(),
Pid :: pid(),
Reason ::
closed | not_owner | badarg |
inet:posix().
Change the controlling process (owner) of a socket.
Assigns a new controlling process Pid to Socket. The controlling process is the process that the socket sends messages to. If this function is called from any other process than the current controlling process, {error, not_owner} is returned. so like, you can't just grab someone else's socket without permission, bestie. boundaries. π§
If the process identified by Pid is not an existing local pid/0, {error, badarg} is returned. {error, badarg} may also be returned in some cases when Socket is closed during the execution of this function. π
If the socket is in active mode, this function will transfer any messages from the socket in the mailbox of the caller to the new controlling process. okay so it also does a lil mailbox cleanup, generous.
If any other process is interacting with the socket during the transfer, it may not work correctly and messages may remain in the caller's mailbox. For instance, changing the sockets active mode during the transfer could cause this.
β οΈ so don't be juggling the socket while you hand it off, okay? touch nothing while the transfer is happening or messages get left behind. fr fr.
open/2 β the big one π
open(Port, Opts) -> {ok, Socket} | {error, Reason}
when
Port :: inet:port_number(),
Opts :: [inet:inet_backend() | open_option()],
Socket :: socket(),
Reason :: system_limit | inet:posix().
Open a UDP socket. this is the moment, the creation, the birth of a socket π¦
The created socket is bound to the UDP port number Port. If Port == 0, the underlying OS assigns a free (ephemeral) UDP port; use inet:port/1 to retrieve it. so 0 means "you pick for me, i trust you" β¨
The process that calls this function becomes the Socket's controlling process (socket owner). you made it, you own it, that's the rule. π
UDP socket options
listβ Received Packet is delivered as a list.binaryβ Received Packet is delivered as a binary.{ip, Address}β If the local host has many IP addresses, this option specifies which one to use.{ifaddr, Address}β Same as{ip, Address}.
However, if this instead is a socket:sockaddr_in/0 or socket:sockaddr_in6/0 this takes precedence over any value previously set with the ip options. If the ip option comes after the ifaddr option, it may be used to update its corresponding field of the ifaddr option (the addr field). order matters here, keep your eyes on it π
{fd, integer() >= 0}β If a socket has somehow been opened without usinggen_udp, use this option to pass the file descriptor for it. IfPortis not set to0and/or{ip, ip_address()}is combined with this option, thefdis bound to the specified interface and port after it is being opened. If these options are not specified, it is assumed that thefdis already bound appropriately.inet6β Sets up the socket for IPv6.inetβ Sets up the socket for IPv4.localβ Sets up a Unix Domain Socket. Seeinet:local_address/0{udp_module, module()}β Overrides which callback module is used. Defaults toinet_udpfor IPv4 andinet6_udpfor IPv6.{multicast_if, Address}β Sets the local device for a multicast socket.{multicast_loop, true | false}β Whentrue, sent multicast packets are looped back to the local sockets.{multicast_ttl, Integer}β Optionmulticast_ttlchanges the time-to-live (TTL) for outgoing multicast datagrams to control the scope of the multicasts.
Datagrams with a TTL of 1 are not forwarded beyond the local network. Defaults to 1.
{add_membership, {MultiAddress, InterfaceAddress}}β Joins a multicast group. welcome to the club π{drop_membership, {MultiAddress, InterfaceAddress}}β Leaves a multicast group. it's not you, it's the group πoption/0β Seeinet:setopts/2.
UDP packets are sent with this socket using send(Socket, ...). When UDP packets arrive to the Socket's UDP port, and the socket is in an active mode, the packets are delivered as messages to the controlling process (socket owner):
okay bestie, so the active-mode messages are giving main character energy and they land in your controlling process's inbox like this:
{udp, Socket, PeerIP, PeerPort, Packet} % Without ancillary data
{udp, Socket, PeerIP, PeerPort, AncData, Packet} % With ancillary data
PeerIP and PeerPort are the address from which Packet was sent, no cap. And Packet? It's a list of bytes ([byte/0] if option list is active and a binary/0 if option binary is active β and yeah, those two are mutually exclusive, so don't try to make them happen at the same time π
).
The message only has an AncData field if any of the socket options recvtos, recvtclass or recvttl are active. it's giving conditional, and we love that for us.
When a socket in {active, N} mode (see inet:setopts/2 for the deets) transitions to passive ({active, false}) mode β that's when N counts down to 0 β the controlling process gets a message on this form:
{udp_passive, Socket}
And if the OS protocol stack reports an error for the socket, this little drama lands in the controlling process's messages:
{udp_error, Socket, Reason}
Reason is mostly a POSIX Error Code. periodt.
If the socket is in passive mode (not in an active mode), received data can be retrieved with the recv/2,3 (recv/2) calls. But be warned bestie: incoming UDP packets that are longer than the receive buffer option specifies can be truncated without warning π. The default value for the receive buffer option is {recbuf, 9216}. remember that number.
Now let's get into recv:
recv(Socket, Length, Timeout) -> {ok, RecvData} | {error, Reason}
when
Socket :: socket(),
Length :: non_neg_integer(),
Timeout :: timeout(),
RecvData ::
{Address, Port, Packet} |
{Address, Port, AncData, Packet},
Address ::
inet:ip_address() |
inet:returned_non_ip_address(),
Port :: inet:port_number(),
AncData :: inet:ancillary_data(),
Packet :: string() | binary(),
Reason :: not_owner | timeout | inet:posix().
Receive a packet from a socket in passive mode. Timeout specifies a time-out in milliseconds. If any of the socket options recvtos, recvtclass or recvttl are active, the RecvData tuple contains an AncData field, otherwise it doesn't. slay.
Next up, the send girlies:
send(Socket, Packet) -> ok | {error, Reason}
when
Socket :: socket(),
Packet :: iodata(),
Reason :: not_owner | inet:posix().
Send a packet on a connected UDP socket. To connect a UDP socket, use connect/2 or connect/3. iykyk.
send(Socket, Destination, Packet) -> ok | {error, Reason}
when
Socket :: socket(),
Destination ::
{inet:ip_address(), inet:port_number()} |
inet:family_address() |
socket:sockaddr_in() |
socket:sockaddr_in6(),
Packet :: iodata(),
Reason :: not_owner | inet:posix().
Equivalent to send(Socket, Destination, [], Packet). simple as that.
Now the chameleon clause, send/4 with multiple forms:
send(Socket, Host, Port, Packet) -> ok | {error, Reason}
when
Socket :: socket(),
Host :: inet:hostname() | inet:ip_address(),
Port :: inet:port_number() | atom(),
Packet :: iodata(),
Reason :: not_owner | inet:posix();
(Socket, Destination, AncData, Packet) -> ok | {error, Reason}
when
Socket :: socket(),
Destination ::
{inet:ip_address(), inet:port_number()} |
inet:family_address() |
socket:sockaddr_in() |
socket:sockaddr_in6(),
AncData :: inet:ancillary_data(),
Packet :: iodata(),
Reason :: not_owner | inet:posix();
(Socket, Destination, PortZero, Packet) ->
ok | {error, Reason}
when
Socket :: socket(),
Destination ::
{inet:ip_address(), inet:port_number()} |
inet:family_address(),
PortZero :: inet:port_number(),
Packet :: iodata(),
Reason :: not_owner | inet:posix().
Send a UDP packet to the specified destination.
With arguments Host and Port: Argument Host can be a hostname or a socket address, and Port can be a port number or a service name atom. These are resolved to a Destination and then this function is equivalent to send(Socket, Destination, [], Packet) just below.
With arguments Destination and AncData (since OTP 22.1): Sends a packet to the specified Destination with ancillary data AncData.
NOTE: The ancillary dataAncDatacontains options that for this single message override the default options for the socket, an operation that may not be supported on all platforms, and if so return{error, einval}. Using more than one of an ancillary data item type may also not be supported.AncData =:= []is always supported.
With arguments Destination and PortZero (since OTP 22.1): Sends a packet to the specified Destination. Since Destination is a complete address, PortZero is redundant and has to be 0. This is a legacy clause mostly for Destination = {local, Binary} where PortZero is superfluous. Equivalent to send(Socket, Destination, [], Packet), right above here.
send(Socket, Host, Port, AncData, Packet) -> ok | {error, Reason}
when
Socket :: socket(),
Host ::
inet:hostname() |
inet:ip_address() |
inet:local_address(),
Port :: inet:port_number() | atom(),
AncData :: inet:ancillary_data(),
Packet :: iodata(),
Reason :: not_owner | inet:posix().
Send a packet to the specified destination, with ancillary data. Equivalent to send(Socket, Host, Port, Packet) regarding Host and Port and also equivalent to send(Socket, Destination, AncData, Packet) regarding the ancillary data: AncData. obsessed.
That's the whole glow-up from Ericsson AB, kernel 11.0.3, gen_udp(3). go off, bestie π
β¨
π¦ where this page comes from
- ubuntu stonking: erlang-doc 1:29.0.6+dfsg-1 Β· Apache-2.0
- π girlified by deepseek-ai/DeepSeek-V4.1-Flash on : a language model's rewrite of the original, under its terms