1SYSTEMD.SOCKET(5) systemd.socket SYSTEMD.SOCKET(5)
2
6 systemd.socket - Socket unit configuration
7
9 socket.socket
10
12 A unit configuration file whose name ends in ".socket" encodes
13 information about an IPC or network socket or a file system FIFO
14 controlled and supervised by systemd, for socket-based activation.
15 This man page lists the configuration options specific to this unit
17 type. See systemd.unit(5) for the common options of all unit
18 configuration files. The common configuration items are configured in
19 the generic [Unit] and [Install] sections. The socket specific
20 configuration options are configured in the [Socket] section.
21 Additional options are listed in systemd.exec(5), which define the
23 execution environment the ExecStartPre=, ExecStartPost=, ExecStopPre=
24 and ExecStopPost= commands are executed in, and in systemd.kill(5),
25 which define the way the processes are terminated, and in
26 systemd.resource-control(5), which configure resource control settings
27 for the processes of the socket.
28 For each socket file, a matching service file must exist, describing
30 the service to start on incoming traffic on the socket (see
31 systemd.service(5) for more information about .service files). The name
32 of the .service unit is by default the same as the name of the .socket
33 unit, but can be altered with the Service= option described below.
34 Depending on the setting of the Accept= option described below, this
35 .service unit must either be named like the .socket unit, but with the
36 suffix replaced, unless overridden with Service=; or it must be a
37 template unit named the same way. Example: a socket file foo.socket
38 needs a matching service foo.service if Accept=false is set. If
39 Accept=true is set, a service template file foo@.service must exist
40 from which services are instantiated for each incoming connection.
41 No implicit WantedBy= or RequiredBy= dependency from the socket to the
43 service is added. This means that the service may be started without
44 the socket, in which case it must be able to open sockets by itself. To
45 prevent this, an explicit Requires= dependency may be added.
46 Socket units may be used to implement on-demand starting of services,
48 as well as parallelized starting of services. See the blog stories
49 linked at the end for an introduction.
50 Note that the daemon software configured for socket activation with
52 socket units needs to be able to accept sockets from systemd, either
53 via systemd's native socket passing interface (see sd_listen_fds(3) for
54 details) or via the traditional inetd(8)-style socket passing (i.e.
55 sockets passed in via standard input and output, using
56 StandardInput=socket in the service file).
57
59 Implicit Dependencies
60 The following dependencies are implicitly added:
61 · Socket units automatically gain a Before= dependency on the service
63 units they activate.
64 · Socket units referring to file system paths (such as AF_UNIX
66 sockets or FIFOs) implicitly gain Requires= and After= dependencies
67 on all mount units necessary to access those paths.
68 · Socket units using the BindToDevice= setting automatically gain a
70 BindsTo= and After= dependency on the device unit encapsulating the
71 specified network interface.
72 Additional implicit dependencies may be added as result of execution
74 and resource control parameters as documented in systemd.exec(5) and
75 systemd.resource-control(5).
76 Default Dependencies
78 The following dependencies are added unless DefaultDependencies=no is
79 set:
80 · Socket units automatically gain a Before= dependency on
82 sockets.target.
83 · Socket units automatically gain a pair of After= and Requires=
85 dependency on sysinit.target, and a pair of Before= and Conflicts=
86 dependencies on shutdown.target. These dependencies ensure that the
87 socket unit is started before normal services at boot, and is
88 stopped on shutdown. Only sockets involved with early boot or late
89 system shutdown should disable DefaultDependencies= option.
90
92 Socket files must include a [Socket] section, which carries information
93 about the socket or FIFO it supervises. A number of options that may be
94 used in this section are shared with other unit types. These options
95 are documented in systemd.exec(5) and systemd.kill(5). The options
96 specific to the [Socket] section of socket units are the following:
97 ListenStream=, ListenDatagram=, ListenSequentialPacket=
99 Specifies an address to listen on for a stream (SOCK_STREAM),
100 datagram (SOCK_DGRAM), or sequential packet (SOCK_SEQPACKET)
101 socket, respectively. The address can be written in various
102 formats:
103 If the address starts with a slash ("/"), it is read as file system
105 socket in the AF_UNIX socket family.
106 If the address starts with an at symbol ("@"), it is read as
108 abstract namespace socket in the AF_UNIX family. The "@" is
109 replaced with a NUL character before binding. For details, see
110 unix(7).
111 If the address string is a single number, it is read as port number
113 to listen on via IPv6. Depending on the value of BindIPv6Only= (see
114 below) this might result in the service being available via both
115 IPv6 and IPv4 (default) or just via IPv6.
116 If the address string is a string in the format v.w.x.y:z, it is
118 read as IPv4 specifier for listening on an address v.w.x.y on a
119 port z.
120 If the address string is a string in the format [x]:y, it is read
122 as IPv6 address x on a port y. Note that this might make the
123 service available via IPv4, too, depending on the BindIPv6Only=
124 setting (see below).
125 If the address string is a string in the format "vsock:x:y", it is
127 read as CID "x" on a port "y" address in the AF_VSOCK family. The
128 CID is a unique 32-bit integer identifier in AF_VSOCK analogous to
129 an IP address. Specifying the CID is optional, and may be set to
130 the empty string.
131 Note that SOCK_SEQPACKET (i.e. ListenSequentialPacket=) is only
133 available for AF_UNIX sockets. SOCK_STREAM (i.e. ListenStream=)
134 when used for IP sockets refers to TCP sockets, SOCK_DGRAM (i.e.
135 ListenDatagram=) to UDP.
136 These options may be specified more than once, in which case
138 incoming traffic on any of the sockets will trigger service
139 activation, and all listed sockets will be passed to the service,
140 regardless of whether there is incoming traffic on them or not. If
141 the empty string is assigned to any of these options, the list of
142 addresses to listen on is reset, all prior uses of any of these
143 options will have no effect.
144 It is also possible to have more than one socket unit for the same
146 service when using Service=, and the service will receive all the
147 sockets configured in all the socket units. Sockets configured in
148 one unit are passed in the order of configuration, but no ordering
149 between socket units is specified.
150 If an IP address is used here, it is often desirable to listen on
152 it before the interface it is configured on is up and running, and
153 even regardless of whether it will be up and running at any point.
154 To deal with this, it is recommended to set the FreeBind= option
155 described below.
156 ListenFIFO=
158 Specifies a file system FIFO to listen on. This expects an absolute
159 file system path as argument. Behavior otherwise is very similar to
160 the ListenDatagram= directive above.
161 ListenSpecial=
163 Specifies a special file in the file system to listen on. This
164 expects an absolute file system path as argument. Behavior
165 otherwise is very similar to the ListenFIFO= directive above. Use
166 this to open character device nodes as well as special files in
167 /proc and /sys.
168 ListenNetlink=
170 Specifies a Netlink family to create a socket for to listen on.
171 This expects a short string referring to the AF_NETLINK family name
172 (such as audit or kobject-uevent) as argument, optionally suffixed
173 by a whitespace followed by a multicast group integer. Behavior
174 otherwise is very similar to the ListenDatagram= directive above.
175 ListenMessageQueue=
177 Specifies a POSIX message queue name to listen on. This expects a
178 valid message queue name (i.e. beginning with /). Behavior
179 otherwise is very similar to the ListenFIFO= directive above. On
180 Linux message queue descriptors are actually file descriptors and
181 can be inherited between processes.
182 ListenUSBFunction=
184 Specifies a USB FunctionFS[1] endpoints location to listen on, for
185 implementation of USB gadget functions. This expects an absolute
186 file system path of functionfs mount point as the argument.
187 Behavior otherwise is very similar to the ListenFIFO= directive
188 above. Use this to open the FunctionFS endpoint ep0. When using
189 this option, the activated service has to have the
190 USBFunctionDescriptors= and USBFunctionStrings= options set.
191 SocketProtocol=
193 Takes a one of udplite or sctp. Specifies a socket protocol
194 (IPPROTO_UDPLITE) UDP-Lite (IPPROTO_SCTP) SCTP socket respectively.
195 BindIPv6Only=
197 Takes a one of default, both or ipv6-only. Controls the IPV6_V6ONLY
198 socket option (see ipv6(7) for details). If both, IPv6 sockets
199 bound will be accessible via both IPv4 and IPv6. If ipv6-only, they
200 will be accessible via IPv6 only. If default (which is the default,
201 surprise!), the system wide default setting is used, as controlled
202 by /proc/sys/net/ipv6/bindv6only, which in turn defaults to the
203 equivalent of both.
204 Backlog=
206 Takes an unsigned integer argument. Specifies the number of
207 connections to queue that have not been accepted yet. This setting
208 matters only for stream and sequential packet sockets. See
209 listen(2) for details. Defaults to SOMAXCONN (128).
210 BindToDevice=
212 Specifies a network interface name to bind this socket to. If set,
213 traffic will only be accepted from the specified network
214 interfaces. This controls the SO_BINDTODEVICE socket option (see
215 socket(7) for details). If this option is used, an implicit
216 dependency from this socket unit on the network interface device
217 unit (systemd.device(5) is created. Note that setting this
218 parameter might result in additional dependencies to be added to
219 the unit (see above).
220 SocketUser=, SocketGroup=
222 Takes a UNIX user/group name. When specified, all AF_UNIX sockets
223 and FIFO nodes in the file system are owned by the specified user
224 and group. If unset (the default), the nodes are owned by the root
225 user/group (if run in system context) or the invoking user/group
226 (if run in user context). If only a user is specified but no group,
227 then the group is derived from the user's default group.
228 SocketMode=
230 If listening on a file system socket or FIFO, this option specifies
231 the file system access mode used when creating the file node. Takes
232 an access mode in octal notation. Defaults to 0666.
233 DirectoryMode=
235 If listening on a file system socket or FIFO, the parent
236 directories are automatically created if needed. This option
237 specifies the file system access mode used when creating these
238 directories. Takes an access mode in octal notation. Defaults to
239 0755.
240 Accept=
242 Takes a boolean argument. If true, a service instance is spawned
243 for each incoming connection and only the connection socket is
244 passed to it. If false, all listening sockets themselves are passed
245 to the started service unit, and only one service unit is spawned
246 for all connections (also see above). This value is ignored for
247 datagram sockets and FIFOs where a single service unit
248 unconditionally handles all incoming traffic. Defaults to false.
249 For performance reasons, it is recommended to write new daemons
250 only in a way that is suitable for Accept=false. A daemon listening
251 on an AF_UNIX socket may, but does not need to, call close(2) on
252 the received socket before exiting. However, it must not unlink the
253 socket from a file system. It should not invoke shutdown(2) on
254 sockets it got with Accept=false, but it may do so for sockets it
255 got with Accept=true set. Setting Accept=true is mostly useful to
256 allow daemons designed for usage with inetd(8) to work unmodified
257 with systemd socket activation.
258 For IPv4 and IPv6 connections, the REMOTE_ADDR environment variable
260 will contain the remote IP address, and REMOTE_PORT will contain
261 the remote port. This is the same as the format used by CGI. For
262 SOCK_RAW, the port is the IP protocol.
263 Writable=
265 Takes a boolean argument. May only be used in conjunction with
266 ListenSpecial=. If true, the specified special file is opened in
267 read-write mode, if false, in read-only mode. Defaults to false.
268 MaxConnections=
270 The maximum number of connections to simultaneously run services
271 instances for, when Accept=true is set. If more concurrent
272 connections are coming in, they will be refused until at least one
273 existing connection is terminated. This setting has no effect on
274 sockets configured with Accept=false or datagram sockets. Defaults
275 to 64.
276 MaxConnectionsPerSource=
278 The maximum number of connections for a service per source IP
279 address. This is very similar to the MaxConnections= directive
280 above. Disabled by default.
281 KeepAlive=
283 Takes a boolean argument. If true, the TCP/IP stack will send a
284 keep alive message after 2h (depending on the configuration of
285 /proc/sys/net/ipv4/tcp_keepalive_time) for all TCP streams accepted
286 on this socket. This controls the SO_KEEPALIVE socket option (see
287 socket(7) and the TCP Keepalive HOWTO[2] for details.) Defaults to
288 false.
289 KeepAliveTimeSec=
291 Takes time (in seconds) as argument. The connection needs to remain
292 idle before TCP starts sending keepalive probes. This controls the
293 TCP_KEEPIDLE socket option (see socket(7) and the TCP Keepalive
294 HOWTO[2] for details.) Defaults value is 7200 seconds (2 hours).
295 KeepAliveIntervalSec=
297 Takes time (in seconds) as argument between individual keepalive
298 probes, if the socket option SO_KEEPALIVE has been set on this
299 socket. This controls the TCP_KEEPINTVL socket option (see
300 socket(7) and the TCP Keepalive HOWTO[2] for details.) Defaults
301 value is 75 seconds.
302 KeepAliveProbes=
304 Takes an integer as argument. It is the number of unacknowledged
305 probes to send before considering the connection dead and notifying
306 the application layer. This controls the TCP_KEEPCNT socket option
307 (see socket(7) and the TCP Keepalive HOWTO[2] for details.)
308 Defaults value is 9.
309 NoDelay=
311 Takes a boolean argument. TCP Nagle's algorithm works by combining
312 a number of small outgoing messages, and sending them all at once.
313 This controls the TCP_NODELAY socket option (see tcp(7) Defaults to
314 false.
315 Priority=
317 Takes an integer argument controlling the priority for all traffic
318 sent from this socket. This controls the SO_PRIORITY socket option
319 (see socket(7) for details.).
320 DeferAcceptSec=
322 Takes time (in seconds) as argument. If set, the listening process
323 will be awakened only when data arrives on the socket, and not
324 immediately when connection is established. When this option is
325 set, the TCP_DEFER_ACCEPT socket option will be used (see tcp(7)),
326 and the kernel will ignore initial ACK packets without any data.
327 The argument specifies the approximate amount of time the kernel
328 should wait for incoming data before falling back to the normal
329 behavior of honoring empty ACK packets. This option is beneficial
330 for protocols where the client sends the data first (e.g. HTTP, in
331 contrast to SMTP), because the server process will not be woken up
332 unnecessarily before it can take any action.
333 If the client also uses the TCP_DEFER_ACCEPT option, the latency of
335 the initial connection may be reduced, because the kernel will send
336 data in the final packet establishing the connection (the third
337 packet in the "three-way handshake").
338 Disabled by default.
340 ReceiveBuffer=, SendBuffer=
342 Takes an integer argument controlling the receive or send buffer
343 sizes of this socket, respectively. This controls the SO_RCVBUF and
344 SO_SNDBUF socket options (see socket(7) for details.). The usual
345 suffixes K, M, G are supported and are understood to the base of
346 1024.
347 IPTOS=
349 Takes an integer argument controlling the IP Type-Of-Service field
350 for packets generated from this socket. This controls the IP_TOS
351 socket option (see ip(7) for details.). Either a numeric string or
352 one of low-delay, throughput, reliability or low-cost may be
353 specified.
354 IPTTL=
356 Takes an integer argument controlling the IPv4 Time-To-Live/IPv6
357 Hop-Count field for packets generated from this socket. This sets
358 the IP_TTL/IPV6_UNICAST_HOPS socket options (see ip(7) and ipv6(7)
359 for details.)
360 Mark=
362 Takes an integer value. Controls the firewall mark of packets
363 generated by this socket. This can be used in the firewall logic to
364 filter packets from this socket. This sets the SO_MARK socket
365 option. See iptables(8) for details.
366 ReusePort=
368 Takes a boolean value. If true, allows multiple bind(2)s to this
369 TCP or UDP port. This controls the SO_REUSEPORT socket option. See
370 socket(7) for details.
371 SmackLabel=, SmackLabelIPIn=, SmackLabelIPOut=
373 Takes a string value. Controls the extended attributes
374 "security.SMACK64", "security.SMACK64IPIN" and
375 "security.SMACK64IPOUT", respectively, i.e. the security label of
376 the FIFO, or the security label for the incoming or outgoing
377 connections of the socket, respectively. See Smack.txt[3] for
378 details.
379 SELinuxContextFromNet=
381 Takes a boolean argument. When true, systemd will attempt to figure
382 out the SELinux label used for the instantiated service from the
383 information handed by the peer over the network. Note that only the
384 security level is used from the information provided by the peer.
385 Other parts of the resulting SELinux context originate from either
386 the target binary that is effectively triggered by socket unit or
387 from the value of the SELinuxContext= option. This configuration
388 option only affects sockets with Accept= mode set to "true". Also
389 note that this option is useful only when MLS/MCS SELinux policy is
390 deployed. Defaults to "false".
391 PipeSize=
393 Takes a size in bytes. Controls the pipe buffer size of FIFOs
394 configured in this socket unit. See fcntl(2) for details. The usual
395 suffixes K, M, G are supported and are understood to the base of
396 1024.
397 MessageQueueMaxMessages=, MessageQueueMessageSize=
399 These two settings take integer values and control the mq_maxmsg
400 field or the mq_msgsize field, respectively, when creating the
401 message queue. Note that either none or both of these variables
402 need to be set. See mq_setattr(3) for details.
403 FreeBind=
405 Takes a boolean value. Controls whether the socket can be bound to
406 non-local IP addresses. This is useful to configure sockets
407 listening on specific IP addresses before those IP addresses are
408 successfully configured on a network interface. This sets the
409 IP_FREEBIND socket option. For robustness reasons it is recommended
410 to use this option whenever you bind a socket to a specific IP
411 address. Defaults to false.
412 Transparent=
414 Takes a boolean value. Controls the IP_TRANSPARENT socket option.
415 Defaults to false.
416 Broadcast=
418 Takes a boolean value. This controls the SO_BROADCAST socket
419 option, which allows broadcast datagrams to be sent from this
420 socket. Defaults to false.
421 PassCredentials=
423 Takes a boolean value. This controls the SO_PASSCRED socket option,
424 which allows AF_UNIX sockets to receive the credentials of the
425 sending process in an ancillary message. Defaults to false.
426 PassSecurity=
428 Takes a boolean value. This controls the SO_PASSSEC socket option,
429 which allows AF_UNIX sockets to receive the security context of the
430 sending process in an ancillary message. Defaults to false.
431 TCPCongestion=
433 Takes a string value. Controls the TCP congestion algorithm used by
434 this socket. Should be one of "westwood", "veno", "cubic", "lp" or
435 any other available algorithm supported by the IP stack. This
436 setting applies only to stream sockets.
437 ExecStartPre=, ExecStartPost=
439 Takes one or more command lines, which are executed before or after
440 the listening sockets/FIFOs are created and bound, respectively.
441 The first token of the command line must be an absolute filename,
442 then followed by arguments for the process. Multiple command lines
443 may be specified following the same scheme as used for
444 ExecStartPre= of service unit files.
445 ExecStopPre=, ExecStopPost=
447 Additional commands that are executed before or after the listening
448 sockets/FIFOs are closed and removed, respectively. Multiple
449 command lines may be specified following the same scheme as used
450 for ExecStartPre= of service unit files.
451 TimeoutSec=
453 Configures the time to wait for the commands specified in
454 ExecStartPre=, ExecStartPost=, ExecStopPre= and ExecStopPost= to
455 finish. If a command does not exit within the configured time, the
456 socket will be considered failed and be shut down again. All
457 commands still running will be terminated forcibly via SIGTERM, and
458 after another delay of this time with SIGKILL. (See KillMode= in
459 systemd.kill(5).) Takes a unit-less value in seconds, or a time
460 span value such as "5min 20s". Pass "0" to disable the timeout
461 logic. Defaults to DefaultTimeoutStartSec= from the manager
462 configuration file (see systemd-system.conf(5)).
463 Service=
465 Specifies the service unit name to activate on incoming traffic.
466 This setting is only allowed for sockets with Accept=no. It
467 defaults to the service that bears the same name as the socket
468 (with the suffix replaced). In most cases, it should not be
469 necessary to use this option. Note that setting this parameter
470 might result in additional dependencies to be added to the unit
471 (see above).
472 RemoveOnStop=
474 Takes a boolean argument. If enabled, any file nodes created by
475 this socket unit are removed when it is stopped. This applies to
476 AF_UNIX sockets in the file system, POSIX message queues, FIFOs, as
477 well as any symlinks to them configured with Symlinks=. Normally,
478 it should not be necessary to use this option, and is not
479 recommended as services might continue to run after the socket unit
480 has been terminated and it should still be possible to communicate
481 with them via their file system node. Defaults to off.
482 Symlinks=
484 Takes a list of file system paths. The specified paths will be
485 created as symlinks to the AF_UNIX socket path or FIFO path of this
486 socket unit. If this setting is used, only one AF_UNIX socket in
487 the file system or one FIFO may be configured for the socket unit.
488 Use this option to manage one or more symlinked alias names for a
489 socket, binding their lifecycle together. Note that if creation of
490 a symlink fails this is not considered fatal for the socket unit,
491 and the socket unit may still start. If an empty string is
492 assigned, the list of paths is reset. Defaults to an empty list.
493 FileDescriptorName=
495 Assigns a name to all file descriptors this socket unit
496 encapsulates. This is useful to help activated services identify
497 specific file descriptors, if multiple fds are passed. Services may
498 use the sd_listen_fds_with_names(3) call to acquire the names
499 configured for the received file descriptors. Names may contain any
500 ASCII character, but must exclude control characters and ":", and
501 must be at most 255 characters in length. If this setting is not
502 used, the file descriptor name defaults to the name of the socket
503 unit, including its .socket suffix.
504 TriggerLimitIntervalSec=, TriggerLimitBurst=
506 Configures a limit on how often this socket unit my be activated
507 within a specific time interval. The TriggerLimitIntervalSec= may
508 be used to configure the length of the time interval in the usual
509 time units "us", "ms", "s", "min", "h", ... and defaults to 2s (See
510 systemd.time(7) for details on the various time units understood).
511 The TriggerLimitBurst= setting takes a positive integer value and
512 specifies the number of permitted activations per time interval,
513 and defaults to 200 for Accept=yes sockets (thus by default
514 permitting 200 activations per 2s), and 20 otherwise (20
515 activations per 2s). Set either to 0 to disable any form of trigger
516 rate limiting. If the limit is hit, the socket unit is placed into
517 a failure mode, and will not be connectible anymore until
518 restarted. Note that this limit is enforced before the service
519 activation is enqueued.
520 Check systemd.exec(5) and systemd.kill(5) for more settings.
522
524 systemd(1), systemctl(1), systemd.unit(5), systemd.exec(5),
525 systemd.kill(5), systemd.resource-control(5), systemd.service(5),
526 systemd.directives(7), sd_listen_fds(3), sd_listen_fds_with_names(3)
527 For more extensive descriptions see the "systemd for Developers"
529 series: Socket Activation[4], Socket Activation, part II[5], Converting
530 inetd Services[6], Socket Activated Internet Services and OS
531 Containers[7].
532
534 1. USB FunctionFS
535 https://www.kernel.org/doc/Documentation/usb/functionfs.txt
536 2. TCP Keepalive HOWTO
538 http://www.tldp.org/HOWTO/html_single/TCP-Keepalive-HOWTO/
539 3. Smack.txt
541 https://www.kernel.org/doc/Documentation/security/Smack.txt
542 4. Socket Activation
544 http://0pointer.de/blog/projects/socket-activation.html
545 5. Socket Activation, part II
547 http://0pointer.de/blog/projects/socket-activation2.html
548 6. Converting inetd Services
550 http://0pointer.de/blog/projects/inetd.html
551 7. Socket Activated Internet Services and OS Containers
553 http://0pointer.de/blog/projects/socket-activated-containers.html
554systemd 239 SYSTEMD.SOCKET(5)