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 unit, a matching service unit must exist, describing
30 the service to start on incoming traffic on the socket (see
31 systemd.service(5) for more information about .service units). 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=no is set. If Accept=yes
39 is set, a service template foo@.service must exist from which services
40 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 about the precise protocol used and the order in which the file
55 descriptors are passed) or via traditional inetd(8)-style socket
56 passing (i.e. sockets passed in via standard input and output, using
57 StandardInput=socket in the service file).
58 All network sockets allocated through .socket units are allocated in
60 the host's network namespace (see network_namespaces(7)). This does not
61 mean however that the service activated by a configured socket unit has
62 to be part of the host's network namespace as well. It is supported and
63 even good practice to run services in their own network namespace (for
64 example through PrivateNetwork=, see systemd.exec(5)), receiving only
65 the sockets configured through socket-activation from the host's
66 namespace. In such a set-up communication within the host's network
67 namespace is only permitted through the activation sockets passed in
68 while all sockets allocated from the service code itself will be
69 associated with the service's own namespace, and thus possibly subject
70 to a a much more restrictive configuration.
71
73 Implicit Dependencies
74 The following dependencies are implicitly added:
75 · Socket units automatically gain a Before= dependency on the service
77 units they activate.
78 · Socket units referring to file system paths (such as AF_UNIX
80 sockets or FIFOs) implicitly gain Requires= and After= dependencies
81 on all mount units necessary to access those paths.
82 · Socket units using the BindToDevice= setting automatically gain a
84 BindsTo= and After= dependency on the device unit encapsulating the
85 specified network interface.
86 Additional implicit dependencies may be added as result of execution
88 and resource control parameters as documented in systemd.exec(5) and
89 systemd.resource-control(5).
90 Default Dependencies
92 The following dependencies are added unless DefaultDependencies=no is
93 set:
94 · Socket units automatically gain a Before= dependency on
96 sockets.target.
97 · Socket units automatically gain a pair of After= and Requires=
99 dependency on sysinit.target, and a pair of Before= and Conflicts=
100 dependencies on shutdown.target. These dependencies ensure that the
101 socket unit is started before normal services at boot, and is
102 stopped on shutdown. Only sockets involved with early boot or late
103 system shutdown should disable DefaultDependencies= option.
104
106 Socket files must include a [Socket] section, which carries information
107 about the socket or FIFO it supervises. A number of options that may be
108 used in this section are shared with other unit types. These options
109 are documented in systemd.exec(5) and systemd.kill(5). The options
110 specific to the [Socket] section of socket units are the following:
111 ListenStream=, ListenDatagram=, ListenSequentialPacket=
113 Specifies an address to listen on for a stream (SOCK_STREAM),
114 datagram (SOCK_DGRAM), or sequential packet (SOCK_SEQPACKET)
115 socket, respectively. The address can be written in various
116 formats:
117 If the address starts with a slash ("/"), it is read as file system
119 socket in the AF_UNIX socket family.
120 If the address starts with an at symbol ("@"), it is read as
122 abstract namespace socket in the AF_UNIX family. The "@" is
123 replaced with a NUL character before binding. For details, see
124 unix(7).
125 If the address string is a single number, it is read as port number
127 to listen on via IPv6. Depending on the value of BindIPv6Only= (see
128 below) this might result in the service being available via both
129 IPv6 and IPv4 (default) or just via IPv6.
130 If the address string is a string in the format v.w.x.y:z, it is
132 read as IPv4 specifier for listening on an address v.w.x.y on a
133 port z.
134 If the address string is a string in the format [x]:y, it is read
136 as IPv6 address x on a port y. Note that this might make the
137 service available via IPv4, too, depending on the BindIPv6Only=
138 setting (see below).
139 If the address string is a string in the format "vsock:x:y", it is
141 read as CID "x" on a port "y" address in the AF_VSOCK family. The
142 CID is a unique 32-bit integer identifier in AF_VSOCK analogous to
143 an IP address. Specifying the CID is optional, and may be set to
144 the empty string.
145 Note that SOCK_SEQPACKET (i.e. ListenSequentialPacket=) is only
147 available for AF_UNIX sockets. SOCK_STREAM (i.e. ListenStream=)
148 when used for IP sockets refers to TCP sockets, SOCK_DGRAM (i.e.
149 ListenDatagram=) to UDP.
150 These options may be specified more than once, in which case
152 incoming traffic on any of the sockets will trigger service
153 activation, and all listed sockets will be passed to the service,
154 regardless of whether there is incoming traffic on them or not. If
155 the empty string is assigned to any of these options, the list of
156 addresses to listen on is reset, all prior uses of any of these
157 options will have no effect.
158 It is also possible to have more than one socket unit for the same
160 service when using Service=, and the service will receive all the
161 sockets configured in all the socket units. Sockets configured in
162 one unit are passed in the order of configuration, but no ordering
163 between socket units is specified.
164 If an IP address is used here, it is often desirable to listen on
166 it before the interface it is configured on is up and running, and
167 even regardless of whether it will be up and running at any point.
168 To deal with this, it is recommended to set the FreeBind= option
169 described below.
170 ListenFIFO=
172 Specifies a file system FIFO to listen on. This expects an absolute
173 file system path as argument. Behavior otherwise is very similar to
174 the ListenDatagram= directive above.
175 ListenSpecial=
177 Specifies a special file in the file system to listen on. This
178 expects an absolute file system path as argument. Behavior
179 otherwise is very similar to the ListenFIFO= directive above. Use
180 this to open character device nodes as well as special files in
181 /proc and /sys.
182 ListenNetlink=
184 Specifies a Netlink family to create a socket for to listen on.
185 This expects a short string referring to the AF_NETLINK family name
186 (such as audit or kobject-uevent) as argument, optionally suffixed
187 by a whitespace followed by a multicast group integer. Behavior
188 otherwise is very similar to the ListenDatagram= directive above.
189 ListenMessageQueue=
191 Specifies a POSIX message queue name to listen on. This expects a
192 valid message queue name (i.e. beginning with /). Behavior
193 otherwise is very similar to the ListenFIFO= directive above. On
194 Linux message queue descriptors are actually file descriptors and
195 can be inherited between processes.
196 ListenUSBFunction=
198 Specifies a USB FunctionFS[1] endpoints location to listen on, for
199 implementation of USB gadget functions. This expects an absolute
200 file system path of FunctionFS mount point as the argument.
201 Behavior otherwise is very similar to the ListenFIFO= directive
202 above. Use this to open the FunctionFS endpoint ep0. When using
203 this option, the activated service has to have the
204 USBFunctionDescriptors= and USBFunctionStrings= options set.
205 SocketProtocol=
207 Takes one of udplite or sctp. The socket will use the UDP-Lite
208 (IPPROTO_UDPLITE) or SCTP (IPPROTO_SCTP) protocol, respectively.
209 BindIPv6Only=
211 Takes one of default, both or ipv6-only. Controls the IPV6_V6ONLY
212 socket option (see ipv6(7) for details). If both, IPv6 sockets
213 bound will be accessible via both IPv4 and IPv6. If ipv6-only, they
214 will be accessible via IPv6 only. If default (which is the default,
215 surprise!), the system wide default setting is used, as controlled
216 by /proc/sys/net/ipv6/bindv6only, which in turn defaults to the
217 equivalent of both.
218 Backlog=
220 Takes an unsigned integer argument. Specifies the number of
221 connections to queue that have not been accepted yet. This setting
222 matters only for stream and sequential packet sockets. See
223 listen(2) for details. Defaults to SOMAXCONN (128).
224 BindToDevice=
226 Specifies a network interface name to bind this socket to. If set,
227 traffic will only be accepted from the specified network
228 interfaces. This controls the SO_BINDTODEVICE socket option (see
229 socket(7) for details). If this option is used, an implicit
230 dependency from this socket unit on the network interface device
231 unit is created (see systemd.device(5)). Note that setting this
232 parameter might result in additional dependencies to be added to
233 the unit (see above).
234 SocketUser=, SocketGroup=
236 Takes a UNIX user/group name. When specified, all AF_UNIX sockets
237 and FIFO nodes in the file system are owned by the specified user
238 and group. If unset (the default), the nodes are owned by the root
239 user/group (if run in system context) or the invoking user/group
240 (if run in user context). If only a user is specified but no group,
241 then the group is derived from the user's default group.
242 SocketMode=
244 If listening on a file system socket or FIFO, this option specifies
245 the file system access mode used when creating the file node. Takes
246 an access mode in octal notation. Defaults to 0666.
247 DirectoryMode=
249 If listening on a file system socket or FIFO, the parent
250 directories are automatically created if needed. This option
251 specifies the file system access mode used when creating these
252 directories. Takes an access mode in octal notation. Defaults to
253 0755.
254 Accept=
256 Takes a boolean argument. If true, a service instance is spawned
257 for each incoming connection and only the connection socket is
258 passed to it. If false, all listening sockets themselves are passed
259 to the started service unit, and only one service unit is spawned
260 for all connections (also see above). This value is ignored for
261 datagram sockets and FIFOs where a single service unit
262 unconditionally handles all incoming traffic. Defaults to false.
263 For performance reasons, it is recommended to write new daemons
264 only in a way that is suitable for Accept=no. A daemon listening on
265 an AF_UNIX socket may, but does not need to, call close(2) on the
266 received socket before exiting. However, it must not unlink the
267 socket from a file system. It should not invoke shutdown(2) on
268 sockets it got with Accept=no, but it may do so for sockets it got
269 with Accept=yes set. Setting Accept=yes is mostly useful to allow
270 daemons designed for usage with inetd(8) to work unmodified with
271 systemd socket activation.
272 For IPv4 and IPv6 connections, the REMOTE_ADDR environment variable
274 will contain the remote IP address, and REMOTE_PORT will contain
275 the remote port. This is the same as the format used by CGI. For
276 SOCK_RAW, the port is the IP protocol.
277 Writable=
279 Takes a boolean argument. May only be used in conjunction with
280 ListenSpecial=. If true, the specified special file is opened in
281 read-write mode, if false, in read-only mode. Defaults to false.
282 MaxConnections=
284 The maximum number of connections to simultaneously run services
285 instances for, when Accept=yes is set. If more concurrent
286 connections are coming in, they will be refused until at least one
287 existing connection is terminated. This setting has no effect on
288 sockets configured with Accept=no or datagram sockets. Defaults to
289 64.
290 MaxConnectionsPerSource=
292 The maximum number of connections for a service per source IP
293 address. This is very similar to the MaxConnections= directive
294 above. Disabled by default.
295 KeepAlive=
297 Takes a boolean argument. If true, the TCP/IP stack will send a
298 keep alive message after 2h (depending on the configuration of
299 /proc/sys/net/ipv4/tcp_keepalive_time) for all TCP streams accepted
300 on this socket. This controls the SO_KEEPALIVE socket option (see
301 socket(7) and the TCP Keepalive HOWTO[2] for details.) Defaults to
302 false.
303 KeepAliveTimeSec=
305 Takes time (in seconds) as argument. The connection needs to remain
306 idle before TCP starts sending keepalive probes. This controls the
307 TCP_KEEPIDLE socket option (see socket(7) and the TCP Keepalive
308 HOWTO[2] for details.) Defaults value is 7200 seconds (2 hours).
309 KeepAliveIntervalSec=
311 Takes time (in seconds) as argument between individual keepalive
312 probes, if the socket option SO_KEEPALIVE has been set on this
313 socket. This controls the TCP_KEEPINTVL socket option (see
314 socket(7) and the TCP Keepalive HOWTO[2] for details.) Defaults
315 value is 75 seconds.
316 KeepAliveProbes=
318 Takes an integer as argument. It is the number of unacknowledged
319 probes to send before considering the connection dead and notifying
320 the application layer. This controls the TCP_KEEPCNT socket option
321 (see socket(7) and the TCP Keepalive HOWTO[2] for details.)
322 Defaults value is 9.
323 NoDelay=
325 Takes a boolean argument. TCP Nagle's algorithm works by combining
326 a number of small outgoing messages, and sending them all at once.
327 This controls the TCP_NODELAY socket option (see tcp(7)). Defaults
328 to false.
329 Priority=
331 Takes an integer argument controlling the priority for all traffic
332 sent from this socket. This controls the SO_PRIORITY socket option
333 (see socket(7) for details.).
334 DeferAcceptSec=
336 Takes time (in seconds) as argument. If set, the listening process
337 will be awakened only when data arrives on the socket, and not
338 immediately when connection is established. When this option is
339 set, the TCP_DEFER_ACCEPT socket option will be used (see tcp(7)),
340 and the kernel will ignore initial ACK packets without any data.
341 The argument specifies the approximate amount of time the kernel
342 should wait for incoming data before falling back to the normal
343 behavior of honoring empty ACK packets. This option is beneficial
344 for protocols where the client sends the data first (e.g. HTTP, in
345 contrast to SMTP), because the server process will not be woken up
346 unnecessarily before it can take any action.
347 If the client also uses the TCP_DEFER_ACCEPT option, the latency of
349 the initial connection may be reduced, because the kernel will send
350 data in the final packet establishing the connection (the third
351 packet in the "three-way handshake").
352 Disabled by default.
354 ReceiveBuffer=, SendBuffer=
356 Takes an integer argument controlling the receive or send buffer
357 sizes of this socket, respectively. This controls the SO_RCVBUF and
358 SO_SNDBUF socket options (see socket(7) for details.). The usual
359 suffixes K, M, G are supported and are understood to the base of
360 1024.
361 IPTOS=
363 Takes an integer argument controlling the IP Type-Of-Service field
364 for packets generated from this socket. This controls the IP_TOS
365 socket option (see ip(7) for details.). Either a numeric string or
366 one of low-delay, throughput, reliability or low-cost may be
367 specified.
368 IPTTL=
370 Takes an integer argument controlling the IPv4 Time-To-Live/IPv6
371 Hop-Count field for packets generated from this socket. This sets
372 the IP_TTL/IPV6_UNICAST_HOPS socket options (see ip(7) and ipv6(7)
373 for details.)
374 Mark=
376 Takes an integer value. Controls the firewall mark of packets
377 generated by this socket. This can be used in the firewall logic to
378 filter packets from this socket. This sets the SO_MARK socket
379 option. See iptables(8) for details.
380 ReusePort=
382 Takes a boolean value. If true, allows multiple bind(2)s to this
383 TCP or UDP port. This controls the SO_REUSEPORT socket option. See
384 socket(7) for details.
385 SmackLabel=, SmackLabelIPIn=, SmackLabelIPOut=
387 Takes a string value. Controls the extended attributes
388 "security.SMACK64", "security.SMACK64IPIN" and
389 "security.SMACK64IPOUT", respectively, i.e. the security label of
390 the FIFO, or the security label for the incoming or outgoing
391 connections of the socket, respectively. See Smack.txt[3] for
392 details.
393 SELinuxContextFromNet=
395 Takes a boolean argument. When true, systemd will attempt to figure
396 out the SELinux label used for the instantiated service from the
397 information handed by the peer over the network. Note that only the
398 security level is used from the information provided by the peer.
399 Other parts of the resulting SELinux context originate from either
400 the target binary that is effectively triggered by socket unit or
401 from the value of the SELinuxContext= option. This configuration
402 option only affects sockets with Accept= mode set to "true". Also
403 note that this option is useful only when MLS/MCS SELinux policy is
404 deployed. Defaults to "false".
405 PipeSize=
407 Takes a size in bytes. Controls the pipe buffer size of FIFOs
408 configured in this socket unit. See fcntl(2) for details. The usual
409 suffixes K, M, G are supported and are understood to the base of
410 1024.
411 MessageQueueMaxMessages=, MessageQueueMessageSize=
413 These two settings take integer values and control the mq_maxmsg
414 field or the mq_msgsize field, respectively, when creating the
415 message queue. Note that either none or both of these variables
416 need to be set. See mq_setattr(3) for details.
417 FreeBind=
419 Takes a boolean value. Controls whether the socket can be bound to
420 non-local IP addresses. This is useful to configure sockets
421 listening on specific IP addresses before those IP addresses are
422 successfully configured on a network interface. This sets the
423 IP_FREEBIND socket option. For robustness reasons it is recommended
424 to use this option whenever you bind a socket to a specific IP
425 address. Defaults to false.
426 Transparent=
428 Takes a boolean value. Controls the IP_TRANSPARENT socket option.
429 Defaults to false.
430 Broadcast=
432 Takes a boolean value. This controls the SO_BROADCAST socket
433 option, which allows broadcast datagrams to be sent from this
434 socket. Defaults to false.
435 PassCredentials=
437 Takes a boolean value. This controls the SO_PASSCRED socket option,
438 which allows AF_UNIX sockets to receive the credentials of the
439 sending process in an ancillary message. Defaults to false.
440 PassSecurity=
442 Takes a boolean value. This controls the SO_PASSSEC socket option,
443 which allows AF_UNIX sockets to receive the security context of the
444 sending process in an ancillary message. Defaults to false.
445 PassPacketInfo=
447 Takes a boolean value. This controls the IP_PKTINFO,
448 IPV6_RECVPKTINFO and NETLINK_PKTINFO socket options, which enable
449 reception of additional per-packet metadata as ancillary message,
450 on AF_INET, AF_INET6 and AF_UNIX sockets. Defaults to false.
451 TCPCongestion=
453 Takes a string value. Controls the TCP congestion algorithm used by
454 this socket. Should be one of "westwood", "veno", "cubic", "lp" or
455 any other available algorithm supported by the IP stack. This
456 setting applies only to stream sockets.
457 ExecStartPre=, ExecStartPost=
459 Takes one or more command lines, which are executed before or after
460 the listening sockets/FIFOs are created and bound, respectively.
461 The first token of the command line must be an absolute filename,
462 then followed by arguments for the process. Multiple command lines
463 may be specified following the same scheme as used for
464 ExecStartPre= of service unit files.
465 ExecStopPre=, ExecStopPost=
467 Additional commands that are executed before or after the listening
468 sockets/FIFOs are closed and removed, respectively. Multiple
469 command lines may be specified following the same scheme as used
470 for ExecStartPre= of service unit files.
471 TimeoutSec=
473 Configures the time to wait for the commands specified in
474 ExecStartPre=, ExecStartPost=, ExecStopPre= and ExecStopPost= to
475 finish. If a command does not exit within the configured time, the
476 socket will be considered failed and be shut down again. All
477 commands still running will be terminated forcibly via SIGTERM, and
478 after another delay of this time with SIGKILL. (See KillMode= in
479 systemd.kill(5).) Takes a unit-less value in seconds, or a time
480 span value such as "5min 20s". Pass "0" to disable the timeout
481 logic. Defaults to DefaultTimeoutStartSec= from the manager
482 configuration file (see systemd-system.conf(5)).
483 Service=
485 Specifies the service unit name to activate on incoming traffic.
486 This setting is only allowed for sockets with Accept=no. It
487 defaults to the service that bears the same name as the socket
488 (with the suffix replaced). In most cases, it should not be
489 necessary to use this option. Note that setting this parameter
490 might result in additional dependencies to be added to the unit
491 (see above).
492 RemoveOnStop=
494 Takes a boolean argument. If enabled, any file nodes created by
495 this socket unit are removed when it is stopped. This applies to
496 AF_UNIX sockets in the file system, POSIX message queues, FIFOs, as
497 well as any symlinks to them configured with Symlinks=. Normally,
498 it should not be necessary to use this option, and is not
499 recommended as services might continue to run after the socket unit
500 has been terminated and it should still be possible to communicate
501 with them via their file system node. Defaults to off.
502 Symlinks=
504 Takes a list of file system paths. The specified paths will be
505 created as symlinks to the AF_UNIX socket path or FIFO path of this
506 socket unit. If this setting is used, only one AF_UNIX socket in
507 the file system or one FIFO may be configured for the socket unit.
508 Use this option to manage one or more symlinked alias names for a
509 socket, binding their lifecycle together. Note that if creation of
510 a symlink fails this is not considered fatal for the socket unit,
511 and the socket unit may still start. If an empty string is
512 assigned, the list of paths is reset. Defaults to an empty list.
513 FileDescriptorName=
515 Assigns a name to all file descriptors this socket unit
516 encapsulates. This is useful to help activated services identify
517 specific file descriptors, if multiple fds are passed. Services may
518 use the sd_listen_fds_with_names(3) call to acquire the names
519 configured for the received file descriptors. Names may contain any
520 ASCII character, but must exclude control characters and ":", and
521 must be at most 255 characters in length. If this setting is not
522 used, the file descriptor name defaults to the name of the socket
523 unit, including its .socket suffix.
524 TriggerLimitIntervalSec=, TriggerLimitBurst=
526 Configures a limit on how often this socket unit my be activated
527 within a specific time interval. The TriggerLimitIntervalSec= may
528 be used to configure the length of the time interval in the usual
529 time units "us", "ms", "s", "min", "h", ... and defaults to 2s (See
530 systemd.time(7) for details on the various time units understood).
531 The TriggerLimitBurst= setting takes a positive integer value and
532 specifies the number of permitted activations per time interval,
533 and defaults to 200 for Accept=yes sockets (thus by default
534 permitting 200 activations per 2s), and 20 otherwise (20
535 activations per 2s). Set either to 0 to disable any form of trigger
536 rate limiting. If the limit is hit, the socket unit is placed into
537 a failure mode, and will not be connectible anymore until
538 restarted. Note that this limit is enforced before the service
539 activation is enqueued.
540 Check systemd.exec(5) and systemd.kill(5) for more settings.
542
544 systemd(1), systemctl(1), systemd-system.conf(5), systemd.unit(5),
545 systemd.exec(5), systemd.kill(5), systemd.resource-control(5),
546 systemd.service(5), systemd.directives(7), sd_listen_fds(3),
547 sd_listen_fds_with_names(3)
548 For more extensive descriptions see the "systemd for Developers"
550 series: Socket Activation[4], Socket Activation, part II[5], Converting
551 inetd Services[6], Socket Activated Internet Services and OS
552 Containers[7].
553
555 1. USB FunctionFS
556 https://www.kernel.org/doc/Documentation/usb/functionfs.txt
557 2. TCP Keepalive HOWTO
559 http://www.tldp.org/HOWTO/html_single/TCP-Keepalive-HOWTO/
560 3. Smack.txt
562 https://www.kernel.org/doc/Documentation/security/Smack.txt
563 4. Socket Activation
565 http://0pointer.de/blog/projects/socket-activation.html
566 5. Socket Activation, part II
568 http://0pointer.de/blog/projects/socket-activation2.html
569 6. Converting inetd Services
571 http://0pointer.de/blog/projects/inetd.html
572 7. Socket Activated Internet Services and OS Containers
574 http://0pointer.de/blog/projects/socket-activated-containers.html
575systemd 246 SYSTEMD.SOCKET(5)