1DAEMON(7) daemon DAEMON(7)
2
6 daemon - Writing and packaging system daemons
7
9 A daemon is a service process that runs in the background and
10 supervises the system or provides functionality to other processes.
11 Traditionally, daemons are implemented following a scheme originating
12 in SysV Unix. Modern daemons should follow a simpler yet more powerful
13 scheme (here called "new-style" daemons), as implemented by systemd(1).
14 This manual page covers both schemes, and in particular includes
15 recommendations for daemons that shall be included in the systemd init
16 system.
17 SysV Daemons
19 When a traditional SysV daemon starts, it should execute the following
20 steps as part of the initialization. Note that these steps are
21 unnecessary for new-style daemons (see below), and should only be
22 implemented if compatibility with SysV is essential.
23 1. Close all open file descriptors except standard input, output, and
25 error (i.e. the first three file descriptors 0, 1, 2). This ensures
26 that no accidentally passed file descriptor stays around in the
27 daemon process. On Linux, this is best implemented by iterating
28 through /proc/self/fd, with a fallback of iterating from file
29 descriptor 3 to the value returned by getrlimit() for
30 RLIMIT_NOFILE.
31 2. Reset all signal handlers to their default. This is best done by
33 iterating through the available signals up to the limit of _NSIG
34 and resetting them to SIG_DFL.
35 3. Reset the signal mask using sigprocmask().
37 4. Sanitize the environment block, removing or resetting environment
39 variables that might negatively impact daemon runtime.
40 5. Call fork(), to create a background process.
42 6. In the child, call setsid() to detach from any terminal and create
44 an independent session.
45 7. In the child, call fork() again, to ensure that the daemon can
47 never re-acquire a terminal again. (This relevant if the program —
48 and all its dependencies — does not carefully specify `O_NOCTTY` on
49 each and every single `open()` call that might potentially open a
50 TTY device node.)
51 8. Call exit() in the first child, so that only the second child (the
53 actual daemon process) stays around. This ensures that the daemon
54 process is re-parented to init/PID 1, as all daemons should be.
55 9. In the daemon process, connect /dev/null to standard input, output,
57 and error.
58 10. In the daemon process, reset the umask to 0, so that the file modes
60 passed to open(), mkdir() and suchlike directly control the access
61 mode of the created files and directories.
62 11. In the daemon process, change the current directory to the root
64 directory (/), in order to avoid that the daemon involuntarily
65 blocks mount points from being unmounted.
66 12. In the daemon process, write the daemon PID (as returned by
68 getpid()) to a PID file, for example /run/foobar.pid (for a
69 hypothetical daemon "foobar") to ensure that the daemon cannot be
70 started more than once. This must be implemented in race-free
71 fashion so that the PID file is only updated when it is verified at
72 the same time that the PID previously stored in the PID file no
73 longer exists or belongs to a foreign process.
74 13. In the daemon process, drop privileges, if possible and applicable.
76 14. From the daemon process, notify the original process started that
78 initialization is complete. This can be implemented via an unnamed
79 pipe or similar communication channel that is created before the
80 first fork() and hence available in both the original and the
81 daemon process.
82 15. Call exit() in the original process. The process that invoked the
84 daemon must be able to rely on that this exit() happens after
85 initialization is complete and all external communication channels
86 are established and accessible.
87 The BSD daemon() function should not be used, as it implements only a
89 subset of these steps.
90 A daemon that needs to provide compatibility with SysV systems should
92 implement the scheme pointed out above. However, it is recommended to
93 make this behavior optional and configurable via a command line
94 argument to ease debugging as well as to simplify integration into
95 systems using systemd.
96 New-Style Daemons
98 Modern services for Linux should be implemented as new-style daemons.
99 This makes it easier to supervise and control them at runtime and
100 simplifies their implementation.
101 For developing a new-style daemon, none of the initialization steps
103 recommended for SysV daemons need to be implemented. New-style init
104 systems such as systemd make all of them redundant. Moreover, since
105 some of these steps interfere with process monitoring, file descriptor
106 passing and other functionality of the init system, it is recommended
107 not to execute them when run as new-style service.
108 Note that new-style init systems guarantee execution of daemon
110 processes in a clean process context: it is guaranteed that the
111 environment block is sanitized, that the signal handlers and mask is
112 reset and that no left-over file descriptors are passed. Daemons will
113 be executed in their own session, with standard input connected to
114 /dev/null and standard output/error connected to the systemd-
115 journald.service(8) logging service, unless otherwise configured. The
116 umask is reset.
117 It is recommended for new-style daemons to implement the following:
119 1. If SIGTERM is received, shut down the daemon and exit cleanly.
121 2. If SIGHUP is received, reload the configuration files, if this
123 applies.
124 3. Provide a correct exit code from the main daemon process, as this
126 is used by the init system to detect service errors and problems.
127 It is recommended to follow the exit code scheme as defined in the
128 LSB recommendations for SysV init scripts[1].
129 4. If possible and applicable, expose the daemon's control interface
131 via the D-Bus IPC system and grab a bus name as last step of
132 initialization.
133 5. For integration in systemd, provide a .service unit file that
135 carries information about starting, stopping and otherwise
136 maintaining the daemon. See systemd.service(5) for details.
137 6. As much as possible, rely on the init system's functionality to
139 limit the access of the daemon to files, services and other
140 resources, i.e. in the case of systemd, rely on systemd's resource
141 limit control instead of implementing your own, rely on systemd's
142 privilege dropping code instead of implementing it in the daemon,
143 and similar. See systemd.exec(5) for the available controls.
144 7. If D-Bus is used, make your daemon bus-activatable by supplying a
146 D-Bus service activation configuration file. This has multiple
147 advantages: your daemon may be started lazily on-demand; it may be
148 started in parallel to other daemons requiring it — which maximizes
149 parallelization and boot-up speed; your daemon can be restarted on
150 failure without losing any bus requests, as the bus queues requests
151 for activatable services. See below for details.
152 8. If your daemon provides services to other local processes or remote
154 clients via a socket, it should be made socket-activatable
155 following the scheme pointed out below. Like D-Bus activation, this
156 enables on-demand starting of services as well as it allows
157 improved parallelization of service start-up. Also, for state-less
158 protocols (such as syslog, DNS), a daemon implementing socket-based
159 activation can be restarted without losing a single request. See
160 below for details.
161 9. If applicable, a daemon should notify the init system about startup
163 completion or status updates via the sd_notify(3) interface.
164 10. Instead of using the syslog() call to log directly to the system
166 syslog service, a new-style daemon may choose to simply log to
167 standard error via fprintf(), which is then forwarded to syslog by
168 the init system. If log levels are necessary, these can be encoded
169 by prefixing individual log lines with strings like "<4>" (for log
170 level 4 "WARNING" in the syslog priority scheme), following a
171 similar style as the Linux kernel's printk() level system. For
172 details, see sd-daemon(3) and systemd.exec(5).
173 11. As new-style daemons are invoked without a controlling TTY (but as
175 their own session leaders) care should be taken to always specify
176 `O_NOCTTY` on `open()` calls that possibly reference a TTY device
177 node, so that no controlling TTY is accidentally acquired.
178 These recommendations are similar but not identical to the Apple MacOS
180 X Daemon Requirements[2].
181
183 New-style init systems provide multiple additional mechanisms to
184 activate services, as detailed below. It is common that services are
185 configured to be activated via more than one mechanism at the same
186 time. An example for systemd: bluetoothd.service might get activated
187 either when Bluetooth hardware is plugged in, or when an application
188 accesses its programming interfaces via D-Bus. Or, a print server
189 daemon might get activated when traffic arrives at an IPP port, or when
190 a printer is plugged in, or when a file is queued in the printer spool
191 directory. Even for services that are intended to be started on system
192 bootup unconditionally, it is a good idea to implement some of the
193 various activation schemes outlined below, in order to maximize
194 parallelization. If a daemon implements a D-Bus service or listening
195 socket, implementing the full bus and socket activation scheme allows
196 starting of the daemon with its clients in parallel (which speeds up
197 boot-up), since all its communication channels are established already,
198 and no request is lost because client requests will be queued by the
199 bus system (in case of D-Bus) or the kernel (in case of sockets) until
200 the activation is completed.
201 Activation on Boot
203 Old-style daemons are usually activated exclusively on boot (and
204 manually by the administrator) via SysV init scripts, as detailed in
205 the LSB Linux Standard Base Core Specification[1]. This method of
206 activation is supported ubiquitously on Linux init systems, both
207 old-style and new-style systems. Among other issues, SysV init scripts
208 have the disadvantage of involving shell scripts in the boot process.
209 New-style init systems generally employ updated versions of activation,
210 both during boot-up and during runtime and using more minimal service
211 description files.
212 In systemd, if the developer or administrator wants to make sure that a
214 service or other unit is activated automatically on boot, it is
215 recommended to place a symlink to the unit file in the .wants/
216 directory of either multi-user.target or graphical.target, which are
217 normally used as boot targets at system startup. See systemd.unit(5)
218 for details about the .wants/ directories, and systemd.special(7) for
219 details about the two boot targets.
220 Socket-Based Activation
222 In order to maximize the possible parallelization and robustness and
223 simplify configuration and development, it is recommended for all
224 new-style daemons that communicate via listening sockets to employ
225 socket-based activation. In a socket-based activation scheme, the
226 creation and binding of the listening socket as primary communication
227 channel of daemons to local (and sometimes remote) clients is moved out
228 of the daemon code and into the init system. Based on per-daemon
229 configuration, the init system installs the sockets and then hands them
230 off to the spawned process as soon as the respective daemon is to be
231 started. Optionally, activation of the service can be delayed until the
232 first inbound traffic arrives at the socket to implement on-demand
233 activation of daemons. However, the primary advantage of this scheme is
234 that all providers and all consumers of the sockets can be started in
235 parallel as soon as all sockets are established. In addition to that,
236 daemons can be restarted with losing only a minimal number of client
237 transactions, or even any client request at all (the latter is
238 particularly true for state-less protocols, such as DNS or syslog),
239 because the socket stays bound and accessible during the restart, and
240 all requests are queued while the daemon cannot process them.
241 New-style daemons which support socket activation must be able to
243 receive their sockets from the init system instead of creating and
244 binding them themselves. For details about the programming interfaces
245 for this scheme provided by systemd, see sd_listen_fds(3) and sd-
246 daemon(3). For details about porting existing daemons to socket-based
247 activation, see below. With minimal effort, it is possible to implement
248 socket-based activation in addition to traditional internal socket
249 creation in the same codebase in order to support both new-style and
250 old-style init systems from the same daemon binary.
251 systemd implements socket-based activation via .socket units, which are
253 described in systemd.socket(5). When configuring socket units for
254 socket-based activation, it is essential that all listening sockets are
255 pulled in by the special target unit sockets.target. It is recommended
256 to place a WantedBy=sockets.target directive in the [Install] section
257 to automatically add such a dependency on installation of a socket
258 unit. Unless DefaultDependencies=no is set, the necessary ordering
259 dependencies are implicitly created for all socket units. For more
260 information about sockets.target, see systemd.special(7). It is not
261 necessary or recommended to place any additional dependencies on socket
262 units (for example from multi-user.target or suchlike) when one is
263 installed in sockets.target.
264 Bus-Based Activation
266 When the D-Bus IPC system is used for communication with clients,
267 new-style daemons should employ bus activation so that they are
268 automatically activated when a client application accesses their IPC
269 interfaces. This is configured in D-Bus service files (not to be
270 confused with systemd service unit files!). To ensure that D-Bus uses
271 systemd to start-up and maintain the daemon, use the SystemdService=
272 directive in these service files to configure the matching systemd
273 service for a D-Bus service. e.g.: For a D-Bus service whose D-Bus
274 activation file is named org.freedesktop.RealtimeKit.service, make sure
275 to set SystemdService=rtkit-daemon.service in that file to bind it to
276 the systemd service rtkit-daemon.service. This is needed to make sure
277 that the daemon is started in a race-free fashion when activated via
278 multiple mechanisms simultaneously.
279 Device-Based Activation
281 Often, daemons that manage a particular type of hardware should be
282 activated only when the hardware of the respective kind is plugged in
283 or otherwise becomes available. In a new-style init system, it is
284 possible to bind activation to hardware plug/unplug events. In systemd,
285 kernel devices appearing in the sysfs/udev device tree can be exposed
286 as units if they are tagged with the string "systemd". Like any other
287 kind of unit, they may then pull in other units when activated (i.e.
288 plugged in) and thus implement device-based activation. systemd
289 dependencies may be encoded in the udev database via the SYSTEMD_WANTS=
290 property. See systemd.device(5) for details. Often, it is nicer to pull
291 in services from devices only indirectly via dedicated targets.
292 Example: Instead of pulling in bluetoothd.service from all the various
293 bluetooth dongles and other hardware available, pull in
294 bluetooth.target from them and bluetoothd.service from that target.
295 This provides for nicer abstraction and gives administrators the option
296 to enable bluetoothd.service via controlling a bluetooth.target.wants/
297 symlink uniformly with a command like enable of systemctl(1) instead of
298 manipulating the udev ruleset.
299 Path-Based Activation
301 Often, runtime of daemons processing spool files or directories (such
302 as a printing system) can be delayed until these file system objects
303 change state, or become non-empty. New-style init systems provide a way
304 to bind service activation to file system changes. systemd implements
305 this scheme via path-based activation configured in .path units, as
306 outlined in systemd.path(5).
307 Timer-Based Activation
309 Some daemons that implement clean-up jobs that are intended to be
310 executed in regular intervals benefit from timer-based activation. In
311 systemd, this is implemented via .timer units, as described in
312 systemd.timer(5).
313 Other Forms of Activation
315 Other forms of activation have been suggested and implemented in some
316 systems. However, there are often simpler or better alternatives, or
317 they can be put together of combinations of the schemes above. Example:
318 Sometimes, it appears useful to start daemons or .socket units when a
319 specific IP address is configured on a network interface, because
320 network sockets shall be bound to the address. However, an alternative
321 to implement this is by utilizing the Linux IP_FREEBIND socket option,
322 as accessible via FreeBind=yes in systemd socket files (see
323 systemd.socket(5) for details). This option, when enabled, allows
324 sockets to be bound to a non-local, not configured IP address, and
325 hence allows bindings to a particular IP address before it actually
326 becomes available, making such an explicit dependency to the configured
327 address redundant. Another often suggested trigger for service
328 activation is low system load. However, here too, a more convincing
329 approach might be to make proper use of features of the operating
330 system, in particular, the CPU or I/O scheduler of Linux. Instead of
331 scheduling jobs from userspace based on monitoring the OS scheduler, it
332 is advisable to leave the scheduling of processes to the OS scheduler
333 itself. systemd provides fine-grained access to the CPU and I/O
334 schedulers. If a process executed by the init system shall not
335 negatively impact the amount of CPU or I/O bandwidth available to other
336 processes, it should be configured with CPUSchedulingPolicy=idle and/or
337 IOSchedulingClass=idle. Optionally, this may be combined with
338 timer-based activation to schedule background jobs during runtime and
339 with minimal impact on the system, and remove it from the boot phase
340 itself.
341
343 Writing systemd Unit Files
344 When writing systemd unit files, it is recommended to consider the
345 following suggestions:
346 1. If possible, do not use the Type=forking setting in service files.
348 But if you do, make sure to set the PID file path using PIDFile=.
349 See systemd.service(5) for details.
350 2. If your daemon registers a D-Bus name on the bus, make sure to use
352 Type=dbus in the service file if possible.
353 3. Make sure to set a good human-readable description string with
355 Description=.
356 4. Do not disable DefaultDependencies=, unless you really know what
358 you do and your unit is involved in early boot or late system
359 shutdown.
360 5. Normally, little if any dependencies should need to be defined
362 explicitly. However, if you do configure explicit dependencies,
363 only refer to unit names listed on systemd.special(7) or names
364 introduced by your own package to keep the unit file operating
365 system-independent.
366 6. Make sure to include an [Install] section including installation
368 information for the unit file. See systemd.unit(5) for details. To
369 activate your service on boot, make sure to add a
370 WantedBy=multi-user.target or WantedBy=graphical.target directive.
371 To activate your socket on boot, make sure to add
372 WantedBy=sockets.target. Usually, you also want to make sure that
373 when your service is installed, your socket is installed too, hence
374 add Also=foo.socket in your service file foo.service, for a
375 hypothetical program foo.
376 Installing systemd Service Files
378 At the build installation time (e.g. make install during package
379 build), packages are recommended to install their systemd unit files in
380 the directory returned by pkg-config systemd
381 --variable=systemdsystemunitdir (for system services) or pkg-config
382 systemd --variable=systemduserunitdir (for user services). This will
383 make the services available in the system on explicit request but not
384 activate them automatically during boot. Optionally, during package
385 installation (e.g. rpm -i by the administrator), symlinks should be
386 created in the systemd configuration directories via the enable command
387 of the systemctl(1) tool to activate them automatically on boot.
388 Packages using autoconf(1) are recommended to use a configure script
390 excerpt like the following to determine the unit installation path
391 during source configuration:
392 PKG_PROG_PKG_CONFIG
394 AC_ARG_WITH([systemdsystemunitdir],
395 [AS_HELP_STRING([--with-systemdsystemunitdir=DIR], [Directory for systemd service files])],,
396 [with_systemdsystemunitdir=auto])
397 AS_IF([test "x$with_systemdsystemunitdir" = "xyes" -o "x$with_systemdsystemunitdir" = "xauto"], [
398 def_systemdsystemunitdir=$($PKG_CONFIG --variable=systemdsystemunitdir systemd)
399 AS_IF([test "x$def_systemdsystemunitdir" = "x"],
401 [AS_IF([test "x$with_systemdsystemunitdir" = "xyes"],
402 [AC_MSG_ERROR([systemd support requested but pkg-config unable to query systemd package])])
403 with_systemdsystemunitdir=no],
404 [with_systemdsystemunitdir="$def_systemdsystemunitdir"])])
405 AS_IF([test "x$with_systemdsystemunitdir" != "xno"],
406 [AC_SUBST([systemdsystemunitdir], [$with_systemdsystemunitdir])])
407 AM_CONDITIONAL([HAVE_SYSTEMD], [test "x$with_systemdsystemunitdir" != "xno"])
408 This snippet allows automatic installation of the unit files on systemd
410 machines, and optionally allows their installation even on machines
411 lacking systemd. (Modification of this snippet for the user unit
412 directory is left as an exercise for the reader.)
413 Additionally, to ensure that make distcheck continues to work, it is
415 recommended to add the following to the top-level Makefile.am file in
416 automake(1)-based projects:
417 AM_DISTCHECK_CONFIGURE_FLAGS = \
419 --with-systemdsystemunitdir=$$dc_install_base/$(systemdsystemunitdir)
420 Finally, unit files should be installed in the system with an automake
422 excerpt like the following:
423 if HAVE_SYSTEMD
425 systemdsystemunit_DATA = \
426 foobar.socket \
427 foobar.service
428 endif
429 In the rpm(8) .spec file, use snippets like the following to
431 enable/disable the service during installation/deinstallation. This
432 makes use of the RPM macros shipped along systemd. Consult the
433 packaging guidelines of your distribution for details and the
434 equivalent for other package managers.
435 At the top of the file:
437 BuildRequires: systemd
439 %{?systemd_requires}
440 And as scriptlets, further down:
442 %post
444 %systemd_post foobar.service foobar.socket
445 %preun
447 %systemd_preun foobar.service foobar.socket
448 %postun
450 %systemd_postun
451 If the service shall be restarted during upgrades, replace the
453 "%postun" scriptlet above with the following:
454 %postun
456 %systemd_postun_with_restart foobar.service
457 Note that "%systemd_post" and "%systemd_preun" expect the names of all
459 units that are installed/removed as arguments, separated by spaces.
460 "%systemd_postun" expects no arguments. "%systemd_postun_with_restart"
461 expects the units to restart as arguments.
462 To facilitate upgrades from a package version that shipped only SysV
464 init scripts to a package version that ships both a SysV init script
465 and a native systemd service file, use a fragment like the following:
466 %triggerun -- foobar < 0.47.11-1
468 if /sbin/chkconfig --level 5 foobar ; then
469 /bin/systemctl --no-reload enable foobar.service foobar.socket >/dev/null 2>&1 || :
470 fi
471 Where 0.47.11-1 is the first package version that includes the native
473 unit file. This fragment will ensure that the first time the unit file
474 is installed, it will be enabled if and only if the SysV init script is
475 enabled, thus making sure that the enable status is not changed. Note
476 that chkconfig is a command specific to Fedora which can be used to
477 check whether a SysV init script is enabled. Other operating systems
478 will have to use different commands here.
479
481 Since new-style init systems such as systemd are compatible with
482 traditional SysV init systems, it is not strictly necessary to port
483 existing daemons to the new style. However, doing so offers additional
484 functionality to the daemons as well as simplifying integration into
485 new-style init systems.
486 To port an existing SysV compatible daemon, the following steps are
488 recommended:
489 1. If not already implemented, add an optional command line switch to
491 the daemon to disable daemonization. This is useful not only for
492 using the daemon in new-style init systems, but also to ease
493 debugging.
494 2. If the daemon offers interfaces to other software running on the
496 local system via local AF_UNIX sockets, consider implementing
497 socket-based activation (see above). Usually, a minimal patch is
498 sufficient to implement this: Extend the socket creation in the
499 daemon code so that sd_listen_fds(3) is checked for already passed
500 sockets first. If sockets are passed (i.e. when sd_listen_fds()
501 returns a positive value), skip the socket creation step and use
502 the passed sockets. Secondly, ensure that the file system socket
503 nodes for local AF_UNIX sockets used in the socket-based activation
504 are not removed when the daemon shuts down, if sockets have been
505 passed. Third, if the daemon normally closes all remaining open
506 file descriptors as part of its initialization, the sockets passed
507 from the init system must be spared. Since new-style init systems
508 guarantee that no left-over file descriptors are passed to executed
509 processes, it might be a good choice to simply skip the closing of
510 all remaining open file descriptors if sockets are passed.
511 3. Write and install a systemd unit file for the service (and the
513 sockets if socket-based activation is used, as well as a path unit
514 file, if the daemon processes a spool directory), see above for
515 details.
516 4. If the daemon exposes interfaces via D-Bus, write and install a
518 D-Bus activation file for the service, see above for details.
519
521 It is recommended to follow the general guidelines for placing package
522 files, as discussed in file-hierarchy(7).
523
525 systemd(1), sd-daemon(3), sd_listen_fds(3), sd_notify(3), daemon(3),
526 systemd.service(5), file-hierarchy(7)
527
529 1. LSB recommendations for SysV init scripts
530 http://refspecs.linuxbase.org/LSB_3.1.1/LSB-Core-generic/LSB-Core-generic/iniscrptact.html
531 2. Apple MacOS X Daemon Requirements
533 https://developer.apple.com/library/mac/documentation/MacOSX/Conceptual/BPSystemStartup/Chapters/CreatingLaunchdJobs.html
534systemd 246 DAEMON(7)