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.
48 8. Call exit() in the first child, so that only the second child (the
50 actual daemon process) stays around. This ensures that the daemon
51 process is re-parented to init/PID 1, as all daemons should be.
52 9. In the daemon process, connect /dev/null to standard input, output,
54 and error.
55 10. In the daemon process, reset the umask to 0, so that the file modes
57 passed to open(), mkdir() and suchlike directly control the access
58 mode of the created files and directories.
59 11. In the daemon process, change the current directory to the root
61 directory (/), in order to avoid that the daemon involuntarily
62 blocks mount points from being unmounted.
63 12. In the daemon process, write the daemon PID (as returned by
65 getpid()) to a PID file, for example /run/foobar.pid (for a
66 hypothetical daemon "foobar") to ensure that the daemon cannot be
67 started more than once. This must be implemented in race-free
68 fashion so that the PID file is only updated when it is verified at
69 the same time that the PID previously stored in the PID file no
70 longer exists or belongs to a foreign process.
71 13. In the daemon process, drop privileges, if possible and applicable.
73 14. From the daemon process, notify the original process started that
75 initialization is complete. This can be implemented via an unnamed
76 pipe or similar communication channel that is created before the
77 first fork() and hence available in both the original and the
78 daemon process.
79 15. Call exit() in the original process. The process that invoked the
81 daemon must be able to rely on that this exit() happens after
82 initialization is complete and all external communication channels
83 are established and accessible.
84 The BSD daemon() function should not be used, as it implements only a
86 subset of these steps.
87 A daemon that needs to provide compatibility with SysV systems should
89 implement the scheme pointed out above. However, it is recommended to
90 make this behavior optional and configurable via a command line
91 argument to ease debugging as well as to simplify integration into
92 systems using systemd.
93 New-Style Daemons
95 Modern services for Linux should be implemented as new-style daemons.
96 This makes it easier to supervise and control them at runtime and
97 simplifies their implementation.
98 For developing a new-style daemon, none of the initialization steps
100 recommended for SysV daemons need to be implemented. New-style init
101 systems such as systemd make all of them redundant. Moreover, since
102 some of these steps interfere with process monitoring, file descriptor
103 passing and other functionality of the init system, it is recommended
104 not to execute them when run as new-style service.
105 Note that new-style init systems guarantee execution of daemon
107 processes in a clean process context: it is guaranteed that the
108 environment block is sanitized, that the signal handlers and mask is
109 reset and that no left-over file descriptors are passed. Daemons will
110 be executed in their own session, with standard input/output/error
111 connected to /dev/null unless otherwise configured. The umask is reset.
112 It is recommended for new-style daemons to implement the following:
114 1. If SIGTERM is received, shut down the daemon and exit cleanly.
116 2. If SIGHUP is received, reload the configuration files, if this
118 applies.
119 3. Provide a correct exit code from the main daemon process, as this
121 is used by the init system to detect service errors and problems.
122 It is recommended to follow the exit code scheme as defined in the
123 LSB recommendations for SysV init scripts[1].
124 4. If possible and applicable, expose the daemon's control interface
126 via the D-Bus IPC system and grab a bus name as last step of
127 initialization.
128 5. For integration in systemd, provide a .service unit file that
130 carries information about starting, stopping and otherwise
131 maintaining the daemon. See systemd.service(5) for details.
132 6. As much as possible, rely on the init system's functionality to
134 limit the access of the daemon to files, services and other
135 resources, i.e. in the case of systemd, rely on systemd's resource
136 limit control instead of implementing your own, rely on systemd's
137 privilege dropping code instead of implementing it in the daemon,
138 and similar. See systemd.exec(5) for the available controls.
139 7. If D-Bus is used, make your daemon bus-activatable by supplying a
141 D-Bus service activation configuration file. This has multiple
142 advantages: your daemon may be started lazily on-demand; it may be
143 started in parallel to other daemons requiring it -- which
144 maximizes parallelization and boot-up speed; your daemon can be
145 restarted on failure without losing any bus requests, as the bus
146 queues requests for activatable services. See below for details.
147 8. If your daemon provides services to other local processes or remote
149 clients via a socket, it should be made socket-activatable
150 following the scheme pointed out below. Like D-Bus activation, this
151 enables on-demand starting of services as well as it allows
152 improved parallelization of service start-up. Also, for state-less
153 protocols (such as syslog, DNS), a daemon implementing socket-based
154 activation can be restarted without losing a single request. See
155 below for details.
156 9. If applicable, a daemon should notify the init system about startup
158 completion or status updates via the sd_notify(3) interface.
159 10. Instead of using the syslog() call to log directly to the system
161 syslog service, a new-style daemon may choose to simply log to
162 standard error via fprintf(), which is then forwarded to syslog by
163 the init system. If log levels are necessary, these can be encoded
164 by prefixing individual log lines with strings like "<4>" (for log
165 level 4 "WARNING" in the syslog priority scheme), following a
166 similar style as the Linux kernel's printk() level system. For
167 details, see sd-daemon(3) and systemd.exec(5).
168 These recommendations are similar but not identical to the Apple MacOS
170 X Daemon Requirements[2].
171
173 New-style init systems provide multiple additional mechanisms to
174 activate services, as detailed below. It is common that services are
175 configured to be activated via more than one mechanism at the same
176 time. An example for systemd: bluetoothd.service might get activated
177 either when Bluetooth hardware is plugged in, or when an application
178 accesses its programming interfaces via D-Bus. Or, a print server
179 daemon might get activated when traffic arrives at an IPP port, or when
180 a printer is plugged in, or when a file is queued in the printer spool
181 directory. Even for services that are intended to be started on system
182 bootup unconditionally, it is a good idea to implement some of the
183 various activation schemes outlined below, in order to maximize
184 parallelization. If a daemon implements a D-Bus service or listening
185 socket, implementing the full bus and socket activation scheme allows
186 starting of the daemon with its clients in parallel (which speeds up
187 boot-up), since all its communication channels are established already,
188 and no request is lost because client requests will be queued by the
189 bus system (in case of D-Bus) or the kernel (in case of sockets) until
190 the activation is completed.
191 Activation on Boot
193 Old-style daemons are usually activated exclusively on boot (and
194 manually by the administrator) via SysV init scripts, as detailed in
195 the LSB Linux Standard Base Core Specification[1]. This method of
196 activation is supported ubiquitously on Linux init systems, both
197 old-style and new-style systems. Among other issues, SysV init scripts
198 have the disadvantage of involving shell scripts in the boot process.
199 New-style init systems generally employ updated versions of activation,
200 both during boot-up and during runtime and using more minimal service
201 description files.
202 In systemd, if the developer or administrator wants to make sure that a
204 service or other unit is activated automatically on boot, it is
205 recommended to place a symlink to the unit file in the .wants/
206 directory of either multi-user.target or graphical.target, which are
207 normally used as boot targets at system startup. See systemd.unit(5)
208 for details about the .wants/ directories, and systemd.special(7) for
209 details about the two boot targets.
210 Socket-Based Activation
212 In order to maximize the possible parallelization and robustness and
213 simplify configuration and development, it is recommended for all
214 new-style daemons that communicate via listening sockets to employ
215 socket-based activation. In a socket-based activation scheme, the
216 creation and binding of the listening socket as primary communication
217 channel of daemons to local (and sometimes remote) clients is moved out
218 of the daemon code and into the init system. Based on per-daemon
219 configuration, the init system installs the sockets and then hands them
220 off to the spawned process as soon as the respective daemon is to be
221 started. Optionally, activation of the service can be delayed until the
222 first inbound traffic arrives at the socket to implement on-demand
223 activation of daemons. However, the primary advantage of this scheme is
224 that all providers and all consumers of the sockets can be started in
225 parallel as soon as all sockets are established. In addition to that,
226 daemons can be restarted with losing only a minimal number of client
227 transactions, or even any client request at all (the latter is
228 particularly true for state-less protocols, such as DNS or syslog),
229 because the socket stays bound and accessible during the restart, and
230 all requests are queued while the daemon cannot process them.
231 New-style daemons which support socket activation must be able to
233 receive their sockets from the init system instead of creating and
234 binding them themselves. For details about the programming interfaces
235 for this scheme provided by systemd, see sd_listen_fds(3) and sd-
236 daemon(3). For details about porting existing daemons to socket-based
237 activation, see below. With minimal effort, it is possible to implement
238 socket-based activation in addition to traditional internal socket
239 creation in the same codebase in order to support both new-style and
240 old-style init systems from the same daemon binary.
241 systemd implements socket-based activation via .socket units, which are
243 described in systemd.socket(5). When configuring socket units for
244 socket-based activation, it is essential that all listening sockets are
245 pulled in by the special target unit sockets.target. It is recommended
246 to place a WantedBy=sockets.target directive in the "[Install]" section
247 to automatically add such a dependency on installation of a socket
248 unit. Unless DefaultDependencies=no is set, the necessary ordering
249 dependencies are implicitly created for all socket units. For more
250 information about sockets.target, see systemd.special(7). It is not
251 necessary or recommended to place any additional dependencies on socket
252 units (for example from multi-user.target or suchlike) when one is
253 installed in sockets.target.
254 Bus-Based Activation
256 When the D-Bus IPC system is used for communication with clients,
257 new-style daemons should employ bus activation so that they are
258 automatically activated when a client application accesses their IPC
259 interfaces. This is configured in D-Bus service files (not to be
260 confused with systemd service unit files!). To ensure that D-Bus uses
261 systemd to start-up and maintain the daemon, use the SystemdService=
262 directive in these service files to configure the matching systemd
263 service for a D-Bus service. e.g.: For a D-Bus service whose D-Bus
264 activation file is named org.freedesktop.RealtimeKit.service, make sure
265 to set SystemdService=rtkit-daemon.service in that file to bind it to
266 the systemd service rtkit-daemon.service. This is needed to make sure
267 that the daemon is started in a race-free fashion when activated via
268 multiple mechanisms simultaneously.
269 Device-Based Activation
271 Often, daemons that manage a particular type of hardware should be
272 activated only when the hardware of the respective kind is plugged in
273 or otherwise becomes available. In a new-style init system, it is
274 possible to bind activation to hardware plug/unplug events. In systemd,
275 kernel devices appearing in the sysfs/udev device tree can be exposed
276 as units if they are tagged with the string "systemd". Like any other
277 kind of unit, they may then pull in other units when activated (i.e.
278 plugged in) and thus implement device-based activation. systemd
279 dependencies may be encoded in the udev database via the SYSTEMD_WANTS=
280 property. See systemd.device(5) for details. Often, it is nicer to pull
281 in services from devices only indirectly via dedicated targets.
282 Example: Instead of pulling in bluetoothd.service from all the various
283 bluetooth dongles and other hardware available, pull in
284 bluetooth.target from them and bluetoothd.service from that target.
285 This provides for nicer abstraction and gives administrators the option
286 to enable bluetoothd.service via controlling a bluetooth.target.wants/
287 symlink uniformly with a command like enable of systemctl(1) instead of
288 manipulating the udev ruleset.
289 Path-Based Activation
291 Often, runtime of daemons processing spool files or directories (such
292 as a printing system) can be delayed until these file system objects
293 change state, or become non-empty. New-style init systems provide a way
294 to bind service activation to file system changes. systemd implements
295 this scheme via path-based activation configured in .path units, as
296 outlined in systemd.path(5).
297 Timer-Based Activation
299 Some daemons that implement clean-up jobs that are intended to be
300 executed in regular intervals benefit from timer-based activation. In
301 systemd, this is implemented via .timer units, as described in
302 systemd.timer(5).
303 Other Forms of Activation
305 Other forms of activation have been suggested and implemented in some
306 systems. However, there are often simpler or better alternatives, or
307 they can be put together of combinations of the schemes above. Example:
308 Sometimes, it appears useful to start daemons or .socket units when a
309 specific IP address is configured on a network interface, because
310 network sockets shall be bound to the address. However, an alternative
311 to implement this is by utilizing the Linux IP_FREEBIND socket option,
312 as accessible via FreeBind=yes in systemd socket files (see
313 systemd.socket(5) for details). This option, when enabled, allows
314 sockets to be bound to a non-local, not configured IP address, and
315 hence allows bindings to a particular IP address before it actually
316 becomes available, making such an explicit dependency to the configured
317 address redundant. Another often suggested trigger for service
318 activation is low system load. However, here too, a more convincing
319 approach might be to make proper use of features of the operating
320 system, in particular, the CPU or IO scheduler of Linux. Instead of
321 scheduling jobs from userspace based on monitoring the OS scheduler, it
322 is advisable to leave the scheduling of processes to the OS scheduler
323 itself. systemd provides fine-grained access to the CPU and IO
324 schedulers. If a process executed by the init system shall not
325 negatively impact the amount of CPU or IO bandwidth available to other
326 processes, it should be configured with CPUSchedulingPolicy=idle and/or
327 IOSchedulingClass=idle. Optionally, this may be combined with
328 timer-based activation to schedule background jobs during runtime and
329 with minimal impact on the system, and remove it from the boot phase
330 itself.
331
333 Writing Systemd Unit Files
334 When writing systemd unit files, it is recommended to consider the
335 following suggestions:
336 1. If possible, do not use the Type=forking setting in service files.
338 But if you do, make sure to set the PID file path using PIDFile=.
339 See systemd.service(5) for details.
340 2. If your daemon registers a D-Bus name on the bus, make sure to use
342 Type=dbus in the service file if possible.
343 3. Make sure to set a good human-readable description string with
345 Description=.
346 4. Do not disable DefaultDependencies=, unless you really know what
348 you do and your unit is involved in early boot or late system
349 shutdown.
350 5. Normally, little if any dependencies should need to be defined
352 explicitly. However, if you do configure explicit dependencies,
353 only refer to unit names listed on systemd.special(7) or names
354 introduced by your own package to keep the unit file operating
355 system-independent.
356 6. Make sure to include an "[Install]" section including installation
358 information for the unit file. See systemd.unit(5) for details. To
359 activate your service on boot, make sure to add a
360 WantedBy=multi-user.target or WantedBy=graphical.target directive.
361 To activate your socket on boot, make sure to add
362 WantedBy=sockets.target. Usually, you also want to make sure that
363 when your service is installed, your socket is installed too, hence
364 add Also=foo.socket in your service file foo.service, for a
365 hypothetical program foo.
366 Installing Systemd Service Files
368 At the build installation time (e.g. make install during package
369 build), packages are recommended to install their systemd unit files in
370 the directory returned by pkg-config systemd
371 --variable=systemdsystemunitdir (for system services) or pkg-config
372 systemd --variable=systemduserunitdir (for user services). This will
373 make the services available in the system on explicit request but not
374 activate them automatically during boot. Optionally, during package
375 installation (e.g. rpm -i by the administrator), symlinks should be
376 created in the systemd configuration directories via the enable command
377 of the systemctl(1) tool to activate them automatically on boot.
378 Packages using autoconf(1) are recommended to use a configure script
380 excerpt like the following to determine the unit installation path
381 during source configuration:
382 PKG_PROG_PKG_CONFIG
384 AC_ARG_WITH([systemdsystemunitdir],
385 [AS_HELP_STRING([--with-systemdsystemunitdir=DIR], [Directory for systemd service files])],,
386 [with_systemdsystemunitdir=auto])
387 AS_IF([test "x$with_systemdsystemunitdir" = "xyes" -o "x$with_systemdsystemunitdir" = "xauto"], [
388 def_systemdsystemunitdir=$($PKG_CONFIG --variable=systemdsystemunitdir systemd)
389 AS_IF([test "x$def_systemdsystemunitdir" = "x"],
391 [AS_IF([test "x$with_systemdsystemunitdir" = "xyes"],
392 [AC_MSG_ERROR([systemd support requested but pkg-config unable to query systemd package])])
393 with_systemdsystemunitdir=no],
394 [with_systemdsystemunitdir="$def_systemdsystemunitdir"])])
395 AS_IF([test "x$with_systemdsystemunitdir" != "xno"],
396 [AC_SUBST([systemdsystemunitdir], [$with_systemdsystemunitdir])])
397 AM_CONDITIONAL([HAVE_SYSTEMD], [test "x$with_systemdsystemunitdir" != "xno"])
398 This snippet allows automatic installation of the unit files on systemd
400 machines, and optionally allows their installation even on machines
401 lacking systemd. (Modification of this snippet for the user unit
402 directory is left as an exercise for the reader.)
403 Additionally, to ensure that make distcheck continues to work, it is
405 recommended to add the following to the top-level Makefile.am file in
406 automake(1)-based projects:
407 DISTCHECK_CONFIGURE_FLAGS = \
409 --with-systemdsystemunitdir=$$dc_install_base/$(systemdsystemunitdir)
410 Finally, unit files should be installed in the system with an automake
412 excerpt like the following:
413 if HAVE_SYSTEMD
415 systemdsystemunit_DATA = \
416 foobar.socket \
417 foobar.service
418 endif
419 In the rpm(8).spec file, use snippets like the following to
421 enable/disable the service during installation/deinstallation. This
422 makes use of the RPM macros shipped along systemd. Consult the
423 packaging guidelines of your distribution for details and the
424 equivalent for other package managers.
425 At the top of the file:
427 BuildRequires: systemd
429 %{?systemd_requires}
430 And as scriptlets, further down:
432 %post
434 %systemd_post foobar.service foobar.socket
435 %preun
437 %systemd_preun foobar.service foobar.socket
438 %postun
440 %systemd_postun
441 If the service shall be restarted during upgrades, replace the
443 "%postun" scriptlet above with the following:
444 %postun
446 %systemd_postun_with_restart foobar.service
447 Note that "%systemd_post" and "%systemd_preun" expect the names of all
449 units that are installed/removed as arguments, separated by spaces.
450 "%systemd_postun" expects no arguments. "%systemd_postun_with_restart"
451 expects the units to restart as arguments.
452 To facilitate upgrades from a package version that shipped only SysV
454 init scripts to a package version that ships both a SysV init script
455 and a native systemd service file, use a fragment like the following:
456 %triggerun -- foobar < 0.47.11-1
458 if /sbin/chkconfig --level 5 foobar ; then
459 /bin/systemctl --no-reload enable foobar.service foobar.socket >/dev/null 2>&1 || :
460 fi
461 Where 0.47.11-1 is the first package version that includes the native
463 unit file. This fragment will ensure that the first time the unit file
464 is installed, it will be enabled if and only if the SysV init script is
465 enabled, thus making sure that the enable status is not changed. Note
466 that chkconfig is a command specific to Fedora which can be used to
467 check whether a SysV init script is enabled. Other operating systems
468 will have to use different commands here.
469
471 Since new-style init systems such as systemd are compatible with
472 traditional SysV init systems, it is not strictly necessary to port
473 existing daemons to the new style. However, doing so offers additional
474 functionality to the daemons as well as simplifying integration into
475 new-style init systems.
476 To port an existing SysV compatible daemon, the following steps are
478 recommended:
479 1. If not already implemented, add an optional command line switch to
481 the daemon to disable daemonization. This is useful not only for
482 using the daemon in new-style init systems, but also to ease
483 debugging.
484 2. If the daemon offers interfaces to other software running on the
486 local system via local AF_UNIX sockets, consider implementing
487 socket-based activation (see above). Usually, a minimal patch is
488 sufficient to implement this: Extend the socket creation in the
489 daemon code so that sd_listen_fds(3) is checked for already passed
490 sockets first. If sockets are passed (i.e. when sd_listen_fds()
491 returns a positive value), skip the socket creation step and use
492 the passed sockets. Secondly, ensure that the file system socket
493 nodes for local AF_UNIX sockets used in the socket-based activation
494 are not removed when the daemon shuts down, if sockets have been
495 passed. Third, if the daemon normally closes all remaining open
496 file descriptors as part of its initialization, the sockets passed
497 from the init system must be spared. Since new-style init systems
498 guarantee that no left-over file descriptors are passed to executed
499 processes, it might be a good choice to simply skip the closing of
500 all remaining open file descriptors if sockets are passed.
501 3. Write and install a systemd unit file for the service (and the
503 sockets if socket-based activation is used, as well as a path unit
504 file, if the daemon processes a spool directory), see above for
505 details.
506 4. If the daemon exposes interfaces via D-Bus, write and install a
508 D-Bus activation file for the service, see above for details.
509
511 It is recommended to follow the general guidelines for placing package
512 files, as discussed in file-hierarchy(7).
513
515 systemd(1), sd-daemon(3), sd_listen_fds(3), sd_notify(3), daemon(3),
516 systemd.service(5), file-hierarchy(7)
517
519 1. LSB recommendations for SysV init scripts
520 http://refspecs.linuxbase.org/LSB_3.1.1/LSB-Core-generic/LSB-Core-generic/iniscrptact.html
521 2. Apple MacOS X Daemon Requirements
523 https://developer.apple.com/library/mac/documentation/MacOSX/Conceptual/BPSystemStartup/Chapters/CreatingLaunchdJobs.html
524systemd 219 DAEMON(7)