1SD_NOTIFY(3) sd_notify SD_NOTIFY(3)
2
6 sd_notify, sd_notifyf, sd_pid_notify, sd_pid_notifyf,
7 sd_pid_notify_with_fds, sd_pid_notifyf_with_fds, sd_notify_barrier,
8 sd_pid_notify_barrier - Notify service manager about start-up
9 completion and other service status changes
10
12 #include <systemd/sd-daemon.h>
13 int sd_notify(int unset_environment, const char *state);
15 int sd_notifyf(int unset_environment, const char *format, ...);
17 int sd_pid_notify(pid_t pid, int unset_environment, const char *state);
19 int sd_pid_notifyf(pid_t pid, int unset_environment,
21 const char *format, ...);
22 int sd_pid_notify_with_fds(pid_t pid, int unset_environment,
24 const char *state, const int *fds,
25 unsigned n_fds);
26 int sd_pid_notifyf_with_fds(pid_t pid, int unset_environment,
28 const int *fds, size_t n_fds,
29 const char *format, ...);
30 int sd_notify_barrier(int unset_environment, uint64_t timeout);
32 int sd_pid_notify_barrier(pid_t pid, int unset_environment,
34 uint64_t timeout);
35
37 sd_notify() may be called by a service to notify the service manager
38 about state changes. It can be used to send arbitrary information,
39 encoded in an environment-block-like string. Most importantly, it can
40 be used for start-up completion notification.
41 If the unset_environment parameter is non-zero, sd_notify() will unset
43 the $NOTIFY_SOCKET environment variable before returning (regardless of
44 whether the function call itself succeeded or not). Further calls to
45 sd_notify() will then fail, but the variable is no longer inherited by
46 child processes.
47 The state parameter should contain a newline-separated list of variable
49 assignments, similar in style to an environment block. A trailing
50 newline is implied if none is specified. The string may contain any
51 kind of variable assignments, but the following shall be considered
52 well-known:
53 READY=1
55 Tells the service manager that service startup is finished, or the
56 service finished re-loading its configuration. This is only used by
57 systemd if the service definition file has Type=notify or
58 Type=notify-reload set. Since there is little value in signaling
59 non-readiness, the only value services should send is "READY=1"
60 (i.e. "READY=0" is not defined).
61 RELOADING=1
63 Tells the service manager that the service is beginning to reload
64 its configuration. This is useful to allow the service manager to
65 track the service's internal state, and present it to the user.
66 Note that a service that sends this notification must also send a
67 "READY=1" notification when it completed reloading its
68 configuration. Reloads the service manager is notified about with
69 this mechanisms are propagated in the same way as they are when
70 originally initiated through the service manager. This message is
71 particularly relevant for Type=notify-reload services, to inform
72 the service manager that the request to reload the service has been
73 received and is now being processed.
74 MONOTONIC_USEC=...
76 A field carrying the monotonic timestamp (as per CLOCK_MONOTONIC)
77 formatted in decimal in μs, when the notification message was
78 generated by the client. This is typically used in combination with
79 "RELOADING=1", to allow the service manager to properly synchronize
80 reload cycles. See systemd.service(5) for details, specifically
81 "Type=notify-reload".
82 STOPPING=1
84 Tells the service manager that the service is beginning its
85 shutdown. This is useful to allow the service manager to track the
86 service's internal state, and present it to the user.
87 STATUS=...
89 Passes a single-line UTF-8 status string back to the service
90 manager that describes the service state. This is free-form and can
91 be used for various purposes: general state feedback, fsck-like
92 programs could pass completion percentages and failing programs
93 could pass a human-readable error message. Example:
94 "STATUS=Completed 66% of file system check..."
95 NOTIFYACCESS=...
97 Reset the access to the service status notification socket during
98 runtime, overriding NotifyAccess= setting in the service unit file.
99 See systemd.service(5) for details, specifically "NotifyAccess="
100 for a list of accepted values.
101 ERRNO=...
103 If a service fails, the errno-style error code, formatted as
104 string. Example: "ERRNO=2" for ENOENT.
105 BUSERROR=...
107 If a service fails, the D-Bus error-style error code. Example:
108 "BUSERROR=org.freedesktop.DBus.Error.TimedOut"
109 EXIT_STATUS=...
111 If a service exits, the return value of its main() function.
112 MAINPID=...
114 The main process ID (PID) of the service, in case the service
115 manager did not fork off the process itself. Example:
116 "MAINPID=4711"
117 WATCHDOG=1
119 Tells the service manager to update the watchdog timestamp. This is
120 the keep-alive ping that services need to issue in regular
121 intervals if WatchdogSec= is enabled for it. See systemd.service(5)
122 for information how to enable this functionality and
123 sd_watchdog_enabled(3) for the details of how the service can check
124 whether the watchdog is enabled.
125 WATCHDOG=trigger
127 Tells the service manager that the service detected an internal
128 error that should be handled by the configured watchdog options.
129 This will trigger the same behaviour as if WatchdogSec= is enabled
130 and the service did not send "WATCHDOG=1" in time. Note that
131 WatchdogSec= does not need to be enabled for "WATCHDOG=trigger" to
132 trigger the watchdog action. See systemd.service(5) for information
133 about the watchdog behavior.
134 WATCHDOG_USEC=...
136 Reset watchdog_usec value during runtime. Notice that this is not
137 available when using sd_event_set_watchdog() or
138 sd_watchdog_enabled(). Example : "WATCHDOG_USEC=20000000"
139 EXTEND_TIMEOUT_USEC=...
141 Tells the service manager to extend the startup, runtime or
142 shutdown service timeout corresponding the current state. The value
143 specified is a time in microseconds during which the service must
144 send a new message. A service timeout will occur if the message
145 isn't received, but only if the runtime of the current state is
146 beyond the original maximum times of TimeoutStartSec=,
147 RuntimeMaxSec=, and TimeoutStopSec=. See systemd.service(5) for
148 effects on the service timeouts.
149 FDSTORE=1
151 Stores additional file descriptors in the service manager. File
152 descriptors sent this way will be maintained per-service by the
153 service manager and will later be handed back using the usual file
154 descriptor passing logic at the next invocation of the service
155 (e.g. when it is restarted), see sd_listen_fds(3). This is useful
156 for implementing services that can restart after an explicit
157 request or a crash without losing state. Any open sockets and other
158 file descriptors which should not be closed during the restart may
159 be stored this way. Application state can either be serialized to a
160 file in /run/, or better, stored in a memfd_create(2) memory file
161 descriptor. Note that the service manager will accept messages for
162 a service only if its FileDescriptorStoreMax= setting is non-zero
163 (defaults to zero, see systemd.service(5)). If FDPOLL=0 is not set
164 and the file descriptors sent are pollable (see epoll_ctl(2)), then
165 any EPOLLHUP or EPOLLERR event seen on them will result in their
166 automatic removal from the store. Multiple arrays of file
167 descriptors may be sent in separate messages, in which case the
168 arrays are combined. Note that the service manager removes
169 duplicate (pointing to the same object) file descriptors before
170 passing them to the service. When a service is stopped, its file
171 descriptor store is discarded and all file descriptors in it are
172 closed. Use sd_pid_notify_with_fds() to send messages with
173 "FDSTORE=1", see below. The service manager will set the $FDSTORE
174 environment variable for services that have the file descriptor
175 store enabled.
176 For further information on the file descriptor store see the File
178 Descriptor Store[1] overview.
179 FDSTOREREMOVE=1
181 Removes file descriptors from the file descriptor store. This field
182 needs to be combined with FDNAME= to specify the name of the file
183 descriptors to remove.
184 FDNAME=...
186 When used in combination with FDSTORE=1, specifies a name for the
187 submitted file descriptors. When used with FDSTOREREMOVE=1,
188 specifies the name for the file descriptors to remove. This name is
189 passed to the service during activation, and may be queried using
190 sd_listen_fds_with_names(3). File descriptors submitted without
191 this field set, will implicitly get the name "stored" assigned.
192 Note that, if multiple file descriptors are submitted at once, the
193 specified name will be assigned to all of them. In order to assign
194 different names to submitted file descriptors, submit them in
195 separate invocations of sd_pid_notify_with_fds(). The name may
196 consist of arbitrary ASCII characters except control characters or
197 ":". It may not be longer than 255 characters. If a submitted name
198 does not follow these restrictions, it is ignored.
199 FDPOLL=0
201 When used in combination with FDSTORE=1, disables polling of the
202 stored file descriptors regardless of whether or not they are
203 pollable. As this option disables automatic cleanup of the stored
204 file descriptors on EPOLLERR and EPOLLHUP, care must be taken to
205 ensure proper manual cleanup. Use of this option is not generally
206 recommended except for when automatic cleanup has unwanted behavior
207 such as prematurely discarding file descriptors from the store.
208 BARRIER=1
210 Tells the service manager that the client is explicitly requesting
211 synchronization by means of closing the file descriptor sent with
212 this command. The service manager guarantees that the processing of
213 a BARRIER=1 command will only happen after all previous
214 notification messages sent before this command have been processed.
215 Hence, this command accompanied with a single file descriptor can
216 be used to synchronize against reception of all previous status
217 messages. Note that this command cannot be mixed with other
218 notifications, and has to be sent in a separate message to the
219 service manager, otherwise all assignments will be ignored. Note
220 that sending 0 or more than 1 file descriptor with this command is
221 a violation of the protocol.
222 It is recommended to prefix variable names that are not listed above
224 with X_ to avoid namespace clashes.
225 Note that systemd will accept status data sent from a service only if
227 the NotifyAccess= option is correctly set in the service definition
228 file. See systemd.service(5) for details.
229 Note that sd_notify() notifications may be attributed to units
231 correctly only if either the sending process is still around at the
232 time PID 1 processes the message, or if the sending process is
233 explicitly runtime-tracked by the service manager. The latter is the
234 case if the service manager originally forked off the process, i.e. on
235 all processes that match NotifyAccess=main or NotifyAccess=exec.
236 Conversely, if an auxiliary process of the unit sends an sd_notify()
237 message and immediately exits, the service manager might not be able to
238 properly attribute the message to the unit, and thus will ignore it,
239 even if NotifyAccess=all is set for it.
240 Hence, to eliminate all race conditions involving lookup of the
242 client's unit and attribution of notifications to units correctly,
243 sd_notify_barrier() may be used. This call acts as a synchronization
244 point and ensures all notifications sent before this call have been
245 picked up by the service manager when it returns successfully. Use of
246 sd_notify_barrier() is needed for clients which are not invoked by the
247 service manager, otherwise this synchronization mechanism is
248 unnecessary for attribution of notifications to the unit.
249 sd_notifyf() is similar to sd_notify() but takes a printf()-like format
251 string plus arguments.
252 sd_pid_notify() and sd_pid_notifyf() are similar to sd_notify() and
254 sd_notifyf() but take a process ID (PID) to use as originating PID for
255 the message as first argument. This is useful to send notification
256 messages on behalf of other processes, provided the appropriate
257 privileges are available. If the PID argument is specified as 0, the
258 process ID of the calling process is used, in which case the calls are
259 fully equivalent to sd_notify() and sd_notifyf().
260 sd_pid_notify_with_fds() is similar to sd_pid_notify() but takes an
262 additional array of file descriptors. These file descriptors are sent
263 along the notification message to the service manager. This is
264 particularly useful for sending "FDSTORE=1" messages, as described
265 above. The additional arguments are a pointer to the file descriptor
266 array plus the number of file descriptors in the array. If the number
267 of file descriptors is passed as 0, the call is fully equivalent to
268 sd_pid_notify(), i.e. no file descriptors are passed. Note that file
269 descriptors sent to the service manager on a message without
270 "FDSTORE=1" are immediately closed on reception.
271 sd_pid_notifyf_with_fds() is a combination of sd_pid_notify_with_fds()
273 and sd_notifyf(), i.e. it accepts both a PID and a set of file
274 descriptors as input, and processes a format string to generate the
275 state string.
276 sd_notify_barrier() allows the caller to synchronize against reception
278 of previously sent notification messages and uses the BARRIER=1
279 command. It takes a relative timeout value in microseconds which is
280 passed to ppoll(2). A value of UINT64_MAX is interpreted as infinite
281 timeout.
282 sd_pid_notify_barrier() is just like sd_notify_barrier(), but allows
284 specifying the originating PID for the notification message.
285
287 On failure, these calls return a negative errno-style error code. If
288 $NOTIFY_SOCKET was not set and hence no status message could be sent, 0
289 is returned. If the status was sent, these functions return a positive
290 value. In order to support both service managers that implement this
291 scheme and those which do not, it is generally recommended to ignore
292 the return value of this call. Note that the return value simply
293 indicates whether the notification message was enqueued properly, it
294 does not reflect whether the message could be processed successfully.
295 Specifically, no error is returned when a file descriptor is attempted
296 to be stored using FDSTORE=1 but the service is not actually configured
297 to permit storing of file descriptors (see above).
298
300 Functions described here are available as a shared library, which can
301 be compiled against and linked to with the libsystemd pkg-config(1)
302 file.
303 The code described here uses getenv(3), which is declared to be not
305 multi-thread-safe. This means that the code calling the functions
306 described here must not call setenv(3) from a parallel thread. It is
307 recommended to only do calls to setenv() from an early phase of the
308 program when no other threads have been started.
309 These functions send a single datagram with the state string as payload
311 to the socket referenced in the $NOTIFY_SOCKET environment variable. If
312 the first character of $NOTIFY_SOCKET is "/" or "@", the string is
313 understood as an AF_UNIX or Linux abstract namespace socket
314 (respectively), and in both cases the datagram is accompanied by the
315 process credentials of the sending service, using SCM_CREDENTIALS. If
316 the string starts with "vsock:" then the string is understood as an
317 AF_VSOCK address, which is useful for hypervisors/VMMs or other
318 processes on the host to receive a notification when a virtual machine
319 has finished booting. Note that in case the hypervisor does not support
320 SOCK_DGRAM over AF_VSOCK, SOCK_SEQPACKET will be used instead. The
321 address should be in the form: "vsock:CID:PORT". Note that unlike other
322 uses of vsock, the CID is mandatory and cannot be "VMADDR_CID_ANY".
323 Note that PID1 will send the VSOCK packets from a privileged port
324 (i.e.: lower than 1024), as an attempt to address concerns that
325 unprivileged processes in the guest might try to send malicious
326 notifications to the host, driving it to make destructive decisions
327 based on them.
328
330 $NOTIFY_SOCKET
331 Set by the service manager for supervised processes for status and
332 start-up completion notification. This environment variable
333 specifies the socket sd_notify() talks to. See above for details.
334
336 Example 1. Start-up Notification
337 When a service finished starting up, it might issue the following call
339 to notify the service manager:
340 sd_notify(0, "READY=1");
342 Example 2. Extended Start-up Notification
344 A service could send the following after completing initialization:
346 sd_notifyf(0, "READY=1\n"
348 "STATUS=Processing requests...\n"
349 "MAINPID=%lu",
350 (unsigned long) getpid());
351 Example 3. Error Cause Notification
353 A service could send the following shortly before exiting, on failure:
355 sd_notifyf(0, "STATUS=Failed to start up: %s\n"
357 "ERRNO=%i",
358 strerror_r(errnum, (char[1024]){}, 1024),
359 errnum);
360 Example 4. Store a File Descriptor in the Service Manager
362 To store an open file descriptor in the service manager, in order to
364 continue operation after a service restart without losing state, use
365 "FDSTORE=1":
366 sd_pid_notify_with_fds(0, 0, "FDSTORE=1\nFDNAME=foobar", &fd, 1);
368 Example 5. Eliminating race conditions
370 When the client sending the notifications is not spawned by the service
372 manager, it may exit too quickly and the service manager may fail to
373 attribute them correctly to the unit. To prevent such races, use
374 sd_notify_barrier() to synchronize against reception of all
375 notifications sent before this call is made.
376 sd_notify(0, "READY=1");
378 /* set timeout to 5 seconds */
379 sd_notify_barrier(0, 5 * 1000000);
380
383 systemd(1), sd-daemon(3), sd_listen_fds(3),
384 sd_listen_fds_with_names(3), sd_watchdog_enabled(3), daemon(7),
385 systemd.service(5)
386
388 1. File Descriptor Store
389 https://systemd.io/FILE_DESCRIPTOR_STORE
390systemd 254 SD_NOTIFY(3)