sd_notify, sd_notifyf, sd_pid_notify, sd_pid_notifyf, sd_pid_notify_with_fds, sd_pid_notifyf_with_fds, sd_notify_barrier, sd_pid_notify_barrier — Notify service manager about start-up completion and other service status changes
#include <systemd/sd-daemon.h>
int sd_notify( | int unset_environment, |
const char *state); |
int sd_notifyf( | int unset_environment, |
| const char *format, | |
…); |
int sd_pid_notify( | pid_t pid, |
| int unset_environment, | |
const char *state); |
int sd_pid_notifyf( | pid_t pid, |
| int unset_environment, | |
| const char *format, | |
…); |
int sd_pid_notify_with_fds( | pid_t pid, |
| int unset_environment, | |
| const char *state, | |
| const int *fds, | |
unsigned n_fds); |
int sd_pid_notifyf_with_fds( | pid_t pid, |
| int unset_environment, | |
| const int *fds, | |
| size_t n_fds, | |
| const char *format, | |
…); |
int sd_notify_barrier( | int unset_environment, |
uint64_t timeout); |
int sd_pid_notify_barrier( | pid_t pid, |
| int unset_environment, | |
uint64_t timeout); |
sd_notify() may be called by a service to notify the service manager about
state changes. It can be used to send arbitrary information, encoded in an environment-block-like string.
Most importantly, it can be used for start-up or reload completion notifications.
If the unset_environment parameter is non-zero,
sd_notify() will unset the $NOTIFY_SOCKET environment variable
before returning (regardless of whether the function call itself succeeded or not). Further calls to
sd_notify() will then silently do nothing, and the variable is no longer inherited
by child processes.
The state parameter should contain a newline-separated list of variable
assignments, similar in style to an environment block. A trailing newline is implied if none is
specified. The string may contain any kind of variable assignments, but see the next section
for a list of assignments understood by the service manager.
Note that systemd will accept status data sent from a service only if the
NotifyAccess= option is correctly set in the service definition file. See
systemd.service(5) for
details.
Note that sd_notify() notifications may be attributed to units correctly only
if either the sending process is still around at the time PID 1 processes the message, or if the sending
process is explicitly runtime-tracked by the service manager. The latter is the case if the service
manager originally forked off the process, i.e. on all processes that match
NotifyAccess=main or
NotifyAccess=exec. Conversely, if an auxiliary process of the unit
sends an sd_notify() message and immediately exits, the service manager might not be
able to properly attribute the message to the unit, and thus will ignore it, even if
NotifyAccess=all is set for it.
Hence, to eliminate all race conditions involving lookup of the client's unit and attribution of
notifications to units correctly, sd_notify_barrier() may be used. This call acts as
a synchronization point and ensures all notifications sent before this call have been picked up by the
service manager when it returns successfully. Use of sd_notify_barrier() is needed
for clients which are not invoked by the service manager, otherwise this synchronization mechanism is
unnecessary for attribution of notifications to the unit.
sd_notifyf() is similar to sd_notify() but takes a
printf()-like format string plus arguments.
sd_pid_notify() and sd_pid_notifyf() are similar to
sd_notify() and sd_notifyf() but take a process ID (PID) to use
as originating PID for the message as first argument. This is useful to send notification messages on
behalf of other processes, provided the appropriate privileges are available. If the PID argument is
specified as 0, the process ID of the calling process is used, in which case the calls are fully
equivalent to sd_notify() and sd_notifyf().
sd_pid_notify_with_fds() is similar to sd_pid_notify()
but takes an additional array of file descriptors. These file descriptors are sent along the notification
message to the service manager. This is particularly useful for sending "FDSTORE=1"
messages, as described above. The additional arguments are a pointer to the file descriptor array plus
the number of file descriptors in the array. If the number of file descriptors is passed as 0, the call
is fully equivalent to sd_pid_notify(), i.e. no file descriptors are passed. Note
that file descriptors sent to the service manager on a message without "FDSTORE=1" are
immediately closed on reception.
sd_pid_notifyf_with_fds() is a combination of
sd_pid_notify_with_fds() and sd_notifyf(), i.e. it accepts both
a PID and a set of file descriptors as input, and processes a format string to generate the state
string.
sd_notify_barrier() allows the caller to synchronize against reception of
previously sent notification messages and uses the BARRIER=1 command. It takes a
relative timeout value in microseconds which is passed to
ppoll(2). A value of UINT64_MAX is interpreted as infinite timeout.
sd_pid_notify_barrier() is just like sd_notify_barrier(),
but allows specifying the originating PID for the notification message.
The following assignments have a defined meaning:
Tells the service manager that service startup is finished, or the service finished
re-loading its configuration. This is only used by systemd if the service definition file has
Type=notify or Type=notify-reload set. Since there is little
value in signaling non-readiness, the only value services should send is "READY=1"
(i.e. "READY=0" is not defined).
Tells the service manager that the service is beginning to reload its configuration.
This is useful to allow the service manager to track the service's internal state, and present it to
the user. Note that a service that sends this notification must also send a
"READY=1" notification when it completed reloading its configuration. Reloads the
service manager is notified about with this mechanisms are propagated in the same way as they are
when originally initiated through the service manager. This message is particularly relevant for
Type=notify-reload services, to inform the service manager that the request to
reload the service has been received and is now being processed.
Tells the service manager that the service is beginning its shutdown. This is useful to allow the service manager to track the service's internal state, and present it to the user.
A field carrying the monotonic timestamp (as per
CLOCK_MONOTONIC) formatted in decimal in μs, when the notification message was
generated by the client. This is typically used in combination with "RELOADING=1",
to allow the service manager to properly synchronize reload cycles. See
systemd.service(5)
for details, specifically "Type=notify-reload".
Passes a single-line UTF-8 status string back to the service manager that describes
the service state. This is free-form and can be used for various purposes: general state feedback,
fsck-like programs could pass completion percentages and failing programs could pass a human-readable
error message. Example: "STATUS=Completed 66% of file system check…"
Reset the access to the service status notification socket during runtime, overriding
NotifyAccess= setting in the service unit file. See
systemd.service(5)
for details, specifically "NotifyAccess=" for a list of accepted values.
If a service fails, the errno-style error code, formatted as string. Example:
"ERRNO=2" for ENOENT.
If a service fails, the D-Bus error-style error code. Example:
"BUSERROR=org.freedesktop.DBus.Error.TimedOut".
If a service fails, the Varlink error-style error code. Example:
"VARLINKERROR=org.varlink.service.InvalidParameter".
The exit status of a service or the manager itself. Note that systemd currently does not consume this value when sent by services, so this assignment is only informational. The manager will send this notification to its notification socket, which may be used to collect an exit status from the system (a container or VM) as it shuts down. For example, mkosi(1) makes use of this. The value to return may be set via the systemctl(1) exit verb.
Change the main process ID (PID) of the service. This is especially useful in the case
where the real main process isn't directly forked off by the service manager.
Example: "MAINPID=4711".
The pidfd inode number of the new main process (specified through MAINPID=).
This information can be acquired through
fstat(2)
on the pidfd and is used to identify the process in a race-free fashion. Alternatively,
a pidfd can be sent directly to the service manager (see MAINPIDFD=1 below).
Similar to MAINPID= with MAINPIDFDID=, but
the process is referenced directly by the pidfd passed to the service manager. This is useful
if pidfd id is not supported on the system. Exactly one fd is expected for this notification.
Tells the service manager to update the watchdog timestamp. This is the keep-alive
ping that services need to issue in regular intervals if WatchdogSec= is enabled
for it. See
systemd.service(5)
for information how to enable this functionality and
sd_watchdog_enabled(3)
for the details of how the service can check whether the watchdog is enabled.
Tells the service manager that the service detected an internal error that should be
handled by the configured watchdog options. This will trigger the same behaviour as if
WatchdogSec= is enabled and the service did not send "WATCHDOG=1"
in time. Note that WatchdogSec= does not need to be enabled for
"WATCHDOG=trigger" to trigger the watchdog action. See
systemd.service(5)
for information about the watchdog behavior.
Reset watchdog_usec value during runtime. Notice that this is not
available when using sd_event_set_watchdog() or
sd_watchdog_enabled(). Example :
"WATCHDOG_USEC=20000000"
Tells the service manager to extend the startup, runtime or shutdown service timeout
corresponding the current state. The value specified is a time in microseconds during which the
service must send a new message. A service timeout will occur if the message isn't received, but only
if the runtime of the current state is beyond the original maximum times of
TimeoutStartSec=, RuntimeMaxSec=, and
TimeoutStopSec=. See
systemd.service(5)
for effects on the service timeouts.
Store file descriptors in the service manager. File descriptors sent this way will be
held for the service by the service manager and will later be handed back using the usual file
descriptor passing logic at the next start or restart of the service, see
sd_listen_fds(3).
Any open sockets and other file descriptors which should not be closed during a restart may be stored
this way. When a service is stopped, its file descriptor store is discarded and all file descriptors
in it are closed, except when overridden with FileDescriptorStorePreserve=, see
systemd.service(5).
The service manager will accept messages for a service only if its
FileDescriptorStoreMax= setting is non-zero (defaults to zero, see
systemd.service(5)).
The service manager will set the $FDSTORE environment variable for services that
have the file descriptor store enabled, see
systemd.exec(5).
If FDPOLL=0 is not set and the file descriptors are pollable (see
epoll_ctl(2)), then
any EPOLLHUP or EPOLLERR event seen on them will result in
their automatic removal from the store.
Multiple sets of file descriptors may be sent in separate messages, in which case the sets are combined. The service manager removes duplicate file descriptors (those pointing to the same object) before passing them to the service.
This functionality should be used to implement services that can restart after an explicit
request or a crash without losing state. Application state can either be serialized to a file in
/run/, or better, stored in a
memfd_create(2)
memory file descriptor. Use sd_pid_notify_with_fds() to send messages with
"FDSTORE=1". It is recommended to combine FDSTORE= with
FDNAME= to make it easier to manage the stored file descriptors.
For further information on the file descriptor store see the File Descriptor Store overview.
Removes file descriptors from the file descriptor store. This field needs to be
combined with FDNAME= to specify the name of the file descriptors to
remove.
When used in combination with FDSTORE=1, specifies a name for the
submitted file descriptors. When used with FDSTOREREMOVE=1, specifies the name for
the file descriptors to remove. This name is passed to the service during activation, and may be
queried using
sd_listen_fds_with_names(3).
File descriptors submitted without this field will be called "stored".
The name may consist of arbitrary ASCII characters except control characters or
":". It may not be longer than 255 characters. If a submitted name does not follow
these restrictions, it is ignored.
Note that if multiple file descriptors are submitted in a single message, the specified name will be used for all of them. In order to assign different names to submitted file descriptors, submit them in separate messages.
When used in combination with FDSTORE=1, disables polling of the
submitted file descriptors regardless of whether or not they are pollable. As this option disables
automatic cleanup of the submitted file descriptors on EPOLLERR and EPOLLHUP, care must be taken to
ensure proper manual cleanup. Use of this option is not generally recommended except for when
automatic cleanup has unwanted behavior such as prematurely discarding file descriptors from the
store.
Tells the service manager that the client is explicitly requesting synchronization by
means of closing the file descriptor sent with this command. The service manager guarantees that the
processing of a BARRIER=1 command will only happen after all previous notification
messages sent before this command have been processed. Hence, this command accompanied with a single
file descriptor can be used to synchronize against reception of all previous status messages. Note
that this command cannot be mixed with other notifications, and has to be sent in a separate message
to the service manager, otherwise all assignments will be ignored. Note that sending 0 or more than 1
file descriptor with this command is a violation of the protocol.
The notification messages sent by services are interpreted by the service manager. Unknown
assignments are ignored. Thus, it is safe (but often without effect) to send assignments which are not
in this list. The protocol is extensible, but care should be taken to ensure private extensions are
recognizable as such. Specifically, it is recommend to prefix them with "X_" followed by
some namespace identifier. The service manager also sends some messages to its
notification socket, which may then consumed by a supervising machine or container manager further up the
stack. The service manager sends a number of extension fields, for example
X_SYSTEMD_UNIT_ACTIVE=, for details see
systemd(1).
On failure, these calls return a negative errno-style error code. If
$NOTIFY_SOCKET was not set and hence no status message could be sent, 0 is
returned. If the status was sent, these functions return a positive value. In order to support both
service managers that implement this scheme and those which do not, it is generally recommended to ignore
the return value of this call. Note that the return value simply indicates whether the notification
message was enqueued properly, it does not reflect whether the message could be processed
successfully. Specifically, no error is returned when a file descriptor is attempted to be stored using
FDSTORE=1 but the service is not actually configured to permit storing of file
descriptors (see above).
Functions described here are available as a shared
library, which can be compiled against and linked to with the
libsystemd pkg-config(1)
file.
The code described here uses
getenv(3),
which is declared to be not multi-thread-safe. This means that the code calling the functions described
here must not call
setenv(3)
from a parallel thread. It is recommended to only do calls to setenv()
from an early phase of the program when no other threads have been started.
These functions send a single datagram with the state string as payload to the socket referenced in
the $NOTIFY_SOCKET environment variable. If the first character of
$NOTIFY_SOCKET is "/" or "@", the string is
understood as an AF_UNIX or Linux abstract namespace socket (respectively), and in
both cases the datagram is accompanied by the process credentials of the sending service, using
SCM_CREDENTIALS. If the string starts with "vsock:" then the string is understood as an
AF_VSOCK address, which is useful for hypervisors/VMMs or other processes on the
host to receive a notification when a virtual machine has finished booting. Note that in case the
hypervisor does not support SOCK_DGRAM over AF_VSOCK,
SOCK_SEQPACKET will be used instead. "vsock-stream",
"vsock-dgram" and "vsock-seqpacket" can be used instead of
"vsock" to force usage of the corresponding socket type. The address should be in the
form: "vsock:CID:PORT". Note that unlike other uses of vsock, the CID is mandatory and
cannot be "VMADDR_CID_ANY". Note that PID1 will send the VSOCK packets from a
privileged port (i.e.: lower than 1024), as an attempt to address concerns that unprivileged processes in
the guest might try to send malicious notifications to the host, driving it to make destructive decisions
based on them.
Note that, while using this library should be preferred in order to avoid code duplication, it is also possible to reimplement the simple readiness notification protocol without external dependencies, as demonstrated in the following self-contained examples from several languages:
/* SPDX-License-Identifier: MIT-0 */
/* Implement the systemd notify protocol without external dependencies.
* Supports both readiness notification on startup and on reloading,
* according to the protocol defined at:
* https://www.freedesktop.org/software/systemd/man/latest/sd_notify.html
* This protocol is guaranteed to be stable as per:
* https://systemd.io/PORTABILITY_AND_STABILITY/ */
#define _GNU_SOURCE 1
#include <errno.h>
#include <inttypes.h>
#include <signal.h>
#include <stdbool.h>
#include <stddef.h>
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <sys/socket.h>
#include <sys/un.h>
#include <time.h>
#include <unistd.h>
#define _cleanup_(f) __attribute__((cleanup(f)))
static void closep(int *fd) {
if (!fd || *fd < 0)
return;
close(*fd);
*fd = -1;
}
static int notify(const char *message) {
union sockaddr_union {
struct sockaddr sa;
struct sockaddr_un sun;
} socket_addr = {
.sun.sun_family = AF_UNIX,
};
size_t path_length, message_length;
_cleanup_(closep) int fd = -1;
const char *socket_path;
/* Verify the argument first */
if (!message)
return -EINVAL;
message_length = strlen(message);
if (message_length == 0)
return -EINVAL;
/* If the variable is not set, the protocol is a noop */
socket_path = getenv("NOTIFY_SOCKET");
if (!socket_path)
return 0; /* Not set? Nothing to do */
/* Only AF_UNIX is supported, with path or abstract sockets */
if (socket_path[0] != '/' && socket_path[0] != '@')
return -EAFNOSUPPORT;
path_length = strlen(socket_path);
/* Ensure there is room for NUL byte */
if (path_length >= sizeof(socket_addr.sun.sun_path))
return -E2BIG;
memcpy(socket_addr.sun.sun_path, socket_path, path_length);
/* Support for abstract socket */
if (socket_addr.sun.sun_path[0] == '@')
socket_addr.sun.sun_path[0] = 0;
fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0);
if (fd < 0)
return -errno;
if (connect(fd, &socket_addr.sa, offsetof(struct sockaddr_un, sun_path) + path_length) != 0)
return -errno;
ssize_t written = write(fd, message, message_length);
if (written != (ssize_t) message_length)
return written < 0 ? -errno : -EPROTO;
return 1; /* Notified! */
}
static int notify_ready(void) {
return notify("READY=1");
}
static int notify_reloading(void) {
/* A buffer with length sufficient to format the maximum UINT64 value. */
char reload_message[sizeof("RELOADING=1\nMONOTONIC_USEC=18446744073709551615")];
struct timespec ts;
uint64_t now;
/* Notify systemd that we are reloading, including a CLOCK_MONOTONIC timestamp in usec
* so that the program is compatible with a Type=notify-reload service. */
if (clock_gettime(CLOCK_MONOTONIC, &ts) < 0)
return -errno;
if (ts.tv_sec < 0 || ts.tv_nsec < 0 ||
(uint64_t) ts.tv_sec > (UINT64_MAX - (ts.tv_nsec / 1000ULL)) / 1000000ULL)
return -EINVAL;
now = (uint64_t) ts.tv_sec * 1000000ULL + (uint64_t) ts.tv_nsec / 1000ULL;
if (snprintf(reload_message, sizeof(reload_message), "RELOADING=1\nMONOTONIC_USEC=%" PRIu64, now) < 0)
return -EINVAL;
return notify(reload_message);
}
static int notify_stopping(void) {
return notify("STOPPING=1");
}
static volatile sig_atomic_t reloading = 0;
static volatile sig_atomic_t terminating = 0;
static void signal_handler(int sig) {
if (sig == SIGHUP)
reloading = 1;
else if (sig == SIGINT || sig == SIGTERM)
terminating = 1;
}
int main(int argc, char **argv) {
struct sigaction sa = {
.sa_handler = signal_handler,
.sa_flags = SA_RESTART,
};
int r;
/* Setup signal handlers */
sigemptyset(&sa.sa_mask);
sigaction(SIGHUP, &sa, NULL);
sigaction(SIGINT, &sa, NULL);
sigaction(SIGTERM, &sa, NULL);
/* Do more service initialization work here … */
/* Now that all the preparations steps are done, signal readiness */
r = notify_ready();
if (r < 0) {
fprintf(stderr, "Failed to notify readiness to $NOTIFY_SOCKET: %s\n", strerror(-r));
return EXIT_FAILURE;
}
while (!terminating) {
if (reloading) {
reloading = false;
/* As a separate but related feature, we can also notify the manager
* when reloading configuration. This allows accurate state-tracking,
* and also automated hook-in of 'systemctl reload' without having to
* specify manually an ExecReload= line in the unit file. */
r = notify_reloading();
if (r < 0) {
fprintf(stderr, "Failed to notify reloading to $NOTIFY_SOCKET: %s\n", strerror(-r));
return EXIT_FAILURE;
}
/* Do some reconfiguration work here … */
r = notify_ready();
if (r < 0) {
fprintf(stderr, "Failed to notify readiness to $NOTIFY_SOCKET: %s\n", strerror(-r));
return EXIT_FAILURE;
}
}
/* Do some daemon work here … */
sleep(5);
}
r = notify_stopping();
if (r < 0) {
fprintf(stderr, "Failed to report termination to $NOTIFY_SOCKET: %s\n", strerror(-r));
return EXIT_FAILURE;
}
/* Do some shutdown work here … */
return EXIT_SUCCESS;
}
#!/usr/bin/env python3
# SPDX-License-Identifier: MIT-0
#
# Implement the systemd notify protocol without external dependencies.
# Supports both readiness notification on startup and on reloading,
# according to the protocol defined at:
# https://www.freedesktop.org/software/systemd/man/latest/sd_notify.html
# This protocol is guaranteed to be stable as per:
# https://systemd.io/PORTABILITY_AND_STABILITY/
import errno
import os
import signal
import socket
import sys
import time
reloading = False
terminating = False
def notify(message):
if not message:
raise ValueError("notify() requires a message")
socket_path = os.environ.get("NOTIFY_SOCKET")
if not socket_path:
return
if socket_path[0] not in ("/", "@"):
raise OSError(errno.EAFNOSUPPORT, "Unsupported socket type")
# Handle abstract socket.
if socket_path[0] == "@":
socket_path = "\0" + socket_path[1:]
with socket.socket(socket.AF_UNIX, socket.SOCK_DGRAM | socket.SOCK_CLOEXEC) as sock:
sock.connect(socket_path)
sock.sendall(message)
def notify_ready():
notify(b"READY=1")
def notify_reloading():
microsecs = time.clock_gettime_ns(time.CLOCK_MONOTONIC) // 1000
notify(f"RELOADING=1\nMONOTONIC_USEC={microsecs}".encode())
def notify_stopping():
notify(b"STOPPING=1")
def reload(signum, frame):
global reloading
reloading = True
def terminate(signum, frame):
global terminating
terminating = True
def main():
print("Doing initial setup")
global reloading, terminating
# Set up signal handlers.
print("Setting up signal handlers")
signal.signal(signal.SIGHUP, reload)
signal.signal(signal.SIGINT, terminate)
signal.signal(signal.SIGTERM, terminate)
# Do any other setup work here.
# Once all setup is done, signal readiness.
print("Done setting up")
notify_ready()
print("Starting loop")
while not terminating:
if reloading:
print("Reloading")
reloading = False
# Support notifying the manager when reloading configuration.
# This allows accurate state tracking as well as automatically
# enabling 'systemctl reload' without needing to manually
# specify an ExecReload= line in the unit file.
notify_reloading()
# Do some reconfiguration work here.
print("Done reloading")
notify_ready()
# Do the real work here ...
print("Sleeping for five seconds")
time.sleep(5)
print("Terminating")
notify_stopping()
if __name__ == "__main__":
sys.stdout.reconfigure(line_buffering=True)
print("Starting app")
main()
print("Stopped app")
$NOTIFY_SOCKET¶Set by the service manager for supervised processes for status and start-up
completion notification. This environment variable specifies the socket
sd_notify() talks to. See above for details.
Example 1. Start-up Notification
When a service finished starting up, it might issue the following call to notify the service manager:
sd_notify(0, "READY=1");
Example 2. Extended Start-up Notification
A service could send the following after completing initialization:
sd_notifyf(0, "READY=1\n"
"STATUS=Processing requests…\n"
"MAINPID=%lu",
(unsigned long) getpid());Example 3. Error Cause Notification
A service could send the following shortly before exiting, on failure:
sd_notifyf(0, "STATUS=Failed to start up: %s\n"
"ERRNO=%i",
strerror_r(errnum, (char[1024]){}, 1024),
errnum);Example 4. Store a File Descriptor in the Service Manager
To store an open file descriptor in the service manager, in order to continue operation after a
service restart without losing state, use "FDSTORE=1":
sd_pid_notify_with_fds(0, 0, "FDSTORE=1\nFDNAME=foobar", &fd, 1);
Example 5. Eliminating race conditions
When the client sending the notifications is not spawned by the service manager, it may exit too
quickly and the service manager may fail to attribute them correctly to the unit. To prevent such
races, use sd_notify_barrier() to synchronize against reception of all
notifications sent before this call is made.
sd_notify(0, "READY=1"); /* set timeout to 5 seconds */ sd_notify_barrier(0, 5 * 1000000);
sd_pid_notify(),
sd_pid_notifyf(), and
sd_pid_notify_with_fds() were added in version 219.
sd_notify_barrier() was added in version 246.
sd_pid_notifyf_with_fds() and
sd_pid_notify_barrier() were added in version 254.