[rrd-developers] [RFC] implement rrd_daemon socket activation for systemd
Tomasz Torcz
tomek at pipebreaker.pl
Sun Apr 15 20:32:59 CEST 2012
Greetings,
I've implemented socket-activitability for Linux' systemd service management
daemon. For background, please visit:
http://0pointer.de/blog/projects/socket-activation.html
http://0pointer.de/blog/projects/socket-activation2.html
Socket management by systemd brings very important feature: no connection are
lost, even if made during daemon restart or if daemon crashes. In this case
systemd holds the connection, restarts service and handles connection to fresh
rrdcached instance.
Also, this scheme gives rrdcached ability to do on-demand startup and as
a side effect gives support for few more socket families than AF_UNIX and
network sockets. Systemd's limits also apply. More details can be found
at systemd.socket man page or at
http://www.freedesktop.org/software/systemd/man/systemd.socket.html
One big missing piece is authorisation. It completely not working with
current patch, so caveat emptor.
First patch add the code support for socket-activation, and adds two
liberally licensed files (sd-daemon.{c,h}). Functions implemented there
return 0 when systemd is not in use and are NOP on non-Linux platforms.
Optionally, systemd-devel library can be use for those function, but
at cost of few more ifdefs.
Please note, if systemd passed sockets are detected, sockets
given on commandline and default socket are ignored.
No autotools integration is provided (yet), so sd-daemon.c must be
manually compiled with rrd_daemon.c. ('man 7 daemon' contains autotools
scriplets if someone is interested).
Second patch sticks sd_notify in RRD_LOG macro, to augment status messages
in systemctl. Third patch *would be* converting rrdcached to Type=notify
type of service (again, man 7 daemon). Third patch is not done.
Please let me know if this way is acceptable and should I bother with
updating the manual page and wrestling with autotools.
So how does it look like?
1) Let's see what happens with socket & service shut down:
# systemctl status rrdcached.socket rrdcached.service
rrdcached.socket - sockets to activate rrdcached
Loaded: loaded (/etc/systemd/system/rrdcached.socket; disabled)
Active: inactive (dead) since Sun, 15 Apr 2012 20:18:57 +0200; 5s ago
rrdcached.service - Data caching daemon for rrdtool
Loaded: loaded (/etc/systemd/system/rrdcached.service; disabled)
Active: inactive (dead) since Fri, 13 Apr 2012 23:14:54 +0200; 1 day and 21h ago
2) check with rrdupdate:
$ rrdupdate --daemon unix:/tmp/rrdcached.sock whatever.rrd N:2
ERROR: Unable to connect to rrdcached: Connection refused
3) let's start socket
# systemctl start rrdcached.socket; systemctl status rrdcached.socket rrdcached.service
rrdcached.socket - sockets to activate rrdcached
Loaded: loaded (/etc/systemd/system/rrdcached.socket; disabled)
Active: active (listening) since Sun, 15 Apr 2012 20:22:25 +0200; 2s ago
rrdcached.service - Data caching daemon for rrdtool
Loaded: loaded (/etc/systemd/system/rrdcached.service; disabled)
Active: inactive (dead)
4) do the test again
$ rrdupdate --daemon unix:/tmp/rrdcached.sock whatever.rrd N:2
ERROR: realpath(whatever.rrd): No such file or director
Notice, it's no longer Conn refused. The error message comes from daemon.
5) see that daemon was started on-demand:
# systemctl status rrdcached.socket rrdcached.service
rrdcached.socket - sockets to activate rrdcached
Loaded: loaded (/etc/systemd/system/rrdcached.socket; disabled)
Active: active (running) since Sun, 15 Apr 2012 20:22:25 +0200; 2min 0s ago
rrdcached.service - Data caching daemon for rrdtool
Loaded: loaded (/etc/systemd/system/rrdcached.service; disabled)
Active: active (running) since Sun, 15 Apr 2012 20:23:26 +0200; 1min 0s ago
Process: 7926 ExecStart=/tmp/rrdcached (code=exited, status=0/SUCCESS)
Status: "listening for connections"
Main PID: 7950 (lt-rrdcached)
└ 7950 /home/tomek/Dokumenty/dev/rrdtool-trunk/src/.libs/lt-rrdcached
Apr 15 20:23:26 sandtrout.local rrdcached[7950]: starting up
Apr 15 20:23:26 sandtrout.local rrdcached[7950]: listening for connections
6) stop the daemon
# systemctl stop rrdcache.service
# systemctl status rrdcached.socket rrdcached.service
rrdcached.socket - sockets to activate rrdcached
Loaded: loaded (/etc/systemd/system/rrdcached.socket; disabled)
Active: active (listening) since Sun, 15 Apr 2012 20:22:25 +0200; 3min 42s ago
rrdcached.service - Data caching daemon for rrdtool
Loaded: loaded (/etc/systemd/system/rrdcached.service; disabled)
Active: inactive (dead) since Sun, 15 Apr 2012 20:25:13 +0200; 53s ago
Process: 7926 ExecStart=/tmp/rrdcached (code=exited, status=0/SUCCESS)
Status: "goodbye"
Main PID: 7950 (code=exited, status=0/SUCCESS)
Apr 15 20:23:26 sandtrout.local rrdcached[7950]: listening for connections
Apr 15 20:25:13 sandtrout.local rrdcached[7950]: caught SIGTERM
Apr 15 20:25:13 sandtrout.local rrdcached[7950]: starting shutdown
Apr 15 20:25:13 sandtrout.local rrdcached[7950]: clean shutdown; all RRDs flushed
Apr 15 20:25:13 sandtrout.local rrdcached[7950]: goodbye
Attached are patch 1 & 2 and unit definition files for systemd.
--
Tomasz Torcz "God, root, what's the difference?"
xmpp: zdzichubg at chrome.pl "God is more forgiving."
-------------- next part --------------
From d2ed090be0e221dd35dfd4c477e9c7d1ed3e986b Mon Sep 17 00:00:00 2001
From: Tomasz Torcz <tomek at pipebreaker.pl>
Date: Fri, 13 Apr 2012 14:30:01 +0200
Subject: [PATCH 1/2] - add socket-activation for systemd
---
src/rrd_daemon.c | 81 +++++++--
src/sd-daemon.c | 530 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
src/sd-daemon.h | 282 +++++++++++++++++++++++++++++
3 files changed, 875 insertions(+), 18 deletions(-)
create mode 100644 src/sd-daemon.c
create mode 100644 src/sd-daemon.h
diff --git a/src/rrd_daemon.c b/src/rrd_daemon.c
index 2c81424..f1e79e1 100644
--- a/src/rrd_daemon.c
+++ b/src/rrd_daemon.c
@@ -64,6 +64,7 @@
*/
/* {{{ */
+#include "sd-daemon.h"
#include "rrd_tool.h"
#include "rrd_client.h"
#include "unused.h"
@@ -2975,6 +2976,63 @@ static int open_listen_socket (const listen_socket_t *sock) /* {{{ */
return (open_listen_socket_network(sock));
} /* }}} int open_listen_socket */
+static void open_listen_sockets_systemd(int number) /* {{{ */
+{
+ listen_socket_t *temp;
+ struct sockaddr_un sa;
+ socklen_t l;
+ int sd_fd;
+
+ temp = (listen_socket_t *) rrd_realloc (listen_fds,
+ sizeof (listen_fds[0]) * (listen_fds_num + number));
+ if (temp == NULL)
+ {
+ fprintf (stderr, "rrdcached: open_listen_socket_systemd: realloc failed.\n");
+ return;
+ }
+ listen_fds = temp;
+
+ for (int i = 0; i < number; i++)
+ {
+ sd_fd = SD_LISTEN_FDS_START + i;
+
+ l = sizeof(sa);
+ memset(&sa, 0, l);
+ if (getsockname(sd_fd, &sa, &l) < 0)
+ {
+ fprintf(stderr, "open_listen_sockets_systemd: problem getting fd %d: %s\n", sd_fd, rrd_strerror (errno));
+ return;
+ }
+
+ listen_fds[listen_fds_num].fd = sd_fd;
+ listen_fds[listen_fds_num].family = sa.sun_family;
+ listen_fds_num++;
+ }
+} /* }}} open_listen_sockets_systemd */
+
+static void open_listen_sockets_traditional(void) /* {{{ */
+{
+ if (config_listen_address_list_len > 0)
+ {
+ for (size_t i = 0; i < config_listen_address_list_len; i++)
+ open_listen_socket (config_listen_address_list[i]);
+
+ rrd_free_ptrs((void ***) &config_listen_address_list,
+ &config_listen_address_list_len);
+ }
+ else
+ {
+ strncpy(default_socket.addr, RRDCACHED_DEFAULT_ADDRESS,
+ sizeof(default_socket.addr) - 1);
+ default_socket.addr[sizeof(default_socket.addr) - 1] = '\0';
+
+ if (default_socket.permissions == 0)
+ socket_permission_set_all (&default_socket);
+
+ open_listen_socket (&default_socket);
+ }
+} /* }}} open_list_sockets_traditional */
+
static int close_listen_sockets (void) /* {{{ */
{
size_t i;
@@ -3120,26 +3178,13 @@ static int daemonize (void) /* {{{ */
if (pid_fd < 0)
return pid_fd;
- /* open all the listen sockets */
- if (config_listen_address_list_len > 0)
- {
- for (size_t i = 0; i < config_listen_address_list_len; i++)
- open_listen_socket (config_listen_address_list[i]);
+ /* gather sockets passed from systemd;
+ * if none, open all the listen sockets from config or default */
- rrd_free_ptrs((void ***) &config_listen_address_list,
- &config_listen_address_list_len);
- }
+ if (sd_listen_fds(0) > 0)
+ open_listen_sockets_systemd(sd_listen_fds(0));
else
- {
- strncpy(default_socket.addr, RRDCACHED_DEFAULT_ADDRESS,
- sizeof(default_socket.addr) - 1);
- default_socket.addr[sizeof(default_socket.addr) - 1] = '\0';
-
- if (default_socket.permissions == 0)
- socket_permission_set_all (&default_socket);
-
- open_listen_socket (&default_socket);
- }
+ open_listen_sockets_traditional();
if (listen_fds_num < 1)
{
diff --git a/src/sd-daemon.c b/src/sd-daemon.c
new file mode 100644
index 0000000..763e079
--- /dev/null
+++ b/src/sd-daemon.c
@@ -0,0 +1,530 @@
+/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
+
+/***
+ Copyright 2010 Lennart Poettering
+
+ Permission is hereby granted, free of charge, to any person
+ obtaining a copy of this software and associated documentation files
+ (the "Software"), to deal in the Software without restriction,
+ including without limitation the rights to use, copy, modify, merge,
+ publish, distribute, sublicense, and/or sell copies of the Software,
+ and to permit persons to whom the Software is furnished to do so,
+ subject to the following conditions:
+
+ The above copyright notice and this permission notice shall be
+ included in all copies or substantial portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ SOFTWARE.
+***/
+
+#ifndef _GNU_SOURCE
+#define _GNU_SOURCE
+#endif
+
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/socket.h>
+#include <sys/un.h>
+#ifdef __BIONIC__
+#include <linux/fcntl.h>
+#else
+#include <sys/fcntl.h>
+#endif
+#include <netinet/in.h>
+#include <stdlib.h>
+#include <errno.h>
+#include <unistd.h>
+#include <string.h>
+#include <stdarg.h>
+#include <stdio.h>
+#include <stddef.h>
+#include <limits.h>
+
+#if defined(__linux__)
+#include <mqueue.h>
+#endif
+
+#include "sd-daemon.h"
+
+#if (__GNUC__ >= 4)
+#ifdef SD_EXPORT_SYMBOLS
+/* Export symbols */
+#define _sd_export_ __attribute__ ((visibility("default")))
+#else
+/* Don't export the symbols */
+#define _sd_export_ __attribute__ ((visibility("hidden")))
+#endif
+#else
+#define _sd_export_
+#endif
+
+_sd_export_ int sd_listen_fds(int unset_environment) {
+
+#if defined(DISABLE_SYSTEMD) || !defined(__linux__)
+ return 0;
+#else
+ int r, fd;
+ const char *e;
+ char *p = NULL;
+ unsigned long l;
+
+ if (!(e = getenv("LISTEN_PID"))) {
+ r = 0;
+ goto finish;
+ }
+
+ errno = 0;
+ l = strtoul(e, &p, 10);
+
+ if (errno != 0) {
+ r = -errno;
+ goto finish;
+ }
+
+ if (!p || *p || l <= 0) {
+ r = -EINVAL;
+ goto finish;
+ }
+
+ /* Is this for us? */
+ if (getpid() != (pid_t) l) {
+ r = 0;
+ goto finish;
+ }
+
+ if (!(e = getenv("LISTEN_FDS"))) {
+ r = 0;
+ goto finish;
+ }
+
+ errno = 0;
+ l = strtoul(e, &p, 10);
+
+ if (errno != 0) {
+ r = -errno;
+ goto finish;
+ }
+
+ if (!p || *p) {
+ r = -EINVAL;
+ goto finish;
+ }
+
+ for (fd = SD_LISTEN_FDS_START; fd < SD_LISTEN_FDS_START + (int) l; fd ++) {
+ int flags;
+
+ if ((flags = fcntl(fd, F_GETFD)) < 0) {
+ r = -errno;
+ goto finish;
+ }
+
+ if (flags & FD_CLOEXEC)
+ continue;
+
+ if (fcntl(fd, F_SETFD, flags | FD_CLOEXEC) < 0) {
+ r = -errno;
+ goto finish;
+ }
+ }
+
+ r = (int) l;
+
+finish:
+ if (unset_environment) {
+ unsetenv("LISTEN_PID");
+ unsetenv("LISTEN_FDS");
+ }
+
+ return r;
+#endif
+}
+
+_sd_export_ int sd_is_fifo(int fd, const char *path) {
+ struct stat st_fd;
+
+ if (fd < 0)
+ return -EINVAL;
+
+ memset(&st_fd, 0, sizeof(st_fd));
+ if (fstat(fd, &st_fd) < 0)
+ return -errno;
+
+ if (!S_ISFIFO(st_fd.st_mode))
+ return 0;
+
+ if (path) {
+ struct stat st_path;
+
+ memset(&st_path, 0, sizeof(st_path));
+ if (stat(path, &st_path) < 0) {
+
+ if (errno == ENOENT || errno == ENOTDIR)
+ return 0;
+
+ return -errno;
+ }
+
+ return
+ st_path.st_dev == st_fd.st_dev &&
+ st_path.st_ino == st_fd.st_ino;
+ }
+
+ return 1;
+}
+
+_sd_export_ int sd_is_special(int fd, const char *path) {
+ struct stat st_fd;
+
+ if (fd < 0)
+ return -EINVAL;
+
+ if (fstat(fd, &st_fd) < 0)
+ return -errno;
+
+ if (!S_ISREG(st_fd.st_mode) && !S_ISCHR(st_fd.st_mode))
+ return 0;
+
+ if (path) {
+ struct stat st_path;
+
+ if (stat(path, &st_path) < 0) {
+
+ if (errno == ENOENT || errno == ENOTDIR)
+ return 0;
+
+ return -errno;
+ }
+
+ if (S_ISREG(st_fd.st_mode) && S_ISREG(st_path.st_mode))
+ return
+ st_path.st_dev == st_fd.st_dev &&
+ st_path.st_ino == st_fd.st_ino;
+ else if (S_ISCHR(st_fd.st_mode) && S_ISCHR(st_path.st_mode))
+ return st_path.st_rdev == st_fd.st_rdev;
+ else
+ return 0;
+ }
+
+ return 1;
+}
+
+static int sd_is_socket_internal(int fd, int type, int listening) {
+ struct stat st_fd;
+
+ if (fd < 0 || type < 0)
+ return -EINVAL;
+
+ if (fstat(fd, &st_fd) < 0)
+ return -errno;
+
+ if (!S_ISSOCK(st_fd.st_mode))
+ return 0;
+
+ if (type != 0) {
+ int other_type = 0;
+ socklen_t l = sizeof(other_type);
+
+ if (getsockopt(fd, SOL_SOCKET, SO_TYPE, &other_type, &l) < 0)
+ return -errno;
+
+ if (l != sizeof(other_type))
+ return -EINVAL;
+
+ if (other_type != type)
+ return 0;
+ }
+
+ if (listening >= 0) {
+ int accepting = 0;
+ socklen_t l = sizeof(accepting);
+
+ if (getsockopt(fd, SOL_SOCKET, SO_ACCEPTCONN, &accepting, &l) < 0)
+ return -errno;
+
+ if (l != sizeof(accepting))
+ return -EINVAL;
+
+ if (!accepting != !listening)
+ return 0;
+ }
+
+ return 1;
+}
+
+union sockaddr_union {
+ struct sockaddr sa;
+ struct sockaddr_in in4;
+ struct sockaddr_in6 in6;
+ struct sockaddr_un un;
+ struct sockaddr_storage storage;
+};
+
+_sd_export_ int sd_is_socket(int fd, int family, int type, int listening) {
+ int r;
+
+ if (family < 0)
+ return -EINVAL;
+
+ if ((r = sd_is_socket_internal(fd, type, listening)) <= 0)
+ return r;
+
+ if (family > 0) {
+ union sockaddr_union sockaddr;
+ socklen_t l;
+
+ memset(&sockaddr, 0, sizeof(sockaddr));
+ l = sizeof(sockaddr);
+
+ if (getsockname(fd, &sockaddr.sa, &l) < 0)
+ return -errno;
+
+ if (l < sizeof(sa_family_t))
+ return -EINVAL;
+
+ return sockaddr.sa.sa_family == family;
+ }
+
+ return 1;
+}
+
+_sd_export_ int sd_is_socket_inet(int fd, int family, int type, int listening, uint16_t port) {
+ union sockaddr_union sockaddr;
+ socklen_t l;
+ int r;
+
+ if (family != 0 && family != AF_INET && family != AF_INET6)
+ return -EINVAL;
+
+ if ((r = sd_is_socket_internal(fd, type, listening)) <= 0)
+ return r;
+
+ memset(&sockaddr, 0, sizeof(sockaddr));
+ l = sizeof(sockaddr);
+
+ if (getsockname(fd, &sockaddr.sa, &l) < 0)
+ return -errno;
+
+ if (l < sizeof(sa_family_t))
+ return -EINVAL;
+
+ if (sockaddr.sa.sa_family != AF_INET &&
+ sockaddr.sa.sa_family != AF_INET6)
+ return 0;
+
+ if (family > 0)
+ if (sockaddr.sa.sa_family != family)
+ return 0;
+
+ if (port > 0) {
+ if (sockaddr.sa.sa_family == AF_INET) {
+ if (l < sizeof(struct sockaddr_in))
+ return -EINVAL;
+
+ return htons(port) == sockaddr.in4.sin_port;
+ } else {
+ if (l < sizeof(struct sockaddr_in6))
+ return -EINVAL;
+
+ return htons(port) == sockaddr.in6.sin6_port;
+ }
+ }
+
+ return 1;
+}
+
+_sd_export_ int sd_is_socket_unix(int fd, int type, int listening, const char *path, size_t length) {
+ union sockaddr_union sockaddr;
+ socklen_t l;
+ int r;
+
+ if ((r = sd_is_socket_internal(fd, type, listening)) <= 0)
+ return r;
+
+ memset(&sockaddr, 0, sizeof(sockaddr));
+ l = sizeof(sockaddr);
+
+ if (getsockname(fd, &sockaddr.sa, &l) < 0)
+ return -errno;
+
+ if (l < sizeof(sa_family_t))
+ return -EINVAL;
+
+ if (sockaddr.sa.sa_family != AF_UNIX)
+ return 0;
+
+ if (path) {
+ if (length <= 0)
+ length = strlen(path);
+
+ if (length <= 0)
+ /* Unnamed socket */
+ return l == offsetof(struct sockaddr_un, sun_path);
+
+ if (path[0])
+ /* Normal path socket */
+ return
+ (l >= offsetof(struct sockaddr_un, sun_path) + length + 1) &&
+ memcmp(path, sockaddr.un.sun_path, length+1) == 0;
+ else
+ /* Abstract namespace socket */
+ return
+ (l == offsetof(struct sockaddr_un, sun_path) + length) &&
+ memcmp(path, sockaddr.un.sun_path, length) == 0;
+ }
+
+ return 1;
+}
+
+_sd_export_ int sd_is_mq(int fd, const char *path) {
+#if !defined(__linux__)
+ return 0;
+#else
+ struct mq_attr attr;
+
+ if (fd < 0)
+ return -EINVAL;
+
+ if (mq_getattr(fd, &attr) < 0)
+ return -errno;
+
+ if (path) {
+ char fpath[PATH_MAX];
+ struct stat a, b;
+
+ if (path[0] != '/')
+ return -EINVAL;
+
+ if (fstat(fd, &a) < 0)
+ return -errno;
+
+ strncpy(stpcpy(fpath, "/dev/mqueue"), path, sizeof(fpath) - 12);
+ fpath[sizeof(fpath)-1] = 0;
+
+ if (stat(fpath, &b) < 0)
+ return -errno;
+
+ if (a.st_dev != b.st_dev ||
+ a.st_ino != b.st_ino)
+ return 0;
+ }
+
+ return 1;
+#endif
+}
+
+_sd_export_ int sd_notify(int unset_environment, const char *state) {
+#if defined(DISABLE_SYSTEMD) || !defined(__linux__) || !defined(SOCK_CLOEXEC)
+ return 0;
+#else
+ int fd = -1, r;
+ struct msghdr msghdr;
+ struct iovec iovec;
+ union sockaddr_union sockaddr;
+ const char *e;
+
+ if (!state) {
+ r = -EINVAL;
+ goto finish;
+ }
+
+ if (!(e = getenv("NOTIFY_SOCKET")))
+ return 0;
+
+ /* Must be an abstract socket, or an absolute path */
+ if ((e[0] != '@' && e[0] != '/') || e[1] == 0) {
+ r = -EINVAL;
+ goto finish;
+ }
+
+ if ((fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0)) < 0) {
+ r = -errno;
+ goto finish;
+ }
+
+ memset(&sockaddr, 0, sizeof(sockaddr));
+ sockaddr.sa.sa_family = AF_UNIX;
+ strncpy(sockaddr.un.sun_path, e, sizeof(sockaddr.un.sun_path));
+
+ if (sockaddr.un.sun_path[0] == '@')
+ sockaddr.un.sun_path[0] = 0;
+
+ memset(&iovec, 0, sizeof(iovec));
+ iovec.iov_base = (char*) state;
+ iovec.iov_len = strlen(state);
+
+ memset(&msghdr, 0, sizeof(msghdr));
+ msghdr.msg_name = &sockaddr;
+ msghdr.msg_namelen = offsetof(struct sockaddr_un, sun_path) + strlen(e);
+
+ if (msghdr.msg_namelen > sizeof(struct sockaddr_un))
+ msghdr.msg_namelen = sizeof(struct sockaddr_un);
+
+ msghdr.msg_iov = &iovec;
+ msghdr.msg_iovlen = 1;
+
+ if (sendmsg(fd, &msghdr, MSG_NOSIGNAL) < 0) {
+ r = -errno;
+ goto finish;
+ }
+
+ r = 1;
+
+finish:
+ if (unset_environment)
+ unsetenv("NOTIFY_SOCKET");
+
+ if (fd >= 0)
+ close(fd);
+
+ return r;
+#endif
+}
+
+_sd_export_ int sd_notifyf(int unset_environment, const char *format, ...) {
+#if defined(DISABLE_SYSTEMD) || !defined(__linux__)
+ return 0;
+#else
+ va_list ap;
+ char *p = NULL;
+ int r;
+
+ va_start(ap, format);
+ r = vasprintf(&p, format, ap);
+ va_end(ap);
+
+ if (r < 0 || !p)
+ return -ENOMEM;
+
+ r = sd_notify(unset_environment, p);
+ free(p);
+
+ return r;
+#endif
+}
+
+_sd_export_ int sd_booted(void) {
+#if defined(DISABLE_SYSTEMD) || !defined(__linux__)
+ return 0;
+#else
+
+ struct stat a, b;
+
+ /* We simply test whether the systemd cgroup hierarchy is
+ * mounted */
+
+ if (lstat("/sys/fs/cgroup", &a) < 0)
+ return 0;
+
+ if (lstat("/sys/fs/cgroup/systemd", &b) < 0)
+ return 0;
+
+ return a.st_dev != b.st_dev;
+#endif
+}
diff --git a/src/sd-daemon.h b/src/sd-daemon.h
new file mode 100644
index 0000000..fe51159
--- /dev/null
+++ b/src/sd-daemon.h
@@ -0,0 +1,282 @@
+/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
+
+#ifndef foosddaemonhfoo
+#define foosddaemonhfoo
+
+/***
+ Copyright 2010 Lennart Poettering
+
+ Permission is hereby granted, free of charge, to any person
+ obtaining a copy of this software and associated documentation files
+ (the "Software"), to deal in the Software without restriction,
+ including without limitation the rights to use, copy, modify, merge,
+ publish, distribute, sublicense, and/or sell copies of the Software,
+ and to permit persons to whom the Software is furnished to do so,
+ subject to the following conditions:
+
+ The above copyright notice and this permission notice shall be
+ included in all copies or substantial portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ SOFTWARE.
+***/
+
+#include <sys/types.h>
+#include <inttypes.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ Reference implementation of a few systemd related interfaces for
+ writing daemons. These interfaces are trivial to implement. To
+ simplify porting we provide this reference implementation.
+ Applications are welcome to reimplement the algorithms described
+ here if they do not want to include these two source files.
+
+ The following functionality is provided:
+
+ - Support for logging with log levels on stderr
+ - File descriptor passing for socket-based activation
+ - Daemon startup and status notification
+ - Detection of systemd boots
+
+ You may compile this with -DDISABLE_SYSTEMD to disable systemd
+ support. This makes all those calls NOPs that are directly related to
+ systemd (i.e. only sd_is_xxx() will stay useful).
+
+ Since this is drop-in code we don't want any of our symbols to be
+ exported in any case. Hence we declare hidden visibility for all of
+ them.
+
+ You may find an up-to-date version of these source files online:
+
+ http://cgit.freedesktop.org/systemd/systemd/plain/src/systemd/sd-daemon.h
+ http://cgit.freedesktop.org/systemd/systemd/plain/src/sd-daemon.c
+
+ This should compile on non-Linux systems, too, but with the
+ exception of the sd_is_xxx() calls all functions will become NOPs.
+
+ See sd-daemon(7) for more information.
+*/
+
+#ifndef _sd_printf_attr_
+#if __GNUC__ >= 4
+#define _sd_printf_attr_(a,b) __attribute__ ((format (printf, a, b)))
+#else
+#define _sd_printf_attr_(a,b)
+#endif
+#endif
+
+/*
+ Log levels for usage on stderr:
+
+ fprintf(stderr, SD_NOTICE "Hello World!\n");
+
+ This is similar to printk() usage in the kernel.
+*/
+#define SD_EMERG "<0>" /* system is unusable */
+#define SD_ALERT "<1>" /* action must be taken immediately */
+#define SD_CRIT "<2>" /* critical conditions */
+#define SD_ERR "<3>" /* error conditions */
+#define SD_WARNING "<4>" /* warning conditions */
+#define SD_NOTICE "<5>" /* normal but significant condition */
+#define SD_INFO "<6>" /* informational */
+#define SD_DEBUG "<7>" /* debug-level messages */
+
+/* The first passed file descriptor is fd 3 */
+#define SD_LISTEN_FDS_START 3
+
+/*
+ Returns how many file descriptors have been passed, or a negative
+ errno code on failure. Optionally, removes the $LISTEN_FDS and
+ $LISTEN_PID file descriptors from the environment (recommended, but
+ problematic in threaded environments). If r is the return value of
+ this function you'll find the file descriptors passed as fds
+ SD_LISTEN_FDS_START to SD_LISTEN_FDS_START+r-1. Returns a negative
+ errno style error code on failure. This function call ensures that
+ the FD_CLOEXEC flag is set for the passed file descriptors, to make
+ sure they are not passed on to child processes. If FD_CLOEXEC shall
+ not be set, the caller needs to unset it after this call for all file
+ descriptors that are used.
+
+ See sd_listen_fds(3) for more information.
+*/
+int sd_listen_fds(int unset_environment);
+
+/*
+ Helper call for identifying a passed file descriptor. Returns 1 if
+ the file descriptor is a FIFO in the file system stored under the
+ specified path, 0 otherwise. If path is NULL a path name check will
+ not be done and the call only verifies if the file descriptor
+ refers to a FIFO. Returns a negative errno style error code on
+ failure.
+
+ See sd_is_fifo(3) for more information.
+*/
+int sd_is_fifo(int fd, const char *path);
+
+/*
+ Helper call for identifying a passed file descriptor. Returns 1 if
+ the file descriptor is a special character device on the file
+ system stored under the specified path, 0 otherwise.
+ If path is NULL a path name check will not be done and the call
+ only verifies if the file descriptor refers to a special character.
+ Returns a negative errno style error code on failure.
+
+ See sd_is_special(3) for more information.
+*/
+int sd_is_special(int fd, const char *path);
+
+/*
+ Helper call for identifying a passed file descriptor. Returns 1 if
+ the file descriptor is a socket of the specified family (AF_INET,
+ ...) and type (SOCK_DGRAM, SOCK_STREAM, ...), 0 otherwise. If
+ family is 0 a socket family check will not be done. If type is 0 a
+ socket type check will not be done and the call only verifies if
+ the file descriptor refers to a socket. If listening is > 0 it is
+ verified that the socket is in listening mode. (i.e. listen() has
+ been called) If listening is == 0 it is verified that the socket is
+ not in listening mode. If listening is < 0 no listening mode check
+ is done. Returns a negative errno style error code on failure.
+
+ See sd_is_socket(3) for more information.
+*/
+int sd_is_socket(int fd, int family, int type, int listening);
+
+/*
+ Helper call for identifying a passed file descriptor. Returns 1 if
+ the file descriptor is an Internet socket, of the specified family
+ (either AF_INET or AF_INET6) and the specified type (SOCK_DGRAM,
+ SOCK_STREAM, ...), 0 otherwise. If version is 0 a protocol version
+ check is not done. If type is 0 a socket type check will not be
+ done. If port is 0 a socket port check will not be done. The
+ listening flag is used the same way as in sd_is_socket(). Returns a
+ negative errno style error code on failure.
+
+ See sd_is_socket_inet(3) for more information.
+*/
+int sd_is_socket_inet(int fd, int family, int type, int listening, uint16_t port);
+
+/*
+ Helper call for identifying a passed file descriptor. Returns 1 if
+ the file descriptor is an AF_UNIX socket of the specified type
+ (SOCK_DGRAM, SOCK_STREAM, ...) and path, 0 otherwise. If type is 0
+ a socket type check will not be done. If path is NULL a socket path
+ check will not be done. For normal AF_UNIX sockets set length to
+ 0. For abstract namespace sockets set length to the length of the
+ socket name (including the initial 0 byte), and pass the full
+ socket path in path (including the initial 0 byte). The listening
+ flag is used the same way as in sd_is_socket(). Returns a negative
+ errno style error code on failure.
+
+ See sd_is_socket_unix(3) for more information.
+*/
+int sd_is_socket_unix(int fd, int type, int listening, const char *path, size_t length);
+
+/*
+ Helper call for identifying a passed file descriptor. Returns 1 if
+ the file descriptor is a POSIX Message Queue of the specified name,
+ 0 otherwise. If path is NULL a message queue name check is not
+ done. Returns a negative errno style error code on failure.
+*/
+int sd_is_mq(int fd, const char *path);
+
+/*
+ Informs systemd about changed daemon state. This takes a number of
+ newline separated environment-style variable assignments in a
+ string. The following variables are known:
+
+ READY=1 Tells systemd that daemon startup is finished (only
+ relevant for services of Type=notify). The passed
+ argument is a boolean "1" or "0". Since there is
+ little value in signaling non-readiness the only
+ value daemons should send is "READY=1".
+
+ STATUS=... Passes a single-line status string back to systemd
+ that describes the daemon state. This is free-from
+ 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..."
+
+ ERRNO=... If a daemon fails, the errno-style error code,
+ formatted as string. Example: "ERRNO=2" for ENOENT.
+
+ BUSERROR=... If a daemon fails, the D-Bus error-style error
+ code. Example: "BUSERROR=org.freedesktop.DBus.Error.TimedOut"
+
+ MAINPID=... The main pid of a daemon, in case systemd did not
+ fork off the process itself. Example: "MAINPID=4711"
+
+ WATCHDOG=1 Tells systemd to update the watchdog timestamp.
+ Services using this feature should do this in
+ regular intervals. A watchdog framework can use the
+ timestamps to detect failed services.
+
+ Daemons can choose to send additional variables. However, it is
+ recommended to prefix variable names not listed above with X_.
+
+ Returns a negative errno-style error code on failure. Returns > 0
+ if systemd could be notified, 0 if it couldn't possibly because
+ systemd is not running.
+
+ Example: When a daemon finished starting up, it could issue this
+ call to notify systemd about it:
+
+ sd_notify(0, "READY=1");
+
+ See sd_notifyf() for more complete examples.
+
+ See sd_notify(3) for more information.
+*/
+int sd_notify(int unset_environment, const char *state);
+
+/*
+ Similar to sd_notify() but takes a format string.
+
+ Example 1: A daemon could send the following after initialization:
+
+ sd_notifyf(0, "READY=1\n"
+ "STATUS=Processing requests...\n"
+ "MAINPID=%lu",
+ (unsigned long) getpid());
+
+ Example 2: A daemon could send the following shortly before
+ exiting, on failure:
+
+ sd_notifyf(0, "STATUS=Failed to start up: %s\n"
+ "ERRNO=%i",
+ strerror(errno),
+ errno);
+
+ See sd_notifyf(3) for more information.
+*/
+int sd_notifyf(int unset_environment, const char *format, ...) _sd_printf_attr_(2,3);
+
+/*
+ Returns > 0 if the system was booted with systemd. Returns < 0 on
+ error. Returns 0 if the system was not booted with systemd. Note
+ that all of the functions above handle non-systemd boots just
+ fine. You should NOT protect them with a call to this function. Also
+ note that this function checks whether the system, not the user
+ session is controlled by systemd. However the functions above work
+ for both user and system services.
+
+ See sd_booted(3) for more information.
+*/
+int sd_booted(void);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
--
1.7.10
-------------- next part --------------
From 7071fda883d4c2d55310f593ad876562a96f8089 Mon Sep 17 00:00:00 2001
From: Tomasz Torcz <tomek at pipebreaker.pl>
Date: Fri, 13 Apr 2012 23:16:20 +0200
Subject: [PATCH 2/2] add some sd_notify()s
---
src/rrd_daemon.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/src/rrd_daemon.c b/src/rrd_daemon.c
index f1e79e1..f1e6cec 100644
--- a/src/rrd_daemon.c
+++ b/src/rrd_daemon.c
@@ -112,6 +112,7 @@
#define RRDD_LOG(severity, ...) \
do { \
+ sd_notifyf(0, "STATUS=" __VA_ARGS__); \
if (stay_foreground) { \
fprintf(stderr, __VA_ARGS__); \
fprintf(stderr, "\n"); } \
@@ -3075,6 +3076,7 @@ static void *listen_thread_main (void UNUSED(*args)) /* {{{ */
memset (pollfds, 0, sizeof (*pollfds) * pollfds_num);
RRDD_LOG(LOG_INFO, "listening for connections");
+ sd_notify(0, "READY=1\n");
while (state == RUNNING)
{
@@ -3217,6 +3219,7 @@ static int daemonize (void) /* {{{ */
if (dup(0) == -1 || dup(0) == -1){
RRDD_LOG (LOG_ERR, "faild to run dup.\n");
}
+ sd_notifyf(0, "MAINPID=%lu\n", (unsigned long)getpid());
} /* if (!stay_foreground) */
/* Change into the /tmp directory. */
--
1.7.10
-------------- next part --------------
#
# To customize, copy this file to /etc/systemd/system/
# and edit there
#
[Unit]
Description=sockets to activate rrdcached
[Socket]
ListenStream=/tmp/rrdcached.sock
#ListenStream=[::1]:42217
#DirectoryMode=0755
#SocketMode=0666
#BindToDevice=
[Install]
WantedBy=sockets.target
-------------- next part --------------
#
# To customize, copy this file to /etc/systemd/system/
# and edit there
#
[Unit]
Description=Data caching daemon for rrdtool
[Service]
# If you enable socket-activable rrdcached.socket,
# command line socket declarations will be ignored
ExecStart=/usr/bin/rrdcached -g
#WorkingDirectory=
#User=
#Group=
[Install]
WantedBy=multi-user.target
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 270 bytes
Desc: not available
Url : http://lists.oetiker.ch/pipermail/rrd-developers/attachments/20120415/b5e5e9c1/attachment-0001.pgp
More information about the rrd-developers
mailing list