openrc/src/supervise-daemon/supervise-daemon.c

1263 lines
32 KiB
C
Raw Normal View History

/*
* supervise-daemon
* This is a supervisor for daemons.
* It will start a daemon and make sure it restarts if it crashes.
*/
/*
* Copyright (c) 2016 The OpenRC Authors.
* See the Authors file at the top-level directory of this distribution and
* https://github.com/OpenRC/openrc/blob/HEAD/AUTHORS
*
* This file is part of OpenRC. It is subject to the license terms in
* the LICENSE file found in the top-level directory of this
* distribution and at https://github.com/OpenRC/openrc/blob/HEAD/LICENSE
* This file may not be copied, modified, propagated, or distributed
* except according to the terms contained in the LICENSE file.
*/
/* nano seconds */
#define POLL_INTERVAL 20000000
#define WAIT_PIDFILE 500000000
#define ONE_SECOND 1000000000
#define ONE_MS 1000000
#ifdef HAVE_CLOSE_RANGE_CLOEXEC
/* For close_range() */
# define _GNU_SOURCE
#endif
#include <errno.h>
#include <fcntl.h>
#include <getopt.h>
#include <limits.h>
#include <grp.h>
#include <pwd.h>
#include <signal.h>
#include <stddef.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <stdbool.h>
#include <strings.h>
#ifdef __linux__
# include <sys/syscall.h> /* For io priority */
# include <sys/prctl.h> /* For prctl */
#endif
#include <syslog.h>
#include <sys/ioctl.h>
#include <sys/resource.h>
#include <sys/stat.h>
#include <sys/wait.h>
#include <time.h>
#include <unistd.h>
#ifdef HAVE_PAM
#include <security/pam_appl.h>
/* We are not supporting authentication conversations */
static struct pam_conv conv = { NULL, NULL};
#endif
#ifdef HAVE_CAP
#include <sys/capability.h>
#endif
#include "einfo.h"
#include "queue.h"
#include "rc.h"
#include "misc.h"
#include "plugin.h"
#include "schedules.h"
#include "_usage.h"
#include "helpers.h"
/* Use long option value that is out of range for 8 bit getopt values.
* The exact enum value is internal and can freely change, so we keep the
* options sorted.
*/
enum {
/* This has to come first so following values stay in the 0x100+ range. */
LONGOPT_BASE = 0x100,
LONGOPT_CAPABILITIES,
LONGOPT_OOM_SCORE_ADJ,
LONGOPT_NO_NEW_PRIVS,
LONGOPT_SECBITS,
};
const char *applet = NULL;
const char *extraopts = NULL;
const char getoptstring[] = "A:a:D:d:e:g:H:I:Kk:m:N:p:R:r:s:Su:1:2:3" \
getoptstring_COMMON;
const struct option longopts[] = {
{ "healthcheck-timer", 1, NULL, 'a'},
{ "healthcheck-delay", 1, NULL, 'A'},
{ "capabilities", 1, NULL, LONGOPT_CAPABILITIES},
{ "secbits", 1, NULL, LONGOPT_SECBITS},
{ "no-new-privs", 0, NULL, LONGOPT_NO_NEW_PRIVS},
{ "respawn-delay", 1, NULL, 'D'},
{ "chdir", 1, NULL, 'd'},
{ "env", 1, NULL, 'e'},
{ "group", 1, NULL, 'g'},
{ "ionice", 1, NULL, 'I'},
{ "stop", 0, NULL, 'K'},
{ "umask", 1, NULL, 'k'},
{ "respawn-max", 1, NULL, 'm'},
{ "nicelevel", 1, NULL, 'N'},
{ "oom-score-adj",1, NULL, LONGOPT_OOM_SCORE_ADJ},
{ "pidfile", 1, NULL, 'p'},
{ "respawn-period", 1, NULL, 'P'},
{ "retry", 1, NULL, 'R'},
{ "chroot", 1, NULL, 'r'},
{ "signal", 1, NULL, 's'},
{ "start", 0, NULL, 'S'},
{ "user", 1, NULL, 'u'},
{ "stdout", 1, NULL, '1'},
{ "stderr", 1, NULL, '2'},
{ "reexec", 0, NULL, '3'},
longopts_COMMON
};
const char * const longopts_help[] = {
"set an initial health check delay",
"set a health check timer",
"Set the inheritable, ambient and bounding capabilities",
"Set the security-bits for the program",
"Set the No New Privs flag for the program",
"Set a respawn delay",
"Change the PWD",
"Set an environment string",
"Change the process group",
"Set an ionice class:data when starting",
"Stop daemon",
"Set the umask for the daemon",
"set maximum number of respawn attempts",
"Set a nicelevel when starting",
"Set OOM score adjustment when starting",
"Match pid found in this file",
"Set respawn time period",
"Retry schedule to use when stopping",
"Chroot to this directory",
"Send a signal to the daemon",
"Start daemon",
"Change the process user",
"Redirect stdout to file",
"Redirect stderr to file",
"reexec (used internally)",
longopts_help_COMMON
};
const char *usagestring = NULL;
static int healthcheckdelay = 0;
static int healthchecktimer = 0;
static volatile sig_atomic_t do_healthcheck = 0;
static volatile sig_atomic_t exiting = 0;
static int nicelevel = INT_MIN;
static int ionicec = -1;
static int ioniced = 0;
static int oom_score_adj = INT_MIN;
static char *changeuser, *ch_root, *ch_dir;
static uid_t uid = 0;
static gid_t gid = 0;
static int devnull_fd = -1;
static int stdin_fd;
static int stdout_fd;
static int stderr_fd;
static char *redirect_stderr = NULL;
static char *redirect_stdout = NULL;
#ifdef TIOCNOTTY
static int tty_fd = -1;
#endif
static pid_t child_pid;
static int respawn_count = 0;
static int respawn_delay = 0;
static int respawn_max = 10;
static int respawn_period = 0;
static char *fifopath = NULL;
static int fifo_fd = 0;
static char *pidfile = NULL;
static char *svcname = NULL;
static bool verbose = false;
#ifdef HAVE_CAP
static cap_iab_t cap_iab = NULL;
static unsigned secbits = 0;
#endif
#ifdef PR_SET_NO_NEW_PRIVS
static bool no_new_privs = false;
#endif
extern char **environ;
#if !defined(SYS_ioprio_set) && defined(__NR_ioprio_set)
# define SYS_ioprio_set __NR_ioprio_set
#endif
#if !defined(__DragonFly__)
static inline int ioprio_set(int which RC_UNUSED, int who RC_UNUSED,
int ioprio RC_UNUSED)
{
#ifdef SYS_ioprio_set
return syscall(SYS_ioprio_set, which, who, ioprio);
#else
return 0;
#endif
}
#endif
static void cleanup(void)
{
free(changeuser);
}
RC_NORETURN static void re_exec_supervisor(void)
{
syslog(LOG_WARNING, "Re-executing for %s", svcname);
2017-10-30 18:21:39 -05:00
execlp("supervise-daemon", "supervise-daemon", svcname, "--reexec",
(char *) NULL);
syslog(LOG_ERR, "Unable to execute supervise-daemon: %s",
strerror(errno));
exit(EXIT_FAILURE);
}
static void handle_signal(int sig)
{
int serrno = errno;
pid_t pid;
switch (sig) {
case SIGALRM:
do_healthcheck = 1;
break;
case SIGCHLD:
if (exiting)
while (waitpid((pid_t)(-1), NULL, WNOHANG) > 0) {}
else {
while ((pid = waitpid((pid_t)(-1), NULL, WNOHANG|WNOWAIT)) > 0) {
if (pid == child_pid)
break;
pid = waitpid(pid, NULL, WNOHANG);
}
}
break;
case SIGTERM:
exiting = 1;
break;
default:
syslog(LOG_WARNING, "caught signal %d", sig);
2017-11-27 11:09:43 -06:00
re_exec_supervisor();
}
/* Restore errno */
errno = serrno;
}
static char * expand_home(const char *home, const char *path)
{
char *opath, *ppath, *p, *nh;
struct passwd *pw;
if (!path || *path != '~')
return xstrdup(path);
opath = ppath = xstrdup(path);
if (ppath[1] != '/' && ppath[1] != '\0') {
p = strchr(ppath + 1, '/');
if (p)
*p = '\0';
pw = getpwnam(ppath + 1);
if (pw) {
home = pw->pw_dir;
ppath = p;
if (ppath)
*ppath = '/';
} else
home = NULL;
} else
ppath++;
if (!home) {
free(opath);
return xstrdup(path);
}
if (!ppath) {
free(opath);
return xstrdup(home);
}
xasprintf(&nh, "%s%s", home, ppath);
free(opath);
return nh;
}
static char *make_cmdline(char **argv)
{
char **c;
char *cmdline = NULL;
size_t len = 0;
for (c = argv; c && *c; c++)
len += (strlen(*c) + 1);
2018-03-14 20:05:45 -05:00
cmdline = xmalloc(len+1);
memset(cmdline, 0, len+1);
for (c = argv; c && *c; c++) {
strcat(cmdline, *c);
strcat(cmdline, " ");
}
return cmdline;
}
2018-12-21 12:06:15 -06:00
static pid_t exec_command(const char *cmd)
{
char *file;
pid_t pid = -1;
sigset_t full;
sigset_t old;
struct sigaction sa;
file = rc_service_resolve(svcname);
if (!exists(file)) {
free(file);
return 0;
}
/* We need to block signals until we have forked */
memset(&sa, 0, sizeof (sa));
sa.sa_handler = SIG_DFL;
sigemptyset(&sa.sa_mask);
sigfillset(&full);
sigprocmask(SIG_SETMASK, &full, &old);
pid = fork();
if (pid == 0) {
/* Restore default handlers */
sigaction(SIGCHLD, &sa, NULL);
sigaction(SIGHUP, &sa, NULL);
sigaction(SIGINT, &sa, NULL);
sigaction(SIGQUIT, &sa, NULL);
sigaction(SIGTERM, &sa, NULL);
sigaction(SIGUSR1, &sa, NULL);
sigaction(SIGWINCH, &sa, NULL);
/* Unmask signals */
sigprocmask(SIG_SETMASK, &old, NULL);
/* Safe to run now */
execl(file, file, cmd, (char *) NULL);
syslog(LOG_ERR, "unable to exec `%s': %s\n",
file, strerror(errno));
_exit(EXIT_FAILURE);
}
if (pid == -1)
syslog(LOG_ERR, "fork: %s\n",strerror (errno));
sigprocmask(SIG_SETMASK, &old, NULL);
free(file);
return pid;
}
RC_NORETURN static void child_process(char *exec, char **argv)
{
RC_STRINGLIST *env_list;
RC_STRING *env;
int i;
char *p;
char *token;
size_t len;
char *newpath;
char *np;
char *cmdline = NULL;
time_t start_time;
char start_count_string[20];
2017-05-11 16:18:41 -05:00
char start_time_string[20];
FILE *fp;
#ifdef HAVE_PAM
pam_handle_t *pamh = NULL;
int pamr;
const char *const *pamenv = NULL;
#endif
setsid();
if (svcname) {
start_time = time(NULL);
from_time_t(start_time_string, start_time);
rc_service_value_set(svcname, "start_time", start_time_string);
sprintf(start_count_string, "%i", respawn_count);
rc_service_value_set(svcname, "start_count", start_count_string);
sprintf(start_count_string, "%d", getpid());
rc_service_value_set(svcname, "child_pid", start_count_string);
}
if (nicelevel != INT_MIN) {
if (setpriority(PRIO_PROCESS, getpid(), nicelevel) == -1)
eerrorx("%s: setpriority %d: %s", applet, nicelevel,
strerror(errno));
start-stop-daemon, supervise-daemon: set autogroup nicelevel As described in "Why nice levels are a placebo and have been for a very long time, and no one seems to have noticed"[1], the Linux kernel in its default configuration on many Linux distributions autogroups tasks by session ID and "fairly" allocates CPU time among such autogroups. The nice levels of tasks within each autogroup are only relative to other tasks within the same autogroup. Effectively, this means that the traditional nice level is rendered moot for tools like start-stop-daemon and supervise-daemon, which start each daemon in its own session and thus in its own autogroup. Linux does provide a means to change the niceness of autogroups relative to each other, so let's have start-stop- daemon and supervise-daemon make use of this feature where available so that -N,--nicelevel/SSD_NICELEVEL will actually do what the user intends. On systems where autogroups are not supported or are disabled, this commit introduces no change in behavior. Note that the setsid() call in the child process of start-stop-daemon is moved to much earlier. This is necessary so that the new process will be assigned to a new autogroup before the autogroup nicelevel is set. To avoid inadvertently acquiring /dev/tty as the controlling terminal of the new session after setsid() has given up the controlling terminal inherited from the parent process, tty_fd is opened before the call to setsid(). [1] https://www.reddit.com/r/linux/comments/d7hx2c/why_nice_levels_are_a_placebo_and_have_been_for_a/ This fixes #542.
2022-08-21 09:55:04 -04:00
/* Open in "r+" mode to avoid creating if non-existent. */
fp = fopen("/proc/self/autogroup", "r+");
if (fp) {
fprintf(fp, "%d\n", nicelevel);
fclose(fp);
} else if (errno != ENOENT)
eerrorx("%s: autogroup nice %d: %s", applet,
nicelevel, strerror(errno));
}
if (ionicec != -1 && ioprio_set(1, getpid(), ionicec | ioniced) == -1)
eerrorx("%s: ioprio_set %d %d: %s", applet, ionicec, ioniced,
strerror(errno));
if (oom_score_adj != INT_MIN) {
fp = fopen("/proc/self/oom_score_adj", "w");
if (!fp)
eerrorx("%s: oom_score_adj %d: %s", applet,
oom_score_adj, strerror(errno));
fprintf(fp, "%d\n", oom_score_adj);
fclose(fp);
}
if (ch_root && chroot(ch_root) < 0)
eerrorx("%s: chroot `%s': %s", applet, ch_root, strerror(errno));
if (ch_dir && chdir(ch_dir) < 0)
eerrorx("%s: chdir `%s': %s", applet, ch_dir, strerror(errno));
#ifdef HAVE_PAM
if (changeuser != NULL) {
2016-05-13 12:05:23 -05:00
pamr = pam_start("supervise-daemon",
changeuser, &conv, &pamh);
if (pamr == PAM_SUCCESS)
pamr = pam_acct_mgmt(pamh, PAM_SILENT);
if (pamr == PAM_SUCCESS)
pamr = pam_open_session(pamh, PAM_SILENT);
if (pamr != PAM_SUCCESS)
eerrorx("%s: pam error: %s", applet, pam_strerror(pamh, pamr));
}
#endif
if (gid && setgid(gid))
eerrorx("%s: unable to set groupid to %d", applet, gid);
if (changeuser && initgroups(changeuser, gid))
eerrorx("%s: initgroups (%s, %d)", applet, changeuser, gid);
#ifdef HAVE_CAP
if (uid && cap_setuid(uid))
#else
if (uid && setuid(uid))
#endif
eerrorx ("%s: unable to set userid to %d", applet, uid);
/* Close any fd's to the passwd database */
endpwent();
#ifdef HAVE_CAP
if (cap_iab != NULL) {
i = cap_iab_set_proc(cap_iab);
if (cap_free(cap_iab) != 0)
eerrorx("Could not releasable memory: %s", strerror(errno));
if (i != 0)
eerrorx("Could not set iab: %s", strerror(errno));
}
if (secbits != 0) {
if (cap_set_secbits(secbits) < 0)
eerrorx("Could not set securebits to 0x%x: %s", secbits, strerror(errno));
}
#endif
#ifdef PR_SET_NO_NEW_PRIVS
if (no_new_privs) {
if (prctl(PR_SET_NO_NEW_PRIVS, 1, 0, 0, 0) == -1)
eerrorx("Could not set No New Privs flag: %s", strerror(errno));
}
#endif
/* remove the controlling tty */
#ifdef TIOCNOTTY
ioctl(tty_fd, TIOCNOTTY, 0);
close(tty_fd);
#endif
/* Clean the environment of any RC_ variables */
env_list = rc_stringlist_new();
i = 0;
while (environ[i])
rc_stringlist_add(env_list, environ[i++]);
#ifdef HAVE_PAM
if (changeuser != NULL) {
pamenv = (const char *const *)pam_getenvlist(pamh);
if (pamenv) {
while (*pamenv) {
/* Don't add strings unless they set a var */
if (strchr(*pamenv, '='))
putenv(xstrdup(*pamenv));
else
unsetenv(*pamenv);
pamenv++;
}
}
}
#endif
TAILQ_FOREACH(env, env_list, entries) {
if ((strncmp(env->value, "RC_", 3) == 0 &&
strncmp(env->value, "RC_SERVICE=", 11) != 0 &&
strncmp(env->value, "RC_SVCNAME=", 11) != 0) ||
strncmp(env->value, "SSD_NICELEVEL=", 14) == 0 ||
strncmp(env->value, "SSD_IONICELEVEL=", 16) == 0 ||
strncmp(env->value, "SSD_OOM_SCORE_ADJ=", 18) == 0)
{
p = strchr(env->value, '=');
*p = '\0';
unsetenv(env->value);
continue;
}
}
rc_stringlist_free(env_list);
/* For the path, remove the rcscript bin dir from it */
if ((token = getenv("PATH"))) {
len = strlen(token);
newpath = np = xmalloc(len + 1);
while (token && *token) {
p = strchr(token, ':');
if (p) {
*p++ = '\0';
while (*p == ':')
p++;
}
if (strcmp(token, RC_LIBEXECDIR "/bin") != 0 &&
strcmp(token, RC_LIBEXECDIR "/sbin") != 0)
{
len = strlen(token);
if (np != newpath)
*np++ = ':';
memcpy(np, token, len);
np += len;
}
token = p;
}
*np = '\0';
unsetenv("PATH");
setenv("PATH", newpath, 1);
}
stdin_fd = devnull_fd;
stdout_fd = devnull_fd;
stderr_fd = devnull_fd;
if (redirect_stdout) {
if ((stdout_fd = open(redirect_stdout,
O_WRONLY | O_CREAT | O_APPEND,
S_IRUSR | S_IWUSR)) == -1)
eerrorx("%s: unable to open the logfile"
" for stdout `%s': %s",
applet, redirect_stdout, strerror(errno));
}
if (redirect_stderr) {
if ((stderr_fd = open(redirect_stderr,
O_WRONLY | O_CREAT | O_APPEND,
S_IRUSR | S_IWUSR)) == -1)
eerrorx("%s: unable to open the logfile"
" for stderr `%s': %s",
applet, redirect_stderr, strerror(errno));
}
dup2(stdin_fd, STDIN_FILENO);
if (redirect_stdout || rc_yesno(getenv("EINFO_QUIET")))
dup2(stdout_fd, STDOUT_FILENO);
if (redirect_stderr || rc_yesno(getenv("EINFO_QUIET")))
dup2(stderr_fd, STDERR_FILENO);
#ifdef HAVE_CLOSE_RANGE_CLOEXEC
if (close_range(3, UINT_MAX, CLOSE_RANGE_CLOEXEC) < 0)
#endif
for (i = getdtablesize() - 1; i >= 3; --i)
fcntl(i, F_SETFD, FD_CLOEXEC);
cmdline = make_cmdline(argv);
syslog(LOG_INFO, "Child command line: %s", cmdline);
free(cmdline);
execvp(exec, argv);
#ifdef HAVE_PAM
if (changeuser != NULL && pamr == PAM_SUCCESS)
pam_close_session(pamh, PAM_SILENT);
#endif
eerrorx("%s: failed to exec `%s': %s", applet, exec,strerror(errno));
}
RC_NORETURN static void supervisor(char *exec, char **argv)
{
FILE *fp;
char buf[2048];
char cmd[2048];
int count;
int failing;
int health_status;
int healthcheck_respawn;
int i;
int nkilled;
int sig_send;
pid_t health_pid;
pid_t wait_pid;
sigset_t old_signals;
sigset_t signals;
struct sigaction sa;
struct timespec ts;
time_t respawn_now= 0;
time_t first_spawn= 0;
/* block all signals we do not handle */
sigfillset(&signals);
sigdelset(&signals, SIGALRM);
sigdelset(&signals, SIGCHLD);
sigdelset(&signals, SIGTERM);
sigprocmask(SIG_SETMASK, &signals, &old_signals);
/* install signal handler */
memset(&sa, 0, sizeof(sa));
sa.sa_handler = handle_signal;
sigaction(SIGALRM, &sa, NULL);
sigaction(SIGCHLD, &sa, NULL);
sigaction(SIGTERM, &sa, NULL);
fp = fopen(pidfile, "w");
if (!fp)
eerrorx("%s: fopen `%s': %s", applet, pidfile, strerror(errno));
fprintf(fp, "%d\n", getpid());
fclose(fp);
if (svcname)
rc_service_daemon_set(svcname, exec, (const char * const *) argv,
pidfile, true);
/* remove the controlling tty */
#ifdef TIOCNOTTY
ioctl(tty_fd, TIOCNOTTY, 0);
close(tty_fd);
#endif
/*
* Supervisor main loop
*/
if (healthcheckdelay)
alarm(healthcheckdelay);
else if (healthchecktimer)
alarm(healthchecktimer);
failing = 0;
while (!exiting) {
healthcheck_respawn = 0;
fifo_fd = open(fifopath, O_RDONLY);
if (fifo_fd > 0) {
memset(buf, 0, sizeof(buf));
count = read(fifo_fd, buf, sizeof(buf) - 1);
close(fifo_fd);
if (count != -1)
buf[count] = 0;
if (count == 0)
continue;
if (verbose)
syslog(LOG_DEBUG, "Received %s from fifo", buf);
if (strncasecmp(buf, "sig", 3) == 0) {
if ((sscanf(buf, "%s %d", cmd, &sig_send) == 2)
&& (sig_send >= 0 && sig_send < NSIG)) {
syslog(LOG_INFO, "Sending signal %d to %d", sig_send,
child_pid);
if (kill(child_pid, sig_send) == -1)
syslog(LOG_ERR, "Unable to send signal %d to %d",
sig_send, child_pid);
}
}
continue;
}
if (do_healthcheck) {
do_healthcheck = 0;
alarm(0);
if (verbose)
syslog(LOG_DEBUG, "running health check for %s", svcname);
health_pid = exec_command("healthcheck");
health_status = rc_waitpid(health_pid);
if (WIFEXITED(health_status) && WEXITSTATUS(health_status) == 0)
alarm(healthchecktimer);
else {
syslog(LOG_WARNING, "health check for %s failed", svcname);
health_pid = exec_command("unhealthy");
rc_waitpid(health_pid);
syslog(LOG_INFO, "stopping %s, pid %d", exec, child_pid);
nkilled = run_stop_schedule(applet, NULL, NULL, child_pid, 0,
false, false, true);
if (nkilled < 0)
syslog(LOG_INFO, "Unable to kill %d: %s",
child_pid, strerror(errno));
else
healthcheck_respawn = 1;
}
}
if (exiting) {
alarm(0);
syslog(LOG_INFO, "stopping %s, pid %d", exec, child_pid);
nkilled = run_stop_schedule(applet, NULL, NULL, child_pid, 0,
false, false, true);
if (nkilled > 0)
syslog(LOG_INFO, "killed %d processes", nkilled);
continue;
}
wait_pid = waitpid(child_pid, &i, WNOHANG);
if (wait_pid == child_pid) {
if (WIFEXITED(i))
syslog(LOG_WARNING, "%s, pid %d, exited with return code %d",
exec, child_pid, WEXITSTATUS(i));
else if (WIFSIGNALED(i))
syslog(LOG_WARNING, "%s, pid %d, terminated by signal %d",
exec, child_pid, WTERMSIG(i));
}
if (wait_pid == child_pid || healthcheck_respawn) {
do_healthcheck = 0;
healthcheck_respawn = 0;
alarm(0);
respawn_now = time(NULL);
if (first_spawn == 0)
first_spawn = respawn_now;
if ((respawn_period > 0)
&& (respawn_now - first_spawn > respawn_period)) {
respawn_count = 0;
first_spawn = 0;
} else
respawn_count++;
if (respawn_max > 0 && respawn_count > respawn_max) {
syslog(LOG_WARNING, "respawned \"%s\" too many times, exiting",
exec);
exiting = 1;
failing = 1;
continue;
}
ts.tv_sec = respawn_delay;
ts.tv_nsec = 0;
nanosleep(&ts, NULL);
if (exiting)
continue;
child_pid = fork();
if (child_pid == -1) {
syslog(LOG_ERR, "%s: fork: %s", applet, strerror(errno));
exit(EXIT_FAILURE);
}
if (child_pid == 0) {
sigprocmask(SIG_SETMASK, &old_signals, NULL);
memset(&sa, 0, sizeof(sa));
sa.sa_handler = SIG_DFL;
sigaction(SIGALRM, &sa, NULL);
sigaction(SIGCHLD, &sa, NULL);
sigaction(SIGTERM, &sa, NULL);
child_process(exec, argv);
}
if (healthcheckdelay)
alarm(healthcheckdelay);
else if (healthchecktimer)
alarm(healthchecktimer);
}
}
if (svcname) {
rc_service_daemon_set(svcname, exec, (const char *const *)argv,
pidfile, false);
rc_service_value_set(svcname, "child_pid", NULL);
rc_service_mark(svcname, RC_SERVICE_STOPPED);
if (failing)
rc_service_mark(svcname, RC_SERVICE_FAILED);
}
if (pidfile && exists(pidfile))
unlink(pidfile);
if (fifopath && exists(fifopath))
unlink(fifopath);
exit(EXIT_SUCCESS);
}
int main(int argc, char **argv)
{
int opt;
char **c;
int x;
bool start = false;
bool stop = false;
bool reexec = false;
bool sendsig = false;
char *exec = NULL;
char *retry = NULL;
int sig = SIGTERM;
char *home = NULL;
int tid = 0;
pid_t pid;
char *tmp;
char *p;
char *token;
int i;
int n;
char *exec_file = NULL;
char *varbuf = NULL;
struct timespec ts;
struct passwd *pw;
struct group *gr;
FILE *fp;
mode_t numask = 022;
int child_argc = 0;
char **child_argv = NULL;
char *str = NULL;
char *cmdline = NULL;
applet = basename_c(argv[0]);
atexit(cleanup);
svcname = getenv("RC_SVCNAME");
if (!svcname)
eerrorx("%s: The RC_SVCNAME environment variable is not set", applet);
openlog(applet, LOG_PID, LOG_DAEMON);
if (argc <= 1 || strcmp(argv[1], svcname))
eerrorx("%s: the first argument is %s and must be %s",
applet, argv[1], svcname);
if ((tmp = getenv("SSD_NICELEVEL")))
if (sscanf(tmp, "%d", &nicelevel) != 1)
eerror("%s: invalid nice level `%s' (SSD_NICELEVEL)",
applet, tmp);
if ((tmp = getenv("SSD_IONICELEVEL"))) {
n = sscanf(tmp, "%d:%d", &ionicec, &ioniced);
if (n != 1 && n != 2)
eerror("%s: invalid ionice level `%s' (SSD_IONICELEVEL)",
applet, tmp);
if (ionicec == 0)
ioniced = 0;
else if (ionicec == 3)
ioniced = 7;
ionicec <<= 13; /* class shift */
}
if ((tmp = getenv("SSD_OOM_SCORE_ADJ")))
if (sscanf(tmp, "%d", &oom_score_adj) != 1)
eerror("%s: invalid oom_score_adj `%s' (SSD_OOM_SCORE_ADJ)",
applet, tmp);
/* Get our user name and initial dir */
p = getenv("USER");
home = getenv("HOME");
if (home == NULL || p == NULL) {
pw = getpwuid(getuid());
if (pw != NULL) {
if (p == NULL)
setenv("USER", pw->pw_name, 1);
if (home == NULL) {
setenv("HOME", pw->pw_dir, 1);
home = pw->pw_dir;
}
}
}
cmdline = make_cmdline(argv);
if (svcname) {
argc--;
argv++;
}
while ((opt = getopt_long(argc, argv, getoptstring, longopts,
(int *) 0)) != -1)
switch (opt) {
case 'a': /* --healthcheck-timer <time> */
if (sscanf(optarg, "%d", &healthchecktimer) != 1 || healthchecktimer < 1)
eerrorx("%s: invalid health check timer %s", applet, optarg);
break;
case 'A': /* --healthcheck-delay <time> */
if (sscanf(optarg, "%d", &healthcheckdelay) != 1 || healthcheckdelay < 1)
eerrorx("%s: invalid health check delay %s", applet, optarg);
break;
case LONGOPT_CAPABILITIES:
#ifdef HAVE_CAP
cap_iab = cap_iab_from_text(optarg);
if (cap_iab == NULL)
eerrorx("Could not parse iab: %s", strerror(errno));
#else
eerrorx("Capabilities support not enabled");
#endif
break;
case LONGOPT_SECBITS:
#ifdef HAVE_CAP
if (*optarg == '\0')
eerrorx("Secbits are empty");
tmp = NULL;
secbits = strtoul(optarg, &tmp, 0);
if (*tmp != '\0')
eerrorx("Could not parse secbits: invalid char %c", *tmp);
#else
eerrorx("Capabilities support not enabled");
#endif
break;
case LONGOPT_NO_NEW_PRIVS:
#ifdef PR_SET_NO_NEW_PRIVS
no_new_privs = true;
#else
eerrorx("The No New Privs flag is only supported by Linux (since 3.5)");
#endif
break;
case 'D': /* --respawn-delay time */
n = sscanf(optarg, "%d", &respawn_delay);
if (n != 1 || respawn_delay < 1)
eerrorx("Invalid respawn-delay value '%s'", optarg);
break;
case 'I': /* --ionice */
if (sscanf(optarg, "%d:%d", &ionicec, &ioniced) == 0)
eerrorx("%s: invalid ionice `%s'",
applet, optarg);
if (ionicec == 0)
ioniced = 0;
else if (ionicec == 3)
ioniced = 7;
ionicec <<= 13; /* class shift */
break;
case 'K': /* --stop */
stop = true;
break;
case 'N': /* --nice */
if (sscanf(optarg, "%d", &nicelevel) != 1)
eerrorx("%s: invalid nice level `%s'",
applet, optarg);
break;
case LONGOPT_OOM_SCORE_ADJ: /* --oom-score-adj */
if (sscanf(optarg, "%d", &oom_score_adj) != 1)
eerrorx("%s: invalid oom-score-adj `%s'",
applet, optarg);
break;
case 'P': /* --respawn-period time */
n = sscanf(optarg, "%d", &respawn_period);
if (n != 1 || respawn_period < 1)
eerrorx("Invalid respawn-period value '%s'", optarg);
break;
case 's': /* --signal */
sig = parse_signal(applet, optarg);
sendsig = true;
break;
case 'S': /* --start */
start = true;
break;
case 'd': /* --chdir /new/dir */
ch_dir = optarg;
break;
case 'e': /* --env */
putenv(optarg);
break;
case 'g': /* --group <group>|<gid> */
if (sscanf(optarg, "%d", &tid) != 1)
gr = getgrnam(optarg);
else
gr = getgrgid((gid_t)tid);
if (gr == NULL)
eerrorx("%s: group `%s' not found",
applet, optarg);
gid = gr->gr_gid;
break;
case 'H': /* --healthcheck-timer <minutes> */
if (sscanf(optarg, "%d", &healthchecktimer) != 1 || healthchecktimer < 1)
eerrorx("%s: invalid health check timer %s", applet, optarg);
break;
case 'k':
if (parse_mode(&numask, optarg))
eerrorx("%s: invalid mode `%s'",
applet, optarg);
break;
case 'm': /* --respawn-max count */
n = sscanf(optarg, "%d", &respawn_max);
if (n != 1 || respawn_max < 0)
eerrorx("Invalid respawn-max value '%s'", optarg);
break;
case 'p': /* --pidfile <pid-file> */
pidfile = optarg;
break;
case 'R': /* --retry <schedule>|timeout */
retry = optarg;
break;
case 'r': /* --chroot /new/root */
ch_root = optarg;
break;
case 'u': /* --user <username>|<uid> */
{
char dummy[2];
p = optarg;
tmp = strsep(&p, ":");
changeuser = xstrdup(tmp);
if (sscanf(tmp, "%d%1s", &tid, dummy) != 1)
pw = getpwnam(tmp);
else
pw = getpwuid((uid_t)tid);
if (pw == NULL)
eerrorx("%s: user `%s' not found",
applet, tmp);
uid = pw->pw_uid;
home = pw->pw_dir;
unsetenv("HOME");
if (pw->pw_dir)
setenv("HOME", pw->pw_dir, 1);
unsetenv("USER");
if (pw->pw_name)
setenv("USER", pw->pw_name, 1);
if (gid == 0)
gid = pw->pw_gid;
if (p) {
tmp = strsep (&p, ":");
if (sscanf(tmp, "%d%1s", &tid, dummy) != 1)
gr = getgrnam(tmp);
else
gr = getgrgid((gid_t) tid);
if (gr == NULL)
eerrorx("%s: group `%s'"
" not found",
applet, tmp);
gid = gr->gr_gid;
}
}
break;
case '1': /* --stdout /path/to/stdout.lgfile */
redirect_stdout = optarg;
break;
case '2': /* --stderr /path/to/stderr.logfile */
redirect_stderr = optarg;
break;
case '3': /* --reexec */
reexec = true;
break;
case_RC_COMMON_GETOPT
}
verbose = rc_yesno(getenv ("EINFO_VERBOSE"));
endpwent();
argc -= optind;
argv += optind;
exec = *argv;
/* Expand ~ */
if (ch_dir && *ch_dir == '~')
ch_dir = expand_home(home, ch_dir);
if (ch_root && *ch_root == '~')
ch_root = expand_home(home, ch_root);
umask(numask);
if (!pidfile)
xasprintf(&pidfile, "/var/run/supervise-%s.pid", svcname);
xasprintf(&fifopath, "%s/supervise-%s.ctl", RC_SVCDIR, svcname);
if (mkfifo(fifopath, 0600) == -1 && errno != EEXIST)
eerrorx("%s: unable to create control fifo: %s",
applet, strerror(errno));
if (reexec) {
str = rc_service_value_get(svcname, "argc");
sscanf(str, "%d", &child_argc);
child_argv = xmalloc((child_argc + 1) * sizeof(char *));
memset(child_argv, 0, (child_argc + 1) * sizeof(char *));
for (x = 0; x < child_argc; x++) {
xasprintf(&varbuf, "argv_%d", x);
str = rc_service_value_get(svcname, varbuf);
child_argv[x] = str;
free(varbuf);
varbuf = NULL;
}
free(str);
str = rc_service_value_get(svcname, "child_pid");
sscanf(str, "%d", &child_pid);
free(str);
exec = rc_service_value_get(svcname, "exec");
pidfile = rc_service_value_get(svcname, "pidfile");
retry = rc_service_value_get(svcname, "retry");
if (retry) {
parse_schedule(applet, retry, sig);
rc_service_value_set(svcname, "retry", retry);
} else
parse_schedule(applet, NULL, sig);
str = rc_service_value_get(svcname, "respawn_delay");
sscanf(str, "%d", &respawn_delay);
str = rc_service_value_get(svcname, "respawn_max");
sscanf(str, "%d", &respawn_max);
supervisor(exec, child_argv);
} else if (start) {
if (exec) {
if (*exec == '~')
exec = expand_home(home, exec);
/* Validate that the binary exists if we are starting */
if (*exec == '/' || *exec == '.') {
/* Full or relative path */
if (ch_root)
xasprintf(&exec_file, "%s/%s", ch_root, exec);
else
xasprintf(&exec_file, "%s", exec);
} else {
/* Something in $PATH */
p = tmp = xstrdup(getenv("PATH"));
exec_file = NULL;
while ((token = strsep(&p, ":"))) {
if (ch_root)
xasprintf(&exec_file, "%s/%s/%s", ch_root, token, exec);
else
xasprintf(&exec_file, "%s/%s", token, exec);
if (exec_file && exists(exec_file))
break;
free(exec_file);
exec_file = NULL;
}
free(tmp);
}
if (!exists(exec_file)) {
eerror("%s: %s does not exist", applet,
exec_file ? exec_file : exec);
free(exec_file);
exit(EXIT_FAILURE);
}
} else
eerrorx("%s: nothing to start", applet);
2017-08-23 14:16:49 -05:00
pid = get_pid(applet, pidfile);
if (pid != -1)
if (do_stop(applet, exec, (const char * const *)argv, pid, uid,
0, false, true) > 0)
eerrorx("%s: %s is already running", applet, exec);
if (respawn_period > 0 && respawn_delay * respawn_max > respawn_period)
ewarn("%s: Please increase the value of --respawn-period to more "
2018-10-12 17:16:23 -05:00
"than %d to avoid infinite respawning", applet,
respawn_delay * respawn_max);
if (retry) {
parse_schedule(applet, retry, sig);
rc_service_value_set(svcname, "retry", retry);
} else
parse_schedule(applet, NULL, sig);
einfov("Detaching to start `%s'", exec);
syslog(LOG_INFO, "Supervisor command line: %s", cmdline);
free(cmdline);
cmdline = NULL;
/* Remove existing pidfile */
if (pidfile)
unlink(pidfile);
/* Make sure we can write a pid file */
fp = fopen(pidfile, "w");
if (!fp)
eerrorx("%s: fopen `%s': %s", applet, pidfile, strerror(errno));
fclose(fp);
rc_service_value_set(svcname, "pidfile", pidfile);
varbuf = NULL;
xasprintf(&varbuf, "%i", respawn_delay);
rc_service_value_set(svcname, "respawn_delay", varbuf);
free(varbuf);
xasprintf(&varbuf, "%i", respawn_max);
rc_service_value_set(svcname, "respawn_max", varbuf);
free(varbuf);
xasprintf(&varbuf, "%i", respawn_period);
rc_service_value_set(svcname, "respawn_period", varbuf);
free(varbuf);
child_pid = fork();
if (child_pid == -1)
eerrorx("%s: fork: %s", applet, strerror(errno));
if (child_pid != 0)
/* first parent process, do nothing. */
exit(EXIT_SUCCESS);
#ifdef TIOCNOTTY
tty_fd = open("/dev/tty", O_RDWR);
#endif
devnull_fd = open("/dev/null", O_RDWR);
dup2(devnull_fd, STDIN_FILENO);
dup2(devnull_fd, STDOUT_FILENO);
dup2(devnull_fd, STDERR_FILENO);
child_pid = fork();
if (child_pid == -1)
eerrorx("%s: fork: %s", applet, strerror(errno));
else if (child_pid != 0) {
c = argv;
x = 0;
while (c && *c) {
varbuf = NULL;
xasprintf(&varbuf, "argv_%-d",x);
rc_service_value_set(svcname, varbuf, *c);
free(varbuf);
varbuf = NULL;
x++;
c++;
}
xasprintf(&varbuf, "%d", x);
rc_service_value_set(svcname, "argc", varbuf);
free(varbuf);
rc_service_value_set(svcname, "exec", exec);
supervisor(exec, argv);
} else
child_process(exec, argv);
} else if (stop) {
pid = get_pid(applet, pidfile);
if (pid != -1) {
i = kill(pid, SIGTERM);
if (i != 0)
/* We failed to send the signal */
ewarn("Unable to shut down the supervisor");
else {
/* wait for the supervisor to go down */
while (kill(pid, 0) == 0) {
ts.tv_sec = 0;
ts.tv_nsec = 1;
nanosleep(&ts, NULL);
}
}
}
/* Even if we have not actually killed anything, we should
* remove information about it as it may have unexpectedly
* crashed out. We should also return success as the end
* result would be the same. */
if (pidfile && exists(pidfile))
unlink(pidfile);
if (svcname) {
rc_service_daemon_set(svcname, exec,
(const char *const *)argv,
pidfile, false);
rc_service_mark(svcname, RC_SERVICE_STOPPED);
}
exit(EXIT_SUCCESS);
} else if (sendsig) {
fifo_fd = open(fifopath, O_WRONLY |O_NONBLOCK);
if (fifo_fd < 0)
eerrorx("%s: unable to open control fifo %s", applet, strerror(errno));
xasprintf(&str, "sig %d", sig);
x = write(fifo_fd, str, strlen(str));
if (x == -1) {
free(tmp);
eerrorx("%s: error writing to control fifo: %s", applet,
strerror(errno));
}
free(tmp);
exit(EXIT_SUCCESS);
}
}