static void down_u (void)
{
status.flagwant = 1 ;
status.flagwantup = 1 ;
announce() ;
trystart() ;
}
int main(int argc,char **argv)
{
struct stat st;
dir = argv[1];
if (!dir || argv[2])
strerr_die1x(100,"supervise: usage: supervise dir");
if (pipe(selfpipe) == -1)
strerr_die4sys(111,FATAL,"unable to create pipe for ",dir,": ");
coe(selfpipe[0]);
coe(selfpipe[1]);
ndelay_on(selfpipe[0]);
ndelay_on(selfpipe[1]);
sig_block(sig_child);
sig_catch(sig_child,trigger);
if (chdir(dir) == -1)
strerr_die4sys(111,FATAL,"unable to chdir to ",dir,": ");
if (stat("down",&st) != -1)
flagwantup = 0;
else
if (errno != error_noent)
strerr_die4sys(111,FATAL,"unable to stat ",dir,"/down: ");
mkdir("supervise",0700);
fdlock = open_append("supervise/lock");
if ((fdlock == -1) || (lock_exnb(fdlock) == -1))
strerr_die4sys(111,FATAL,"unable to acquire ",dir,"/supervise/lock: ");
coe(fdlock);
fifo_make("supervise/control",0600);
fdcontrol = open_read("supervise/control");
if (fdcontrol == -1)
strerr_die4sys(111,FATAL,"unable to read ",dir,"/supervise/control: ");
coe(fdcontrol);
ndelay_on(fdcontrol); /* shouldn't be necessary */
fdcontrolwrite = open_write("supervise/control");
if (fdcontrolwrite == -1)
strerr_die4sys(111,FATAL,"unable to write ",dir,"/supervise/control: ");
coe(fdcontrolwrite);
pidchange();
announce();
fifo_make("supervise/ok",0600);
fdok = open_read("supervise/ok");
if (fdok == -1)
strerr_die4sys(111,FATAL,"unable to read ",dir,"/supervise/ok: ");
coe(fdok);
if (!flagwant || flagwantup) trystart();
doit();
announce();
_exit(0);
}
int supervise_run(void)
{
g_pid = 0;
g_flagexit = 0;
g_flagwant = 1;
g_flagwantup = opt_auto_start;
g_flagpaused = 0;
sig_block(SIGCHLD);
sig_catch(SIGCHLD, sigchild_handler);
stat_pidchange();
stat_update();
if(g_flagwant && g_flagwantup) {
trystart();
}
while(1) {
char c;
ssize_t rl;
if(g_flagexit && !g_pid) { return 0; }
printf("waiting pid pid=%d\n",g_pid);
while(1) {
int stat;
int r = waitpid(-1, &stat, WNOHANG);
if(r == 0) { break; }
if(r < 0 && errno != EAGAIN && errno != EINTR) { break; }
if(r == g_pid) {
g_pid = 0;
stat_pidchange();
stat_update();
if(g_flagexit) { return 0; }
if(g_flagwant && g_flagwantup) {
trystart();
break;
}
}
}
printf("reading... pid=%d\n",g_pid);
sig_unblock(SIGCHLD);
rl = read(g_ctl_rfd, &c, 1);
if(rl <= 0) {
if(errno == EAGAIN || errno == EINTR) {
continue;
}
return -1;
}
sig_block(SIGCHLD);
switch(c) {
case 'd': /* down */
printf("down %d\n",g_pid);
g_flagwant = 1;
g_flagwantup = 0;
if(g_pid) {
kill(g_pid, SIGTERM);
kill(g_pid, SIGCONT);
g_flagpaused = 0;
}
stat_update();
break;
case 'u': /* up */
printf("up %d\n",g_pid);
g_flagwant = 1;
g_flagwantup = 1;
if(!g_pid) { trystart(); }
stat_update();
break;
case 'o': /* once */
printf("once %d\n",g_pid);
g_flagwant = 0;
if(!g_pid) { trystart(); }
stat_update();
break;
case 'x': /* exit */
printf("exit %d\n",g_pid);
g_flagexit = 1;
stat_update();
break;
case 'p': /* pause */
printf("pause %d\n",g_pid);
g_flagpaused = 1;
if(g_pid) { kill(g_pid, SIGSTOP); }
stat_update();
break;
case 'c': /* continue */
printf("continue %d\n",g_pid);
g_flagpaused = 0;
if(g_pid) { kill(g_pid, SIGCONT); }
stat_update();
break;
case 'h': /* hup */
printf("hup %d\n",g_pid);
if(g_pid) { kill(g_pid, SIGHUP); }
break;
case 'a': /* alarm */
printf("alarm %d\n",g_pid);
if(g_pid) { kill(g_pid, SIGALRM); }
break;
case 'i': /* interrupt */
printf("interrupt %d\n",g_pid);
if(g_pid) { kill(g_pid, SIGINT); }
break;
case 't': /* term */
printf("term %d\n",g_pid);
if(g_pid) { kill(g_pid, SIGTERM); }
break;
case 'k': /* kill */
printf("kill %d\n",g_pid);
if(g_pid) { kill(g_pid, SIGKILL); }
break;
case '1': /* usr1 */
printf("usr1 %d\n",g_pid);
if(g_pid) { kill(g_pid, SIGUSR1); }
break;
case '2': /* usr2 */
printf("usr2 %d\n",g_pid);
if(g_pid) { kill(g_pid, SIGUSR2); }
break;
case 's': /* sigchld */
printf("sigchld %d\n",g_pid);
if(!opt_auto_restart) {
g_flagwant = 0;
stat_update();
}
break;
case ' ': /* ping */
// FIXME touch status file
break;
}
}
}
static void check (char const *name)
{
struct stat st ;
unsigned int namelen ;
unsigned int i = 0 ;
if (name[0] == '.') return ;
if (stat(name, &st) == -1)
{
strerr_warnwu2sys("stat ", name) ;
retrydirlater() ;
return ;
}
if (!S_ISDIR(st.st_mode)) return ;
namelen = str_len(name) ;
for (; i < n ; i++) if ((services[i].ino == st.st_ino) && (services[i].dev == st.st_dev)) break ;
if (i < n)
{
if (services[i].flaglog && (services[i].p[0] < 0))
{
/* See BLACK MAGIC above. */
services[i].p[0] = -2 ;
return ;
}
}
else
{
if (n >= max)
{
strerr_warnwu3x("start supervisor for ", name, ": too many services") ;
return ;
}
else
{
struct stat su ;
char tmp[namelen + 5] ;
byte_copy(tmp, namelen, name) ;
byte_copy(tmp + namelen, 5, "/log") ;
if (stat(tmp, &su) < 0)
if (errno == ENOENT) services[i].flaglog = 0 ;
else
{
strerr_warnwu2sys("stat ", tmp) ;
retrydirlater() ;
return ;
}
else if (!S_ISDIR(su.st_mode))
services[i].flaglog = 0 ;
else
{
if (pipecoe(services[i].p) < 0)
{
strerr_warnwu1sys("pipecoe") ;
retrydirlater() ;
return ;
}
services[i].flaglog = 1 ;
}
services[i].ino = st.st_ino ;
services[i].dev = st.st_dev ;
tain_copynow(&services[i].restartafter[0]) ;
tain_copynow(&services[i].restartafter[1]) ;
services[i].pid[0] = 0 ;
services[i].pid[1] = 0 ;
n++ ;
}
}
services[i].flagactive = 1 ;
if (services[i].flaglog && !services[i].pid[1])
{
if (!tain_future(&services[i].restartafter[1]))
{
char tmp[namelen + 5] ;
byte_copy(tmp, namelen, name) ;
byte_copy(tmp + namelen, 5, "/log") ;
trystart(i, tmp, 1) ;
}
else if (tain_less(&services[i].restartafter[1], &deadline))
deadline = services[i].restartafter[1] ;
}
if (!services[i].pid[0])
{
if (!tain_future(&services[i].restartafter[0]))
trystart(i, name, 0) ;
else if (tain_less(&services[i].restartafter[0], &deadline))
deadline = services[i].restartafter[0] ;
}
}
void doit(void)
{
iopause_fd x[2];
struct taia deadline;
struct taia stamp;
int wstat;
int r;
char ch;
announce();
for (;;) {
if (flagexit && !pid) return;
sig_unblock(sig_child);
x[0].fd = selfpipe[0];
x[0].events = IOPAUSE_READ;
x[1].fd = fdcontrol;
x[1].events = IOPAUSE_READ;
taia_now(&stamp);
taia_uint(&deadline,3600);
taia_add(&deadline,&stamp,&deadline);
iopause(x,2,&deadline,&stamp);
sig_block(sig_child);
while (read(selfpipe[0],&ch,1) == 1)
;
for (;;) {
r = wait_nohang(&wstat);
if (!r) break;
if ((r == -1) && (errno != error_intr)) break;
if (r == pid) {
pid = 0;
pidchange();
announce();
if (flagexit) return;
if (flagwant && flagwantup) trystart();
break;
}
}
if (read(fdcontrol,&ch,1) == 1)
switch(ch) {
case 'd':
flagwant = 1;
flagwantup = 0;
if (pid) { kill(pid,SIGTERM); kill(pid,SIGCONT); flagpaused = 0; }
announce();
break;
case 'u':
flagwant = 1;
flagwantup = 1;
announce();
if (!pid) trystart();
break;
case 'o':
flagwant = 0;
announce();
if (!pid) trystart();
break;
case 'a':
if (pid) kill(pid,SIGALRM);
break;
case 'h':
if (pid) kill(pid,SIGHUP);
break;
case 'k':
if (pid) kill(pid,SIGKILL);
break;
case 't':
if (pid) kill(pid,SIGTERM);
break;
case 'i':
if (pid) kill(pid,SIGINT);
break;
case 'p':
flagpaused = 1;
announce();
if (pid) kill(pid,SIGSTOP);
break;
case 'c':
flagpaused = 0;
announce();
if (pid) kill(pid,SIGCONT);
break;
case 'x':
flagexit = 1;
announce();
break;
}
}
}
int main(int argc, char **argv)
{
struct sigaction sa;
if (parse_argv(argc, argv) < 0)
_exit(100);
snprintf(status_files[0], 1024, "%s/lock", status_dir);
snprintf(status_files[1], 1024, "%s/control", status_dir);
snprintf(status_files[2], 1024, "%s/ok", status_dir);
snprintf(status_files[3], 1024, "%s/status", status_dir);
snprintf(status_files[4], 1024, "%s/status.new", status_dir);
snprintf(status_files[5], 1024, "%s/supervise.log", status_dir);
snprintf(status_files[6], 1024, "%s/supervise.log.wf", status_dir);
sa.sa_handler = SIG_IGN;
sigemptyset(&sa.sa_mask);
if (sigaction(SIGHUP, &sa, NULL) < 0)
{
printf("unable to ignore SIGHUP for %s\n", service);
_exit(110);
}
if (mkdir(status_dir, 0700) < 0 && errno != EEXIST)
{
printf("unable to create dir: %s\n", status_dir);
_exit(110);
}
fdlog = open_append(status_files[5]);
if (fdlog == -1)
{
printf("unable to open %s%s", status_dir, "/supervise.log");
_exit(111);
}
coe(fdlog);
fdlogwf = open_append(status_files[6]);
if (fdlogwf == -1)
{
printf("unable to open %s%s", status_dir, "/supervise.log.wf");
_exit(1);
}
coe(fdlogwf);
if (daemon(1, 0) < 0)
{
printf("failed to daemonize supervise!\n");
_exit(111);
}
if (pipe(selfpipe) == -1)
{
write_log(fdlogwf, FATAL, "unable to create pipe for ", service, "\n");
_exit(111);
}
coe(selfpipe[0]);
coe(selfpipe[1]);
ndelay_on(selfpipe[0]);
ndelay_on(selfpipe[1]);
sig_block(sig_child);
sig_catch(sig_child, trigger);
sig_block(sig_alarm);
sig_catch(sig_alarm, timer_handler);
sig_unblock(sig_alarm);
fdlock = open_append(status_files[0]);
if ((fdlock == -1) || (lock_exnb(fdlock) == -1))
{
write_log(fdlogwf, FATAL, "Unable to acquier ", status_dir, "/lock\n");
_exit(111);
}
coe(fdlock);
fifo_make(status_files[1], 0600);
fdcontrol = open_read(status_files[1]);
if (fdcontrol == -1)
{
write_log(fdlogwf, FATAL, "unable to read ", status_dir, "/control\n");
_exit(1);
}
coe(fdcontrol);
ndelay_on(fdcontrol);
fdcontrolwrite = open_write(status_files[1]);
if (fdcontrolwrite == -1)
{
write_log(fdlogwf, FATAL, "unable to write ", status_dir, "/control\n");
_exit(1);
}
coe(fdcontrolwrite);
fifo_make(status_files[2], 0600);
fdok = open_read(status_files[2]);
if (fdok == -1)
{
write_log(fdlogwf, FATAL, "unable to read ", status_dir, "/ok\n");
_exit(1);
}
coe(fdok);
if (!restart_sh[0])
{
parse_conf();
}
pidchange();
announce();
if (!flagwant || flagwantup)
trystart();
doit();
announce();
_exit(0);
}
static void doit()
{
iopause_fd x[2];
struct taia deadline;
struct taia stamp;
int wstat;
long int time_cmp = 0;
int r;
char ch;
char warn_message[2048];
int coredumped = 0;
char restart_cmd[2048];
announce();
x[0].fd = selfpipe[0];
x[0].events = IOPAUSE_READ;
x[1].fd = fdcontrol;
x[1].events = IOPAUSE_READ;
while (1)
{
if (flagexit && !pid)
break;
taia_now(&stamp);
taia_uint(&deadline, 3600);
taia_add(&deadline, &stamp, &deadline);
sig_unblock(sig_child);
iopause(x, 2, &deadline, &stamp);
sig_block(sig_child);
while (read(selfpipe[0], &ch, 1) == 1);
while (1)
{
//waitpid(-1,&wstat,WNOHANG);WNOHANG : return immediately if no child has exited
r = wait_nohang(&wstat);
//r==0 if one or more child(ren) exit(s) but have not yet changed state
if (!r)
break;
//r == -1 && errno == error_intr means waitpid is interrupted by a signal, we should call waitpid again.
//here is not necessary cause we call waitpid with a WNOHANG argument.
if (r == -1 && errno != error_intr)
break;
if (r == pid)
{
pid = 0;
pidchange();
announce();
time_now = time((time_t *)0);
if(time_old)
{
time_cmp = time_now - time_old;
if(time_cmp >= time_alarm) //cmp
{
num = 0;
}
time_old = time_now;
}
else
{
time_cmp = 0;
time_old = time_now;
}
if (0 != restart_sh[0])
{
if (num == INT_MAX)
num = 0;
num++;
if (snprintf(restart_cmd, sizeof(restart_cmd), "%s %d", restart_sh, num) > 1)
{
system(restart_cmd);
write_log(fdlog, NOTICE, "restart_cmd: ", restart_cmd, " is called.\n");
}
}
else
{
if (WCOREDUMP(wstat))
{
have_coredumped++;
coredumped = 1;
}
bzero(warn_message, 2048);
create_warn_message(warn_message, coredumped);
write_log(fdlog, NOTICE, "service exited", coredumped ? " with coredump" : "", "!\n");
coredumped = 0;
if (!closealarm && alarm_interval > 0 && have_tried++ == 0)
{
alarm(alarm_interval);
do_alarm(warn_message);
}
if (flagexit || (alarm_interval > 0 && have_tried > max_tries && max_tries > 0) ||
(alarm_interval > 0 && have_coredumped > max_tries_if_coredumped && max_tries_if_coredumped > 0))
{
write_log(fdlog, NOTICE, "supervise refused to restart ", service,
" any more and exited itself!\n");
alarm(0);
return;
}
}
if (flagwant && flagwantup)
{
write_log(fdlog, NOTICE, "supervise is trying to restart ", service, "...\n");
trystart();
}
break;
}
}
if (read(fdcontrol, &ch, 1) != 1)
continue;
switch(ch)
{
// -s: Silent. Do not alarm any more.
case 's':
closealarm = 1;
announce();
break;
case 'n':
closealarm = 0;
announce();
break;
case 'r':
parse_conf();
break;
// -d: Down. If the service is running, send it a TERM signal and then a CONT signal.
// After it stops, do not restart it.
case 'd':
flagwant = 1;
flagwantup = 0;
if (pid) { kill(pid, SIGTERM); kill(pid, SIGCONT); flagpaused = 0; }
announce();
break;
// -u: Up. If the service is not running, start it. If the service stops, restart it.
case 'u':
flagwant = 1;
flagwantup = 1;
announce();
if (!pid) trystart();
break;
// -o: Once. If the service is not running, start it. Do not restart it if it stops.
case 'o':
flagwant = 0;
announce();
if (!pid) trystart();
break;
// -a: Alarm. Send the service an ALRM signal.
case 'a':
if (pid) kill(pid, SIGALRM);
break;
// -h: Hangup. Send the service a HUP signal.
case 'h':
if (pid) kill(pid, SIGHUP);
break;
// -k: Kill. Send the service a KILL signal.
case 'k':
if (pid) kill(pid, SIGKILL);
break;
//* -t: Terminate. Send the service a TERM signal.
case 't':
if (pid) kill(pid, SIGTERM);
break;
// -i: Interrupt. Send the service an INT signal.
case 'i':
if (pid) kill(pid, SIGINT);
break;
// -p: Pause. Send the service a STOP signal.
case 'p':
flagpaused = 1;
announce();
if (pid) kill(pid, SIGSTOP);
break;
// -c: Continue. Send the service a CONT signal.
case 'c':
flagpaused = 0;
announce();
if (pid) kill(pid, SIGCONT);
break;
// -x: Exit. supervise will exit as soon as the service is down. If you use this option on a
//stable system, you're doing something wrong; supervise is designed to run forever.
case 'x':
flagexit = 1;
announce();
break;
} //end switch
} //end while
}
static void down_o (void)
{
down_O() ;
trystart() ;
}
static void downtimeout (void)
{
if (status.flagwant && status.flagwantup) trystart() ;
else settimeout_infinite() ;
}
