]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * Monitor status of frr daemons and restart if necessary. | |
3 | * | |
4 | * Copyright (C) 2004 Andrew J. Schorr | |
5 | * | |
6 | * This program is free software; you can redistribute it and/or modify | |
7 | * it under the terms of the GNU General Public License as published by | |
8 | * the Free Software Foundation; either version 2 of the License, or | |
9 | * (at your option) any later version. | |
10 | * | |
11 | * This program is distributed in the hope that it will be useful, | |
12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
14 | * GNU General Public License for more details. | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License along | |
17 | * with this program; see the file COPYING; if not, write to the Free Software | |
18 | * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
19 | */ | |
20 | ||
21 | #include <zebra.h> | |
22 | #include <thread.h> | |
23 | #include <log.h> | |
24 | #include <network.h> | |
25 | #include <sigevent.h> | |
26 | #include <lib/version.h> | |
27 | #include "command.h" | |
28 | #include "memory_vty.h" | |
29 | #include "libfrr.h" | |
30 | ||
31 | #include <getopt.h> | |
32 | #include <sys/un.h> | |
33 | #include <sys/wait.h> | |
34 | #include <memory.h> | |
35 | #include <systemd.h> | |
36 | ||
37 | #include "watchfrr.h" | |
38 | ||
39 | #ifndef MIN | |
40 | #define MIN(X,Y) (((X) <= (Y)) ? (X) : (Y)) | |
41 | #endif | |
42 | ||
43 | /* Macros to help randomize timers. */ | |
44 | #define JITTER(X) ((random() % ((X)+1))-((X)/2)) | |
45 | #define FUZZY(X) ((X)+JITTER((X)/20)) | |
46 | ||
47 | #define DEFAULT_PERIOD 5 | |
48 | #define DEFAULT_TIMEOUT 90 | |
49 | #define DEFAULT_RESTART_TIMEOUT 20 | |
50 | #define DEFAULT_LOGLEVEL LOG_INFO | |
51 | #define DEFAULT_MIN_RESTART 60 | |
52 | #define DEFAULT_MAX_RESTART 600 | |
53 | ||
54 | #define PING_TOKEN "PING" | |
55 | ||
56 | /* Needs to be global, referenced somewhere inside libfrr. */ | |
57 | struct thread_master *master; | |
58 | static char pidfile_default[256]; | |
59 | ||
60 | static bool watch_only = false; | |
61 | ||
62 | typedef enum { | |
63 | PHASE_NONE = 0, | |
64 | PHASE_STOPS_PENDING, | |
65 | PHASE_WAITING_DOWN, | |
66 | PHASE_ZEBRA_RESTART_PENDING, | |
67 | PHASE_WAITING_ZEBRA_UP | |
68 | } restart_phase_t; | |
69 | ||
70 | static const char *phase_str[] = { | |
71 | "None", | |
72 | "Stop jobs running", | |
73 | "Waiting for other daemons to come down", | |
74 | "Zebra restart job running", | |
75 | "Waiting for zebra to come up", | |
76 | "Start jobs running", | |
77 | }; | |
78 | ||
79 | #define PHASE_TIMEOUT (3*gs.restart_timeout) | |
80 | ||
81 | struct restart_info { | |
82 | const char *name; | |
83 | const char *what; | |
84 | pid_t pid; | |
85 | struct timeval time; | |
86 | long interval; | |
87 | struct thread *t_kill; | |
88 | int kills; | |
89 | }; | |
90 | ||
91 | static struct global_state { | |
92 | restart_phase_t phase; | |
93 | struct thread *t_phase_hanging; | |
94 | const char *vtydir; | |
95 | long period; | |
96 | long timeout; | |
97 | long restart_timeout; | |
98 | long min_restart_interval; | |
99 | long max_restart_interval; | |
100 | struct daemon *daemons; | |
101 | const char *restart_command; | |
102 | const char *start_command; | |
103 | const char *stop_command; | |
104 | struct restart_info restart; | |
105 | int loglevel; | |
106 | struct daemon *special; /* points to zebra when doing phased restart */ | |
107 | int numdaemons; | |
108 | int numpids; | |
109 | int numdown; /* # of daemons that are not UP or UNRESPONSIVE */ | |
110 | } gs = { | |
111 | .phase = PHASE_NONE, | |
112 | .vtydir = frr_vtydir, | |
113 | .period = 1000 * DEFAULT_PERIOD, | |
114 | .timeout = DEFAULT_TIMEOUT, | |
115 | .restart_timeout = DEFAULT_RESTART_TIMEOUT, | |
116 | .loglevel = DEFAULT_LOGLEVEL, | |
117 | .min_restart_interval = DEFAULT_MIN_RESTART, | |
118 | .max_restart_interval = DEFAULT_MAX_RESTART, | |
119 | }; | |
120 | ||
121 | typedef enum { | |
122 | DAEMON_INIT, | |
123 | DAEMON_DOWN, | |
124 | DAEMON_CONNECTING, | |
125 | DAEMON_UP, | |
126 | DAEMON_UNRESPONSIVE | |
127 | } daemon_state_t; | |
128 | ||
129 | #define IS_UP(DMN) \ | |
130 | (((DMN)->state == DAEMON_UP) || ((DMN)->state == DAEMON_UNRESPONSIVE)) | |
131 | ||
132 | static const char *state_str[] = { | |
133 | "Init", "Down", "Connecting", "Up", "Unresponsive", | |
134 | }; | |
135 | ||
136 | struct daemon { | |
137 | const char *name; | |
138 | daemon_state_t state; | |
139 | int fd; | |
140 | struct timeval echo_sent; | |
141 | u_int connect_tries; | |
142 | struct thread *t_wakeup; | |
143 | struct thread *t_read; | |
144 | struct thread *t_write; | |
145 | struct daemon *next; | |
146 | struct restart_info restart; | |
147 | }; | |
148 | ||
149 | #define OPTION_MINRESTART 2000 | |
150 | #define OPTION_MAXRESTART 2001 | |
151 | #define OPTION_DRY 2002 | |
152 | ||
153 | static const struct option longopts[] = { | |
154 | {"daemon", no_argument, NULL, 'd'}, | |
155 | {"statedir", required_argument, NULL, 'S'}, | |
156 | {"loglevel", required_argument, NULL, 'l'}, | |
157 | {"interval", required_argument, NULL, 'i'}, | |
158 | {"timeout", required_argument, NULL, 't'}, | |
159 | {"restart-timeout", required_argument, NULL, 'T'}, | |
160 | {"restart", required_argument, NULL, 'r'}, | |
161 | {"start-command", required_argument, NULL, 's'}, | |
162 | {"kill-command", required_argument, NULL, 'k'}, | |
163 | {"dry", no_argument, NULL, OPTION_DRY}, | |
164 | {"min-restart-interval", required_argument, NULL, OPTION_MINRESTART}, | |
165 | {"max-restart-interval", required_argument, NULL, OPTION_MAXRESTART}, | |
166 | {"pid-file", required_argument, NULL, 'p'}, | |
167 | {"blank-string", required_argument, NULL, 'b'}, | |
168 | {"help", no_argument, NULL, 'h'}, | |
169 | {"version", no_argument, NULL, 'v'}, | |
170 | {NULL, 0, NULL, 0}}; | |
171 | ||
172 | static int try_connect(struct daemon *dmn); | |
173 | static int wakeup_send_echo(struct thread *t_wakeup); | |
174 | static void try_restart(struct daemon *dmn); | |
175 | static void phase_check(void); | |
176 | ||
177 | static const char *progname; | |
178 | static void printhelp(FILE *target) | |
179 | { | |
180 | fprintf(target, | |
181 | "Usage : %s [OPTION...] <daemon name> ...\n\n\ | |
182 | Watchdog program to monitor status of frr daemons and try to restart\n\ | |
183 | them if they are down or unresponsive. It determines whether a daemon is\n\ | |
184 | up based on whether it can connect to the daemon's vty unix stream socket.\n\ | |
185 | It then repeatedly sends echo commands over that socket to determine whether\n\ | |
186 | the daemon is responsive. If the daemon crashes, we will receive an EOF\n\ | |
187 | on the socket connection and know immediately that the daemon is down.\n\n\ | |
188 | The daemons to be monitored should be listed on the command line.\n\n\ | |
189 | In order to avoid attempting to restart the daemons in a fast loop,\n\ | |
190 | the -m and -M options allow you to control the minimum delay between\n\ | |
191 | restart commands. The minimum restart delay is recalculated each time\n\ | |
192 | a restart is attempted: if the time since the last restart attempt exceeds\n\ | |
193 | twice the -M value, then the restart delay is set to the -m value.\n\ | |
194 | Otherwise, the interval is doubled (but capped at the -M value).\n\n", | |
195 | progname); | |
196 | ||
197 | fprintf(target, | |
198 | "Options:\n\ | |
199 | -d, --daemon Run in daemon mode. In this mode, error messages are sent\n\ | |
200 | to syslog instead of stdout.\n\ | |
201 | -S, --statedir Set the vty socket directory (default is %s)\n\ | |
202 | -l, --loglevel Set the logging level (default is %d).\n\ | |
203 | The value should range from %d (LOG_EMERG) to %d (LOG_DEBUG),\n\ | |
204 | but it can be set higher than %d if extra-verbose debugging\n\ | |
205 | messages are desired.\n\ | |
206 | --min-restart-interval\n\ | |
207 | Set the minimum seconds to wait between invocations of daemon\n\ | |
208 | restart commands (default is %d).\n\ | |
209 | --max-restart-interval\n\ | |
210 | Set the maximum seconds to wait between invocations of daemon\n\ | |
211 | restart commands (default is %d).\n\ | |
212 | -i, --interval Set the status polling interval in seconds (default is %d)\n\ | |
213 | -t, --timeout Set the unresponsiveness timeout in seconds (default is %d)\n\ | |
214 | -T, --restart-timeout\n\ | |
215 | Set the restart (kill) timeout in seconds (default is %d).\n\ | |
216 | If any background jobs are still running after this much\n\ | |
217 | time has elapsed, they will be killed.\n\ | |
218 | -r, --restart Supply a Bourne shell command to use to restart a single\n\ | |
219 | daemon. The command string should include '%%s' where the\n\ | |
220 | name of the daemon should be substituted.\n\ | |
221 | -s, --start-command\n\ | |
222 | Supply a Bourne shell to command to use to start a single\n\ | |
223 | daemon. The command string should include '%%s' where the\n\ | |
224 | name of the daemon should be substituted.\n\ | |
225 | -k, --kill-command\n\ | |
226 | Supply a Bourne shell to command to use to stop a single\n\ | |
227 | daemon. The command string should include '%%s' where the\n\ | |
228 | name of the daemon should be substituted.\n\ | |
229 | --dry Do not start or restart anything, just log.\n\ | |
230 | -p, --pid-file Set process identifier file name\n\ | |
231 | (default is %s).\n\ | |
232 | -b, --blank-string\n\ | |
233 | When the supplied argument string is found in any of the\n\ | |
234 | various shell command arguments (-r, -s, or -k), replace\n\ | |
235 | it with a space. This is an ugly hack to circumvent problems\n\ | |
236 | passing command-line arguments with embedded spaces.\n\ | |
237 | -v, --version Print program version\n\ | |
238 | -h, --help Display this help and exit\n", | |
239 | frr_vtydir, DEFAULT_LOGLEVEL, LOG_EMERG, LOG_DEBUG, LOG_DEBUG, | |
240 | DEFAULT_MIN_RESTART, DEFAULT_MAX_RESTART, DEFAULT_PERIOD, | |
241 | DEFAULT_TIMEOUT, DEFAULT_RESTART_TIMEOUT, pidfile_default); | |
242 | } | |
243 | ||
244 | static pid_t run_background(char *shell_cmd) | |
245 | { | |
246 | pid_t child; | |
247 | ||
248 | switch (child = fork()) { | |
249 | case -1: | |
250 | zlog_err("fork failed, cannot run command [%s]: %s", shell_cmd, | |
251 | safe_strerror(errno)); | |
252 | return -1; | |
253 | case 0: | |
254 | /* Child process. */ | |
255 | /* Use separate process group so child processes can be killed | |
256 | * easily. */ | |
257 | if (setpgid(0, 0) < 0) | |
258 | zlog_warn("warning: setpgid(0,0) failed: %s", | |
259 | safe_strerror(errno)); | |
260 | { | |
261 | char shell[] = "sh"; | |
262 | char dashc[] = "-c"; | |
263 | char *const argv[4] = {shell, dashc, shell_cmd, NULL}; | |
264 | execv("/bin/sh", argv); | |
265 | zlog_err("execv(/bin/sh -c '%s') failed: %s", shell_cmd, | |
266 | safe_strerror(errno)); | |
267 | _exit(127); | |
268 | } | |
269 | default: | |
270 | /* Parent process: we will reap the child later. */ | |
271 | zlog_err("Forked background command [pid %d]: %s", (int)child, | |
272 | shell_cmd); | |
273 | return child; | |
274 | } | |
275 | } | |
276 | ||
277 | static struct timeval *time_elapsed(struct timeval *result, | |
278 | const struct timeval *start_time) | |
279 | { | |
280 | gettimeofday(result, NULL); | |
281 | result->tv_sec -= start_time->tv_sec; | |
282 | result->tv_usec -= start_time->tv_usec; | |
283 | while (result->tv_usec < 0) { | |
284 | result->tv_usec += 1000000L; | |
285 | result->tv_sec--; | |
286 | } | |
287 | return result; | |
288 | } | |
289 | ||
290 | static int restart_kill(struct thread *t_kill) | |
291 | { | |
292 | struct restart_info *restart = THREAD_ARG(t_kill); | |
293 | struct timeval delay; | |
294 | ||
295 | time_elapsed(&delay, &restart->time); | |
296 | zlog_warn( | |
297 | "Warning: %s %s child process %d still running after " | |
298 | "%ld seconds, sending signal %d", | |
299 | restart->what, restart->name, (int)restart->pid, | |
300 | (long)delay.tv_sec, (restart->kills ? SIGKILL : SIGTERM)); | |
301 | kill(-restart->pid, (restart->kills ? SIGKILL : SIGTERM)); | |
302 | restart->kills++; | |
303 | restart->t_kill = NULL; | |
304 | thread_add_timer(master, restart_kill, restart, gs.restart_timeout, | |
305 | &restart->t_kill); | |
306 | return 0; | |
307 | } | |
308 | ||
309 | static struct restart_info *find_child(pid_t child) | |
310 | { | |
311 | struct daemon *dmn; | |
312 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { | |
313 | if (dmn->restart.pid == child) | |
314 | return &dmn->restart; | |
315 | } | |
316 | return NULL; | |
317 | } | |
318 | ||
319 | static void sigchild(void) | |
320 | { | |
321 | pid_t child; | |
322 | int status; | |
323 | const char *name; | |
324 | const char *what; | |
325 | struct restart_info *restart; | |
326 | ||
327 | switch (child = waitpid(-1, &status, WNOHANG)) { | |
328 | case -1: | |
329 | zlog_err("waitpid failed: %s", safe_strerror(errno)); | |
330 | return; | |
331 | case 0: | |
332 | zlog_warn("SIGCHLD received, but waitpid did not reap a child"); | |
333 | return; | |
334 | } | |
335 | ||
336 | if (child == integrated_write_pid) { | |
337 | integrated_write_sigchld(status); | |
338 | return; | |
339 | } | |
340 | ||
341 | if ((restart = find_child(child)) != NULL) { | |
342 | name = restart->name; | |
343 | what = restart->what; | |
344 | restart->pid = 0; | |
345 | gs.numpids--; | |
346 | thread_cancel(restart->t_kill); | |
347 | restart->t_kill = NULL; | |
348 | /* Update restart time to reflect the time the command | |
349 | * completed. */ | |
350 | gettimeofday(&restart->time, NULL); | |
351 | } else { | |
352 | zlog_err( | |
353 | "waitpid returned status for an unknown child process %d", | |
354 | (int)child); | |
355 | name = "(unknown)"; | |
356 | what = "background"; | |
357 | } | |
358 | if (WIFSTOPPED(status)) | |
359 | zlog_warn("warning: %s %s process %d is stopped", what, name, | |
360 | (int)child); | |
361 | else if (WIFSIGNALED(status)) | |
362 | zlog_warn("%s %s process %d terminated due to signal %d", what, | |
363 | name, (int)child, WTERMSIG(status)); | |
364 | else if (WIFEXITED(status)) { | |
365 | if (WEXITSTATUS(status) != 0) | |
366 | zlog_warn( | |
367 | "%s %s process %d exited with non-zero status %d", | |
368 | what, name, (int)child, WEXITSTATUS(status)); | |
369 | else | |
370 | zlog_debug("%s %s process %d exited normally", what, | |
371 | name, (int)child); | |
372 | } else | |
373 | zlog_err("cannot interpret %s %s process %d wait status 0x%x", | |
374 | what, name, (int)child, status); | |
375 | phase_check(); | |
376 | } | |
377 | ||
378 | static int run_job(struct restart_info *restart, const char *cmdtype, | |
379 | const char *command, int force, int update_interval) | |
380 | { | |
381 | struct timeval delay; | |
382 | ||
383 | if (gs.loglevel > LOG_DEBUG + 1) | |
384 | zlog_debug("attempting to %s %s", cmdtype, restart->name); | |
385 | ||
386 | if (restart->pid) { | |
387 | if (gs.loglevel > LOG_DEBUG + 1) | |
388 | zlog_debug( | |
389 | "cannot %s %s, previous pid %d still running", | |
390 | cmdtype, restart->name, (int)restart->pid); | |
391 | return -1; | |
392 | } | |
393 | ||
394 | /* Note: time_elapsed test must come before the force test, since we | |
395 | need | |
396 | to make sure that delay is initialized for use below in updating the | |
397 | restart interval. */ | |
398 | if ((time_elapsed(&delay, &restart->time)->tv_sec < restart->interval) | |
399 | && !force) { | |
400 | if (gs.loglevel > LOG_DEBUG + 1) | |
401 | zlog_debug( | |
402 | "postponing %s %s: " | |
403 | "elapsed time %ld < retry interval %ld", | |
404 | cmdtype, restart->name, (long)delay.tv_sec, | |
405 | restart->interval); | |
406 | return -1; | |
407 | } | |
408 | ||
409 | gettimeofday(&restart->time, NULL); | |
410 | restart->kills = 0; | |
411 | { | |
412 | char cmd[strlen(command) + strlen(restart->name) + 1]; | |
413 | snprintf(cmd, sizeof(cmd), command, restart->name); | |
414 | if ((restart->pid = run_background(cmd)) > 0) { | |
415 | restart->t_kill = NULL; | |
416 | thread_add_timer(master, restart_kill, restart, | |
417 | gs.restart_timeout, &restart->t_kill); | |
418 | restart->what = cmdtype; | |
419 | gs.numpids++; | |
420 | } else | |
421 | restart->pid = 0; | |
422 | } | |
423 | ||
424 | /* Calculate the new restart interval. */ | |
425 | if (update_interval) { | |
426 | if (delay.tv_sec > 2 * gs.max_restart_interval) | |
427 | restart->interval = gs.min_restart_interval; | |
428 | else if ((restart->interval *= 2) > gs.max_restart_interval) | |
429 | restart->interval = gs.max_restart_interval; | |
430 | if (gs.loglevel > LOG_DEBUG + 1) | |
431 | zlog_debug("restart %s interval is now %ld", | |
432 | restart->name, restart->interval); | |
433 | } | |
434 | return restart->pid; | |
435 | } | |
436 | ||
437 | #define SET_READ_HANDLER(DMN) \ | |
438 | do { \ | |
439 | (DMN)->t_read = NULL; \ | |
440 | thread_add_read(master, handle_read, (DMN), (DMN)->fd, \ | |
441 | &(DMN)->t_read); \ | |
442 | } while (0); | |
443 | ||
444 | #define SET_WAKEUP_DOWN(DMN) \ | |
445 | do { \ | |
446 | (DMN)->t_wakeup = NULL; \ | |
447 | thread_add_timer_msec(master, wakeup_down, (DMN), \ | |
448 | FUZZY(gs.period), &(DMN)->t_wakeup); \ | |
449 | } while (0); | |
450 | ||
451 | #define SET_WAKEUP_UNRESPONSIVE(DMN) \ | |
452 | do { \ | |
453 | (DMN)->t_wakeup = NULL; \ | |
454 | thread_add_timer_msec(master, wakeup_unresponsive, (DMN), \ | |
455 | FUZZY(gs.period), &(DMN)->t_wakeup); \ | |
456 | } while (0); | |
457 | ||
458 | #define SET_WAKEUP_ECHO(DMN) \ | |
459 | do { \ | |
460 | (DMN)->t_wakeup = NULL; \ | |
461 | thread_add_timer_msec(master, wakeup_send_echo, (DMN), \ | |
462 | FUZZY(gs.period), &(DMN)->t_wakeup); \ | |
463 | } while (0); | |
464 | ||
465 | static int wakeup_down(struct thread *t_wakeup) | |
466 | { | |
467 | struct daemon *dmn = THREAD_ARG(t_wakeup); | |
468 | ||
469 | dmn->t_wakeup = NULL; | |
470 | if (try_connect(dmn) < 0) | |
471 | SET_WAKEUP_DOWN(dmn); | |
472 | if ((dmn->connect_tries > 1) && (dmn->state != DAEMON_UP)) | |
473 | try_restart(dmn); | |
474 | return 0; | |
475 | } | |
476 | ||
477 | static int wakeup_init(struct thread *t_wakeup) | |
478 | { | |
479 | struct daemon *dmn = THREAD_ARG(t_wakeup); | |
480 | ||
481 | dmn->t_wakeup = NULL; | |
482 | if (try_connect(dmn) < 0) { | |
483 | SET_WAKEUP_DOWN(dmn); | |
484 | zlog_err("%s state -> down : initial connection attempt failed", | |
485 | dmn->name); | |
486 | dmn->state = DAEMON_DOWN; | |
487 | } | |
488 | return 0; | |
489 | } | |
490 | ||
491 | static void daemon_down(struct daemon *dmn, const char *why) | |
492 | { | |
493 | if (IS_UP(dmn) || (dmn->state == DAEMON_INIT)) | |
494 | zlog_err("%s state -> down : %s", dmn->name, why); | |
495 | else if (gs.loglevel > LOG_DEBUG) | |
496 | zlog_debug("%s still down : %s", dmn->name, why); | |
497 | if (IS_UP(dmn)) | |
498 | gs.numdown++; | |
499 | dmn->state = DAEMON_DOWN; | |
500 | if (dmn->fd >= 0) { | |
501 | close(dmn->fd); | |
502 | dmn->fd = -1; | |
503 | } | |
504 | THREAD_OFF(dmn->t_read); | |
505 | THREAD_OFF(dmn->t_write); | |
506 | THREAD_OFF(dmn->t_wakeup); | |
507 | if (try_connect(dmn) < 0) | |
508 | SET_WAKEUP_DOWN(dmn); | |
509 | phase_check(); | |
510 | } | |
511 | ||
512 | static int handle_read(struct thread *t_read) | |
513 | { | |
514 | struct daemon *dmn = THREAD_ARG(t_read); | |
515 | static const char resp[sizeof(PING_TOKEN) + 4] = PING_TOKEN "\n"; | |
516 | char buf[sizeof(resp) + 100]; | |
517 | ssize_t rc; | |
518 | struct timeval delay; | |
519 | ||
520 | dmn->t_read = NULL; | |
521 | if ((rc = read(dmn->fd, buf, sizeof(buf))) < 0) { | |
522 | char why[100]; | |
523 | ||
524 | if (ERRNO_IO_RETRY(errno)) { | |
525 | /* Pretend it never happened. */ | |
526 | SET_READ_HANDLER(dmn); | |
527 | return 0; | |
528 | } | |
529 | snprintf(why, sizeof(why), "unexpected read error: %s", | |
530 | safe_strerror(errno)); | |
531 | daemon_down(dmn, why); | |
532 | return 0; | |
533 | } | |
534 | if (rc == 0) { | |
535 | daemon_down(dmn, "read returned EOF"); | |
536 | return 0; | |
537 | } | |
538 | if (!dmn->echo_sent.tv_sec) { | |
539 | char why[sizeof(buf) + 100]; | |
540 | snprintf(why, sizeof(why), | |
541 | "unexpected read returns %d bytes: %.*s", (int)rc, | |
542 | (int)rc, buf); | |
543 | daemon_down(dmn, why); | |
544 | return 0; | |
545 | } | |
546 | ||
547 | /* We are expecting an echo response: is there any chance that the | |
548 | response would not be returned entirely in the first read? That | |
549 | seems inconceivable... */ | |
550 | if ((rc != sizeof(resp)) || memcmp(buf, resp, sizeof(resp))) { | |
551 | char why[100 + sizeof(buf)]; | |
552 | snprintf(why, sizeof(why), | |
553 | "read returned bad echo response of %d bytes " | |
554 | "(expecting %u): %.*s", | |
555 | (int)rc, (u_int)sizeof(resp), (int)rc, buf); | |
556 | daemon_down(dmn, why); | |
557 | return 0; | |
558 | } | |
559 | ||
560 | time_elapsed(&delay, &dmn->echo_sent); | |
561 | dmn->echo_sent.tv_sec = 0; | |
562 | if (dmn->state == DAEMON_UNRESPONSIVE) { | |
563 | if (delay.tv_sec < gs.timeout) { | |
564 | dmn->state = DAEMON_UP; | |
565 | zlog_warn( | |
566 | "%s state -> up : echo response received after %ld.%06ld " | |
567 | "seconds", | |
568 | dmn->name, (long)delay.tv_sec, | |
569 | (long)delay.tv_usec); | |
570 | } else | |
571 | zlog_warn( | |
572 | "%s: slow echo response finally received after %ld.%06ld " | |
573 | "seconds", | |
574 | dmn->name, (long)delay.tv_sec, | |
575 | (long)delay.tv_usec); | |
576 | } else if (gs.loglevel > LOG_DEBUG + 1) | |
577 | zlog_debug("%s: echo response received after %ld.%06ld seconds", | |
578 | dmn->name, (long)delay.tv_sec, (long)delay.tv_usec); | |
579 | ||
580 | SET_READ_HANDLER(dmn); | |
581 | if (dmn->t_wakeup) | |
582 | thread_cancel(dmn->t_wakeup); | |
583 | SET_WAKEUP_ECHO(dmn); | |
584 | ||
585 | return 0; | |
586 | } | |
587 | ||
588 | /* | |
589 | * Wait till we notice that all daemons are ready before | |
590 | * we send we are ready to systemd | |
591 | */ | |
592 | static void daemon_send_ready(void) | |
593 | { | |
594 | static int sent = 0; | |
595 | if (!sent && gs.numdown == 0) { | |
596 | FILE *fp; | |
597 | ||
598 | fp = fopen(DAEMON_VTY_DIR "/watchfrr.started", "w"); | |
599 | if (fp) | |
600 | fclose(fp); | |
601 | #if defined HAVE_SYSTEMD | |
602 | zlog_notice( | |
603 | "Watchfrr: Notifying Systemd we are up and running"); | |
604 | systemd_send_started(master, 0); | |
605 | #endif | |
606 | sent = 1; | |
607 | } | |
608 | } | |
609 | ||
610 | static void daemon_up(struct daemon *dmn, const char *why) | |
611 | { | |
612 | dmn->state = DAEMON_UP; | |
613 | gs.numdown--; | |
614 | dmn->connect_tries = 0; | |
615 | zlog_notice("%s state -> up : %s", dmn->name, why); | |
616 | daemon_send_ready(); | |
617 | SET_WAKEUP_ECHO(dmn); | |
618 | phase_check(); | |
619 | } | |
620 | ||
621 | static int check_connect(struct thread *t_write) | |
622 | { | |
623 | struct daemon *dmn = THREAD_ARG(t_write); | |
624 | int sockerr; | |
625 | socklen_t reslen = sizeof(sockerr); | |
626 | ||
627 | dmn->t_write = NULL; | |
628 | if (getsockopt(dmn->fd, SOL_SOCKET, SO_ERROR, (char *)&sockerr, &reslen) | |
629 | < 0) { | |
630 | zlog_warn("%s: check_connect: getsockopt failed: %s", dmn->name, | |
631 | safe_strerror(errno)); | |
632 | daemon_down(dmn, | |
633 | "getsockopt failed checking connection success"); | |
634 | return 0; | |
635 | } | |
636 | if ((reslen == sizeof(sockerr)) && sockerr) { | |
637 | char why[100]; | |
638 | snprintf( | |
639 | why, sizeof(why), | |
640 | "getsockopt reports that connection attempt failed: %s", | |
641 | safe_strerror(sockerr)); | |
642 | daemon_down(dmn, why); | |
643 | return 0; | |
644 | } | |
645 | ||
646 | daemon_up(dmn, "delayed connect succeeded"); | |
647 | return 0; | |
648 | } | |
649 | ||
650 | static int wakeup_connect_hanging(struct thread *t_wakeup) | |
651 | { | |
652 | struct daemon *dmn = THREAD_ARG(t_wakeup); | |
653 | char why[100]; | |
654 | ||
655 | dmn->t_wakeup = NULL; | |
656 | snprintf(why, sizeof(why), | |
657 | "connection attempt timed out after %ld seconds", gs.timeout); | |
658 | daemon_down(dmn, why); | |
659 | return 0; | |
660 | } | |
661 | ||
662 | /* Making connection to protocol daemon. */ | |
663 | static int try_connect(struct daemon *dmn) | |
664 | { | |
665 | int sock; | |
666 | struct sockaddr_un addr; | |
667 | socklen_t len; | |
668 | ||
669 | if (gs.loglevel > LOG_DEBUG + 1) | |
670 | zlog_debug("%s: attempting to connect", dmn->name); | |
671 | dmn->connect_tries++; | |
672 | ||
673 | memset(&addr, 0, sizeof(struct sockaddr_un)); | |
674 | addr.sun_family = AF_UNIX; | |
675 | snprintf(addr.sun_path, sizeof(addr.sun_path), "%s/%s.vty", gs.vtydir, | |
676 | dmn->name); | |
677 | #ifdef HAVE_STRUCT_SOCKADDR_UN_SUN_LEN | |
678 | len = addr.sun_len = SUN_LEN(&addr); | |
679 | #else | |
680 | len = sizeof(addr.sun_family) + strlen(addr.sun_path); | |
681 | #endif /* HAVE_STRUCT_SOCKADDR_UN_SUN_LEN */ | |
682 | ||
683 | /* Quick check to see if we might succeed before we go to the trouble | |
684 | of creating a socket. */ | |
685 | if (access(addr.sun_path, W_OK) < 0) { | |
686 | if (errno != ENOENT) | |
687 | zlog_err("%s: access to socket %s denied: %s", | |
688 | dmn->name, addr.sun_path, | |
689 | safe_strerror(errno)); | |
690 | return -1; | |
691 | } | |
692 | ||
693 | if ((sock = socket(AF_UNIX, SOCK_STREAM, 0)) < 0) { | |
694 | zlog_err("%s(%s): cannot make socket: %s", __func__, | |
695 | addr.sun_path, safe_strerror(errno)); | |
696 | return -1; | |
697 | } | |
698 | ||
699 | if (set_nonblocking(sock) < 0 || set_cloexec(sock) < 0) { | |
700 | zlog_err("%s(%s): set_nonblocking/cloexec(%d) failed", __func__, | |
701 | addr.sun_path, sock); | |
702 | close(sock); | |
703 | return -1; | |
704 | } | |
705 | ||
706 | if (connect(sock, (struct sockaddr *)&addr, len) < 0) { | |
707 | if ((errno != EINPROGRESS) && (errno != EWOULDBLOCK)) { | |
708 | if (gs.loglevel > LOG_DEBUG) | |
709 | zlog_debug("%s(%s): connect failed: %s", | |
710 | __func__, addr.sun_path, | |
711 | safe_strerror(errno)); | |
712 | close(sock); | |
713 | return -1; | |
714 | } | |
715 | if (gs.loglevel > LOG_DEBUG) | |
716 | zlog_debug("%s: connection in progress", dmn->name); | |
717 | dmn->state = DAEMON_CONNECTING; | |
718 | dmn->fd = sock; | |
719 | dmn->t_write = NULL; | |
720 | thread_add_write(master, check_connect, dmn, dmn->fd, | |
721 | &dmn->t_write); | |
722 | dmn->t_wakeup = NULL; | |
723 | thread_add_timer(master, wakeup_connect_hanging, dmn, | |
724 | gs.timeout, &dmn->t_wakeup); | |
725 | SET_READ_HANDLER(dmn); | |
726 | return 0; | |
727 | } | |
728 | ||
729 | dmn->fd = sock; | |
730 | SET_READ_HANDLER(dmn); | |
731 | daemon_up(dmn, "connect succeeded"); | |
732 | return 1; | |
733 | } | |
734 | ||
735 | static int phase_hanging(struct thread *t_hanging) | |
736 | { | |
737 | gs.t_phase_hanging = NULL; | |
738 | zlog_err("Phase [%s] hanging for %ld seconds, aborting phased restart", | |
739 | phase_str[gs.phase], PHASE_TIMEOUT); | |
740 | gs.phase = PHASE_NONE; | |
741 | return 0; | |
742 | } | |
743 | ||
744 | static void set_phase(restart_phase_t new_phase) | |
745 | { | |
746 | gs.phase = new_phase; | |
747 | if (gs.t_phase_hanging) | |
748 | thread_cancel(gs.t_phase_hanging); | |
749 | gs.t_phase_hanging = NULL; | |
750 | thread_add_timer(master, phase_hanging, NULL, PHASE_TIMEOUT, | |
751 | &gs.t_phase_hanging); | |
752 | } | |
753 | ||
754 | static void phase_check(void) | |
755 | { | |
756 | switch (gs.phase) { | |
757 | case PHASE_NONE: | |
758 | break; | |
759 | case PHASE_STOPS_PENDING: | |
760 | if (gs.numpids) | |
761 | break; | |
762 | zlog_info( | |
763 | "Phased restart: all routing daemon stop jobs have completed."); | |
764 | set_phase(PHASE_WAITING_DOWN); | |
765 | ||
766 | /*FALLTHRU*/ | |
767 | case PHASE_WAITING_DOWN: | |
768 | if (gs.numdown + IS_UP(gs.special) < gs.numdaemons) | |
769 | break; | |
770 | zlog_info("Phased restart: all routing daemons now down."); | |
771 | run_job(&gs.special->restart, "restart", gs.restart_command, 1, | |
772 | 1); | |
773 | set_phase(PHASE_ZEBRA_RESTART_PENDING); | |
774 | ||
775 | /*FALLTHRU*/ | |
776 | case PHASE_ZEBRA_RESTART_PENDING: | |
777 | if (gs.special->restart.pid) | |
778 | break; | |
779 | zlog_info("Phased restart: %s restart job completed.", | |
780 | gs.special->name); | |
781 | set_phase(PHASE_WAITING_ZEBRA_UP); | |
782 | ||
783 | /*FALLTHRU*/ | |
784 | case PHASE_WAITING_ZEBRA_UP: | |
785 | if (!IS_UP(gs.special)) | |
786 | break; | |
787 | zlog_info("Phased restart: %s is now up.", gs.special->name); | |
788 | { | |
789 | struct daemon *dmn; | |
790 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { | |
791 | if (dmn != gs.special) | |
792 | run_job(&dmn->restart, "start", | |
793 | gs.start_command, 1, 0); | |
794 | } | |
795 | } | |
796 | gs.phase = PHASE_NONE; | |
797 | THREAD_OFF(gs.t_phase_hanging); | |
798 | zlog_notice("Phased global restart has completed."); | |
799 | break; | |
800 | } | |
801 | } | |
802 | ||
803 | static void try_restart(struct daemon *dmn) | |
804 | { | |
805 | if (watch_only) | |
806 | return; | |
807 | ||
808 | if (dmn != gs.special) { | |
809 | if ((gs.special->state == DAEMON_UP) | |
810 | && (gs.phase == PHASE_NONE)) | |
811 | run_job(&dmn->restart, "restart", gs.restart_command, 0, | |
812 | 1); | |
813 | else | |
814 | zlog_debug( | |
815 | "%s: postponing restart attempt because master %s daemon " | |
816 | "not up [%s], or phased restart in progress", | |
817 | dmn->name, gs.special->name, | |
818 | state_str[gs.special->state]); | |
819 | return; | |
820 | } | |
821 | ||
822 | if ((gs.phase != PHASE_NONE) || gs.numpids) { | |
823 | if (gs.loglevel > LOG_DEBUG + 1) | |
824 | zlog_debug( | |
825 | "postponing phased global restart: restart already in " | |
826 | "progress [%s], or outstanding child processes [%d]", | |
827 | phase_str[gs.phase], gs.numpids); | |
828 | return; | |
829 | } | |
830 | /* Is it too soon for a restart? */ | |
831 | { | |
832 | struct timeval delay; | |
833 | if (time_elapsed(&delay, &gs.special->restart.time)->tv_sec | |
834 | < gs.special->restart.interval) { | |
835 | if (gs.loglevel > LOG_DEBUG + 1) | |
836 | zlog_debug( | |
837 | "postponing phased global restart: " | |
838 | "elapsed time %ld < retry interval %ld", | |
839 | (long)delay.tv_sec, | |
840 | gs.special->restart.interval); | |
841 | return; | |
842 | } | |
843 | } | |
844 | run_job(&gs.restart, "restart", gs.restart_command, 0, 1); | |
845 | } | |
846 | ||
847 | static int wakeup_unresponsive(struct thread *t_wakeup) | |
848 | { | |
849 | struct daemon *dmn = THREAD_ARG(t_wakeup); | |
850 | ||
851 | dmn->t_wakeup = NULL; | |
852 | if (dmn->state != DAEMON_UNRESPONSIVE) | |
853 | zlog_err( | |
854 | "%s: no longer unresponsive (now %s), " | |
855 | "wakeup should have been cancelled!", | |
856 | dmn->name, state_str[dmn->state]); | |
857 | else { | |
858 | SET_WAKEUP_UNRESPONSIVE(dmn); | |
859 | try_restart(dmn); | |
860 | } | |
861 | return 0; | |
862 | } | |
863 | ||
864 | static int wakeup_no_answer(struct thread *t_wakeup) | |
865 | { | |
866 | struct daemon *dmn = THREAD_ARG(t_wakeup); | |
867 | ||
868 | dmn->t_wakeup = NULL; | |
869 | dmn->state = DAEMON_UNRESPONSIVE; | |
870 | zlog_err( | |
871 | "%s state -> unresponsive : no response yet to ping " | |
872 | "sent %ld seconds ago", | |
873 | dmn->name, gs.timeout); | |
874 | SET_WAKEUP_UNRESPONSIVE(dmn); | |
875 | try_restart(dmn); | |
876 | return 0; | |
877 | } | |
878 | ||
879 | static int wakeup_send_echo(struct thread *t_wakeup) | |
880 | { | |
881 | static const char echocmd[] = "echo " PING_TOKEN; | |
882 | ssize_t rc; | |
883 | struct daemon *dmn = THREAD_ARG(t_wakeup); | |
884 | ||
885 | dmn->t_wakeup = NULL; | |
886 | if (((rc = write(dmn->fd, echocmd, sizeof(echocmd))) < 0) | |
887 | || ((size_t)rc != sizeof(echocmd))) { | |
888 | char why[100 + sizeof(echocmd)]; | |
889 | snprintf(why, sizeof(why), | |
890 | "write '%s' returned %d instead of %u", echocmd, | |
891 | (int)rc, (u_int)sizeof(echocmd)); | |
892 | daemon_down(dmn, why); | |
893 | } else { | |
894 | gettimeofday(&dmn->echo_sent, NULL); | |
895 | dmn->t_wakeup = NULL; | |
896 | thread_add_timer(master, wakeup_no_answer, dmn, gs.timeout, | |
897 | &dmn->t_wakeup); | |
898 | } | |
899 | return 0; | |
900 | } | |
901 | ||
902 | static void sigint(void) | |
903 | { | |
904 | zlog_notice("Terminating on signal"); | |
905 | systemd_send_stopping(); | |
906 | exit(0); | |
907 | } | |
908 | ||
909 | static int valid_command(const char *cmd) | |
910 | { | |
911 | char *p; | |
912 | ||
913 | return ((p = strchr(cmd, '%')) != NULL) && (*(p + 1) == 's') | |
914 | && !strchr(p + 1, '%'); | |
915 | } | |
916 | ||
917 | /* This is an ugly hack to circumvent problems with passing command-line | |
918 | arguments that contain spaces. The fix is to use a configuration file. */ | |
919 | static char *translate_blanks(const char *cmd, const char *blankstr) | |
920 | { | |
921 | char *res; | |
922 | char *p; | |
923 | size_t bslen = strlen(blankstr); | |
924 | ||
925 | if (!(res = strdup(cmd))) { | |
926 | perror("strdup"); | |
927 | exit(1); | |
928 | } | |
929 | while ((p = strstr(res, blankstr)) != NULL) { | |
930 | *p = ' '; | |
931 | if (bslen != 1) | |
932 | memmove(p + 1, p + bslen, strlen(p + bslen) + 1); | |
933 | } | |
934 | return res; | |
935 | } | |
936 | ||
937 | struct zebra_privs_t watchfrr_privs = { | |
938 | #ifdef VTY_GROUP | |
939 | .vty_group = VTY_GROUP, | |
940 | #endif | |
941 | }; | |
942 | ||
943 | static struct quagga_signal_t watchfrr_signals[] = { | |
944 | { | |
945 | .signal = SIGINT, | |
946 | .handler = sigint, | |
947 | }, | |
948 | { | |
949 | .signal = SIGTERM, | |
950 | .handler = sigint, | |
951 | }, | |
952 | { | |
953 | .signal = SIGCHLD, | |
954 | .handler = sigchild, | |
955 | }, | |
956 | }; | |
957 | ||
958 | FRR_DAEMON_INFO(watchfrr, WATCHFRR, | |
959 | .flags = FRR_NO_PRIVSEP | FRR_NO_TCPVTY | FRR_LIMITED_CLI | |
960 | | FRR_NO_CFG_PID_DRY | FRR_NO_ZCLIENT, | |
961 | ||
962 | .printhelp = printhelp, | |
963 | .copyright = "Copyright 2004 Andrew J. Schorr", | |
964 | ||
965 | .signals = watchfrr_signals, | |
966 | .n_signals = array_size(watchfrr_signals), | |
967 | ||
968 | .privs = &watchfrr_privs, ) | |
969 | ||
970 | #define DEPRECATED_OPTIONS "aAezR:" | |
971 | ||
972 | int main(int argc, char **argv) | |
973 | { | |
974 | int opt; | |
975 | const char *pidfile = pidfile_default; | |
976 | const char *special = "zebra"; | |
977 | const char *blankstr = NULL; | |
978 | ||
979 | snprintf(pidfile_default, sizeof(pidfile_default), "%s/watchfrr.pid", | |
980 | frr_vtydir); | |
981 | ||
982 | frr_preinit(&watchfrr_di, argc, argv); | |
983 | progname = watchfrr_di.progname; | |
984 | ||
985 | frr_opt_add("b:dk:l:i:p:r:S:s:t:T:" DEPRECATED_OPTIONS, longopts, ""); | |
986 | ||
987 | gs.restart.name = "all"; | |
988 | while ((opt = frr_getopt(argc, argv, NULL)) != EOF) { | |
989 | if (opt && opt < 128 && strchr(DEPRECATED_OPTIONS, opt)) { | |
990 | fprintf(stderr, | |
991 | "The -%c option no longer exists.\n" | |
992 | "Please refer to the watchfrr(8) man page.\n", | |
993 | opt); | |
994 | exit(1); | |
995 | } | |
996 | ||
997 | switch (opt) { | |
998 | case 0: | |
999 | break; | |
1000 | case 'b': | |
1001 | blankstr = optarg; | |
1002 | break; | |
1003 | case OPTION_DRY: | |
1004 | watch_only = true; | |
1005 | break; | |
1006 | case 'k': | |
1007 | if (!valid_command(optarg)) { | |
1008 | fprintf(stderr, | |
1009 | "Invalid kill command, must contain '%%s': %s\n", | |
1010 | optarg); | |
1011 | frr_help_exit(1); | |
1012 | } | |
1013 | gs.stop_command = optarg; | |
1014 | break; | |
1015 | case 'l': { | |
1016 | char garbage[3]; | |
1017 | if ((sscanf(optarg, "%d%1s", &gs.loglevel, garbage) | |
1018 | != 1) | |
1019 | || (gs.loglevel < LOG_EMERG)) { | |
1020 | fprintf(stderr, | |
1021 | "Invalid loglevel argument: %s\n", | |
1022 | optarg); | |
1023 | frr_help_exit(1); | |
1024 | } | |
1025 | } break; | |
1026 | case OPTION_MINRESTART: { | |
1027 | char garbage[3]; | |
1028 | if ((sscanf(optarg, "%ld%1s", &gs.min_restart_interval, | |
1029 | garbage) | |
1030 | != 1) | |
1031 | || (gs.min_restart_interval < 0)) { | |
1032 | fprintf(stderr, | |
1033 | "Invalid min_restart_interval argument: %s\n", | |
1034 | optarg); | |
1035 | frr_help_exit(1); | |
1036 | } | |
1037 | } break; | |
1038 | case OPTION_MAXRESTART: { | |
1039 | char garbage[3]; | |
1040 | if ((sscanf(optarg, "%ld%1s", &gs.max_restart_interval, | |
1041 | garbage) | |
1042 | != 1) | |
1043 | || (gs.max_restart_interval < 0)) { | |
1044 | fprintf(stderr, | |
1045 | "Invalid max_restart_interval argument: %s\n", | |
1046 | optarg); | |
1047 | frr_help_exit(1); | |
1048 | } | |
1049 | } break; | |
1050 | case 'i': { | |
1051 | char garbage[3]; | |
1052 | int period; | |
1053 | if ((sscanf(optarg, "%d%1s", &period, garbage) != 1) | |
1054 | || (gs.period < 1)) { | |
1055 | fprintf(stderr, | |
1056 | "Invalid interval argument: %s\n", | |
1057 | optarg); | |
1058 | frr_help_exit(1); | |
1059 | } | |
1060 | gs.period = 1000 * period; | |
1061 | } break; | |
1062 | case 'p': | |
1063 | pidfile = optarg; | |
1064 | break; | |
1065 | case 'r': | |
1066 | if (!valid_command(optarg)) { | |
1067 | fprintf(stderr, | |
1068 | "Invalid restart command, must contain '%%s': %s\n", | |
1069 | optarg); | |
1070 | frr_help_exit(1); | |
1071 | } | |
1072 | gs.restart_command = optarg; | |
1073 | break; | |
1074 | case 's': | |
1075 | if (!valid_command(optarg)) { | |
1076 | fprintf(stderr, | |
1077 | "Invalid start command, must contain '%%s': %s\n", | |
1078 | optarg); | |
1079 | frr_help_exit(1); | |
1080 | } | |
1081 | gs.start_command = optarg; | |
1082 | break; | |
1083 | case 'S': | |
1084 | gs.vtydir = optarg; | |
1085 | break; | |
1086 | case 't': { | |
1087 | char garbage[3]; | |
1088 | if ((sscanf(optarg, "%ld%1s", &gs.timeout, garbage) | |
1089 | != 1) | |
1090 | || (gs.timeout < 1)) { | |
1091 | fprintf(stderr, | |
1092 | "Invalid timeout argument: %s\n", | |
1093 | optarg); | |
1094 | frr_help_exit(1); | |
1095 | } | |
1096 | } break; | |
1097 | case 'T': { | |
1098 | char garbage[3]; | |
1099 | if ((sscanf(optarg, "%ld%1s", &gs.restart_timeout, | |
1100 | garbage) | |
1101 | != 1) | |
1102 | || (gs.restart_timeout < 1)) { | |
1103 | fprintf(stderr, | |
1104 | "Invalid restart timeout argument: %s\n", | |
1105 | optarg); | |
1106 | frr_help_exit(1); | |
1107 | } | |
1108 | } break; | |
1109 | default: | |
1110 | fputs("Invalid option.\n", stderr); | |
1111 | frr_help_exit(1); | |
1112 | } | |
1113 | } | |
1114 | ||
1115 | if (watch_only | |
1116 | && (gs.start_command || gs.stop_command || gs.restart_command)) { | |
1117 | fputs("Options -r/-s/-k are not used when --dry is active.\n", | |
1118 | stderr); | |
1119 | } | |
1120 | if (!watch_only | |
1121 | && (!gs.restart_command || !gs.start_command || !gs.stop_command)) { | |
1122 | fprintf(stderr, | |
1123 | "Options -s (start), -k (kill), and -r (restart) are required.\n"); | |
1124 | frr_help_exit(1); | |
1125 | } | |
1126 | ||
1127 | if (blankstr) { | |
1128 | if (gs.restart_command) | |
1129 | gs.restart_command = | |
1130 | translate_blanks(gs.restart_command, blankstr); | |
1131 | if (gs.start_command) | |
1132 | gs.start_command = | |
1133 | translate_blanks(gs.start_command, blankstr); | |
1134 | if (gs.stop_command) | |
1135 | gs.stop_command = | |
1136 | translate_blanks(gs.stop_command, blankstr); | |
1137 | } | |
1138 | ||
1139 | gs.restart.interval = gs.min_restart_interval; | |
1140 | ||
1141 | master = frr_init(); | |
1142 | ||
1143 | zlog_set_level(ZLOG_DEST_MONITOR, ZLOG_DISABLED); | |
1144 | if (watchfrr_di.daemon_mode) { | |
1145 | zlog_set_level(ZLOG_DEST_SYSLOG, MIN(gs.loglevel, LOG_DEBUG)); | |
1146 | if (daemon(0, 0) < 0) { | |
1147 | fprintf(stderr, "Watchfrr daemon failed: %s", | |
1148 | strerror(errno)); | |
1149 | exit(1); | |
1150 | } | |
1151 | } else | |
1152 | zlog_set_level(ZLOG_DEST_STDOUT, MIN(gs.loglevel, LOG_DEBUG)); | |
1153 | ||
1154 | watchfrr_vty_init(); | |
1155 | ||
1156 | frr_vty_serv(); | |
1157 | ||
1158 | { | |
1159 | int i; | |
1160 | struct daemon *tail = NULL; | |
1161 | ||
1162 | for (i = optind; i < argc; i++) { | |
1163 | struct daemon *dmn; | |
1164 | ||
1165 | if (!(dmn = (struct daemon *)calloc(1, sizeof(*dmn)))) { | |
1166 | fprintf(stderr, "calloc(1,%u) failed: %s\n", | |
1167 | (u_int)sizeof(*dmn), | |
1168 | safe_strerror(errno)); | |
1169 | return 1; | |
1170 | } | |
1171 | dmn->name = dmn->restart.name = argv[i]; | |
1172 | dmn->state = DAEMON_INIT; | |
1173 | gs.numdaemons++; | |
1174 | gs.numdown++; | |
1175 | dmn->fd = -1; | |
1176 | dmn->t_wakeup = NULL; | |
1177 | thread_add_timer_msec(master, wakeup_init, dmn, | |
1178 | 100 + (random() % 900), | |
1179 | &dmn->t_wakeup); | |
1180 | dmn->restart.interval = gs.min_restart_interval; | |
1181 | if (tail) | |
1182 | tail->next = dmn; | |
1183 | else | |
1184 | gs.daemons = dmn; | |
1185 | tail = dmn; | |
1186 | ||
1187 | if (!strcmp(dmn->name, special)) | |
1188 | gs.special = dmn; | |
1189 | } | |
1190 | } | |
1191 | if (!gs.daemons) { | |
1192 | fputs("Must specify one or more daemons to monitor.\n", stderr); | |
1193 | frr_help_exit(1); | |
1194 | } | |
1195 | if (!watch_only && !gs.special) { | |
1196 | fprintf(stderr, "\"%s\" daemon must be in daemon list\n", | |
1197 | special); | |
1198 | frr_help_exit(1); | |
1199 | } | |
1200 | ||
1201 | /* Make sure we're not already running. */ | |
1202 | pid_output(pidfile); | |
1203 | ||
1204 | /* Announce which daemons are being monitored. */ | |
1205 | { | |
1206 | struct daemon *dmn; | |
1207 | size_t len = 0; | |
1208 | ||
1209 | for (dmn = gs.daemons; dmn; dmn = dmn->next) | |
1210 | len += strlen(dmn->name) + 1; | |
1211 | ||
1212 | { | |
1213 | char buf[len + 1]; | |
1214 | char *p = buf; | |
1215 | ||
1216 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { | |
1217 | if (p != buf) | |
1218 | *p++ = ' '; | |
1219 | strcpy(p, dmn->name); | |
1220 | p += strlen(p); | |
1221 | } | |
1222 | zlog_notice("%s %s watching [%s]%s", progname, | |
1223 | FRR_VERSION, buf, | |
1224 | watch_only ? ", monitor mode" : ""); | |
1225 | } | |
1226 | } | |
1227 | ||
1228 | { | |
1229 | struct thread thread; | |
1230 | ||
1231 | while (thread_fetch(master, &thread)) | |
1232 | thread_call(&thread); | |
1233 | } | |
1234 | ||
1235 | systemd_send_stopping(); | |
1236 | /* Not reached. */ | |
1237 | return 0; | |
1238 | } |