]>
Commit | Line | Data |
---|---|---|
8b886ca7 | 1 | /* |
896014f4 DL |
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 | |
8b886ca7 | 19 | */ |
20 | ||
a365534f | 21 | #include <zebra.h> |
8b886ca7 | 22 | #include <thread.h> |
23 | #include <log.h> | |
52e66296 | 24 | #include <network.h> |
8b886ca7 | 25 | #include <sigevent.h> |
a365534f | 26 | #include <lib/version.h> |
95c4aff2 | 27 | #include "command.h" |
87f44e2f | 28 | #include "memory_vty.h" |
4f04a76b | 29 | #include "libfrr.h" |
95c4aff2 | 30 | |
6f594023 | 31 | #include <getopt.h> |
a365534f | 32 | #include <sys/un.h> |
33 | #include <sys/wait.h> | |
837d16cc | 34 | #include <memory.h> |
651415bd | 35 | #include <systemd.h> |
8b886ca7 | 36 | |
9473e340 | 37 | #include "watchfrr.h" |
95c4aff2 | 38 | |
8b886ca7 | 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 10 | |
49 | #define DEFAULT_RESTART_TIMEOUT 20 | |
50 | #define DEFAULT_LOGLEVEL LOG_INFO | |
51 | #define DEFAULT_MIN_RESTART 60 | |
52 | #define DEFAULT_MAX_RESTART 600 | |
9473e340 DS |
53 | #ifdef PATH_WATCHFRR_PID |
54 | #define DEFAULT_PIDFILE PATH_WATCHFRR_PID | |
6028df52 | 55 | #else |
9473e340 | 56 | #define DEFAULT_PIDFILE STATEDIR "/watchfrr.pid" |
6028df52 | 57 | #endif |
16f6511e | 58 | #ifdef DAEMON_VTY_DIR |
59 | #define VTYDIR DAEMON_VTY_DIR | |
60 | #else | |
61 | #define VTYDIR STATEDIR | |
62 | #endif | |
8b886ca7 | 63 | |
64 | #define PING_TOKEN "PING" | |
65 | ||
55c72803 | 66 | /* Needs to be global, referenced somewhere inside libfrr. */ |
8b886ca7 | 67 | struct thread_master *master; |
68 | ||
f168b713 | 69 | static bool watch_only = false; |
8b886ca7 | 70 | |
a6810074 DL |
71 | typedef enum { |
72 | PHASE_NONE = 0, | |
73 | PHASE_STOPS_PENDING, | |
74 | PHASE_WAITING_DOWN, | |
75 | PHASE_ZEBRA_RESTART_PENDING, | |
76 | PHASE_WAITING_ZEBRA_UP | |
8b886ca7 | 77 | } restart_phase_t; |
78 | ||
a6810074 DL |
79 | static const char *phase_str[] = { |
80 | "None", | |
81 | "Stop jobs running", | |
82 | "Waiting for other daemons to come down", | |
83 | "Zebra restart job running", | |
84 | "Waiting for zebra to come up", | |
85 | "Start jobs running", | |
8b886ca7 | 86 | }; |
87 | ||
88 | #define PHASE_TIMEOUT (3*gs.restart_timeout) | |
89 | ||
a6810074 DL |
90 | struct restart_info { |
91 | const char *name; | |
92 | const char *what; | |
93 | pid_t pid; | |
94 | struct timeval time; | |
95 | long interval; | |
96 | struct thread *t_kill; | |
97 | int kills; | |
098e240f | 98 | }; |
99 | ||
a6810074 | 100 | static struct global_state { |
a6810074 DL |
101 | restart_phase_t phase; |
102 | struct thread *t_phase_hanging; | |
103 | const char *vtydir; | |
104 | long period; | |
105 | long timeout; | |
106 | long restart_timeout; | |
107 | long min_restart_interval; | |
108 | long max_restart_interval; | |
109 | int do_ping; | |
110 | struct daemon *daemons; | |
111 | const char *restart_command; | |
112 | const char *start_command; | |
113 | const char *stop_command; | |
114 | struct restart_info restart; | |
115 | int unresponsive_restart; | |
116 | int loglevel; | |
d62a17ae | 117 | struct daemon *special; /* points to zebra when doing phased restart */ |
a6810074 DL |
118 | int numdaemons; |
119 | int numpids; | |
d62a17ae | 120 | int numdown; /* # of daemons that are not UP or UNRESPONSIVE */ |
8b886ca7 | 121 | } gs = { |
d62a17ae | 122 | .phase = PHASE_NONE, |
123 | .vtydir = VTYDIR, | |
124 | .period = 1000 * DEFAULT_PERIOD, | |
125 | .timeout = DEFAULT_TIMEOUT, | |
126 | .restart_timeout = DEFAULT_RESTART_TIMEOUT, | |
127 | .loglevel = DEFAULT_LOGLEVEL, | |
128 | .min_restart_interval = DEFAULT_MIN_RESTART, | |
129 | .max_restart_interval = DEFAULT_MAX_RESTART, | |
130 | .do_ping = 1, | |
131 | }; | |
a6810074 DL |
132 | |
133 | typedef enum { | |
134 | DAEMON_INIT, | |
135 | DAEMON_DOWN, | |
136 | DAEMON_CONNECTING, | |
137 | DAEMON_UP, | |
138 | DAEMON_UNRESPONSIVE | |
8b886ca7 | 139 | } daemon_state_t; |
140 | ||
d62a17ae | 141 | #define IS_UP(DMN) \ |
142 | (((DMN)->state == DAEMON_UP) || ((DMN)->state == DAEMON_UNRESPONSIVE)) | |
8b886ca7 | 143 | |
a6810074 | 144 | static const char *state_str[] = { |
d62a17ae | 145 | "Init", "Down", "Connecting", "Up", "Unresponsive", |
8b886ca7 | 146 | }; |
147 | ||
148 | struct daemon { | |
a6810074 DL |
149 | const char *name; |
150 | daemon_state_t state; | |
151 | int fd; | |
152 | struct timeval echo_sent; | |
153 | u_int connect_tries; | |
154 | struct thread *t_wakeup; | |
155 | struct thread *t_read; | |
156 | struct thread *t_write; | |
157 | struct daemon *next; | |
158 | struct restart_info restart; | |
8b886ca7 | 159 | }; |
160 | ||
9272302b DL |
161 | #define OPTION_MINRESTART 2000 |
162 | #define OPTION_MAXRESTART 2001 | |
f168b713 | 163 | #define OPTION_DRY 2002 |
9272302b | 164 | |
a6810074 DL |
165 | static const struct option longopts[] = { |
166 | {"daemon", no_argument, NULL, 'd'}, | |
167 | {"statedir", required_argument, NULL, 'S'}, | |
168 | {"no-echo", no_argument, NULL, 'e'}, | |
169 | {"loglevel", required_argument, NULL, 'l'}, | |
170 | {"interval", required_argument, NULL, 'i'}, | |
171 | {"timeout", required_argument, NULL, 't'}, | |
172 | {"restart-timeout", required_argument, NULL, 'T'}, | |
173 | {"restart", required_argument, NULL, 'r'}, | |
174 | {"start-command", required_argument, NULL, 's'}, | |
175 | {"kill-command", required_argument, NULL, 'k'}, | |
a6810074 | 176 | {"unresponsive-restart", no_argument, NULL, 'z'}, |
f168b713 | 177 | {"dry", no_argument, NULL, OPTION_DRY}, |
d62a17ae | 178 | {"min-restart-interval", required_argument, NULL, OPTION_MINRESTART}, |
179 | {"max-restart-interval", required_argument, NULL, OPTION_MAXRESTART}, | |
a6810074 DL |
180 | {"pid-file", required_argument, NULL, 'p'}, |
181 | {"blank-string", required_argument, NULL, 'b'}, | |
182 | {"help", no_argument, NULL, 'h'}, | |
183 | {"version", no_argument, NULL, 'v'}, | |
d62a17ae | 184 | {NULL, 0, NULL, 0}}; |
8b886ca7 | 185 | |
186 | static int try_connect(struct daemon *dmn); | |
187 | static int wakeup_send_echo(struct thread *t_wakeup); | |
188 | static void try_restart(struct daemon *dmn); | |
189 | static void phase_check(void); | |
190 | ||
4f04a76b DL |
191 | static const char *progname; |
192 | static void printhelp(FILE *target) | |
8b886ca7 | 193 | { |
d62a17ae | 194 | fprintf(target, |
195 | "Usage : %s [OPTION...] <daemon name> ...\n\n\ | |
9473e340 | 196 | Watchdog program to monitor status of frr daemons and try to restart\n\ |
8b886ca7 | 197 | them if they are down or unresponsive. It determines whether a daemon is\n\ |
198 | up based on whether it can connect to the daemon's vty unix stream socket.\n\ | |
199 | It then repeatedly sends echo commands over that socket to determine whether\n\ | |
200 | the daemon is responsive. If the daemon crashes, we will receive an EOF\n\ | |
201 | on the socket connection and know immediately that the daemon is down.\n\n\ | |
202 | The daemons to be monitored should be listed on the command line.\n\n\ | |
8b886ca7 | 203 | In order to avoid attempting to restart the daemons in a fast loop,\n\ |
204 | the -m and -M options allow you to control the minimum delay between\n\ | |
205 | restart commands. The minimum restart delay is recalculated each time\n\ | |
206 | a restart is attempted: if the time since the last restart attempt exceeds\n\ | |
207 | twice the -M value, then the restart delay is set to the -m value.\n\ | |
d62a17ae | 208 | Otherwise, the interval is doubled (but capped at the -M value).\n\n", |
f168b713 | 209 | progname); |
e757c940 | 210 | |
d62a17ae | 211 | fprintf(target, |
212 | "Options:\n\ | |
8b886ca7 | 213 | -d, --daemon Run in daemon mode. In this mode, error messages are sent\n\ |
214 | to syslog instead of stdout.\n\ | |
215 | -S, --statedir Set the vty socket directory (default is %s)\n\ | |
216 | -e, --no-echo Do not ping the daemons to test responsiveness (this\n\ | |
217 | option is necessary if the daemons do not support the\n\ | |
218 | echo command)\n\ | |
219 | -l, --loglevel Set the logging level (default is %d).\n\ | |
220 | The value should range from %d (LOG_EMERG) to %d (LOG_DEBUG),\n\ | |
221 | but it can be set higher than %d if extra-verbose debugging\n\ | |
222 | messages are desired.\n\ | |
9272302b | 223 | --min-restart-interval\n\ |
8b886ca7 | 224 | Set the minimum seconds to wait between invocations of daemon\n\ |
225 | restart commands (default is %d).\n\ | |
9272302b | 226 | --max-restart-interval\n\ |
8b886ca7 | 227 | Set the maximum seconds to wait between invocations of daemon\n\ |
228 | restart commands (default is %d).\n\ | |
229 | -i, --interval Set the status polling interval in seconds (default is %d)\n\ | |
230 | -t, --timeout Set the unresponsiveness timeout in seconds (default is %d)\n\ | |
231 | -T, --restart-timeout\n\ | |
232 | Set the restart (kill) timeout in seconds (default is %d).\n\ | |
233 | If any background jobs are still running after this much\n\ | |
234 | time has elapsed, they will be killed.\n\ | |
235 | -r, --restart Supply a Bourne shell command to use to restart a single\n\ | |
236 | daemon. The command string should include '%%s' where the\n\ | |
237 | name of the daemon should be substituted.\n\ | |
8b886ca7 | 238 | -s, --start-command\n\ |
239 | Supply a Bourne shell to command to use to start a single\n\ | |
240 | daemon. The command string should include '%%s' where the\n\ | |
241 | name of the daemon should be substituted.\n\ | |
242 | -k, --kill-command\n\ | |
243 | Supply a Bourne shell to command to use to stop a single\n\ | |
244 | daemon. The command string should include '%%s' where the\n\ | |
245 | name of the daemon should be substituted.\n\ | |
8b886ca7 | 246 | -z, --unresponsive-restart\n\ |
247 | When a daemon is unresponsive, treat it as being down for\n\ | |
248 | restart purposes.\n\ | |
f168b713 | 249 | --dry Do not start or restart anything, just log.\n\ |
8b886ca7 | 250 | -p, --pid-file Set process identifier file name\n\ |
251 | (default is %s).\n\ | |
c8b40f86 | 252 | -b, --blank-string\n\ |
253 | When the supplied argument string is found in any of the\n\ | |
f168b713 | 254 | various shell command arguments (-r, -s, or -k), replace\n\ |
c8b40f86 | 255 | it with a space. This is an ugly hack to circumvent problems\n\ |
256 | passing command-line arguments with embedded spaces.\n\ | |
8b886ca7 | 257 | -v, --version Print program version\n\ |
d62a17ae | 258 | -h, --help Display this help and exit\n", |
259 | VTYDIR, DEFAULT_LOGLEVEL, LOG_EMERG, LOG_DEBUG, LOG_DEBUG, | |
260 | DEFAULT_MIN_RESTART, DEFAULT_MAX_RESTART, DEFAULT_PERIOD, | |
261 | DEFAULT_TIMEOUT, DEFAULT_RESTART_TIMEOUT, DEFAULT_PIDFILE); | |
8b886ca7 | 262 | } |
263 | ||
a6810074 | 264 | static pid_t run_background(char *shell_cmd) |
8b886ca7 | 265 | { |
a6810074 DL |
266 | pid_t child; |
267 | ||
268 | switch (child = fork()) { | |
269 | case -1: | |
d62a17ae | 270 | zlog_err("fork failed, cannot run command [%s]: %s", shell_cmd, |
271 | safe_strerror(errno)); | |
a6810074 DL |
272 | return -1; |
273 | case 0: | |
274 | /* Child process. */ | |
d62a17ae | 275 | /* Use separate process group so child processes can be killed |
276 | * easily. */ | |
a6810074 DL |
277 | if (setpgid(0, 0) < 0) |
278 | zlog_warn("warning: setpgid(0,0) failed: %s", | |
279 | safe_strerror(errno)); | |
280 | { | |
281 | char shell[] = "sh"; | |
282 | char dashc[] = "-c"; | |
d62a17ae | 283 | char *const argv[4] = {shell, dashc, shell_cmd, NULL}; |
a6810074 | 284 | execv("/bin/sh", argv); |
d62a17ae | 285 | zlog_err("execv(/bin/sh -c '%s') failed: %s", shell_cmd, |
286 | safe_strerror(errno)); | |
a6810074 DL |
287 | _exit(127); |
288 | } | |
289 | default: | |
290 | /* Parent process: we will reap the child later. */ | |
291 | zlog_err("Forked background command [pid %d]: %s", (int)child, | |
292 | shell_cmd); | |
293 | return child; | |
294 | } | |
8b886ca7 | 295 | } |
296 | ||
a6810074 DL |
297 | static struct timeval *time_elapsed(struct timeval *result, |
298 | const struct timeval *start_time) | |
8b886ca7 | 299 | { |
a6810074 DL |
300 | gettimeofday(result, NULL); |
301 | result->tv_sec -= start_time->tv_sec; | |
302 | result->tv_usec -= start_time->tv_usec; | |
303 | while (result->tv_usec < 0) { | |
304 | result->tv_usec += 1000000L; | |
305 | result->tv_sec--; | |
306 | } | |
307 | return result; | |
8b886ca7 | 308 | } |
309 | ||
a6810074 | 310 | static int restart_kill(struct thread *t_kill) |
8b886ca7 | 311 | { |
a6810074 DL |
312 | struct restart_info *restart = THREAD_ARG(t_kill); |
313 | struct timeval delay; | |
314 | ||
315 | time_elapsed(&delay, &restart->time); | |
d62a17ae | 316 | zlog_warn( |
317 | "Warning: %s %s child process %d still running after " | |
318 | "%ld seconds, sending signal %d", | |
319 | restart->what, restart->name, (int)restart->pid, | |
320 | (long)delay.tv_sec, (restart->kills ? SIGKILL : SIGTERM)); | |
a6810074 DL |
321 | kill(-restart->pid, (restart->kills ? SIGKILL : SIGTERM)); |
322 | restart->kills++; | |
66e78ae6 QY |
323 | restart->t_kill = NULL; |
324 | thread_add_timer(master, restart_kill, restart, gs.restart_timeout, | |
325 | &restart->t_kill); | |
a6810074 | 326 | return 0; |
8b886ca7 | 327 | } |
328 | ||
a6810074 | 329 | static struct restart_info *find_child(pid_t child) |
8b886ca7 | 330 | { |
f168b713 DL |
331 | struct daemon *dmn; |
332 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { | |
333 | if (dmn->restart.pid == child) | |
334 | return &dmn->restart; | |
a6810074 DL |
335 | } |
336 | return NULL; | |
8b886ca7 | 337 | } |
338 | ||
a6810074 | 339 | static void sigchild(void) |
8b886ca7 | 340 | { |
a6810074 DL |
341 | pid_t child; |
342 | int status; | |
343 | const char *name; | |
344 | const char *what; | |
345 | struct restart_info *restart; | |
346 | ||
347 | switch (child = waitpid(-1, &status, WNOHANG)) { | |
348 | case -1: | |
349 | zlog_err("waitpid failed: %s", safe_strerror(errno)); | |
350 | return; | |
351 | case 0: | |
352 | zlog_warn("SIGCHLD received, but waitpid did not reap a child"); | |
353 | return; | |
354 | } | |
355 | ||
356 | if (child == integrated_write_pid) { | |
357 | integrated_write_sigchld(status); | |
358 | return; | |
359 | } | |
360 | ||
361 | if ((restart = find_child(child)) != NULL) { | |
362 | name = restart->name; | |
363 | what = restart->what; | |
364 | restart->pid = 0; | |
365 | gs.numpids--; | |
366 | thread_cancel(restart->t_kill); | |
367 | restart->t_kill = NULL; | |
d62a17ae | 368 | /* Update restart time to reflect the time the command |
369 | * completed. */ | |
a6810074 DL |
370 | gettimeofday(&restart->time, NULL); |
371 | } else { | |
d62a17ae | 372 | zlog_err( |
373 | "waitpid returned status for an unknown child process %d", | |
374 | (int)child); | |
a6810074 DL |
375 | name = "(unknown)"; |
376 | what = "background"; | |
377 | } | |
378 | if (WIFSTOPPED(status)) | |
d62a17ae | 379 | zlog_warn("warning: %s %s process %d is stopped", what, name, |
380 | (int)child); | |
a6810074 | 381 | else if (WIFSIGNALED(status)) |
d62a17ae | 382 | zlog_warn("%s %s process %d terminated due to signal %d", what, |
383 | name, (int)child, WTERMSIG(status)); | |
a6810074 DL |
384 | else if (WIFEXITED(status)) { |
385 | if (WEXITSTATUS(status) != 0) | |
d62a17ae | 386 | zlog_warn( |
387 | "%s %s process %d exited with non-zero status %d", | |
388 | what, name, (int)child, WEXITSTATUS(status)); | |
a6810074 DL |
389 | else |
390 | zlog_debug("%s %s process %d exited normally", what, | |
391 | name, (int)child); | |
392 | } else | |
393 | zlog_err("cannot interpret %s %s process %d wait status 0x%x", | |
394 | what, name, (int)child, status); | |
395 | phase_check(); | |
8b886ca7 | 396 | } |
397 | ||
d62a17ae | 398 | static int run_job(struct restart_info *restart, const char *cmdtype, |
399 | const char *command, int force, int update_interval) | |
8b886ca7 | 400 | { |
a6810074 DL |
401 | struct timeval delay; |
402 | ||
403 | if (gs.loglevel > LOG_DEBUG + 1) | |
404 | zlog_debug("attempting to %s %s", cmdtype, restart->name); | |
405 | ||
406 | if (restart->pid) { | |
407 | if (gs.loglevel > LOG_DEBUG + 1) | |
d62a17ae | 408 | zlog_debug( |
409 | "cannot %s %s, previous pid %d still running", | |
410 | cmdtype, restart->name, (int)restart->pid); | |
a6810074 DL |
411 | return -1; |
412 | } | |
413 | ||
d62a17ae | 414 | /* Note: time_elapsed test must come before the force test, since we |
415 | need | |
a6810074 DL |
416 | to make sure that delay is initialized for use below in updating the |
417 | restart interval. */ | |
418 | if ((time_elapsed(&delay, &restart->time)->tv_sec < restart->interval) | |
419 | && !force) { | |
420 | if (gs.loglevel > LOG_DEBUG + 1) | |
d62a17ae | 421 | zlog_debug( |
422 | "postponing %s %s: " | |
423 | "elapsed time %ld < retry interval %ld", | |
424 | cmdtype, restart->name, (long)delay.tv_sec, | |
425 | restart->interval); | |
a6810074 DL |
426 | return -1; |
427 | } | |
428 | ||
429 | gettimeofday(&restart->time, NULL); | |
430 | restart->kills = 0; | |
431 | { | |
432 | char cmd[strlen(command) + strlen(restart->name) + 1]; | |
433 | snprintf(cmd, sizeof(cmd), command, restart->name); | |
434 | if ((restart->pid = run_background(cmd)) > 0) { | |
66e78ae6 | 435 | restart->t_kill = NULL; |
d62a17ae | 436 | thread_add_timer(master, restart_kill, restart, |
437 | gs.restart_timeout, &restart->t_kill); | |
a6810074 DL |
438 | restart->what = cmdtype; |
439 | gs.numpids++; | |
440 | } else | |
441 | restart->pid = 0; | |
442 | } | |
443 | ||
444 | /* Calculate the new restart interval. */ | |
445 | if (update_interval) { | |
446 | if (delay.tv_sec > 2 * gs.max_restart_interval) | |
447 | restart->interval = gs.min_restart_interval; | |
448 | else if ((restart->interval *= 2) > gs.max_restart_interval) | |
449 | restart->interval = gs.max_restart_interval; | |
450 | if (gs.loglevel > LOG_DEBUG + 1) | |
451 | zlog_debug("restart %s interval is now %ld", | |
452 | restart->name, restart->interval); | |
453 | } | |
454 | return restart->pid; | |
8b886ca7 | 455 | } |
456 | ||
d62a17ae | 457 | #define SET_READ_HANDLER(DMN) \ |
458 | do { \ | |
459 | (DMN)->t_read = NULL; \ | |
460 | thread_add_read(master, handle_read, (DMN), (DMN)->fd, \ | |
461 | &(DMN)->t_read); \ | |
462 | } while (0); | |
463 | ||
464 | #define SET_WAKEUP_DOWN(DMN) \ | |
465 | do { \ | |
466 | (DMN)->t_wakeup = NULL; \ | |
467 | thread_add_timer_msec(master, wakeup_down, (DMN), \ | |
468 | FUZZY(gs.period), &(DMN)->t_wakeup); \ | |
469 | } while (0); | |
470 | ||
471 | #define SET_WAKEUP_UNRESPONSIVE(DMN) \ | |
472 | do { \ | |
473 | (DMN)->t_wakeup = NULL; \ | |
474 | thread_add_timer_msec(master, wakeup_unresponsive, (DMN), \ | |
475 | FUZZY(gs.period), &(DMN)->t_wakeup); \ | |
476 | } while (0); | |
477 | ||
478 | #define SET_WAKEUP_ECHO(DMN) \ | |
479 | do { \ | |
480 | (DMN)->t_wakeup = NULL; \ | |
481 | thread_add_timer_msec(master, wakeup_send_echo, (DMN), \ | |
482 | FUZZY(gs.period), &(DMN)->t_wakeup); \ | |
483 | } while (0); | |
8b886ca7 | 484 | |
a6810074 | 485 | static int wakeup_down(struct thread *t_wakeup) |
8b886ca7 | 486 | { |
a6810074 DL |
487 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
488 | ||
489 | dmn->t_wakeup = NULL; | |
490 | if (try_connect(dmn) < 0) | |
491 | SET_WAKEUP_DOWN(dmn); | |
492 | if ((dmn->connect_tries > 1) && (dmn->state != DAEMON_UP)) | |
493 | try_restart(dmn); | |
494 | return 0; | |
8b886ca7 | 495 | } |
496 | ||
a6810074 | 497 | static int wakeup_init(struct thread *t_wakeup) |
8b886ca7 | 498 | { |
a6810074 DL |
499 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
500 | ||
501 | dmn->t_wakeup = NULL; | |
502 | if (try_connect(dmn) < 0) { | |
503 | SET_WAKEUP_DOWN(dmn); | |
504 | zlog_err("%s state -> down : initial connection attempt failed", | |
505 | dmn->name); | |
506 | dmn->state = DAEMON_DOWN; | |
507 | } | |
508 | return 0; | |
8b886ca7 | 509 | } |
510 | ||
a6810074 | 511 | static void daemon_down(struct daemon *dmn, const char *why) |
8b886ca7 | 512 | { |
a6810074 DL |
513 | if (IS_UP(dmn) || (dmn->state == DAEMON_INIT)) |
514 | zlog_err("%s state -> down : %s", dmn->name, why); | |
515 | else if (gs.loglevel > LOG_DEBUG) | |
516 | zlog_debug("%s still down : %s", dmn->name, why); | |
517 | if (IS_UP(dmn)) | |
518 | gs.numdown++; | |
519 | dmn->state = DAEMON_DOWN; | |
520 | if (dmn->fd >= 0) { | |
521 | close(dmn->fd); | |
522 | dmn->fd = -1; | |
523 | } | |
524 | THREAD_OFF(dmn->t_read); | |
525 | THREAD_OFF(dmn->t_write); | |
526 | THREAD_OFF(dmn->t_wakeup); | |
527 | if (try_connect(dmn) < 0) | |
528 | SET_WAKEUP_DOWN(dmn); | |
529 | phase_check(); | |
8b886ca7 | 530 | } |
531 | ||
a6810074 | 532 | static int handle_read(struct thread *t_read) |
8b886ca7 | 533 | { |
a6810074 DL |
534 | struct daemon *dmn = THREAD_ARG(t_read); |
535 | static const char resp[sizeof(PING_TOKEN) + 4] = PING_TOKEN "\n"; | |
536 | char buf[sizeof(resp) + 100]; | |
537 | ssize_t rc; | |
538 | struct timeval delay; | |
539 | ||
540 | dmn->t_read = NULL; | |
541 | if ((rc = read(dmn->fd, buf, sizeof(buf))) < 0) { | |
542 | char why[100]; | |
543 | ||
544 | if (ERRNO_IO_RETRY(errno)) { | |
545 | /* Pretend it never happened. */ | |
546 | SET_READ_HANDLER(dmn); | |
547 | return 0; | |
548 | } | |
549 | snprintf(why, sizeof(why), "unexpected read error: %s", | |
550 | safe_strerror(errno)); | |
551 | daemon_down(dmn, why); | |
552 | return 0; | |
8b886ca7 | 553 | } |
a6810074 DL |
554 | if (rc == 0) { |
555 | daemon_down(dmn, "read returned EOF"); | |
556 | return 0; | |
557 | } | |
558 | if (!dmn->echo_sent.tv_sec) { | |
559 | char why[sizeof(buf) + 100]; | |
560 | snprintf(why, sizeof(why), | |
561 | "unexpected read returns %d bytes: %.*s", (int)rc, | |
562 | (int)rc, buf); | |
563 | daemon_down(dmn, why); | |
564 | return 0; | |
8b886ca7 | 565 | } |
a6810074 DL |
566 | |
567 | /* We are expecting an echo response: is there any chance that the | |
568 | response would not be returned entirely in the first read? That | |
569 | seems inconceivable... */ | |
570 | if ((rc != sizeof(resp)) || memcmp(buf, resp, sizeof(resp))) { | |
571 | char why[100 + sizeof(buf)]; | |
572 | snprintf(why, sizeof(why), | |
573 | "read returned bad echo response of %d bytes " | |
d62a17ae | 574 | "(expecting %u): %.*s", |
575 | (int)rc, (u_int)sizeof(resp), (int)rc, buf); | |
a6810074 DL |
576 | daemon_down(dmn, why); |
577 | return 0; | |
578 | } | |
579 | ||
580 | time_elapsed(&delay, &dmn->echo_sent); | |
581 | dmn->echo_sent.tv_sec = 0; | |
582 | if (dmn->state == DAEMON_UNRESPONSIVE) { | |
583 | if (delay.tv_sec < gs.timeout) { | |
584 | dmn->state = DAEMON_UP; | |
d62a17ae | 585 | zlog_warn( |
586 | "%s state -> up : echo response received after %ld.%06ld " | |
587 | "seconds", | |
588 | dmn->name, (long)delay.tv_sec, | |
589 | (long)delay.tv_usec); | |
a6810074 | 590 | } else |
d62a17ae | 591 | zlog_warn( |
592 | "%s: slow echo response finally received after %ld.%06ld " | |
593 | "seconds", | |
594 | dmn->name, (long)delay.tv_sec, | |
595 | (long)delay.tv_usec); | |
a6810074 DL |
596 | } else if (gs.loglevel > LOG_DEBUG + 1) |
597 | zlog_debug("%s: echo response received after %ld.%06ld seconds", | |
598 | dmn->name, (long)delay.tv_sec, (long)delay.tv_usec); | |
599 | ||
600 | SET_READ_HANDLER(dmn); | |
601 | if (dmn->t_wakeup) | |
602 | thread_cancel(dmn->t_wakeup); | |
603 | SET_WAKEUP_ECHO(dmn); | |
604 | ||
605 | return 0; | |
8b886ca7 | 606 | } |
607 | ||
207e0d7a DS |
608 | /* |
609 | * Wait till we notice that all daemons are ready before | |
610 | * we send we are ready to systemd | |
611 | */ | |
a6810074 | 612 | static void daemon_send_ready(void) |
207e0d7a | 613 | { |
a6810074 DL |
614 | static int sent = 0; |
615 | if (!sent && gs.numdown == 0) { | |
a6810074 | 616 | FILE *fp; |
207e0d7a | 617 | |
a6810074 DL |
618 | fp = fopen(DAEMON_VTY_DIR "/watchfrr.started", "w"); |
619 | fclose(fp); | |
d62a17ae | 620 | zlog_notice( |
621 | "Watchfrr: Notifying Systemd we are up and running"); | |
a6810074 DL |
622 | systemd_send_started(master, 0); |
623 | sent = 1; | |
624 | } | |
207e0d7a DS |
625 | } |
626 | ||
a6810074 | 627 | static void daemon_up(struct daemon *dmn, const char *why) |
8b886ca7 | 628 | { |
a6810074 DL |
629 | dmn->state = DAEMON_UP; |
630 | gs.numdown--; | |
631 | dmn->connect_tries = 0; | |
632 | zlog_notice("%s state -> up : %s", dmn->name, why); | |
633 | daemon_send_ready(); | |
634 | if (gs.do_ping) | |
635 | SET_WAKEUP_ECHO(dmn); | |
636 | phase_check(); | |
8b886ca7 | 637 | } |
638 | ||
a6810074 | 639 | static int check_connect(struct thread *t_write) |
8b886ca7 | 640 | { |
a6810074 DL |
641 | struct daemon *dmn = THREAD_ARG(t_write); |
642 | int sockerr; | |
643 | socklen_t reslen = sizeof(sockerr); | |
644 | ||
645 | dmn->t_write = NULL; | |
646 | if (getsockopt(dmn->fd, SOL_SOCKET, SO_ERROR, (char *)&sockerr, &reslen) | |
647 | < 0) { | |
648 | zlog_warn("%s: check_connect: getsockopt failed: %s", dmn->name, | |
649 | safe_strerror(errno)); | |
650 | daemon_down(dmn, | |
651 | "getsockopt failed checking connection success"); | |
652 | return 0; | |
653 | } | |
654 | if ((reslen == sizeof(sockerr)) && sockerr) { | |
655 | char why[100]; | |
d62a17ae | 656 | snprintf( |
657 | why, sizeof(why), | |
658 | "getsockopt reports that connection attempt failed: %s", | |
659 | safe_strerror(sockerr)); | |
a6810074 DL |
660 | daemon_down(dmn, why); |
661 | return 0; | |
662 | } | |
663 | ||
664 | daemon_up(dmn, "delayed connect succeeded"); | |
665 | return 0; | |
8b886ca7 | 666 | } |
667 | ||
a6810074 | 668 | static int wakeup_connect_hanging(struct thread *t_wakeup) |
8b886ca7 | 669 | { |
a6810074 DL |
670 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
671 | char why[100]; | |
672 | ||
673 | dmn->t_wakeup = NULL; | |
674 | snprintf(why, sizeof(why), | |
675 | "connection attempt timed out after %ld seconds", gs.timeout); | |
676 | daemon_down(dmn, why); | |
677 | return 0; | |
8b886ca7 | 678 | } |
679 | ||
680 | /* Making connection to protocol daemon. */ | |
a6810074 | 681 | static int try_connect(struct daemon *dmn) |
8b886ca7 | 682 | { |
a6810074 DL |
683 | int sock; |
684 | struct sockaddr_un addr; | |
685 | socklen_t len; | |
686 | ||
687 | if (gs.loglevel > LOG_DEBUG + 1) | |
688 | zlog_debug("%s: attempting to connect", dmn->name); | |
689 | dmn->connect_tries++; | |
690 | ||
691 | memset(&addr, 0, sizeof(struct sockaddr_un)); | |
692 | addr.sun_family = AF_UNIX; | |
d62a17ae | 693 | snprintf(addr.sun_path, sizeof(addr.sun_path), "%s/%s.vty", gs.vtydir, |
694 | dmn->name); | |
6f0e3f6e | 695 | #ifdef HAVE_STRUCT_SOCKADDR_UN_SUN_LEN |
a6810074 | 696 | len = addr.sun_len = SUN_LEN(&addr); |
8b886ca7 | 697 | #else |
a6810074 | 698 | len = sizeof(addr.sun_family) + strlen(addr.sun_path); |
d62a17ae | 699 | #endif /* HAVE_STRUCT_SOCKADDR_UN_SUN_LEN */ |
a6810074 DL |
700 | |
701 | /* Quick check to see if we might succeed before we go to the trouble | |
702 | of creating a socket. */ | |
703 | if (access(addr.sun_path, W_OK) < 0) { | |
704 | if (errno != ENOENT) | |
705 | zlog_err("%s: access to socket %s denied: %s", | |
706 | dmn->name, addr.sun_path, | |
707 | safe_strerror(errno)); | |
708 | return -1; | |
709 | } | |
710 | ||
711 | if ((sock = socket(AF_UNIX, SOCK_STREAM, 0)) < 0) { | |
d62a17ae | 712 | zlog_err("%s(%s): cannot make socket: %s", __func__, |
713 | addr.sun_path, safe_strerror(errno)); | |
a6810074 DL |
714 | return -1; |
715 | } | |
716 | ||
717 | if (set_nonblocking(sock) < 0 || set_cloexec(sock) < 0) { | |
d62a17ae | 718 | zlog_err("%s(%s): set_nonblocking/cloexec(%d) failed", __func__, |
719 | addr.sun_path, sock); | |
a6810074 DL |
720 | close(sock); |
721 | return -1; | |
8b886ca7 | 722 | } |
a6810074 DL |
723 | |
724 | if (connect(sock, (struct sockaddr *)&addr, len) < 0) { | |
725 | if ((errno != EINPROGRESS) && (errno != EWOULDBLOCK)) { | |
726 | if (gs.loglevel > LOG_DEBUG) | |
727 | zlog_debug("%s(%s): connect failed: %s", | |
728 | __func__, addr.sun_path, | |
729 | safe_strerror(errno)); | |
730 | close(sock); | |
731 | return -1; | |
732 | } | |
733 | if (gs.loglevel > LOG_DEBUG) | |
734 | zlog_debug("%s: connection in progress", dmn->name); | |
735 | dmn->state = DAEMON_CONNECTING; | |
736 | dmn->fd = sock; | |
66e78ae6 QY |
737 | dmn->t_write = NULL; |
738 | thread_add_write(master, check_connect, dmn, dmn->fd, | |
d62a17ae | 739 | &dmn->t_write); |
740 | dmn->t_wakeup = NULL; | |
741 | thread_add_timer(master, wakeup_connect_hanging, dmn, | |
742 | gs.timeout, &dmn->t_wakeup); | |
a6810074 DL |
743 | SET_READ_HANDLER(dmn); |
744 | return 0; | |
745 | } | |
746 | ||
747 | dmn->fd = sock; | |
748 | SET_READ_HANDLER(dmn); | |
749 | daemon_up(dmn, "connect succeeded"); | |
750 | return 1; | |
8b886ca7 | 751 | } |
752 | ||
a6810074 | 753 | static int phase_hanging(struct thread *t_hanging) |
8b886ca7 | 754 | { |
a6810074 DL |
755 | gs.t_phase_hanging = NULL; |
756 | zlog_err("Phase [%s] hanging for %ld seconds, aborting phased restart", | |
757 | phase_str[gs.phase], PHASE_TIMEOUT); | |
758 | gs.phase = PHASE_NONE; | |
759 | return 0; | |
8b886ca7 | 760 | } |
761 | ||
a6810074 | 762 | static void set_phase(restart_phase_t new_phase) |
8b886ca7 | 763 | { |
a6810074 DL |
764 | gs.phase = new_phase; |
765 | if (gs.t_phase_hanging) | |
766 | thread_cancel(gs.t_phase_hanging); | |
66e78ae6 QY |
767 | gs.t_phase_hanging = NULL; |
768 | thread_add_timer(master, phase_hanging, NULL, PHASE_TIMEOUT, | |
769 | &gs.t_phase_hanging); | |
8b886ca7 | 770 | } |
771 | ||
a6810074 | 772 | static void phase_check(void) |
8b886ca7 | 773 | { |
a6810074 DL |
774 | switch (gs.phase) { |
775 | case PHASE_NONE: | |
776 | break; | |
777 | case PHASE_STOPS_PENDING: | |
778 | if (gs.numpids) | |
779 | break; | |
d62a17ae | 780 | zlog_info( |
781 | "Phased restart: all routing daemon stop jobs have completed."); | |
a6810074 DL |
782 | set_phase(PHASE_WAITING_DOWN); |
783 | ||
d62a17ae | 784 | /*FALLTHRU*/ |
a6810074 DL |
785 | case PHASE_WAITING_DOWN: |
786 | if (gs.numdown + IS_UP(gs.special) < gs.numdaemons) | |
787 | break; | |
788 | zlog_info("Phased restart: all routing daemons now down."); | |
789 | run_job(&gs.special->restart, "restart", gs.restart_command, 1, | |
790 | 1); | |
791 | set_phase(PHASE_ZEBRA_RESTART_PENDING); | |
792 | ||
d62a17ae | 793 | /*FALLTHRU*/ |
a6810074 DL |
794 | case PHASE_ZEBRA_RESTART_PENDING: |
795 | if (gs.special->restart.pid) | |
796 | break; | |
797 | zlog_info("Phased restart: %s restart job completed.", | |
798 | gs.special->name); | |
799 | set_phase(PHASE_WAITING_ZEBRA_UP); | |
800 | ||
d62a17ae | 801 | /*FALLTHRU*/ |
a6810074 DL |
802 | case PHASE_WAITING_ZEBRA_UP: |
803 | if (!IS_UP(gs.special)) | |
804 | break; | |
805 | zlog_info("Phased restart: %s is now up.", gs.special->name); | |
806 | { | |
807 | struct daemon *dmn; | |
808 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { | |
809 | if (dmn != gs.special) | |
810 | run_job(&dmn->restart, "start", | |
811 | gs.start_command, 1, 0); | |
812 | } | |
813 | } | |
814 | gs.phase = PHASE_NONE; | |
815 | THREAD_OFF(gs.t_phase_hanging); | |
816 | zlog_notice("Phased global restart has completed."); | |
817 | break; | |
818 | } | |
8b886ca7 | 819 | } |
820 | ||
a6810074 | 821 | static void try_restart(struct daemon *dmn) |
8b886ca7 | 822 | { |
f168b713 | 823 | if (watch_only) |
a6810074 | 824 | return; |
a6810074 | 825 | |
f168b713 DL |
826 | if (dmn != gs.special) { |
827 | if ((gs.special->state == DAEMON_UP) | |
828 | && (gs.phase == PHASE_NONE)) | |
829 | run_job(&dmn->restart, "restart", gs.restart_command, 0, | |
830 | 1); | |
831 | else | |
832 | zlog_debug( | |
833 | "%s: postponing restart attempt because master %s daemon " | |
834 | "not up [%s], or phased restart in progress", | |
835 | dmn->name, gs.special->name, | |
836 | state_str[gs.special->state]); | |
837 | return; | |
838 | } | |
839 | ||
840 | if ((gs.phase != PHASE_NONE) || gs.numpids) { | |
841 | if (gs.loglevel > LOG_DEBUG + 1) | |
842 | zlog_debug( | |
843 | "postponing phased global restart: restart already in " | |
844 | "progress [%s], or outstanding child processes [%d]", | |
845 | phase_str[gs.phase], gs.numpids); | |
846 | return; | |
847 | } | |
848 | /* Is it too soon for a restart? */ | |
849 | { | |
850 | struct timeval delay; | |
851 | if (time_elapsed(&delay, &gs.special->restart.time)->tv_sec | |
852 | < gs.special->restart.interval) { | |
a6810074 | 853 | if (gs.loglevel > LOG_DEBUG + 1) |
d62a17ae | 854 | zlog_debug( |
f168b713 DL |
855 | "postponing phased global restart: " |
856 | "elapsed time %ld < retry interval %ld", | |
857 | (long)delay.tv_sec, | |
858 | gs.special->restart.interval); | |
859 | return; | |
a6810074 | 860 | } |
8b886ca7 | 861 | } |
f168b713 | 862 | run_job(&gs.restart, "restart", gs.restart_command, 0, 1); |
8b886ca7 | 863 | } |
864 | ||
a6810074 | 865 | static int wakeup_unresponsive(struct thread *t_wakeup) |
8b886ca7 | 866 | { |
a6810074 DL |
867 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
868 | ||
869 | dmn->t_wakeup = NULL; | |
870 | if (dmn->state != DAEMON_UNRESPONSIVE) | |
d62a17ae | 871 | zlog_err( |
872 | "%s: no longer unresponsive (now %s), " | |
873 | "wakeup should have been cancelled!", | |
874 | dmn->name, state_str[dmn->state]); | |
a6810074 DL |
875 | else { |
876 | SET_WAKEUP_UNRESPONSIVE(dmn); | |
877 | try_restart(dmn); | |
878 | } | |
879 | return 0; | |
8b886ca7 | 880 | } |
881 | ||
a6810074 | 882 | static int wakeup_no_answer(struct thread *t_wakeup) |
8b886ca7 | 883 | { |
a6810074 DL |
884 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
885 | ||
886 | dmn->t_wakeup = NULL; | |
887 | dmn->state = DAEMON_UNRESPONSIVE; | |
d62a17ae | 888 | zlog_err( |
889 | "%s state -> unresponsive : no response yet to ping " | |
890 | "sent %ld seconds ago", | |
891 | dmn->name, gs.timeout); | |
a6810074 DL |
892 | if (gs.unresponsive_restart) { |
893 | SET_WAKEUP_UNRESPONSIVE(dmn); | |
894 | try_restart(dmn); | |
895 | } | |
896 | return 0; | |
8b886ca7 | 897 | } |
898 | ||
a6810074 | 899 | static int wakeup_send_echo(struct thread *t_wakeup) |
8b886ca7 | 900 | { |
a6810074 DL |
901 | static const char echocmd[] = "echo " PING_TOKEN; |
902 | ssize_t rc; | |
903 | struct daemon *dmn = THREAD_ARG(t_wakeup); | |
904 | ||
905 | dmn->t_wakeup = NULL; | |
d62a17ae | 906 | if (((rc = write(dmn->fd, echocmd, sizeof(echocmd))) < 0) |
907 | || ((size_t)rc != sizeof(echocmd))) { | |
a6810074 DL |
908 | char why[100 + sizeof(echocmd)]; |
909 | snprintf(why, sizeof(why), | |
910 | "write '%s' returned %d instead of %u", echocmd, | |
d62a17ae | 911 | (int)rc, (u_int)sizeof(echocmd)); |
a6810074 DL |
912 | daemon_down(dmn, why); |
913 | } else { | |
914 | gettimeofday(&dmn->echo_sent, NULL); | |
66e78ae6 QY |
915 | dmn->t_wakeup = NULL; |
916 | thread_add_timer(master, wakeup_no_answer, dmn, gs.timeout, | |
917 | &dmn->t_wakeup); | |
a6810074 DL |
918 | } |
919 | return 0; | |
8b886ca7 | 920 | } |
921 | ||
a6810074 | 922 | static void sigint(void) |
8b886ca7 | 923 | { |
a6810074 DL |
924 | zlog_notice("Terminating on signal"); |
925 | systemd_send_stopping(); | |
926 | exit(0); | |
8b886ca7 | 927 | } |
928 | ||
a6810074 | 929 | static int valid_command(const char *cmd) |
8b886ca7 | 930 | { |
a6810074 | 931 | char *p; |
8b886ca7 | 932 | |
a6810074 | 933 | return ((p = strchr(cmd, '%')) != NULL) && (*(p + 1) == 's') |
d62a17ae | 934 | && !strchr(p + 1, '%'); |
8b886ca7 | 935 | } |
936 | ||
c8b40f86 | 937 | /* This is an ugly hack to circumvent problems with passing command-line |
938 | arguments that contain spaces. The fix is to use a configuration file. */ | |
a6810074 | 939 | static char *translate_blanks(const char *cmd, const char *blankstr) |
c8b40f86 | 940 | { |
a6810074 DL |
941 | char *res; |
942 | char *p; | |
943 | size_t bslen = strlen(blankstr); | |
944 | ||
945 | if (!(res = strdup(cmd))) { | |
946 | perror("strdup"); | |
947 | exit(1); | |
948 | } | |
949 | while ((p = strstr(res, blankstr)) != NULL) { | |
950 | *p = ' '; | |
951 | if (bslen != 1) | |
952 | memmove(p + 1, p + bslen, strlen(p + bslen) + 1); | |
953 | } | |
954 | return res; | |
c8b40f86 | 955 | } |
956 | ||
a6810074 | 957 | struct zebra_privs_t watchfrr_privs = { |
95c4aff2 | 958 | #ifdef VTY_GROUP |
a6810074 | 959 | .vty_group = VTY_GROUP, |
95c4aff2 DL |
960 | #endif |
961 | }; | |
962 | ||
4f04a76b DL |
963 | static struct quagga_signal_t watchfrr_signals[] = { |
964 | { | |
965 | .signal = SIGINT, | |
966 | .handler = sigint, | |
967 | }, | |
968 | { | |
969 | .signal = SIGTERM, | |
970 | .handler = sigint, | |
971 | }, | |
972 | { | |
973 | .signal = SIGCHLD, | |
974 | .handler = sigchild, | |
975 | }, | |
976 | }; | |
977 | ||
978 | FRR_DAEMON_INFO(watchfrr, WATCHFRR, | |
d62a17ae | 979 | .flags = FRR_NO_PRIVSEP | FRR_NO_TCPVTY | FRR_LIMITED_CLI |
980 | | FRR_NO_CFG_PID_DRY | FRR_NO_ZCLIENT, | |
4f04a76b | 981 | |
d62a17ae | 982 | .printhelp = printhelp, |
983 | .copyright = "Copyright 2004 Andrew J. Schorr", | |
4f04a76b | 984 | |
d62a17ae | 985 | .signals = watchfrr_signals, |
986 | .n_signals = array_size(watchfrr_signals), | |
4f04a76b | 987 | |
d62a17ae | 988 | .privs = &watchfrr_privs, ) |
4f04a76b | 989 | |
a6810074 | 990 | int main(int argc, char **argv) |
8b886ca7 | 991 | { |
a6810074 | 992 | int opt; |
a6810074 DL |
993 | const char *pidfile = DEFAULT_PIDFILE; |
994 | const char *special = "zebra"; | |
995 | const char *blankstr = NULL; | |
a6810074 | 996 | |
4f04a76b DL |
997 | frr_preinit(&watchfrr_di, argc, argv); |
998 | progname = watchfrr_di.progname; | |
999 | ||
f168b713 | 1000 | frr_opt_add("b:dek:l:i:p:r:S:s:t:T:z", longopts, ""); |
a6810074 DL |
1001 | |
1002 | gs.restart.name = "all"; | |
4f04a76b | 1003 | while ((opt = frr_getopt(argc, argv, NULL)) != EOF) { |
a6810074 DL |
1004 | switch (opt) { |
1005 | case 0: | |
1006 | break; | |
a6810074 DL |
1007 | case 'b': |
1008 | blankstr = optarg; | |
1009 | break; | |
f168b713 DL |
1010 | case OPTION_DRY: |
1011 | watch_only = true; | |
1012 | break; | |
a6810074 DL |
1013 | case 'e': |
1014 | gs.do_ping = 0; | |
1015 | break; | |
1016 | case 'k': | |
1017 | if (!valid_command(optarg)) { | |
1018 | fprintf(stderr, | |
1019 | "Invalid kill command, must contain '%%s': %s\n", | |
1020 | optarg); | |
4f04a76b | 1021 | frr_help_exit(1); |
a6810074 DL |
1022 | } |
1023 | gs.stop_command = optarg; | |
1024 | break; | |
d62a17ae | 1025 | case 'l': { |
1026 | char garbage[3]; | |
1027 | if ((sscanf(optarg, "%d%1s", &gs.loglevel, garbage) | |
1028 | != 1) | |
1029 | || (gs.loglevel < LOG_EMERG)) { | |
1030 | fprintf(stderr, | |
1031 | "Invalid loglevel argument: %s\n", | |
1032 | optarg); | |
1033 | frr_help_exit(1); | |
a6810074 | 1034 | } |
d62a17ae | 1035 | } break; |
1036 | case OPTION_MINRESTART: { | |
1037 | char garbage[3]; | |
1038 | if ((sscanf(optarg, "%ld%1s", &gs.min_restart_interval, | |
1039 | garbage) | |
1040 | != 1) | |
1041 | || (gs.min_restart_interval < 0)) { | |
1042 | fprintf(stderr, | |
1043 | "Invalid min_restart_interval argument: %s\n", | |
1044 | optarg); | |
1045 | frr_help_exit(1); | |
a6810074 | 1046 | } |
d62a17ae | 1047 | } break; |
1048 | case OPTION_MAXRESTART: { | |
1049 | char garbage[3]; | |
1050 | if ((sscanf(optarg, "%ld%1s", &gs.max_restart_interval, | |
1051 | garbage) | |
1052 | != 1) | |
1053 | || (gs.max_restart_interval < 0)) { | |
1054 | fprintf(stderr, | |
1055 | "Invalid max_restart_interval argument: %s\n", | |
1056 | optarg); | |
1057 | frr_help_exit(1); | |
a6810074 | 1058 | } |
d62a17ae | 1059 | } break; |
1060 | case 'i': { | |
1061 | char garbage[3]; | |
1062 | int period; | |
1063 | if ((sscanf(optarg, "%d%1s", &period, garbage) != 1) | |
1064 | || (gs.period < 1)) { | |
1065 | fprintf(stderr, | |
1066 | "Invalid interval argument: %s\n", | |
1067 | optarg); | |
1068 | frr_help_exit(1); | |
a6810074 | 1069 | } |
d62a17ae | 1070 | gs.period = 1000 * period; |
1071 | } break; | |
a6810074 DL |
1072 | case 'p': |
1073 | pidfile = optarg; | |
1074 | break; | |
1075 | case 'r': | |
a6810074 DL |
1076 | if (!valid_command(optarg)) { |
1077 | fprintf(stderr, | |
1078 | "Invalid restart command, must contain '%%s': %s\n", | |
1079 | optarg); | |
4f04a76b | 1080 | frr_help_exit(1); |
a6810074 DL |
1081 | } |
1082 | gs.restart_command = optarg; | |
a6810074 DL |
1083 | break; |
1084 | case 's': | |
1085 | if (!valid_command(optarg)) { | |
1086 | fprintf(stderr, | |
1087 | "Invalid start command, must contain '%%s': %s\n", | |
1088 | optarg); | |
4f04a76b | 1089 | frr_help_exit(1); |
a6810074 DL |
1090 | } |
1091 | gs.start_command = optarg; | |
1092 | break; | |
1093 | case 'S': | |
1094 | gs.vtydir = optarg; | |
1095 | break; | |
d62a17ae | 1096 | case 't': { |
1097 | char garbage[3]; | |
1098 | if ((sscanf(optarg, "%ld%1s", &gs.timeout, garbage) | |
1099 | != 1) | |
1100 | || (gs.timeout < 1)) { | |
1101 | fprintf(stderr, | |
1102 | "Invalid timeout argument: %s\n", | |
1103 | optarg); | |
1104 | frr_help_exit(1); | |
a6810074 | 1105 | } |
d62a17ae | 1106 | } break; |
1107 | case 'T': { | |
1108 | char garbage[3]; | |
1109 | if ((sscanf(optarg, "%ld%1s", &gs.restart_timeout, | |
1110 | garbage) | |
1111 | != 1) | |
1112 | || (gs.restart_timeout < 1)) { | |
1113 | fprintf(stderr, | |
1114 | "Invalid restart timeout argument: %s\n", | |
1115 | optarg); | |
1116 | frr_help_exit(1); | |
a6810074 | 1117 | } |
d62a17ae | 1118 | } break; |
a6810074 DL |
1119 | case 'z': |
1120 | gs.unresponsive_restart = 1; | |
1121 | break; | |
a6810074 DL |
1122 | default: |
1123 | fputs("Invalid option.\n", stderr); | |
4f04a76b | 1124 | frr_help_exit(1); |
a6810074 | 1125 | } |
8b886ca7 | 1126 | } |
a6810074 | 1127 | |
f168b713 DL |
1128 | if (watch_only && (gs.unresponsive_restart || gs.start_command |
1129 | || gs.stop_command || gs.restart_command)) { | |
1130 | fputs("Options -z/-r/-s/-k make no sense combined with -D.\n", | |
a6810074 | 1131 | stderr); |
4f04a76b | 1132 | frr_help_exit(1); |
8b886ca7 | 1133 | } |
f168b713 DL |
1134 | if (!watch_only |
1135 | && (!gs.restart_command || !gs.start_command || !gs.stop_command)) { | |
1136 | fprintf(stderr, | |
1137 | "Options -s (start), -k (kill), and -r (restart) are required.\n"); | |
1138 | frr_help_exit(1); | |
8b886ca7 | 1139 | } |
8b886ca7 | 1140 | |
a6810074 DL |
1141 | if (blankstr) { |
1142 | if (gs.restart_command) | |
1143 | gs.restart_command = | |
d62a17ae | 1144 | translate_blanks(gs.restart_command, blankstr); |
a6810074 DL |
1145 | if (gs.start_command) |
1146 | gs.start_command = | |
d62a17ae | 1147 | translate_blanks(gs.start_command, blankstr); |
a6810074 DL |
1148 | if (gs.stop_command) |
1149 | gs.stop_command = | |
d62a17ae | 1150 | translate_blanks(gs.stop_command, blankstr); |
065de903 | 1151 | } |
8b886ca7 | 1152 | |
a6810074 | 1153 | gs.restart.interval = gs.min_restart_interval; |
8b886ca7 | 1154 | |
4f04a76b DL |
1155 | master = frr_init(); |
1156 | ||
dd8376fe | 1157 | zlog_set_level(ZLOG_DEST_MONITOR, ZLOG_DISABLED); |
eb05883f | 1158 | if (watchfrr_di.daemon_mode) { |
dd8376fe | 1159 | zlog_set_level(ZLOG_DEST_SYSLOG, MIN(gs.loglevel, LOG_DEBUG)); |
d62a17ae | 1160 | if (daemon(0, 0) < 0) { |
2f4f11fa | 1161 | fprintf(stderr, "Watchfrr daemon failed: %s", |
d62a17ae | 1162 | strerror(errno)); |
1163 | exit(1); | |
4f04a76b DL |
1164 | } |
1165 | } else | |
dd8376fe | 1166 | zlog_set_level(ZLOG_DEST_STDOUT, MIN(gs.loglevel, LOG_DEBUG)); |
8b886ca7 | 1167 | |
a6810074 | 1168 | watchfrr_vty_init(); |
8b886ca7 | 1169 | |
eb05883f | 1170 | frr_vty_serv(); |
8b886ca7 | 1171 | |
8b886ca7 | 1172 | { |
a6810074 DL |
1173 | int i; |
1174 | struct daemon *tail = NULL; | |
1175 | ||
1176 | for (i = optind; i < argc; i++) { | |
1177 | struct daemon *dmn; | |
1178 | ||
1179 | if (!(dmn = (struct daemon *)calloc(1, sizeof(*dmn)))) { | |
1180 | fprintf(stderr, "calloc(1,%u) failed: %s\n", | |
d62a17ae | 1181 | (u_int)sizeof(*dmn), |
a6810074 DL |
1182 | safe_strerror(errno)); |
1183 | return 1; | |
1184 | } | |
1185 | dmn->name = dmn->restart.name = argv[i]; | |
1186 | dmn->state = DAEMON_INIT; | |
1187 | gs.numdaemons++; | |
1188 | gs.numdown++; | |
1189 | dmn->fd = -1; | |
66e78ae6 | 1190 | dmn->t_wakeup = NULL; |
d62a17ae | 1191 | thread_add_timer_msec(master, wakeup_init, dmn, |
1192 | 100 + (random() % 900), | |
66e78ae6 | 1193 | &dmn->t_wakeup); |
a6810074 DL |
1194 | dmn->restart.interval = gs.min_restart_interval; |
1195 | if (tail) | |
1196 | tail->next = dmn; | |
1197 | else | |
1198 | gs.daemons = dmn; | |
1199 | tail = dmn; | |
1200 | ||
f168b713 | 1201 | if (!strcmp(dmn->name, special)) |
a6810074 DL |
1202 | gs.special = dmn; |
1203 | } | |
1204 | } | |
1205 | if (!gs.daemons) { | |
1206 | fputs("Must specify one or more daemons to monitor.\n", stderr); | |
4f04a76b | 1207 | frr_help_exit(1); |
a6810074 | 1208 | } |
f168b713 DL |
1209 | if (!watch_only && !gs.special) { |
1210 | fprintf(stderr, "\"%s\" daemon must be in daemon list\n", | |
1211 | special); | |
4f04a76b | 1212 | frr_help_exit(1); |
8b886ca7 | 1213 | } |
8b886ca7 | 1214 | |
a6810074 DL |
1215 | /* Make sure we're not already running. */ |
1216 | pid_output(pidfile); | |
1217 | ||
1218 | /* Announce which daemons are being monitored. */ | |
1219 | { | |
1220 | struct daemon *dmn; | |
1221 | size_t len = 0; | |
1222 | ||
1223 | for (dmn = gs.daemons; dmn; dmn = dmn->next) | |
1224 | len += strlen(dmn->name) + 1; | |
1225 | ||
1226 | { | |
1227 | char buf[len + 1]; | |
1228 | char *p = buf; | |
1229 | ||
1230 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { | |
1231 | if (p != buf) | |
1232 | *p++ = ' '; | |
1233 | strcpy(p, dmn->name); | |
1234 | p += strlen(p); | |
1235 | } | |
f168b713 DL |
1236 | zlog_notice("%s %s watching [%s]%s", progname, |
1237 | FRR_VERSION, buf, | |
1238 | watch_only ? ", monitor mode" : ""); | |
a6810074 DL |
1239 | } |
1240 | } | |
8b886ca7 | 1241 | |
a6810074 DL |
1242 | { |
1243 | struct thread thread; | |
1244 | ||
1245 | while (thread_fetch(master, &thread)) | |
1246 | thread_call(&thread); | |
1247 | } | |
8b886ca7 | 1248 | |
a6810074 DL |
1249 | systemd_send_stopping(); |
1250 | /* Not reached. */ | |
1251 | return 0; | |
8b886ca7 | 1252 | } |