]>
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" |
4f04a76b | 28 | #include "libfrr.h" |
b647dc2a | 29 | #include "lib_errors.h" |
0bdeb5e5 | 30 | #include "zlog_targets.h" |
5920b3eb | 31 | #include "network.h" |
33606a15 | 32 | #include "printfrr.h" |
95c4aff2 | 33 | |
6f594023 | 34 | #include <getopt.h> |
a365534f | 35 | #include <sys/un.h> |
36 | #include <sys/wait.h> | |
837d16cc | 37 | #include <memory.h> |
651415bd | 38 | #include <systemd.h> |
8b886ca7 | 39 | |
9473e340 | 40 | #include "watchfrr.h" |
b647dc2a | 41 | #include "watchfrr_errors.h" |
95c4aff2 | 42 | |
8b886ca7 | 43 | #ifndef MIN |
44 | #define MIN(X,Y) (((X) <= (Y)) ? (X) : (Y)) | |
45 | #endif | |
46 | ||
47 | /* Macros to help randomize timers. */ | |
5920b3eb | 48 | #define JITTER(X) ((frr_weak_random() % ((X)+1))-((X)/2)) |
8b886ca7 | 49 | #define FUZZY(X) ((X)+JITTER((X)/20)) |
50 | ||
51 | #define DEFAULT_PERIOD 5 | |
0a64aff6 | 52 | #define DEFAULT_TIMEOUT 90 |
8b886ca7 | 53 | #define DEFAULT_RESTART_TIMEOUT 20 |
54 | #define DEFAULT_LOGLEVEL LOG_INFO | |
55 | #define DEFAULT_MIN_RESTART 60 | |
56 | #define DEFAULT_MAX_RESTART 600 | |
8b886ca7 | 57 | |
3ec95567 DL |
58 | #define DEFAULT_RESTART_CMD WATCHFRR_SH_PATH " restart %s" |
59 | #define DEFAULT_START_CMD WATCHFRR_SH_PATH " start %s" | |
60 | #define DEFAULT_STOP_CMD WATCHFRR_SH_PATH " stop %s" | |
61 | ||
8b886ca7 | 62 | #define PING_TOKEN "PING" |
63 | ||
bf8d3d6a DL |
64 | DEFINE_MGROUP(WATCHFRR, "watchfrr"); |
65 | DEFINE_MTYPE_STATIC(WATCHFRR, WATCHFRR_DAEMON, "watchfrr daemon entry"); | |
0a7c7856 | 66 | |
55c72803 | 67 | /* Needs to be global, referenced somewhere inside libfrr. */ |
8b886ca7 | 68 | struct thread_master *master; |
69 | ||
f168b713 | 70 | static bool watch_only = false; |
8b886ca7 | 71 | |
a6810074 DL |
72 | typedef enum { |
73 | PHASE_NONE = 0, | |
c0e5cb52 | 74 | PHASE_INIT, |
a6810074 DL |
75 | PHASE_STOPS_PENDING, |
76 | PHASE_WAITING_DOWN, | |
77 | PHASE_ZEBRA_RESTART_PENDING, | |
78 | PHASE_WAITING_ZEBRA_UP | |
8b886ca7 | 79 | } restart_phase_t; |
80 | ||
2b64873d | 81 | static const char *const phase_str[] = { |
af568444 | 82 | "Idle", |
c0e5cb52 | 83 | "Startup", |
a6810074 DL |
84 | "Stop jobs running", |
85 | "Waiting for other daemons to come down", | |
86 | "Zebra restart job running", | |
87 | "Waiting for zebra to come up", | |
88 | "Start jobs running", | |
8b886ca7 | 89 | }; |
90 | ||
91 | #define PHASE_TIMEOUT (3*gs.restart_timeout) | |
5c9d1c83 | 92 | #define STARTUP_TIMEOUT 55 * 1000 |
8b886ca7 | 93 | |
a6810074 DL |
94 | struct restart_info { |
95 | const char *name; | |
96 | const char *what; | |
97 | pid_t pid; | |
98 | struct timeval time; | |
99 | long interval; | |
100 | struct thread *t_kill; | |
101 | int kills; | |
098e240f | 102 | }; |
103 | ||
a6810074 | 104 | static struct global_state { |
a6810074 DL |
105 | restart_phase_t phase; |
106 | struct thread *t_phase_hanging; | |
5c9d1c83 | 107 | struct thread *t_startup_timeout; |
a6810074 DL |
108 | const char *vtydir; |
109 | long period; | |
110 | long timeout; | |
111 | long restart_timeout; | |
112 | long min_restart_interval; | |
113 | long max_restart_interval; | |
a6810074 DL |
114 | struct daemon *daemons; |
115 | const char *restart_command; | |
116 | const char *start_command; | |
117 | const char *stop_command; | |
118 | struct restart_info restart; | |
a6810074 | 119 | int loglevel; |
d62a17ae | 120 | struct daemon *special; /* points to zebra when doing phased restart */ |
a6810074 DL |
121 | int numdaemons; |
122 | int numpids; | |
d62a17ae | 123 | int numdown; /* # of daemons that are not UP or UNRESPONSIVE */ |
8b886ca7 | 124 | } gs = { |
c0e5cb52 | 125 | .phase = PHASE_INIT, |
64a249ad | 126 | .vtydir = frr_vtydir, |
d62a17ae | 127 | .period = 1000 * DEFAULT_PERIOD, |
128 | .timeout = DEFAULT_TIMEOUT, | |
129 | .restart_timeout = DEFAULT_RESTART_TIMEOUT, | |
130 | .loglevel = DEFAULT_LOGLEVEL, | |
131 | .min_restart_interval = DEFAULT_MIN_RESTART, | |
132 | .max_restart_interval = DEFAULT_MAX_RESTART, | |
3ec95567 DL |
133 | .restart_command = DEFAULT_RESTART_CMD, |
134 | .start_command = DEFAULT_START_CMD, | |
135 | .stop_command = DEFAULT_STOP_CMD, | |
d62a17ae | 136 | }; |
a6810074 DL |
137 | |
138 | typedef enum { | |
139 | DAEMON_INIT, | |
140 | DAEMON_DOWN, | |
141 | DAEMON_CONNECTING, | |
142 | DAEMON_UP, | |
143 | DAEMON_UNRESPONSIVE | |
8b886ca7 | 144 | } daemon_state_t; |
145 | ||
d62a17ae | 146 | #define IS_UP(DMN) \ |
147 | (((DMN)->state == DAEMON_UP) || ((DMN)->state == DAEMON_UNRESPONSIVE)) | |
8b886ca7 | 148 | |
2b64873d | 149 | static const char *const state_str[] = { |
d62a17ae | 150 | "Init", "Down", "Connecting", "Up", "Unresponsive", |
8b886ca7 | 151 | }; |
152 | ||
153 | struct daemon { | |
a6810074 DL |
154 | const char *name; |
155 | daemon_state_t state; | |
156 | int fd; | |
157 | struct timeval echo_sent; | |
d7c0a89a | 158 | unsigned int connect_tries; |
a6810074 DL |
159 | struct thread *t_wakeup; |
160 | struct thread *t_read; | |
161 | struct thread *t_write; | |
162 | struct daemon *next; | |
163 | struct restart_info restart; | |
cc53b605 DS |
164 | |
165 | /* | |
166 | * For a given daemon, if we've turned on ignore timeouts | |
167 | * ignore the timeout value and assume everything is ok | |
168 | * This is for daemon debugging w/ gdb after we have started | |
169 | * FRR and realize we have something that needs to be looked | |
170 | * at | |
171 | */ | |
172 | bool ignore_timeout; | |
8b886ca7 | 173 | }; |
174 | ||
9272302b DL |
175 | #define OPTION_MINRESTART 2000 |
176 | #define OPTION_MAXRESTART 2001 | |
f168b713 | 177 | #define OPTION_DRY 2002 |
33606a15 | 178 | #define OPTION_NETNS 2003 |
9272302b | 179 | |
a6810074 DL |
180 | static const struct option longopts[] = { |
181 | {"daemon", no_argument, NULL, 'd'}, | |
182 | {"statedir", required_argument, NULL, 'S'}, | |
a6810074 DL |
183 | {"loglevel", required_argument, NULL, 'l'}, |
184 | {"interval", required_argument, NULL, 'i'}, | |
185 | {"timeout", required_argument, NULL, 't'}, | |
186 | {"restart-timeout", required_argument, NULL, 'T'}, | |
187 | {"restart", required_argument, NULL, 'r'}, | |
188 | {"start-command", required_argument, NULL, 's'}, | |
189 | {"kill-command", required_argument, NULL, 'k'}, | |
f168b713 | 190 | {"dry", no_argument, NULL, OPTION_DRY}, |
d62a17ae | 191 | {"min-restart-interval", required_argument, NULL, OPTION_MINRESTART}, |
192 | {"max-restart-interval", required_argument, NULL, OPTION_MAXRESTART}, | |
a6810074 DL |
193 | {"pid-file", required_argument, NULL, 'p'}, |
194 | {"blank-string", required_argument, NULL, 'b'}, | |
33606a15 DL |
195 | #ifdef GNU_LINUX |
196 | {"netns", optional_argument, NULL, OPTION_NETNS}, | |
197 | #endif | |
a6810074 DL |
198 | {"help", no_argument, NULL, 'h'}, |
199 | {"version", no_argument, NULL, 'v'}, | |
d62a17ae | 200 | {NULL, 0, NULL, 0}}; |
8b886ca7 | 201 | |
202 | static int try_connect(struct daemon *dmn); | |
203 | static int wakeup_send_echo(struct thread *t_wakeup); | |
204 | static void try_restart(struct daemon *dmn); | |
205 | static void phase_check(void); | |
75f8b0e4 | 206 | static void restart_done(struct daemon *dmn); |
8b886ca7 | 207 | |
4f04a76b | 208 | static const char *progname; |
cc53b605 DS |
209 | |
210 | void watchfrr_set_ignore_daemon(struct vty *vty, const char *dname, bool ignore) | |
211 | { | |
212 | struct daemon *dmn; | |
213 | ||
214 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { | |
215 | if (strncmp(dmn->name, dname, strlen(dmn->name)) == 0) | |
216 | break; | |
217 | } | |
218 | ||
219 | if (dmn) { | |
220 | dmn->ignore_timeout = ignore; | |
221 | vty_out(vty, "%s switching to %s\n", dmn->name, | |
222 | ignore ? "ignore" : "watch"); | |
223 | } else | |
224 | vty_out(vty, "%s is not configured for running at the moment", | |
225 | dname); | |
226 | } | |
227 | ||
4f04a76b | 228 | static void printhelp(FILE *target) |
8b886ca7 | 229 | { |
d62a17ae | 230 | fprintf(target, |
231 | "Usage : %s [OPTION...] <daemon name> ...\n\n\ | |
9473e340 | 232 | Watchdog program to monitor status of frr daemons and try to restart\n\ |
8b886ca7 | 233 | them if they are down or unresponsive. It determines whether a daemon is\n\ |
234 | up based on whether it can connect to the daemon's vty unix stream socket.\n\ | |
235 | It then repeatedly sends echo commands over that socket to determine whether\n\ | |
236 | the daemon is responsive. If the daemon crashes, we will receive an EOF\n\ | |
237 | on the socket connection and know immediately that the daemon is down.\n\n\ | |
238 | The daemons to be monitored should be listed on the command line.\n\n\ | |
8b886ca7 | 239 | In order to avoid attempting to restart the daemons in a fast loop,\n\ |
240 | the -m and -M options allow you to control the minimum delay between\n\ | |
241 | restart commands. The minimum restart delay is recalculated each time\n\ | |
242 | a restart is attempted: if the time since the last restart attempt exceeds\n\ | |
243 | twice the -M value, then the restart delay is set to the -m value.\n\ | |
d62a17ae | 244 | Otherwise, the interval is doubled (but capped at the -M value).\n\n", |
f168b713 | 245 | progname); |
e757c940 | 246 | |
d62a17ae | 247 | fprintf(target, |
248 | "Options:\n\ | |
8b886ca7 | 249 | -d, --daemon Run in daemon mode. In this mode, error messages are sent\n\ |
250 | to syslog instead of stdout.\n\ | |
251 | -S, --statedir Set the vty socket directory (default is %s)\n\ | |
33606a15 DL |
252 | -N, --pathspace Insert prefix into config & socket paths\n" |
253 | #ifdef GNU_LINUX | |
254 | " --netns Create and/or use Linux network namespace. If no name is\n" | |
255 | " given, uses the value from `-N`.\n" | |
256 | #endif | |
257 | "-l, --loglevel Set the logging level (default is %d).\n\ | |
8b886ca7 | 258 | The value should range from %d (LOG_EMERG) to %d (LOG_DEBUG),\n\ |
259 | but it can be set higher than %d if extra-verbose debugging\n\ | |
260 | messages are desired.\n\ | |
9272302b | 261 | --min-restart-interval\n\ |
8b886ca7 | 262 | Set the minimum seconds to wait between invocations of daemon\n\ |
263 | restart commands (default is %d).\n\ | |
9272302b | 264 | --max-restart-interval\n\ |
8b886ca7 | 265 | Set the maximum seconds to wait between invocations of daemon\n\ |
266 | restart commands (default is %d).\n\ | |
267 | -i, --interval Set the status polling interval in seconds (default is %d)\n\ | |
268 | -t, --timeout Set the unresponsiveness timeout in seconds (default is %d)\n\ | |
269 | -T, --restart-timeout\n\ | |
270 | Set the restart (kill) timeout in seconds (default is %d).\n\ | |
271 | If any background jobs are still running after this much\n\ | |
272 | time has elapsed, they will be killed.\n\ | |
273 | -r, --restart Supply a Bourne shell command to use to restart a single\n\ | |
274 | daemon. The command string should include '%%s' where the\n\ | |
275 | name of the daemon should be substituted.\n\ | |
3ec95567 | 276 | (default: '%s')\n\ |
8b886ca7 | 277 | -s, --start-command\n\ |
278 | Supply a Bourne shell to command to use to start a single\n\ | |
279 | daemon. The command string should include '%%s' where the\n\ | |
280 | name of the daemon should be substituted.\n\ | |
3ec95567 | 281 | (default: '%s')\n\ |
8b886ca7 | 282 | -k, --kill-command\n\ |
283 | Supply a Bourne shell to command to use to stop a single\n\ | |
284 | daemon. The command string should include '%%s' where the\n\ | |
285 | name of the daemon should be substituted.\n\ | |
3ec95567 | 286 | (default: '%s')\n\ |
f168b713 | 287 | --dry Do not start or restart anything, just log.\n\ |
8b886ca7 | 288 | -p, --pid-file Set process identifier file name\n\ |
0a7c7856 | 289 | (default is %s/watchfrr.pid).\n\ |
c8b40f86 | 290 | -b, --blank-string\n\ |
291 | When the supplied argument string is found in any of the\n\ | |
f168b713 | 292 | various shell command arguments (-r, -s, or -k), replace\n\ |
c8b40f86 | 293 | it with a space. This is an ugly hack to circumvent problems\n\ |
294 | passing command-line arguments with embedded spaces.\n\ | |
8b886ca7 | 295 | -v, --version Print program version\n\ |
d62a17ae | 296 | -h, --help Display this help and exit\n", |
64a249ad | 297 | frr_vtydir, DEFAULT_LOGLEVEL, LOG_EMERG, LOG_DEBUG, LOG_DEBUG, |
d62a17ae | 298 | DEFAULT_MIN_RESTART, DEFAULT_MAX_RESTART, DEFAULT_PERIOD, |
3ec95567 DL |
299 | DEFAULT_TIMEOUT, DEFAULT_RESTART_TIMEOUT, |
300 | DEFAULT_RESTART_CMD, DEFAULT_START_CMD, DEFAULT_STOP_CMD, | |
301 | frr_vtydir); | |
8b886ca7 | 302 | } |
303 | ||
a6810074 | 304 | static pid_t run_background(char *shell_cmd) |
8b886ca7 | 305 | { |
a6810074 DL |
306 | pid_t child; |
307 | ||
308 | switch (child = fork()) { | |
309 | case -1: | |
450971aa | 310 | flog_err_sys(EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
311 | "fork failed, cannot run command [%s]: %s", |
312 | shell_cmd, safe_strerror(errno)); | |
a6810074 DL |
313 | return -1; |
314 | case 0: | |
315 | /* Child process. */ | |
d62a17ae | 316 | /* Use separate process group so child processes can be killed |
317 | * easily. */ | |
a6810074 | 318 | if (setpgid(0, 0) < 0) |
957cfa24 | 319 | zlog_warn("setpgid(0,0) failed: %s", |
a6810074 DL |
320 | safe_strerror(errno)); |
321 | { | |
322 | char shell[] = "sh"; | |
323 | char dashc[] = "-c"; | |
d62a17ae | 324 | char *const argv[4] = {shell, dashc, shell_cmd, NULL}; |
a6810074 | 325 | execv("/bin/sh", argv); |
450971aa | 326 | flog_err_sys(EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
327 | "execv(/bin/sh -c '%s') failed: %s", |
328 | shell_cmd, safe_strerror(errno)); | |
a6810074 DL |
329 | _exit(127); |
330 | } | |
331 | default: | |
332 | /* Parent process: we will reap the child later. */ | |
c3f65458 QY |
333 | zlog_info("Forked background command [pid %d]: %s", (int)child, |
334 | shell_cmd); | |
a6810074 DL |
335 | return child; |
336 | } | |
8b886ca7 | 337 | } |
338 | ||
a6810074 DL |
339 | static struct timeval *time_elapsed(struct timeval *result, |
340 | const struct timeval *start_time) | |
8b886ca7 | 341 | { |
a6810074 DL |
342 | gettimeofday(result, NULL); |
343 | result->tv_sec -= start_time->tv_sec; | |
344 | result->tv_usec -= start_time->tv_usec; | |
345 | while (result->tv_usec < 0) { | |
346 | result->tv_usec += 1000000L; | |
347 | result->tv_sec--; | |
348 | } | |
349 | return result; | |
8b886ca7 | 350 | } |
351 | ||
a6810074 | 352 | static int restart_kill(struct thread *t_kill) |
8b886ca7 | 353 | { |
a6810074 DL |
354 | struct restart_info *restart = THREAD_ARG(t_kill); |
355 | struct timeval delay; | |
356 | ||
357 | time_elapsed(&delay, &restart->time); | |
d62a17ae | 358 | zlog_warn( |
957cfa24 | 359 | "%s %s child process %d still running after %ld seconds, sending signal %d", |
d62a17ae | 360 | restart->what, restart->name, (int)restart->pid, |
361 | (long)delay.tv_sec, (restart->kills ? SIGKILL : SIGTERM)); | |
a6810074 DL |
362 | kill(-restart->pid, (restart->kills ? SIGKILL : SIGTERM)); |
363 | restart->kills++; | |
66e78ae6 QY |
364 | restart->t_kill = NULL; |
365 | thread_add_timer(master, restart_kill, restart, gs.restart_timeout, | |
366 | &restart->t_kill); | |
a6810074 | 367 | return 0; |
8b886ca7 | 368 | } |
369 | ||
a6810074 | 370 | static struct restart_info *find_child(pid_t child) |
8b886ca7 | 371 | { |
f168b713 | 372 | struct daemon *dmn; |
7c265f7d CF |
373 | if (gs.restart.pid == child) |
374 | return &gs.restart; | |
375 | ||
f168b713 DL |
376 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { |
377 | if (dmn->restart.pid == child) | |
378 | return &dmn->restart; | |
a6810074 DL |
379 | } |
380 | return NULL; | |
8b886ca7 | 381 | } |
382 | ||
a6810074 | 383 | static void sigchild(void) |
8b886ca7 | 384 | { |
a6810074 DL |
385 | pid_t child; |
386 | int status; | |
387 | const char *name; | |
388 | const char *what; | |
389 | struct restart_info *restart; | |
75f8b0e4 | 390 | struct daemon *dmn; |
a6810074 DL |
391 | |
392 | switch (child = waitpid(-1, &status, WNOHANG)) { | |
393 | case -1: | |
450971aa | 394 | flog_err_sys(EC_LIB_SYSTEM_CALL, "waitpid failed: %s", |
09c866e3 | 395 | safe_strerror(errno)); |
a6810074 DL |
396 | return; |
397 | case 0: | |
398 | zlog_warn("SIGCHLD received, but waitpid did not reap a child"); | |
399 | return; | |
400 | } | |
401 | ||
402 | if (child == integrated_write_pid) { | |
403 | integrated_write_sigchld(status); | |
404 | return; | |
405 | } | |
406 | ||
407 | if ((restart = find_child(child)) != NULL) { | |
408 | name = restart->name; | |
409 | what = restart->what; | |
410 | restart->pid = 0; | |
411 | gs.numpids--; | |
b3d6bc6e MS |
412 | thread_cancel(&restart->t_kill); |
413 | ||
d62a17ae | 414 | /* Update restart time to reflect the time the command |
415 | * completed. */ | |
a6810074 DL |
416 | gettimeofday(&restart->time, NULL); |
417 | } else { | |
09c866e3 | 418 | flog_err_sys( |
450971aa | 419 | EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
420 | "waitpid returned status for an unknown child process %d", |
421 | (int)child); | |
a6810074 DL |
422 | name = "(unknown)"; |
423 | what = "background"; | |
424 | } | |
425 | if (WIFSTOPPED(status)) | |
957cfa24 | 426 | zlog_warn("%s %s process %d is stopped", what, name, |
d62a17ae | 427 | (int)child); |
a6810074 | 428 | else if (WIFSIGNALED(status)) |
d62a17ae | 429 | zlog_warn("%s %s process %d terminated due to signal %d", what, |
430 | name, (int)child, WTERMSIG(status)); | |
a6810074 DL |
431 | else if (WIFEXITED(status)) { |
432 | if (WEXITSTATUS(status) != 0) | |
d62a17ae | 433 | zlog_warn( |
434 | "%s %s process %d exited with non-zero status %d", | |
435 | what, name, (int)child, WEXITSTATUS(status)); | |
75f8b0e4 | 436 | else { |
a6810074 DL |
437 | zlog_debug("%s %s process %d exited normally", what, |
438 | name, (int)child); | |
75f8b0e4 DL |
439 | |
440 | if (restart && restart != &gs.restart) { | |
441 | dmn = container_of(restart, struct daemon, | |
442 | restart); | |
443 | restart_done(dmn); | |
444 | } else if (restart) | |
445 | for (dmn = gs.daemons; dmn; dmn = dmn->next) | |
446 | restart_done(dmn); | |
447 | } | |
a6810074 | 448 | } else |
09c866e3 | 449 | flog_err_sys( |
450971aa | 450 | EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
451 | "cannot interpret %s %s process %d wait status 0x%x", |
452 | what, name, (int)child, status); | |
a6810074 | 453 | phase_check(); |
8b886ca7 | 454 | } |
455 | ||
d62a17ae | 456 | static int run_job(struct restart_info *restart, const char *cmdtype, |
457 | const char *command, int force, int update_interval) | |
8b886ca7 | 458 | { |
a6810074 DL |
459 | struct timeval delay; |
460 | ||
461 | if (gs.loglevel > LOG_DEBUG + 1) | |
462 | zlog_debug("attempting to %s %s", cmdtype, restart->name); | |
463 | ||
464 | if (restart->pid) { | |
465 | if (gs.loglevel > LOG_DEBUG + 1) | |
d62a17ae | 466 | zlog_debug( |
467 | "cannot %s %s, previous pid %d still running", | |
468 | cmdtype, restart->name, (int)restart->pid); | |
a6810074 DL |
469 | return -1; |
470 | } | |
471 | ||
b3ee8bcc DS |
472 | #if defined HAVE_SYSTEMD |
473 | char buffer[512]; | |
474 | ||
475 | snprintf(buffer, sizeof(buffer), "restarting %s", restart->name); | |
476 | systemd_send_status(buffer); | |
477 | #endif | |
478 | ||
d62a17ae | 479 | /* Note: time_elapsed test must come before the force test, since we |
480 | need | |
a6810074 DL |
481 | to make sure that delay is initialized for use below in updating the |
482 | restart interval. */ | |
483 | if ((time_elapsed(&delay, &restart->time)->tv_sec < restart->interval) | |
484 | && !force) { | |
b3ee8bcc | 485 | |
a6810074 | 486 | if (gs.loglevel > LOG_DEBUG + 1) |
d62a17ae | 487 | zlog_debug( |
3efd0893 | 488 | "postponing %s %s: elapsed time %ld < retry interval %ld", |
d62a17ae | 489 | cmdtype, restart->name, (long)delay.tv_sec, |
490 | restart->interval); | |
a6810074 DL |
491 | return -1; |
492 | } | |
493 | ||
494 | gettimeofday(&restart->time, NULL); | |
495 | restart->kills = 0; | |
496 | { | |
497 | char cmd[strlen(command) + strlen(restart->name) + 1]; | |
498 | snprintf(cmd, sizeof(cmd), command, restart->name); | |
499 | if ((restart->pid = run_background(cmd)) > 0) { | |
66e78ae6 | 500 | restart->t_kill = NULL; |
d62a17ae | 501 | thread_add_timer(master, restart_kill, restart, |
502 | gs.restart_timeout, &restart->t_kill); | |
a6810074 DL |
503 | restart->what = cmdtype; |
504 | gs.numpids++; | |
505 | } else | |
506 | restart->pid = 0; | |
507 | } | |
508 | ||
b3ee8bcc DS |
509 | #if defined HAVE_SYSTEMD |
510 | systemd_send_status("FRR Operational"); | |
511 | #endif | |
a6810074 DL |
512 | /* Calculate the new restart interval. */ |
513 | if (update_interval) { | |
514 | if (delay.tv_sec > 2 * gs.max_restart_interval) | |
515 | restart->interval = gs.min_restart_interval; | |
516 | else if ((restart->interval *= 2) > gs.max_restart_interval) | |
517 | restart->interval = gs.max_restart_interval; | |
518 | if (gs.loglevel > LOG_DEBUG + 1) | |
519 | zlog_debug("restart %s interval is now %ld", | |
520 | restart->name, restart->interval); | |
521 | } | |
522 | return restart->pid; | |
8b886ca7 | 523 | } |
524 | ||
d62a17ae | 525 | #define SET_READ_HANDLER(DMN) \ |
526 | do { \ | |
527 | (DMN)->t_read = NULL; \ | |
528 | thread_add_read(master, handle_read, (DMN), (DMN)->fd, \ | |
529 | &(DMN)->t_read); \ | |
530 | } while (0); | |
531 | ||
532 | #define SET_WAKEUP_DOWN(DMN) \ | |
533 | do { \ | |
534 | (DMN)->t_wakeup = NULL; \ | |
535 | thread_add_timer_msec(master, wakeup_down, (DMN), \ | |
536 | FUZZY(gs.period), &(DMN)->t_wakeup); \ | |
537 | } while (0); | |
538 | ||
539 | #define SET_WAKEUP_UNRESPONSIVE(DMN) \ | |
540 | do { \ | |
541 | (DMN)->t_wakeup = NULL; \ | |
542 | thread_add_timer_msec(master, wakeup_unresponsive, (DMN), \ | |
543 | FUZZY(gs.period), &(DMN)->t_wakeup); \ | |
544 | } while (0); | |
545 | ||
546 | #define SET_WAKEUP_ECHO(DMN) \ | |
547 | do { \ | |
548 | (DMN)->t_wakeup = NULL; \ | |
549 | thread_add_timer_msec(master, wakeup_send_echo, (DMN), \ | |
550 | FUZZY(gs.period), &(DMN)->t_wakeup); \ | |
551 | } while (0); | |
8b886ca7 | 552 | |
a6810074 | 553 | static int wakeup_down(struct thread *t_wakeup) |
8b886ca7 | 554 | { |
a6810074 DL |
555 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
556 | ||
557 | dmn->t_wakeup = NULL; | |
558 | if (try_connect(dmn) < 0) | |
559 | SET_WAKEUP_DOWN(dmn); | |
560 | if ((dmn->connect_tries > 1) && (dmn->state != DAEMON_UP)) | |
561 | try_restart(dmn); | |
562 | return 0; | |
8b886ca7 | 563 | } |
564 | ||
a6810074 | 565 | static int wakeup_init(struct thread *t_wakeup) |
8b886ca7 | 566 | { |
a6810074 DL |
567 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
568 | ||
569 | dmn->t_wakeup = NULL; | |
570 | if (try_connect(dmn) < 0) { | |
c3f65458 QY |
571 | zlog_info( |
572 | "%s state -> down : initial connection attempt failed", | |
573 | dmn->name); | |
a6810074 DL |
574 | dmn->state = DAEMON_DOWN; |
575 | } | |
c0e5cb52 | 576 | phase_check(); |
a6810074 | 577 | return 0; |
8b886ca7 | 578 | } |
579 | ||
75f8b0e4 DL |
580 | static void restart_done(struct daemon *dmn) |
581 | { | |
582 | if (dmn->state != DAEMON_DOWN) { | |
3f391bec DS |
583 | zlog_warn( |
584 | "Daemon: %s: is in %s state but expected it to be in DAEMON_DOWN state", | |
585 | dmn->name, state_str[dmn->state]); | |
75f8b0e4 DL |
586 | return; |
587 | } | |
28ef0ee1 | 588 | THREAD_OFF(dmn->t_wakeup); |
50478845 | 589 | |
75f8b0e4 DL |
590 | if (try_connect(dmn) < 0) |
591 | SET_WAKEUP_DOWN(dmn); | |
592 | } | |
593 | ||
a6810074 | 594 | static void daemon_down(struct daemon *dmn, const char *why) |
8b886ca7 | 595 | { |
a6810074 | 596 | if (IS_UP(dmn) || (dmn->state == DAEMON_INIT)) |
1c50c1c0 QY |
597 | flog_err(EC_WATCHFRR_CONNECTION, "%s state -> down : %s", |
598 | dmn->name, why); | |
a6810074 DL |
599 | else if (gs.loglevel > LOG_DEBUG) |
600 | zlog_debug("%s still down : %s", dmn->name, why); | |
601 | if (IS_UP(dmn)) | |
602 | gs.numdown++; | |
603 | dmn->state = DAEMON_DOWN; | |
604 | if (dmn->fd >= 0) { | |
605 | close(dmn->fd); | |
606 | dmn->fd = -1; | |
607 | } | |
608 | THREAD_OFF(dmn->t_read); | |
609 | THREAD_OFF(dmn->t_write); | |
610 | THREAD_OFF(dmn->t_wakeup); | |
611 | if (try_connect(dmn) < 0) | |
612 | SET_WAKEUP_DOWN(dmn); | |
613 | phase_check(); | |
8b886ca7 | 614 | } |
615 | ||
a6810074 | 616 | static int handle_read(struct thread *t_read) |
8b886ca7 | 617 | { |
a6810074 DL |
618 | struct daemon *dmn = THREAD_ARG(t_read); |
619 | static const char resp[sizeof(PING_TOKEN) + 4] = PING_TOKEN "\n"; | |
620 | char buf[sizeof(resp) + 100]; | |
621 | ssize_t rc; | |
622 | struct timeval delay; | |
623 | ||
624 | dmn->t_read = NULL; | |
625 | if ((rc = read(dmn->fd, buf, sizeof(buf))) < 0) { | |
626 | char why[100]; | |
627 | ||
628 | if (ERRNO_IO_RETRY(errno)) { | |
629 | /* Pretend it never happened. */ | |
630 | SET_READ_HANDLER(dmn); | |
631 | return 0; | |
632 | } | |
633 | snprintf(why, sizeof(why), "unexpected read error: %s", | |
634 | safe_strerror(errno)); | |
635 | daemon_down(dmn, why); | |
636 | return 0; | |
8b886ca7 | 637 | } |
a6810074 DL |
638 | if (rc == 0) { |
639 | daemon_down(dmn, "read returned EOF"); | |
640 | return 0; | |
641 | } | |
642 | if (!dmn->echo_sent.tv_sec) { | |
643 | char why[sizeof(buf) + 100]; | |
644 | snprintf(why, sizeof(why), | |
645 | "unexpected read returns %d bytes: %.*s", (int)rc, | |
646 | (int)rc, buf); | |
647 | daemon_down(dmn, why); | |
648 | return 0; | |
8b886ca7 | 649 | } |
a6810074 DL |
650 | |
651 | /* We are expecting an echo response: is there any chance that the | |
652 | response would not be returned entirely in the first read? That | |
653 | seems inconceivable... */ | |
654 | if ((rc != sizeof(resp)) || memcmp(buf, resp, sizeof(resp))) { | |
655 | char why[100 + sizeof(buf)]; | |
656 | snprintf(why, sizeof(why), | |
3efd0893 | 657 | "read returned bad echo response of %d bytes (expecting %u): %.*s", |
d7c0a89a | 658 | (int)rc, (unsigned int)sizeof(resp), (int)rc, buf); |
a6810074 DL |
659 | daemon_down(dmn, why); |
660 | return 0; | |
661 | } | |
662 | ||
663 | time_elapsed(&delay, &dmn->echo_sent); | |
664 | dmn->echo_sent.tv_sec = 0; | |
665 | if (dmn->state == DAEMON_UNRESPONSIVE) { | |
666 | if (delay.tv_sec < gs.timeout) { | |
667 | dmn->state = DAEMON_UP; | |
d62a17ae | 668 | zlog_warn( |
3efd0893 | 669 | "%s state -> up : echo response received after %ld.%06ld seconds", |
d62a17ae | 670 | dmn->name, (long)delay.tv_sec, |
671 | (long)delay.tv_usec); | |
a6810074 | 672 | } else |
d62a17ae | 673 | zlog_warn( |
3efd0893 | 674 | "%s: slow echo response finally received after %ld.%06ld seconds", |
d62a17ae | 675 | dmn->name, (long)delay.tv_sec, |
676 | (long)delay.tv_usec); | |
a6810074 DL |
677 | } else if (gs.loglevel > LOG_DEBUG + 1) |
678 | zlog_debug("%s: echo response received after %ld.%06ld seconds", | |
679 | dmn->name, (long)delay.tv_sec, (long)delay.tv_usec); | |
680 | ||
681 | SET_READ_HANDLER(dmn); | |
b3d6bc6e | 682 | thread_cancel(&dmn->t_wakeup); |
a6810074 DL |
683 | SET_WAKEUP_ECHO(dmn); |
684 | ||
685 | return 0; | |
8b886ca7 | 686 | } |
687 | ||
207e0d7a DS |
688 | /* |
689 | * Wait till we notice that all daemons are ready before | |
690 | * we send we are ready to systemd | |
691 | */ | |
5c9d1c83 | 692 | static void daemon_send_ready(int exitcode) |
207e0d7a | 693 | { |
5c9d1c83 | 694 | FILE *fp; |
a6810074 | 695 | static int sent = 0; |
43e587c1 | 696 | char started[1024]; |
207e0d7a | 697 | |
5c9d1c83 DL |
698 | if (sent) |
699 | return; | |
700 | ||
701 | if (exitcode == 0) | |
0a7c7856 | 702 | zlog_notice("all daemons up, doing startup-complete notify"); |
5c9d1c83 DL |
703 | else if (gs.numdown < gs.numdaemons) |
704 | flog_err(EC_WATCHFRR_CONNECTION, | |
3efd0893 | 705 | "startup did not complete within timeout (%d/%d daemons running)", |
5c9d1c83 DL |
706 | gs.numdaemons - gs.numdown, gs.numdaemons); |
707 | else { | |
708 | flog_err(EC_WATCHFRR_CONNECTION, | |
3efd0893 | 709 | "all configured daemons failed to start -- exiting watchfrr"); |
5c9d1c83 DL |
710 | exit(exitcode); |
711 | ||
712 | } | |
0a7c7856 | 713 | |
5c9d1c83 DL |
714 | frr_detach(); |
715 | ||
33606a15 | 716 | snprintf(started, sizeof(started), "%s/%s", frr_vtydir, |
3c649c71 DS |
717 | "watchfrr.started"); |
718 | fp = fopen(started, "w"); | |
5c9d1c83 DL |
719 | if (fp) |
720 | fclose(fp); | |
60bd2534 | 721 | #if defined HAVE_SYSTEMD |
5c9d1c83 | 722 | systemd_send_started(master, 0); |
b3ee8bcc | 723 | systemd_send_status("FRR Operational"); |
60bd2534 | 724 | #endif |
5c9d1c83 | 725 | sent = 1; |
207e0d7a DS |
726 | } |
727 | ||
a6810074 | 728 | static void daemon_up(struct daemon *dmn, const char *why) |
8b886ca7 | 729 | { |
a6810074 DL |
730 | dmn->state = DAEMON_UP; |
731 | gs.numdown--; | |
732 | dmn->connect_tries = 0; | |
733 | zlog_notice("%s state -> up : %s", dmn->name, why); | |
5c9d1c83 DL |
734 | if (gs.numdown == 0) |
735 | daemon_send_ready(0); | |
a8cbb8b3 | 736 | SET_WAKEUP_ECHO(dmn); |
a6810074 | 737 | phase_check(); |
8b886ca7 | 738 | } |
739 | ||
a6810074 | 740 | static int check_connect(struct thread *t_write) |
8b886ca7 | 741 | { |
a6810074 DL |
742 | struct daemon *dmn = THREAD_ARG(t_write); |
743 | int sockerr; | |
744 | socklen_t reslen = sizeof(sockerr); | |
745 | ||
746 | dmn->t_write = NULL; | |
747 | if (getsockopt(dmn->fd, SOL_SOCKET, SO_ERROR, (char *)&sockerr, &reslen) | |
748 | < 0) { | |
749 | zlog_warn("%s: check_connect: getsockopt failed: %s", dmn->name, | |
750 | safe_strerror(errno)); | |
751 | daemon_down(dmn, | |
752 | "getsockopt failed checking connection success"); | |
753 | return 0; | |
754 | } | |
755 | if ((reslen == sizeof(sockerr)) && sockerr) { | |
756 | char why[100]; | |
d62a17ae | 757 | snprintf( |
758 | why, sizeof(why), | |
759 | "getsockopt reports that connection attempt failed: %s", | |
760 | safe_strerror(sockerr)); | |
a6810074 DL |
761 | daemon_down(dmn, why); |
762 | return 0; | |
763 | } | |
764 | ||
765 | daemon_up(dmn, "delayed connect succeeded"); | |
766 | return 0; | |
8b886ca7 | 767 | } |
768 | ||
a6810074 | 769 | static int wakeup_connect_hanging(struct thread *t_wakeup) |
8b886ca7 | 770 | { |
a6810074 DL |
771 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
772 | char why[100]; | |
773 | ||
774 | dmn->t_wakeup = NULL; | |
775 | snprintf(why, sizeof(why), | |
776 | "connection attempt timed out after %ld seconds", gs.timeout); | |
777 | daemon_down(dmn, why); | |
778 | return 0; | |
8b886ca7 | 779 | } |
780 | ||
781 | /* Making connection to protocol daemon. */ | |
a6810074 | 782 | static int try_connect(struct daemon *dmn) |
8b886ca7 | 783 | { |
a6810074 DL |
784 | int sock; |
785 | struct sockaddr_un addr; | |
786 | socklen_t len; | |
787 | ||
788 | if (gs.loglevel > LOG_DEBUG + 1) | |
789 | zlog_debug("%s: attempting to connect", dmn->name); | |
790 | dmn->connect_tries++; | |
791 | ||
792 | memset(&addr, 0, sizeof(struct sockaddr_un)); | |
793 | addr.sun_family = AF_UNIX; | |
d62a17ae | 794 | snprintf(addr.sun_path, sizeof(addr.sun_path), "%s/%s.vty", gs.vtydir, |
795 | dmn->name); | |
6f0e3f6e | 796 | #ifdef HAVE_STRUCT_SOCKADDR_UN_SUN_LEN |
a6810074 | 797 | len = addr.sun_len = SUN_LEN(&addr); |
8b886ca7 | 798 | #else |
a6810074 | 799 | len = sizeof(addr.sun_family) + strlen(addr.sun_path); |
d62a17ae | 800 | #endif /* HAVE_STRUCT_SOCKADDR_UN_SUN_LEN */ |
a6810074 DL |
801 | |
802 | /* Quick check to see if we might succeed before we go to the trouble | |
803 | of creating a socket. */ | |
804 | if (access(addr.sun_path, W_OK) < 0) { | |
805 | if (errno != ENOENT) | |
450971aa | 806 | flog_err_sys(EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
807 | "%s: access to socket %s denied: %s", |
808 | dmn->name, addr.sun_path, | |
809 | safe_strerror(errno)); | |
a6810074 DL |
810 | return -1; |
811 | } | |
812 | ||
813 | if ((sock = socket(AF_UNIX, SOCK_STREAM, 0)) < 0) { | |
450971aa | 814 | flog_err_sys(EC_LIB_SOCKET, "%s(%s): cannot make socket: %s", |
09c866e3 | 815 | __func__, addr.sun_path, safe_strerror(errno)); |
a6810074 DL |
816 | return -1; |
817 | } | |
818 | ||
819 | if (set_nonblocking(sock) < 0 || set_cloexec(sock) < 0) { | |
450971aa | 820 | flog_err_sys(EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
821 | "%s(%s): set_nonblocking/cloexec(%d) failed", |
822 | __func__, addr.sun_path, sock); | |
a6810074 DL |
823 | close(sock); |
824 | return -1; | |
8b886ca7 | 825 | } |
a6810074 DL |
826 | |
827 | if (connect(sock, (struct sockaddr *)&addr, len) < 0) { | |
828 | if ((errno != EINPROGRESS) && (errno != EWOULDBLOCK)) { | |
829 | if (gs.loglevel > LOG_DEBUG) | |
830 | zlog_debug("%s(%s): connect failed: %s", | |
831 | __func__, addr.sun_path, | |
832 | safe_strerror(errno)); | |
833 | close(sock); | |
834 | return -1; | |
835 | } | |
836 | if (gs.loglevel > LOG_DEBUG) | |
837 | zlog_debug("%s: connection in progress", dmn->name); | |
838 | dmn->state = DAEMON_CONNECTING; | |
839 | dmn->fd = sock; | |
66e78ae6 QY |
840 | dmn->t_write = NULL; |
841 | thread_add_write(master, check_connect, dmn, dmn->fd, | |
d62a17ae | 842 | &dmn->t_write); |
843 | dmn->t_wakeup = NULL; | |
844 | thread_add_timer(master, wakeup_connect_hanging, dmn, | |
845 | gs.timeout, &dmn->t_wakeup); | |
a6810074 DL |
846 | SET_READ_HANDLER(dmn); |
847 | return 0; | |
848 | } | |
849 | ||
850 | dmn->fd = sock; | |
851 | SET_READ_HANDLER(dmn); | |
852 | daemon_up(dmn, "connect succeeded"); | |
853 | return 1; | |
8b886ca7 | 854 | } |
855 | ||
a6810074 | 856 | static int phase_hanging(struct thread *t_hanging) |
8b886ca7 | 857 | { |
a6810074 | 858 | gs.t_phase_hanging = NULL; |
f74ae2bb | 859 | flog_err(EC_WATCHFRR_CONNECTION, |
1c50c1c0 QY |
860 | "Phase [%s] hanging for %ld seconds, aborting phased restart", |
861 | phase_str[gs.phase], PHASE_TIMEOUT); | |
a6810074 DL |
862 | gs.phase = PHASE_NONE; |
863 | return 0; | |
8b886ca7 | 864 | } |
865 | ||
a6810074 | 866 | static void set_phase(restart_phase_t new_phase) |
8b886ca7 | 867 | { |
a6810074 | 868 | gs.phase = new_phase; |
b3d6bc6e MS |
869 | thread_cancel(&gs.t_phase_hanging); |
870 | ||
66e78ae6 QY |
871 | thread_add_timer(master, phase_hanging, NULL, PHASE_TIMEOUT, |
872 | &gs.t_phase_hanging); | |
8b886ca7 | 873 | } |
874 | ||
a6810074 | 875 | static void phase_check(void) |
8b886ca7 | 876 | { |
c0e5cb52 DL |
877 | struct daemon *dmn; |
878 | ||
a6810074 DL |
879 | switch (gs.phase) { |
880 | case PHASE_NONE: | |
881 | break; | |
c0e5cb52 DL |
882 | |
883 | case PHASE_INIT: | |
884 | for (dmn = gs.daemons; dmn; dmn = dmn->next) | |
885 | if (dmn->state == DAEMON_INIT) | |
886 | return; | |
887 | ||
888 | /* startup complete, everything out of INIT */ | |
889 | gs.phase = PHASE_NONE; | |
890 | for (dmn = gs.daemons; dmn; dmn = dmn->next) | |
891 | if (dmn->state == DAEMON_DOWN) { | |
892 | SET_WAKEUP_DOWN(dmn); | |
893 | try_restart(dmn); | |
894 | } | |
895 | break; | |
a6810074 DL |
896 | case PHASE_STOPS_PENDING: |
897 | if (gs.numpids) | |
898 | break; | |
d62a17ae | 899 | zlog_info( |
900 | "Phased restart: all routing daemon stop jobs have completed."); | |
a6810074 DL |
901 | set_phase(PHASE_WAITING_DOWN); |
902 | ||
d62a17ae | 903 | /*FALLTHRU*/ |
a6810074 DL |
904 | case PHASE_WAITING_DOWN: |
905 | if (gs.numdown + IS_UP(gs.special) < gs.numdaemons) | |
906 | break; | |
907 | zlog_info("Phased restart: all routing daemons now down."); | |
908 | run_job(&gs.special->restart, "restart", gs.restart_command, 1, | |
909 | 1); | |
910 | set_phase(PHASE_ZEBRA_RESTART_PENDING); | |
911 | ||
d62a17ae | 912 | /*FALLTHRU*/ |
a6810074 DL |
913 | case PHASE_ZEBRA_RESTART_PENDING: |
914 | if (gs.special->restart.pid) | |
915 | break; | |
916 | zlog_info("Phased restart: %s restart job completed.", | |
917 | gs.special->name); | |
918 | set_phase(PHASE_WAITING_ZEBRA_UP); | |
919 | ||
d62a17ae | 920 | /*FALLTHRU*/ |
a6810074 DL |
921 | case PHASE_WAITING_ZEBRA_UP: |
922 | if (!IS_UP(gs.special)) | |
923 | break; | |
924 | zlog_info("Phased restart: %s is now up.", gs.special->name); | |
925 | { | |
926 | struct daemon *dmn; | |
927 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { | |
928 | if (dmn != gs.special) | |
929 | run_job(&dmn->restart, "start", | |
930 | gs.start_command, 1, 0); | |
931 | } | |
932 | } | |
933 | gs.phase = PHASE_NONE; | |
934 | THREAD_OFF(gs.t_phase_hanging); | |
935 | zlog_notice("Phased global restart has completed."); | |
936 | break; | |
937 | } | |
8b886ca7 | 938 | } |
939 | ||
a6810074 | 940 | static void try_restart(struct daemon *dmn) |
8b886ca7 | 941 | { |
f168b713 | 942 | if (watch_only) |
a6810074 | 943 | return; |
a6810074 | 944 | |
f168b713 DL |
945 | if (dmn != gs.special) { |
946 | if ((gs.special->state == DAEMON_UP) | |
947 | && (gs.phase == PHASE_NONE)) | |
948 | run_job(&dmn->restart, "restart", gs.restart_command, 0, | |
949 | 1); | |
950 | else | |
951 | zlog_debug( | |
3efd0893 | 952 | "%s: postponing restart attempt because master %s daemon not up [%s], or phased restart in progress", |
f168b713 DL |
953 | dmn->name, gs.special->name, |
954 | state_str[gs.special->state]); | |
955 | return; | |
956 | } | |
957 | ||
958 | if ((gs.phase != PHASE_NONE) || gs.numpids) { | |
959 | if (gs.loglevel > LOG_DEBUG + 1) | |
960 | zlog_debug( | |
3efd0893 | 961 | "postponing phased global restart: restart already in progress [%s], or outstanding child processes [%d]", |
f168b713 DL |
962 | phase_str[gs.phase], gs.numpids); |
963 | return; | |
964 | } | |
965 | /* Is it too soon for a restart? */ | |
966 | { | |
967 | struct timeval delay; | |
968 | if (time_elapsed(&delay, &gs.special->restart.time)->tv_sec | |
969 | < gs.special->restart.interval) { | |
a6810074 | 970 | if (gs.loglevel > LOG_DEBUG + 1) |
d62a17ae | 971 | zlog_debug( |
3efd0893 | 972 | "postponing phased global restart: elapsed time %ld < retry interval %ld", |
f168b713 DL |
973 | (long)delay.tv_sec, |
974 | gs.special->restart.interval); | |
975 | return; | |
a6810074 | 976 | } |
8b886ca7 | 977 | } |
f168b713 | 978 | run_job(&gs.restart, "restart", gs.restart_command, 0, 1); |
8b886ca7 | 979 | } |
980 | ||
a6810074 | 981 | static int wakeup_unresponsive(struct thread *t_wakeup) |
8b886ca7 | 982 | { |
a6810074 DL |
983 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
984 | ||
985 | dmn->t_wakeup = NULL; | |
986 | if (dmn->state != DAEMON_UNRESPONSIVE) | |
f74ae2bb | 987 | flog_err(EC_WATCHFRR_CONNECTION, |
3efd0893 | 988 | "%s: no longer unresponsive (now %s), wakeup should have been cancelled!", |
1c50c1c0 | 989 | dmn->name, state_str[dmn->state]); |
a6810074 DL |
990 | else { |
991 | SET_WAKEUP_UNRESPONSIVE(dmn); | |
992 | try_restart(dmn); | |
993 | } | |
994 | return 0; | |
8b886ca7 | 995 | } |
996 | ||
a6810074 | 997 | static int wakeup_no_answer(struct thread *t_wakeup) |
8b886ca7 | 998 | { |
a6810074 DL |
999 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
1000 | ||
1001 | dmn->t_wakeup = NULL; | |
1002 | dmn->state = DAEMON_UNRESPONSIVE; | |
cc53b605 DS |
1003 | if (dmn->ignore_timeout) |
1004 | return 0; | |
f74ae2bb | 1005 | flog_err(EC_WATCHFRR_CONNECTION, |
3efd0893 | 1006 | "%s state -> unresponsive : no response yet to ping sent %ld seconds ago", |
1c50c1c0 | 1007 | dmn->name, gs.timeout); |
71e7975a DL |
1008 | SET_WAKEUP_UNRESPONSIVE(dmn); |
1009 | try_restart(dmn); | |
a6810074 | 1010 | return 0; |
8b886ca7 | 1011 | } |
1012 | ||
a6810074 | 1013 | static int wakeup_send_echo(struct thread *t_wakeup) |
8b886ca7 | 1014 | { |
a6810074 DL |
1015 | static const char echocmd[] = "echo " PING_TOKEN; |
1016 | ssize_t rc; | |
1017 | struct daemon *dmn = THREAD_ARG(t_wakeup); | |
1018 | ||
1019 | dmn->t_wakeup = NULL; | |
d62a17ae | 1020 | if (((rc = write(dmn->fd, echocmd, sizeof(echocmd))) < 0) |
1021 | || ((size_t)rc != sizeof(echocmd))) { | |
a6810074 DL |
1022 | char why[100 + sizeof(echocmd)]; |
1023 | snprintf(why, sizeof(why), | |
1024 | "write '%s' returned %d instead of %u", echocmd, | |
d7c0a89a | 1025 | (int)rc, (unsigned int)sizeof(echocmd)); |
a6810074 DL |
1026 | daemon_down(dmn, why); |
1027 | } else { | |
1028 | gettimeofday(&dmn->echo_sent, NULL); | |
66e78ae6 QY |
1029 | dmn->t_wakeup = NULL; |
1030 | thread_add_timer(master, wakeup_no_answer, dmn, gs.timeout, | |
1031 | &dmn->t_wakeup); | |
a6810074 DL |
1032 | } |
1033 | return 0; | |
8b886ca7 | 1034 | } |
1035 | ||
470bc619 QY |
1036 | bool check_all_up(void) |
1037 | { | |
1038 | struct daemon *dmn; | |
1039 | ||
1040 | for (dmn = gs.daemons; dmn; dmn = dmn->next) | |
1041 | if (dmn->state != DAEMON_UP) | |
1042 | return false; | |
1043 | return true; | |
1044 | } | |
1045 | ||
af568444 DL |
1046 | void watchfrr_status(struct vty *vty) |
1047 | { | |
1048 | struct daemon *dmn; | |
1049 | struct timeval delay; | |
1050 | ||
1051 | vty_out(vty, "watchfrr global phase: %s\n", phase_str[gs.phase]); | |
1052 | if (gs.restart.pid) | |
1053 | vty_out(vty, " global restart running, pid %ld\n", | |
1054 | (long)gs.restart.pid); | |
1055 | ||
1056 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { | |
cc53b605 DS |
1057 | vty_out(vty, " %-20s %s%s", dmn->name, state_str[dmn->state], |
1058 | dmn->ignore_timeout ? "/Ignoring Timeout\n" : "\n"); | |
af568444 DL |
1059 | if (dmn->restart.pid) |
1060 | vty_out(vty, " restart running, pid %ld\n", | |
1061 | (long)dmn->restart.pid); | |
1062 | else if (dmn->state == DAEMON_DOWN && | |
1063 | time_elapsed(&delay, &dmn->restart.time)->tv_sec | |
1064 | < dmn->restart.interval) | |
3efd0893 | 1065 | vty_out(vty, " restarting in %jd seconds (%jds backoff interval)\n", |
051a0be4 DL |
1066 | (intmax_t)dmn->restart.interval |
1067 | - (intmax_t)delay.tv_sec, | |
1068 | (intmax_t)dmn->restart.interval); | |
af568444 DL |
1069 | } |
1070 | } | |
1071 | ||
a6810074 | 1072 | static void sigint(void) |
8b886ca7 | 1073 | { |
a6810074 DL |
1074 | zlog_notice("Terminating on signal"); |
1075 | systemd_send_stopping(); | |
1076 | exit(0); | |
8b886ca7 | 1077 | } |
1078 | ||
a6810074 | 1079 | static int valid_command(const char *cmd) |
8b886ca7 | 1080 | { |
a6810074 | 1081 | char *p; |
8b886ca7 | 1082 | |
53a78fc1 RZ |
1083 | if (cmd == NULL) |
1084 | return 0; | |
1085 | ||
a6810074 | 1086 | return ((p = strchr(cmd, '%')) != NULL) && (*(p + 1) == 's') |
d62a17ae | 1087 | && !strchr(p + 1, '%'); |
8b886ca7 | 1088 | } |
1089 | ||
c8b40f86 | 1090 | /* This is an ugly hack to circumvent problems with passing command-line |
1091 | arguments that contain spaces. The fix is to use a configuration file. */ | |
a6810074 | 1092 | static char *translate_blanks(const char *cmd, const char *blankstr) |
c8b40f86 | 1093 | { |
a6810074 DL |
1094 | char *res; |
1095 | char *p; | |
1096 | size_t bslen = strlen(blankstr); | |
1097 | ||
1098 | if (!(res = strdup(cmd))) { | |
1099 | perror("strdup"); | |
1100 | exit(1); | |
1101 | } | |
1102 | while ((p = strstr(res, blankstr)) != NULL) { | |
1103 | *p = ' '; | |
1104 | if (bslen != 1) | |
1105 | memmove(p + 1, p + bslen, strlen(p + bslen) + 1); | |
1106 | } | |
1107 | return res; | |
c8b40f86 | 1108 | } |
1109 | ||
5c9d1c83 DL |
1110 | static int startup_timeout(struct thread *t_wakeup) |
1111 | { | |
1112 | daemon_send_ready(1); | |
1113 | return 0; | |
1114 | } | |
1115 | ||
33606a15 DL |
1116 | #ifdef GNU_LINUX |
1117 | ||
1118 | #include <sys/mount.h> | |
1119 | #include <sched.h> | |
1120 | ||
1121 | #define NETNS_RUN_DIR "/var/run/netns" | |
1122 | ||
1123 | static void netns_create(int dirfd, const char *nsname) | |
1124 | { | |
1125 | /* make /var/run/netns shared between mount namespaces | |
1126 | * just like iproute2 sets it up | |
1127 | */ | |
1128 | if (mount("", NETNS_RUN_DIR, "none", MS_SHARED | MS_REC, NULL)) { | |
1129 | if (errno != EINVAL) { | |
1130 | perror("mount"); | |
1131 | exit(1); | |
1132 | } | |
1133 | ||
1134 | if (mount(NETNS_RUN_DIR, NETNS_RUN_DIR, "none", | |
1135 | MS_BIND | MS_REC, NULL)) { | |
1136 | perror("mount"); | |
1137 | exit(1); | |
1138 | } | |
1139 | ||
1140 | if (mount("", NETNS_RUN_DIR, "none", MS_SHARED | MS_REC, | |
1141 | NULL)) { | |
1142 | perror("mount"); | |
1143 | exit(1); | |
1144 | } | |
1145 | } | |
1146 | ||
1147 | /* need an empty file to mount on top of */ | |
1148 | int nsfd = openat(dirfd, nsname, O_CREAT | O_RDONLY | O_EXCL, 0); | |
1149 | ||
1150 | if (nsfd < 0) { | |
1151 | fprintf(stderr, "failed to create \"%s/%s\": %s\n", | |
1152 | NETNS_RUN_DIR, nsname, strerror(errno)); | |
1153 | exit(1); | |
1154 | } | |
1155 | close(nsfd); | |
1156 | ||
1157 | if (unshare(CLONE_NEWNET)) { | |
1158 | perror("unshare"); | |
1159 | unlinkat(dirfd, nsname, 0); | |
1160 | exit(1); | |
1161 | } | |
1162 | ||
1163 | char *dstpath = asprintfrr(MTYPE_TMP, "%s/%s", NETNS_RUN_DIR, nsname); | |
1164 | ||
1165 | /* bind-mount so the namespace has a name and is persistent */ | |
1166 | if (mount("/proc/self/ns/net", dstpath, "none", MS_BIND, NULL) < 0) { | |
1167 | fprintf(stderr, "failed to bind-mount netns to \"%s\": %s\n", | |
1168 | dstpath, strerror(errno)); | |
1169 | unlinkat(dirfd, nsname, 0); | |
1170 | exit(1); | |
1171 | } | |
1172 | ||
1173 | XFREE(MTYPE_TMP, dstpath); | |
1174 | } | |
1175 | ||
1176 | static void netns_setup(const char *nsname) | |
1177 | { | |
1178 | int dirfd, nsfd; | |
1179 | ||
1180 | dirfd = open(NETNS_RUN_DIR, O_DIRECTORY | O_RDONLY); | |
1181 | if (dirfd < 0) { | |
1182 | if (errno == ENOTDIR) { | |
1183 | fprintf(stderr, "error: \"%s\" is not a directory!\n", | |
1184 | NETNS_RUN_DIR); | |
1185 | exit(1); | |
1186 | } else if (errno == ENOENT) { | |
1187 | if (mkdir(NETNS_RUN_DIR, 0755)) { | |
1188 | fprintf(stderr, "error: \"%s\": mkdir: %s\n", | |
1189 | NETNS_RUN_DIR, strerror(errno)); | |
1190 | exit(1); | |
1191 | } | |
1192 | dirfd = open(NETNS_RUN_DIR, O_DIRECTORY | O_RDONLY); | |
1193 | if (dirfd < 0) { | |
1194 | fprintf(stderr, "error: \"%s\": opendir: %s\n", | |
1195 | NETNS_RUN_DIR, strerror(errno)); | |
1196 | exit(1); | |
1197 | } | |
1198 | } else { | |
1199 | fprintf(stderr, "error: \"%s\": %s\n", | |
1200 | NETNS_RUN_DIR, strerror(errno)); | |
1201 | exit(1); | |
1202 | } | |
1203 | } | |
1204 | ||
1205 | nsfd = openat(dirfd, nsname, O_RDONLY); | |
1206 | if (nsfd < 0 && errno != ENOENT) { | |
1207 | fprintf(stderr, "error: \"%s/%s\": %s\n", | |
1208 | NETNS_RUN_DIR, nsname, strerror(errno)); | |
1209 | exit(1); | |
1210 | } | |
1211 | if (nsfd < 0) | |
1212 | netns_create(dirfd, nsname); | |
1213 | else { | |
1214 | if (setns(nsfd, CLONE_NEWNET)) { | |
1215 | perror("setns"); | |
1216 | exit(1); | |
1217 | } | |
1218 | close(nsfd); | |
1219 | } | |
1220 | close(dirfd); | |
1221 | ||
1222 | /* make sure loopback is up... weird things happen otherwise. | |
1223 | * ioctl is perfectly fine for this, don't need netlink... | |
1224 | */ | |
1225 | int sockfd; | |
1226 | struct ifreq ifr = { }; | |
1227 | ||
1228 | strlcpy(ifr.ifr_name, "lo", sizeof(ifr.ifr_name)); | |
1229 | ||
1230 | sockfd = socket(AF_INET, SOCK_DGRAM, 0); | |
1231 | if (sockfd < 0) { | |
1232 | perror("socket"); | |
1233 | exit(1); | |
1234 | } | |
1235 | if (ioctl(sockfd, SIOCGIFFLAGS, &ifr)) { | |
1236 | perror("ioctl(SIOCGIFFLAGS, \"lo\")"); | |
1237 | exit(1); | |
1238 | } | |
1239 | if (!(ifr.ifr_flags & IFF_UP)) { | |
1240 | ifr.ifr_flags |= IFF_UP; | |
1241 | if (ioctl(sockfd, SIOCSIFFLAGS, &ifr)) { | |
1242 | perror("ioctl(SIOCSIFFLAGS, \"lo\")"); | |
1243 | exit(1); | |
1244 | } | |
1245 | } | |
1246 | close(sockfd); | |
1247 | } | |
1248 | ||
1249 | #else /* !GNU_LINUX */ | |
1250 | ||
1251 | static void netns_setup(const char *nsname) | |
1252 | { | |
1253 | fprintf(stderr, "network namespaces are only available on Linux\n"); | |
1254 | exit(1); | |
1255 | } | |
1256 | #endif | |
1257 | ||
0a7c7856 DL |
1258 | static void watchfrr_init(int argc, char **argv) |
1259 | { | |
1260 | const char *special = "zebra"; | |
1261 | int i; | |
1262 | struct daemon *dmn, **add = &gs.daemons; | |
1263 | char alldaemons[512] = "", *p = alldaemons; | |
1264 | ||
5c9d1c83 DL |
1265 | thread_add_timer_msec(master, startup_timeout, NULL, STARTUP_TIMEOUT, |
1266 | &gs.t_startup_timeout); | |
1267 | ||
0a7c7856 DL |
1268 | for (i = optind; i < argc; i++) { |
1269 | dmn = XCALLOC(MTYPE_WATCHFRR_DAEMON, sizeof(*dmn)); | |
1270 | ||
1271 | dmn->name = dmn->restart.name = argv[i]; | |
1272 | dmn->state = DAEMON_INIT; | |
1273 | gs.numdaemons++; | |
1274 | gs.numdown++; | |
1275 | dmn->fd = -1; | |
1276 | dmn->t_wakeup = NULL; | |
c0e5cb52 | 1277 | thread_add_timer_msec(master, wakeup_init, dmn, 0, |
0a7c7856 DL |
1278 | &dmn->t_wakeup); |
1279 | dmn->restart.interval = gs.min_restart_interval; | |
1280 | *add = dmn; | |
1281 | add = &dmn->next; | |
1282 | ||
1283 | if (!strcmp(dmn->name, special)) | |
1284 | gs.special = dmn; | |
1285 | } | |
1286 | ||
1287 | if (!gs.daemons) { | |
1288 | fprintf(stderr, | |
1289 | "Must specify one or more daemons to monitor.\n\n"); | |
1290 | frr_help_exit(1); | |
1291 | } | |
1292 | if (!watch_only && !gs.special) { | |
1293 | fprintf(stderr, "\"%s\" daemon must be in daemon lists\n\n", | |
1294 | special); | |
1295 | frr_help_exit(1); | |
1296 | } | |
1297 | ||
1298 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { | |
1299 | snprintf(p, alldaemons + sizeof(alldaemons) - p, "%s%s", | |
1300 | (p == alldaemons) ? "" : " ", dmn->name); | |
1301 | p += strlen(p); | |
1302 | } | |
1303 | zlog_notice("%s %s watching [%s]%s", progname, FRR_VERSION, alldaemons, | |
1304 | watch_only ? ", monitor mode" : ""); | |
1305 | } | |
1306 | ||
a6810074 | 1307 | struct zebra_privs_t watchfrr_privs = { |
95c4aff2 | 1308 | #ifdef VTY_GROUP |
a6810074 | 1309 | .vty_group = VTY_GROUP, |
95c4aff2 DL |
1310 | #endif |
1311 | }; | |
1312 | ||
4f04a76b DL |
1313 | static struct quagga_signal_t watchfrr_signals[] = { |
1314 | { | |
1315 | .signal = SIGINT, | |
1316 | .handler = sigint, | |
1317 | }, | |
1318 | { | |
1319 | .signal = SIGTERM, | |
1320 | .handler = sigint, | |
1321 | }, | |
1322 | { | |
1323 | .signal = SIGCHLD, | |
1324 | .handler = sigchild, | |
1325 | }, | |
1326 | }; | |
1327 | ||
1328 | FRR_DAEMON_INFO(watchfrr, WATCHFRR, | |
d62a17ae | 1329 | .flags = FRR_NO_PRIVSEP | FRR_NO_TCPVTY | FRR_LIMITED_CLI |
0a7c7856 DL |
1330 | | FRR_NO_CFG_PID_DRY | FRR_NO_ZCLIENT |
1331 | | FRR_DETACH_LATER, | |
4f04a76b | 1332 | |
d62a17ae | 1333 | .printhelp = printhelp, |
1334 | .copyright = "Copyright 2004 Andrew J. Schorr", | |
4f04a76b | 1335 | |
d62a17ae | 1336 | .signals = watchfrr_signals, |
1337 | .n_signals = array_size(watchfrr_signals), | |
4f04a76b | 1338 | |
80413c20 DL |
1339 | .privs = &watchfrr_privs, |
1340 | ); | |
4f04a76b | 1341 | |
999f153e DL |
1342 | #define DEPRECATED_OPTIONS "aAezR:" |
1343 | ||
a6810074 | 1344 | int main(int argc, char **argv) |
8b886ca7 | 1345 | { |
a6810074 | 1346 | int opt; |
a6810074 | 1347 | const char *blankstr = NULL; |
33606a15 DL |
1348 | const char *netns = NULL; |
1349 | bool netns_en = false; | |
a6810074 | 1350 | |
4f04a76b DL |
1351 | frr_preinit(&watchfrr_di, argc, argv); |
1352 | progname = watchfrr_di.progname; | |
1353 | ||
33606a15 | 1354 | frr_opt_add("b:di:k:l:N:p:r:S:s:t:T:" DEPRECATED_OPTIONS, longopts, ""); |
a6810074 DL |
1355 | |
1356 | gs.restart.name = "all"; | |
4f04a76b | 1357 | while ((opt = frr_getopt(argc, argv, NULL)) != EOF) { |
999f153e DL |
1358 | if (opt && opt < 128 && strchr(DEPRECATED_OPTIONS, opt)) { |
1359 | fprintf(stderr, | |
1360 | "The -%c option no longer exists.\n" | |
1361 | "Please refer to the watchfrr(8) man page.\n", | |
1362 | opt); | |
1363 | exit(1); | |
1364 | } | |
1365 | ||
a6810074 DL |
1366 | switch (opt) { |
1367 | case 0: | |
1368 | break; | |
a6810074 DL |
1369 | case 'b': |
1370 | blankstr = optarg; | |
1371 | break; | |
f168b713 DL |
1372 | case OPTION_DRY: |
1373 | watch_only = true; | |
a6810074 DL |
1374 | break; |
1375 | case 'k': | |
1376 | if (!valid_command(optarg)) { | |
1377 | fprintf(stderr, | |
1378 | "Invalid kill command, must contain '%%s': %s\n", | |
1379 | optarg); | |
4f04a76b | 1380 | frr_help_exit(1); |
a6810074 DL |
1381 | } |
1382 | gs.stop_command = optarg; | |
1383 | break; | |
d62a17ae | 1384 | case 'l': { |
1385 | char garbage[3]; | |
1386 | if ((sscanf(optarg, "%d%1s", &gs.loglevel, garbage) | |
1387 | != 1) | |
1388 | || (gs.loglevel < LOG_EMERG)) { | |
1389 | fprintf(stderr, | |
1390 | "Invalid loglevel argument: %s\n", | |
1391 | optarg); | |
1392 | frr_help_exit(1); | |
a6810074 | 1393 | } |
d62a17ae | 1394 | } break; |
1395 | case OPTION_MINRESTART: { | |
1396 | char garbage[3]; | |
1397 | if ((sscanf(optarg, "%ld%1s", &gs.min_restart_interval, | |
1398 | garbage) | |
1399 | != 1) | |
1400 | || (gs.min_restart_interval < 0)) { | |
1401 | fprintf(stderr, | |
1402 | "Invalid min_restart_interval argument: %s\n", | |
1403 | optarg); | |
1404 | frr_help_exit(1); | |
a6810074 | 1405 | } |
d62a17ae | 1406 | } break; |
1407 | case OPTION_MAXRESTART: { | |
1408 | char garbage[3]; | |
1409 | if ((sscanf(optarg, "%ld%1s", &gs.max_restart_interval, | |
1410 | garbage) | |
1411 | != 1) | |
1412 | || (gs.max_restart_interval < 0)) { | |
1413 | fprintf(stderr, | |
1414 | "Invalid max_restart_interval argument: %s\n", | |
1415 | optarg); | |
1416 | frr_help_exit(1); | |
a6810074 | 1417 | } |
d62a17ae | 1418 | } break; |
33606a15 DL |
1419 | case OPTION_NETNS: |
1420 | netns_en = true; | |
b12bc77c | 1421 | if (optarg && strchr(optarg, '/')) { |
33606a15 DL |
1422 | fprintf(stderr, |
1423 | "invalid network namespace name \"%s\" (may not contain slashes)\n", | |
1424 | optarg); | |
1425 | frr_help_exit(1); | |
1426 | } | |
1427 | netns = optarg; | |
1428 | break; | |
d62a17ae | 1429 | case 'i': { |
1430 | char garbage[3]; | |
1431 | int period; | |
1432 | if ((sscanf(optarg, "%d%1s", &period, garbage) != 1) | |
1433 | || (gs.period < 1)) { | |
1434 | fprintf(stderr, | |
1435 | "Invalid interval argument: %s\n", | |
1436 | optarg); | |
1437 | frr_help_exit(1); | |
a6810074 | 1438 | } |
d62a17ae | 1439 | gs.period = 1000 * period; |
1440 | } break; | |
a6810074 | 1441 | case 'p': |
0a7c7856 | 1442 | watchfrr_di.pid_file = optarg; |
a6810074 DL |
1443 | break; |
1444 | case 'r': | |
a6810074 DL |
1445 | if (!valid_command(optarg)) { |
1446 | fprintf(stderr, | |
1447 | "Invalid restart command, must contain '%%s': %s\n", | |
1448 | optarg); | |
4f04a76b | 1449 | frr_help_exit(1); |
a6810074 DL |
1450 | } |
1451 | gs.restart_command = optarg; | |
a6810074 DL |
1452 | break; |
1453 | case 's': | |
1454 | if (!valid_command(optarg)) { | |
1455 | fprintf(stderr, | |
1456 | "Invalid start command, must contain '%%s': %s\n", | |
1457 | optarg); | |
4f04a76b | 1458 | frr_help_exit(1); |
a6810074 DL |
1459 | } |
1460 | gs.start_command = optarg; | |
1461 | break; | |
1462 | case 'S': | |
1463 | gs.vtydir = optarg; | |
1464 | break; | |
d62a17ae | 1465 | case 't': { |
1466 | char garbage[3]; | |
1467 | if ((sscanf(optarg, "%ld%1s", &gs.timeout, garbage) | |
1468 | != 1) | |
1469 | || (gs.timeout < 1)) { | |
1470 | fprintf(stderr, | |
1471 | "Invalid timeout argument: %s\n", | |
1472 | optarg); | |
1473 | frr_help_exit(1); | |
a6810074 | 1474 | } |
d62a17ae | 1475 | } break; |
1476 | case 'T': { | |
1477 | char garbage[3]; | |
1478 | if ((sscanf(optarg, "%ld%1s", &gs.restart_timeout, | |
1479 | garbage) | |
1480 | != 1) | |
1481 | || (gs.restart_timeout < 1)) { | |
1482 | fprintf(stderr, | |
1483 | "Invalid restart timeout argument: %s\n", | |
1484 | optarg); | |
1485 | frr_help_exit(1); | |
a6810074 | 1486 | } |
d62a17ae | 1487 | } break; |
a6810074 DL |
1488 | default: |
1489 | fputs("Invalid option.\n", stderr); | |
4f04a76b | 1490 | frr_help_exit(1); |
a6810074 | 1491 | } |
8b886ca7 | 1492 | } |
a6810074 | 1493 | |
71e7975a DL |
1494 | if (watch_only |
1495 | && (gs.start_command || gs.stop_command || gs.restart_command)) { | |
d87ae5cc | 1496 | fputs("Options -r/-s/-k are not used when --dry is active.\n", |
a6810074 | 1497 | stderr); |
8b886ca7 | 1498 | } |
f168b713 DL |
1499 | if (!watch_only |
1500 | && (!gs.restart_command || !gs.start_command || !gs.stop_command)) { | |
1501 | fprintf(stderr, | |
1502 | "Options -s (start), -k (kill), and -r (restart) are required.\n"); | |
1503 | frr_help_exit(1); | |
8b886ca7 | 1504 | } |
8b886ca7 | 1505 | |
a6810074 DL |
1506 | if (blankstr) { |
1507 | if (gs.restart_command) | |
1508 | gs.restart_command = | |
d62a17ae | 1509 | translate_blanks(gs.restart_command, blankstr); |
a6810074 DL |
1510 | if (gs.start_command) |
1511 | gs.start_command = | |
d62a17ae | 1512 | translate_blanks(gs.start_command, blankstr); |
a6810074 DL |
1513 | if (gs.stop_command) |
1514 | gs.stop_command = | |
d62a17ae | 1515 | translate_blanks(gs.stop_command, blankstr); |
065de903 | 1516 | } |
8b886ca7 | 1517 | |
a6810074 | 1518 | gs.restart.interval = gs.min_restart_interval; |
8b886ca7 | 1519 | |
33606a15 DL |
1520 | /* env variable for the processes that we start */ |
1521 | if (watchfrr_di.pathspace) | |
1522 | setenv("FRR_PATHSPACE", watchfrr_di.pathspace, 1); | |
1523 | else | |
1524 | unsetenv("FRR_PATHSPACE"); | |
1525 | ||
1526 | if (netns_en && !netns) | |
1527 | netns = watchfrr_di.pathspace; | |
1528 | if (netns_en && netns && netns[0]) | |
1529 | netns_setup(netns); | |
1530 | ||
4f04a76b | 1531 | master = frr_init(); |
b647dc2a | 1532 | watchfrr_error_init(); |
0a7c7856 DL |
1533 | watchfrr_init(argc, argv); |
1534 | watchfrr_vty_init(); | |
1535 | ||
1536 | frr_config_fork(); | |
4f04a76b | 1537 | |
0a7c7856 | 1538 | if (watchfrr_di.daemon_mode) |
0bdeb5e5 | 1539 | zlog_syslog_set_prio_min(MIN(gs.loglevel, LOG_DEBUG)); |
0a7c7856 | 1540 | else |
0bdeb5e5 | 1541 | zlog_aux_init(NULL, MIN(gs.loglevel, LOG_DEBUG)); |
8b886ca7 | 1542 | |
0a7c7856 | 1543 | frr_run(master); |
8b886ca7 | 1544 | |
a6810074 DL |
1545 | systemd_send_stopping(); |
1546 | /* Not reached. */ | |
1547 | return 0; | |
8b886ca7 | 1548 | } |