]>
Commit | Line | Data |
---|---|---|
acddc0ed | 1 | // SPDX-License-Identifier: GPL-2.0-or-later |
8b886ca7 | 2 | /* |
896014f4 DL |
3 | * Monitor status of frr daemons and restart if necessary. |
4 | * | |
5 | * Copyright (C) 2004 Andrew J. Schorr | |
8b886ca7 | 6 | */ |
7 | ||
a365534f | 8 | #include <zebra.h> |
8b886ca7 | 9 | #include <thread.h> |
10 | #include <log.h> | |
52e66296 | 11 | #include <network.h> |
8b886ca7 | 12 | #include <sigevent.h> |
a365534f | 13 | #include <lib/version.h> |
95c4aff2 | 14 | #include "command.h" |
4f04a76b | 15 | #include "libfrr.h" |
b647dc2a | 16 | #include "lib_errors.h" |
0bdeb5e5 | 17 | #include "zlog_targets.h" |
5920b3eb | 18 | #include "network.h" |
33606a15 | 19 | #include "printfrr.h" |
95c4aff2 | 20 | |
6f594023 | 21 | #include <getopt.h> |
a365534f | 22 | #include <sys/un.h> |
23 | #include <sys/wait.h> | |
837d16cc | 24 | #include <memory.h> |
651415bd | 25 | #include <systemd.h> |
8b886ca7 | 26 | |
9473e340 | 27 | #include "watchfrr.h" |
b647dc2a | 28 | #include "watchfrr_errors.h" |
95c4aff2 | 29 | |
8b886ca7 | 30 | #ifndef MIN |
31 | #define MIN(X,Y) (((X) <= (Y)) ? (X) : (Y)) | |
32 | #endif | |
33 | ||
34 | /* Macros to help randomize timers. */ | |
5920b3eb | 35 | #define JITTER(X) ((frr_weak_random() % ((X)+1))-((X)/2)) |
8b886ca7 | 36 | #define FUZZY(X) ((X)+JITTER((X)/20)) |
37 | ||
38 | #define DEFAULT_PERIOD 5 | |
0a64aff6 | 39 | #define DEFAULT_TIMEOUT 90 |
8b886ca7 | 40 | #define DEFAULT_RESTART_TIMEOUT 20 |
41 | #define DEFAULT_LOGLEVEL LOG_INFO | |
42 | #define DEFAULT_MIN_RESTART 60 | |
43 | #define DEFAULT_MAX_RESTART 600 | |
6d0fa5c2 | 44 | #define DEFAULT_OPERATIONAL_TIMEOUT 60 |
8b886ca7 | 45 | |
3ec95567 DL |
46 | #define DEFAULT_RESTART_CMD WATCHFRR_SH_PATH " restart %s" |
47 | #define DEFAULT_START_CMD WATCHFRR_SH_PATH " start %s" | |
48 | #define DEFAULT_STOP_CMD WATCHFRR_SH_PATH " stop %s" | |
49 | ||
8b886ca7 | 50 | #define PING_TOKEN "PING" |
51 | ||
bf8d3d6a DL |
52 | DEFINE_MGROUP(WATCHFRR, "watchfrr"); |
53 | DEFINE_MTYPE_STATIC(WATCHFRR, WATCHFRR_DAEMON, "watchfrr daemon entry"); | |
0a7c7856 | 54 | |
55c72803 | 55 | /* Needs to be global, referenced somewhere inside libfrr. */ |
8b886ca7 | 56 | struct thread_master *master; |
57 | ||
f168b713 | 58 | static bool watch_only = false; |
a91f5417 | 59 | const char *pathspace; |
8b886ca7 | 60 | |
f1692c51 | 61 | enum restart_phase { |
a6810074 | 62 | PHASE_NONE = 0, |
c0e5cb52 | 63 | PHASE_INIT, |
a6810074 DL |
64 | PHASE_STOPS_PENDING, |
65 | PHASE_WAITING_DOWN, | |
66 | PHASE_ZEBRA_RESTART_PENDING, | |
67 | PHASE_WAITING_ZEBRA_UP | |
f1692c51 | 68 | }; |
8b886ca7 | 69 | |
2b64873d | 70 | static const char *const phase_str[] = { |
af568444 | 71 | "Idle", |
c0e5cb52 | 72 | "Startup", |
a6810074 DL |
73 | "Stop jobs running", |
74 | "Waiting for other daemons to come down", | |
75 | "Zebra restart job running", | |
76 | "Waiting for zebra to come up", | |
77 | "Start jobs running", | |
8b886ca7 | 78 | }; |
79 | ||
80 | #define PHASE_TIMEOUT (3*gs.restart_timeout) | |
5c9d1c83 | 81 | #define STARTUP_TIMEOUT 55 * 1000 |
8b886ca7 | 82 | |
a6810074 DL |
83 | struct restart_info { |
84 | const char *name; | |
85 | const char *what; | |
86 | pid_t pid; | |
87 | struct timeval time; | |
88 | long interval; | |
89 | struct thread *t_kill; | |
90 | int kills; | |
098e240f | 91 | }; |
92 | ||
a6810074 | 93 | static struct global_state { |
f1692c51 | 94 | enum restart_phase phase; |
a6810074 | 95 | struct thread *t_phase_hanging; |
5c9d1c83 | 96 | struct thread *t_startup_timeout; |
6d0fa5c2 | 97 | struct thread *t_operational; |
a6810074 DL |
98 | const char *vtydir; |
99 | long period; | |
100 | long timeout; | |
101 | long restart_timeout; | |
2ab760f0 | 102 | bool reading_configuration; |
a6810074 DL |
103 | long min_restart_interval; |
104 | long max_restart_interval; | |
6d0fa5c2 | 105 | long operational_timeout; |
a6810074 DL |
106 | struct daemon *daemons; |
107 | const char *restart_command; | |
108 | const char *start_command; | |
109 | const char *stop_command; | |
110 | struct restart_info restart; | |
a6810074 | 111 | int loglevel; |
d62a17ae | 112 | struct daemon *special; /* points to zebra when doing phased restart */ |
a6810074 DL |
113 | int numdaemons; |
114 | int numpids; | |
d62a17ae | 115 | int numdown; /* # of daemons that are not UP or UNRESPONSIVE */ |
8b886ca7 | 116 | } gs = { |
c0e5cb52 | 117 | .phase = PHASE_INIT, |
64a249ad | 118 | .vtydir = frr_vtydir, |
d62a17ae | 119 | .period = 1000 * DEFAULT_PERIOD, |
120 | .timeout = DEFAULT_TIMEOUT, | |
121 | .restart_timeout = DEFAULT_RESTART_TIMEOUT, | |
122 | .loglevel = DEFAULT_LOGLEVEL, | |
123 | .min_restart_interval = DEFAULT_MIN_RESTART, | |
124 | .max_restart_interval = DEFAULT_MAX_RESTART, | |
6d0fa5c2 | 125 | .operational_timeout = DEFAULT_OPERATIONAL_TIMEOUT, |
3ec95567 DL |
126 | .restart_command = DEFAULT_RESTART_CMD, |
127 | .start_command = DEFAULT_START_CMD, | |
128 | .stop_command = DEFAULT_STOP_CMD, | |
d62a17ae | 129 | }; |
a6810074 | 130 | |
c3db4ca8 | 131 | enum daemon_state { |
a6810074 DL |
132 | DAEMON_INIT, |
133 | DAEMON_DOWN, | |
134 | DAEMON_CONNECTING, | |
135 | DAEMON_UP, | |
136 | DAEMON_UNRESPONSIVE | |
c3db4ca8 | 137 | }; |
8b886ca7 | 138 | |
d62a17ae | 139 | #define IS_UP(DMN) \ |
140 | (((DMN)->state == DAEMON_UP) || ((DMN)->state == DAEMON_UNRESPONSIVE)) | |
8b886ca7 | 141 | |
2b64873d | 142 | static const char *const state_str[] = { |
d62a17ae | 143 | "Init", "Down", "Connecting", "Up", "Unresponsive", |
8b886ca7 | 144 | }; |
145 | ||
146 | struct daemon { | |
a6810074 | 147 | const char *name; |
c3db4ca8 | 148 | enum daemon_state state; |
a6810074 DL |
149 | int fd; |
150 | struct timeval echo_sent; | |
d7c0a89a | 151 | unsigned int connect_tries; |
a6810074 DL |
152 | struct thread *t_wakeup; |
153 | struct thread *t_read; | |
154 | struct thread *t_write; | |
155 | struct daemon *next; | |
156 | struct restart_info restart; | |
cc53b605 DS |
157 | |
158 | /* | |
159 | * For a given daemon, if we've turned on ignore timeouts | |
160 | * ignore the timeout value and assume everything is ok | |
161 | * This is for daemon debugging w/ gdb after we have started | |
162 | * FRR and realize we have something that needs to be looked | |
163 | * at | |
164 | */ | |
165 | bool ignore_timeout; | |
8b886ca7 | 166 | }; |
167 | ||
9272302b DL |
168 | #define OPTION_MINRESTART 2000 |
169 | #define OPTION_MAXRESTART 2001 | |
f168b713 | 170 | #define OPTION_DRY 2002 |
33606a15 | 171 | #define OPTION_NETNS 2003 |
6d0fa5c2 | 172 | #define OPTION_MAXOPERATIONAL 2004 |
9272302b | 173 | |
a6810074 DL |
174 | static const struct option longopts[] = { |
175 | {"daemon", no_argument, NULL, 'd'}, | |
176 | {"statedir", required_argument, NULL, 'S'}, | |
a6810074 DL |
177 | {"loglevel", required_argument, NULL, 'l'}, |
178 | {"interval", required_argument, NULL, 'i'}, | |
179 | {"timeout", required_argument, NULL, 't'}, | |
180 | {"restart-timeout", required_argument, NULL, 'T'}, | |
181 | {"restart", required_argument, NULL, 'r'}, | |
182 | {"start-command", required_argument, NULL, 's'}, | |
183 | {"kill-command", required_argument, NULL, 'k'}, | |
f168b713 | 184 | {"dry", no_argument, NULL, OPTION_DRY}, |
d62a17ae | 185 | {"min-restart-interval", required_argument, NULL, OPTION_MINRESTART}, |
186 | {"max-restart-interval", required_argument, NULL, OPTION_MAXRESTART}, | |
6d0fa5c2 | 187 | {"operational-timeout", required_argument, NULL, OPTION_MAXOPERATIONAL}, |
a6810074 DL |
188 | {"pid-file", required_argument, NULL, 'p'}, |
189 | {"blank-string", required_argument, NULL, 'b'}, | |
33606a15 DL |
190 | #ifdef GNU_LINUX |
191 | {"netns", optional_argument, NULL, OPTION_NETNS}, | |
192 | #endif | |
a6810074 DL |
193 | {"help", no_argument, NULL, 'h'}, |
194 | {"version", no_argument, NULL, 'v'}, | |
d62a17ae | 195 | {NULL, 0, NULL, 0}}; |
8b886ca7 | 196 | |
197 | static int try_connect(struct daemon *dmn); | |
cc9f21da | 198 | static void wakeup_send_echo(struct thread *t_wakeup); |
8b886ca7 | 199 | static void try_restart(struct daemon *dmn); |
200 | static void phase_check(void); | |
75f8b0e4 | 201 | static void restart_done(struct daemon *dmn); |
8b886ca7 | 202 | |
4f04a76b | 203 | static const char *progname; |
cc53b605 DS |
204 | |
205 | void watchfrr_set_ignore_daemon(struct vty *vty, const char *dname, bool ignore) | |
206 | { | |
207 | struct daemon *dmn; | |
208 | ||
209 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { | |
210 | if (strncmp(dmn->name, dname, strlen(dmn->name)) == 0) | |
211 | break; | |
212 | } | |
213 | ||
214 | if (dmn) { | |
215 | dmn->ignore_timeout = ignore; | |
216 | vty_out(vty, "%s switching to %s\n", dmn->name, | |
217 | ignore ? "ignore" : "watch"); | |
218 | } else | |
219 | vty_out(vty, "%s is not configured for running at the moment", | |
220 | dname); | |
221 | } | |
222 | ||
4f04a76b | 223 | static void printhelp(FILE *target) |
8b886ca7 | 224 | { |
d62a17ae | 225 | fprintf(target, |
226 | "Usage : %s [OPTION...] <daemon name> ...\n\n\ | |
9473e340 | 227 | Watchdog program to monitor status of frr daemons and try to restart\n\ |
8b886ca7 | 228 | them if they are down or unresponsive. It determines whether a daemon is\n\ |
229 | up based on whether it can connect to the daemon's vty unix stream socket.\n\ | |
230 | It then repeatedly sends echo commands over that socket to determine whether\n\ | |
231 | the daemon is responsive. If the daemon crashes, we will receive an EOF\n\ | |
232 | on the socket connection and know immediately that the daemon is down.\n\n\ | |
233 | The daemons to be monitored should be listed on the command line.\n\n\ | |
8b886ca7 | 234 | In order to avoid attempting to restart the daemons in a fast loop,\n\ |
235 | the -m and -M options allow you to control the minimum delay between\n\ | |
236 | restart commands. The minimum restart delay is recalculated each time\n\ | |
237 | a restart is attempted: if the time since the last restart attempt exceeds\n\ | |
238 | twice the -M value, then the restart delay is set to the -m value.\n\ | |
d62a17ae | 239 | Otherwise, the interval is doubled (but capped at the -M value).\n\n", |
f168b713 | 240 | progname); |
e757c940 | 241 | |
d62a17ae | 242 | fprintf(target, |
243 | "Options:\n\ | |
8b886ca7 | 244 | -d, --daemon Run in daemon mode. In this mode, error messages are sent\n\ |
245 | to syslog instead of stdout.\n\ | |
246 | -S, --statedir Set the vty socket directory (default is %s)\n\ | |
33606a15 DL |
247 | -N, --pathspace Insert prefix into config & socket paths\n" |
248 | #ifdef GNU_LINUX | |
249 | " --netns Create and/or use Linux network namespace. If no name is\n" | |
250 | " given, uses the value from `-N`.\n" | |
251 | #endif | |
252 | "-l, --loglevel Set the logging level (default is %d).\n\ | |
8b886ca7 | 253 | The value should range from %d (LOG_EMERG) to %d (LOG_DEBUG),\n\ |
254 | but it can be set higher than %d if extra-verbose debugging\n\ | |
255 | messages are desired.\n\ | |
9272302b | 256 | --min-restart-interval\n\ |
8b886ca7 | 257 | Set the minimum seconds to wait between invocations of daemon\n\ |
258 | restart commands (default is %d).\n\ | |
9272302b | 259 | --max-restart-interval\n\ |
8b886ca7 | 260 | Set the maximum seconds to wait between invocations of daemon\n\ |
261 | restart commands (default is %d).\n\ | |
6d0fa5c2 DS |
262 | --operational-timeout\n\ |
263 | Set the time before systemd is notified that we are considered\n\ | |
264 | operational again after a daemon restart (default is %d).\n\ | |
8b886ca7 | 265 | -i, --interval Set the status polling interval in seconds (default is %d)\n\ |
266 | -t, --timeout Set the unresponsiveness timeout in seconds (default is %d)\n\ | |
267 | -T, --restart-timeout\n\ | |
268 | Set the restart (kill) timeout in seconds (default is %d).\n\ | |
269 | If any background jobs are still running after this much\n\ | |
270 | time has elapsed, they will be killed.\n\ | |
271 | -r, --restart Supply a Bourne shell command to use to restart a single\n\ | |
272 | daemon. The command string should include '%%s' where the\n\ | |
273 | name of the daemon should be substituted.\n\ | |
3ec95567 | 274 | (default: '%s')\n\ |
8b886ca7 | 275 | -s, --start-command\n\ |
276 | Supply a Bourne shell to command to use to start a single\n\ | |
277 | daemon. The command string should include '%%s' where the\n\ | |
278 | name of the daemon should be substituted.\n\ | |
3ec95567 | 279 | (default: '%s')\n\ |
8b886ca7 | 280 | -k, --kill-command\n\ |
281 | Supply a Bourne shell to command to use to stop a single\n\ | |
282 | daemon. The command string should include '%%s' where the\n\ | |
283 | name of the daemon should be substituted.\n\ | |
3ec95567 | 284 | (default: '%s')\n\ |
f168b713 | 285 | --dry Do not start or restart anything, just log.\n\ |
8b886ca7 | 286 | -p, --pid-file Set process identifier file name\n\ |
0a7c7856 | 287 | (default is %s/watchfrr.pid).\n\ |
c8b40f86 | 288 | -b, --blank-string\n\ |
289 | When the supplied argument string is found in any of the\n\ | |
f168b713 | 290 | various shell command arguments (-r, -s, or -k), replace\n\ |
c8b40f86 | 291 | it with a space. This is an ugly hack to circumvent problems\n\ |
292 | passing command-line arguments with embedded spaces.\n\ | |
8b886ca7 | 293 | -v, --version Print program version\n\ |
d62a17ae | 294 | -h, --help Display this help and exit\n", |
64a249ad | 295 | frr_vtydir, DEFAULT_LOGLEVEL, LOG_EMERG, LOG_DEBUG, LOG_DEBUG, |
6d0fa5c2 DS |
296 | DEFAULT_MIN_RESTART, DEFAULT_MAX_RESTART, |
297 | DEFAULT_OPERATIONAL_TIMEOUT, DEFAULT_PERIOD, DEFAULT_TIMEOUT, | |
298 | DEFAULT_RESTART_TIMEOUT, DEFAULT_RESTART_CMD, DEFAULT_START_CMD, | |
299 | DEFAULT_STOP_CMD, frr_vtydir); | |
8b886ca7 | 300 | } |
301 | ||
a6810074 | 302 | static pid_t run_background(char *shell_cmd) |
8b886ca7 | 303 | { |
a6810074 DL |
304 | pid_t child; |
305 | ||
306 | switch (child = fork()) { | |
307 | case -1: | |
450971aa | 308 | flog_err_sys(EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
309 | "fork failed, cannot run command [%s]: %s", |
310 | shell_cmd, safe_strerror(errno)); | |
a6810074 DL |
311 | return -1; |
312 | case 0: | |
313 | /* Child process. */ | |
d62a17ae | 314 | /* Use separate process group so child processes can be killed |
315 | * easily. */ | |
a6810074 | 316 | if (setpgid(0, 0) < 0) |
957cfa24 | 317 | zlog_warn("setpgid(0,0) failed: %s", |
a6810074 DL |
318 | safe_strerror(errno)); |
319 | { | |
320 | char shell[] = "sh"; | |
321 | char dashc[] = "-c"; | |
d62a17ae | 322 | char *const argv[4] = {shell, dashc, shell_cmd, NULL}; |
a6810074 | 323 | execv("/bin/sh", argv); |
450971aa | 324 | flog_err_sys(EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
325 | "execv(/bin/sh -c '%s') failed: %s", |
326 | shell_cmd, safe_strerror(errno)); | |
a6810074 DL |
327 | _exit(127); |
328 | } | |
329 | default: | |
330 | /* Parent process: we will reap the child later. */ | |
c3f65458 QY |
331 | zlog_info("Forked background command [pid %d]: %s", (int)child, |
332 | shell_cmd); | |
a6810074 DL |
333 | return child; |
334 | } | |
8b886ca7 | 335 | } |
336 | ||
a6810074 DL |
337 | static struct timeval *time_elapsed(struct timeval *result, |
338 | const struct timeval *start_time) | |
8b886ca7 | 339 | { |
a6810074 DL |
340 | gettimeofday(result, NULL); |
341 | result->tv_sec -= start_time->tv_sec; | |
342 | result->tv_usec -= start_time->tv_usec; | |
343 | while (result->tv_usec < 0) { | |
344 | result->tv_usec += 1000000L; | |
345 | result->tv_sec--; | |
346 | } | |
347 | return result; | |
8b886ca7 | 348 | } |
349 | ||
cc9f21da | 350 | static void restart_kill(struct thread *t_kill) |
8b886ca7 | 351 | { |
a6810074 DL |
352 | struct restart_info *restart = THREAD_ARG(t_kill); |
353 | struct timeval delay; | |
354 | ||
355 | time_elapsed(&delay, &restart->time); | |
2ab760f0 DA |
356 | |
357 | if (gs.reading_configuration) { | |
358 | zlog_err( | |
359 | "%s %s child process appears to still be reading configuration, delaying for another %lu time", | |
360 | restart->what, restart->name, gs.restart_timeout); | |
361 | thread_add_timer(master, restart_kill, restart, | |
362 | gs.restart_timeout, &restart->t_kill); | |
363 | return; | |
364 | } | |
365 | ||
d62a17ae | 366 | zlog_warn( |
957cfa24 | 367 | "%s %s child process %d still running after %ld seconds, sending signal %d", |
d62a17ae | 368 | restart->what, restart->name, (int)restart->pid, |
369 | (long)delay.tv_sec, (restart->kills ? SIGKILL : SIGTERM)); | |
a6810074 DL |
370 | kill(-restart->pid, (restart->kills ? SIGKILL : SIGTERM)); |
371 | restart->kills++; | |
66e78ae6 QY |
372 | thread_add_timer(master, restart_kill, restart, gs.restart_timeout, |
373 | &restart->t_kill); | |
8b886ca7 | 374 | } |
375 | ||
a6810074 | 376 | static struct restart_info *find_child(pid_t child) |
8b886ca7 | 377 | { |
f168b713 | 378 | struct daemon *dmn; |
7c265f7d CF |
379 | if (gs.restart.pid == child) |
380 | return &gs.restart; | |
381 | ||
f168b713 DL |
382 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { |
383 | if (dmn->restart.pid == child) | |
384 | return &dmn->restart; | |
a6810074 DL |
385 | } |
386 | return NULL; | |
8b886ca7 | 387 | } |
388 | ||
a6810074 | 389 | static void sigchild(void) |
8b886ca7 | 390 | { |
a6810074 DL |
391 | pid_t child; |
392 | int status; | |
393 | const char *name; | |
394 | const char *what; | |
395 | struct restart_info *restart; | |
75f8b0e4 | 396 | struct daemon *dmn; |
a6810074 DL |
397 | |
398 | switch (child = waitpid(-1, &status, WNOHANG)) { | |
399 | case -1: | |
450971aa | 400 | flog_err_sys(EC_LIB_SYSTEM_CALL, "waitpid failed: %s", |
09c866e3 | 401 | safe_strerror(errno)); |
a6810074 DL |
402 | return; |
403 | case 0: | |
404 | zlog_warn("SIGCHLD received, but waitpid did not reap a child"); | |
405 | return; | |
406 | } | |
407 | ||
408 | if (child == integrated_write_pid) { | |
409 | integrated_write_sigchld(status); | |
410 | return; | |
411 | } | |
412 | ||
413 | if ((restart = find_child(child)) != NULL) { | |
414 | name = restart->name; | |
415 | what = restart->what; | |
416 | restart->pid = 0; | |
417 | gs.numpids--; | |
b3d6bc6e MS |
418 | thread_cancel(&restart->t_kill); |
419 | ||
d62a17ae | 420 | /* Update restart time to reflect the time the command |
421 | * completed. */ | |
a6810074 DL |
422 | gettimeofday(&restart->time, NULL); |
423 | } else { | |
09c866e3 | 424 | flog_err_sys( |
450971aa | 425 | EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
426 | "waitpid returned status for an unknown child process %d", |
427 | (int)child); | |
a6810074 DL |
428 | name = "(unknown)"; |
429 | what = "background"; | |
430 | } | |
431 | if (WIFSTOPPED(status)) | |
957cfa24 | 432 | zlog_warn("%s %s process %d is stopped", what, name, |
d62a17ae | 433 | (int)child); |
a6810074 | 434 | else if (WIFSIGNALED(status)) |
d62a17ae | 435 | zlog_warn("%s %s process %d terminated due to signal %d", what, |
436 | name, (int)child, WTERMSIG(status)); | |
a6810074 DL |
437 | else if (WIFEXITED(status)) { |
438 | if (WEXITSTATUS(status) != 0) | |
d62a17ae | 439 | zlog_warn( |
440 | "%s %s process %d exited with non-zero status %d", | |
441 | what, name, (int)child, WEXITSTATUS(status)); | |
75f8b0e4 | 442 | else { |
a6810074 DL |
443 | zlog_debug("%s %s process %d exited normally", what, |
444 | name, (int)child); | |
75f8b0e4 DL |
445 | |
446 | if (restart && restart != &gs.restart) { | |
447 | dmn = container_of(restart, struct daemon, | |
448 | restart); | |
449 | restart_done(dmn); | |
450 | } else if (restart) | |
451 | for (dmn = gs.daemons; dmn; dmn = dmn->next) | |
452 | restart_done(dmn); | |
453 | } | |
a6810074 | 454 | } else |
09c866e3 | 455 | flog_err_sys( |
450971aa | 456 | EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
457 | "cannot interpret %s %s process %d wait status 0x%x", |
458 | what, name, (int)child, status); | |
a6810074 | 459 | phase_check(); |
8b886ca7 | 460 | } |
461 | ||
d62a17ae | 462 | static int run_job(struct restart_info *restart, const char *cmdtype, |
463 | const char *command, int force, int update_interval) | |
8b886ca7 | 464 | { |
a6810074 DL |
465 | struct timeval delay; |
466 | ||
467 | if (gs.loglevel > LOG_DEBUG + 1) | |
468 | zlog_debug("attempting to %s %s", cmdtype, restart->name); | |
469 | ||
470 | if (restart->pid) { | |
471 | if (gs.loglevel > LOG_DEBUG + 1) | |
d62a17ae | 472 | zlog_debug( |
473 | "cannot %s %s, previous pid %d still running", | |
474 | cmdtype, restart->name, (int)restart->pid); | |
a6810074 DL |
475 | return -1; |
476 | } | |
477 | ||
b3ee8bcc DS |
478 | char buffer[512]; |
479 | ||
480 | snprintf(buffer, sizeof(buffer), "restarting %s", restart->name); | |
481 | systemd_send_status(buffer); | |
b3ee8bcc | 482 | |
d62a17ae | 483 | /* Note: time_elapsed test must come before the force test, since we |
484 | need | |
a6810074 DL |
485 | to make sure that delay is initialized for use below in updating the |
486 | restart interval. */ | |
487 | if ((time_elapsed(&delay, &restart->time)->tv_sec < restart->interval) | |
488 | && !force) { | |
b3ee8bcc | 489 | |
a6810074 | 490 | if (gs.loglevel > LOG_DEBUG + 1) |
d62a17ae | 491 | zlog_debug( |
3efd0893 | 492 | "postponing %s %s: elapsed time %ld < retry interval %ld", |
d62a17ae | 493 | cmdtype, restart->name, (long)delay.tv_sec, |
494 | restart->interval); | |
a6810074 DL |
495 | return -1; |
496 | } | |
497 | ||
498 | gettimeofday(&restart->time, NULL); | |
499 | restart->kills = 0; | |
500 | { | |
501 | char cmd[strlen(command) + strlen(restart->name) + 1]; | |
c84e5187 DL |
502 | #pragma GCC diagnostic push |
503 | #pragma GCC diagnostic ignored "-Wformat-nonliteral" | |
504 | /* user supplied command string has a %s for the daemon name */ | |
a6810074 | 505 | snprintf(cmd, sizeof(cmd), command, restart->name); |
c84e5187 | 506 | #pragma GCC diagnostic pop |
a6810074 | 507 | if ((restart->pid = run_background(cmd)) > 0) { |
d62a17ae | 508 | thread_add_timer(master, restart_kill, restart, |
509 | gs.restart_timeout, &restart->t_kill); | |
a6810074 DL |
510 | restart->what = cmdtype; |
511 | gs.numpids++; | |
512 | } else | |
513 | restart->pid = 0; | |
514 | } | |
515 | ||
516 | /* Calculate the new restart interval. */ | |
517 | if (update_interval) { | |
518 | if (delay.tv_sec > 2 * gs.max_restart_interval) | |
519 | restart->interval = gs.min_restart_interval; | |
520 | else if ((restart->interval *= 2) > gs.max_restart_interval) | |
521 | restart->interval = gs.max_restart_interval; | |
522 | if (gs.loglevel > LOG_DEBUG + 1) | |
523 | zlog_debug("restart %s interval is now %ld", | |
524 | restart->name, restart->interval); | |
525 | } | |
526 | return restart->pid; | |
8b886ca7 | 527 | } |
528 | ||
d62a17ae | 529 | #define SET_READ_HANDLER(DMN) \ |
530 | do { \ | |
531 | (DMN)->t_read = NULL; \ | |
532 | thread_add_read(master, handle_read, (DMN), (DMN)->fd, \ | |
533 | &(DMN)->t_read); \ | |
534 | } while (0); | |
535 | ||
536 | #define SET_WAKEUP_DOWN(DMN) \ | |
537 | do { \ | |
538 | (DMN)->t_wakeup = NULL; \ | |
539 | thread_add_timer_msec(master, wakeup_down, (DMN), \ | |
540 | FUZZY(gs.period), &(DMN)->t_wakeup); \ | |
541 | } while (0); | |
542 | ||
543 | #define SET_WAKEUP_UNRESPONSIVE(DMN) \ | |
544 | do { \ | |
545 | (DMN)->t_wakeup = NULL; \ | |
546 | thread_add_timer_msec(master, wakeup_unresponsive, (DMN), \ | |
547 | FUZZY(gs.period), &(DMN)->t_wakeup); \ | |
548 | } while (0); | |
549 | ||
550 | #define SET_WAKEUP_ECHO(DMN) \ | |
551 | do { \ | |
552 | (DMN)->t_wakeup = NULL; \ | |
553 | thread_add_timer_msec(master, wakeup_send_echo, (DMN), \ | |
554 | FUZZY(gs.period), &(DMN)->t_wakeup); \ | |
555 | } while (0); | |
8b886ca7 | 556 | |
cc9f21da | 557 | static void wakeup_down(struct thread *t_wakeup) |
8b886ca7 | 558 | { |
a6810074 DL |
559 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
560 | ||
561 | dmn->t_wakeup = NULL; | |
562 | if (try_connect(dmn) < 0) | |
563 | SET_WAKEUP_DOWN(dmn); | |
564 | if ((dmn->connect_tries > 1) && (dmn->state != DAEMON_UP)) | |
565 | try_restart(dmn); | |
8b886ca7 | 566 | } |
567 | ||
cc9f21da | 568 | static void wakeup_init(struct thread *t_wakeup) |
8b886ca7 | 569 | { |
a6810074 DL |
570 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
571 | ||
572 | dmn->t_wakeup = NULL; | |
573 | if (try_connect(dmn) < 0) { | |
c3f65458 QY |
574 | zlog_info( |
575 | "%s state -> down : initial connection attempt failed", | |
576 | dmn->name); | |
a6810074 DL |
577 | dmn->state = DAEMON_DOWN; |
578 | } | |
c0e5cb52 | 579 | phase_check(); |
8b886ca7 | 580 | } |
581 | ||
75f8b0e4 DL |
582 | static void restart_done(struct daemon *dmn) |
583 | { | |
584 | if (dmn->state != DAEMON_DOWN) { | |
3f391bec DS |
585 | zlog_warn( |
586 | "Daemon: %s: is in %s state but expected it to be in DAEMON_DOWN state", | |
587 | dmn->name, state_str[dmn->state]); | |
75f8b0e4 DL |
588 | return; |
589 | } | |
28ef0ee1 | 590 | THREAD_OFF(dmn->t_wakeup); |
50478845 | 591 | |
75f8b0e4 DL |
592 | if (try_connect(dmn) < 0) |
593 | SET_WAKEUP_DOWN(dmn); | |
594 | } | |
595 | ||
6d0fa5c2 DS |
596 | static void daemon_restarting_operational(struct thread *thread) |
597 | { | |
598 | systemd_send_status("FRR Operational"); | |
599 | } | |
600 | ||
a6810074 | 601 | static void daemon_down(struct daemon *dmn, const char *why) |
8b886ca7 | 602 | { |
a6810074 | 603 | if (IS_UP(dmn) || (dmn->state == DAEMON_INIT)) |
1c50c1c0 QY |
604 | flog_err(EC_WATCHFRR_CONNECTION, "%s state -> down : %s", |
605 | dmn->name, why); | |
a6810074 DL |
606 | else if (gs.loglevel > LOG_DEBUG) |
607 | zlog_debug("%s still down : %s", dmn->name, why); | |
608 | if (IS_UP(dmn)) | |
609 | gs.numdown++; | |
610 | dmn->state = DAEMON_DOWN; | |
611 | if (dmn->fd >= 0) { | |
612 | close(dmn->fd); | |
613 | dmn->fd = -1; | |
614 | } | |
615 | THREAD_OFF(dmn->t_read); | |
616 | THREAD_OFF(dmn->t_write); | |
617 | THREAD_OFF(dmn->t_wakeup); | |
618 | if (try_connect(dmn) < 0) | |
619 | SET_WAKEUP_DOWN(dmn); | |
6d0fa5c2 DS |
620 | |
621 | systemd_send_status("FRR partially operational"); | |
a6810074 | 622 | phase_check(); |
8b886ca7 | 623 | } |
624 | ||
cc9f21da | 625 | static void handle_read(struct thread *t_read) |
8b886ca7 | 626 | { |
a6810074 DL |
627 | struct daemon *dmn = THREAD_ARG(t_read); |
628 | static const char resp[sizeof(PING_TOKEN) + 4] = PING_TOKEN "\n"; | |
629 | char buf[sizeof(resp) + 100]; | |
630 | ssize_t rc; | |
631 | struct timeval delay; | |
632 | ||
633 | dmn->t_read = NULL; | |
634 | if ((rc = read(dmn->fd, buf, sizeof(buf))) < 0) { | |
635 | char why[100]; | |
636 | ||
637 | if (ERRNO_IO_RETRY(errno)) { | |
638 | /* Pretend it never happened. */ | |
639 | SET_READ_HANDLER(dmn); | |
cc9f21da | 640 | return; |
a6810074 DL |
641 | } |
642 | snprintf(why, sizeof(why), "unexpected read error: %s", | |
643 | safe_strerror(errno)); | |
644 | daemon_down(dmn, why); | |
cc9f21da | 645 | return; |
8b886ca7 | 646 | } |
a6810074 DL |
647 | if (rc == 0) { |
648 | daemon_down(dmn, "read returned EOF"); | |
cc9f21da | 649 | return; |
a6810074 DL |
650 | } |
651 | if (!dmn->echo_sent.tv_sec) { | |
652 | char why[sizeof(buf) + 100]; | |
653 | snprintf(why, sizeof(why), | |
654 | "unexpected read returns %d bytes: %.*s", (int)rc, | |
655 | (int)rc, buf); | |
656 | daemon_down(dmn, why); | |
cc9f21da | 657 | return; |
8b886ca7 | 658 | } |
a6810074 DL |
659 | |
660 | /* We are expecting an echo response: is there any chance that the | |
661 | response would not be returned entirely in the first read? That | |
662 | seems inconceivable... */ | |
663 | if ((rc != sizeof(resp)) || memcmp(buf, resp, sizeof(resp))) { | |
664 | char why[100 + sizeof(buf)]; | |
665 | snprintf(why, sizeof(why), | |
3efd0893 | 666 | "read returned bad echo response of %d bytes (expecting %u): %.*s", |
d7c0a89a | 667 | (int)rc, (unsigned int)sizeof(resp), (int)rc, buf); |
a6810074 | 668 | daemon_down(dmn, why); |
cc9f21da | 669 | return; |
a6810074 DL |
670 | } |
671 | ||
672 | time_elapsed(&delay, &dmn->echo_sent); | |
673 | dmn->echo_sent.tv_sec = 0; | |
674 | if (dmn->state == DAEMON_UNRESPONSIVE) { | |
675 | if (delay.tv_sec < gs.timeout) { | |
676 | dmn->state = DAEMON_UP; | |
d62a17ae | 677 | zlog_warn( |
3efd0893 | 678 | "%s state -> up : echo response received after %ld.%06ld seconds", |
d62a17ae | 679 | dmn->name, (long)delay.tv_sec, |
680 | (long)delay.tv_usec); | |
a6810074 | 681 | } else |
d62a17ae | 682 | zlog_warn( |
3efd0893 | 683 | "%s: slow echo response finally received after %ld.%06ld seconds", |
d62a17ae | 684 | dmn->name, (long)delay.tv_sec, |
685 | (long)delay.tv_usec); | |
a6810074 DL |
686 | } else if (gs.loglevel > LOG_DEBUG + 1) |
687 | zlog_debug("%s: echo response received after %ld.%06ld seconds", | |
688 | dmn->name, (long)delay.tv_sec, (long)delay.tv_usec); | |
689 | ||
690 | SET_READ_HANDLER(dmn); | |
b3d6bc6e | 691 | thread_cancel(&dmn->t_wakeup); |
a6810074 | 692 | SET_WAKEUP_ECHO(dmn); |
8b886ca7 | 693 | } |
694 | ||
207e0d7a DS |
695 | /* |
696 | * Wait till we notice that all daemons are ready before | |
697 | * we send we are ready to systemd | |
698 | */ | |
5c9d1c83 | 699 | static void daemon_send_ready(int exitcode) |
207e0d7a | 700 | { |
5c9d1c83 | 701 | FILE *fp; |
a6810074 | 702 | static int sent = 0; |
43e587c1 | 703 | char started[1024]; |
207e0d7a | 704 | |
5c9d1c83 DL |
705 | if (sent) |
706 | return; | |
707 | ||
708 | if (exitcode == 0) | |
0a7c7856 | 709 | zlog_notice("all daemons up, doing startup-complete notify"); |
5c9d1c83 DL |
710 | else if (gs.numdown < gs.numdaemons) |
711 | flog_err(EC_WATCHFRR_CONNECTION, | |
3efd0893 | 712 | "startup did not complete within timeout (%d/%d daemons running)", |
5c9d1c83 DL |
713 | gs.numdaemons - gs.numdown, gs.numdaemons); |
714 | else { | |
715 | flog_err(EC_WATCHFRR_CONNECTION, | |
3efd0893 | 716 | "all configured daemons failed to start -- exiting watchfrr"); |
5c9d1c83 DL |
717 | exit(exitcode); |
718 | ||
719 | } | |
0a7c7856 | 720 | |
5c9d1c83 DL |
721 | frr_detach(); |
722 | ||
33606a15 | 723 | snprintf(started, sizeof(started), "%s/%s", frr_vtydir, |
3c649c71 DS |
724 | "watchfrr.started"); |
725 | fp = fopen(started, "w"); | |
5c9d1c83 DL |
726 | if (fp) |
727 | fclose(fp); | |
247898d5 DL |
728 | |
729 | systemd_send_started(master); | |
b3ee8bcc | 730 | systemd_send_status("FRR Operational"); |
5c9d1c83 | 731 | sent = 1; |
207e0d7a DS |
732 | } |
733 | ||
a6810074 | 734 | static void daemon_up(struct daemon *dmn, const char *why) |
8b886ca7 | 735 | { |
a6810074 DL |
736 | dmn->state = DAEMON_UP; |
737 | gs.numdown--; | |
738 | dmn->connect_tries = 0; | |
739 | zlog_notice("%s state -> up : %s", dmn->name, why); | |
6d0fa5c2 | 740 | if (gs.numdown == 0) { |
5c9d1c83 | 741 | daemon_send_ready(0); |
6d0fa5c2 DS |
742 | |
743 | THREAD_OFF(gs.t_operational); | |
744 | ||
745 | thread_add_timer(master, daemon_restarting_operational, NULL, | |
746 | gs.operational_timeout, &gs.t_operational); | |
747 | } | |
748 | ||
a8cbb8b3 | 749 | SET_WAKEUP_ECHO(dmn); |
a6810074 | 750 | phase_check(); |
8b886ca7 | 751 | } |
752 | ||
cc9f21da | 753 | static void check_connect(struct thread *t_write) |
8b886ca7 | 754 | { |
a6810074 DL |
755 | struct daemon *dmn = THREAD_ARG(t_write); |
756 | int sockerr; | |
757 | socklen_t reslen = sizeof(sockerr); | |
758 | ||
759 | dmn->t_write = NULL; | |
760 | if (getsockopt(dmn->fd, SOL_SOCKET, SO_ERROR, (char *)&sockerr, &reslen) | |
761 | < 0) { | |
762 | zlog_warn("%s: check_connect: getsockopt failed: %s", dmn->name, | |
763 | safe_strerror(errno)); | |
764 | daemon_down(dmn, | |
765 | "getsockopt failed checking connection success"); | |
cc9f21da | 766 | return; |
a6810074 DL |
767 | } |
768 | if ((reslen == sizeof(sockerr)) && sockerr) { | |
769 | char why[100]; | |
d62a17ae | 770 | snprintf( |
771 | why, sizeof(why), | |
772 | "getsockopt reports that connection attempt failed: %s", | |
773 | safe_strerror(sockerr)); | |
a6810074 | 774 | daemon_down(dmn, why); |
cc9f21da | 775 | return; |
a6810074 DL |
776 | } |
777 | ||
778 | daemon_up(dmn, "delayed connect succeeded"); | |
8b886ca7 | 779 | } |
780 | ||
cc9f21da | 781 | static void wakeup_connect_hanging(struct thread *t_wakeup) |
8b886ca7 | 782 | { |
a6810074 DL |
783 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
784 | char why[100]; | |
785 | ||
786 | dmn->t_wakeup = NULL; | |
787 | snprintf(why, sizeof(why), | |
788 | "connection attempt timed out after %ld seconds", gs.timeout); | |
789 | daemon_down(dmn, why); | |
8b886ca7 | 790 | } |
791 | ||
792 | /* Making connection to protocol daemon. */ | |
a6810074 | 793 | static int try_connect(struct daemon *dmn) |
8b886ca7 | 794 | { |
a6810074 DL |
795 | int sock; |
796 | struct sockaddr_un addr; | |
797 | socklen_t len; | |
798 | ||
799 | if (gs.loglevel > LOG_DEBUG + 1) | |
800 | zlog_debug("%s: attempting to connect", dmn->name); | |
801 | dmn->connect_tries++; | |
802 | ||
6006b807 | 803 | memset(&addr, 0, sizeof(addr)); |
a6810074 | 804 | addr.sun_family = AF_UNIX; |
d62a17ae | 805 | snprintf(addr.sun_path, sizeof(addr.sun_path), "%s/%s.vty", gs.vtydir, |
806 | dmn->name); | |
6f0e3f6e | 807 | #ifdef HAVE_STRUCT_SOCKADDR_UN_SUN_LEN |
a6810074 | 808 | len = addr.sun_len = SUN_LEN(&addr); |
8b886ca7 | 809 | #else |
a6810074 | 810 | len = sizeof(addr.sun_family) + strlen(addr.sun_path); |
d62a17ae | 811 | #endif /* HAVE_STRUCT_SOCKADDR_UN_SUN_LEN */ |
a6810074 DL |
812 | |
813 | /* Quick check to see if we might succeed before we go to the trouble | |
814 | of creating a socket. */ | |
815 | if (access(addr.sun_path, W_OK) < 0) { | |
816 | if (errno != ENOENT) | |
450971aa | 817 | flog_err_sys(EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
818 | "%s: access to socket %s denied: %s", |
819 | dmn->name, addr.sun_path, | |
820 | safe_strerror(errno)); | |
a6810074 DL |
821 | return -1; |
822 | } | |
823 | ||
824 | if ((sock = socket(AF_UNIX, SOCK_STREAM, 0)) < 0) { | |
450971aa | 825 | flog_err_sys(EC_LIB_SOCKET, "%s(%s): cannot make socket: %s", |
09c866e3 | 826 | __func__, addr.sun_path, safe_strerror(errno)); |
a6810074 DL |
827 | return -1; |
828 | } | |
829 | ||
830 | if (set_nonblocking(sock) < 0 || set_cloexec(sock) < 0) { | |
450971aa | 831 | flog_err_sys(EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
832 | "%s(%s): set_nonblocking/cloexec(%d) failed", |
833 | __func__, addr.sun_path, sock); | |
a6810074 DL |
834 | close(sock); |
835 | return -1; | |
8b886ca7 | 836 | } |
a6810074 DL |
837 | |
838 | if (connect(sock, (struct sockaddr *)&addr, len) < 0) { | |
839 | if ((errno != EINPROGRESS) && (errno != EWOULDBLOCK)) { | |
840 | if (gs.loglevel > LOG_DEBUG) | |
841 | zlog_debug("%s(%s): connect failed: %s", | |
842 | __func__, addr.sun_path, | |
843 | safe_strerror(errno)); | |
844 | close(sock); | |
845 | return -1; | |
846 | } | |
847 | if (gs.loglevel > LOG_DEBUG) | |
848 | zlog_debug("%s: connection in progress", dmn->name); | |
849 | dmn->state = DAEMON_CONNECTING; | |
850 | dmn->fd = sock; | |
66e78ae6 | 851 | thread_add_write(master, check_connect, dmn, dmn->fd, |
d62a17ae | 852 | &dmn->t_write); |
d62a17ae | 853 | thread_add_timer(master, wakeup_connect_hanging, dmn, |
854 | gs.timeout, &dmn->t_wakeup); | |
a6810074 DL |
855 | SET_READ_HANDLER(dmn); |
856 | return 0; | |
857 | } | |
858 | ||
859 | dmn->fd = sock; | |
860 | SET_READ_HANDLER(dmn); | |
861 | daemon_up(dmn, "connect succeeded"); | |
862 | return 1; | |
8b886ca7 | 863 | } |
864 | ||
cc9f21da | 865 | static void phase_hanging(struct thread *t_hanging) |
8b886ca7 | 866 | { |
a6810074 | 867 | gs.t_phase_hanging = NULL; |
f74ae2bb | 868 | flog_err(EC_WATCHFRR_CONNECTION, |
1c50c1c0 QY |
869 | "Phase [%s] hanging for %ld seconds, aborting phased restart", |
870 | phase_str[gs.phase], PHASE_TIMEOUT); | |
a6810074 | 871 | gs.phase = PHASE_NONE; |
8b886ca7 | 872 | } |
873 | ||
f1692c51 | 874 | static void set_phase(enum restart_phase new_phase) |
8b886ca7 | 875 | { |
a6810074 | 876 | gs.phase = new_phase; |
b3d6bc6e MS |
877 | thread_cancel(&gs.t_phase_hanging); |
878 | ||
66e78ae6 QY |
879 | thread_add_timer(master, phase_hanging, NULL, PHASE_TIMEOUT, |
880 | &gs.t_phase_hanging); | |
8b886ca7 | 881 | } |
882 | ||
a6810074 | 883 | static void phase_check(void) |
8b886ca7 | 884 | { |
c0e5cb52 DL |
885 | struct daemon *dmn; |
886 | ||
a6810074 DL |
887 | switch (gs.phase) { |
888 | case PHASE_NONE: | |
889 | break; | |
c0e5cb52 DL |
890 | |
891 | case PHASE_INIT: | |
892 | for (dmn = gs.daemons; dmn; dmn = dmn->next) | |
893 | if (dmn->state == DAEMON_INIT) | |
894 | return; | |
895 | ||
896 | /* startup complete, everything out of INIT */ | |
897 | gs.phase = PHASE_NONE; | |
898 | for (dmn = gs.daemons; dmn; dmn = dmn->next) | |
899 | if (dmn->state == DAEMON_DOWN) { | |
900 | SET_WAKEUP_DOWN(dmn); | |
901 | try_restart(dmn); | |
902 | } | |
903 | break; | |
a6810074 DL |
904 | case PHASE_STOPS_PENDING: |
905 | if (gs.numpids) | |
906 | break; | |
d62a17ae | 907 | zlog_info( |
908 | "Phased restart: all routing daemon stop jobs have completed."); | |
a6810074 DL |
909 | set_phase(PHASE_WAITING_DOWN); |
910 | ||
d62a17ae | 911 | /*FALLTHRU*/ |
a6810074 DL |
912 | case PHASE_WAITING_DOWN: |
913 | if (gs.numdown + IS_UP(gs.special) < gs.numdaemons) | |
914 | break; | |
6d0fa5c2 | 915 | systemd_send_status("Phased Restart"); |
a6810074 DL |
916 | zlog_info("Phased restart: all routing daemons now down."); |
917 | run_job(&gs.special->restart, "restart", gs.restart_command, 1, | |
918 | 1); | |
919 | set_phase(PHASE_ZEBRA_RESTART_PENDING); | |
920 | ||
d62a17ae | 921 | /*FALLTHRU*/ |
a6810074 DL |
922 | case PHASE_ZEBRA_RESTART_PENDING: |
923 | if (gs.special->restart.pid) | |
924 | break; | |
6d0fa5c2 | 925 | systemd_send_status("Zebra Restarting"); |
a6810074 DL |
926 | zlog_info("Phased restart: %s restart job completed.", |
927 | gs.special->name); | |
928 | set_phase(PHASE_WAITING_ZEBRA_UP); | |
929 | ||
d62a17ae | 930 | /*FALLTHRU*/ |
a6810074 DL |
931 | case PHASE_WAITING_ZEBRA_UP: |
932 | if (!IS_UP(gs.special)) | |
933 | break; | |
934 | zlog_info("Phased restart: %s is now up.", gs.special->name); | |
9ca6d3db DS |
935 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { |
936 | if (dmn != gs.special) | |
937 | run_job(&dmn->restart, "start", | |
938 | gs.start_command, 1, 0); | |
a6810074 DL |
939 | } |
940 | gs.phase = PHASE_NONE; | |
941 | THREAD_OFF(gs.t_phase_hanging); | |
942 | zlog_notice("Phased global restart has completed."); | |
943 | break; | |
944 | } | |
8b886ca7 | 945 | } |
946 | ||
a6810074 | 947 | static void try_restart(struct daemon *dmn) |
8b886ca7 | 948 | { |
f168b713 | 949 | if (watch_only) |
a6810074 | 950 | return; |
a6810074 | 951 | |
f168b713 DL |
952 | if (dmn != gs.special) { |
953 | if ((gs.special->state == DAEMON_UP) | |
954 | && (gs.phase == PHASE_NONE)) | |
955 | run_job(&dmn->restart, "restart", gs.restart_command, 0, | |
956 | 1); | |
957 | else | |
958 | zlog_debug( | |
3efd0893 | 959 | "%s: postponing restart attempt because master %s daemon not up [%s], or phased restart in progress", |
f168b713 DL |
960 | dmn->name, gs.special->name, |
961 | state_str[gs.special->state]); | |
962 | return; | |
963 | } | |
964 | ||
965 | if ((gs.phase != PHASE_NONE) || gs.numpids) { | |
966 | if (gs.loglevel > LOG_DEBUG + 1) | |
967 | zlog_debug( | |
3efd0893 | 968 | "postponing phased global restart: restart already in progress [%s], or outstanding child processes [%d]", |
f168b713 DL |
969 | phase_str[gs.phase], gs.numpids); |
970 | return; | |
971 | } | |
972 | /* Is it too soon for a restart? */ | |
973 | { | |
974 | struct timeval delay; | |
975 | if (time_elapsed(&delay, &gs.special->restart.time)->tv_sec | |
976 | < gs.special->restart.interval) { | |
a6810074 | 977 | if (gs.loglevel > LOG_DEBUG + 1) |
d62a17ae | 978 | zlog_debug( |
3efd0893 | 979 | "postponing phased global restart: elapsed time %ld < retry interval %ld", |
f168b713 DL |
980 | (long)delay.tv_sec, |
981 | gs.special->restart.interval); | |
982 | return; | |
a6810074 | 983 | } |
8b886ca7 | 984 | } |
f168b713 | 985 | run_job(&gs.restart, "restart", gs.restart_command, 0, 1); |
8b886ca7 | 986 | } |
987 | ||
cc9f21da | 988 | static void wakeup_unresponsive(struct thread *t_wakeup) |
8b886ca7 | 989 | { |
a6810074 DL |
990 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
991 | ||
992 | dmn->t_wakeup = NULL; | |
993 | if (dmn->state != DAEMON_UNRESPONSIVE) | |
f74ae2bb | 994 | flog_err(EC_WATCHFRR_CONNECTION, |
3efd0893 | 995 | "%s: no longer unresponsive (now %s), wakeup should have been cancelled!", |
1c50c1c0 | 996 | dmn->name, state_str[dmn->state]); |
a6810074 DL |
997 | else { |
998 | SET_WAKEUP_UNRESPONSIVE(dmn); | |
999 | try_restart(dmn); | |
1000 | } | |
8b886ca7 | 1001 | } |
1002 | ||
cc9f21da | 1003 | static void wakeup_no_answer(struct thread *t_wakeup) |
8b886ca7 | 1004 | { |
a6810074 DL |
1005 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
1006 | ||
1007 | dmn->t_wakeup = NULL; | |
1008 | dmn->state = DAEMON_UNRESPONSIVE; | |
cc53b605 | 1009 | if (dmn->ignore_timeout) |
cc9f21da | 1010 | return; |
f74ae2bb | 1011 | flog_err(EC_WATCHFRR_CONNECTION, |
3efd0893 | 1012 | "%s state -> unresponsive : no response yet to ping sent %ld seconds ago", |
1c50c1c0 | 1013 | dmn->name, gs.timeout); |
71e7975a DL |
1014 | SET_WAKEUP_UNRESPONSIVE(dmn); |
1015 | try_restart(dmn); | |
8b886ca7 | 1016 | } |
1017 | ||
cc9f21da | 1018 | static void wakeup_send_echo(struct thread *t_wakeup) |
8b886ca7 | 1019 | { |
a6810074 DL |
1020 | static const char echocmd[] = "echo " PING_TOKEN; |
1021 | ssize_t rc; | |
1022 | struct daemon *dmn = THREAD_ARG(t_wakeup); | |
1023 | ||
1024 | dmn->t_wakeup = NULL; | |
d62a17ae | 1025 | if (((rc = write(dmn->fd, echocmd, sizeof(echocmd))) < 0) |
1026 | || ((size_t)rc != sizeof(echocmd))) { | |
a6810074 DL |
1027 | char why[100 + sizeof(echocmd)]; |
1028 | snprintf(why, sizeof(why), | |
1029 | "write '%s' returned %d instead of %u", echocmd, | |
d7c0a89a | 1030 | (int)rc, (unsigned int)sizeof(echocmd)); |
a6810074 DL |
1031 | daemon_down(dmn, why); |
1032 | } else { | |
1033 | gettimeofday(&dmn->echo_sent, NULL); | |
66e78ae6 QY |
1034 | thread_add_timer(master, wakeup_no_answer, dmn, gs.timeout, |
1035 | &dmn->t_wakeup); | |
a6810074 | 1036 | } |
8b886ca7 | 1037 | } |
1038 | ||
470bc619 QY |
1039 | bool check_all_up(void) |
1040 | { | |
1041 | struct daemon *dmn; | |
1042 | ||
1043 | for (dmn = gs.daemons; dmn; dmn = dmn->next) | |
1044 | if (dmn->state != DAEMON_UP) | |
1045 | return false; | |
1046 | return true; | |
1047 | } | |
1048 | ||
af568444 DL |
1049 | void watchfrr_status(struct vty *vty) |
1050 | { | |
1051 | struct daemon *dmn; | |
1052 | struct timeval delay; | |
1053 | ||
1054 | vty_out(vty, "watchfrr global phase: %s\n", phase_str[gs.phase]); | |
603fef0e DS |
1055 | vty_out(vty, " Restart Command: %pSQq\n", gs.restart_command); |
1056 | vty_out(vty, " Start Command: %pSQq\n", gs.start_command); | |
1057 | vty_out(vty, " Stop Command: %pSQq\n", gs.stop_command); | |
1058 | vty_out(vty, " Min Restart Interval: %ld\n", gs.min_restart_interval); | |
1059 | vty_out(vty, " Max Restart Interval: %ld\n", gs.max_restart_interval); | |
1060 | vty_out(vty, " Restart Timeout: %ld\n", gs.restart_timeout); | |
2ab760f0 DA |
1061 | vty_out(vty, " Reading Configuration: %s\n", |
1062 | gs.reading_configuration ? "yes" : "no"); | |
af568444 DL |
1063 | if (gs.restart.pid) |
1064 | vty_out(vty, " global restart running, pid %ld\n", | |
1065 | (long)gs.restart.pid); | |
1066 | ||
1067 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { | |
cc53b605 DS |
1068 | vty_out(vty, " %-20s %s%s", dmn->name, state_str[dmn->state], |
1069 | dmn->ignore_timeout ? "/Ignoring Timeout\n" : "\n"); | |
af568444 DL |
1070 | if (dmn->restart.pid) |
1071 | vty_out(vty, " restart running, pid %ld\n", | |
1072 | (long)dmn->restart.pid); | |
1073 | else if (dmn->state == DAEMON_DOWN && | |
1074 | time_elapsed(&delay, &dmn->restart.time)->tv_sec | |
1075 | < dmn->restart.interval) | |
3efd0893 | 1076 | vty_out(vty, " restarting in %jd seconds (%jds backoff interval)\n", |
051a0be4 DL |
1077 | (intmax_t)dmn->restart.interval |
1078 | - (intmax_t)delay.tv_sec, | |
1079 | (intmax_t)dmn->restart.interval); | |
af568444 DL |
1080 | } |
1081 | } | |
1082 | ||
a6810074 | 1083 | static void sigint(void) |
8b886ca7 | 1084 | { |
a6810074 DL |
1085 | zlog_notice("Terminating on signal"); |
1086 | systemd_send_stopping(); | |
1087 | exit(0); | |
8b886ca7 | 1088 | } |
1089 | ||
a6810074 | 1090 | static int valid_command(const char *cmd) |
8b886ca7 | 1091 | { |
a6810074 | 1092 | char *p; |
8b886ca7 | 1093 | |
53a78fc1 RZ |
1094 | if (cmd == NULL) |
1095 | return 0; | |
1096 | ||
a6810074 | 1097 | return ((p = strchr(cmd, '%')) != NULL) && (*(p + 1) == 's') |
d62a17ae | 1098 | && !strchr(p + 1, '%'); |
8b886ca7 | 1099 | } |
1100 | ||
c8b40f86 | 1101 | /* This is an ugly hack to circumvent problems with passing command-line |
1102 | arguments that contain spaces. The fix is to use a configuration file. */ | |
a6810074 | 1103 | static char *translate_blanks(const char *cmd, const char *blankstr) |
c8b40f86 | 1104 | { |
a6810074 DL |
1105 | char *res; |
1106 | char *p; | |
1107 | size_t bslen = strlen(blankstr); | |
1108 | ||
1109 | if (!(res = strdup(cmd))) { | |
1110 | perror("strdup"); | |
1111 | exit(1); | |
1112 | } | |
1113 | while ((p = strstr(res, blankstr)) != NULL) { | |
1114 | *p = ' '; | |
1115 | if (bslen != 1) | |
1116 | memmove(p + 1, p + bslen, strlen(p + bslen) + 1); | |
1117 | } | |
1118 | return res; | |
c8b40f86 | 1119 | } |
1120 | ||
cc9f21da | 1121 | static void startup_timeout(struct thread *t_wakeup) |
5c9d1c83 DL |
1122 | { |
1123 | daemon_send_ready(1); | |
5c9d1c83 DL |
1124 | } |
1125 | ||
33606a15 DL |
1126 | #ifdef GNU_LINUX |
1127 | ||
1128 | #include <sys/mount.h> | |
1129 | #include <sched.h> | |
1130 | ||
1131 | #define NETNS_RUN_DIR "/var/run/netns" | |
1132 | ||
1133 | static void netns_create(int dirfd, const char *nsname) | |
1134 | { | |
1135 | /* make /var/run/netns shared between mount namespaces | |
1136 | * just like iproute2 sets it up | |
1137 | */ | |
1138 | if (mount("", NETNS_RUN_DIR, "none", MS_SHARED | MS_REC, NULL)) { | |
1139 | if (errno != EINVAL) { | |
1140 | perror("mount"); | |
1141 | exit(1); | |
1142 | } | |
1143 | ||
1144 | if (mount(NETNS_RUN_DIR, NETNS_RUN_DIR, "none", | |
1145 | MS_BIND | MS_REC, NULL)) { | |
1146 | perror("mount"); | |
1147 | exit(1); | |
1148 | } | |
1149 | ||
1150 | if (mount("", NETNS_RUN_DIR, "none", MS_SHARED | MS_REC, | |
1151 | NULL)) { | |
1152 | perror("mount"); | |
1153 | exit(1); | |
1154 | } | |
1155 | } | |
1156 | ||
1157 | /* need an empty file to mount on top of */ | |
1158 | int nsfd = openat(dirfd, nsname, O_CREAT | O_RDONLY | O_EXCL, 0); | |
1159 | ||
1160 | if (nsfd < 0) { | |
1161 | fprintf(stderr, "failed to create \"%s/%s\": %s\n", | |
1162 | NETNS_RUN_DIR, nsname, strerror(errno)); | |
1163 | exit(1); | |
1164 | } | |
1165 | close(nsfd); | |
1166 | ||
1167 | if (unshare(CLONE_NEWNET)) { | |
1168 | perror("unshare"); | |
1169 | unlinkat(dirfd, nsname, 0); | |
1170 | exit(1); | |
1171 | } | |
1172 | ||
1173 | char *dstpath = asprintfrr(MTYPE_TMP, "%s/%s", NETNS_RUN_DIR, nsname); | |
1174 | ||
1175 | /* bind-mount so the namespace has a name and is persistent */ | |
1176 | if (mount("/proc/self/ns/net", dstpath, "none", MS_BIND, NULL) < 0) { | |
1177 | fprintf(stderr, "failed to bind-mount netns to \"%s\": %s\n", | |
1178 | dstpath, strerror(errno)); | |
1179 | unlinkat(dirfd, nsname, 0); | |
1180 | exit(1); | |
1181 | } | |
1182 | ||
1183 | XFREE(MTYPE_TMP, dstpath); | |
1184 | } | |
1185 | ||
1186 | static void netns_setup(const char *nsname) | |
1187 | { | |
1188 | int dirfd, nsfd; | |
1189 | ||
1190 | dirfd = open(NETNS_RUN_DIR, O_DIRECTORY | O_RDONLY); | |
1191 | if (dirfd < 0) { | |
1192 | if (errno == ENOTDIR) { | |
1193 | fprintf(stderr, "error: \"%s\" is not a directory!\n", | |
1194 | NETNS_RUN_DIR); | |
1195 | exit(1); | |
1196 | } else if (errno == ENOENT) { | |
1197 | if (mkdir(NETNS_RUN_DIR, 0755)) { | |
1198 | fprintf(stderr, "error: \"%s\": mkdir: %s\n", | |
1199 | NETNS_RUN_DIR, strerror(errno)); | |
1200 | exit(1); | |
1201 | } | |
1202 | dirfd = open(NETNS_RUN_DIR, O_DIRECTORY | O_RDONLY); | |
1203 | if (dirfd < 0) { | |
1204 | fprintf(stderr, "error: \"%s\": opendir: %s\n", | |
1205 | NETNS_RUN_DIR, strerror(errno)); | |
1206 | exit(1); | |
1207 | } | |
1208 | } else { | |
1209 | fprintf(stderr, "error: \"%s\": %s\n", | |
1210 | NETNS_RUN_DIR, strerror(errno)); | |
1211 | exit(1); | |
1212 | } | |
1213 | } | |
1214 | ||
1215 | nsfd = openat(dirfd, nsname, O_RDONLY); | |
1216 | if (nsfd < 0 && errno != ENOENT) { | |
1217 | fprintf(stderr, "error: \"%s/%s\": %s\n", | |
1218 | NETNS_RUN_DIR, nsname, strerror(errno)); | |
1219 | exit(1); | |
1220 | } | |
1221 | if (nsfd < 0) | |
1222 | netns_create(dirfd, nsname); | |
1223 | else { | |
1224 | if (setns(nsfd, CLONE_NEWNET)) { | |
1225 | perror("setns"); | |
1226 | exit(1); | |
1227 | } | |
1228 | close(nsfd); | |
1229 | } | |
1230 | close(dirfd); | |
1231 | ||
1232 | /* make sure loopback is up... weird things happen otherwise. | |
1233 | * ioctl is perfectly fine for this, don't need netlink... | |
1234 | */ | |
1235 | int sockfd; | |
1236 | struct ifreq ifr = { }; | |
1237 | ||
1238 | strlcpy(ifr.ifr_name, "lo", sizeof(ifr.ifr_name)); | |
1239 | ||
1240 | sockfd = socket(AF_INET, SOCK_DGRAM, 0); | |
1241 | if (sockfd < 0) { | |
1242 | perror("socket"); | |
1243 | exit(1); | |
1244 | } | |
1245 | if (ioctl(sockfd, SIOCGIFFLAGS, &ifr)) { | |
1246 | perror("ioctl(SIOCGIFFLAGS, \"lo\")"); | |
1247 | exit(1); | |
1248 | } | |
1249 | if (!(ifr.ifr_flags & IFF_UP)) { | |
1250 | ifr.ifr_flags |= IFF_UP; | |
1251 | if (ioctl(sockfd, SIOCSIFFLAGS, &ifr)) { | |
1252 | perror("ioctl(SIOCSIFFLAGS, \"lo\")"); | |
1253 | exit(1); | |
1254 | } | |
1255 | } | |
1256 | close(sockfd); | |
1257 | } | |
1258 | ||
1259 | #else /* !GNU_LINUX */ | |
1260 | ||
1261 | static void netns_setup(const char *nsname) | |
1262 | { | |
1263 | fprintf(stderr, "network namespaces are only available on Linux\n"); | |
1264 | exit(1); | |
1265 | } | |
1266 | #endif | |
1267 | ||
2ab760f0 DA |
1268 | static void watchfrr_start_config(void) |
1269 | { | |
1270 | gs.reading_configuration = true; | |
1271 | } | |
1272 | ||
1273 | static void watchfrr_end_config(void) | |
1274 | { | |
1275 | gs.reading_configuration = false; | |
1276 | } | |
1277 | ||
0a7c7856 DL |
1278 | static void watchfrr_init(int argc, char **argv) |
1279 | { | |
1280 | const char *special = "zebra"; | |
1281 | int i; | |
1282 | struct daemon *dmn, **add = &gs.daemons; | |
1283 | char alldaemons[512] = "", *p = alldaemons; | |
1284 | ||
5c9d1c83 DL |
1285 | thread_add_timer_msec(master, startup_timeout, NULL, STARTUP_TIMEOUT, |
1286 | &gs.t_startup_timeout); | |
1287 | ||
0a7c7856 DL |
1288 | for (i = optind; i < argc; i++) { |
1289 | dmn = XCALLOC(MTYPE_WATCHFRR_DAEMON, sizeof(*dmn)); | |
1290 | ||
1291 | dmn->name = dmn->restart.name = argv[i]; | |
1292 | dmn->state = DAEMON_INIT; | |
1293 | gs.numdaemons++; | |
1294 | gs.numdown++; | |
1295 | dmn->fd = -1; | |
c0e5cb52 | 1296 | thread_add_timer_msec(master, wakeup_init, dmn, 0, |
0a7c7856 DL |
1297 | &dmn->t_wakeup); |
1298 | dmn->restart.interval = gs.min_restart_interval; | |
1299 | *add = dmn; | |
1300 | add = &dmn->next; | |
1301 | ||
1302 | if (!strcmp(dmn->name, special)) | |
1303 | gs.special = dmn; | |
1304 | } | |
1305 | ||
1306 | if (!gs.daemons) { | |
1307 | fprintf(stderr, | |
1308 | "Must specify one or more daemons to monitor.\n\n"); | |
1309 | frr_help_exit(1); | |
1310 | } | |
1311 | if (!watch_only && !gs.special) { | |
1312 | fprintf(stderr, "\"%s\" daemon must be in daemon lists\n\n", | |
1313 | special); | |
1314 | frr_help_exit(1); | |
1315 | } | |
1316 | ||
1317 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { | |
1318 | snprintf(p, alldaemons + sizeof(alldaemons) - p, "%s%s", | |
1319 | (p == alldaemons) ? "" : " ", dmn->name); | |
1320 | p += strlen(p); | |
1321 | } | |
1322 | zlog_notice("%s %s watching [%s]%s", progname, FRR_VERSION, alldaemons, | |
1323 | watch_only ? ", monitor mode" : ""); | |
1324 | } | |
1325 | ||
a6810074 | 1326 | struct zebra_privs_t watchfrr_privs = { |
95c4aff2 | 1327 | #ifdef VTY_GROUP |
a6810074 | 1328 | .vty_group = VTY_GROUP, |
95c4aff2 DL |
1329 | #endif |
1330 | }; | |
1331 | ||
7cc91e67 | 1332 | static struct frr_signal_t watchfrr_signals[] = { |
4f04a76b DL |
1333 | { |
1334 | .signal = SIGINT, | |
1335 | .handler = sigint, | |
1336 | }, | |
1337 | { | |
1338 | .signal = SIGTERM, | |
1339 | .handler = sigint, | |
1340 | }, | |
1341 | { | |
1342 | .signal = SIGCHLD, | |
1343 | .handler = sigchild, | |
1344 | }, | |
1345 | }; | |
1346 | ||
1347 | FRR_DAEMON_INFO(watchfrr, WATCHFRR, | |
d62a17ae | 1348 | .flags = FRR_NO_PRIVSEP | FRR_NO_TCPVTY | FRR_LIMITED_CLI |
0a7c7856 DL |
1349 | | FRR_NO_CFG_PID_DRY | FRR_NO_ZCLIENT |
1350 | | FRR_DETACH_LATER, | |
4f04a76b | 1351 | |
d62a17ae | 1352 | .printhelp = printhelp, |
1353 | .copyright = "Copyright 2004 Andrew J. Schorr", | |
4f04a76b | 1354 | |
d62a17ae | 1355 | .signals = watchfrr_signals, |
1356 | .n_signals = array_size(watchfrr_signals), | |
4f04a76b | 1357 | |
80413c20 DL |
1358 | .privs = &watchfrr_privs, |
1359 | ); | |
4f04a76b | 1360 | |
999f153e DL |
1361 | #define DEPRECATED_OPTIONS "aAezR:" |
1362 | ||
a6810074 | 1363 | int main(int argc, char **argv) |
8b886ca7 | 1364 | { |
a6810074 | 1365 | int opt; |
a6810074 | 1366 | const char *blankstr = NULL; |
33606a15 DL |
1367 | const char *netns = NULL; |
1368 | bool netns_en = false; | |
a6810074 | 1369 | |
4f04a76b DL |
1370 | frr_preinit(&watchfrr_di, argc, argv); |
1371 | progname = watchfrr_di.progname; | |
1372 | ||
33606a15 | 1373 | frr_opt_add("b:di:k:l:N:p:r:S:s:t:T:" DEPRECATED_OPTIONS, longopts, ""); |
a6810074 DL |
1374 | |
1375 | gs.restart.name = "all"; | |
4f04a76b | 1376 | while ((opt = frr_getopt(argc, argv, NULL)) != EOF) { |
999f153e DL |
1377 | if (opt && opt < 128 && strchr(DEPRECATED_OPTIONS, opt)) { |
1378 | fprintf(stderr, | |
1379 | "The -%c option no longer exists.\n" | |
1380 | "Please refer to the watchfrr(8) man page.\n", | |
1381 | opt); | |
1382 | exit(1); | |
1383 | } | |
1384 | ||
a6810074 DL |
1385 | switch (opt) { |
1386 | case 0: | |
1387 | break; | |
a6810074 DL |
1388 | case 'b': |
1389 | blankstr = optarg; | |
1390 | break; | |
f168b713 DL |
1391 | case OPTION_DRY: |
1392 | watch_only = true; | |
a6810074 DL |
1393 | break; |
1394 | case 'k': | |
1395 | if (!valid_command(optarg)) { | |
1396 | fprintf(stderr, | |
1397 | "Invalid kill command, must contain '%%s': %s\n", | |
1398 | optarg); | |
4f04a76b | 1399 | frr_help_exit(1); |
a6810074 DL |
1400 | } |
1401 | gs.stop_command = optarg; | |
1402 | break; | |
d62a17ae | 1403 | case 'l': { |
1404 | char garbage[3]; | |
1405 | if ((sscanf(optarg, "%d%1s", &gs.loglevel, garbage) | |
1406 | != 1) | |
1407 | || (gs.loglevel < LOG_EMERG)) { | |
1408 | fprintf(stderr, | |
1409 | "Invalid loglevel argument: %s\n", | |
1410 | optarg); | |
1411 | frr_help_exit(1); | |
a6810074 | 1412 | } |
d62a17ae | 1413 | } break; |
1414 | case OPTION_MINRESTART: { | |
1415 | char garbage[3]; | |
1416 | if ((sscanf(optarg, "%ld%1s", &gs.min_restart_interval, | |
1417 | garbage) | |
1418 | != 1) | |
1419 | || (gs.min_restart_interval < 0)) { | |
1420 | fprintf(stderr, | |
1421 | "Invalid min_restart_interval argument: %s\n", | |
1422 | optarg); | |
1423 | frr_help_exit(1); | |
a6810074 | 1424 | } |
d62a17ae | 1425 | } break; |
1426 | case OPTION_MAXRESTART: { | |
1427 | char garbage[3]; | |
1428 | if ((sscanf(optarg, "%ld%1s", &gs.max_restart_interval, | |
1429 | garbage) | |
1430 | != 1) | |
1431 | || (gs.max_restart_interval < 0)) { | |
1432 | fprintf(stderr, | |
1433 | "Invalid max_restart_interval argument: %s\n", | |
1434 | optarg); | |
1435 | frr_help_exit(1); | |
a6810074 | 1436 | } |
d62a17ae | 1437 | } break; |
6d0fa5c2 DS |
1438 | case OPTION_MAXOPERATIONAL: { |
1439 | char garbage[3]; | |
1440 | ||
1441 | if ((sscanf(optarg, "%ld%1s", &gs.operational_timeout, | |
1442 | garbage) != 1) || | |
7a8120da | 1443 | (gs.operational_timeout < 0)) { |
6d0fa5c2 DS |
1444 | fprintf(stderr, |
1445 | "Invalid Operational_timeout argument: %s\n", | |
1446 | optarg); | |
1447 | frr_help_exit(1); | |
1448 | } | |
1449 | } break; | |
33606a15 DL |
1450 | case OPTION_NETNS: |
1451 | netns_en = true; | |
b12bc77c | 1452 | if (optarg && strchr(optarg, '/')) { |
33606a15 DL |
1453 | fprintf(stderr, |
1454 | "invalid network namespace name \"%s\" (may not contain slashes)\n", | |
1455 | optarg); | |
1456 | frr_help_exit(1); | |
1457 | } | |
1458 | netns = optarg; | |
1459 | break; | |
d62a17ae | 1460 | case 'i': { |
1461 | char garbage[3]; | |
1462 | int period; | |
1463 | if ((sscanf(optarg, "%d%1s", &period, garbage) != 1) | |
1464 | || (gs.period < 1)) { | |
1465 | fprintf(stderr, | |
1466 | "Invalid interval argument: %s\n", | |
1467 | optarg); | |
1468 | frr_help_exit(1); | |
a6810074 | 1469 | } |
d62a17ae | 1470 | gs.period = 1000 * period; |
1471 | } break; | |
a6810074 | 1472 | case 'p': |
0a7c7856 | 1473 | watchfrr_di.pid_file = optarg; |
a6810074 DL |
1474 | break; |
1475 | case 'r': | |
a6810074 DL |
1476 | if (!valid_command(optarg)) { |
1477 | fprintf(stderr, | |
1478 | "Invalid restart command, must contain '%%s': %s\n", | |
1479 | optarg); | |
4f04a76b | 1480 | frr_help_exit(1); |
a6810074 DL |
1481 | } |
1482 | gs.restart_command = optarg; | |
a6810074 DL |
1483 | break; |
1484 | case 's': | |
1485 | if (!valid_command(optarg)) { | |
1486 | fprintf(stderr, | |
1487 | "Invalid start command, must contain '%%s': %s\n", | |
1488 | optarg); | |
4f04a76b | 1489 | frr_help_exit(1); |
a6810074 DL |
1490 | } |
1491 | gs.start_command = optarg; | |
1492 | break; | |
1493 | case 'S': | |
1494 | gs.vtydir = optarg; | |
1495 | break; | |
d62a17ae | 1496 | case 't': { |
1497 | char garbage[3]; | |
1498 | if ((sscanf(optarg, "%ld%1s", &gs.timeout, garbage) | |
1499 | != 1) | |
1500 | || (gs.timeout < 1)) { | |
1501 | fprintf(stderr, | |
1502 | "Invalid timeout argument: %s\n", | |
1503 | optarg); | |
1504 | frr_help_exit(1); | |
a6810074 | 1505 | } |
d62a17ae | 1506 | } break; |
1507 | case 'T': { | |
1508 | char garbage[3]; | |
1509 | if ((sscanf(optarg, "%ld%1s", &gs.restart_timeout, | |
1510 | garbage) | |
1511 | != 1) | |
1512 | || (gs.restart_timeout < 1)) { | |
1513 | fprintf(stderr, | |
1514 | "Invalid restart timeout argument: %s\n", | |
1515 | optarg); | |
1516 | frr_help_exit(1); | |
a6810074 | 1517 | } |
d62a17ae | 1518 | } break; |
a6810074 DL |
1519 | default: |
1520 | fputs("Invalid option.\n", stderr); | |
4f04a76b | 1521 | frr_help_exit(1); |
a6810074 | 1522 | } |
8b886ca7 | 1523 | } |
a6810074 | 1524 | |
71e7975a DL |
1525 | if (watch_only |
1526 | && (gs.start_command || gs.stop_command || gs.restart_command)) { | |
d87ae5cc | 1527 | fputs("Options -r/-s/-k are not used when --dry is active.\n", |
a6810074 | 1528 | stderr); |
8b886ca7 | 1529 | } |
f168b713 DL |
1530 | if (!watch_only |
1531 | && (!gs.restart_command || !gs.start_command || !gs.stop_command)) { | |
1532 | fprintf(stderr, | |
1533 | "Options -s (start), -k (kill), and -r (restart) are required.\n"); | |
1534 | frr_help_exit(1); | |
8b886ca7 | 1535 | } |
8b886ca7 | 1536 | |
a6810074 DL |
1537 | if (blankstr) { |
1538 | if (gs.restart_command) | |
1539 | gs.restart_command = | |
d62a17ae | 1540 | translate_blanks(gs.restart_command, blankstr); |
a6810074 DL |
1541 | if (gs.start_command) |
1542 | gs.start_command = | |
d62a17ae | 1543 | translate_blanks(gs.start_command, blankstr); |
a6810074 DL |
1544 | if (gs.stop_command) |
1545 | gs.stop_command = | |
d62a17ae | 1546 | translate_blanks(gs.stop_command, blankstr); |
065de903 | 1547 | } |
8b886ca7 | 1548 | |
a6810074 | 1549 | gs.restart.interval = gs.min_restart_interval; |
8b886ca7 | 1550 | |
33606a15 DL |
1551 | /* env variable for the processes that we start */ |
1552 | if (watchfrr_di.pathspace) | |
1553 | setenv("FRR_PATHSPACE", watchfrr_di.pathspace, 1); | |
1554 | else | |
1555 | unsetenv("FRR_PATHSPACE"); | |
1556 | ||
a91f5417 DS |
1557 | /* |
1558 | * when watchfrr_di.pathspace is read, if it is not specified | |
1559 | * pathspace is NULL as expected | |
1560 | */ | |
1561 | pathspace = watchfrr_di.pathspace; | |
1562 | ||
33606a15 DL |
1563 | if (netns_en && !netns) |
1564 | netns = watchfrr_di.pathspace; | |
a91f5417 | 1565 | |
33606a15 DL |
1566 | if (netns_en && netns && netns[0]) |
1567 | netns_setup(netns); | |
1568 | ||
4f04a76b | 1569 | master = frr_init(); |
b647dc2a | 1570 | watchfrr_error_init(); |
0a7c7856 | 1571 | watchfrr_init(argc, argv); |
2ab760f0 | 1572 | cmd_init_config_callbacks(watchfrr_start_config, watchfrr_end_config); |
0a7c7856 DL |
1573 | watchfrr_vty_init(); |
1574 | ||
1575 | frr_config_fork(); | |
4f04a76b | 1576 | |
0a7c7856 | 1577 | if (watchfrr_di.daemon_mode) |
0bdeb5e5 | 1578 | zlog_syslog_set_prio_min(MIN(gs.loglevel, LOG_DEBUG)); |
0a7c7856 | 1579 | else |
0bdeb5e5 | 1580 | zlog_aux_init(NULL, MIN(gs.loglevel, LOG_DEBUG)); |
8b886ca7 | 1581 | |
0a7c7856 | 1582 | frr_run(master); |
8b886ca7 | 1583 | |
a6810074 DL |
1584 | systemd_send_stopping(); |
1585 | /* Not reached. */ | |
1586 | return 0; | |
8b886ca7 | 1587 | } |