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