]>
Commit | Line | Data |
---|---|---|
8b886ca7 | 1 | /* |
896014f4 DL |
2 | * Monitor status of frr daemons and restart if necessary. |
3 | * | |
4 | * Copyright (C) 2004 Andrew J. Schorr | |
5 | * | |
6 | * This program is free software; you can redistribute it and/or modify | |
7 | * it under the terms of the GNU General Public License as published by | |
8 | * the Free Software Foundation; either version 2 of the License, or | |
9 | * (at your option) any later version. | |
10 | * | |
11 | * This program is distributed in the hope that it will be useful, | |
12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
14 | * GNU General Public License for more details. | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License along | |
17 | * with this program; see the file COPYING; if not, write to the Free Software | |
18 | * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
8b886ca7 | 19 | */ |
20 | ||
a365534f | 21 | #include <zebra.h> |
8b886ca7 | 22 | #include <thread.h> |
23 | #include <log.h> | |
52e66296 | 24 | #include <network.h> |
8b886ca7 | 25 | #include <sigevent.h> |
a365534f | 26 | #include <lib/version.h> |
95c4aff2 | 27 | #include "command.h" |
87f44e2f | 28 | #include "memory_vty.h" |
4f04a76b | 29 | #include "libfrr.h" |
b647dc2a | 30 | #include "lib_errors.h" |
95c4aff2 | 31 | |
6f594023 | 32 | #include <getopt.h> |
a365534f | 33 | #include <sys/un.h> |
34 | #include <sys/wait.h> | |
837d16cc | 35 | #include <memory.h> |
651415bd | 36 | #include <systemd.h> |
8b886ca7 | 37 | |
9473e340 | 38 | #include "watchfrr.h" |
b647dc2a | 39 | #include "watchfrr_errors.h" |
95c4aff2 | 40 | |
8b886ca7 | 41 | #ifndef MIN |
42 | #define MIN(X,Y) (((X) <= (Y)) ? (X) : (Y)) | |
43 | #endif | |
44 | ||
45 | /* Macros to help randomize timers. */ | |
46 | #define JITTER(X) ((random() % ((X)+1))-((X)/2)) | |
47 | #define FUZZY(X) ((X)+JITTER((X)/20)) | |
48 | ||
49 | #define DEFAULT_PERIOD 5 | |
0a64aff6 | 50 | #define DEFAULT_TIMEOUT 90 |
8b886ca7 | 51 | #define DEFAULT_RESTART_TIMEOUT 20 |
52 | #define DEFAULT_LOGLEVEL LOG_INFO | |
53 | #define DEFAULT_MIN_RESTART 60 | |
54 | #define DEFAULT_MAX_RESTART 600 | |
8b886ca7 | 55 | |
56 | #define PING_TOKEN "PING" | |
57 | ||
0a7c7856 DL |
58 | DEFINE_MGROUP(WATCHFRR, "watchfrr") |
59 | DEFINE_MTYPE_STATIC(WATCHFRR, WATCHFRR_DAEMON, "watchfrr daemon entry") | |
60 | ||
55c72803 | 61 | /* Needs to be global, referenced somewhere inside libfrr. */ |
8b886ca7 | 62 | struct thread_master *master; |
63 | ||
f168b713 | 64 | static bool watch_only = false; |
8b886ca7 | 65 | |
a6810074 DL |
66 | typedef enum { |
67 | PHASE_NONE = 0, | |
c0e5cb52 | 68 | PHASE_INIT, |
a6810074 DL |
69 | PHASE_STOPS_PENDING, |
70 | PHASE_WAITING_DOWN, | |
71 | PHASE_ZEBRA_RESTART_PENDING, | |
72 | PHASE_WAITING_ZEBRA_UP | |
8b886ca7 | 73 | } restart_phase_t; |
74 | ||
a6810074 | 75 | static const char *phase_str[] = { |
af568444 | 76 | "Idle", |
c0e5cb52 | 77 | "Startup", |
a6810074 DL |
78 | "Stop jobs running", |
79 | "Waiting for other daemons to come down", | |
80 | "Zebra restart job running", | |
81 | "Waiting for zebra to come up", | |
82 | "Start jobs running", | |
8b886ca7 | 83 | }; |
84 | ||
85 | #define PHASE_TIMEOUT (3*gs.restart_timeout) | |
86 | ||
a6810074 DL |
87 | struct restart_info { |
88 | const char *name; | |
89 | const char *what; | |
90 | pid_t pid; | |
91 | struct timeval time; | |
92 | long interval; | |
93 | struct thread *t_kill; | |
94 | int kills; | |
098e240f | 95 | }; |
96 | ||
a6810074 | 97 | static struct global_state { |
a6810074 DL |
98 | restart_phase_t phase; |
99 | struct thread *t_phase_hanging; | |
100 | const char *vtydir; | |
101 | long period; | |
102 | long timeout; | |
103 | long restart_timeout; | |
104 | long min_restart_interval; | |
105 | long max_restart_interval; | |
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, | |
d62a17ae | 125 | }; |
a6810074 DL |
126 | |
127 | typedef enum { | |
128 | DAEMON_INIT, | |
129 | DAEMON_DOWN, | |
130 | DAEMON_CONNECTING, | |
131 | DAEMON_UP, | |
132 | DAEMON_UNRESPONSIVE | |
8b886ca7 | 133 | } daemon_state_t; |
134 | ||
d62a17ae | 135 | #define IS_UP(DMN) \ |
136 | (((DMN)->state == DAEMON_UP) || ((DMN)->state == DAEMON_UNRESPONSIVE)) | |
8b886ca7 | 137 | |
a6810074 | 138 | static const char *state_str[] = { |
d62a17ae | 139 | "Init", "Down", "Connecting", "Up", "Unresponsive", |
8b886ca7 | 140 | }; |
141 | ||
142 | struct daemon { | |
a6810074 DL |
143 | const char *name; |
144 | daemon_state_t state; | |
145 | int fd; | |
146 | struct timeval echo_sent; | |
d7c0a89a | 147 | unsigned int connect_tries; |
a6810074 DL |
148 | struct thread *t_wakeup; |
149 | struct thread *t_read; | |
150 | struct thread *t_write; | |
151 | struct daemon *next; | |
152 | struct restart_info restart; | |
8b886ca7 | 153 | }; |
154 | ||
9272302b DL |
155 | #define OPTION_MINRESTART 2000 |
156 | #define OPTION_MAXRESTART 2001 | |
f168b713 | 157 | #define OPTION_DRY 2002 |
9272302b | 158 | |
a6810074 DL |
159 | static const struct option longopts[] = { |
160 | {"daemon", no_argument, NULL, 'd'}, | |
161 | {"statedir", required_argument, NULL, 'S'}, | |
a6810074 DL |
162 | {"loglevel", required_argument, NULL, 'l'}, |
163 | {"interval", required_argument, NULL, 'i'}, | |
164 | {"timeout", required_argument, NULL, 't'}, | |
165 | {"restart-timeout", required_argument, NULL, 'T'}, | |
166 | {"restart", required_argument, NULL, 'r'}, | |
167 | {"start-command", required_argument, NULL, 's'}, | |
168 | {"kill-command", required_argument, NULL, 'k'}, | |
f168b713 | 169 | {"dry", no_argument, NULL, OPTION_DRY}, |
d62a17ae | 170 | {"min-restart-interval", required_argument, NULL, OPTION_MINRESTART}, |
171 | {"max-restart-interval", required_argument, NULL, OPTION_MAXRESTART}, | |
a6810074 DL |
172 | {"pid-file", required_argument, NULL, 'p'}, |
173 | {"blank-string", required_argument, NULL, 'b'}, | |
174 | {"help", no_argument, NULL, 'h'}, | |
175 | {"version", no_argument, NULL, 'v'}, | |
d62a17ae | 176 | {NULL, 0, NULL, 0}}; |
8b886ca7 | 177 | |
178 | static int try_connect(struct daemon *dmn); | |
179 | static int wakeup_send_echo(struct thread *t_wakeup); | |
180 | static void try_restart(struct daemon *dmn); | |
181 | static void phase_check(void); | |
75f8b0e4 | 182 | static void restart_done(struct daemon *dmn); |
8b886ca7 | 183 | |
4f04a76b DL |
184 | static const char *progname; |
185 | static void printhelp(FILE *target) | |
8b886ca7 | 186 | { |
d62a17ae | 187 | fprintf(target, |
188 | "Usage : %s [OPTION...] <daemon name> ...\n\n\ | |
9473e340 | 189 | Watchdog program to monitor status of frr daemons and try to restart\n\ |
8b886ca7 | 190 | them if they are down or unresponsive. It determines whether a daemon is\n\ |
191 | up based on whether it can connect to the daemon's vty unix stream socket.\n\ | |
192 | It then repeatedly sends echo commands over that socket to determine whether\n\ | |
193 | the daemon is responsive. If the daemon crashes, we will receive an EOF\n\ | |
194 | on the socket connection and know immediately that the daemon is down.\n\n\ | |
195 | The daemons to be monitored should be listed on the command line.\n\n\ | |
8b886ca7 | 196 | In order to avoid attempting to restart the daemons in a fast loop,\n\ |
197 | the -m and -M options allow you to control the minimum delay between\n\ | |
198 | restart commands. The minimum restart delay is recalculated each time\n\ | |
199 | a restart is attempted: if the time since the last restart attempt exceeds\n\ | |
200 | twice the -M value, then the restart delay is set to the -m value.\n\ | |
d62a17ae | 201 | Otherwise, the interval is doubled (but capped at the -M value).\n\n", |
f168b713 | 202 | progname); |
e757c940 | 203 | |
d62a17ae | 204 | fprintf(target, |
205 | "Options:\n\ | |
8b886ca7 | 206 | -d, --daemon Run in daemon mode. In this mode, error messages are sent\n\ |
207 | to syslog instead of stdout.\n\ | |
208 | -S, --statedir Set the vty socket directory (default is %s)\n\ | |
8b886ca7 | 209 | -l, --loglevel Set the logging level (default is %d).\n\ |
210 | The value should range from %d (LOG_EMERG) to %d (LOG_DEBUG),\n\ | |
211 | but it can be set higher than %d if extra-verbose debugging\n\ | |
212 | messages are desired.\n\ | |
9272302b | 213 | --min-restart-interval\n\ |
8b886ca7 | 214 | Set the minimum seconds to wait between invocations of daemon\n\ |
215 | restart commands (default is %d).\n\ | |
9272302b | 216 | --max-restart-interval\n\ |
8b886ca7 | 217 | Set the maximum seconds to wait between invocations of daemon\n\ |
218 | restart commands (default is %d).\n\ | |
219 | -i, --interval Set the status polling interval in seconds (default is %d)\n\ | |
220 | -t, --timeout Set the unresponsiveness timeout in seconds (default is %d)\n\ | |
221 | -T, --restart-timeout\n\ | |
222 | Set the restart (kill) timeout in seconds (default is %d).\n\ | |
223 | If any background jobs are still running after this much\n\ | |
224 | time has elapsed, they will be killed.\n\ | |
225 | -r, --restart Supply a Bourne shell command to use to restart a single\n\ | |
226 | daemon. The command string should include '%%s' where the\n\ | |
227 | name of the daemon should be substituted.\n\ | |
8b886ca7 | 228 | -s, --start-command\n\ |
229 | Supply a Bourne shell to command to use to start a single\n\ | |
230 | daemon. The command string should include '%%s' where the\n\ | |
231 | name of the daemon should be substituted.\n\ | |
232 | -k, --kill-command\n\ | |
233 | Supply a Bourne shell to command to use to stop a single\n\ | |
234 | daemon. The command string should include '%%s' where the\n\ | |
235 | name of the daemon should be substituted.\n\ | |
f168b713 | 236 | --dry Do not start or restart anything, just log.\n\ |
8b886ca7 | 237 | -p, --pid-file Set process identifier file name\n\ |
0a7c7856 | 238 | (default is %s/watchfrr.pid).\n\ |
c8b40f86 | 239 | -b, --blank-string\n\ |
240 | When the supplied argument string is found in any of the\n\ | |
f168b713 | 241 | various shell command arguments (-r, -s, or -k), replace\n\ |
c8b40f86 | 242 | it with a space. This is an ugly hack to circumvent problems\n\ |
243 | passing command-line arguments with embedded spaces.\n\ | |
8b886ca7 | 244 | -v, --version Print program version\n\ |
d62a17ae | 245 | -h, --help Display this help and exit\n", |
64a249ad | 246 | frr_vtydir, DEFAULT_LOGLEVEL, LOG_EMERG, LOG_DEBUG, LOG_DEBUG, |
d62a17ae | 247 | DEFAULT_MIN_RESTART, DEFAULT_MAX_RESTART, DEFAULT_PERIOD, |
0a7c7856 | 248 | DEFAULT_TIMEOUT, DEFAULT_RESTART_TIMEOUT, frr_vtydir); |
8b886ca7 | 249 | } |
250 | ||
a6810074 | 251 | static pid_t run_background(char *shell_cmd) |
8b886ca7 | 252 | { |
a6810074 DL |
253 | pid_t child; |
254 | ||
255 | switch (child = fork()) { | |
256 | case -1: | |
450971aa | 257 | flog_err_sys(EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
258 | "fork failed, cannot run command [%s]: %s", |
259 | shell_cmd, safe_strerror(errno)); | |
a6810074 DL |
260 | return -1; |
261 | case 0: | |
262 | /* Child process. */ | |
d62a17ae | 263 | /* Use separate process group so child processes can be killed |
264 | * easily. */ | |
a6810074 DL |
265 | if (setpgid(0, 0) < 0) |
266 | zlog_warn("warning: setpgid(0,0) failed: %s", | |
267 | safe_strerror(errno)); | |
268 | { | |
269 | char shell[] = "sh"; | |
270 | char dashc[] = "-c"; | |
d62a17ae | 271 | char *const argv[4] = {shell, dashc, shell_cmd, NULL}; |
a6810074 | 272 | execv("/bin/sh", argv); |
450971aa | 273 | flog_err_sys(EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
274 | "execv(/bin/sh -c '%s') failed: %s", |
275 | shell_cmd, safe_strerror(errno)); | |
a6810074 DL |
276 | _exit(127); |
277 | } | |
278 | default: | |
279 | /* Parent process: we will reap the child later. */ | |
450971aa | 280 | flog_err_sys(EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
281 | "Forked background command [pid %d]: %s", |
282 | (int)child, shell_cmd); | |
a6810074 DL |
283 | return child; |
284 | } | |
8b886ca7 | 285 | } |
286 | ||
a6810074 DL |
287 | static struct timeval *time_elapsed(struct timeval *result, |
288 | const struct timeval *start_time) | |
8b886ca7 | 289 | { |
a6810074 DL |
290 | gettimeofday(result, NULL); |
291 | result->tv_sec -= start_time->tv_sec; | |
292 | result->tv_usec -= start_time->tv_usec; | |
293 | while (result->tv_usec < 0) { | |
294 | result->tv_usec += 1000000L; | |
295 | result->tv_sec--; | |
296 | } | |
297 | return result; | |
8b886ca7 | 298 | } |
299 | ||
a6810074 | 300 | static int restart_kill(struct thread *t_kill) |
8b886ca7 | 301 | { |
a6810074 DL |
302 | struct restart_info *restart = THREAD_ARG(t_kill); |
303 | struct timeval delay; | |
304 | ||
305 | time_elapsed(&delay, &restart->time); | |
d62a17ae | 306 | zlog_warn( |
307 | "Warning: %s %s child process %d still running after " | |
308 | "%ld seconds, sending signal %d", | |
309 | restart->what, restart->name, (int)restart->pid, | |
310 | (long)delay.tv_sec, (restart->kills ? SIGKILL : SIGTERM)); | |
a6810074 DL |
311 | kill(-restart->pid, (restart->kills ? SIGKILL : SIGTERM)); |
312 | restart->kills++; | |
66e78ae6 QY |
313 | restart->t_kill = NULL; |
314 | thread_add_timer(master, restart_kill, restart, gs.restart_timeout, | |
315 | &restart->t_kill); | |
a6810074 | 316 | return 0; |
8b886ca7 | 317 | } |
318 | ||
a6810074 | 319 | static struct restart_info *find_child(pid_t child) |
8b886ca7 | 320 | { |
f168b713 | 321 | struct daemon *dmn; |
7c265f7d CF |
322 | if (gs.restart.pid == child) |
323 | return &gs.restart; | |
324 | ||
f168b713 DL |
325 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { |
326 | if (dmn->restart.pid == child) | |
327 | return &dmn->restart; | |
a6810074 DL |
328 | } |
329 | return NULL; | |
8b886ca7 | 330 | } |
331 | ||
a6810074 | 332 | static void sigchild(void) |
8b886ca7 | 333 | { |
a6810074 DL |
334 | pid_t child; |
335 | int status; | |
336 | const char *name; | |
337 | const char *what; | |
338 | struct restart_info *restart; | |
75f8b0e4 | 339 | struct daemon *dmn; |
a6810074 DL |
340 | |
341 | switch (child = waitpid(-1, &status, WNOHANG)) { | |
342 | case -1: | |
450971aa | 343 | flog_err_sys(EC_LIB_SYSTEM_CALL, "waitpid failed: %s", |
09c866e3 | 344 | safe_strerror(errno)); |
a6810074 DL |
345 | return; |
346 | case 0: | |
347 | zlog_warn("SIGCHLD received, but waitpid did not reap a child"); | |
348 | return; | |
349 | } | |
350 | ||
351 | if (child == integrated_write_pid) { | |
352 | integrated_write_sigchld(status); | |
353 | return; | |
354 | } | |
355 | ||
356 | if ((restart = find_child(child)) != NULL) { | |
357 | name = restart->name; | |
358 | what = restart->what; | |
359 | restart->pid = 0; | |
360 | gs.numpids--; | |
361 | thread_cancel(restart->t_kill); | |
362 | restart->t_kill = NULL; | |
d62a17ae | 363 | /* Update restart time to reflect the time the command |
364 | * completed. */ | |
a6810074 DL |
365 | gettimeofday(&restart->time, NULL); |
366 | } else { | |
09c866e3 | 367 | flog_err_sys( |
450971aa | 368 | EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
369 | "waitpid returned status for an unknown child process %d", |
370 | (int)child); | |
a6810074 DL |
371 | name = "(unknown)"; |
372 | what = "background"; | |
373 | } | |
374 | if (WIFSTOPPED(status)) | |
d62a17ae | 375 | zlog_warn("warning: %s %s process %d is stopped", what, name, |
376 | (int)child); | |
a6810074 | 377 | else if (WIFSIGNALED(status)) |
d62a17ae | 378 | zlog_warn("%s %s process %d terminated due to signal %d", what, |
379 | name, (int)child, WTERMSIG(status)); | |
a6810074 DL |
380 | else if (WIFEXITED(status)) { |
381 | if (WEXITSTATUS(status) != 0) | |
d62a17ae | 382 | zlog_warn( |
383 | "%s %s process %d exited with non-zero status %d", | |
384 | what, name, (int)child, WEXITSTATUS(status)); | |
75f8b0e4 | 385 | else { |
a6810074 DL |
386 | zlog_debug("%s %s process %d exited normally", what, |
387 | name, (int)child); | |
75f8b0e4 DL |
388 | |
389 | if (restart && restart != &gs.restart) { | |
390 | dmn = container_of(restart, struct daemon, | |
391 | restart); | |
392 | restart_done(dmn); | |
393 | } else if (restart) | |
394 | for (dmn = gs.daemons; dmn; dmn = dmn->next) | |
395 | restart_done(dmn); | |
396 | } | |
a6810074 | 397 | } else |
09c866e3 | 398 | flog_err_sys( |
450971aa | 399 | EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
400 | "cannot interpret %s %s process %d wait status 0x%x", |
401 | what, name, (int)child, status); | |
a6810074 | 402 | phase_check(); |
8b886ca7 | 403 | } |
404 | ||
d62a17ae | 405 | static int run_job(struct restart_info *restart, const char *cmdtype, |
406 | const char *command, int force, int update_interval) | |
8b886ca7 | 407 | { |
a6810074 DL |
408 | struct timeval delay; |
409 | ||
410 | if (gs.loglevel > LOG_DEBUG + 1) | |
411 | zlog_debug("attempting to %s %s", cmdtype, restart->name); | |
412 | ||
413 | if (restart->pid) { | |
414 | if (gs.loglevel > LOG_DEBUG + 1) | |
d62a17ae | 415 | zlog_debug( |
416 | "cannot %s %s, previous pid %d still running", | |
417 | cmdtype, restart->name, (int)restart->pid); | |
a6810074 DL |
418 | return -1; |
419 | } | |
420 | ||
d62a17ae | 421 | /* Note: time_elapsed test must come before the force test, since we |
422 | need | |
a6810074 DL |
423 | to make sure that delay is initialized for use below in updating the |
424 | restart interval. */ | |
425 | if ((time_elapsed(&delay, &restart->time)->tv_sec < restart->interval) | |
426 | && !force) { | |
427 | if (gs.loglevel > LOG_DEBUG + 1) | |
d62a17ae | 428 | zlog_debug( |
429 | "postponing %s %s: " | |
430 | "elapsed time %ld < retry interval %ld", | |
431 | cmdtype, restart->name, (long)delay.tv_sec, | |
432 | restart->interval); | |
a6810074 DL |
433 | return -1; |
434 | } | |
435 | ||
436 | gettimeofday(&restart->time, NULL); | |
437 | restart->kills = 0; | |
438 | { | |
439 | char cmd[strlen(command) + strlen(restart->name) + 1]; | |
440 | snprintf(cmd, sizeof(cmd), command, restart->name); | |
441 | if ((restart->pid = run_background(cmd)) > 0) { | |
66e78ae6 | 442 | restart->t_kill = NULL; |
d62a17ae | 443 | thread_add_timer(master, restart_kill, restart, |
444 | gs.restart_timeout, &restart->t_kill); | |
a6810074 DL |
445 | restart->what = cmdtype; |
446 | gs.numpids++; | |
447 | } else | |
448 | restart->pid = 0; | |
449 | } | |
450 | ||
451 | /* Calculate the new restart interval. */ | |
452 | if (update_interval) { | |
453 | if (delay.tv_sec > 2 * gs.max_restart_interval) | |
454 | restart->interval = gs.min_restart_interval; | |
455 | else if ((restart->interval *= 2) > gs.max_restart_interval) | |
456 | restart->interval = gs.max_restart_interval; | |
457 | if (gs.loglevel > LOG_DEBUG + 1) | |
458 | zlog_debug("restart %s interval is now %ld", | |
459 | restart->name, restart->interval); | |
460 | } | |
461 | return restart->pid; | |
8b886ca7 | 462 | } |
463 | ||
d62a17ae | 464 | #define SET_READ_HANDLER(DMN) \ |
465 | do { \ | |
466 | (DMN)->t_read = NULL; \ | |
467 | thread_add_read(master, handle_read, (DMN), (DMN)->fd, \ | |
468 | &(DMN)->t_read); \ | |
469 | } while (0); | |
470 | ||
471 | #define SET_WAKEUP_DOWN(DMN) \ | |
472 | do { \ | |
473 | (DMN)->t_wakeup = NULL; \ | |
474 | thread_add_timer_msec(master, wakeup_down, (DMN), \ | |
475 | FUZZY(gs.period), &(DMN)->t_wakeup); \ | |
476 | } while (0); | |
477 | ||
478 | #define SET_WAKEUP_UNRESPONSIVE(DMN) \ | |
479 | do { \ | |
480 | (DMN)->t_wakeup = NULL; \ | |
481 | thread_add_timer_msec(master, wakeup_unresponsive, (DMN), \ | |
482 | FUZZY(gs.period), &(DMN)->t_wakeup); \ | |
483 | } while (0); | |
484 | ||
485 | #define SET_WAKEUP_ECHO(DMN) \ | |
486 | do { \ | |
487 | (DMN)->t_wakeup = NULL; \ | |
488 | thread_add_timer_msec(master, wakeup_send_echo, (DMN), \ | |
489 | FUZZY(gs.period), &(DMN)->t_wakeup); \ | |
490 | } while (0); | |
8b886ca7 | 491 | |
a6810074 | 492 | static int wakeup_down(struct thread *t_wakeup) |
8b886ca7 | 493 | { |
a6810074 DL |
494 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
495 | ||
496 | dmn->t_wakeup = NULL; | |
497 | if (try_connect(dmn) < 0) | |
498 | SET_WAKEUP_DOWN(dmn); | |
499 | if ((dmn->connect_tries > 1) && (dmn->state != DAEMON_UP)) | |
500 | try_restart(dmn); | |
501 | return 0; | |
8b886ca7 | 502 | } |
503 | ||
a6810074 | 504 | static int wakeup_init(struct thread *t_wakeup) |
8b886ca7 | 505 | { |
a6810074 DL |
506 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
507 | ||
508 | dmn->t_wakeup = NULL; | |
509 | if (try_connect(dmn) < 0) { | |
f74ae2bb | 510 | flog_err(EC_WATCHFRR_CONNECTION, |
1c50c1c0 QY |
511 | "%s state -> down : initial connection attempt failed", |
512 | dmn->name); | |
a6810074 DL |
513 | dmn->state = DAEMON_DOWN; |
514 | } | |
c0e5cb52 | 515 | phase_check(); |
a6810074 | 516 | return 0; |
8b886ca7 | 517 | } |
518 | ||
75f8b0e4 DL |
519 | static void restart_done(struct daemon *dmn) |
520 | { | |
521 | if (dmn->state != DAEMON_DOWN) { | |
522 | zlog_warn("wtf?"); | |
523 | return; | |
524 | } | |
525 | if (dmn->t_wakeup) | |
526 | THREAD_OFF(dmn->t_wakeup); | |
527 | if (try_connect(dmn) < 0) | |
528 | SET_WAKEUP_DOWN(dmn); | |
529 | } | |
530 | ||
a6810074 | 531 | static void daemon_down(struct daemon *dmn, const char *why) |
8b886ca7 | 532 | { |
a6810074 | 533 | if (IS_UP(dmn) || (dmn->state == DAEMON_INIT)) |
1c50c1c0 QY |
534 | flog_err(EC_WATCHFRR_CONNECTION, "%s state -> down : %s", |
535 | dmn->name, why); | |
a6810074 DL |
536 | else if (gs.loglevel > LOG_DEBUG) |
537 | zlog_debug("%s still down : %s", dmn->name, why); | |
538 | if (IS_UP(dmn)) | |
539 | gs.numdown++; | |
540 | dmn->state = DAEMON_DOWN; | |
541 | if (dmn->fd >= 0) { | |
542 | close(dmn->fd); | |
543 | dmn->fd = -1; | |
544 | } | |
545 | THREAD_OFF(dmn->t_read); | |
546 | THREAD_OFF(dmn->t_write); | |
547 | THREAD_OFF(dmn->t_wakeup); | |
548 | if (try_connect(dmn) < 0) | |
549 | SET_WAKEUP_DOWN(dmn); | |
550 | phase_check(); | |
8b886ca7 | 551 | } |
552 | ||
a6810074 | 553 | static int handle_read(struct thread *t_read) |
8b886ca7 | 554 | { |
a6810074 DL |
555 | struct daemon *dmn = THREAD_ARG(t_read); |
556 | static const char resp[sizeof(PING_TOKEN) + 4] = PING_TOKEN "\n"; | |
557 | char buf[sizeof(resp) + 100]; | |
558 | ssize_t rc; | |
559 | struct timeval delay; | |
560 | ||
561 | dmn->t_read = NULL; | |
562 | if ((rc = read(dmn->fd, buf, sizeof(buf))) < 0) { | |
563 | char why[100]; | |
564 | ||
565 | if (ERRNO_IO_RETRY(errno)) { | |
566 | /* Pretend it never happened. */ | |
567 | SET_READ_HANDLER(dmn); | |
568 | return 0; | |
569 | } | |
570 | snprintf(why, sizeof(why), "unexpected read error: %s", | |
571 | safe_strerror(errno)); | |
572 | daemon_down(dmn, why); | |
573 | return 0; | |
8b886ca7 | 574 | } |
a6810074 DL |
575 | if (rc == 0) { |
576 | daemon_down(dmn, "read returned EOF"); | |
577 | return 0; | |
578 | } | |
579 | if (!dmn->echo_sent.tv_sec) { | |
580 | char why[sizeof(buf) + 100]; | |
581 | snprintf(why, sizeof(why), | |
582 | "unexpected read returns %d bytes: %.*s", (int)rc, | |
583 | (int)rc, buf); | |
584 | daemon_down(dmn, why); | |
585 | return 0; | |
8b886ca7 | 586 | } |
a6810074 DL |
587 | |
588 | /* We are expecting an echo response: is there any chance that the | |
589 | response would not be returned entirely in the first read? That | |
590 | seems inconceivable... */ | |
591 | if ((rc != sizeof(resp)) || memcmp(buf, resp, sizeof(resp))) { | |
592 | char why[100 + sizeof(buf)]; | |
593 | snprintf(why, sizeof(why), | |
594 | "read returned bad echo response of %d bytes " | |
d62a17ae | 595 | "(expecting %u): %.*s", |
d7c0a89a | 596 | (int)rc, (unsigned int)sizeof(resp), (int)rc, buf); |
a6810074 DL |
597 | daemon_down(dmn, why); |
598 | return 0; | |
599 | } | |
600 | ||
601 | time_elapsed(&delay, &dmn->echo_sent); | |
602 | dmn->echo_sent.tv_sec = 0; | |
603 | if (dmn->state == DAEMON_UNRESPONSIVE) { | |
604 | if (delay.tv_sec < gs.timeout) { | |
605 | dmn->state = DAEMON_UP; | |
d62a17ae | 606 | zlog_warn( |
607 | "%s state -> up : echo response received after %ld.%06ld " | |
608 | "seconds", | |
609 | dmn->name, (long)delay.tv_sec, | |
610 | (long)delay.tv_usec); | |
a6810074 | 611 | } else |
d62a17ae | 612 | zlog_warn( |
613 | "%s: slow echo response finally received after %ld.%06ld " | |
614 | "seconds", | |
615 | dmn->name, (long)delay.tv_sec, | |
616 | (long)delay.tv_usec); | |
a6810074 DL |
617 | } else if (gs.loglevel > LOG_DEBUG + 1) |
618 | zlog_debug("%s: echo response received after %ld.%06ld seconds", | |
619 | dmn->name, (long)delay.tv_sec, (long)delay.tv_usec); | |
620 | ||
621 | SET_READ_HANDLER(dmn); | |
622 | if (dmn->t_wakeup) | |
623 | thread_cancel(dmn->t_wakeup); | |
624 | SET_WAKEUP_ECHO(dmn); | |
625 | ||
626 | return 0; | |
8b886ca7 | 627 | } |
628 | ||
207e0d7a DS |
629 | /* |
630 | * Wait till we notice that all daemons are ready before | |
631 | * we send we are ready to systemd | |
632 | */ | |
a6810074 | 633 | static void daemon_send_ready(void) |
207e0d7a | 634 | { |
a6810074 DL |
635 | static int sent = 0; |
636 | if (!sent && gs.numdown == 0) { | |
a6810074 | 637 | FILE *fp; |
207e0d7a | 638 | |
0a7c7856 DL |
639 | zlog_notice("all daemons up, doing startup-complete notify"); |
640 | frr_detach(); | |
641 | ||
a6810074 | 642 | fp = fopen(DAEMON_VTY_DIR "/watchfrr.started", "w"); |
f5ba21fc DS |
643 | if (fp) |
644 | fclose(fp); | |
60bd2534 | 645 | #if defined HAVE_SYSTEMD |
a6810074 | 646 | systemd_send_started(master, 0); |
60bd2534 | 647 | #endif |
a6810074 DL |
648 | sent = 1; |
649 | } | |
207e0d7a DS |
650 | } |
651 | ||
a6810074 | 652 | static void daemon_up(struct daemon *dmn, const char *why) |
8b886ca7 | 653 | { |
a6810074 DL |
654 | dmn->state = DAEMON_UP; |
655 | gs.numdown--; | |
656 | dmn->connect_tries = 0; | |
657 | zlog_notice("%s state -> up : %s", dmn->name, why); | |
658 | daemon_send_ready(); | |
a8cbb8b3 | 659 | SET_WAKEUP_ECHO(dmn); |
a6810074 | 660 | phase_check(); |
8b886ca7 | 661 | } |
662 | ||
a6810074 | 663 | static int check_connect(struct thread *t_write) |
8b886ca7 | 664 | { |
a6810074 DL |
665 | struct daemon *dmn = THREAD_ARG(t_write); |
666 | int sockerr; | |
667 | socklen_t reslen = sizeof(sockerr); | |
668 | ||
669 | dmn->t_write = NULL; | |
670 | if (getsockopt(dmn->fd, SOL_SOCKET, SO_ERROR, (char *)&sockerr, &reslen) | |
671 | < 0) { | |
672 | zlog_warn("%s: check_connect: getsockopt failed: %s", dmn->name, | |
673 | safe_strerror(errno)); | |
674 | daemon_down(dmn, | |
675 | "getsockopt failed checking connection success"); | |
676 | return 0; | |
677 | } | |
678 | if ((reslen == sizeof(sockerr)) && sockerr) { | |
679 | char why[100]; | |
d62a17ae | 680 | snprintf( |
681 | why, sizeof(why), | |
682 | "getsockopt reports that connection attempt failed: %s", | |
683 | safe_strerror(sockerr)); | |
a6810074 DL |
684 | daemon_down(dmn, why); |
685 | return 0; | |
686 | } | |
687 | ||
688 | daemon_up(dmn, "delayed connect succeeded"); | |
689 | return 0; | |
8b886ca7 | 690 | } |
691 | ||
a6810074 | 692 | static int wakeup_connect_hanging(struct thread *t_wakeup) |
8b886ca7 | 693 | { |
a6810074 DL |
694 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
695 | char why[100]; | |
696 | ||
697 | dmn->t_wakeup = NULL; | |
698 | snprintf(why, sizeof(why), | |
699 | "connection attempt timed out after %ld seconds", gs.timeout); | |
700 | daemon_down(dmn, why); | |
701 | return 0; | |
8b886ca7 | 702 | } |
703 | ||
704 | /* Making connection to protocol daemon. */ | |
a6810074 | 705 | static int try_connect(struct daemon *dmn) |
8b886ca7 | 706 | { |
a6810074 DL |
707 | int sock; |
708 | struct sockaddr_un addr; | |
709 | socklen_t len; | |
710 | ||
711 | if (gs.loglevel > LOG_DEBUG + 1) | |
712 | zlog_debug("%s: attempting to connect", dmn->name); | |
713 | dmn->connect_tries++; | |
714 | ||
715 | memset(&addr, 0, sizeof(struct sockaddr_un)); | |
716 | addr.sun_family = AF_UNIX; | |
d62a17ae | 717 | snprintf(addr.sun_path, sizeof(addr.sun_path), "%s/%s.vty", gs.vtydir, |
718 | dmn->name); | |
6f0e3f6e | 719 | #ifdef HAVE_STRUCT_SOCKADDR_UN_SUN_LEN |
a6810074 | 720 | len = addr.sun_len = SUN_LEN(&addr); |
8b886ca7 | 721 | #else |
a6810074 | 722 | len = sizeof(addr.sun_family) + strlen(addr.sun_path); |
d62a17ae | 723 | #endif /* HAVE_STRUCT_SOCKADDR_UN_SUN_LEN */ |
a6810074 DL |
724 | |
725 | /* Quick check to see if we might succeed before we go to the trouble | |
726 | of creating a socket. */ | |
727 | if (access(addr.sun_path, W_OK) < 0) { | |
728 | if (errno != ENOENT) | |
450971aa | 729 | flog_err_sys(EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
730 | "%s: access to socket %s denied: %s", |
731 | dmn->name, addr.sun_path, | |
732 | safe_strerror(errno)); | |
a6810074 DL |
733 | return -1; |
734 | } | |
735 | ||
736 | if ((sock = socket(AF_UNIX, SOCK_STREAM, 0)) < 0) { | |
450971aa | 737 | flog_err_sys(EC_LIB_SOCKET, "%s(%s): cannot make socket: %s", |
09c866e3 | 738 | __func__, addr.sun_path, safe_strerror(errno)); |
a6810074 DL |
739 | return -1; |
740 | } | |
741 | ||
742 | if (set_nonblocking(sock) < 0 || set_cloexec(sock) < 0) { | |
450971aa | 743 | flog_err_sys(EC_LIB_SYSTEM_CALL, |
09c866e3 QY |
744 | "%s(%s): set_nonblocking/cloexec(%d) failed", |
745 | __func__, addr.sun_path, sock); | |
a6810074 DL |
746 | close(sock); |
747 | return -1; | |
8b886ca7 | 748 | } |
a6810074 DL |
749 | |
750 | if (connect(sock, (struct sockaddr *)&addr, len) < 0) { | |
751 | if ((errno != EINPROGRESS) && (errno != EWOULDBLOCK)) { | |
752 | if (gs.loglevel > LOG_DEBUG) | |
753 | zlog_debug("%s(%s): connect failed: %s", | |
754 | __func__, addr.sun_path, | |
755 | safe_strerror(errno)); | |
756 | close(sock); | |
757 | return -1; | |
758 | } | |
759 | if (gs.loglevel > LOG_DEBUG) | |
760 | zlog_debug("%s: connection in progress", dmn->name); | |
761 | dmn->state = DAEMON_CONNECTING; | |
762 | dmn->fd = sock; | |
66e78ae6 QY |
763 | dmn->t_write = NULL; |
764 | thread_add_write(master, check_connect, dmn, dmn->fd, | |
d62a17ae | 765 | &dmn->t_write); |
766 | dmn->t_wakeup = NULL; | |
767 | thread_add_timer(master, wakeup_connect_hanging, dmn, | |
768 | gs.timeout, &dmn->t_wakeup); | |
a6810074 DL |
769 | SET_READ_HANDLER(dmn); |
770 | return 0; | |
771 | } | |
772 | ||
773 | dmn->fd = sock; | |
774 | SET_READ_HANDLER(dmn); | |
775 | daemon_up(dmn, "connect succeeded"); | |
776 | return 1; | |
8b886ca7 | 777 | } |
778 | ||
a6810074 | 779 | static int phase_hanging(struct thread *t_hanging) |
8b886ca7 | 780 | { |
a6810074 | 781 | gs.t_phase_hanging = NULL; |
f74ae2bb | 782 | flog_err(EC_WATCHFRR_CONNECTION, |
1c50c1c0 QY |
783 | "Phase [%s] hanging for %ld seconds, aborting phased restart", |
784 | phase_str[gs.phase], PHASE_TIMEOUT); | |
a6810074 DL |
785 | gs.phase = PHASE_NONE; |
786 | return 0; | |
8b886ca7 | 787 | } |
788 | ||
a6810074 | 789 | static void set_phase(restart_phase_t new_phase) |
8b886ca7 | 790 | { |
a6810074 DL |
791 | gs.phase = new_phase; |
792 | if (gs.t_phase_hanging) | |
793 | thread_cancel(gs.t_phase_hanging); | |
66e78ae6 QY |
794 | gs.t_phase_hanging = NULL; |
795 | thread_add_timer(master, phase_hanging, NULL, PHASE_TIMEOUT, | |
796 | &gs.t_phase_hanging); | |
8b886ca7 | 797 | } |
798 | ||
a6810074 | 799 | static void phase_check(void) |
8b886ca7 | 800 | { |
c0e5cb52 DL |
801 | struct daemon *dmn; |
802 | ||
a6810074 DL |
803 | switch (gs.phase) { |
804 | case PHASE_NONE: | |
805 | break; | |
c0e5cb52 DL |
806 | |
807 | case PHASE_INIT: | |
808 | for (dmn = gs.daemons; dmn; dmn = dmn->next) | |
809 | if (dmn->state == DAEMON_INIT) | |
810 | return; | |
811 | ||
812 | /* startup complete, everything out of INIT */ | |
813 | gs.phase = PHASE_NONE; | |
814 | for (dmn = gs.daemons; dmn; dmn = dmn->next) | |
815 | if (dmn->state == DAEMON_DOWN) { | |
816 | SET_WAKEUP_DOWN(dmn); | |
817 | try_restart(dmn); | |
818 | } | |
819 | break; | |
a6810074 DL |
820 | case PHASE_STOPS_PENDING: |
821 | if (gs.numpids) | |
822 | break; | |
d62a17ae | 823 | zlog_info( |
824 | "Phased restart: all routing daemon stop jobs have completed."); | |
a6810074 DL |
825 | set_phase(PHASE_WAITING_DOWN); |
826 | ||
d62a17ae | 827 | /*FALLTHRU*/ |
a6810074 DL |
828 | case PHASE_WAITING_DOWN: |
829 | if (gs.numdown + IS_UP(gs.special) < gs.numdaemons) | |
830 | break; | |
831 | zlog_info("Phased restart: all routing daemons now down."); | |
832 | run_job(&gs.special->restart, "restart", gs.restart_command, 1, | |
833 | 1); | |
834 | set_phase(PHASE_ZEBRA_RESTART_PENDING); | |
835 | ||
d62a17ae | 836 | /*FALLTHRU*/ |
a6810074 DL |
837 | case PHASE_ZEBRA_RESTART_PENDING: |
838 | if (gs.special->restart.pid) | |
839 | break; | |
840 | zlog_info("Phased restart: %s restart job completed.", | |
841 | gs.special->name); | |
842 | set_phase(PHASE_WAITING_ZEBRA_UP); | |
843 | ||
d62a17ae | 844 | /*FALLTHRU*/ |
a6810074 DL |
845 | case PHASE_WAITING_ZEBRA_UP: |
846 | if (!IS_UP(gs.special)) | |
847 | break; | |
848 | zlog_info("Phased restart: %s is now up.", gs.special->name); | |
849 | { | |
850 | struct daemon *dmn; | |
851 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { | |
852 | if (dmn != gs.special) | |
853 | run_job(&dmn->restart, "start", | |
854 | gs.start_command, 1, 0); | |
855 | } | |
856 | } | |
857 | gs.phase = PHASE_NONE; | |
858 | THREAD_OFF(gs.t_phase_hanging); | |
859 | zlog_notice("Phased global restart has completed."); | |
860 | break; | |
861 | } | |
8b886ca7 | 862 | } |
863 | ||
a6810074 | 864 | static void try_restart(struct daemon *dmn) |
8b886ca7 | 865 | { |
f168b713 | 866 | if (watch_only) |
a6810074 | 867 | return; |
a6810074 | 868 | |
f168b713 DL |
869 | if (dmn != gs.special) { |
870 | if ((gs.special->state == DAEMON_UP) | |
871 | && (gs.phase == PHASE_NONE)) | |
872 | run_job(&dmn->restart, "restart", gs.restart_command, 0, | |
873 | 1); | |
874 | else | |
875 | zlog_debug( | |
876 | "%s: postponing restart attempt because master %s daemon " | |
877 | "not up [%s], or phased restart in progress", | |
878 | dmn->name, gs.special->name, | |
879 | state_str[gs.special->state]); | |
880 | return; | |
881 | } | |
882 | ||
883 | if ((gs.phase != PHASE_NONE) || gs.numpids) { | |
884 | if (gs.loglevel > LOG_DEBUG + 1) | |
885 | zlog_debug( | |
886 | "postponing phased global restart: restart already in " | |
887 | "progress [%s], or outstanding child processes [%d]", | |
888 | phase_str[gs.phase], gs.numpids); | |
889 | return; | |
890 | } | |
891 | /* Is it too soon for a restart? */ | |
892 | { | |
893 | struct timeval delay; | |
894 | if (time_elapsed(&delay, &gs.special->restart.time)->tv_sec | |
895 | < gs.special->restart.interval) { | |
a6810074 | 896 | if (gs.loglevel > LOG_DEBUG + 1) |
d62a17ae | 897 | zlog_debug( |
f168b713 DL |
898 | "postponing phased global restart: " |
899 | "elapsed time %ld < retry interval %ld", | |
900 | (long)delay.tv_sec, | |
901 | gs.special->restart.interval); | |
902 | return; | |
a6810074 | 903 | } |
8b886ca7 | 904 | } |
f168b713 | 905 | run_job(&gs.restart, "restart", gs.restart_command, 0, 1); |
8b886ca7 | 906 | } |
907 | ||
a6810074 | 908 | static int wakeup_unresponsive(struct thread *t_wakeup) |
8b886ca7 | 909 | { |
a6810074 DL |
910 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
911 | ||
912 | dmn->t_wakeup = NULL; | |
913 | if (dmn->state != DAEMON_UNRESPONSIVE) | |
f74ae2bb | 914 | flog_err(EC_WATCHFRR_CONNECTION, |
1c50c1c0 QY |
915 | "%s: no longer unresponsive (now %s), " |
916 | "wakeup should have been cancelled!", | |
917 | dmn->name, state_str[dmn->state]); | |
a6810074 DL |
918 | else { |
919 | SET_WAKEUP_UNRESPONSIVE(dmn); | |
920 | try_restart(dmn); | |
921 | } | |
922 | return 0; | |
8b886ca7 | 923 | } |
924 | ||
a6810074 | 925 | static int wakeup_no_answer(struct thread *t_wakeup) |
8b886ca7 | 926 | { |
a6810074 DL |
927 | struct daemon *dmn = THREAD_ARG(t_wakeup); |
928 | ||
929 | dmn->t_wakeup = NULL; | |
930 | dmn->state = DAEMON_UNRESPONSIVE; | |
f74ae2bb | 931 | flog_err(EC_WATCHFRR_CONNECTION, |
1c50c1c0 QY |
932 | "%s state -> unresponsive : no response yet to ping " |
933 | "sent %ld seconds ago", | |
934 | dmn->name, gs.timeout); | |
71e7975a DL |
935 | SET_WAKEUP_UNRESPONSIVE(dmn); |
936 | try_restart(dmn); | |
a6810074 | 937 | return 0; |
8b886ca7 | 938 | } |
939 | ||
a6810074 | 940 | static int wakeup_send_echo(struct thread *t_wakeup) |
8b886ca7 | 941 | { |
a6810074 DL |
942 | static const char echocmd[] = "echo " PING_TOKEN; |
943 | ssize_t rc; | |
944 | struct daemon *dmn = THREAD_ARG(t_wakeup); | |
945 | ||
946 | dmn->t_wakeup = NULL; | |
d62a17ae | 947 | if (((rc = write(dmn->fd, echocmd, sizeof(echocmd))) < 0) |
948 | || ((size_t)rc != sizeof(echocmd))) { | |
a6810074 DL |
949 | char why[100 + sizeof(echocmd)]; |
950 | snprintf(why, sizeof(why), | |
951 | "write '%s' returned %d instead of %u", echocmd, | |
d7c0a89a | 952 | (int)rc, (unsigned int)sizeof(echocmd)); |
a6810074 DL |
953 | daemon_down(dmn, why); |
954 | } else { | |
955 | gettimeofday(&dmn->echo_sent, NULL); | |
66e78ae6 QY |
956 | dmn->t_wakeup = NULL; |
957 | thread_add_timer(master, wakeup_no_answer, dmn, gs.timeout, | |
958 | &dmn->t_wakeup); | |
a6810074 DL |
959 | } |
960 | return 0; | |
8b886ca7 | 961 | } |
962 | ||
470bc619 QY |
963 | bool check_all_up(void) |
964 | { | |
965 | struct daemon *dmn; | |
966 | ||
967 | for (dmn = gs.daemons; dmn; dmn = dmn->next) | |
968 | if (dmn->state != DAEMON_UP) | |
969 | return false; | |
970 | return true; | |
971 | } | |
972 | ||
af568444 DL |
973 | void watchfrr_status(struct vty *vty) |
974 | { | |
975 | struct daemon *dmn; | |
976 | struct timeval delay; | |
977 | ||
978 | vty_out(vty, "watchfrr global phase: %s\n", phase_str[gs.phase]); | |
979 | if (gs.restart.pid) | |
980 | vty_out(vty, " global restart running, pid %ld\n", | |
981 | (long)gs.restart.pid); | |
982 | ||
983 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { | |
984 | vty_out(vty, " %-20s %s\n", dmn->name, state_str[dmn->state]); | |
985 | if (dmn->restart.pid) | |
986 | vty_out(vty, " restart running, pid %ld\n", | |
987 | (long)dmn->restart.pid); | |
988 | else if (dmn->state == DAEMON_DOWN && | |
989 | time_elapsed(&delay, &dmn->restart.time)->tv_sec | |
990 | < dmn->restart.interval) | |
991 | vty_out(vty, " restarting in %ld seconds" | |
992 | " (%lds backoff interval)\n", | |
993 | dmn->restart.interval - delay.tv_sec, | |
994 | dmn->restart.interval); | |
995 | } | |
996 | } | |
997 | ||
a6810074 | 998 | static void sigint(void) |
8b886ca7 | 999 | { |
a6810074 DL |
1000 | zlog_notice("Terminating on signal"); |
1001 | systemd_send_stopping(); | |
1002 | exit(0); | |
8b886ca7 | 1003 | } |
1004 | ||
a6810074 | 1005 | static int valid_command(const char *cmd) |
8b886ca7 | 1006 | { |
a6810074 | 1007 | char *p; |
8b886ca7 | 1008 | |
a6810074 | 1009 | return ((p = strchr(cmd, '%')) != NULL) && (*(p + 1) == 's') |
d62a17ae | 1010 | && !strchr(p + 1, '%'); |
8b886ca7 | 1011 | } |
1012 | ||
c8b40f86 | 1013 | /* This is an ugly hack to circumvent problems with passing command-line |
1014 | arguments that contain spaces. The fix is to use a configuration file. */ | |
a6810074 | 1015 | static char *translate_blanks(const char *cmd, const char *blankstr) |
c8b40f86 | 1016 | { |
a6810074 DL |
1017 | char *res; |
1018 | char *p; | |
1019 | size_t bslen = strlen(blankstr); | |
1020 | ||
1021 | if (!(res = strdup(cmd))) { | |
1022 | perror("strdup"); | |
1023 | exit(1); | |
1024 | } | |
1025 | while ((p = strstr(res, blankstr)) != NULL) { | |
1026 | *p = ' '; | |
1027 | if (bslen != 1) | |
1028 | memmove(p + 1, p + bslen, strlen(p + bslen) + 1); | |
1029 | } | |
1030 | return res; | |
c8b40f86 | 1031 | } |
1032 | ||
0a7c7856 DL |
1033 | static void watchfrr_init(int argc, char **argv) |
1034 | { | |
1035 | const char *special = "zebra"; | |
1036 | int i; | |
1037 | struct daemon *dmn, **add = &gs.daemons; | |
1038 | char alldaemons[512] = "", *p = alldaemons; | |
1039 | ||
1040 | for (i = optind; i < argc; i++) { | |
1041 | dmn = XCALLOC(MTYPE_WATCHFRR_DAEMON, sizeof(*dmn)); | |
1042 | ||
1043 | dmn->name = dmn->restart.name = argv[i]; | |
1044 | dmn->state = DAEMON_INIT; | |
1045 | gs.numdaemons++; | |
1046 | gs.numdown++; | |
1047 | dmn->fd = -1; | |
1048 | dmn->t_wakeup = NULL; | |
c0e5cb52 | 1049 | thread_add_timer_msec(master, wakeup_init, dmn, 0, |
0a7c7856 DL |
1050 | &dmn->t_wakeup); |
1051 | dmn->restart.interval = gs.min_restart_interval; | |
1052 | *add = dmn; | |
1053 | add = &dmn->next; | |
1054 | ||
1055 | if (!strcmp(dmn->name, special)) | |
1056 | gs.special = dmn; | |
1057 | } | |
1058 | ||
1059 | if (!gs.daemons) { | |
1060 | fprintf(stderr, | |
1061 | "Must specify one or more daemons to monitor.\n\n"); | |
1062 | frr_help_exit(1); | |
1063 | } | |
1064 | if (!watch_only && !gs.special) { | |
1065 | fprintf(stderr, "\"%s\" daemon must be in daemon lists\n\n", | |
1066 | special); | |
1067 | frr_help_exit(1); | |
1068 | } | |
1069 | ||
1070 | for (dmn = gs.daemons; dmn; dmn = dmn->next) { | |
1071 | snprintf(p, alldaemons + sizeof(alldaemons) - p, "%s%s", | |
1072 | (p == alldaemons) ? "" : " ", dmn->name); | |
1073 | p += strlen(p); | |
1074 | } | |
1075 | zlog_notice("%s %s watching [%s]%s", progname, FRR_VERSION, alldaemons, | |
1076 | watch_only ? ", monitor mode" : ""); | |
1077 | } | |
1078 | ||
a6810074 | 1079 | struct zebra_privs_t watchfrr_privs = { |
95c4aff2 | 1080 | #ifdef VTY_GROUP |
a6810074 | 1081 | .vty_group = VTY_GROUP, |
95c4aff2 DL |
1082 | #endif |
1083 | }; | |
1084 | ||
4f04a76b DL |
1085 | static struct quagga_signal_t watchfrr_signals[] = { |
1086 | { | |
1087 | .signal = SIGINT, | |
1088 | .handler = sigint, | |
1089 | }, | |
1090 | { | |
1091 | .signal = SIGTERM, | |
1092 | .handler = sigint, | |
1093 | }, | |
1094 | { | |
1095 | .signal = SIGCHLD, | |
1096 | .handler = sigchild, | |
1097 | }, | |
1098 | }; | |
1099 | ||
1100 | FRR_DAEMON_INFO(watchfrr, WATCHFRR, | |
d62a17ae | 1101 | .flags = FRR_NO_PRIVSEP | FRR_NO_TCPVTY | FRR_LIMITED_CLI |
0a7c7856 DL |
1102 | | FRR_NO_CFG_PID_DRY | FRR_NO_ZCLIENT |
1103 | | FRR_DETACH_LATER, | |
4f04a76b | 1104 | |
d62a17ae | 1105 | .printhelp = printhelp, |
1106 | .copyright = "Copyright 2004 Andrew J. Schorr", | |
4f04a76b | 1107 | |
d62a17ae | 1108 | .signals = watchfrr_signals, |
1109 | .n_signals = array_size(watchfrr_signals), | |
4f04a76b | 1110 | |
d62a17ae | 1111 | .privs = &watchfrr_privs, ) |
4f04a76b | 1112 | |
999f153e DL |
1113 | #define DEPRECATED_OPTIONS "aAezR:" |
1114 | ||
a6810074 | 1115 | int main(int argc, char **argv) |
8b886ca7 | 1116 | { |
a6810074 | 1117 | int opt; |
a6810074 | 1118 | const char *blankstr = NULL; |
a6810074 | 1119 | |
4f04a76b DL |
1120 | frr_preinit(&watchfrr_di, argc, argv); |
1121 | progname = watchfrr_di.progname; | |
1122 | ||
999f153e | 1123 | frr_opt_add("b:dk:l:i:p:r:S:s:t:T:" DEPRECATED_OPTIONS, longopts, ""); |
a6810074 DL |
1124 | |
1125 | gs.restart.name = "all"; | |
4f04a76b | 1126 | while ((opt = frr_getopt(argc, argv, NULL)) != EOF) { |
999f153e DL |
1127 | if (opt && opt < 128 && strchr(DEPRECATED_OPTIONS, opt)) { |
1128 | fprintf(stderr, | |
1129 | "The -%c option no longer exists.\n" | |
1130 | "Please refer to the watchfrr(8) man page.\n", | |
1131 | opt); | |
1132 | exit(1); | |
1133 | } | |
1134 | ||
a6810074 DL |
1135 | switch (opt) { |
1136 | case 0: | |
1137 | break; | |
a6810074 DL |
1138 | case 'b': |
1139 | blankstr = optarg; | |
1140 | break; | |
f168b713 DL |
1141 | case OPTION_DRY: |
1142 | watch_only = true; | |
a6810074 DL |
1143 | break; |
1144 | case 'k': | |
1145 | if (!valid_command(optarg)) { | |
1146 | fprintf(stderr, | |
1147 | "Invalid kill command, must contain '%%s': %s\n", | |
1148 | optarg); | |
4f04a76b | 1149 | frr_help_exit(1); |
a6810074 DL |
1150 | } |
1151 | gs.stop_command = optarg; | |
1152 | break; | |
d62a17ae | 1153 | case 'l': { |
1154 | char garbage[3]; | |
1155 | if ((sscanf(optarg, "%d%1s", &gs.loglevel, garbage) | |
1156 | != 1) | |
1157 | || (gs.loglevel < LOG_EMERG)) { | |
1158 | fprintf(stderr, | |
1159 | "Invalid loglevel argument: %s\n", | |
1160 | optarg); | |
1161 | frr_help_exit(1); | |
a6810074 | 1162 | } |
d62a17ae | 1163 | } break; |
1164 | case OPTION_MINRESTART: { | |
1165 | char garbage[3]; | |
1166 | if ((sscanf(optarg, "%ld%1s", &gs.min_restart_interval, | |
1167 | garbage) | |
1168 | != 1) | |
1169 | || (gs.min_restart_interval < 0)) { | |
1170 | fprintf(stderr, | |
1171 | "Invalid min_restart_interval argument: %s\n", | |
1172 | optarg); | |
1173 | frr_help_exit(1); | |
a6810074 | 1174 | } |
d62a17ae | 1175 | } break; |
1176 | case OPTION_MAXRESTART: { | |
1177 | char garbage[3]; | |
1178 | if ((sscanf(optarg, "%ld%1s", &gs.max_restart_interval, | |
1179 | garbage) | |
1180 | != 1) | |
1181 | || (gs.max_restart_interval < 0)) { | |
1182 | fprintf(stderr, | |
1183 | "Invalid max_restart_interval argument: %s\n", | |
1184 | optarg); | |
1185 | frr_help_exit(1); | |
a6810074 | 1186 | } |
d62a17ae | 1187 | } break; |
1188 | case 'i': { | |
1189 | char garbage[3]; | |
1190 | int period; | |
1191 | if ((sscanf(optarg, "%d%1s", &period, garbage) != 1) | |
1192 | || (gs.period < 1)) { | |
1193 | fprintf(stderr, | |
1194 | "Invalid interval argument: %s\n", | |
1195 | optarg); | |
1196 | frr_help_exit(1); | |
a6810074 | 1197 | } |
d62a17ae | 1198 | gs.period = 1000 * period; |
1199 | } break; | |
a6810074 | 1200 | case 'p': |
0a7c7856 | 1201 | watchfrr_di.pid_file = optarg; |
a6810074 DL |
1202 | break; |
1203 | case 'r': | |
a6810074 DL |
1204 | if (!valid_command(optarg)) { |
1205 | fprintf(stderr, | |
1206 | "Invalid restart command, must contain '%%s': %s\n", | |
1207 | optarg); | |
4f04a76b | 1208 | frr_help_exit(1); |
a6810074 DL |
1209 | } |
1210 | gs.restart_command = optarg; | |
a6810074 DL |
1211 | break; |
1212 | case 's': | |
1213 | if (!valid_command(optarg)) { | |
1214 | fprintf(stderr, | |
1215 | "Invalid start command, must contain '%%s': %s\n", | |
1216 | optarg); | |
4f04a76b | 1217 | frr_help_exit(1); |
a6810074 DL |
1218 | } |
1219 | gs.start_command = optarg; | |
1220 | break; | |
1221 | case 'S': | |
1222 | gs.vtydir = optarg; | |
1223 | break; | |
d62a17ae | 1224 | case 't': { |
1225 | char garbage[3]; | |
1226 | if ((sscanf(optarg, "%ld%1s", &gs.timeout, garbage) | |
1227 | != 1) | |
1228 | || (gs.timeout < 1)) { | |
1229 | fprintf(stderr, | |
1230 | "Invalid timeout argument: %s\n", | |
1231 | optarg); | |
1232 | frr_help_exit(1); | |
a6810074 | 1233 | } |
d62a17ae | 1234 | } break; |
1235 | case 'T': { | |
1236 | char garbage[3]; | |
1237 | if ((sscanf(optarg, "%ld%1s", &gs.restart_timeout, | |
1238 | garbage) | |
1239 | != 1) | |
1240 | || (gs.restart_timeout < 1)) { | |
1241 | fprintf(stderr, | |
1242 | "Invalid restart timeout argument: %s\n", | |
1243 | optarg); | |
1244 | frr_help_exit(1); | |
a6810074 | 1245 | } |
d62a17ae | 1246 | } break; |
a6810074 DL |
1247 | default: |
1248 | fputs("Invalid option.\n", stderr); | |
4f04a76b | 1249 | frr_help_exit(1); |
a6810074 | 1250 | } |
8b886ca7 | 1251 | } |
a6810074 | 1252 | |
71e7975a DL |
1253 | if (watch_only |
1254 | && (gs.start_command || gs.stop_command || gs.restart_command)) { | |
d87ae5cc | 1255 | fputs("Options -r/-s/-k are not used when --dry is active.\n", |
a6810074 | 1256 | stderr); |
8b886ca7 | 1257 | } |
f168b713 DL |
1258 | if (!watch_only |
1259 | && (!gs.restart_command || !gs.start_command || !gs.stop_command)) { | |
1260 | fprintf(stderr, | |
1261 | "Options -s (start), -k (kill), and -r (restart) are required.\n"); | |
1262 | frr_help_exit(1); | |
8b886ca7 | 1263 | } |
8b886ca7 | 1264 | |
a6810074 DL |
1265 | if (blankstr) { |
1266 | if (gs.restart_command) | |
1267 | gs.restart_command = | |
d62a17ae | 1268 | translate_blanks(gs.restart_command, blankstr); |
a6810074 DL |
1269 | if (gs.start_command) |
1270 | gs.start_command = | |
d62a17ae | 1271 | translate_blanks(gs.start_command, blankstr); |
a6810074 DL |
1272 | if (gs.stop_command) |
1273 | gs.stop_command = | |
d62a17ae | 1274 | translate_blanks(gs.stop_command, blankstr); |
065de903 | 1275 | } |
8b886ca7 | 1276 | |
a6810074 | 1277 | gs.restart.interval = gs.min_restart_interval; |
8b886ca7 | 1278 | |
4f04a76b | 1279 | master = frr_init(); |
b647dc2a | 1280 | watchfrr_error_init(); |
0a7c7856 DL |
1281 | watchfrr_init(argc, argv); |
1282 | watchfrr_vty_init(); | |
1283 | ||
1284 | frr_config_fork(); | |
4f04a76b | 1285 | |
dd8376fe | 1286 | zlog_set_level(ZLOG_DEST_MONITOR, ZLOG_DISABLED); |
0a7c7856 | 1287 | if (watchfrr_di.daemon_mode) |
dd8376fe | 1288 | zlog_set_level(ZLOG_DEST_SYSLOG, MIN(gs.loglevel, LOG_DEBUG)); |
0a7c7856 | 1289 | else |
dd8376fe | 1290 | zlog_set_level(ZLOG_DEST_STDOUT, MIN(gs.loglevel, LOG_DEBUG)); |
8b886ca7 | 1291 | |
0a7c7856 | 1292 | frr_run(master); |
8b886ca7 | 1293 | |
a6810074 DL |
1294 | systemd_send_stopping(); |
1295 | /* Not reached. */ | |
1296 | return 0; | |
8b886ca7 | 1297 | } |