summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--.github/monitoring-plugins.spec13
-rw-r--r--.gitignore1
-rw-r--r--plugins/Makefile.am3
-rw-r--r--plugins/check_ntp.c947
4 files changed, 1 insertions, 963 deletions
diff --git a/.github/monitoring-plugins.spec b/.github/monitoring-plugins.spec
index 5878a862..fa47cdc7 100644
--- a/.github/monitoring-plugins.spec
+++ b/.github/monitoring-plugins.spec
@@ -652,19 +652,6 @@ Provides check_nagios of the Monitoring Plugins.
652 652
653 653
654 654
655# check_ntp
656%package ntp
657Summary: Monitoring Plugins - check_ntp
658Requires: %{name} = %{version}-%{release}
659
660%description ntp
661Provides check_ntp of the Monitoring Plugins.
662
663%files ntp
664%{plugindir}/check_ntp
665
666
667
668# check_ntp_peer 655# check_ntp_peer
669%package ntp_peer 656%package ntp_peer
670Summary: Monitoring Plugins - check_ntp_peer 657Summary: Monitoring Plugins - check_ntp_peer
diff --git a/.gitignore b/.gitignore
index a91edd7c..d0115b05 100644
--- a/.gitignore
+++ b/.gitignore
@@ -174,7 +174,6 @@ plugins/check_disk.d/.dirstamp
174/plugins/check_netsaint 174/plugins/check_netsaint
175/plugins/check_nntp 175/plugins/check_nntp
176/plugins/check_nntps 176/plugins/check_nntps
177/plugins/check_ntp
178/plugins/check_ntp_peer 177/plugins/check_ntp_peer
179/plugins/check_ntp_time 178/plugins/check_ntp_time
180/plugins/check_pgsql 179/plugins/check_pgsql
diff --git a/plugins/Makefile.am b/plugins/Makefile.am
index 877c0947..d098fa8a 100644
--- a/plugins/Makefile.am
+++ b/plugins/Makefile.am
@@ -33,7 +33,7 @@ MATHLIBS = @MATHLIBS@
33#AM_CFLAGS = -Wall 33#AM_CFLAGS = -Wall
34 34
35libexec_PROGRAMS = check_apt check_cluster check_disk check_dummy check_http check_load \ 35libexec_PROGRAMS = check_apt check_cluster check_disk check_dummy check_http check_load \
36 check_mrtg check_mrtgtraf check_ntp check_ntp_peer check_ping \ 36 check_mrtg check_mrtgtraf check_ntp_peer check_ping \
37 check_real check_smtp check_ssh check_tcp check_time check_ntp_time \ 37 check_real check_smtp check_ssh check_tcp check_time check_ntp_time \
38 check_ups check_users negate \ 38 check_ups check_users negate \
39 urlize @EXTRAS@ \ 39 urlize @EXTRAS@ \
@@ -156,7 +156,6 @@ check_mysql_query_CFLAGS = $(AM_CFLAGS) $(MYSQLCFLAGS)
156check_mysql_query_CPPFLAGS = $(AM_CPPFLAGS) $(MYSQLINCLUDE) 156check_mysql_query_CPPFLAGS = $(AM_CPPFLAGS) $(MYSQLINCLUDE)
157check_mysql_query_LDADD = $(NETLIBS) $(MYSQLLIBS) 157check_mysql_query_LDADD = $(NETLIBS) $(MYSQLLIBS)
158check_nagios_LDADD = $(BASEOBJS) 158check_nagios_LDADD = $(BASEOBJS)
159check_ntp_LDADD = $(NETLIBS) $(MATHLIBS)
160check_ntp_peer_LDADD = $(NETLIBS) $(MATHLIBS) 159check_ntp_peer_LDADD = $(NETLIBS) $(MATHLIBS)
161check_pgsql_LDADD = $(NETLIBS) $(PGLIBS) 160check_pgsql_LDADD = $(NETLIBS) $(PGLIBS)
162check_ping_LDADD = $(NETLIBS) 161check_ping_LDADD = $(NETLIBS)
diff --git a/plugins/check_ntp.c b/plugins/check_ntp.c
deleted file mode 100644
index b22cc3c1..00000000
--- a/plugins/check_ntp.c
+++ /dev/null
@@ -1,947 +0,0 @@
1/*****************************************************************************
2 *
3 * Monitoring check_ntp plugin
4 *
5 * License: GPL
6 * Copyright (c) 2006 Sean Finney <seanius@seanius.net>
7 * Copyright (c) 2006-2024 Monitoring Plugins Development Team
8 *
9 * Description:
10 *
11 * This file contains the check_ntp plugin
12 *
13 * This plugin to check ntp servers independent of any commandline
14 * programs or external libraries.
15 *
16 *
17 * This program is free software: you can redistribute it and/or modify
18 * it under the terms of the GNU General Public License as published by
19 * the Free Software Foundation, either version 3 of the License, or
20 * (at your option) any later version.
21 *
22 * This program is distributed in the hope that it will be useful,
23 * but WITHOUT ANY WARRANTY; without even the implied warranty of
24 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25 * GNU General Public License for more details.
26 *
27 * You should have received a copy of the GNU General Public License
28 * along with this program. If not, see <http://www.gnu.org/licenses/>.
29 *
30 *
31 *****************************************************************************/
32
33const char *progname = "check_ntp";
34const char *copyright = "2006-2024";
35const char *email = "devel@monitoring-plugins.org";
36
37#include "common.h"
38#include "netutils.h"
39#include "utils.h"
40
41static char *server_address = NULL;
42static int verbose = 0;
43static bool do_offset = false;
44static char *owarn = "60";
45static char *ocrit = "120";
46static bool do_jitter = false;
47static char *jwarn = "5000";
48static char *jcrit = "10000";
49
50static int process_arguments(int /*argc*/, char ** /*argv*/);
51static thresholds *offset_thresholds = NULL;
52static thresholds *jitter_thresholds = NULL;
53static void print_help(void);
54void print_usage(void);
55
56/* number of times to perform each request to get a good average. */
57#ifndef AVG_NUM
58# define AVG_NUM 4
59#endif
60
61/* max size of control message data */
62#define MAX_CM_SIZE 468
63
64/* this structure holds everything in an ntp request/response as per rfc1305 */
65typedef struct {
66 uint8_t flags; /* byte with leapindicator,vers,mode. see macros */
67 uint8_t stratum; /* clock stratum */
68 int8_t poll; /* polling interval */
69 int8_t precision; /* precision of the local clock */
70 int32_t rtdelay; /* total rt delay, as a fixed point num. see macros */
71 uint32_t rtdisp; /* like above, but for max err to primary src */
72 uint32_t refid; /* ref clock identifier */
73 uint64_t refts; /* reference timestamp. local time local clock */
74 uint64_t origts; /* time at which request departed client */
75 uint64_t rxts; /* time at which request arrived at server */
76 uint64_t txts; /* time at which request departed server */
77} ntp_message;
78
79/* this structure holds data about results from querying offset from a peer */
80typedef struct {
81 time_t waiting; /* ts set when we started waiting for a response */
82 int num_responses; /* number of successfully received responses */
83 uint8_t stratum; /* copied verbatim from the ntp_message */
84 double rtdelay; /* converted from the ntp_message */
85 double rtdisp; /* converted from the ntp_message */
86 double offset[AVG_NUM]; /* offsets from each response */
87 uint8_t flags; /* byte with leapindicator,vers,mode. see macros */
88} ntp_server_results;
89
90/* this structure holds everything in an ntp control message as per rfc1305 */
91typedef struct {
92 uint8_t flags; /* byte with leapindicator,vers,mode. see macros */
93 uint8_t op; /* R,E,M bits and Opcode */
94 uint16_t seq; /* Packet sequence */
95 uint16_t status; /* Clock status */
96 uint16_t assoc; /* Association */
97 uint16_t offset; /* Similar to TCP sequence # */
98 uint16_t count; /* # bytes of data */
99 char data[MAX_CM_SIZE]; /* ASCII data of the request */
100 /* NB: not necessarily NULL terminated! */
101} ntp_control_message;
102
103/* this is an association/status-word pair found in control packet responses */
104typedef struct {
105 uint16_t assoc;
106 uint16_t status;
107} ntp_assoc_status_pair;
108
109/* bits 1,2 are the leap indicator */
110#define LI_MASK 0xc0
111#define LI(x) ((x & LI_MASK) >> 6)
112#define LI_SET(x, y) \
113 do { \
114 x |= ((y << 6) & LI_MASK); \
115 } while (0)
116/* and these are the values of the leap indicator */
117#define LI_NOWARNING 0x00
118#define LI_EXTRASEC 0x01
119#define LI_MISSINGSEC 0x02
120#define LI_ALARM 0x03
121/* bits 3,4,5 are the ntp version */
122#define VN_MASK 0x38
123#define VN(x) ((x & VN_MASK) >> 3)
124#define VN_SET(x, y) \
125 do { \
126 x |= ((y << 3) & VN_MASK); \
127 } while (0)
128#define VN_RESERVED 0x02
129/* bits 6,7,8 are the ntp mode */
130#define MODE_MASK 0x07
131#define MODE(x) (x & MODE_MASK)
132#define MODE_SET(x, y) \
133 do { \
134 x |= (y & MODE_MASK); \
135 } while (0)
136/* here are some values */
137#define MODE_CLIENT 0x03
138#define MODE_CONTROLMSG 0x06
139/* In control message, bits 8-10 are R,E,M bits */
140#define REM_MASK 0xe0
141#define REM_RESP 0x80
142#define REM_ERROR 0x40
143#define REM_MORE 0x20
144/* In control message, bits 11 - 15 are opcode */
145#define OP_MASK 0x1f
146#define OP_SET(x, y) \
147 do { \
148 x |= (y & OP_MASK); \
149 } while (0)
150#define OP_READSTAT 0x01
151#define OP_READVAR 0x02
152/* In peer status bytes, bits 6,7,8 determine clock selection status */
153#define PEER_SEL(x) ((ntohs(x) >> 8) & 0x07)
154#define PEER_INCLUDED 0x04
155#define PEER_SYNCSOURCE 0x06
156
157/**
158 ** a note about the 32-bit "fixed point" numbers:
159 **
160 they are divided into halves, each being a 16-bit int in network byte order:
161 - the first 16 bits are an int on the left side of a decimal point.
162 - the second 16 bits represent a fraction n/(2^16)
163 likewise for the 64-bit "fixed point" numbers with everything doubled :)
164 **/
165
166/* macros to access the left/right 16 bits of a 32-bit ntp "fixed point"
167 number. note that these can be used as lvalues too */
168#define L16(x) (((uint16_t *)&x)[0])
169#define R16(x) (((uint16_t *)&x)[1])
170/* macros to access the left/right 32 bits of a 64-bit ntp "fixed point"
171 number. these too can be used as lvalues */
172#define L32(x) (((uint32_t *)&x)[0])
173#define R32(x) (((uint32_t *)&x)[1])
174
175/* ntp wants seconds since 1/1/00, epoch is 1/1/70. this is the difference */
176#define EPOCHDIFF 0x83aa7e80UL
177
178/* extract a 32-bit ntp fixed point number into a double */
179#define NTP32asDOUBLE(x) (ntohs(L16(x)) + (double)ntohs(R16(x)) / 65536.0)
180
181/* likewise for a 64-bit ntp fp number */
182#define NTP64asDOUBLE(n) \
183 (double)(((uint64_t)n) ? (ntohl(L32(n)) - EPOCHDIFF) + \
184 (.00000001 * (0.5 + (double)(ntohl(R32(n)) / 42.94967296))) \
185 : 0)
186
187/* convert a struct timeval to a double */
188#define TVasDOUBLE(x) (double)(x.tv_sec + (0.000001 * x.tv_usec))
189
190/* convert an ntp 64-bit fp number to a struct timeval */
191#define NTP64toTV(n, t) \
192 do { \
193 if (!n) \
194 t.tv_sec = t.tv_usec = 0; \
195 else { \
196 t.tv_sec = ntohl(L32(n)) - EPOCHDIFF; \
197 t.tv_usec = (int)(0.5 + (double)(ntohl(R32(n)) / 4294.967296)); \
198 } \
199 } while (0)
200
201/* convert a struct timeval to an ntp 64-bit fp number */
202#define TVtoNTP64(t, n) \
203 do { \
204 if (!t.tv_usec && !t.tv_sec) \
205 n = 0x0UL; \
206 else { \
207 L32(n) = htonl(t.tv_sec + EPOCHDIFF); \
208 R32(n) = htonl((uint64_t)((4294.967296 * t.tv_usec) + .5)); \
209 } \
210 } while (0)
211
212/* NTP control message header is 12 bytes, plus any data in the data
213 * field, plus null padding to the nearest 32-bit boundary per rfc.
214 */
215#define SIZEOF_NTPCM(m) \
216 (12 + ntohs(m.count) + ((ntohs(m.count) % 4) ? 4 - (ntohs(m.count) % 4) : 0))
217
218/* finally, a little helper or two for debugging: */
219#define DBG(x) \
220 do { \
221 if (verbose > 1) { \
222 x; \
223 } \
224 } while (0);
225#define PRINTSOCKADDR(x) \
226 do { \
227 printf("%u.%u.%u.%u", (x >> 24) & 0xff, (x >> 16) & 0xff, (x >> 8) & 0xff, x & 0xff); \
228 } while (0);
229
230/* calculate the offset of the local clock */
231static inline double calc_offset(const ntp_message *m, const struct timeval *t) {
232 double client_tx, peer_rx, peer_tx, client_rx;
233 client_tx = NTP64asDOUBLE(m->origts);
234 peer_rx = NTP64asDOUBLE(m->rxts);
235 peer_tx = NTP64asDOUBLE(m->txts);
236 client_rx = TVasDOUBLE((*t));
237 return (.5 * ((peer_tx - client_rx) + (peer_rx - client_tx)));
238}
239
240/* print out a ntp packet in human readable/debuggable format */
241void print_ntp_message(const ntp_message *p) {
242 struct timeval ref, orig, rx, tx;
243
244 NTP64toTV(p->refts, ref);
245 NTP64toTV(p->origts, orig);
246 NTP64toTV(p->rxts, rx);
247 NTP64toTV(p->txts, tx);
248
249 printf("packet contents:\n");
250 printf("\tflags: 0x%.2x\n", p->flags);
251 printf("\t li=%d (0x%.2x)\n", LI(p->flags), p->flags & LI_MASK);
252 printf("\t vn=%d (0x%.2x)\n", VN(p->flags), p->flags & VN_MASK);
253 printf("\t mode=%d (0x%.2x)\n", MODE(p->flags), p->flags & MODE_MASK);
254 printf("\tstratum = %d\n", p->stratum);
255 printf("\tpoll = %g\n", pow(2, p->poll));
256 printf("\tprecision = %g\n", pow(2, p->precision));
257 printf("\trtdelay = %-.16g\n", NTP32asDOUBLE(p->rtdelay));
258 printf("\trtdisp = %-.16g\n", NTP32asDOUBLE(p->rtdisp));
259 printf("\trefid = %x\n", p->refid);
260 printf("\trefts = %-.16g\n", NTP64asDOUBLE(p->refts));
261 printf("\torigts = %-.16g\n", NTP64asDOUBLE(p->origts));
262 printf("\trxts = %-.16g\n", NTP64asDOUBLE(p->rxts));
263 printf("\ttxts = %-.16g\n", NTP64asDOUBLE(p->txts));
264}
265
266void print_ntp_control_message(const ntp_control_message *p) {
267 int i = 0, numpeers = 0;
268 const ntp_assoc_status_pair *peer = NULL;
269
270 printf("control packet contents:\n");
271 printf("\tflags: 0x%.2x , 0x%.2x\n", p->flags, p->op);
272 printf("\t li=%d (0x%.2x)\n", LI(p->flags), p->flags & LI_MASK);
273 printf("\t vn=%d (0x%.2x)\n", VN(p->flags), p->flags & VN_MASK);
274 printf("\t mode=%d (0x%.2x)\n", MODE(p->flags), p->flags & MODE_MASK);
275 printf("\t response=%d (0x%.2x)\n", (p->op & REM_RESP) > 0, p->op & REM_RESP);
276 printf("\t more=%d (0x%.2x)\n", (p->op & REM_MORE) > 0, p->op & REM_MORE);
277 printf("\t error=%d (0x%.2x)\n", (p->op & REM_ERROR) > 0, p->op & REM_ERROR);
278 printf("\t op=%d (0x%.2x)\n", p->op & OP_MASK, p->op & OP_MASK);
279 printf("\tsequence: %d (0x%.2x)\n", ntohs(p->seq), ntohs(p->seq));
280 printf("\tstatus: %d (0x%.2x)\n", ntohs(p->status), ntohs(p->status));
281 printf("\tassoc: %d (0x%.2x)\n", ntohs(p->assoc), ntohs(p->assoc));
282 printf("\toffset: %d (0x%.2x)\n", ntohs(p->offset), ntohs(p->offset));
283 printf("\tcount: %d (0x%.2x)\n", ntohs(p->count), ntohs(p->count));
284 numpeers = ntohs(p->count) / (sizeof(ntp_assoc_status_pair));
285 if (p->op & REM_RESP && p->op & OP_READSTAT) {
286 peer = (ntp_assoc_status_pair *)p->data;
287 for (i = 0; i < numpeers; i++) {
288 printf("\tpeer id %.2x status %.2x", ntohs(peer[i].assoc), ntohs(peer[i].status));
289 if (PEER_SEL(peer[i].status) >= PEER_INCLUDED) {
290 if (PEER_SEL(peer[i].status) >= PEER_SYNCSOURCE) {
291 printf(" <-- current sync source");
292 } else {
293 printf(" <-- current sync candidate");
294 }
295 }
296 printf("\n");
297 }
298 }
299}
300
301void setup_request(ntp_message *p) {
302 struct timeval t;
303
304 memset(p, 0, sizeof(ntp_message));
305 LI_SET(p->flags, LI_ALARM);
306 VN_SET(p->flags, 4);
307 MODE_SET(p->flags, MODE_CLIENT);
308 p->poll = 4;
309 p->precision = (int8_t)0xfa;
310 L16(p->rtdelay) = htons(1);
311 L16(p->rtdisp) = htons(1);
312
313 gettimeofday(&t, NULL);
314 TVtoNTP64(t, p->txts);
315}
316
317/* select the "best" server from a list of servers, and return its index.
318 * this is done by filtering servers based on stratum, dispersion, and
319 * finally round-trip delay. */
320int best_offset_server(const ntp_server_results *slist, int nservers) {
321 int cserver = 0, best_server = -1;
322
323 /* for each server */
324 for (cserver = 0; cserver < nservers; cserver++) {
325 /* We don't want any servers that fails these tests */
326 /* Sort out servers that didn't respond or responede with a 0 stratum;
327 * stratum 0 is for reference clocks so no NTP server should ever report
328 * a stratum 0 */
329 if (slist[cserver].stratum == 0) {
330 if (verbose) {
331 printf("discarding peer %d: stratum=%d\n", cserver, slist[cserver].stratum);
332 }
333 continue;
334 }
335 /* Sort out servers with error flags */
336 if (LI(slist[cserver].flags) == LI_ALARM) {
337 if (verbose) {
338 printf("discarding peer %d: flags=%d\n", cserver, LI(slist[cserver].flags));
339 }
340 continue;
341 }
342
343 /* If we don't have a server yet, use the first one */
344 if (best_server == -1) {
345 best_server = cserver;
346 DBG(printf("using peer %d as our first candidate\n", best_server));
347 continue;
348 }
349
350 /* compare the server to the best one we've seen so far */
351 /* does it have an equal or better stratum? */
352 DBG(printf("comparing peer %d with peer %d\n", cserver, best_server));
353 if (slist[cserver].stratum <= slist[best_server].stratum) {
354 DBG(printf("stratum for peer %d <= peer %d\n", cserver, best_server));
355 /* does it have an equal or better dispersion? */
356 if (slist[cserver].rtdisp <= slist[best_server].rtdisp) {
357 DBG(printf("dispersion for peer %d <= peer %d\n", cserver, best_server));
358 /* does it have a better rtdelay? */
359 if (slist[cserver].rtdelay < slist[best_server].rtdelay) {
360 DBG(printf("rtdelay for peer %d < peer %d\n", cserver, best_server));
361 best_server = cserver;
362 DBG(printf("peer %d is now our best candidate\n", best_server));
363 }
364 }
365 }
366 }
367
368 if (best_server >= 0) {
369 DBG(printf("best server selected: peer %d\n", best_server));
370 return best_server;
371 } else {
372 DBG(printf("no peers meeting synchronization criteria :(\n"));
373 return -1;
374 }
375}
376
377/* do everything we need to get the total average offset
378 * - we use a certain amount of parallelization with poll() to ensure
379 * we don't waste time sitting around waiting for single packets.
380 * - we also "manually" handle resolving host names and connecting, because
381 * we have to do it in a way that our lazy macros don't handle currently :( */
382double offset_request(const char *host, int *status) {
383 int i = 0, ga_result = 0, num_hosts = 0, *socklist = NULL, respnum = 0;
384 int servers_completed = 0, one_read = 0, servers_readable = 0, best_index = -1;
385 time_t now_time = 0, start_ts = 0;
386 ntp_message *req = NULL;
387 double avg_offset = 0.;
388 struct timeval recv_time;
389 struct addrinfo *ai = NULL, *ai_tmp = NULL, hints;
390 struct pollfd *ufds = NULL;
391 ntp_server_results *servers = NULL;
392
393 /* setup hints to only return results from getaddrinfo that we'd like */
394 memset(&hints, 0, sizeof(struct addrinfo));
395 hints.ai_family = address_family;
396 hints.ai_protocol = IPPROTO_UDP;
397 hints.ai_socktype = SOCK_DGRAM;
398
399 /* fill in ai with the list of hosts resolved by the host name */
400 ga_result = getaddrinfo(host, "123", &hints, &ai);
401 if (ga_result != 0) {
402 die(STATE_UNKNOWN, "error getting address for %s: %s\n", host, gai_strerror(ga_result));
403 }
404
405 /* count the number of returned hosts, and allocate stuff accordingly */
406 for (ai_tmp = ai; ai_tmp != NULL; ai_tmp = ai_tmp->ai_next) {
407 num_hosts++;
408 }
409 req = (ntp_message *)malloc(sizeof(ntp_message) * num_hosts);
410 if (req == NULL) {
411 die(STATE_UNKNOWN, "can not allocate ntp message array");
412 }
413 socklist = (int *)malloc(sizeof(int) * num_hosts);
414 if (socklist == NULL) {
415 die(STATE_UNKNOWN, "can not allocate socket array");
416 }
417 ufds = (struct pollfd *)malloc(sizeof(struct pollfd) * num_hosts);
418 if (ufds == NULL) {
419 die(STATE_UNKNOWN, "can not allocate socket array");
420 }
421 servers = (ntp_server_results *)malloc(sizeof(ntp_server_results) * num_hosts);
422 if (servers == NULL) {
423 die(STATE_UNKNOWN, "can not allocate server array");
424 }
425 memset(servers, 0, sizeof(ntp_server_results) * num_hosts);
426 DBG(printf("Found %d peers to check\n", num_hosts));
427
428 /* setup each socket for writing, and the corresponding struct pollfd */
429 ai_tmp = ai;
430 for (i = 0; ai_tmp; i++) {
431 socklist[i] = socket(ai_tmp->ai_family, SOCK_DGRAM, IPPROTO_UDP);
432 if (socklist[i] == -1) {
433 perror(NULL);
434 die(STATE_UNKNOWN, "can not create new socket");
435 }
436 if (connect(socklist[i], ai_tmp->ai_addr, ai_tmp->ai_addrlen)) {
437 /* don't die here, because it is enough if there is one server
438 answering in time. This also would break for dual ipv4/6 stacked
439 ntp servers when the client only supports on of them.
440 */
441 DBG(printf("can't create socket connection on peer %i: %s\n", i, strerror(errno)));
442 } else {
443 ufds[i].fd = socklist[i];
444 ufds[i].events = POLLIN;
445 ufds[i].revents = 0;
446 }
447 ai_tmp = ai_tmp->ai_next;
448 }
449
450 /* now do AVG_NUM checks to each host. we stop before timeout/2 seconds
451 * have passed in order to ensure post-processing and jitter time. */
452 now_time = start_ts = time(NULL);
453 while (servers_completed < num_hosts && now_time - start_ts <= socket_timeout / 2) {
454 /* loop through each server and find each one which hasn't
455 * been touched in the past second or so and is still lacking
456 * some responses. for each of these servers, send a new request,
457 * and update the "waiting" timestamp with the current time. */
458 now_time = time(NULL);
459
460 for (i = 0; i < num_hosts; i++) {
461 if (servers[i].waiting < now_time && servers[i].num_responses < AVG_NUM) {
462 if (verbose && servers[i].waiting != 0) {
463 printf("re-");
464 }
465 if (verbose) {
466 printf("sending request to peer %d\n", i);
467 }
468 setup_request(&req[i]);
469 write(socklist[i], &req[i], sizeof(ntp_message));
470 servers[i].waiting = now_time;
471 break;
472 }
473 }
474
475 /* quickly poll for any sockets with pending data */
476 servers_readable = poll(ufds, num_hosts, 100);
477 if (servers_readable == -1) {
478 perror("polling ntp sockets");
479 die(STATE_UNKNOWN, "communication errors");
480 }
481
482 /* read from any sockets with pending data */
483 for (i = 0; servers_readable && i < num_hosts; i++) {
484 if (ufds[i].revents & POLLIN && servers[i].num_responses < AVG_NUM) {
485 if (verbose) {
486 printf("response from peer %d: ", i);
487 }
488
489 read(ufds[i].fd, &req[i], sizeof(ntp_message));
490 gettimeofday(&recv_time, NULL);
491 DBG(print_ntp_message(&req[i]));
492 respnum = servers[i].num_responses++;
493 servers[i].offset[respnum] = calc_offset(&req[i], &recv_time);
494 if (verbose) {
495 printf("offset %.10g\n", servers[i].offset[respnum]);
496 }
497 servers[i].stratum = req[i].stratum;
498 servers[i].rtdisp = NTP32asDOUBLE(req[i].rtdisp);
499 servers[i].rtdelay = NTP32asDOUBLE(req[i].rtdelay);
500 servers[i].waiting = 0;
501 servers[i].flags = req[i].flags;
502 servers_readable--;
503 one_read = 1;
504 if (servers[i].num_responses == AVG_NUM) {
505 servers_completed++;
506 }
507 }
508 }
509 /* lather, rinse, repeat. */
510 }
511
512 if (one_read == 0) {
513 die(STATE_CRITICAL, "NTP CRITICAL: No response from NTP server\n");
514 }
515
516 /* now, pick the best server from the list */
517 best_index = best_offset_server(servers, num_hosts);
518 if (best_index < 0) {
519 *status = STATE_UNKNOWN;
520 } else {
521 /* finally, calculate the average offset */
522 for (i = 0; i < servers[best_index].num_responses; i++) {
523 avg_offset += servers[best_index].offset[i];
524 }
525 avg_offset /= servers[best_index].num_responses;
526 }
527
528 /* cleanup */
529 /* FIXME: Not closing the socket to avoid reuse of the local port
530 * which can cause old NTP packets to be read instead of NTP control
531 * packets in jitter_request(). THERE MUST BE ANOTHER WAY...
532 * for(j=0; j<num_hosts; j++){ close(socklist[j]); } */
533 free(socklist);
534 free(ufds);
535 free(servers);
536 free(req);
537 freeaddrinfo(ai);
538
539 if (verbose) {
540 printf("overall average offset: %.10g\n", avg_offset);
541 }
542 return avg_offset;
543}
544
545void setup_control_request(ntp_control_message *p, uint8_t opcode, uint16_t seq) {
546 memset(p, 0, sizeof(ntp_control_message));
547 LI_SET(p->flags, LI_NOWARNING);
548 VN_SET(p->flags, VN_RESERVED);
549 MODE_SET(p->flags, MODE_CONTROLMSG);
550 OP_SET(p->op, opcode);
551 p->seq = htons(seq);
552 /* Remaining fields are zero for requests */
553}
554
555/* XXX handle responses with the error bit set */
556double jitter_request(int *status) {
557 int conn = -1, i, npeers = 0, num_candidates = 0;
558 bool syncsource_found = false;
559 int run = 0, min_peer_sel = PEER_INCLUDED, num_selected = 0, num_valid = 0;
560 int peers_size = 0, peer_offset = 0;
561 ntp_assoc_status_pair *peers = NULL;
562 ntp_control_message req;
563 const char *getvar = "jitter";
564 double rval = 0.0, jitter = -1.0;
565 char *startofvalue = NULL, *nptr = NULL;
566 void *tmp;
567
568 /* Long-winded explanation:
569 * Getting the jitter requires a number of steps:
570 * 1) Send a READSTAT request.
571 * 2) Interpret the READSTAT reply
572 * a) The data section contains a list of peer identifiers (16 bits)
573 * and associated status words (16 bits)
574 * b) We want the value of 0x06 in the SEL (peer selection) value,
575 * which means "current synchronizatin source". If that's missing,
576 * we take anything better than 0x04 (see the rfc for details) but
577 * set a minimum of warning.
578 * 3) Send a READVAR request for information on each peer identified
579 * in 2b greater than the minimum selection value.
580 * 4) Extract the jitter value from the data[] (it's ASCII)
581 */
582 my_udp_connect(server_address, 123, &conn);
583
584 /* keep sending requests until the server stops setting the
585 * REM_MORE bit, though usually this is only 1 packet. */
586 do {
587 setup_control_request(&req, OP_READSTAT, 1);
588 DBG(printf("sending READSTAT request"));
589 write(conn, &req, SIZEOF_NTPCM(req));
590 DBG(print_ntp_control_message(&req));
591 /* Attempt to read the largest size packet possible */
592 req.count = htons(MAX_CM_SIZE);
593 DBG(printf("receiving READSTAT response"))
594 read(conn, &req, SIZEOF_NTPCM(req));
595 DBG(print_ntp_control_message(&req));
596 /* Each peer identifier is 4 bytes in the data section, which
597 * we represent as a ntp_assoc_status_pair datatype.
598 */
599 peers_size += ntohs(req.count);
600 if ((tmp = realloc(peers, peers_size)) == NULL) {
601 free(peers), die(STATE_UNKNOWN, "can not (re)allocate 'peers' buffer\n");
602 }
603 peers = tmp;
604 memcpy((void *)((ptrdiff_t)peers + peer_offset), (void *)req.data, ntohs(req.count));
605 npeers = peers_size / sizeof(ntp_assoc_status_pair);
606 peer_offset += ntohs(req.count);
607 } while (req.op & REM_MORE);
608
609 /* first, let's find out if we have a sync source, or if there are
610 * at least some candidates. in the case of the latter we'll issue
611 * a warning but go ahead with the check on them. */
612 for (i = 0; i < npeers; i++) {
613 if (PEER_SEL(peers[i].status) >= PEER_INCLUDED) {
614 num_candidates++;
615 if (PEER_SEL(peers[i].status) >= PEER_SYNCSOURCE) {
616 syncsource_found = true;
617 min_peer_sel = PEER_SYNCSOURCE;
618 }
619 }
620 }
621 if (verbose) {
622 printf("%d candidate peers available\n", num_candidates);
623 }
624 if (verbose && syncsource_found) {
625 printf("synchronization source found\n");
626 }
627 if (!syncsource_found) {
628 *status = STATE_UNKNOWN;
629 if (verbose) {
630 printf("warning: no synchronization source found\n");
631 }
632 }
633
634 for (run = 0; run < AVG_NUM; run++) {
635 if (verbose) {
636 printf("jitter run %d of %d\n", run + 1, AVG_NUM);
637 }
638 for (i = 0; i < npeers; i++) {
639 /* Only query this server if it is the current sync source */
640 if (PEER_SEL(peers[i].status) >= min_peer_sel) {
641 char jitter_data[MAX_CM_SIZE + 1];
642 size_t jitter_data_count;
643
644 num_selected++;
645 setup_control_request(&req, OP_READVAR, 2);
646 req.assoc = peers[i].assoc;
647 /* By spec, putting the variable name "jitter" in the request
648 * should cause the server to provide _only_ the jitter value.
649 * thus reducing net traffic, guaranteeing us only a single
650 * datagram in reply, and making interpretation much simpler
651 */
652 /* Older servers doesn't know what jitter is, so if we get an
653 * error on the first pass we redo it with "dispersion" */
654 strncpy(req.data, getvar, MAX_CM_SIZE - 1);
655 req.count = htons(strlen(getvar));
656 DBG(printf("sending READVAR request...\n"));
657 write(conn, &req, SIZEOF_NTPCM(req));
658 DBG(print_ntp_control_message(&req));
659
660 req.count = htons(MAX_CM_SIZE);
661 DBG(printf("receiving READVAR response...\n"));
662 read(conn, &req, SIZEOF_NTPCM(req));
663 DBG(print_ntp_control_message(&req));
664
665 if (req.op & REM_ERROR && strstr(getvar, "jitter")) {
666 if (verbose) {
667 printf("The 'jitter' command failed (old ntp server?)\nRestarting with "
668 "'dispersion'...\n");
669 }
670 getvar = "dispersion";
671 num_selected--;
672 i--;
673 continue;
674 }
675
676 /* get to the float value */
677 if (verbose) {
678 printf("parsing jitter from peer %.2x: ", ntohs(peers[i].assoc));
679 }
680 if ((jitter_data_count = ntohs(req.count)) >= sizeof(jitter_data)) {
681 die(STATE_UNKNOWN, _("jitter response too large (%lu bytes)\n"),
682 (unsigned long)jitter_data_count);
683 }
684 memcpy(jitter_data, req.data, jitter_data_count);
685 jitter_data[jitter_data_count] = '\0';
686 startofvalue = strchr(jitter_data, '=');
687 if (startofvalue != NULL) {
688 startofvalue++;
689 jitter = strtod(startofvalue, &nptr);
690 }
691 if (startofvalue == NULL || startofvalue == nptr) {
692 printf("warning: unable to read server jitter response.\n");
693 *status = STATE_UNKNOWN;
694 } else {
695 if (verbose) {
696 printf("%g\n", jitter);
697 }
698 num_valid++;
699 rval += jitter;
700 }
701 }
702 }
703 if (verbose) {
704 printf("jitter parsed from %d/%d peers\n", num_valid, num_selected);
705 }
706 }
707
708 rval = num_valid ? rval / num_valid : -1.0;
709
710 close(conn);
711 if (peers != NULL) {
712 free(peers);
713 }
714 /* If we return -1.0, it means no synchronization source was found */
715 return rval;
716}
717
718int process_arguments(int argc, char **argv) {
719 int c;
720 int option = 0;
721 static struct option longopts[] = {
722 {"version", no_argument, 0, 'V'}, {"help", no_argument, 0, 'h'},
723 {"verbose", no_argument, 0, 'v'}, {"use-ipv4", no_argument, 0, '4'},
724 {"use-ipv6", no_argument, 0, '6'}, {"warning", required_argument, 0, 'w'},
725 {"critical", required_argument, 0, 'c'}, {"jwarn", required_argument, 0, 'j'},
726 {"jcrit", required_argument, 0, 'k'}, {"timeout", required_argument, 0, 't'},
727 {"hostname", required_argument, 0, 'H'}, {0, 0, 0, 0}};
728
729 if (argc < 2) {
730 usage("\n");
731 }
732
733 while (1) {
734 c = getopt_long(argc, argv, "Vhv46w:c:j:k:t:H:", longopts, &option);
735 if (c == -1 || c == EOF || c == 1) {
736 break;
737 }
738
739 switch (c) {
740 case 'h':
741 print_help();
742 exit(STATE_UNKNOWN);
743 break;
744 case 'V':
745 print_revision(progname, NP_VERSION);
746 exit(STATE_UNKNOWN);
747 break;
748 case 'v':
749 verbose++;
750 break;
751 case 'w':
752 do_offset = true;
753 owarn = optarg;
754 break;
755 case 'c':
756 do_offset = true;
757 ocrit = optarg;
758 break;
759 case 'j':
760 do_jitter = true;
761 jwarn = optarg;
762 break;
763 case 'k':
764 do_jitter = true;
765 jcrit = optarg;
766 break;
767 case 'H':
768 if (!is_host(optarg)) {
769 usage2(_("Invalid hostname/address"), optarg);
770 }
771 server_address = strdup(optarg);
772 break;
773 case 't':
774 socket_timeout = atoi(optarg);
775 break;
776 case '4':
777 address_family = AF_INET;
778 break;
779 case '6':
780#ifdef USE_IPV6
781 address_family = AF_INET6;
782#else
783 usage4(_("IPv6 support not available"));
784#endif
785 break;
786 case '?':
787 /* print short usage statement if args not parsable */
788 usage5();
789 break;
790 }
791 }
792
793 if (server_address == NULL) {
794 usage4(_("Hostname was not supplied"));
795 }
796
797 return 0;
798}
799
800char *perfd_offset(double offset) {
801 return fperfdata("offset", offset, "s", true, offset_thresholds->warning->end, true,
802 offset_thresholds->critical->end, false, 0, false, 0);
803}
804
805char *perfd_jitter(double jitter) {
806 return fperfdata("jitter", jitter, "s", do_jitter, jitter_thresholds->warning->end, do_jitter,
807 jitter_thresholds->critical->end, true, 0, false, 0);
808}
809
810int main(int argc, char *argv[]) {
811 int result, offset_result, jitter_result;
812 double offset = 0, jitter = 0;
813 char *result_line, *perfdata_line;
814
815 setlocale(LC_ALL, "");
816 bindtextdomain(PACKAGE, LOCALEDIR);
817 textdomain(PACKAGE);
818
819 result = offset_result = jitter_result = STATE_OK;
820
821 /* Parse extra opts if any */
822 argv = np_extra_opts(&argc, argv, progname);
823
824 if (process_arguments(argc, argv) == ERROR) {
825 usage4(_("Could not parse arguments"));
826 }
827
828 set_thresholds(&offset_thresholds, owarn, ocrit);
829 set_thresholds(&jitter_thresholds, jwarn, jcrit);
830
831 /* initialize alarm signal handling */
832 signal(SIGALRM, socket_timeout_alarm_handler);
833
834 /* set socket timeout */
835 alarm(socket_timeout);
836
837 offset = offset_request(server_address, &offset_result);
838 /* check_ntp used to always return CRITICAL if offset_result == STATE_UNKNOWN.
839 * Now we'll only do that is the offset thresholds were set */
840 if (do_offset && offset_result == STATE_UNKNOWN) {
841 result = STATE_CRITICAL;
842 } else {
843 result = get_status(fabs(offset), offset_thresholds);
844 }
845
846 /* If not told to check the jitter, we don't even send packets.
847 * jitter is checked using NTP control packets, which not all
848 * servers recognize. Trying to check the jitter on OpenNTPD
849 * (for example) will result in an error
850 */
851 if (do_jitter) {
852 jitter = jitter_request(&jitter_result);
853 result = max_state_alt(result, get_status(jitter, jitter_thresholds));
854 /* -1 indicates that we couldn't calculate the jitter
855 * Only overrides STATE_OK from the offset */
856 if (jitter == -1.0 && result == STATE_OK) {
857 result = STATE_UNKNOWN;
858 }
859 }
860 result = max_state_alt(result, jitter_result);
861
862 switch (result) {
863 case STATE_CRITICAL:
864 xasprintf(&result_line, _("NTP CRITICAL:"));
865 break;
866 case STATE_WARNING:
867 xasprintf(&result_line, _("NTP WARNING:"));
868 break;
869 case STATE_OK:
870 xasprintf(&result_line, _("NTP OK:"));
871 break;
872 default:
873 xasprintf(&result_line, _("NTP UNKNOWN:"));
874 break;
875 }
876 if (offset_result == STATE_UNKNOWN) {
877 xasprintf(&result_line, "%s %s", result_line, _("Offset unknown"));
878 xasprintf(&perfdata_line, "");
879 } else {
880 xasprintf(&result_line, "%s %s %.10g secs", result_line, _("Offset"), offset);
881 xasprintf(&perfdata_line, "%s", perfd_offset(offset));
882 }
883 if (do_jitter) {
884 xasprintf(&result_line, "%s, jitter=%f", result_line, jitter);
885 xasprintf(&perfdata_line, "%s %s", perfdata_line, perfd_jitter(jitter));
886 }
887 printf("%s|%s\n", result_line, perfdata_line);
888
889 if (server_address != NULL) {
890 free(server_address);
891 }
892 return result;
893}
894
895void print_help(void) {
896 print_revision(progname, NP_VERSION);
897
898 printf("Copyright (c) 2006 Sean Finney\n");
899 printf(COPYRIGHT, copyright, email);
900
901 printf("%s\n", _("This plugin checks the selected ntp server"));
902
903 printf("\n\n");
904
905 print_usage();
906 printf(UT_HELP_VRSN);
907 printf(UT_EXTRA_OPTS);
908 printf(UT_HOST_PORT, 'p', "123");
909 printf(UT_IPv46);
910 printf(" %s\n", "-w, --warning=THRESHOLD");
911 printf(" %s\n", _("Offset to result in warning status (seconds)"));
912 printf(" %s\n", "-c, --critical=THRESHOLD");
913 printf(" %s\n", _("Offset to result in critical status (seconds)"));
914 printf(" %s\n", "-j, --jwarn=THRESHOLD");
915 printf(" %s\n", _("Warning threshold for jitter"));
916 printf(" %s\n", "-k, --jcrit=THRESHOLD");
917 printf(" %s\n", _("Critical threshold for jitter"));
918 printf(UT_CONN_TIMEOUT, DEFAULT_SOCKET_TIMEOUT);
919 printf(UT_VERBOSE);
920
921 printf("\n");
922 printf("%s\n", _("Notes:"));
923 printf(UT_THRESHOLDS_NOTES);
924
925 printf("\n");
926 printf("%s\n", _("Examples:"));
927 printf(" %s\n", _("Normal offset check:"));
928 printf(" %s\n", ("./check_ntp -H ntpserv -w 0.5 -c 1"));
929 printf("\n");
930 printf(" %s\n",
931 _("Check jitter too, avoiding critical notifications if jitter isn't available"));
932 printf(" %s\n", _("(See Notes above for more details on thresholds formats):"));
933 printf(" %s\n", ("./check_ntp -H ntpserv -w 0.5 -c 1 -j -1:100 -k -1:200"));
934
935 printf(UT_SUPPORT);
936
937 printf("%s\n", _("WARNING: check_ntp is deprecated. Please use check_ntp_peer or"));
938 printf("%s\n\n", _("check_ntp_time instead."));
939}
940
941void print_usage(void) {
942 printf("%s\n", _("WARNING: check_ntp is deprecated. Please use check_ntp_peer or"));
943 printf("%s\n\n", _("check_ntp_time instead."));
944 printf("%s\n", _("Usage:"));
945 printf(" %s -H <host> [-w <warn>] [-c <crit>] [-j <warn>] [-k <crit>] [-4|-6] [-v verbose]\n",
946 progname);
947}