2854161687cbce9945b9d8e7a3f22a24ab15b390
1 /******************************************************************************
2 *
3 * Nagios check_ntp plugin
4 *
5 * License: GPL
6 * Copyright (c) 2006 sean finney <seanius@seanius.net>
7 * Copyright (c) 2007 nagios-plugins team
8 *
9 * Last Modified: $Date$
10 *
11 * Description:
12 *
13 * This file contains the check_ntp plugin
14 *
15 * This plugin to check ntp servers independant of any commandline
16 * programs or external libraries.
17 *
18 *
19 * License Information:
20 *
21 * This program is free software; you can redistribute it and/or modify
22 * it under the terms of the GNU General Public License as published by
23 * the Free Software Foundation; either version 2 of the License, or
24 * (at your option) any later version.
25 *
26 * This program is distributed in the hope that it will be useful,
27 * but WITHOUT ANY WARRANTY; without even the implied warranty of
28 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29 * GNU General Public License for more details.
30 *
31 * You should have received a copy of the GNU General Public License
32 * along with this program; if not, write to the Free Software
33 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
35 $Id$
37 *****************************************************************************/
39 const char *progname = "check_ntp";
40 const char *revision = "$Revision$";
41 const char *copyright = "2007";
42 const char *email = "nagiosplug-devel@lists.sourceforge.net";
44 #include "common.h"
45 #include "netutils.h"
46 #include "utils.h"
48 static char *server_address=NULL;
49 static int verbose=0;
50 static short do_offset=0;
51 static char *owarn="60";
52 static char *ocrit="120";
53 static short do_stratum=0;
54 static char *swarn="16";
55 static char *scrit="16";
56 static short do_jitter=0;
57 static char *jwarn="5000";
58 static char *jcrit="10000";
60 int process_arguments (int, char **);
61 thresholds *offset_thresholds = NULL;
62 thresholds *jitter_thresholds = NULL;
63 thresholds *stratum_thresholds = NULL;
64 void print_help (void);
65 void print_usage (void);
67 /* number of times to perform each request to get a good average. */
68 #define AVG_NUM 4
70 /* max size of control message data */
71 #define MAX_CM_SIZE 468
73 /* this structure holds everything in an ntp request/response as per rfc1305 */
74 typedef struct {
75 uint8_t flags; /* byte with leapindicator,vers,mode. see macros */
76 uint8_t stratum; /* clock stratum */
77 int8_t poll; /* polling interval */
78 int8_t precision; /* precision of the local clock */
79 int32_t rtdelay; /* total rt delay, as a fixed point num. see macros */
80 uint32_t rtdisp; /* like above, but for max err to primary src */
81 uint32_t refid; /* ref clock identifier */
82 uint64_t refts; /* reference timestamp. local time local clock */
83 uint64_t origts; /* time at which request departed client */
84 uint64_t rxts; /* time at which request arrived at server */
85 uint64_t txts; /* time at which request departed server */
86 } ntp_message;
88 /* this structure holds data about results from querying offset from a peer */
89 typedef struct {
90 time_t waiting; /* ts set when we started waiting for a response */
91 int num_responses; /* number of successfully recieved responses */
92 uint8_t stratum; /* copied verbatim from the ntp_message */
93 double rtdelay; /* converted from the ntp_message */
94 double rtdisp; /* converted from the ntp_message */
95 double offset[AVG_NUM]; /* offsets from each response */
96 uint8_t flags; /* byte with leapindicator,vers,mode. see macros */
97 } ntp_server_results;
99 /* this structure holds everything in an ntp control message as per rfc1305 */
100 typedef struct {
101 uint8_t flags; /* byte with leapindicator,vers,mode. see macros */
102 uint8_t op; /* R,E,M bits and Opcode */
103 uint16_t seq; /* Packet sequence */
104 uint16_t status; /* Clock status */
105 uint16_t assoc; /* Association */
106 uint16_t offset; /* Similar to TCP sequence # */
107 uint16_t count; /* # bytes of data */
108 char data[MAX_CM_SIZE]; /* ASCII data of the request */
109 /* NB: not necessarily NULL terminated! */
110 } ntp_control_message;
112 /* this is an association/status-word pair found in control packet reponses */
113 typedef struct {
114 uint16_t assoc;
115 uint16_t status;
116 } ntp_assoc_status_pair;
118 /* bits 1,2 are the leap indicator */
119 #define LI_MASK 0xc0
120 #define LI(x) ((x&LI_MASK)>>6)
121 #define LI_SET(x,y) do{ x |= ((y<<6)&LI_MASK); }while(0)
122 /* and these are the values of the leap indicator */
123 #define LI_NOWARNING 0x00
124 #define LI_EXTRASEC 0x01
125 #define LI_MISSINGSEC 0x02
126 #define LI_ALARM 0x03
127 /* bits 3,4,5 are the ntp version */
128 #define VN_MASK 0x38
129 #define VN(x) ((x&VN_MASK)>>3)
130 #define VN_SET(x,y) do{ x |= ((y<<3)&VN_MASK); }while(0)
131 #define VN_RESERVED 0x02
132 /* bits 6,7,8 are the ntp mode */
133 #define MODE_MASK 0x07
134 #define MODE(x) (x&MODE_MASK)
135 #define MODE_SET(x,y) do{ x |= (y&MODE_MASK); }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) do{ x |= (y&OP_MASK); }while(0)
147 #define OP_READSTAT 0x01
148 #define OP_READVAR 0x02
149 /* In peer status bytes, bits 6,7,8 determine clock selection status */
150 #define PEER_SEL(x) ((ntohs(x)>>8)&0x07)
151 #define PEER_INCLUDED 0x04
152 #define PEER_SYNCSOURCE 0x06
154 /**
155 ** a note about the 32-bit "fixed point" numbers:
156 **
157 they are divided into halves, each being a 16-bit int in network byte order:
158 - the first 16 bits are an int on the left side of a decimal point.
159 - the second 16 bits represent a fraction n/(2^16)
160 likewise for the 64-bit "fixed point" numbers with everything doubled :)
161 **/
163 /* macros to access the left/right 16 bits of a 32-bit ntp "fixed point"
164 number. note that these can be used as lvalues too */
165 #define L16(x) (((uint16_t*)&x)[0])
166 #define R16(x) (((uint16_t*)&x)[1])
167 /* macros to access the left/right 32 bits of a 64-bit ntp "fixed point"
168 number. these too can be used as lvalues */
169 #define L32(x) (((uint32_t*)&x)[0])
170 #define R32(x) (((uint32_t*)&x)[1])
172 /* ntp wants seconds since 1/1/00, epoch is 1/1/70. this is the difference */
173 #define EPOCHDIFF 0x83aa7e80UL
175 /* extract a 32-bit ntp fixed point number into a double */
176 #define NTP32asDOUBLE(x) (ntohs(L16(x)) + (double)ntohs(R16(x))/65536.0)
178 /* likewise for a 64-bit ntp fp number */
179 #define NTP64asDOUBLE(n) (double)(((uint64_t)n)?\
180 (ntohl(L32(n))-EPOCHDIFF) + \
181 (.00000001*(0.5+(double)(ntohl(R32(n))/42.94967296))):\
182 0)
184 /* convert a struct timeval to a double */
185 #define TVasDOUBLE(x) (double)(x.tv_sec+(0.000001*x.tv_usec))
187 /* convert an ntp 64-bit fp number to a struct timeval */
188 #define NTP64toTV(n,t) \
189 do{ if(!n) t.tv_sec = t.tv_usec = 0; \
190 else { \
191 t.tv_sec=ntohl(L32(n))-EPOCHDIFF; \
192 t.tv_usec=(int)(0.5+(double)(ntohl(R32(n))/4294.967296)); \
193 } \
194 }while(0)
196 /* convert a struct timeval to an ntp 64-bit fp number */
197 #define TVtoNTP64(t,n) \
198 do{ if(!t.tv_usec && !t.tv_sec) n=0x0UL; \
199 else { \
200 L32(n)=htonl(t.tv_sec + EPOCHDIFF); \
201 R32(n)=htonl((uint64_t)((4294.967296*t.tv_usec)+.5)); \
202 } \
203 } while(0)
205 /* NTP control message header is 12 bytes, plus any data in the data
206 * field, plus null padding to the nearest 32-bit boundary per rfc.
207 */
208 #define SIZEOF_NTPCM(m) (12+ntohs(m.count)+((m.count)?4-(ntohs(m.count)%4):0))
210 /* finally, a little helper or two for debugging: */
211 #define DBG(x) do{if(verbose>1){ x; }}while(0);
212 #define PRINTSOCKADDR(x) \
213 do{ \
214 printf("%u.%u.%u.%u", (x>>24)&0xff, (x>>16)&0xff, (x>>8)&0xff, x&0xff);\
215 }while(0);
217 /* calculate the offset of the local clock */
218 static inline double calc_offset(const ntp_message *m, const struct timeval *t){
219 double client_tx, peer_rx, peer_tx, client_rx;
220 client_tx = NTP64asDOUBLE(m->origts);
221 peer_rx = NTP64asDOUBLE(m->rxts);
222 peer_tx = NTP64asDOUBLE(m->txts);
223 client_rx=TVasDOUBLE((*t));
224 return (.5*((peer_tx-client_rx)+(peer_rx-client_tx)));
225 }
227 /* print out a ntp packet in human readable/debuggable format */
228 void print_ntp_message(const ntp_message *p){
229 struct timeval ref, orig, rx, tx;
231 NTP64toTV(p->refts,ref);
232 NTP64toTV(p->origts,orig);
233 NTP64toTV(p->rxts,rx);
234 NTP64toTV(p->txts,tx);
236 printf("packet contents:\n");
237 printf("\tflags: 0x%.2x\n", p->flags);
238 printf("\t li=%d (0x%.2x)\n", LI(p->flags), p->flags&LI_MASK);
239 printf("\t vn=%d (0x%.2x)\n", VN(p->flags), p->flags&VN_MASK);
240 printf("\t mode=%d (0x%.2x)\n", MODE(p->flags), p->flags&MODE_MASK);
241 printf("\tstratum = %d\n", p->stratum);
242 printf("\tpoll = %g\n", pow(2, p->poll));
243 printf("\tprecision = %g\n", pow(2, p->precision));
244 printf("\trtdelay = %-.16g\n", NTP32asDOUBLE(p->rtdelay));
245 printf("\trtdisp = %-.16g\n", NTP32asDOUBLE(p->rtdisp));
246 printf("\trefid = %x\n", p->refid);
247 printf("\trefts = %-.16g\n", NTP64asDOUBLE(p->refts));
248 printf("\torigts = %-.16g\n", NTP64asDOUBLE(p->origts));
249 printf("\trxts = %-.16g\n", NTP64asDOUBLE(p->rxts));
250 printf("\ttxts = %-.16g\n", NTP64asDOUBLE(p->txts));
251 }
253 void print_ntp_control_message(const ntp_control_message *p){
254 int i=0, numpeers=0;
255 const ntp_assoc_status_pair *peer=NULL;
257 printf("control packet contents:\n");
258 printf("\tflags: 0x%.2x , 0x%.2x\n", p->flags, p->op);
259 printf("\t li=%d (0x%.2x)\n", LI(p->flags), p->flags&LI_MASK);
260 printf("\t vn=%d (0x%.2x)\n", VN(p->flags), p->flags&VN_MASK);
261 printf("\t mode=%d (0x%.2x)\n", MODE(p->flags), p->flags&MODE_MASK);
262 printf("\t response=%d (0x%.2x)\n", (p->op&REM_RESP)>0, p->op&REM_RESP);
263 printf("\t more=%d (0x%.2x)\n", (p->op&REM_MORE)>0, p->op&REM_MORE);
264 printf("\t error=%d (0x%.2x)\n", (p->op&REM_ERROR)>0, p->op&REM_ERROR);
265 printf("\t op=%d (0x%.2x)\n", p->op&OP_MASK, p->op&OP_MASK);
266 printf("\tsequence: %d (0x%.2x)\n", ntohs(p->seq), ntohs(p->seq));
267 printf("\tstatus: %d (0x%.2x)\n", ntohs(p->status), ntohs(p->status));
268 printf("\tassoc: %d (0x%.2x)\n", ntohs(p->assoc), ntohs(p->assoc));
269 printf("\toffset: %d (0x%.2x)\n", ntohs(p->offset), ntohs(p->offset));
270 printf("\tcount: %d (0x%.2x)\n", ntohs(p->count), ntohs(p->count));
271 numpeers=ntohs(p->count)/(sizeof(ntp_assoc_status_pair));
272 if(p->op&REM_RESP && p->op&OP_READSTAT){
273 peer=(ntp_assoc_status_pair*)p->data;
274 for(i=0;i<numpeers;i++){
275 printf("\tpeer id %.2x status %.2x",
276 ntohs(peer[i].assoc), ntohs(peer[i].status));
277 if (PEER_SEL(peer[i].status) >= PEER_INCLUDED){
278 if(PEER_SEL(peer[i].status) >= PEER_SYNCSOURCE){
279 printf(" <-- current sync source");
280 } else {
281 printf(" <-- current sync candidate");
282 }
283 }
284 printf("\n");
285 }
286 }
287 }
289 void setup_request(ntp_message *p){
290 struct timeval t;
292 memset(p, 0, sizeof(ntp_message));
293 LI_SET(p->flags, LI_ALARM);
294 VN_SET(p->flags, 4);
295 MODE_SET(p->flags, MODE_CLIENT);
296 p->poll=4;
297 p->precision=(int8_t)0xfa;
298 L16(p->rtdelay)=htons(1);
299 L16(p->rtdisp)=htons(1);
301 gettimeofday(&t, NULL);
302 TVtoNTP64(t,p->txts);
303 }
305 /* select the "best" server from a list of servers, and return its index.
306 * this is done by filtering servers based on stratum, dispersion, and
307 * finally round-trip delay. */
308 int best_offset_server(const ntp_server_results *slist, int nservers){
309 int i=0, j=0, cserver=0, candidates[5], csize=0;
311 /* for each server */
312 for(cserver=0; cserver<nservers; cserver++){
313 /* sort out servers with error flags */
314 if ( LI(slist[cserver].flags) != LI_NOWARNING ){
315 if (verbose) printf("discarding peer id %d: flags=%d\n", cserver, LI(slist[cserver].flags));
316 break;
317 }
319 /* compare it to each of the servers already in the candidate list */
320 for(i=0; i<csize; i++){
321 /* does it have an equal or better stratum? */
322 if(slist[cserver].stratum <= slist[i].stratum){
323 /* does it have an equal or better dispersion? */
324 if(slist[cserver].rtdisp <= slist[i].rtdisp){
325 /* does it have a better rtdelay? */
326 if(slist[cserver].rtdelay < slist[i].rtdelay){
327 break;
328 }
329 }
330 }
331 }
333 /* if we haven't reached the current list's end, move everyone
334 * over one to the right, and insert the new candidate */
335 if(i<csize){
336 for(j=5; j>i; j--){
337 candidates[j]=candidates[j-1];
338 }
339 }
340 /* regardless, if they should be on the list... */
341 if(i<5) {
342 candidates[i]=cserver;
343 if(csize<5) csize++;
344 /* otherwise discard the server */
345 } else {
346 DBG(printf("discarding peer id %d\n", cserver));
347 }
348 }
350 if(csize>0) {
351 DBG(printf("best server selected: peer %d\n", candidates[0]));
352 return candidates[0];
353 } else {
354 DBG(printf("no peers meeting synchronization criteria :(\n"));
355 return -1;
356 }
357 }
359 /* do everything we need to get the total average offset
360 * - we use a certain amount of parallelization with poll() to ensure
361 * we don't waste time sitting around waiting for single packets.
362 * - we also "manually" handle resolving host names and connecting, because
363 * we have to do it in a way that our lazy macros don't handle currently :( */
364 double offset_request(const char *host, int *stratum, int *status){
365 int i=0, j=0, ga_result=0, num_hosts=0, *socklist=NULL, respnum=0;
366 int servers_completed=0, one_written=0, one_read=0, servers_readable=0, best_index=-1;
367 time_t now_time=0, start_ts=0;
368 ntp_message *req=NULL;
369 double avg_offset=0.;
370 struct timeval recv_time;
371 struct addrinfo *ai=NULL, *ai_tmp=NULL, hints;
372 struct pollfd *ufds=NULL;
373 ntp_server_results *servers=NULL;
375 /* setup hints to only return results from getaddrinfo that we'd like */
376 memset(&hints, 0, sizeof(struct addrinfo));
377 hints.ai_family = address_family;
378 hints.ai_protocol = IPPROTO_UDP;
379 hints.ai_socktype = SOCK_DGRAM;
381 /* fill in ai with the list of hosts resolved by the host name */
382 ga_result = getaddrinfo(host, "123", &hints, &ai);
383 if(ga_result!=0){
384 die(STATE_UNKNOWN, "error getting address for %s: %s\n",
385 host, gai_strerror(ga_result));
386 }
388 /* count the number of returned hosts, and allocate stuff accordingly */
389 for(ai_tmp=ai; ai_tmp!=NULL; ai_tmp=ai_tmp->ai_next){ num_hosts++; }
390 req=(ntp_message*)malloc(sizeof(ntp_message)*num_hosts);
391 if(req==NULL) die(STATE_UNKNOWN, "can not allocate ntp message array");
392 socklist=(int*)malloc(sizeof(int)*num_hosts);
393 if(socklist==NULL) die(STATE_UNKNOWN, "can not allocate socket array");
394 ufds=(struct pollfd*)malloc(sizeof(struct pollfd)*num_hosts);
395 if(ufds==NULL) die(STATE_UNKNOWN, "can not allocate socket array");
396 servers=(ntp_server_results*)malloc(sizeof(ntp_server_results)*num_hosts);
397 if(servers==NULL) die(STATE_UNKNOWN, "can not allocate server array");
398 memset(servers, 0, sizeof(ntp_server_results)*num_hosts);
400 /* setup each socket for writing, and the corresponding struct pollfd */
401 ai_tmp=ai;
402 for(i=0;ai_tmp;i++){
403 socklist[i]=socket(ai_tmp->ai_family, SOCK_DGRAM, IPPROTO_UDP);
404 if(socklist[i] == -1) {
405 perror(NULL);
406 die(STATE_UNKNOWN, "can not create new socket");
407 }
408 if(connect(socklist[i], ai_tmp->ai_addr, ai_tmp->ai_addrlen)){
409 die(STATE_UNKNOWN, "can't create socket connection");
410 } else {
411 ufds[i].fd=socklist[i];
412 ufds[i].events=POLLIN;
413 ufds[i].revents=0;
414 }
415 ai_tmp = ai_tmp->ai_next;
416 }
418 /* now do AVG_NUM checks to each host. we stop before timeout/2 seconds
419 * have passed in order to ensure post-processing and jitter time. */
420 now_time=start_ts=time(NULL);
421 while(servers_completed<num_hosts && now_time-start_ts <= socket_timeout/2){
422 /* loop through each server and find each one which hasn't
423 * been touched in the past second or so and is still lacking
424 * some responses. for each of these servers, send a new request,
425 * and update the "waiting" timestamp with the current time. */
426 one_written=0;
427 now_time=time(NULL);
429 for(i=0; i<num_hosts; i++){
430 if(servers[i].waiting<now_time && servers[i].num_responses<AVG_NUM){
431 if(verbose && servers[i].waiting != 0) printf("re-");
432 if(verbose) printf("sending request to peer %d\n", i);
433 setup_request(&req[i]);
434 write(socklist[i], &req[i], sizeof(ntp_message));
435 servers[i].waiting=now_time;
436 one_written=1;
437 break;
438 }
439 }
441 /* quickly poll for any sockets with pending data */
442 servers_readable=poll(ufds, num_hosts, 100);
443 if(servers_readable==-1){
444 perror("polling ntp sockets");
445 die(STATE_UNKNOWN, "communication errors");
446 }
448 /* read from any sockets with pending data */
449 for(i=0; servers_readable && i<num_hosts; i++){
450 if(ufds[i].revents&POLLIN && servers[i].num_responses < AVG_NUM){
451 if(verbose) {
452 printf("response from peer %d: ", i);
453 }
455 read(ufds[i].fd, &req[i], sizeof(ntp_message));
456 gettimeofday(&recv_time, NULL);
457 DBG(print_ntp_message(&req[i]));
458 respnum=servers[i].num_responses++;
459 servers[i].offset[respnum]=calc_offset(&req[i], &recv_time);
460 if(verbose) {
461 printf("offset %.10g, stratum %i\n", servers[i].offset[respnum], req[i].stratum);
462 }
463 servers[i].stratum=req[i].stratum;
464 servers[i].rtdisp=NTP32asDOUBLE(req[i].rtdisp);
465 servers[i].rtdelay=NTP32asDOUBLE(req[i].rtdelay);
466 servers[i].waiting=0;
467 servers[i].flags=req[i].flags;
468 servers_readable--;
469 one_read = 1;
470 if(servers[i].num_responses==AVG_NUM) servers_completed++;
471 }
472 }
473 /* lather, rinse, repeat. */
474 }
476 if (one_read == 0) {
477 die(STATE_CRITICAL, "NTP CRITICAL: No response from NTP server\n");
478 }
480 /* now, pick the best server from the list */
481 best_index=best_offset_server(servers, num_hosts);
482 if(best_index < 0){
483 *status=STATE_UNKNOWN;
484 } else {
485 /* finally, calculate the average offset */
486 for(i=0; i<servers[best_index].num_responses;i++){
487 avg_offset+=servers[best_index].offset[j];
488 }
489 avg_offset/=servers[best_index].num_responses;
490 /* Stratum sent in normal packets is ingreased by 1 (i.e. stratum that
491 * would be displayed if we were a server) so we decrease it */
492 *stratum = servers[best_index].stratum - 1;
493 }
495 /* cleanup */
496 /* FIXME: Not closing the socket to avoid re-use of the local port
497 * which can cause old NTP packets to be read instead of NTP control
498 * pactets in jitter_request(). THERE MUST BE ANOTHER WAY...
499 * for(j=0; j<num_hosts; j++){ close(socklist[j]); } */
500 free(socklist);
501 free(ufds);
502 free(servers);
503 free(req);
504 freeaddrinfo(ai);
506 if(verbose) printf("overall average offset: %.10g\n", avg_offset);
507 return avg_offset;
508 }
510 void
511 setup_control_request(ntp_control_message *p, uint8_t opcode, uint16_t seq){
512 memset(p, 0, sizeof(ntp_control_message));
513 LI_SET(p->flags, LI_NOWARNING);
514 VN_SET(p->flags, VN_RESERVED);
515 MODE_SET(p->flags, MODE_CONTROLMSG);
516 OP_SET(p->op, opcode);
517 p->seq = htons(seq);
518 /* Remaining fields are zero for requests */
519 }
521 /* XXX handle responses with the error bit set */
522 double jitter_request(const char *host, int *status){
523 int conn=-1, i, npeers=0, num_candidates=0, syncsource_found=0;
524 int run=0, min_peer_sel=PEER_INCLUDED, num_selected=0, num_valid=0;
525 int peers_size=0, peer_offset=0;
526 ntp_assoc_status_pair *peers=NULL;
527 ntp_control_message req;
528 const char *getvar = "jitter";
529 double rval = 0.0, jitter = -1.0;
530 char *startofvalue=NULL, *nptr=NULL;
531 void *tmp;
533 /* Long-winded explanation:
534 * Getting the jitter requires a number of steps:
535 * 1) Send a READSTAT request.
536 * 2) Interpret the READSTAT reply
537 * a) The data section contains a list of peer identifiers (16 bits)
538 * and associated status words (16 bits)
539 * b) We want the value of 0x06 in the SEL (peer selection) value,
540 * which means "current synchronizatin source". If that's missing,
541 * we take anything better than 0x04 (see the rfc for details) but
542 * set a minimum of warning.
543 * 3) Send a READVAR request for information on each peer identified
544 * in 2b greater than the minimum selection value.
545 * 4) Extract the jitter value from the data[] (it's ASCII)
546 */
547 my_udp_connect(server_address, 123, &conn);
549 /* keep sending requests until the server stops setting the
550 * REM_MORE bit, though usually this is only 1 packet. */
551 do{
552 setup_control_request(&req, OP_READSTAT, 1);
553 DBG(printf("sending READSTAT request"));
554 write(conn, &req, SIZEOF_NTPCM(req));
555 DBG(print_ntp_control_message(&req));
556 /* Attempt to read the largest size packet possible */
557 req.count=htons(MAX_CM_SIZE);
558 DBG(printf("recieving READSTAT response"))
559 read(conn, &req, SIZEOF_NTPCM(req));
560 DBG(print_ntp_control_message(&req));
561 /* Each peer identifier is 4 bytes in the data section, which
562 * we represent as a ntp_assoc_status_pair datatype.
563 */
564 peers_size+=ntohs(req.count);
565 if((tmp=realloc(peers, peers_size)) == NULL)
566 free(peers), die(STATE_UNKNOWN, "can not (re)allocate 'peers' buffer\n");
567 peers=tmp;
568 memcpy((void*)((ptrdiff_t)peers+peer_offset), (void*)req.data, ntohs(req.count));
569 npeers=peers_size/sizeof(ntp_assoc_status_pair);
570 peer_offset+=ntohs(req.count);
571 } while(req.op&REM_MORE);
573 /* first, let's find out if we have a sync source, or if there are
574 * at least some candidates. in the case of the latter we'll issue
575 * a warning but go ahead with the check on them. */
576 for (i = 0; i < npeers; i++){
577 if (PEER_SEL(peers[i].status) >= PEER_INCLUDED){
578 num_candidates++;
579 if(PEER_SEL(peers[i].status) >= PEER_SYNCSOURCE){
580 syncsource_found=1;
581 min_peer_sel=PEER_SYNCSOURCE;
582 }
583 }
584 }
585 if(verbose) printf("%d candiate peers available\n", num_candidates);
586 if(verbose && syncsource_found) printf("synchronization source found\n");
587 if(! syncsource_found){
588 *status = STATE_UNKNOWN;
589 if(verbose) printf("warning: no synchronization source found\n");
590 }
593 for (run=0; run<AVG_NUM; run++){
594 if(verbose) printf("jitter run %d of %d\n", run+1, AVG_NUM);
595 for (i = 0; i < npeers; i++){
596 /* Only query this server if it is the current sync source */
597 if (PEER_SEL(peers[i].status) >= min_peer_sel){
598 num_selected++;
599 setup_control_request(&req, OP_READVAR, 2);
600 req.assoc = peers[i].assoc;
601 /* By spec, putting the variable name "jitter" in the request
602 * should cause the server to provide _only_ the jitter value.
603 * thus reducing net traffic, guaranteeing us only a single
604 * datagram in reply, and making intepretation much simpler
605 */
606 /* Older servers doesn't know what jitter is, so if we get an
607 * error on the first pass we redo it with "dispersion" */
608 strncpy(req.data, getvar, MAX_CM_SIZE-1);
609 req.count = htons(strlen(getvar));
610 DBG(printf("sending READVAR request...\n"));
611 write(conn, &req, SIZEOF_NTPCM(req));
612 DBG(print_ntp_control_message(&req));
614 req.count = htons(MAX_CM_SIZE);
615 DBG(printf("recieving READVAR response...\n"));
616 read(conn, &req, SIZEOF_NTPCM(req));
617 DBG(print_ntp_control_message(&req));
619 if(req.op&REM_ERROR && strstr(getvar, "jitter")) {
620 if(verbose) printf("The 'jitter' command failed (old ntp server?)\nRestarting with 'dispersion'...\n");
621 getvar = "dispersion";
622 num_selected--;
623 i--;
624 continue;
625 }
627 /* get to the float value */
628 if(verbose) {
629 printf("parsing jitter from peer %.2x: ", ntohs(peers[i].assoc));
630 }
631 startofvalue = strchr(req.data, '=');
632 if(startofvalue != NULL) {
633 startofvalue++;
634 jitter = strtod(startofvalue, &nptr);
635 }
636 if(startofvalue == NULL || startofvalue==nptr){
637 printf("warning: unable to read server jitter response.\n");
638 *status = STATE_UNKNOWN;
639 } else {
640 if(verbose) printf("%g\n", jitter);
641 num_valid++;
642 rval += jitter;
643 }
644 }
645 }
646 if(verbose){
647 printf("jitter parsed from %d/%d peers\n", num_valid, num_selected);
648 }
649 }
651 rval = num_valid ? rval / num_valid : -1.0;
653 close(conn);
654 if(peers!=NULL) free(peers);
655 /* If we return -1.0, it means no synchronization source was found */
656 return rval;
657 }
659 int process_arguments(int argc, char **argv){
660 int c;
661 int option=0;
662 static struct option longopts[] = {
663 {"version", no_argument, 0, 'V'},
664 {"help", no_argument, 0, 'h'},
665 {"verbose", no_argument, 0, 'v'},
666 {"use-ipv4", no_argument, 0, '4'},
667 {"use-ipv6", no_argument, 0, '6'},
668 {"warning", required_argument, 0, 'w'},
669 {"critical", required_argument, 0, 'c'},
670 {"swarn", required_argument, 0, 'W'},
671 {"scrit", required_argument, 0, 'C'},
672 {"jwarn", required_argument, 0, 'j'},
673 {"jcrit", required_argument, 0, 'k'},
674 {"timeout", required_argument, 0, 't'},
675 {"hostname", required_argument, 0, 'H'},
676 {0, 0, 0, 0}
677 };
680 if (argc < 2)
681 usage ("\n");
683 while (1) {
684 c = getopt_long (argc, argv, "Vhv46w:c:W:C:j:k:t:H:", longopts, &option);
685 if (c == -1 || c == EOF || c == 1)
686 break;
688 switch (c) {
689 case 'h':
690 print_help();
691 exit(STATE_OK);
692 break;
693 case 'V':
694 print_revision(progname, revision);
695 exit(STATE_OK);
696 break;
697 case 'v':
698 verbose++;
699 break;
700 case 'w':
701 do_offset=1;
702 owarn = optarg;
703 break;
704 case 'c':
705 do_offset=1;
706 ocrit = optarg;
707 break;
708 case 'W':
709 do_stratum=1;
710 swarn = optarg;
711 break;
712 case 'C':
713 do_stratum=1;
714 scrit = optarg;
715 break;
716 case 'j':
717 do_jitter=1;
718 jwarn = optarg;
719 break;
720 case 'k':
721 do_jitter=1;
722 jcrit = optarg;
723 break;
724 case 'H':
725 if(is_host(optarg) == FALSE)
726 usage2(_("Invalid hostname/address"), optarg);
727 server_address = strdup(optarg);
728 break;
729 case 't':
730 socket_timeout=atoi(optarg);
731 break;
732 case '4':
733 address_family = AF_INET;
734 break;
735 case '6':
736 #ifdef USE_IPV6
737 address_family = AF_INET6;
738 #else
739 usage4 (_("IPv6 support not available"));
740 #endif
741 break;
742 case '?':
743 /* print short usage statement if args not parsable */
744 usage5 ();
745 break;
746 }
747 }
749 if(server_address == NULL){
750 usage4(_("Hostname was not supplied"));
751 }
753 return 0;
754 }
756 char *perfd_offset (double offset)
757 {
758 return fperfdata ("offset", offset, "s",
759 TRUE, offset_thresholds->warning->end,
760 TRUE, offset_thresholds->critical->end,
761 FALSE, 0, FALSE, 0);
762 }
764 char *perfd_jitter (double jitter)
765 {
766 return fperfdata ("jitter", jitter, "s",
767 do_jitter, jitter_thresholds->warning->end,
768 do_jitter, jitter_thresholds->critical->end,
769 TRUE, 0, FALSE, 0);
770 }
772 char *perfd_stratum (int stratum)
773 {
774 return perfdata ("stratum", stratum, "",
775 do_stratum, (int)stratum_thresholds->warning->end,
776 do_stratum, (int)stratum_thresholds->critical->end,
777 TRUE, 0, TRUE, 16);
778 }
780 int main(int argc, char *argv[]){
781 int result, offset_result, jitter_result, stratum;
782 double offset=0, jitter=0;
783 char *result_line, *perfdata_line;
785 result = offset_result = jitter_result = STATE_OK;
787 if (process_arguments (argc, argv) == ERROR)
788 usage4 (_("Could not parse arguments"));
790 set_thresholds(&offset_thresholds, owarn, ocrit);
791 set_thresholds(&jitter_thresholds, jwarn, jcrit);
792 set_thresholds(&stratum_thresholds, swarn, scrit);
794 /* initialize alarm signal handling */
795 signal (SIGALRM, socket_timeout_alarm_handler);
797 /* set socket timeout */
798 alarm (socket_timeout);
800 offset = offset_request(server_address, &stratum, &offset_result);
801 if (do_offset && offset_result == STATE_UNKNOWN) {
802 result = STATE_CRITICAL;
803 } else {
804 result = get_status(fabs(offset), offset_thresholds);
805 }
806 result = max_state(result, offset_result);
807 if(do_stratum)
808 result = max_state(result, get_status(stratum, stratum_thresholds));
810 /* If not told to check the jitter, we don't even send packets.
811 * jitter is checked using NTP control packets, which not all
812 * servers recognize. Trying to check the jitter on OpenNTPD
813 * (for example) will result in an error
814 */
815 if(do_jitter){
816 jitter=jitter_request(server_address, &jitter_result);
817 result = max_state(result, get_status(jitter, jitter_thresholds));
818 /* -1 indicates that we couldn't calculate the jitter
819 * Only overrides STATE_OK from the offset */
820 if(jitter == -1.0 && result == STATE_OK)
821 result = STATE_UNKNOWN;
822 }
823 result = max_state(result, jitter_result);
825 switch (result) {
826 case STATE_CRITICAL :
827 asprintf(&result_line, "NTP CRITICAL:");
828 break;
829 case STATE_WARNING :
830 asprintf(&result_line, "NTP WARNING:");
831 break;
832 case STATE_OK :
833 asprintf(&result_line, "NTP OK:");
834 break;
835 default :
836 asprintf(&result_line, "NTP UNKNOWN:");
837 break;
838 }
839 if(offset_result == STATE_UNKNOWN){
840 asprintf(&result_line, "%s %s", result_line, _("Offset unknown"));
841 asprintf(&perfdata_line, "");
842 } else {
843 #if 0 /* 2007-10-25 This can't happen. Leftovers or uninplemented? */
844 if(offset_result==STATE_WARNING){
845 asprintf(&result_line, "%s %s", result_line, _("Unable to fully sample sync server"));
846 }
847 #endif
848 asprintf(&result_line, "%s Offset %.10g secs", result_line, offset);
849 asprintf(&perfdata_line, "%s", perfd_offset(offset));
850 }
851 if (do_jitter) {
852 asprintf(&result_line, "%s, jitter=%f", result_line, jitter);
853 asprintf(&perfdata_line, "%s %s", perfdata_line, perfd_jitter(jitter));
854 }
855 if (do_stratum) {
856 asprintf(&result_line, "%s, stratum=%i", result_line, stratum);
857 asprintf(&perfdata_line, "%s %s", perfdata_line, perfd_stratum(stratum));
858 }
859 printf("%s|%s\n", result_line, perfdata_line);
861 if(server_address!=NULL) free(server_address);
862 return result;
863 }
867 void print_help(void){
868 print_revision(progname, revision);
870 printf ("Copyright (c) 2006 Sean Finney\n");
871 printf (COPYRIGHT, copyright, email);
873 printf ("%s\n", _("This plugin checks the selected ntp server"));
875 printf ("\n\n");
877 print_usage();
878 printf (_(UT_HELP_VRSN));
879 printf (_(UT_HOST_PORT), 'p', "123");
880 printf (" %s\n", "-w, --warning=THRESHOLD");
881 printf (" %s\n", _("Offset to result in warning status (seconds)"));
882 printf (" %s\n", "-c, --critical=THRESHOLD");
883 printf (" %s\n", _("Offset to result in critical status (seconds)"));
884 printf (" %s\n", "-W, --warning=THRESHOLD");
885 printf (" %s\n", _("Warning threshold for stratum"));
886 printf (" %s\n", "-W, --critical=THRESHOLD");
887 printf (" %s\n", _("Critical threshold for stratum"));
888 printf (" %s\n", "-j, --warning=THRESHOLD");
889 printf (" %s\n", _("Warning threshold for jitter"));
890 printf (" %s\n", "-k, --critical=THRESHOLD");
891 printf (" %s\n", _("Critical threshold for jitter"));
892 printf (_(UT_TIMEOUT), DEFAULT_SOCKET_TIMEOUT);
893 printf (_(UT_VERBOSE));
895 printf("\n");
896 printf("%s\n", _("Notes:"));
897 printf(" %s\n", _("See:"));
898 printf(" %s\n", ("http://nagiosplug.sourceforge.net/developer-guidelines.html#THRESHOLDFORMAT"));
899 printf(" %s\n", _("for THRESHOLD format and examples."));
901 printf("\n");
902 printf("%s\n", _("Examples:"));
903 printf(" %s\n", _("Normal offset check:"));
904 printf(" %s\n", ("./check_ntp -H ntpserv -w 0.5 -c 1"));
905 printf(" %s\n", _("Check jitter too, avoiding critical notifications if jitter isn't available"));
906 printf(" %s\n", _("(See Notes above for more details on thresholds formats):"));
907 printf(" %s\n", ("./check_ntp -H ntpserv -w 0.5 -c 1 -j -1:100 -k -1:200"));
908 printf(" %s\n", _("Check only stratum:"));
909 printf(" %s\n", ("./check_ntp -H ntpserv -W 4 -C 6"));
911 printf (_(UT_SUPPORT));
912 }
914 void
915 print_usage(void)
916 {
917 printf (_("Usage:"));
918 printf(" %s -H <host> [-w <warn>] [-c <crit>] [-W <warn>] [-C <crit>]\n", progname);
919 printf(" [-j <warn>] [-k <crit>] [-v verbose]\n");
920 }