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 char *owarn="60";
51 static char *ocrit="120";
52 static short do_stratum=0;
53 static char *swarn="16";
54 static char *scrit="16";
55 static short do_jitter=0;
56 static char *jwarn="5000";
57 static char *jcrit="10000";
59 int process_arguments (int, char **);
60 thresholds *offset_thresholds = NULL;
61 thresholds *jitter_thresholds = NULL;
62 thresholds *stratum_thresholds = NULL;
63 void print_help (void);
64 void print_usage (void);
66 /* number of times to perform each request to get a good average. */
67 #define AVG_NUM 4
69 /* max size of control message data */
70 #define MAX_CM_SIZE 468
72 /* this structure holds everything in an ntp request/response as per rfc1305 */
73 typedef struct {
74 uint8_t flags; /* byte with leapindicator,vers,mode. see macros */
75 uint8_t stratum; /* clock stratum */
76 int8_t poll; /* polling interval */
77 int8_t precision; /* precision of the local clock */
78 int32_t rtdelay; /* total rt delay, as a fixed point num. see macros */
79 uint32_t rtdisp; /* like above, but for max err to primary src */
80 uint32_t refid; /* ref clock identifier */
81 uint64_t refts; /* reference timestamp. local time local clock */
82 uint64_t origts; /* time at which request departed client */
83 uint64_t rxts; /* time at which request arrived at server */
84 uint64_t txts; /* time at which request departed server */
85 } ntp_message;
87 /* this structure holds data about results from querying offset from a peer */
88 typedef struct {
89 time_t waiting; /* ts set when we started waiting for a response */
90 int num_responses; /* number of successfully recieved responses */
91 uint8_t stratum; /* copied verbatim from the ntp_message */
92 double rtdelay; /* converted from the ntp_message */
93 double rtdisp; /* converted from the ntp_message */
94 double offset[AVG_NUM]; /* offsets from each response */
95 uint8_t flags; /* byte with leapindicator,vers,mode. see macros */
96 } ntp_server_results;
98 /* this structure holds everything in an ntp control message as per rfc1305 */
99 typedef struct {
100 uint8_t flags; /* byte with leapindicator,vers,mode. see macros */
101 uint8_t op; /* R,E,M bits and Opcode */
102 uint16_t seq; /* Packet sequence */
103 uint16_t status; /* Clock status */
104 uint16_t assoc; /* Association */
105 uint16_t offset; /* Similar to TCP sequence # */
106 uint16_t count; /* # bytes of data */
107 char data[MAX_CM_SIZE]; /* ASCII data of the request */
108 /* NB: not necessarily NULL terminated! */
109 } ntp_control_message;
111 /* this is an association/status-word pair found in control packet reponses */
112 typedef struct {
113 uint16_t assoc;
114 uint16_t status;
115 } ntp_assoc_status_pair;
117 /* bits 1,2 are the leap indicator */
118 #define LI_MASK 0xc0
119 #define LI(x) ((x&LI_MASK)>>6)
120 #define LI_SET(x,y) do{ x |= ((y<<6)&LI_MASK); }while(0)
121 /* and these are the values of the leap indicator */
122 #define LI_NOWARNING 0x00
123 #define LI_EXTRASEC 0x01
124 #define LI_MISSINGSEC 0x02
125 #define LI_ALARM 0x03
126 /* bits 3,4,5 are the ntp version */
127 #define VN_MASK 0x38
128 #define VN(x) ((x&VN_MASK)>>3)
129 #define VN_SET(x,y) do{ x |= ((y<<3)&VN_MASK); }while(0)
130 #define VN_RESERVED 0x02
131 /* bits 6,7,8 are the ntp mode */
132 #define MODE_MASK 0x07
133 #define MODE(x) (x&MODE_MASK)
134 #define MODE_SET(x,y) do{ x |= (y&MODE_MASK); }while(0)
135 /* here are some values */
136 #define MODE_CLIENT 0x03
137 #define MODE_CONTROLMSG 0x06
138 /* In control message, bits 8-10 are R,E,M bits */
139 #define REM_MASK 0xe0
140 #define REM_RESP 0x80
141 #define REM_ERROR 0x40
142 #define REM_MORE 0x20
143 /* In control message, bits 11 - 15 are opcode */
144 #define OP_MASK 0x1f
145 #define OP_SET(x,y) do{ x |= (y&OP_MASK); }while(0)
146 #define OP_READSTAT 0x01
147 #define OP_READVAR 0x02
148 /* In peer status bytes, bits 6,7,8 determine clock selection status */
149 #define PEER_SEL(x) ((ntohs(x)>>8)&0x07)
150 #define PEER_INCLUDED 0x04
151 #define PEER_SYNCSOURCE 0x06
153 /**
154 ** a note about the 32-bit "fixed point" numbers:
155 **
156 they are divided into halves, each being a 16-bit int in network byte order:
157 - the first 16 bits are an int on the left side of a decimal point.
158 - the second 16 bits represent a fraction n/(2^16)
159 likewise for the 64-bit "fixed point" numbers with everything doubled :)
160 **/
162 /* macros to access the left/right 16 bits of a 32-bit ntp "fixed point"
163 number. note that these can be used as lvalues too */
164 #define L16(x) (((uint16_t*)&x)[0])
165 #define R16(x) (((uint16_t*)&x)[1])
166 /* macros to access the left/right 32 bits of a 64-bit ntp "fixed point"
167 number. these too can be used as lvalues */
168 #define L32(x) (((uint32_t*)&x)[0])
169 #define R32(x) (((uint32_t*)&x)[1])
171 /* ntp wants seconds since 1/1/00, epoch is 1/1/70. this is the difference */
172 #define EPOCHDIFF 0x83aa7e80UL
174 /* extract a 32-bit ntp fixed point number into a double */
175 #define NTP32asDOUBLE(x) (ntohs(L16(x)) + (double)ntohs(R16(x))/65536.0)
177 /* likewise for a 64-bit ntp fp number */
178 #define NTP64asDOUBLE(n) (double)(((uint64_t)n)?\
179 (ntohl(L32(n))-EPOCHDIFF) + \
180 (.00000001*(0.5+(double)(ntohl(R32(n))/42.94967296))):\
181 0)
183 /* convert a struct timeval to a double */
184 #define TVasDOUBLE(x) (double)(x.tv_sec+(0.000001*x.tv_usec))
186 /* convert an ntp 64-bit fp number to a struct timeval */
187 #define NTP64toTV(n,t) \
188 do{ if(!n) t.tv_sec = t.tv_usec = 0; \
189 else { \
190 t.tv_sec=ntohl(L32(n))-EPOCHDIFF; \
191 t.tv_usec=(int)(0.5+(double)(ntohl(R32(n))/4294.967296)); \
192 } \
193 }while(0)
195 /* convert a struct timeval to an ntp 64-bit fp number */
196 #define TVtoNTP64(t,n) \
197 do{ if(!t.tv_usec && !t.tv_sec) n=0x0UL; \
198 else { \
199 L32(n)=htonl(t.tv_sec + EPOCHDIFF); \
200 R32(n)=htonl((uint64_t)((4294.967296*t.tv_usec)+.5)); \
201 } \
202 } while(0)
204 /* NTP control message header is 12 bytes, plus any data in the data
205 * field, plus null padding to the nearest 32-bit boundary per rfc.
206 */
207 #define SIZEOF_NTPCM(m) (12+ntohs(m.count)+((m.count)?4-(ntohs(m.count)%4):0))
209 /* finally, a little helper or two for debugging: */
210 #define DBG(x) do{if(verbose>1){ x; }}while(0);
211 #define PRINTSOCKADDR(x) \
212 do{ \
213 printf("%u.%u.%u.%u", (x>>24)&0xff, (x>>16)&0xff, (x>>8)&0xff, x&0xff);\
214 }while(0);
216 /* calculate the offset of the local clock */
217 static inline double calc_offset(const ntp_message *m, const struct timeval *t){
218 double client_tx, peer_rx, peer_tx, client_rx;
219 client_tx = NTP64asDOUBLE(m->origts);
220 peer_rx = NTP64asDOUBLE(m->rxts);
221 peer_tx = NTP64asDOUBLE(m->txts);
222 client_rx=TVasDOUBLE((*t));
223 return (.5*((peer_tx-client_rx)+(peer_rx-client_tx)));
224 }
226 /* print out a ntp packet in human readable/debuggable format */
227 void print_ntp_message(const ntp_message *p){
228 struct timeval ref, orig, rx, tx;
230 NTP64toTV(p->refts,ref);
231 NTP64toTV(p->origts,orig);
232 NTP64toTV(p->rxts,rx);
233 NTP64toTV(p->txts,tx);
235 printf("packet contents:\n");
236 printf("\tflags: 0x%.2x\n", p->flags);
237 printf("\t li=%d (0x%.2x)\n", LI(p->flags), p->flags&LI_MASK);
238 printf("\t vn=%d (0x%.2x)\n", VN(p->flags), p->flags&VN_MASK);
239 printf("\t mode=%d (0x%.2x)\n", MODE(p->flags), p->flags&MODE_MASK);
240 printf("\tstratum = %d\n", p->stratum);
241 printf("\tpoll = %g\n", pow(2, p->poll));
242 printf("\tprecision = %g\n", pow(2, p->precision));
243 printf("\trtdelay = %-.16g\n", NTP32asDOUBLE(p->rtdelay));
244 printf("\trtdisp = %-.16g\n", NTP32asDOUBLE(p->rtdisp));
245 printf("\trefid = %x\n", p->refid);
246 printf("\trefts = %-.16g\n", NTP64asDOUBLE(p->refts));
247 printf("\torigts = %-.16g\n", NTP64asDOUBLE(p->origts));
248 printf("\trxts = %-.16g\n", NTP64asDOUBLE(p->rxts));
249 printf("\ttxts = %-.16g\n", NTP64asDOUBLE(p->txts));
250 }
252 void print_ntp_control_message(const ntp_control_message *p){
253 int i=0, numpeers=0;
254 const ntp_assoc_status_pair *peer=NULL;
256 printf("control packet contents:\n");
257 printf("\tflags: 0x%.2x , 0x%.2x\n", p->flags, p->op);
258 printf("\t li=%d (0x%.2x)\n", LI(p->flags), p->flags&LI_MASK);
259 printf("\t vn=%d (0x%.2x)\n", VN(p->flags), p->flags&VN_MASK);
260 printf("\t mode=%d (0x%.2x)\n", MODE(p->flags), p->flags&MODE_MASK);
261 printf("\t response=%d (0x%.2x)\n", (p->op&REM_RESP)>0, p->op&REM_RESP);
262 printf("\t more=%d (0x%.2x)\n", (p->op&REM_MORE)>0, p->op&REM_MORE);
263 printf("\t error=%d (0x%.2x)\n", (p->op&REM_ERROR)>0, p->op&REM_ERROR);
264 printf("\t op=%d (0x%.2x)\n", p->op&OP_MASK, p->op&OP_MASK);
265 printf("\tsequence: %d (0x%.2x)\n", ntohs(p->seq), ntohs(p->seq));
266 printf("\tstatus: %d (0x%.2x)\n", ntohs(p->status), ntohs(p->status));
267 printf("\tassoc: %d (0x%.2x)\n", ntohs(p->assoc), ntohs(p->assoc));
268 printf("\toffset: %d (0x%.2x)\n", ntohs(p->offset), ntohs(p->offset));
269 printf("\tcount: %d (0x%.2x)\n", ntohs(p->count), ntohs(p->count));
270 numpeers=ntohs(p->count)/(sizeof(ntp_assoc_status_pair));
271 if(p->op&REM_RESP && p->op&OP_READSTAT){
272 peer=(ntp_assoc_status_pair*)p->data;
273 for(i=0;i<numpeers;i++){
274 printf("\tpeer id %.2x status %.2x",
275 ntohs(peer[i].assoc), ntohs(peer[i].status));
276 if (PEER_SEL(peer[i].status) >= PEER_INCLUDED){
277 if(PEER_SEL(peer[i].status) >= PEER_SYNCSOURCE){
278 printf(" <-- current sync source");
279 } else {
280 printf(" <-- current sync candidate");
281 }
282 }
283 printf("\n");
284 }
285 }
286 }
288 void setup_request(ntp_message *p){
289 struct timeval t;
291 memset(p, 0, sizeof(ntp_message));
292 LI_SET(p->flags, LI_ALARM);
293 VN_SET(p->flags, 4);
294 MODE_SET(p->flags, MODE_CLIENT);
295 p->poll=4;
296 p->precision=(int8_t)0xfa;
297 L16(p->rtdelay)=htons(1);
298 L16(p->rtdisp)=htons(1);
300 gettimeofday(&t, NULL);
301 TVtoNTP64(t,p->txts);
302 }
304 /* select the "best" server from a list of servers, and return its index.
305 * this is done by filtering servers based on stratum, dispersion, and
306 * finally round-trip delay. */
307 int best_offset_server(const ntp_server_results *slist, int nservers){
308 int i=0, j=0, cserver=0, candidates[5], csize=0;
310 /* for each server */
311 for(cserver=0; cserver<nservers; cserver++){
312 /* sort out servers with error flags */
313 if ( LI(slist[cserver].flags) != LI_NOWARNING ){
314 if (verbose) printf("discarding peer id %d: flags=%d\n", cserver, LI(slist[cserver].flags));
315 break;
316 }
318 /* compare it to each of the servers already in the candidate list */
319 for(i=0; i<csize; i++){
320 /* does it have an equal or better stratum? */
321 if(slist[cserver].stratum <= slist[i].stratum){
322 /* does it have an equal or better dispersion? */
323 if(slist[cserver].rtdisp <= slist[i].rtdisp){
324 /* does it have a better rtdelay? */
325 if(slist[cserver].rtdelay < slist[i].rtdelay){
326 break;
327 }
328 }
329 }
330 }
332 /* if we haven't reached the current list's end, move everyone
333 * over one to the right, and insert the new candidate */
334 if(i<csize){
335 for(j=5; j>i; j--){
336 candidates[j]=candidates[j-1];
337 }
338 }
339 /* regardless, if they should be on the list... */
340 if(i<5) {
341 candidates[i]=cserver;
342 if(csize<5) csize++;
343 /* otherwise discard the server */
344 } else {
345 DBG(printf("discarding peer id %d\n", cserver));
346 }
347 }
349 if(csize>0) {
350 DBG(printf("best server selected: peer %d\n", candidates[0]));
351 return candidates[0];
352 } else {
353 DBG(printf("no peers meeting synchronization criteria :(\n"));
354 return -1;
355 }
356 }
358 /* do everything we need to get the total average offset
359 * - we use a certain amount of parallelization with poll() to ensure
360 * we don't waste time sitting around waiting for single packets.
361 * - we also "manually" handle resolving host names and connecting, because
362 * we have to do it in a way that our lazy macros don't handle currently :( */
363 double offset_request(const char *host, int *stratum, int *status){
364 int i=0, j=0, ga_result=0, num_hosts=0, *socklist=NULL, respnum=0;
365 int servers_completed=0, one_written=0, one_read=0, servers_readable=0, best_index=-1;
366 time_t now_time=0, start_ts=0;
367 ntp_message *req=NULL;
368 double avg_offset=0.;
369 struct timeval recv_time;
370 struct addrinfo *ai=NULL, *ai_tmp=NULL, hints;
371 struct pollfd *ufds=NULL;
372 ntp_server_results *servers=NULL;
374 /* setup hints to only return results from getaddrinfo that we'd like */
375 memset(&hints, 0, sizeof(struct addrinfo));
376 hints.ai_family = address_family;
377 hints.ai_protocol = IPPROTO_UDP;
378 hints.ai_socktype = SOCK_DGRAM;
380 /* fill in ai with the list of hosts resolved by the host name */
381 ga_result = getaddrinfo(host, "123", &hints, &ai);
382 if(ga_result!=0){
383 die(STATE_UNKNOWN, "error getting address for %s: %s\n",
384 host, gai_strerror(ga_result));
385 }
387 /* count the number of returned hosts, and allocate stuff accordingly */
388 for(ai_tmp=ai; ai_tmp!=NULL; ai_tmp=ai_tmp->ai_next){ num_hosts++; }
389 req=(ntp_message*)malloc(sizeof(ntp_message)*num_hosts);
390 if(req==NULL) die(STATE_UNKNOWN, "can not allocate ntp message array");
391 socklist=(int*)malloc(sizeof(int)*num_hosts);
392 if(socklist==NULL) die(STATE_UNKNOWN, "can not allocate socket array");
393 ufds=(struct pollfd*)malloc(sizeof(struct pollfd)*num_hosts);
394 if(ufds==NULL) die(STATE_UNKNOWN, "can not allocate socket array");
395 servers=(ntp_server_results*)malloc(sizeof(ntp_server_results)*num_hosts);
396 if(servers==NULL) die(STATE_UNKNOWN, "can not allocate server array");
397 memset(servers, 0, sizeof(ntp_server_results)*num_hosts);
399 /* setup each socket for writing, and the corresponding struct pollfd */
400 ai_tmp=ai;
401 for(i=0;ai_tmp;i++){
402 socklist[i]=socket(ai_tmp->ai_family, SOCK_DGRAM, IPPROTO_UDP);
403 if(socklist[i] == -1) {
404 perror(NULL);
405 die(STATE_UNKNOWN, "can not create new socket");
406 }
407 if(connect(socklist[i], ai_tmp->ai_addr, ai_tmp->ai_addrlen)){
408 die(STATE_UNKNOWN, "can't create socket connection");
409 } else {
410 ufds[i].fd=socklist[i];
411 ufds[i].events=POLLIN;
412 ufds[i].revents=0;
413 }
414 ai_tmp = ai_tmp->ai_next;
415 }
417 /* now do AVG_NUM checks to each host. we stop before timeout/2 seconds
418 * have passed in order to ensure post-processing and jitter time. */
419 now_time=start_ts=time(NULL);
420 while(servers_completed<num_hosts && now_time-start_ts <= socket_timeout/2){
421 /* loop through each server and find each one which hasn't
422 * been touched in the past second or so and is still lacking
423 * some responses. for each of these servers, send a new request,
424 * and update the "waiting" timestamp with the current time. */
425 one_written=0;
426 now_time=time(NULL);
428 for(i=0; i<num_hosts; i++){
429 if(servers[i].waiting<now_time && servers[i].num_responses<AVG_NUM){
430 if(verbose && servers[i].waiting != 0) printf("re-");
431 if(verbose) printf("sending request to peer %d\n", i);
432 setup_request(&req[i]);
433 write(socklist[i], &req[i], sizeof(ntp_message));
434 servers[i].waiting=now_time;
435 one_written=1;
436 break;
437 }
438 }
440 /* quickly poll for any sockets with pending data */
441 servers_readable=poll(ufds, num_hosts, 100);
442 if(servers_readable==-1){
443 perror("polling ntp sockets");
444 die(STATE_UNKNOWN, "communication errors");
445 }
447 /* read from any sockets with pending data */
448 for(i=0; servers_readable && i<num_hosts; i++){
449 if(ufds[i].revents&POLLIN && servers[i].num_responses < AVG_NUM){
450 if(verbose) {
451 printf("response from peer %d: ", i);
452 }
454 read(ufds[i].fd, &req[i], sizeof(ntp_message));
455 gettimeofday(&recv_time, NULL);
456 DBG(print_ntp_message(&req[i]));
457 respnum=servers[i].num_responses++;
458 servers[i].offset[respnum]=calc_offset(&req[i], &recv_time);
459 if(verbose) {
460 printf("offset %.10g, stratum %i\n", servers[i].offset[respnum], req[i].stratum);
461 }
462 servers[i].stratum=req[i].stratum;
463 servers[i].rtdisp=NTP32asDOUBLE(req[i].rtdisp);
464 servers[i].rtdelay=NTP32asDOUBLE(req[i].rtdelay);
465 servers[i].waiting=0;
466 servers[i].flags=req[i].flags;
467 servers_readable--;
468 one_read = 1;
469 if(servers[i].num_responses==AVG_NUM) servers_completed++;
470 }
471 }
472 /* lather, rinse, repeat. */
473 }
475 if (one_read == 0) {
476 die(STATE_CRITICAL, "NTP CRITICAL: No response from NTP server\n");
477 }
479 /* now, pick the best server from the list */
480 best_index=best_offset_server(servers, num_hosts);
481 if(best_index < 0){
482 *status=STATE_CRITICAL;
483 } else {
484 /* finally, calculate the average offset */
485 for(i=0; i<servers[best_index].num_responses;i++){
486 avg_offset+=servers[best_index].offset[j];
487 }
488 avg_offset/=servers[best_index].num_responses;
489 *stratum = servers[best_index].stratum;
490 }
492 /* cleanup */
493 /* FIXME: Not closing the socket to avoid re-use of the local port
494 * which can cause old NTP packets to be read instead of NTP control
495 * pactets in jitter_request(). THERE MUST BE ANOTHER WAY...
496 * for(j=0; j<num_hosts; j++){ close(socklist[j]); } */
497 free(socklist);
498 free(ufds);
499 free(servers);
500 free(req);
501 freeaddrinfo(ai);
503 if(verbose) printf("overall average offset: %.10g\n", avg_offset);
504 return avg_offset;
505 }
507 void
508 setup_control_request(ntp_control_message *p, uint8_t opcode, uint16_t seq){
509 memset(p, 0, sizeof(ntp_control_message));
510 LI_SET(p->flags, LI_NOWARNING);
511 VN_SET(p->flags, VN_RESERVED);
512 MODE_SET(p->flags, MODE_CONTROLMSG);
513 OP_SET(p->op, opcode);
514 p->seq = htons(seq);
515 /* Remaining fields are zero for requests */
516 }
518 /* XXX handle responses with the error bit set */
519 double jitter_request(const char *host, int *status){
520 int conn=-1, i, npeers=0, num_candidates=0, syncsource_found=0;
521 int run=0, min_peer_sel=PEER_INCLUDED, num_selected=0, num_valid=0;
522 int peers_size=0, peer_offset=0;
523 ntp_assoc_status_pair *peers=NULL;
524 ntp_control_message req;
525 const char *getvar = "jitter";
526 double rval = 0.0, jitter = -1.0;
527 char *startofvalue=NULL, *nptr=NULL;
528 void *tmp;
530 /* Long-winded explanation:
531 * Getting the jitter requires a number of steps:
532 * 1) Send a READSTAT request.
533 * 2) Interpret the READSTAT reply
534 * a) The data section contains a list of peer identifiers (16 bits)
535 * and associated status words (16 bits)
536 * b) We want the value of 0x06 in the SEL (peer selection) value,
537 * which means "current synchronizatin source". If that's missing,
538 * we take anything better than 0x04 (see the rfc for details) but
539 * set a minimum of warning.
540 * 3) Send a READVAR request for information on each peer identified
541 * in 2b greater than the minimum selection value.
542 * 4) Extract the jitter value from the data[] (it's ASCII)
543 */
544 my_udp_connect(server_address, 123, &conn);
546 /* keep sending requests until the server stops setting the
547 * REM_MORE bit, though usually this is only 1 packet. */
548 do{
549 setup_control_request(&req, OP_READSTAT, 1);
550 DBG(printf("sending READSTAT request"));
551 write(conn, &req, SIZEOF_NTPCM(req));
552 DBG(print_ntp_control_message(&req));
553 /* Attempt to read the largest size packet possible */
554 req.count=htons(MAX_CM_SIZE);
555 DBG(printf("recieving READSTAT response"))
556 read(conn, &req, SIZEOF_NTPCM(req));
557 DBG(print_ntp_control_message(&req));
558 /* Each peer identifier is 4 bytes in the data section, which
559 * we represent as a ntp_assoc_status_pair datatype.
560 */
561 peers_size+=ntohs(req.count);
562 if((tmp=realloc(peers, peers_size)) == NULL)
563 free(peers), die(STATE_UNKNOWN, "can not (re)allocate 'peers' buffer\n");
564 peers=tmp;
565 memcpy((void*)((ptrdiff_t)peers+peer_offset), (void*)req.data, ntohs(req.count));
566 npeers=peers_size/sizeof(ntp_assoc_status_pair);
567 peer_offset+=ntohs(req.count);
568 } while(req.op&REM_MORE);
570 /* first, let's find out if we have a sync source, or if there are
571 * at least some candidates. in the case of the latter we'll issue
572 * a warning but go ahead with the check on them. */
573 for (i = 0; i < npeers; i++){
574 if (PEER_SEL(peers[i].status) >= PEER_INCLUDED){
575 num_candidates++;
576 if(PEER_SEL(peers[i].status) >= PEER_SYNCSOURCE){
577 syncsource_found=1;
578 min_peer_sel=PEER_SYNCSOURCE;
579 }
580 }
581 }
582 if(verbose) printf("%d candiate peers available\n", num_candidates);
583 if(verbose && syncsource_found) printf("synchronization source found\n");
584 if(! syncsource_found){
585 *status = STATE_WARNING;
586 if(verbose) printf("warning: no synchronization source found\n");
587 }
590 for (run=0; run<AVG_NUM; run++){
591 if(verbose) printf("jitter run %d of %d\n", run+1, AVG_NUM);
592 for (i = 0; i < npeers; i++){
593 /* Only query this server if it is the current sync source */
594 if (PEER_SEL(peers[i].status) >= min_peer_sel){
595 num_selected++;
596 setup_control_request(&req, OP_READVAR, 2);
597 req.assoc = peers[i].assoc;
598 /* By spec, putting the variable name "jitter" in the request
599 * should cause the server to provide _only_ the jitter value.
600 * thus reducing net traffic, guaranteeing us only a single
601 * datagram in reply, and making intepretation much simpler
602 */
603 /* Older servers doesn't know what jitter is, so if we get an
604 * error on the first pass we redo it with "dispersion" */
605 strncpy(req.data, getvar, MAX_CM_SIZE-1);
606 req.count = htons(strlen(getvar));
607 DBG(printf("sending READVAR request...\n"));
608 write(conn, &req, SIZEOF_NTPCM(req));
609 DBG(print_ntp_control_message(&req));
611 req.count = htons(MAX_CM_SIZE);
612 DBG(printf("recieving READVAR response...\n"));
613 read(conn, &req, SIZEOF_NTPCM(req));
614 DBG(print_ntp_control_message(&req));
616 if(req.op&REM_ERROR && strstr(getvar, "jitter")) {
617 if(verbose) printf("The 'jitter' command failed (old ntp server?)\nRestarting with 'dispersion'...\n");
618 getvar = "dispersion";
619 num_selected--;
620 i--;
621 continue;
622 }
624 /* get to the float value */
625 if(verbose) {
626 printf("parsing jitter from peer %.2x: ", ntohs(peers[i].assoc));
627 }
628 startofvalue = strchr(req.data, '=');
629 if(startofvalue != NULL) {
630 startofvalue++;
631 jitter = strtod(startofvalue, &nptr);
632 }
633 if(startofvalue == NULL || startofvalue==nptr){
634 printf("warning: unable to read server jitter response.\n");
635 *status = STATE_WARNING;
636 } else {
637 if(verbose) printf("%g\n", jitter);
638 num_valid++;
639 rval += jitter;
640 }
641 }
642 }
643 if(verbose){
644 printf("jitter parsed from %d/%d peers\n", num_valid, num_selected);
645 }
646 }
648 rval = num_valid ? rval / num_valid : -1.0;
650 close(conn);
651 if(peers!=NULL) free(peers);
652 /* If we return -1.0, it means no synchronization source was found */
653 return rval;
654 }
656 int process_arguments(int argc, char **argv){
657 int c;
658 int option=0;
659 static struct option longopts[] = {
660 {"version", no_argument, 0, 'V'},
661 {"help", no_argument, 0, 'h'},
662 {"verbose", no_argument, 0, 'v'},
663 {"use-ipv4", no_argument, 0, '4'},
664 {"use-ipv6", no_argument, 0, '6'},
665 {"warning", required_argument, 0, 'w'},
666 {"critical", required_argument, 0, 'c'},
667 {"swarn", required_argument, 0, 'W'},
668 {"scrit", required_argument, 0, 'C'},
669 {"jwarn", required_argument, 0, 'j'},
670 {"jcrit", required_argument, 0, 'k'},
671 {"timeout", required_argument, 0, 't'},
672 {"hostname", required_argument, 0, 'H'},
673 {0, 0, 0, 0}
674 };
677 if (argc < 2)
678 usage ("\n");
680 while (1) {
681 c = getopt_long (argc, argv, "Vhv46w:c:W:C:j:k:t:H:", longopts, &option);
682 if (c == -1 || c == EOF || c == 1)
683 break;
685 switch (c) {
686 case 'h':
687 print_help();
688 exit(STATE_OK);
689 break;
690 case 'V':
691 print_revision(progname, revision);
692 exit(STATE_OK);
693 break;
694 case 'v':
695 verbose++;
696 break;
697 case 'w':
698 owarn = optarg;
699 break;
700 case 'c':
701 ocrit = optarg;
702 break;
703 case 'W':
704 do_stratum=1;
705 swarn = optarg;
706 break;
707 case 'C':
708 do_stratum=1;
709 scrit = optarg;
710 break;
711 case 'j':
712 do_jitter=1;
713 jwarn = optarg;
714 break;
715 case 'k':
716 do_jitter=1;
717 jcrit = optarg;
718 break;
719 case 'H':
720 if(is_host(optarg) == FALSE)
721 usage2(_("Invalid hostname/address"), optarg);
722 server_address = strdup(optarg);
723 break;
724 case 't':
725 socket_timeout=atoi(optarg);
726 break;
727 case '4':
728 address_family = AF_INET;
729 break;
730 case '6':
731 #ifdef USE_IPV6
732 address_family = AF_INET6;
733 #else
734 usage4 (_("IPv6 support not available"));
735 #endif
736 break;
737 case '?':
738 /* print short usage statement if args not parsable */
739 usage5 ();
740 break;
741 }
742 }
744 if(server_address == NULL){
745 usage4(_("Hostname was not supplied"));
746 }
748 return 0;
749 }
751 char *perfd_offset (double offset)
752 {
753 return fperfdata ("offset", offset, "s",
754 TRUE, offset_thresholds->warning->end,
755 TRUE, offset_thresholds->critical->end,
756 FALSE, 0, FALSE, 0);
757 }
759 char *perfd_jitter (double jitter)
760 {
761 return fperfdata ("jitter", jitter, "s",
762 do_jitter, jitter_thresholds->warning->end,
763 do_jitter, jitter_thresholds->critical->end,
764 TRUE, 0, FALSE, 0);
765 }
767 char *perfd_stratum (int stratum)
768 {
769 return perfdata ("stratum", stratum, "",
770 do_stratum, (int)stratum_thresholds->warning->end,
771 do_stratum, (int)stratum_thresholds->critical->end,
772 TRUE, 0, TRUE, 16);
773 }
775 int main(int argc, char *argv[]){
776 int result, offset_result, jitter_result, stratum;
777 double offset=0, jitter=0;
778 char *result_line, *perfdata_line;
780 result = offset_result = jitter_result= STATE_UNKNOWN;
782 if (process_arguments (argc, argv) == ERROR)
783 usage4 (_("Could not parse arguments"));
785 set_thresholds(&offset_thresholds, owarn, ocrit);
786 set_thresholds(&jitter_thresholds, jwarn, jcrit);
787 set_thresholds(&stratum_thresholds, swarn, scrit);
789 /* initialize alarm signal handling */
790 signal (SIGALRM, socket_timeout_alarm_handler);
792 /* set socket timeout */
793 alarm (socket_timeout);
795 offset = offset_request(server_address, &stratum, &offset_result);
796 result = get_status(fabs(offset), offset_thresholds);
797 result = max_state(result, offset_result);
798 if(do_stratum)
799 result = max_state(result, get_status(stratum, stratum_thresholds));
801 /* If not told to check the jitter, we don't even send packets.
802 * jitter is checked using NTP control packets, which not all
803 * servers recognize. Trying to check the jitter on OpenNTPD
804 * (for example) will result in an error
805 */
806 if(do_jitter){
807 jitter=jitter_request(server_address, &jitter_result);
808 result = max_state(result, get_status(jitter, jitter_thresholds));
809 /* -1 indicates that we couldn't calculate the jitter
810 * Only overrides STATE_OK from the offset */
811 if(jitter == -1.0 && result == STATE_OK)
812 result = STATE_UNKNOWN;
813 }
814 result = max_state(result, jitter_result);
816 switch (result) {
817 case STATE_CRITICAL :
818 asprintf(&result_line, "NTP CRITICAL:");
819 break;
820 case STATE_WARNING :
821 asprintf(&result_line, "NTP WARNING:");
822 break;
823 case STATE_OK :
824 asprintf(&result_line, "NTP OK:");
825 break;
826 default :
827 asprintf(&result_line, "NTP UNKNOWN:");
828 break;
829 }
830 if(offset_result==STATE_CRITICAL){
831 asprintf(&result_line, "%s %s", result_line, _("Offset unknown"));
832 asprintf(&perfdata_line, "");
833 } else {
834 if(offset_result==STATE_WARNING){
835 asprintf(&result_line, "%s %s", result_line, _("Unable to fully sample sync server"));
836 }
837 asprintf(&result_line, "%s Offset %.10g secs", result_line, offset);
838 asprintf(&perfdata_line, "%s", perfd_offset(offset));
839 }
840 if (do_jitter) {
841 asprintf(&result_line, "%s, jitter=%f", result_line, jitter);
842 asprintf(&perfdata_line, "%s %s", perfdata_line, perfd_jitter(jitter));
843 }
844 if (do_stratum) {
845 asprintf(&result_line, "%s, stratum=%i", result_line, stratum);
846 asprintf(&perfdata_line, "%s %s", perfdata_line, perfd_stratum(stratum));
847 }
848 printf("%s|%s\n", result_line, perfdata_line);
850 if(server_address!=NULL) free(server_address);
851 return result;
852 }
856 void print_help(void){
857 print_revision(progname, revision);
859 printf ("Copyright (c) 2006 Sean Finney\n");
860 printf (COPYRIGHT, copyright, email);
862 printf ("%s\n", _("This plugin checks the selected ntp server"));
864 printf ("\n\n");
866 print_usage();
867 printf (_(UT_HELP_VRSN));
868 printf (_(UT_HOST_PORT), 'p', "123");
869 printf (" %s\n", "-w, --warning=THRESHOLD");
870 printf (" %s\n", _("Offset to result in warning status (seconds)"));
871 printf (" %s\n", "-c, --critical=THRESHOLD");
872 printf (" %s\n", _("Offset to result in critical status (seconds)"));
873 printf (" %s\n", "-W, --warning=THRESHOLD");
874 printf (" %s\n", _("Warning threshold for stratum"));
875 printf (" %s\n", "-W, --critical=THRESHOLD");
876 printf (" %s\n", _("Critical threshold for stratum"));
877 printf (" %s\n", "-j, --warning=THRESHOLD");
878 printf (" %s\n", _("Warning threshold for jitter"));
879 printf (" %s\n", "-k, --critical=THRESHOLD");
880 printf (" %s\n", _("Critical threshold for jitter"));
881 printf (_(UT_TIMEOUT), DEFAULT_SOCKET_TIMEOUT);
882 printf (_(UT_VERBOSE));
884 printf("\n");
885 printf("%s\n", _("Notes:"));
886 printf(" %s\n", _("See:"));
887 printf(" %s\n", ("http://nagiosplug.sourceforge.net/developer-guidelines.html#THRESHOLDFORMAT"));
888 printf(" %s\n", _("for THRESHOLD format and examples."));
890 printf (_(UT_SUPPORT));
891 }
893 void
894 print_usage(void)
895 {
896 printf (_("Usage:"));
897 printf(" %s -H <host> [-w <warn>] [-c <crit>] [-W <warn>] [-C <crit>]\n", progname);
898 printf(" [-j <warn>] [-k <crit>] [-v verbose]\n");
899 }