summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
-rw-r--r--plugins/check_ntp.c269
1 files changed, 247 insertions, 22 deletions
diff --git a/plugins/check_ntp.c b/plugins/check_ntp.c
index 5037786..2954aa8 100644
--- a/plugins/check_ntp.c
+++ b/plugins/check_ntp.c
@@ -43,6 +43,12 @@ int process_arguments (int, char **);
43void print_help (void); 43void print_help (void);
44void print_usage (void); 44void print_usage (void);
45 45
46/* number of times to perform each request to get a good average. */
47#define AVG_NUM 4
48
49/* max size of control message data */
50#define MAX_CM_SIZE 468
51
46/* this structure holds everything in an ntp request/response as per rfc1305 */ 52/* this structure holds everything in an ntp request/response as per rfc1305 */
47typedef struct { 53typedef struct {
48 uint8_t flags; /* byte with leapindicator,vers,mode. see macros */ 54 uint8_t flags; /* byte with leapindicator,vers,mode. see macros */
@@ -58,6 +64,25 @@ typedef struct {
58 uint64_t txts; /* time at which request departed server */ 64 uint64_t txts; /* time at which request departed server */
59} ntp_message; 65} ntp_message;
60 66
67/* this structure holds everything in an ntp control message as per rfc1305 */
68typedef struct {
69 uint8_t flags; /* byte with leapindicator,vers,mode. see macros */
70 uint8_t op; /* R,E,M bits and Opcode */
71 uint16_t seq; /* Packet sequence */
72 uint16_t status; /* Clock status */
73 uint16_t assoc; /* Association */
74 uint16_t offset; /* Similar to TCP sequence # */
75 uint16_t count; /* # bytes of data */
76 char data[MAX_CM_SIZE]; /* ASCII data of the request */
77 /* NB: not necessarily NULL terminated! */
78} ntp_control_message;
79
80/* this is an association/status-word pair found in control packet reponses */
81typedef struct {
82 uint16_t assoc;
83 uint16_t status;
84} ntp_assoc_status_pair;
85
61/* bits 1,2 are the leap indicator */ 86/* bits 1,2 are the leap indicator */
62#define LI_MASK 0xc0 87#define LI_MASK 0xc0
63#define LI(x) ((x&LI_MASK)>>6) 88#define LI(x) ((x&LI_MASK)>>6)
@@ -71,12 +96,28 @@ typedef struct {
71#define VN_MASK 0x38 96#define VN_MASK 0x38
72#define VN(x) ((x&VN_MASK)>>3) 97#define VN(x) ((x&VN_MASK)>>3)
73#define VN_SET(x,y) do{ x |= ((y<<3)&VN_MASK); }while(0) 98#define VN_SET(x,y) do{ x |= ((y<<3)&VN_MASK); }while(0)
99#define VN_RESERVED 0x02
74/* bits 6,7,8 are the ntp mode */ 100/* bits 6,7,8 are the ntp mode */
75#define MODE_MASK 0x07 101#define MODE_MASK 0x07
76#define MODE(x) (x&MODE_MASK) 102#define MODE(x) (x&MODE_MASK)
77#define MODE_SET(x,y) do{ x |= (y&MODE_MASK); }while(0) 103#define MODE_SET(x,y) do{ x |= (y&MODE_MASK); }while(0)
78/* here are some values */ 104/* here are some values */
79#define MODE_CLIENT 0x03 105#define MODE_CLIENT 0x03
106#define MODE_CONTROLMSG 0x06
107/* In control message, bits 8-10 are R,E,M bits */
108#define REM_MASK 0xe0
109#define REM_RESP 0x80
110#define REM_ERROR 0x40
111#define REM_MORE 0x20
112/* In control message, bits 11 - 15 are opcode */
113#define OP_MASK 0x1f
114#define OP_SET(x,y) do{ x |= (y&OP_MASK); }while(0)
115#define OP_READSTAT 0x01
116#define OP_READVAR 0x02
117/* In peer status bytes, bytes 6,7,8 determine clock selection status */
118#define PEER_SEL(x) (x&0x07)
119#define PEER_INCLUDED 0x04
120#define PEER_SYNCSOURCE 0x06
80 121
81/** 122/**
82 ** a note about the 32-bit "fixed point" numbers: 123 ** a note about the 32-bit "fixed point" numbers:
@@ -116,7 +157,7 @@ typedef struct {
116 do{ if(!n) t.tv_sec = t.tv_usec = 0; \ 157 do{ if(!n) t.tv_sec = t.tv_usec = 0; \
117 else { \ 158 else { \
118 t.tv_sec=ntohl(L32(n))-EPOCHDIFF; \ 159 t.tv_sec=ntohl(L32(n))-EPOCHDIFF; \
119 t.tv_usec=(int)(0.5+(double)(ntohl(R32(n))/4294.967296)); \ 160 t.tv_usec=(int)(0.5+(double)(ntohl(R32(n))/4294.967296)); \
120 } \ 161 } \
121 }while(0) 162 }while(0)
122 163
@@ -129,6 +170,17 @@ typedef struct {
129 } \ 170 } \
130 } while(0) 171 } while(0)
131 172
173/* NTP control message header is 12 bytes, plus any data in the data
174 * field, plus null padding to the nearest 32-bit boundary per rfc.
175 */
176#define SIZEOF_NTPCM(m) (12+ntohs(m.count)+((m.count)?4-(ntohs(m.count)%4):0))
177
178/* finally, a little helper or two for debugging: */
179#define DBG(x) do{if(verbose>1){ x; }}while(0);
180#define PRINTSOCKADDR(x) \
181 do{ \
182 printf("%u.%u.%u.%u", (x>>24)&0xff, (x>>16)&0xff, (x>>8)&0xff, x&0xff);\
183 }while(0);
132 184
133/* calculate the offset of the local clock */ 185/* calculate the offset of the local clock */
134static inline double calc_offset(const ntp_message *m, const struct timeval *t){ 186static inline double calc_offset(const ntp_message *m, const struct timeval *t){
@@ -142,7 +194,7 @@ static inline double calc_offset(const ntp_message *m, const struct timeval *t){
142} 194}
143 195
144/* print out a ntp packet in human readable/debuggable format */ 196/* print out a ntp packet in human readable/debuggable format */
145void print_packet(const ntp_message *p){ 197void print_ntp_message(const ntp_message *p){
146 struct timeval ref, orig, rx, tx; 198 struct timeval ref, orig, rx, tx;
147 199
148 NTP64toTV(p->refts,ref); 200 NTP64toTV(p->refts,ref);
@@ -167,6 +219,42 @@ void print_packet(const ntp_message *p){
167 printf("\ttxts = %-.16g\n", NTP64asDOUBLE(p->txts)); 219 printf("\ttxts = %-.16g\n", NTP64asDOUBLE(p->txts));
168} 220}
169 221
222void print_ntp_control_message(const ntp_control_message *p){
223 int i=0, numpeers=0;
224 const ntp_assoc_status_pair *peer=NULL;
225
226 printf("control packet contents:\n");
227 printf("\tflags: 0x%.2x , 0x%.2x\n", p->flags, p->op);
228 printf("\t li=%d (0x%.2x)\n", LI(p->flags), p->flags&LI_MASK);
229 printf("\t vn=%d (0x%.2x)\n", VN(p->flags), p->flags&VN_MASK);
230 printf("\t mode=%d (0x%.2x)\n", MODE(p->flags), p->flags&MODE_MASK);
231 printf("\t response=%d (0x%.2x)\n", (p->op&REM_RESP)>0, p->op&REM_RESP);
232 printf("\t more=%d (0x%.2x)\n", (p->op&REM_MORE)>0, p->op&REM_MORE);
233 printf("\t error=%d (0x%.2x)\n", (p->op&REM_ERROR)>0, p->op&REM_ERROR);
234 printf("\t op=%d (0x%.2x)\n", p->op&OP_MASK, p->op&OP_MASK);
235 printf("\tsequence: %d (0x%.2x)\n", ntohs(p->seq), ntohs(p->seq));
236 printf("\tstatus: %d (0x%.2x)\n", ntohs(p->status), ntohs(p->status));
237 printf("\tassoc: %d (0x%.2x)\n", ntohs(p->assoc), ntohs(p->assoc));
238 printf("\toffset: %d (0x%.2x)\n", ntohs(p->offset), ntohs(p->offset));
239 printf("\tcount: %d (0x%.2x)\n", ntohs(p->count), ntohs(p->count));
240 numpeers=ntohs(p->count)/(sizeof(ntp_assoc_status_pair));
241 if(p->op&REM_RESP && p->op&OP_READSTAT){
242 peer=(ntp_assoc_status_pair*)p->data;
243 for(i=0;i<numpeers;i++){
244 printf("\tpeer id %.2x status %.2x",
245 ntohs(peer[i].assoc), ntohs(peer[i].status));
246 if (PEER_SEL(peer[i].status) >= PEER_INCLUDED){
247 if(PEER_SEL(peer[i].status) >= PEER_SYNCSOURCE){
248 printf(" <-- current sync source");
249 } else {
250 printf(" <-- current sync candidate");
251 }
252 }
253 printf("\n");
254 }
255 }
256}
257
170void setup_request(ntp_message *p){ 258void setup_request(ntp_message *p){
171 struct timeval t; 259 struct timeval t;
172 260
@@ -189,7 +277,8 @@ double offset_request(const char *host){
189 double next_offset=0., avg_offset=0.; 277 double next_offset=0., avg_offset=0.;
190 struct timeval recv_time; 278 struct timeval recv_time;
191 279
192 for(i=0; i<4; i++){ 280 for(i=0; i<AVG_NUM; i++){
281 if(verbose) printf("offset run: %d/%d\n", i+1, AVG_NUM);
193 setup_request(&req); 282 setup_request(&req);
194 my_udp_connect(server_address, 123, &conn); 283 my_udp_connect(server_address, 123, &conn);
195 write(conn, &req, sizeof(ntp_message)); 284 write(conn, &req, sizeof(ntp_message));
@@ -201,12 +290,134 @@ double offset_request(const char *host){
201 if(verbose) printf("offset: %g\n", next_offset); 290 if(verbose) printf("offset: %g\n", next_offset);
202 avg_offset+=next_offset; 291 avg_offset+=next_offset;
203 } 292 }
204 return avg_offset/4.; 293 avg_offset/=AVG_NUM;
294 if(verbose) printf("average offset: %g\n", avg_offset);
295 return avg_offset;
296}
297
298void
299setup_control_request(ntp_control_message *p, uint8_t opcode, uint16_t seq){
300 memset(p, 0, sizeof(ntp_control_message));
301 LI_SET(p->flags, LI_NOWARNING);
302 VN_SET(p->flags, VN_RESERVED);
303 MODE_SET(p->flags, MODE_CONTROLMSG);
304 OP_SET(p->op, opcode);
305 p->seq = htons(seq);
306 /* Remaining fields are zero for requests */
205} 307}
206 308
207/* not yet implemented yet */ 309/* XXX handle responses with the error bit set */
208double jitter_request(const char *host){ 310double jitter_request(const char *host){
209 return 0.; 311 int conn=-1, i, npeers=0, num_candidates=0, syncsource_found=0;
312 int run=0, min_peer_sel=PEER_INCLUDED, num_selected=0, num_valid=0;
313 ntp_assoc_status_pair *peers;
314 ntp_control_message req;
315 double rval = 0.0, jitter = -1.0;
316 char *startofvalue=NULL, *nptr=NULL;
317
318 /* Long-winded explanation:
319 * Getting the jitter requires a number of steps:
320 * 1) Send a READSTAT request.
321 * 2) Interpret the READSTAT reply
322 * a) The data section contains a list of peer identifiers (16 bits)
323 * and associated status words (16 bits)
324 * b) We want the value of 0x06 in the SEL (peer selection) value,
325 * which means "current synchronizatin source". If that's missing,
326 * we take anything better than 0x04 (see the rfc for details) but
327 * set a minimum of warning.
328 * 3) Send a READVAR request for information on each peer identified
329 * in 2b greater than the minimum selection value.
330 * 4) Extract the jitter value from the data[] (it's ASCII)
331 */
332 my_udp_connect(server_address, 123, &conn);
333 setup_control_request(&req, OP_READSTAT, 1);
334
335 DBG(printf("sending READSTAT request"));
336 write(conn, &req, SIZEOF_NTPCM(req));
337 DBG(print_ntp_control_message(&req));
338 /* Attempt to read the largest size packet possible
339 * Is it possible for an NTP server to have more than 117 synchronization
340 * sources? If so, we will receive a second datagram with additional
341 * peers listed, since 117 is the maximum number that can fit in a
342 * single NTP control datagram. This code doesn't handle that case */
343 /* XXX check the REM_MORE bit */
344 req.count=htons(MAX_CM_SIZE);
345 DBG(printf("recieving READSTAT response"))
346 read(conn, &req, SIZEOF_NTPCM(req));
347 DBG(print_ntp_control_message(&req));
348 /* Each peer identifier is 4 bytes in the data section, which
349 * we represent as a ntp_assoc_status_pair datatype.
350 */
351 npeers=ntohs(req.count)/sizeof(ntp_assoc_status_pair);
352 peers=(ntp_assoc_status_pair*)malloc(sizeof(ntp_assoc_status_pair)*npeers);
353 memcpy((void*)peers, (void*)req.data, sizeof(ntp_assoc_status_pair)*npeers);
354 /* first, let's find out if we have a sync source, or if there are
355 * at least some candidates. in the case of the latter we'll issue
356 * a warning but go ahead with the check on them. */
357 for (i = 0; i < npeers; i++){
358 if (PEER_SEL(peers[i].status) >= PEER_INCLUDED){
359 num_candidates++;
360 if(PEER_SEL(peers[i].status) >= PEER_SYNCSOURCE){
361 syncsource_found=1;
362 min_peer_sel=PEER_SYNCSOURCE;
363 }
364 }
365 }
366 if(verbose) printf("%d candiate peers available\n", num_candidates);
367 if(verbose && syncsource_found) printf("synchronization source found\n");
368 /* XXX if ! syncsource_found set status to warning */
369
370 for (run=0; run<AVG_NUM; run++){
371 if(verbose) printf("jitter run %d of %d\n", run+1, AVG_NUM);
372 for (i = 0; i < npeers; i++){
373 /* Only query this server if it is the current sync source */
374 if (PEER_SEL(peers[i].status) >= min_peer_sel){
375 setup_control_request(&req, OP_READVAR, 2);
376 req.assoc = peers[i].assoc;
377 /* By spec, putting the variable name "jitter" in the request
378 * should cause the server to provide _only_ the jitter value.
379 * thus reducing net traffic, guaranteeing us only a single
380 * datagram in reply, and making intepretation much simpler
381 */
382 strncpy(req.data, "jitter", 6);
383 req.count = htons(6);
384 DBG(printf("sending READVAR request...\n"));
385 write(conn, &req, SIZEOF_NTPCM(req));
386 DBG(print_ntp_control_message(&req));
387
388 req.count = htons(MAX_CM_SIZE);
389 DBG(printf("recieving READVAR response...\n"));
390 read(conn, &req, SIZEOF_NTPCM(req));
391 DBG(print_ntp_control_message(&req));
392
393 /* get to the float value */
394 if(verbose) {
395 printf("parsing jitter from peer %.2x: ", peers[i].assoc);
396 }
397 startofvalue = strchr(req.data, '=') + 1;
398 jitter = strtod(startofvalue, &nptr);
399 num_selected++;
400 if(jitter == 0 && startofvalue==nptr){
401 printf("warning: unable to parse server response.\n");
402 /* XXX errors value ... */
403 } else {
404 if(verbose) printf("%g\n", jitter);
405 num_valid++;
406 rval += jitter;
407 }
408 }
409 }
410 if(verbose){
411 printf("jitter parsed from %d/%d peers\n", num_selected, num_valid);
412 }
413 }
414
415 rval /= num_valid;
416
417 close(conn);
418 free(peers);
419 /* If we return -1.0, it means no synchronization source was found */
420 return rval;
210} 421}
211 422
212int process_arguments(int argc, char **argv){ 423int process_arguments(int argc, char **argv){
@@ -247,7 +458,7 @@ int process_arguments(int argc, char **argv){
247 exit(STATE_OK); 458 exit(STATE_OK);
248 break; 459 break;
249 case 'v': 460 case 'v':
250 verbose = 1; 461 verbose++;
251 break; 462 break;
252 case 'w': 463 case 'w':
253 owarn = atof(optarg); 464 owarn = atof(optarg);
@@ -321,35 +532,49 @@ int main(int argc, char *argv[]){
321 532
322 offset = offset_request(server_address); 533 offset = offset_request(server_address);
323 if(offset > ocrit){ 534 if(offset > ocrit){
324 printf("NTP CRITICAL: ");
325 result = STATE_CRITICAL; 535 result = STATE_CRITICAL;
326 } else if(offset > owarn) { 536 } else if(offset > owarn) {
327 printf("NTP WARNING: ");
328 result = STATE_WARNING; 537 result = STATE_WARNING;
329 } else { 538 } else {
330 printf("NTP OK: ");
331 result = STATE_OK; 539 result = STATE_OK;
332 } 540 }
333 541
334 /* not implemented yet: */ 542 /* If not told to check the jitter, we don't even send packets.
335 jitter=jitter_request(server_address); 543 * jitter is checked using NTP control packets, which not all
336 544 * servers recognize. Trying to check the jitter on OpenNTPD
337 /* not implemented yet: 545 * (for example) will result in an error
546 */
338 if(do_jitter){ 547 if(do_jitter){
548 jitter=jitter_request(server_address);
339 if(jitter > jcrit){ 549 if(jitter > jcrit){
340 printf("NTP CRITICAL: "); 550 result = max_state(result, STATE_CRITICAL);
341 result = STATE_CRITICAL;
342 } else if(jitter > jwarn) { 551 } else if(jitter > jwarn) {
552 result = max_state(result, STATE_WARNING);
553 } else if(jitter == -1.0 && result == STATE_OK){
554 /* -1 indicates that we couldn't calculate the jitter
555 * Only overrides STATE_OK from the offset */
556 result = STATE_UNKNOWN;
557 }
558 }
559
560 switch (result) {
561 case STATE_CRITICAL :
562 printf("NTP CRITICAL: ");
563 break;
564 case STATE_WARNING :
343 printf("NTP WARNING: "); 565 printf("NTP WARNING: ");
344 result = STATE_WARNING; 566 break;
345 } else { 567 case STATE_OK :
346 printf("NTP OK: "); 568 printf("NTP OK: ");
347 result = STATE_OK; 569 break;
348 } 570 default :
571 printf("NTP UNKNOWN: ");
572 break;
349 } 573 }
350 */
351 574
352 printf("Offset %g secs|offset=%g\n", offset, offset); 575 printf("Offset %g secs|offset=%g", offset, offset);
576 if (do_jitter) printf("|jitter=%f", jitter);
577 printf("\n");
353 578
354 if(server_address!=NULL) free(server_address); 579 if(server_address!=NULL) free(server_address);
355 return result; 580 return result;