xref: /freebsd/contrib/ntp/ntpd/ntp_monitor.c (revision f5f40dd63bc7acbb5312b26ac1ea1103c12352a6)
1 /*
2  * ntp_monitor - monitor ntpd statistics
3  */
4 #ifdef HAVE_CONFIG_H
5 # include <config.h>
6 #endif
7 
8 #include "ntpd.h"
9 #include "ntp_io.h"
10 #include "ntp_if.h"
11 #include "ntp_lists.h"
12 #include "ntp_stdlib.h"
13 #include <ntp_random.h>
14 
15 #include <stdio.h>
16 #include <signal.h>
17 #ifdef HAVE_SYS_IOCTL_H
18 # include <sys/ioctl.h>
19 #endif
20 
21 /*
22  * Record statistics based on source address, mode and version. The
23  * receive procedure calls us with the incoming rbufp before it does
24  * anything else. While at it, implement rate controls for inbound
25  * traffic.
26  *
27  * Each entry is doubly linked into two lists, a hash table and a most-
28  * recently-used (MRU) list. When a packet arrives it is looked up in
29  * the hash table. If found, the statistics are updated and the entry
30  * relinked at the head of the MRU list. If not found, a new entry is
31  * allocated, initialized and linked into both the hash table and at the
32  * head of the MRU list.
33  *
34  * Memory is usually allocated by grabbing a big chunk of new memory and
35  * cutting it up into littler pieces. The exception to this when we hit
36  * the memory limit. Then we free memory by grabbing entries off the
37  * tail for the MRU list, unlinking from the hash table, and
38  * reinitializing.
39  *
40  * INC_MONLIST is the default allocation granularity in entries.
41  * INIT_MONLIST is the default initial allocation in entries.
42  */
43 #ifdef MONMEMINC		/* old name */
44 # define	INC_MONLIST	MONMEMINC
45 #elif !defined(INC_MONLIST)
46 # define	INC_MONLIST	(4 * 1024 / sizeof(mon_entry))
47 #endif
48 #ifndef INIT_MONLIST
49 # define	INIT_MONLIST	(4 * 1024 / sizeof(mon_entry))
50 #endif
51 #ifndef MRU_MAXDEPTH_DEF
52 # define MRU_MAXDEPTH_DEF	(1024 * 1024 / sizeof(mon_entry))
53 #endif
54 
55 /*
56  * Hashing stuff
57  */
58 u_char	mon_hash_bits;
59 
60 /*
61  * Pointers to the hash table and the MRU list.  Memory for the hash
62  * table is allocated only if monitoring is enabled.
63  */
64 mon_entry **	mon_hash;	/* MRU hash table */
65 mon_entry	mon_mru_list;	/* mru listhead */
66 
67 /*
68  * List of free structures structures, and counters of in-use and total
69  * structures. The free structures are linked with the hash_next field.
70  */
71 static  mon_entry *mon_free;		/* free list or null if none */
72 	u_int mru_alloc;		/* mru list + free list count */
73 	u_int mru_entries;		/* mru list count */
74 	u_int mru_peakentries;		/* highest mru_entries seen */
75 	u_int mru_initalloc = INIT_MONLIST;/* entries to preallocate */
76 	u_int mru_incalloc = INC_MONLIST;/* allocation batch factor */
77 static	u_int mon_mem_increments;	/* times called malloc() */
78 
79 /*
80  * Parameters of the RES_LIMITED restriction option. We define headway
81  * as the idle time between packets. A packet is discarded if the
82  * headway is less than the minimum, as well as if the average headway
83  * is less than eight times the increment.
84  */
85 int	ntp_minpkt = NTP_MINPKT;	/* minimum seconds between */
86 					/* requests from a client */
87 u_char	ntp_minpoll = NTP_MINPOLL;	/* minimum average log2 seconds */
88 					/* between client requests */
89 
90 /*
91  * Initialization state.  We may be monitoring, we may not.  If
92  * we aren't, we may not even have allocated any memory yet.
93  */
94 	u_int	mon_enabled;		/* enable switch */
95 	u_int	mru_mindepth = 600;	/* preempt above this */
96 	int	mru_maxage = 64;	/* for entries older than */
97 	u_int	mru_maxdepth = 		/* MRU count hard limit */
98 			MRU_MAXDEPTH_DEF;
99 	int	mon_age = 3000;		/* preemption limit */
100 
101 static	void		mon_getmoremem(void);
102 static	void		remove_from_hash(mon_entry *);
103 static	inline void	mon_free_entry(mon_entry *);
104 static	inline void	mon_reclaim_entry(mon_entry *);
105 
106 
107 /*
108  * init_mon - initialize monitoring global data
109  */
110 void
111 init_mon(void)
112 {
113 	/*
114 	 * Don't do much of anything here.  We don't allocate memory
115 	 * until mon_start().
116 	 */
117 	mon_enabled = MON_OFF;
118 	INIT_DLIST(mon_mru_list, mru);
119 }
120 
121 
122 /*
123  * remove_from_hash - removes an entry from the address hash table and
124  *		      decrements mru_entries.
125  */
126 static void
127 remove_from_hash(
128 	mon_entry *mon
129 	)
130 {
131 	u_int hash;
132 	mon_entry *punlinked;
133 
134 	mru_entries--;
135 	hash = MON_HASH(&mon->rmtadr);
136 	UNLINK_SLIST(punlinked, mon_hash[hash], mon, hash_next,
137 		     mon_entry);
138 	ENSURE(punlinked == mon);
139 }
140 
141 
142 static inline void
143 mon_free_entry(
144 	mon_entry *m
145 	)
146 {
147 	ZERO(*m);
148 	LINK_SLIST(mon_free, m, hash_next);
149 }
150 
151 
152 /*
153  * mon_reclaim_entry - Remove an entry from the MRU list and from the
154  *		       hash array, then zero-initialize it.  Indirectly
155  *		       decrements mru_entries.
156 
157  * The entry is prepared to be reused.  Before return, in
158  * remove_from_hash(), mru_entries is decremented.  It is the caller's
159  * responsibility to increment it again.
160  */
161 static inline void
162 mon_reclaim_entry(
163 	mon_entry *m
164 	)
165 {
166 	DEBUG_INSIST(NULL != m);
167 
168 	UNLINK_DLIST(m, mru);
169 	remove_from_hash(m);
170 	ZERO(*m);
171 }
172 
173 
174 /*
175  * mon_getmoremem - get more memory and put it on the free list
176  */
177 static void
178 mon_getmoremem(void)
179 {
180 	mon_entry *chunk;
181 	u_int entries;
182 
183 	entries = (0 == mon_mem_increments)
184 		      ? mru_initalloc
185 		      : mru_incalloc;
186 
187 	if (entries) {
188 		chunk = eallocarray(entries, sizeof(*chunk));
189 		mru_alloc += entries;
190 		for (chunk += entries; entries; entries--)
191 			mon_free_entry(--chunk);
192 
193 		mon_mem_increments++;
194 	}
195 }
196 
197 
198 /*
199  * mon_start - start up the monitoring software
200  */
201 void
202 mon_start(
203 	int mode
204 	)
205 {
206 	size_t octets;
207 	u_int min_hash_slots;
208 
209 	if (MON_OFF == mode)		/* MON_OFF is 0 */
210 		return;
211 	if (mon_enabled) {
212 		mon_enabled |= mode;
213 		return;
214 	}
215 	if (0 == mon_mem_increments)
216 		mon_getmoremem();
217 	/*
218 	 * Select the MRU hash table size to limit the average count
219 	 * per bucket at capacity (mru_maxdepth) to 8, if possible
220 	 * given our hash is limited to 16 bits.
221 	 */
222 	min_hash_slots = (mru_maxdepth / 8) + 1;
223 	mon_hash_bits = 0;
224 	while (min_hash_slots >>= 1)
225 		mon_hash_bits++;
226 	mon_hash_bits = max(4, mon_hash_bits);
227 	mon_hash_bits = min(16, mon_hash_bits);
228 	octets = sizeof(*mon_hash) * MON_HASH_SIZE;
229 	mon_hash = erealloc_zero(mon_hash, octets, 0);
230 
231 	mon_enabled = mode;
232 }
233 
234 
235 /*
236  * mon_stop - stop the monitoring software
237  */
238 void
239 mon_stop(
240 	int mode
241 	)
242 {
243 	mon_entry *mon;
244 
245 	if (MON_OFF == mon_enabled)
246 		return;
247 	if ((mon_enabled & mode) == 0 || mode == MON_OFF)
248 		return;
249 
250 	mon_enabled &= ~mode;
251 	if (mon_enabled != MON_OFF)
252 		return;
253 
254 	/*
255 	 * Move everything on the MRU list to the free list quickly,
256 	 * without bothering to remove each from either the MRU list or
257 	 * the hash table.
258 	 */
259 	ITER_DLIST_BEGIN(mon_mru_list, mon, mru, mon_entry)
260 		mon_free_entry(mon);
261 	ITER_DLIST_END()
262 
263 	/* empty the MRU list and hash table. */
264 	mru_entries = 0;
265 	INIT_DLIST(mon_mru_list, mru);
266 	zero_mem(mon_hash, sizeof(*mon_hash) * MON_HASH_SIZE);
267 }
268 
269 
270 /*
271  * mon_clearinterface -- remove mru entries referring to a local address
272  *			 which is going away.
273  */
274 void
275 mon_clearinterface(
276 	endpt *lcladr
277 	)
278 {
279 	mon_entry *mon;
280 
281 	/* iterate mon over mon_mru_list */
282 	ITER_DLIST_BEGIN(mon_mru_list, mon, mru, mon_entry)
283 		if (mon->lcladr == lcladr) {
284 			/* remove from mru list */
285 			UNLINK_DLIST(mon, mru);
286 			/* remove from hash list, adjust mru_entries */
287 			remove_from_hash(mon);
288 			/* put on free list */
289 			mon_free_entry(mon);
290 		}
291 	ITER_DLIST_END()
292 }
293 
294 
295 /*
296  * ntp_monitor - record stats about this packet
297  *
298  * Returns supplied restriction flags, with RES_LIMITED and RES_KOD
299  * cleared unless the packet should not be responded to normally
300  * (RES_LIMITED) and possibly should trigger a KoD response (RES_KOD).
301  * The returned flags are saved in the MRU entry, so that it reflects
302  * whether the last packet from that source triggered rate limiting,
303  * and if so, possible KoD response.  This implies you can not tell
304  * whether a given address is eligible for rate limiting/KoD from the
305  * monlist restrict bits, only whether or not the last packet triggered
306  * such responses.  ntpdc -c reslist lets you see whether RES_LIMITED
307  * or RES_KOD is lit for a particular address before ntp_monitor()'s
308  * typical dousing.
309  */
310 u_short
311 ntp_monitor(
312 	struct recvbuf *rbufp,
313 	u_short	flags
314 	)
315 {
316 	l_fp		interval_fp;
317 	struct pkt *	pkt;
318 	mon_entry *	mon;
319 	mon_entry *	oldest;
320 	int		oldest_age;
321 	u_int		hash;
322 	u_short		restrict_mask;
323 	u_char		mode;
324 	u_char		version;
325 	int		interval;
326 	int		head;		/* headway increment */
327 	int		leak;		/* new headway */
328 	int		limit;		/* average threshold */
329 
330 	REQUIRE(rbufp != NULL);
331 
332 	if (mon_enabled == MON_OFF) {
333 		return ~(RES_LIMITED | RES_KOD) & flags;
334 	}
335 	pkt = &rbufp->recv_pkt;
336 	hash = MON_HASH(&rbufp->recv_srcadr);
337 	mode = PKT_MODE(pkt->li_vn_mode);
338 	version = PKT_VERSION(pkt->li_vn_mode);
339 	mon = mon_hash[hash];
340 
341 	/*
342 	 * We keep track of all traffic for a given IP in one entry,
343 	 * otherwise cron'ed ntpdate or similar evades RES_LIMITED.
344 	 */
345 
346 	for (; mon != NULL; mon = mon->hash_next) {
347 		if (SOCK_EQ(&mon->rmtadr, &rbufp->recv_srcadr)) {
348 			break;
349 		}
350 	}
351 	if (mon != NULL) {
352 		interval_fp = rbufp->recv_time;
353 		L_SUB(&interval_fp, &mon->last);
354 		/* add one-half second to round up */
355 		L_ADDUF(&interval_fp, 0x80000000);
356 		interval = interval_fp.l_i;
357 		mon->last = rbufp->recv_time;
358 		NSRCPORT(&mon->rmtadr) = NSRCPORT(&rbufp->recv_srcadr);
359 		mon->count++;
360 		restrict_mask = flags;
361 		mon->vn_mode = VN_MODE(version, mode);
362 
363 		/* Shuffle to the head of the MRU list. */
364 		UNLINK_DLIST(mon, mru);
365 		LINK_DLIST(mon_mru_list, mon, mru);
366 
367 		/*
368 		 * At this point the most recent arrival is first in the
369 		 * MRU list.  Decrease the counter by the headway, but
370 		 * not less than zero.
371 		 */
372 		mon->leak -= interval;
373 		mon->leak = max(0, mon->leak);
374 		head = 1 << ntp_minpoll;
375 		leak = mon->leak + head;
376 		limit = NTP_SHIFT * head;
377 
378 		DPRINTF(2, ("MRU: interval %d headway %d limit %d\n",
379 			    interval, leak, limit));
380 
381 		/*
382 		 * If the minimum and average thresholds are not
383 		 * exceeded, douse the RES_LIMITED and RES_KOD bits and
384 		 * increase the counter by the headway increment.  Note
385 		 * that we give a 1-s grace for the minimum threshold
386 		 * and a 2-s grace for the headway increment.  If one or
387 		 * both thresholds are exceeded and the old counter is
388 		 * less than the average threshold, set the counter to
389 		 * the average threshold plus the increment and leave
390 		 * the RES_LIMITED and RES_KOD bits lit. Otherwise,
391 		 * leave the counter alone and douse the RES_KOD bit.
392 		 * This rate-limits the KoDs to no more often than the
393 		 * average headway.
394 		 */
395 		if (interval + 1 >= ntp_minpkt && leak < limit) {
396 			mon->leak = leak - 2;
397 			restrict_mask &= ~(RES_LIMITED | RES_KOD);
398 		} else if (mon->leak < limit) {
399 			mon->leak = limit + head;
400 		} else {
401 			restrict_mask &= ~RES_KOD;
402 		}
403 		mon->flags = restrict_mask;
404 
405 		return mon->flags;
406 	}
407 
408 	/*
409 	 * If we got here, this is the first we've heard of this
410 	 * guy.  Get him some memory, either from the free list
411 	 * or from the tail of the MRU list.
412 	 *
413 	 * The following ntp.conf "mru" knobs come into play determining
414 	 * the depth (or count) of the MRU list:
415 	 * - mru_mindepth ("mru mindepth") is a floor beneath which
416 	 *   entries are kept without regard to their age.  The
417 	 *   default is 600 which matches the longtime implementation
418 	 *   limit on the total number of entries.
419 	 * - mru_maxage ("mru maxage") is a ceiling on the age in
420 	 *   seconds of entries.  Entries older than this are
421 	 *   reclaimed once mon_mindepth is exceeded.  64s default.
422 	 *   Note that entries older than this can easily survive
423 	 *   as they are reclaimed only as needed.
424 	 * - mru_maxdepth ("mru maxdepth") is a hard limit on the
425 	 *   number of entries.
426 	 * - "mru maxmem" sets mru_maxdepth to the number of entries
427 	 *   which fit in the given number of kilobytes.  The default is
428 	 *   1024, or 1 megabyte.
429 	 * - mru_initalloc ("mru initalloc" sets the count of the
430 	 *   initial allocation of MRU entries.
431 	 * - "mru initmem" sets mru_initalloc in units of kilobytes.
432 	 *   The default is 4.
433 	 * - mru_incalloc ("mru incalloc" sets the number of entries to
434 	 *   allocate on-demand each time the free list is empty.
435 	 * - "mru incmem" sets mru_incalloc in units of kilobytes.
436 	 *   The default is 4.
437 	 * Whichever of "mru maxmem" or "mru maxdepth" occurs last in
438 	 * ntp.conf controls.  Similarly for "mru initalloc" and "mru
439 	 * initmem", and for "mru incalloc" and "mru incmem".
440 	 */
441 	if (mru_entries < mru_mindepth) {
442 		if (NULL == mon_free)
443 			mon_getmoremem();
444 		UNLINK_HEAD_SLIST(mon, mon_free, hash_next);
445 	} else {
446 		oldest = TAIL_DLIST(mon_mru_list, mru);
447 		oldest_age = 0;		/* silence uninit warning */
448 		if (oldest != NULL) {
449 			interval_fp = rbufp->recv_time;
450 			L_SUB(&interval_fp, &oldest->last);
451 			/* add one-half second to round up */
452 			L_ADDUF(&interval_fp, 0x80000000);
453 			oldest_age = interval_fp.l_i;
454 		}
455 		/* note -1 is legal for mru_maxage (disables) */
456 		if (oldest != NULL && mru_maxage < oldest_age) {
457 			mon_reclaim_entry(oldest);
458 			mon = oldest;
459 		} else if (mon_free != NULL || mru_alloc <
460 			   mru_maxdepth) {
461 			if (NULL == mon_free)
462 				mon_getmoremem();
463 			UNLINK_HEAD_SLIST(mon, mon_free, hash_next);
464 		/* Preempt from the MRU list if old enough. */
465 		} else if (ntp_uurandom() >
466 			   (double)oldest_age / mon_age) {
467 			return ~(RES_LIMITED | RES_KOD) & flags;
468 		} else {
469 			mon_reclaim_entry(oldest);
470 			mon = oldest;
471 		}
472 	}
473 
474 	INSIST(mon != NULL);
475 
476 	/*
477 	 * Got one, initialize it
478 	 */
479 	mru_entries++;
480 	mru_peakentries = max(mru_peakentries, mru_entries);
481 	mon->last = rbufp->recv_time;
482 	mon->first = mon->last;
483 	mon->count = 1;
484 	mon->flags = ~(RES_LIMITED | RES_KOD) & flags;
485 	mon->leak = 0;
486 	memcpy(&mon->rmtadr, &rbufp->recv_srcadr, sizeof(mon->rmtadr));
487 	mon->vn_mode = VN_MODE(version, mode);
488 	mon->lcladr = rbufp->dstadr;
489 	mon->cast_flags = (u_char)(((rbufp->dstadr->flags &
490 	    INT_MCASTOPEN) && rbufp->fd == mon->lcladr->fd) ? MDF_MCAST
491 	    : rbufp->fd == mon->lcladr->bfd ? MDF_BCAST : MDF_UCAST);
492 
493 	/*
494 	 * Drop him into front of the hash table. Also put him on top of
495 	 * the MRU list.
496 	 */
497 	LINK_SLIST(mon_hash[hash], mon, hash_next);
498 	LINK_DLIST(mon_mru_list, mon, mru);
499 
500 	return mon->flags;
501 }
502 
503 
504