xref: /freebsd/sys/netinet/igmp.c (revision a9148abd9da5db2f1c682fb17bed791845fc41c9)
1 /*-
2  * Copyright (c) 1988 Stephen Deering.
3  * Copyright (c) 1992, 1993
4  *	The Regents of the University of California.  All rights reserved.
5  *
6  * This code is derived from software contributed to Berkeley by
7  * Stephen Deering of Stanford University.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)igmp.c	8.1 (Berkeley) 7/19/93
34  */
35 
36 /*
37  * Internet Group Management Protocol (IGMP) routines.
38  *
39  * Written by Steve Deering, Stanford, May 1988.
40  * Modified by Rosen Sharma, Stanford, Aug 1994.
41  * Modified by Bill Fenner, Xerox PARC, Feb 1995.
42  * Modified to fully comply to IGMPv2 by Bill Fenner, Oct 1995.
43  *
44  * MULTICAST Revision: 3.5.1.4
45  */
46 
47 #include <sys/cdefs.h>
48 __FBSDID("$FreeBSD$");
49 
50 #include "opt_mac.h"
51 
52 #include <sys/param.h>
53 #include <sys/systm.h>
54 #include <sys/malloc.h>
55 #include <sys/mbuf.h>
56 #include <sys/socket.h>
57 #include <sys/protosw.h>
58 #include <sys/kernel.h>
59 #include <sys/sysctl.h>
60 #include <sys/vimage.h>
61 
62 #include <net/if.h>
63 #include <net/route.h>
64 
65 #include <netinet/in.h>
66 #include <netinet/in_var.h>
67 #include <netinet/in_systm.h>
68 #include <netinet/ip.h>
69 #include <netinet/ip_var.h>
70 #include <netinet/ip_options.h>
71 #include <netinet/igmp.h>
72 #include <netinet/igmp_var.h>
73 
74 #include <machine/in_cksum.h>
75 
76 #include <security/mac/mac_framework.h>
77 
78 static MALLOC_DEFINE(M_IGMP, "igmp", "igmp state");
79 
80 static struct router_info	*find_rti(struct ifnet *ifp);
81 static void	igmp_sendpkt(struct in_multi *, int, unsigned long);
82 
83 static struct igmpstat igmpstat;
84 
85 SYSCTL_V_STRUCT(V_NET, vnet_inet, _net_inet_igmp, IGMPCTL_STATS,
86     stats, CTLFLAG_RW, igmpstat, igmpstat, "");
87 
88 /*
89  * igmp_mtx protects all mutable global variables in igmp.c, as well as the
90  * data fields in struct router_info.  In general, a router_info structure
91  * will be valid as long as the referencing struct in_multi is valid, so no
92  * reference counting is used.  We allow unlocked reads of router_info data
93  * when accessed via an in_multi read-only.
94  */
95 static struct mtx igmp_mtx;
96 static SLIST_HEAD(, router_info) router_info_head;
97 static int igmp_timers_are_running;
98 
99 /*
100  * XXXRW: can we define these such that these can be made const?  In any
101  * case, these shouldn't be changed after igmp_init() and therefore don't
102  * need locking.
103  */
104 static u_long igmp_all_hosts_group;
105 static u_long igmp_all_rtrs_group;
106 
107 static struct mbuf *router_alert;
108 static struct route igmprt;
109 
110 #ifdef IGMP_DEBUG
111 #define	IGMP_PRINTF(x)	printf(x)
112 #else
113 #define	IGMP_PRINTF(x)
114 #endif
115 
116 void
117 igmp_init(void)
118 {
119 	INIT_VNET_INET(curvnet);
120 	struct ipoption *ra;
121 
122 	/*
123 	 * To avoid byte-swapping the same value over and over again.
124 	 */
125 	igmp_all_hosts_group = htonl(INADDR_ALLHOSTS_GROUP);
126 	igmp_all_rtrs_group = htonl(INADDR_ALLRTRS_GROUP);
127 
128 	igmp_timers_are_running = 0;
129 
130 	/*
131 	 * Construct a Router Alert option to use in outgoing packets.
132 	 */
133 	MGET(router_alert, M_DONTWAIT, MT_DATA);
134 	ra = mtod(router_alert, struct ipoption *);
135 	ra->ipopt_dst.s_addr = 0;
136 	ra->ipopt_list[0] = IPOPT_RA;	/* Router Alert Option */
137 	ra->ipopt_list[1] = 0x04;	/* 4 bytes long */
138 	ra->ipopt_list[2] = 0x00;
139 	ra->ipopt_list[3] = 0x00;
140 	router_alert->m_len = sizeof(ra->ipopt_dst) + ra->ipopt_list[1];
141 
142 	mtx_init(&igmp_mtx, "igmp_mtx", NULL, MTX_DEF);
143 	SLIST_INIT(&V_router_info_head);
144 }
145 
146 static struct router_info *
147 find_rti(struct ifnet *ifp)
148 {
149 	INIT_VNET_INET(ifp->if_vnet);
150 	struct router_info *rti;
151 
152 	mtx_assert(&igmp_mtx, MA_OWNED);
153 	IGMP_PRINTF("[igmp.c, _find_rti] --> entering \n");
154 	SLIST_FOREACH(rti, &V_router_info_head, rti_list) {
155 		if (rti->rti_ifp == ifp) {
156 			IGMP_PRINTF(
157 			    "[igmp.c, _find_rti] --> found old entry \n");
158 			return (rti);
159 		}
160 	}
161 	rti = malloc(sizeof *rti, M_IGMP, M_NOWAIT);
162 	if (rti == NULL) {
163 		IGMP_PRINTF("[igmp.c, _find_rti] --> no memory for entry\n");
164 		return (NULL);
165 	}
166 	rti->rti_ifp = ifp;
167 	rti->rti_type = IGMP_V2_ROUTER;
168 	rti->rti_time = 0;
169 	SLIST_INSERT_HEAD(&V_router_info_head, rti, rti_list);
170 	IGMP_PRINTF("[igmp.c, _find_rti] --> created an entry \n");
171 	return (rti);
172 }
173 
174 void
175 igmp_input(register struct mbuf *m, int off)
176 {
177 	register int iphlen = off;
178 	register struct igmp *igmp;
179 	register struct ip *ip;
180 	register int igmplen;
181 	register struct ifnet *ifp = m->m_pkthdr.rcvif;
182 	register int minlen;
183 	register struct in_multi *inm;
184 	register struct in_ifaddr *ia;
185 	struct in_multistep step;
186 	struct router_info *rti;
187 	int timer; /** timer value in the igmp query header **/
188 	INIT_VNET_INET(ifp->if_vnet);
189 
190 	++V_igmpstat.igps_rcv_total;
191 
192 	ip = mtod(m, struct ip *);
193 	igmplen = ip->ip_len;
194 
195 	/*
196 	 * Validate lengths.
197 	 */
198 	if (igmplen < IGMP_MINLEN) {
199 		++V_igmpstat.igps_rcv_tooshort;
200 		m_freem(m);
201 		return;
202 	}
203 	minlen = iphlen + IGMP_MINLEN;
204 	if ((m->m_flags & M_EXT || m->m_len < minlen) &&
205 	    (m = m_pullup(m, minlen)) == 0) {
206 		++V_igmpstat.igps_rcv_tooshort;
207 		return;
208 	}
209 
210 	/*
211 	 * Validate checksum.
212 	 */
213 	m->m_data += iphlen;
214 	m->m_len -= iphlen;
215 	igmp = mtod(m, struct igmp *);
216 	if (in_cksum(m, igmplen)) {
217 		++V_igmpstat.igps_rcv_badsum;
218 		m_freem(m);
219 		return;
220 	}
221 	m->m_data -= iphlen;
222 	m->m_len += iphlen;
223 
224 	ip = mtod(m, struct ip *);
225 	timer = igmp->igmp_code * PR_FASTHZ / IGMP_TIMER_SCALE;
226 	if (timer == 0)
227 		timer = 1;
228 
229 	/*
230 	 * In the IGMPv2 specification, there are 3 states and a flag.
231 	 *
232 	 * In Non-Member state, we simply don't have a membership record.
233 	 * In Delaying Member state, our timer is running (inm->inm_timer).
234 	 * In Idle Member state, our timer is not running (inm->inm_timer==0).
235 	 *
236 	 * The flag is inm->inm_state, it is set to IGMP_OTHERMEMBER if we
237 	 * have heard a report from another member, or IGMP_IREPORTEDLAST if
238 	 * I sent the last report.
239 	 */
240 	switch (igmp->igmp_type) {
241 	case IGMP_MEMBERSHIP_QUERY:
242 		++V_igmpstat.igps_rcv_queries;
243 
244 		if (ifp->if_flags & IFF_LOOPBACK)
245 			break;
246 
247 		if (igmp->igmp_code == 0) {
248 			/*
249 			 * Old router.  Remember that the querier on this
250 			 * interface is old, and set the timer to the value
251 			 * in RFC 1112.
252 			 */
253 
254 			mtx_lock(&igmp_mtx);
255 			rti = find_rti(ifp);
256 			if (rti == NULL) {
257 				mtx_unlock(&igmp_mtx);
258 				m_freem(m);
259 				return;
260 			}
261 			rti->rti_type = IGMP_V1_ROUTER;
262 			rti->rti_time = 0;
263 			mtx_unlock(&igmp_mtx);
264 
265 			timer = IGMP_MAX_HOST_REPORT_DELAY * PR_FASTHZ;
266 
267 			if (ip->ip_dst.s_addr != igmp_all_hosts_group ||
268 			    igmp->igmp_group.s_addr != 0) {
269 				++V_igmpstat.igps_rcv_badqueries;
270 				m_freem(m);
271 				return;
272 			}
273 		} else {
274 			/*
275 			 * New router.  Simply do the new validity check.
276 			 */
277 
278 			if (igmp->igmp_group.s_addr != 0 &&
279 			    !IN_MULTICAST(ntohl(igmp->igmp_group.s_addr))) {
280 				++V_igmpstat.igps_rcv_badqueries;
281 				m_freem(m);
282 				return;
283 			}
284 		}
285 
286 		/*
287 		 * - Start the timers in all of our membership records that
288 		 *   the query applies to for the interface on which the
289 		 *   query arrived excl. those that belong to the "all-hosts"
290 		 *   group (224.0.0.1).
291 		 * - Restart any timer that is already running but has a
292 		 *   value longer than the requested timeout.
293 		 * - Use the value specified in the query message as the
294 		 *   maximum timeout.
295 		 */
296 		IN_MULTI_LOCK();
297 		IN_FIRST_MULTI(step, inm);
298 		while (inm != NULL) {
299 			if (inm->inm_ifp == ifp &&
300 			    inm->inm_addr.s_addr != igmp_all_hosts_group &&
301 			    (igmp->igmp_group.s_addr == 0 ||
302 			     igmp->igmp_group.s_addr == inm->inm_addr.s_addr)) {
303 				if (inm->inm_timer == 0 ||
304 				    inm->inm_timer > timer) {
305 					inm->inm_timer =
306 						IGMP_RANDOM_DELAY(timer);
307 					igmp_timers_are_running = 1;
308 				}
309 			}
310 			IN_NEXT_MULTI(step, inm);
311 		}
312 		IN_MULTI_UNLOCK();
313 		break;
314 
315 	case IGMP_V1_MEMBERSHIP_REPORT:
316 	case IGMP_V2_MEMBERSHIP_REPORT:
317 		/*
318 		 * For fast leave to work, we have to know that we are the
319 		 * last person to send a report for this group.  Reports can
320 		 * potentially get looped back if we are a multicast router,
321 		 * so discard reports sourced by me.
322 		 */
323 		IFP_TO_IA(ifp, ia);
324 		if (ia != NULL &&
325 		    ip->ip_src.s_addr == IA_SIN(ia)->sin_addr.s_addr)
326 			break;
327 
328 		++V_igmpstat.igps_rcv_reports;
329 
330 		if (ifp->if_flags & IFF_LOOPBACK)
331 			break;
332 
333 		if (!IN_MULTICAST(ntohl(igmp->igmp_group.s_addr))) {
334 			++V_igmpstat.igps_rcv_badreports;
335 			m_freem(m);
336 			return;
337 		}
338 
339 		/*
340 		 * KLUDGE: if the IP source address of the report has an
341 		 * unspecified (i.e., zero) subnet number, as is allowed for
342 		 * a booting host, replace it with the correct subnet number
343 		 * so that a process-level multicast routing daemon can
344 		 * determine which subnet it arrived from.  This is necessary
345 		 * to compensate for the lack of any way for a process to
346 		 * determine the arrival interface of an incoming packet.
347 		 */
348 		if ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) == 0) {
349 			if (ia != NULL)
350 				ip->ip_src.s_addr = htonl(ia->ia_subnet);
351 		}
352 
353 		/*
354 		 * If we belong to the group being reported, stop our timer
355 		 * for that group.
356 		 */
357 		IN_MULTI_LOCK();
358 		IN_LOOKUP_MULTI(igmp->igmp_group, ifp, inm);
359 		if (inm != NULL) {
360 			inm->inm_timer = 0;
361 			++V_igmpstat.igps_rcv_ourreports;
362 			inm->inm_state = IGMP_OTHERMEMBER;
363 		}
364 		IN_MULTI_UNLOCK();
365 		break;
366 	}
367 
368 	/*
369 	 * Pass all valid IGMP packets up to any process(es) listening on a
370 	 * raw IGMP socket.
371 	 */
372 	rip_input(m, off);
373 }
374 
375 void
376 igmp_joingroup(struct in_multi *inm)
377 {
378 
379 	IN_MULTI_LOCK_ASSERT();
380 
381 	if (inm->inm_addr.s_addr == igmp_all_hosts_group
382 	    || inm->inm_ifp->if_flags & IFF_LOOPBACK) {
383 		inm->inm_timer = 0;
384 		inm->inm_state = IGMP_OTHERMEMBER;
385 	} else {
386 		mtx_lock(&igmp_mtx);
387 		inm->inm_rti = find_rti(inm->inm_ifp);
388 		mtx_unlock(&igmp_mtx);
389 		if (inm->inm_rti != NULL) {
390 			igmp_sendpkt(inm, inm->inm_rti->rti_type, 0);
391 			inm->inm_timer = IGMP_RANDOM_DELAY(
392 					IGMP_MAX_HOST_REPORT_DELAY*PR_FASTHZ);
393 			inm->inm_state = IGMP_IREPORTEDLAST;
394 			igmp_timers_are_running = 1;
395 		}
396 		/* XXX handling of failure case? */
397 	}
398 }
399 
400 void
401 igmp_leavegroup(struct in_multi *inm)
402 {
403 
404 	IN_MULTI_LOCK_ASSERT();
405 
406 	if (inm->inm_state == IGMP_IREPORTEDLAST &&
407 	    inm->inm_addr.s_addr != igmp_all_hosts_group &&
408 	    !(inm->inm_ifp->if_flags & IFF_LOOPBACK) &&
409 	    inm->inm_rti->rti_type != IGMP_V1_ROUTER)
410 		igmp_sendpkt(inm, IGMP_V2_LEAVE_GROUP, igmp_all_rtrs_group);
411 }
412 
413 void
414 igmp_fasttimo(void)
415 {
416 	VNET_ITERATOR_DECL(vnet_iter);
417 	register struct in_multi *inm;
418 	struct in_multistep step;
419 
420 	/*
421 	 * Quick check to see if any work needs to be done, in order to
422 	 * minimize the overhead of fasttimo processing.
423 	 */
424 
425 	if (!igmp_timers_are_running)
426 		return;
427 
428 	IN_MULTI_LOCK();
429 	igmp_timers_are_running = 0;
430 	VNET_LIST_RLOCK();
431 	VNET_FOREACH(vnet_iter) {
432 		CURVNET_SET(vnet_iter);
433 		INIT_VNET_INET(vnet_iter);
434 		IN_FIRST_MULTI(step, inm);
435 		while (inm != NULL) {
436 			if (inm->inm_timer == 0) {
437 				/* do nothing */
438 			} else if (--inm->inm_timer == 0) {
439 				igmp_sendpkt(inm, inm->inm_rti->rti_type, 0);
440 				inm->inm_state = IGMP_IREPORTEDLAST;
441 			} else {
442 				igmp_timers_are_running = 1;
443 			}
444 			IN_NEXT_MULTI(step, inm);
445 		}
446 		CURVNET_RESTORE();
447 	}
448 	VNET_LIST_RUNLOCK();
449 	IN_MULTI_UNLOCK();
450 }
451 
452 void
453 igmp_slowtimo(void)
454 {
455 	VNET_ITERATOR_DECL(vnet_iter);
456 	struct router_info *rti;
457 
458 	IGMP_PRINTF("[igmp.c,_slowtimo] -- > entering \n");
459 	mtx_lock(&igmp_mtx);
460 	VNET_LIST_RLOCK();
461 	VNET_FOREACH(vnet_iter) {
462 		CURVNET_SET(vnet_iter);
463 		INIT_VNET_INET(vnet_iter);
464 		SLIST_FOREACH(rti, &V_router_info_head, rti_list) {
465 			if (rti->rti_type == IGMP_V1_ROUTER) {
466 				rti->rti_time++;
467 				if (rti->rti_time >= IGMP_AGE_THRESHOLD)
468 					rti->rti_type = IGMP_V2_ROUTER;
469 			}
470 		}
471 		CURVNET_RESTORE();
472 	}
473 	VNET_LIST_RUNLOCK();
474 	mtx_unlock(&igmp_mtx);
475 	IGMP_PRINTF("[igmp.c,_slowtimo] -- > exiting \n");
476 }
477 
478 static void
479 igmp_sendpkt(struct in_multi *inm, int type, unsigned long addr)
480 {
481 	INIT_VNET_NET(curvnet);
482 	INIT_VNET_INET(curvnet);
483 	struct mbuf *m;
484 	struct igmp *igmp;
485 	struct ip *ip;
486 	struct ip_moptions imo;
487 
488 	IN_MULTI_LOCK_ASSERT();
489 
490 	MGETHDR(m, M_DONTWAIT, MT_DATA);
491 	if (m == NULL)
492 		return;
493 
494 	m->m_pkthdr.rcvif = V_loif;
495 #ifdef MAC
496 	mac_netinet_igmp_send(inm->inm_ifp, m);
497 #endif
498 	m->m_pkthdr.len = sizeof(struct ip) + IGMP_MINLEN;
499 	MH_ALIGN(m, IGMP_MINLEN + sizeof(struct ip));
500 	m->m_data += sizeof(struct ip);
501 	m->m_len = IGMP_MINLEN;
502 	igmp = mtod(m, struct igmp *);
503 	igmp->igmp_type = type;
504 	igmp->igmp_code = 0;
505 	igmp->igmp_group = inm->inm_addr;
506 	igmp->igmp_cksum = 0;
507 	igmp->igmp_cksum = in_cksum(m, IGMP_MINLEN);
508 
509 	m->m_data -= sizeof(struct ip);
510 	m->m_len += sizeof(struct ip);
511 	ip = mtod(m, struct ip *);
512 	ip->ip_tos = 0;
513 	ip->ip_len = sizeof(struct ip) + IGMP_MINLEN;
514 	ip->ip_off = 0;
515 	ip->ip_p = IPPROTO_IGMP;
516 	ip->ip_src.s_addr = INADDR_ANY;
517 	ip->ip_dst.s_addr = addr ? addr : igmp->igmp_group.s_addr;
518 
519 	imo.imo_multicast_ifp  = inm->inm_ifp;
520 	imo.imo_multicast_ttl  = 1;
521 	imo.imo_multicast_vif  = -1;
522 	/*
523 	 * Request loopback of the report if we are acting as a multicast
524 	 * router, so that the process-level routing daemon can hear it.
525 	 */
526 	imo.imo_multicast_loop = (V_ip_mrouter != NULL);
527 
528 	/*
529 	 * XXX: Do we have to worry about reentrancy here?  Don't think so.
530 	 */
531 	ip_output(m, router_alert, &igmprt, 0, &imo, NULL);
532 
533 	++V_igmpstat.igps_snd_reports;
534 }
535