xref: /freebsd/sys/netinet/igmp.c (revision 7d0d268b8a67f28ccefdd0b8ce6fb38acac78d80)
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 #include <net/vnet.h>
65 
66 #include <netinet/in.h>
67 #include <netinet/in_var.h>
68 #include <netinet/in_systm.h>
69 #include <netinet/ip.h>
70 #include <netinet/ip_var.h>
71 #include <netinet/ip_options.h>
72 #include <netinet/igmp.h>
73 #include <netinet/igmp_var.h>
74 #include <netinet/vinet.h>
75 
76 #include <machine/in_cksum.h>
77 
78 #include <security/mac/mac_framework.h>
79 
80 static MALLOC_DEFINE(M_IGMP, "igmp", "igmp state");
81 
82 static struct router_info	*find_rti(struct ifnet *ifp);
83 static void	igmp_sendpkt(struct in_multi *, int, unsigned long);
84 
85 #ifdef VIMAGE_GLOBALS
86 static struct igmpstat igmpstat;
87 #endif
88 
89 SYSCTL_V_STRUCT(V_NET, vnet_inet, _net_inet_igmp, IGMPCTL_STATS,
90     stats, CTLFLAG_RW, igmpstat, igmpstat, "");
91 
92 /*
93  * igmp_mtx protects all mutable global variables in igmp.c, as well as the
94  * data fields in struct router_info.  In general, a router_info structure
95  * will be valid as long as the referencing struct in_multi is valid, so no
96  * reference counting is used.  We allow unlocked reads of router_info data
97  * when accessed via an in_multi read-only.
98  */
99 #ifdef VIMAGE_GLOBALS
100 static SLIST_HEAD(, router_info) router_info_head;
101 #endif
102 static struct mtx igmp_mtx;
103 static int igmp_timers_are_running;
104 
105 /*
106  * XXXRW: can we define these such that these can be made const?  In any
107  * case, these shouldn't be changed after igmp_init() and therefore don't
108  * need locking.
109  */
110 static u_long igmp_all_hosts_group;
111 static u_long igmp_all_rtrs_group;
112 
113 static struct mbuf *router_alert;
114 static struct route igmprt;
115 
116 #ifdef IGMP_DEBUG
117 #define	IGMP_PRINTF(x)	printf(x)
118 #else
119 #define	IGMP_PRINTF(x)
120 #endif
121 
122 void
123 igmp_init(void)
124 {
125 	INIT_VNET_INET(curvnet);
126 	struct ipoption *ra;
127 
128 	/*
129 	 * To avoid byte-swapping the same value over and over again.
130 	 */
131 	igmp_all_hosts_group = htonl(INADDR_ALLHOSTS_GROUP);
132 	igmp_all_rtrs_group = htonl(INADDR_ALLRTRS_GROUP);
133 
134 	igmp_timers_are_running = 0;
135 
136 	/*
137 	 * Construct a Router Alert option to use in outgoing packets.
138 	 */
139 	MGET(router_alert, M_DONTWAIT, MT_DATA);
140 	ra = mtod(router_alert, struct ipoption *);
141 	ra->ipopt_dst.s_addr = 0;
142 	ra->ipopt_list[0] = IPOPT_RA;	/* Router Alert Option */
143 	ra->ipopt_list[1] = 0x04;	/* 4 bytes long */
144 	ra->ipopt_list[2] = 0x00;
145 	ra->ipopt_list[3] = 0x00;
146 	router_alert->m_len = sizeof(ra->ipopt_dst) + ra->ipopt_list[1];
147 
148 	mtx_init(&igmp_mtx, "igmp_mtx", NULL, MTX_DEF);
149 	SLIST_INIT(&V_router_info_head);
150 }
151 
152 static struct router_info *
153 find_rti(struct ifnet *ifp)
154 {
155 	INIT_VNET_INET(ifp->if_vnet);
156 	struct router_info *rti;
157 
158 	mtx_assert(&igmp_mtx, MA_OWNED);
159 	IGMP_PRINTF("[igmp.c, _find_rti] --> entering \n");
160 	SLIST_FOREACH(rti, &V_router_info_head, rti_list) {
161 		if (rti->rti_ifp == ifp) {
162 			IGMP_PRINTF(
163 			    "[igmp.c, _find_rti] --> found old entry \n");
164 			return (rti);
165 		}
166 	}
167 	rti = malloc(sizeof *rti, M_IGMP, M_NOWAIT);
168 	if (rti == NULL) {
169 		IGMP_PRINTF("[igmp.c, _find_rti] --> no memory for entry\n");
170 		return (NULL);
171 	}
172 	rti->rti_ifp = ifp;
173 	rti->rti_type = IGMP_V2_ROUTER;
174 	rti->rti_time = 0;
175 	SLIST_INSERT_HEAD(&V_router_info_head, rti, rti_list);
176 	IGMP_PRINTF("[igmp.c, _find_rti] --> created an entry \n");
177 	return (rti);
178 }
179 
180 void
181 igmp_input(register struct mbuf *m, int off)
182 {
183 	register int iphlen = off;
184 	register struct igmp *igmp;
185 	register struct ip *ip;
186 	register int igmplen;
187 	register struct ifnet *ifp = m->m_pkthdr.rcvif;
188 	register int minlen;
189 	register struct in_multi *inm;
190 	register struct in_ifaddr *ia;
191 	struct in_multistep step;
192 	struct router_info *rti;
193 	int timer; /** timer value in the igmp query header **/
194 	INIT_VNET_INET(ifp->if_vnet);
195 
196 	++V_igmpstat.igps_rcv_total;
197 
198 	ip = mtod(m, struct ip *);
199 	igmplen = ip->ip_len;
200 
201 	/*
202 	 * Validate lengths.
203 	 */
204 	if (igmplen < IGMP_MINLEN) {
205 		++V_igmpstat.igps_rcv_tooshort;
206 		m_freem(m);
207 		return;
208 	}
209 	minlen = iphlen + IGMP_MINLEN;
210 	if ((m->m_flags & M_EXT || m->m_len < minlen) &&
211 	    (m = m_pullup(m, minlen)) == 0) {
212 		++V_igmpstat.igps_rcv_tooshort;
213 		return;
214 	}
215 
216 	/*
217 	 * Validate checksum.
218 	 */
219 	m->m_data += iphlen;
220 	m->m_len -= iphlen;
221 	igmp = mtod(m, struct igmp *);
222 	if (in_cksum(m, igmplen)) {
223 		++V_igmpstat.igps_rcv_badsum;
224 		m_freem(m);
225 		return;
226 	}
227 	m->m_data -= iphlen;
228 	m->m_len += iphlen;
229 
230 	ip = mtod(m, struct ip *);
231 	timer = igmp->igmp_code * PR_FASTHZ / IGMP_TIMER_SCALE;
232 	if (timer == 0)
233 		timer = 1;
234 
235 	/*
236 	 * In the IGMPv2 specification, there are 3 states and a flag.
237 	 *
238 	 * In Non-Member state, we simply don't have a membership record.
239 	 * In Delaying Member state, our timer is running (inm->inm_timer).
240 	 * In Idle Member state, our timer is not running (inm->inm_timer==0).
241 	 *
242 	 * The flag is inm->inm_state, it is set to IGMP_OTHERMEMBER if we
243 	 * have heard a report from another member, or IGMP_IREPORTEDLAST if
244 	 * I sent the last report.
245 	 */
246 	switch (igmp->igmp_type) {
247 	case IGMP_MEMBERSHIP_QUERY:
248 		++V_igmpstat.igps_rcv_queries;
249 
250 		if (ifp->if_flags & IFF_LOOPBACK)
251 			break;
252 
253 		if (igmp->igmp_code == 0) {
254 			/*
255 			 * Old router.  Remember that the querier on this
256 			 * interface is old, and set the timer to the value
257 			 * in RFC 1112.
258 			 */
259 
260 			mtx_lock(&igmp_mtx);
261 			rti = find_rti(ifp);
262 			if (rti == NULL) {
263 				mtx_unlock(&igmp_mtx);
264 				m_freem(m);
265 				return;
266 			}
267 			rti->rti_type = IGMP_V1_ROUTER;
268 			rti->rti_time = 0;
269 			mtx_unlock(&igmp_mtx);
270 
271 			timer = IGMP_MAX_HOST_REPORT_DELAY * PR_FASTHZ;
272 
273 			if (ip->ip_dst.s_addr != igmp_all_hosts_group ||
274 			    igmp->igmp_group.s_addr != 0) {
275 				++V_igmpstat.igps_rcv_badqueries;
276 				m_freem(m);
277 				return;
278 			}
279 		} else {
280 			/*
281 			 * New router.  Simply do the new validity check.
282 			 */
283 
284 			if (igmp->igmp_group.s_addr != 0 &&
285 			    !IN_MULTICAST(ntohl(igmp->igmp_group.s_addr))) {
286 				++V_igmpstat.igps_rcv_badqueries;
287 				m_freem(m);
288 				return;
289 			}
290 		}
291 
292 		/*
293 		 * - Start the timers in all of our membership records that
294 		 *   the query applies to for the interface on which the
295 		 *   query arrived excl. those that belong to the "all-hosts"
296 		 *   group (224.0.0.1).
297 		 * - Restart any timer that is already running but has a
298 		 *   value longer than the requested timeout.
299 		 * - Use the value specified in the query message as the
300 		 *   maximum timeout.
301 		 */
302 		IN_MULTI_LOCK();
303 		IN_FIRST_MULTI(step, inm);
304 		while (inm != NULL) {
305 			if (inm->inm_ifp == ifp &&
306 			    inm->inm_addr.s_addr != igmp_all_hosts_group &&
307 			    (igmp->igmp_group.s_addr == 0 ||
308 			     igmp->igmp_group.s_addr == inm->inm_addr.s_addr)) {
309 				if (inm->inm_timer == 0 ||
310 				    inm->inm_timer > timer) {
311 					inm->inm_timer =
312 						IGMP_RANDOM_DELAY(timer);
313 					igmp_timers_are_running = 1;
314 				}
315 			}
316 			IN_NEXT_MULTI(step, inm);
317 		}
318 		IN_MULTI_UNLOCK();
319 		break;
320 
321 	case IGMP_V1_MEMBERSHIP_REPORT:
322 	case IGMP_V2_MEMBERSHIP_REPORT:
323 		/*
324 		 * For fast leave to work, we have to know that we are the
325 		 * last person to send a report for this group.  Reports can
326 		 * potentially get looped back if we are a multicast router,
327 		 * so discard reports sourced by me.
328 		 */
329 		IFP_TO_IA(ifp, ia);
330 		if (ia != NULL &&
331 		    ip->ip_src.s_addr == IA_SIN(ia)->sin_addr.s_addr)
332 			break;
333 
334 		++V_igmpstat.igps_rcv_reports;
335 
336 		if (ifp->if_flags & IFF_LOOPBACK)
337 			break;
338 
339 		if (!IN_MULTICAST(ntohl(igmp->igmp_group.s_addr))) {
340 			++V_igmpstat.igps_rcv_badreports;
341 			m_freem(m);
342 			return;
343 		}
344 
345 		/*
346 		 * KLUDGE: if the IP source address of the report has an
347 		 * unspecified (i.e., zero) subnet number, as is allowed for
348 		 * a booting host, replace it with the correct subnet number
349 		 * so that a process-level multicast routing daemon can
350 		 * determine which subnet it arrived from.  This is necessary
351 		 * to compensate for the lack of any way for a process to
352 		 * determine the arrival interface of an incoming packet.
353 		 */
354 		if ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) == 0) {
355 			if (ia != NULL)
356 				ip->ip_src.s_addr = htonl(ia->ia_subnet);
357 		}
358 
359 		/*
360 		 * If we belong to the group being reported, stop our timer
361 		 * for that group.
362 		 */
363 		IN_MULTI_LOCK();
364 		IN_LOOKUP_MULTI(igmp->igmp_group, ifp, inm);
365 		if (inm != NULL) {
366 			inm->inm_timer = 0;
367 			++V_igmpstat.igps_rcv_ourreports;
368 			inm->inm_state = IGMP_OTHERMEMBER;
369 		}
370 		IN_MULTI_UNLOCK();
371 		break;
372 	}
373 
374 	/*
375 	 * Pass all valid IGMP packets up to any process(es) listening on a
376 	 * raw IGMP socket.
377 	 */
378 	rip_input(m, off);
379 }
380 
381 void
382 igmp_joingroup(struct in_multi *inm)
383 {
384 
385 	IN_MULTI_LOCK_ASSERT();
386 
387 	if (inm->inm_addr.s_addr == igmp_all_hosts_group
388 	    || inm->inm_ifp->if_flags & IFF_LOOPBACK) {
389 		inm->inm_timer = 0;
390 		inm->inm_state = IGMP_OTHERMEMBER;
391 	} else {
392 		mtx_lock(&igmp_mtx);
393 		inm->inm_rti = find_rti(inm->inm_ifp);
394 		mtx_unlock(&igmp_mtx);
395 		if (inm->inm_rti != NULL) {
396 			igmp_sendpkt(inm, inm->inm_rti->rti_type, 0);
397 			inm->inm_timer = IGMP_RANDOM_DELAY(
398 					IGMP_MAX_HOST_REPORT_DELAY*PR_FASTHZ);
399 			inm->inm_state = IGMP_IREPORTEDLAST;
400 			igmp_timers_are_running = 1;
401 		}
402 		/* XXX handling of failure case? */
403 	}
404 }
405 
406 void
407 igmp_leavegroup(struct in_multi *inm)
408 {
409 
410 	IN_MULTI_LOCK_ASSERT();
411 
412 	if (inm->inm_state == IGMP_IREPORTEDLAST &&
413 	    inm->inm_addr.s_addr != igmp_all_hosts_group &&
414 	    !(inm->inm_ifp->if_flags & IFF_LOOPBACK) &&
415 	    inm->inm_rti->rti_type != IGMP_V1_ROUTER)
416 		igmp_sendpkt(inm, IGMP_V2_LEAVE_GROUP, igmp_all_rtrs_group);
417 }
418 
419 void
420 igmp_fasttimo(void)
421 {
422 	VNET_ITERATOR_DECL(vnet_iter);
423 	register struct in_multi *inm;
424 	struct in_multistep step;
425 
426 	/*
427 	 * Quick check to see if any work needs to be done, in order to
428 	 * minimize the overhead of fasttimo processing.
429 	 */
430 
431 	if (!igmp_timers_are_running)
432 		return;
433 
434 	IN_MULTI_LOCK();
435 	igmp_timers_are_running = 0;
436 	VNET_LIST_RLOCK();
437 	VNET_FOREACH(vnet_iter) {
438 		CURVNET_SET(vnet_iter);
439 		INIT_VNET_INET(vnet_iter);
440 		IN_FIRST_MULTI(step, inm);
441 		while (inm != NULL) {
442 			if (inm->inm_timer == 0) {
443 				/* do nothing */
444 			} else if (--inm->inm_timer == 0) {
445 				igmp_sendpkt(inm, inm->inm_rti->rti_type, 0);
446 				inm->inm_state = IGMP_IREPORTEDLAST;
447 			} else {
448 				igmp_timers_are_running = 1;
449 			}
450 			IN_NEXT_MULTI(step, inm);
451 		}
452 		CURVNET_RESTORE();
453 	}
454 	VNET_LIST_RUNLOCK();
455 	IN_MULTI_UNLOCK();
456 }
457 
458 void
459 igmp_slowtimo(void)
460 {
461 	VNET_ITERATOR_DECL(vnet_iter);
462 	struct router_info *rti;
463 
464 	IGMP_PRINTF("[igmp.c,_slowtimo] -- > entering \n");
465 	mtx_lock(&igmp_mtx);
466 	VNET_LIST_RLOCK();
467 	VNET_FOREACH(vnet_iter) {
468 		CURVNET_SET(vnet_iter);
469 		INIT_VNET_INET(vnet_iter);
470 		SLIST_FOREACH(rti, &V_router_info_head, rti_list) {
471 			if (rti->rti_type == IGMP_V1_ROUTER) {
472 				rti->rti_time++;
473 				if (rti->rti_time >= IGMP_AGE_THRESHOLD)
474 					rti->rti_type = IGMP_V2_ROUTER;
475 			}
476 		}
477 		CURVNET_RESTORE();
478 	}
479 	VNET_LIST_RUNLOCK();
480 	mtx_unlock(&igmp_mtx);
481 	IGMP_PRINTF("[igmp.c,_slowtimo] -- > exiting \n");
482 }
483 
484 static void
485 igmp_sendpkt(struct in_multi *inm, int type, unsigned long addr)
486 {
487 	INIT_VNET_NET(curvnet);
488 	INIT_VNET_INET(curvnet);
489 	struct mbuf *m;
490 	struct igmp *igmp;
491 	struct ip *ip;
492 	struct ip_moptions imo;
493 
494 	IN_MULTI_LOCK_ASSERT();
495 
496 	MGETHDR(m, M_DONTWAIT, MT_DATA);
497 	if (m == NULL)
498 		return;
499 
500 	m->m_pkthdr.rcvif = V_loif;
501 #ifdef MAC
502 	mac_netinet_igmp_send(inm->inm_ifp, m);
503 #endif
504 	m->m_pkthdr.len = sizeof(struct ip) + IGMP_MINLEN;
505 	MH_ALIGN(m, IGMP_MINLEN + sizeof(struct ip));
506 	m->m_data += sizeof(struct ip);
507 	m->m_len = IGMP_MINLEN;
508 	igmp = mtod(m, struct igmp *);
509 	igmp->igmp_type = type;
510 	igmp->igmp_code = 0;
511 	igmp->igmp_group = inm->inm_addr;
512 	igmp->igmp_cksum = 0;
513 	igmp->igmp_cksum = in_cksum(m, IGMP_MINLEN);
514 
515 	m->m_data -= sizeof(struct ip);
516 	m->m_len += sizeof(struct ip);
517 	ip = mtod(m, struct ip *);
518 	ip->ip_tos = 0;
519 	ip->ip_len = sizeof(struct ip) + IGMP_MINLEN;
520 	ip->ip_off = 0;
521 	ip->ip_p = IPPROTO_IGMP;
522 	ip->ip_src.s_addr = INADDR_ANY;
523 	ip->ip_dst.s_addr = addr ? addr : igmp->igmp_group.s_addr;
524 
525 	imo.imo_multicast_ifp  = inm->inm_ifp;
526 	imo.imo_multicast_ttl  = 1;
527 	imo.imo_multicast_vif  = -1;
528 	/*
529 	 * Request loopback of the report if we are acting as a multicast
530 	 * router, so that the process-level routing daemon can hear it.
531 	 */
532 	imo.imo_multicast_loop = (V_ip_mrouter != NULL);
533 
534 	/*
535 	 * XXX: Do we have to worry about reentrancy here?  Don't think so.
536 	 */
537 	ip_output(m, router_alert, &igmprt, 0, &imo, NULL);
538 
539 	++V_igmpstat.igps_snd_reports;
540 }
541