xref: /netbsd-src/sys/netinet6/mld6.c (revision c38e7cc395b1472a774ff828e46123de44c628e9)
1 /*	$NetBSD: mld6.c,v 1.92 2018/05/01 07:21:39 maxv Exp $	*/
2 /*	$KAME: mld6.c,v 1.25 2001/01/16 14:14:18 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1992, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * This code is derived from software contributed to Berkeley by
38  * Stephen Deering of Stanford University.
39  *
40  * Redistribution and use in source and binary forms, with or without
41  * modification, are permitted provided that the following conditions
42  * are met:
43  * 1. Redistributions of source code must retain the above copyright
44  *    notice, this list of conditions and the following disclaimer.
45  * 2. Redistributions in binary form must reproduce the above copyright
46  *    notice, this list of conditions and the following disclaimer in the
47  *    documentation and/or other materials provided with the distribution.
48  * 3. Neither the name of the University nor the names of its contributors
49  *    may be used to endorse or promote products derived from this software
50  *    without specific prior written permission.
51  *
52  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
53  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
54  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
55  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
56  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
57  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
58  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
59  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
60  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
61  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62  * SUCH DAMAGE.
63  *
64  *	@(#)igmp.c	8.1 (Berkeley) 7/19/93
65  */
66 
67 /*
68  * Copyright (c) 1988 Stephen Deering.
69  *
70  * This code is derived from software contributed to Berkeley by
71  * Stephen Deering of Stanford University.
72  *
73  * Redistribution and use in source and binary forms, with or without
74  * modification, are permitted provided that the following conditions
75  * are met:
76  * 1. Redistributions of source code must retain the above copyright
77  *    notice, this list of conditions and the following disclaimer.
78  * 2. Redistributions in binary form must reproduce the above copyright
79  *    notice, this list of conditions and the following disclaimer in the
80  *    documentation and/or other materials provided with the distribution.
81  * 3. All advertising materials mentioning features or use of this software
82  *    must display the following acknowledgement:
83  *	This product includes software developed by the University of
84  *	California, Berkeley and its contributors.
85  * 4. Neither the name of the University nor the names of its contributors
86  *    may be used to endorse or promote products derived from this software
87  *    without specific prior written permission.
88  *
89  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
90  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
91  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
92  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
93  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
94  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
95  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
96  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
97  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
98  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
99  * SUCH DAMAGE.
100  *
101  *	@(#)igmp.c	8.1 (Berkeley) 7/19/93
102  */
103 
104 #include <sys/cdefs.h>
105 __KERNEL_RCSID(0, "$NetBSD: mld6.c,v 1.92 2018/05/01 07:21:39 maxv Exp $");
106 
107 #ifdef _KERNEL_OPT
108 #include "opt_inet.h"
109 #include "opt_net_mpsafe.h"
110 #endif
111 
112 #include <sys/param.h>
113 #include <sys/systm.h>
114 #include <sys/mbuf.h>
115 #include <sys/socket.h>
116 #include <sys/socketvar.h>
117 #include <sys/syslog.h>
118 #include <sys/sysctl.h>
119 #include <sys/kernel.h>
120 #include <sys/callout.h>
121 #include <sys/cprng.h>
122 #include <sys/rwlock.h>
123 
124 #include <net/if.h>
125 
126 #include <netinet/in.h>
127 #include <netinet/in_var.h>
128 #include <netinet6/in6_var.h>
129 #include <netinet/ip6.h>
130 #include <netinet6/ip6_var.h>
131 #include <netinet6/scope6_var.h>
132 #include <netinet/icmp6.h>
133 #include <netinet6/icmp6_private.h>
134 #include <netinet6/mld6_var.h>
135 
136 static krwlock_t	in6_multilock __cacheline_aligned;
137 
138 /*
139  * Protocol constants
140  */
141 
142 /*
143  * time between repetitions of a node's initial report of interest in a
144  * multicast address(in seconds)
145  */
146 #define MLD_UNSOLICITED_REPORT_INTERVAL	10
147 
148 static struct ip6_pktopts ip6_opts;
149 
150 static void mld_start_listening(struct in6_multi *);
151 static void mld_stop_listening(struct in6_multi *);
152 
153 static struct mld_hdr *mld_allocbuf(struct mbuf **, struct in6_multi *, int);
154 static void mld_sendpkt(struct in6_multi *, int, const struct in6_addr *);
155 static void mld_starttimer(struct in6_multi *);
156 static void mld_stoptimer(struct in6_multi *);
157 static u_long mld_timerresid(struct in6_multi *);
158 
159 static void in6m_ref(struct in6_multi *);
160 static void in6m_unref(struct in6_multi *);
161 static void in6m_destroy(struct in6_multi *);
162 
163 void
164 mld_init(void)
165 {
166 	static u_int8_t hbh_buf[8];
167 	struct ip6_hbh *hbh = (struct ip6_hbh *)hbh_buf;
168 	u_int16_t rtalert_code = htons((u_int16_t)IP6OPT_RTALERT_MLD);
169 
170 	/* ip6h_nxt will be fill in later */
171 	hbh->ip6h_len = 0;	/* (8 >> 3) - 1 */
172 
173 	/* XXX: grotty hard coding... */
174 	hbh_buf[2] = IP6OPT_PADN;	/* 2 byte padding */
175 	hbh_buf[3] = 0;
176 	hbh_buf[4] = IP6OPT_RTALERT;
177 	hbh_buf[5] = IP6OPT_RTALERT_LEN - 2;
178 	memcpy(&hbh_buf[6], (void *)&rtalert_code, sizeof(u_int16_t));
179 
180 	ip6_opts.ip6po_hbh = hbh;
181 	/* We will specify the hoplimit by a multicast option. */
182 	ip6_opts.ip6po_hlim = -1;
183 	ip6_opts.ip6po_prefer_tempaddr = IP6PO_TEMPADDR_NOTPREFER;
184 
185 	rw_init(&in6_multilock);
186 }
187 
188 static void
189 mld_starttimer(struct in6_multi *in6m)
190 {
191 	struct timeval now;
192 
193 	KASSERT(rw_write_held(&in6_multilock));
194 	KASSERT(in6m->in6m_timer != IN6M_TIMER_UNDEF);
195 
196 	microtime(&now);
197 	in6m->in6m_timer_expire.tv_sec = now.tv_sec + in6m->in6m_timer / hz;
198 	in6m->in6m_timer_expire.tv_usec = now.tv_usec +
199 	    (in6m->in6m_timer % hz) * (1000000 / hz);
200 	if (in6m->in6m_timer_expire.tv_usec > 1000000) {
201 		in6m->in6m_timer_expire.tv_sec++;
202 		in6m->in6m_timer_expire.tv_usec -= 1000000;
203 	}
204 
205 	/* start or restart the timer */
206 	callout_schedule(&in6m->in6m_timer_ch, in6m->in6m_timer);
207 }
208 
209 /*
210  * mld_stoptimer releases in6_multilock when calling callout_halt.
211  * The caller must ensure in6m won't be freed while releasing the lock.
212  */
213 static void
214 mld_stoptimer(struct in6_multi *in6m)
215 {
216 
217 	KASSERT(rw_write_held(&in6_multilock));
218 
219 	if (in6m->in6m_timer == IN6M_TIMER_UNDEF)
220 		return;
221 
222 	rw_exit(&in6_multilock);
223 
224 	if (mutex_owned(softnet_lock))
225 		callout_halt(&in6m->in6m_timer_ch, softnet_lock);
226 	else
227 		callout_halt(&in6m->in6m_timer_ch, NULL);
228 
229 	rw_enter(&in6_multilock, RW_WRITER);
230 
231 	in6m->in6m_timer = IN6M_TIMER_UNDEF;
232 }
233 
234 static void
235 mld_timeo(void *arg)
236 {
237 	struct in6_multi *in6m = arg;
238 
239 	KASSERT(in6m->in6m_refcount > 0);
240 
241 	SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE();
242 	rw_enter(&in6_multilock, RW_WRITER);
243 	if (in6m->in6m_timer == IN6M_TIMER_UNDEF)
244 		goto out;
245 
246 	in6m->in6m_timer = IN6M_TIMER_UNDEF;
247 
248 	switch (in6m->in6m_state) {
249 	case MLD_REPORTPENDING:
250 		mld_start_listening(in6m);
251 		break;
252 	default:
253 		mld_sendpkt(in6m, MLD_LISTENER_REPORT, NULL);
254 		break;
255 	}
256 
257 out:
258 	rw_exit(&in6_multilock);
259 	SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
260 }
261 
262 static u_long
263 mld_timerresid(struct in6_multi *in6m)
264 {
265 	struct timeval now, diff;
266 
267 	microtime(&now);
268 
269 	if (now.tv_sec > in6m->in6m_timer_expire.tv_sec ||
270 	    (now.tv_sec == in6m->in6m_timer_expire.tv_sec &&
271 	    now.tv_usec > in6m->in6m_timer_expire.tv_usec)) {
272 		return (0);
273 	}
274 	diff = in6m->in6m_timer_expire;
275 	diff.tv_sec -= now.tv_sec;
276 	diff.tv_usec -= now.tv_usec;
277 	if (diff.tv_usec < 0) {
278 		diff.tv_sec--;
279 		diff.tv_usec += 1000000;
280 	}
281 
282 	/* return the remaining time in milliseconds */
283 	return diff.tv_sec * 1000 + diff.tv_usec / 1000;
284 }
285 
286 static void
287 mld_start_listening(struct in6_multi *in6m)
288 {
289 	struct in6_addr all_in6;
290 
291 	KASSERT(rw_write_held(&in6_multilock));
292 
293 	/*
294 	 * RFC2710 page 10:
295 	 * The node never sends a Report or Done for the link-scope all-nodes
296 	 * address.
297 	 * MLD messages are never sent for multicast addresses whose scope is 0
298 	 * (reserved) or 1 (node-local).
299 	 */
300 	all_in6 = in6addr_linklocal_allnodes;
301 	if (in6_setscope(&all_in6, in6m->in6m_ifp, NULL)) {
302 		/* XXX: this should not happen! */
303 		in6m->in6m_timer = 0;
304 		in6m->in6m_state = MLD_OTHERLISTENER;
305 	}
306 	if (IN6_ARE_ADDR_EQUAL(&in6m->in6m_addr, &all_in6) ||
307 	    IPV6_ADDR_MC_SCOPE(&in6m->in6m_addr) < IPV6_ADDR_SCOPE_LINKLOCAL) {
308 		in6m->in6m_timer = IN6M_TIMER_UNDEF;
309 		in6m->in6m_state = MLD_OTHERLISTENER;
310 	} else {
311 		mld_sendpkt(in6m, MLD_LISTENER_REPORT, NULL);
312 		in6m->in6m_timer = cprng_fast32() %
313 		    (MLD_UNSOLICITED_REPORT_INTERVAL * hz);
314 		in6m->in6m_state = MLD_IREPORTEDLAST;
315 
316 		mld_starttimer(in6m);
317 	}
318 }
319 
320 static void
321 mld_stop_listening(struct in6_multi *in6m)
322 {
323 	struct in6_addr allnode, allrouter;
324 
325 	KASSERT(rw_lock_held(&in6_multilock));
326 
327 	allnode = in6addr_linklocal_allnodes;
328 	if (in6_setscope(&allnode, in6m->in6m_ifp, NULL)) {
329 		/* XXX: this should not happen! */
330 		return;
331 	}
332 	allrouter = in6addr_linklocal_allrouters;
333 	if (in6_setscope(&allrouter, in6m->in6m_ifp, NULL)) {
334 		/* XXX impossible */
335 		return;
336 	}
337 
338 	if (in6m->in6m_state == MLD_IREPORTEDLAST &&
339 	    (!IN6_ARE_ADDR_EQUAL(&in6m->in6m_addr, &allnode)) &&
340 	    IPV6_ADDR_MC_SCOPE(&in6m->in6m_addr) >
341 	    IPV6_ADDR_SCOPE_INTFACELOCAL) {
342 		mld_sendpkt(in6m, MLD_LISTENER_DONE, &allrouter);
343 	}
344 }
345 
346 void
347 mld_input(struct mbuf *m, int off)
348 {
349 	struct ip6_hdr *ip6;
350 	struct mld_hdr *mldh;
351 	struct ifnet *ifp;
352 	struct in6_multi *in6m = NULL;
353 	struct in6_addr mld_addr, all_in6;
354 	u_long timer = 0;	/* timer value in the MLD query header */
355 	struct psref psref;
356 
357 	ifp = m_get_rcvif_psref(m, &psref);
358 	if (__predict_false(ifp == NULL))
359 		goto out;
360 	IP6_EXTHDR_GET(mldh, struct mld_hdr *, m, off, sizeof(*mldh));
361 	if (mldh == NULL) {
362 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
363 		goto out_nodrop;
364 	}
365 
366 	ip6 = mtod(m, struct ip6_hdr *);
367 
368 	/* source address validation */
369 	if (!IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src)) {
370 		/*
371 		 * RFC3590 allows the IPv6 unspecified address as the source
372 		 * address of MLD report and done messages.  However, as this
373 		 * same document says, this special rule is for snooping
374 		 * switches and the RFC requires routers to discard MLD packets
375 		 * with the unspecified source address.  The RFC only talks
376 		 * about hosts receiving an MLD query or report in Security
377 		 * Considerations, but this is probably the correct intention.
378 		 * RFC3590 does not talk about other cases than link-local and
379 		 * the unspecified source addresses, but we believe the same
380 		 * rule should be applied.
381 		 * As a result, we only allow link-local addresses as the
382 		 * source address; otherwise, simply discard the packet.
383 		 */
384 #if 0
385 		/*
386 		 * XXX: do not log in an input path to avoid log flooding,
387 		 * though RFC3590 says "SHOULD log" if the source of a query
388 		 * is the unspecified address.
389 		 */
390 		char ip6bufs[INET6_ADDRSTRLEN];
391 		char ip6bufm[INET6_ADDRSTRLEN];
392 		log(LOG_INFO,
393 		    "mld_input: src %s is not link-local (grp=%s)\n",
394 		    IN6_PRINT(ip6bufs,&ip6->ip6_src),
395 		    IN6_PRINT(ip6bufm, &mldh->mld_addr));
396 #endif
397 		goto out;
398 	}
399 
400 	/*
401 	 * make a copy for local work (in6_setscope() may modify the 1st arg)
402 	 */
403 	mld_addr = mldh->mld_addr;
404 	if (in6_setscope(&mld_addr, ifp, NULL)) {
405 		/* XXX: this should not happen! */
406 		goto out;
407 	}
408 
409 	/*
410 	 * In the MLD specification, there are 3 states and a flag.
411 	 *
412 	 * In Non-Listener state, we simply don't have a membership record.
413 	 * In Delaying Listener state, our timer is running (in6m->in6m_timer)
414 	 * In Idle Listener state, our timer is not running
415 	 * (in6m->in6m_timer==IN6M_TIMER_UNDEF)
416 	 *
417 	 * The flag is in6m->in6m_state, it is set to MLD_OTHERLISTENER if
418 	 * we have heard a report from another member, or MLD_IREPORTEDLAST
419 	 * if we sent the last report.
420 	 */
421 	switch (mldh->mld_type) {
422 	case MLD_LISTENER_QUERY: {
423 		struct in6_multi *next;
424 
425 		if (ifp->if_flags & IFF_LOOPBACK)
426 			break;
427 
428 		if (!IN6_IS_ADDR_UNSPECIFIED(&mld_addr) &&
429 		    !IN6_IS_ADDR_MULTICAST(&mld_addr))
430 			break;	/* print error or log stat? */
431 
432 		all_in6 = in6addr_linklocal_allnodes;
433 		if (in6_setscope(&all_in6, ifp, NULL)) {
434 			/* XXX: this should not happen! */
435 			break;
436 		}
437 
438 		/*
439 		 * - Start the timers in all of our membership records
440 		 *   that the query applies to for the interface on
441 		 *   which the query arrived excl. those that belong
442 		 *   to the "all-nodes" group (ff02::1).
443 		 * - Restart any timer that is already running but has
444 		 *   a value longer than the requested timeout.
445 		 * - Use the value specified in the query message as
446 		 *   the maximum timeout.
447 		 */
448 		timer = ntohs(mldh->mld_maxdelay);
449 
450 		rw_enter(&in6_multilock, RW_WRITER);
451 		/*
452 		 * mld_stoptimer and mld_sendpkt release in6_multilock
453 		 * temporarily, so we have to prevent in6m from being freed
454 		 * while releasing the lock by having an extra reference to it.
455 		 *
456 		 * Also in6_purge_multi might remove items from the list of the
457 		 * ifp while releasing the lock. Fortunately in6_purge_multi is
458 		 * never executed as long as we have a psref of the ifp.
459 		 */
460 		LIST_FOREACH_SAFE(in6m, &ifp->if_multiaddrs, in6m_entry, next) {
461 			if (IN6_ARE_ADDR_EQUAL(&in6m->in6m_addr, &all_in6) ||
462 			    IPV6_ADDR_MC_SCOPE(&in6m->in6m_addr) <
463 			    IPV6_ADDR_SCOPE_LINKLOCAL)
464 				continue;
465 
466 			if (in6m->in6m_state == MLD_REPORTPENDING)
467 				continue; /* we are not yet ready */
468 
469 			if (!IN6_IS_ADDR_UNSPECIFIED(&mld_addr) &&
470 			    !IN6_ARE_ADDR_EQUAL(&mld_addr, &in6m->in6m_addr))
471 				continue;
472 
473 			if (timer == 0) {
474 				in6m_ref(in6m);
475 
476 				/* send a report immediately */
477 				mld_stoptimer(in6m);
478 				mld_sendpkt(in6m, MLD_LISTENER_REPORT, NULL);
479 				in6m->in6m_state = MLD_IREPORTEDLAST;
480 
481 				in6m_unref(in6m); /* May free in6m */
482 			} else if (in6m->in6m_timer == IN6M_TIMER_UNDEF ||
483 			    mld_timerresid(in6m) > timer) {
484 				in6m->in6m_timer =
485 				   1 + (cprng_fast32() % timer) * hz / 1000;
486 				mld_starttimer(in6m);
487 			}
488 		}
489 		rw_exit(&in6_multilock);
490 		break;
491 	    }
492 
493 	case MLD_LISTENER_REPORT:
494 		/*
495 		 * For fast leave to work, we have to know that we are the
496 		 * last person to send a report for this group.  Reports
497 		 * can potentially get looped back if we are a multicast
498 		 * router, so discard reports sourced by me.
499 		 * Note that it is impossible to check IFF_LOOPBACK flag of
500 		 * ifp for this purpose, since ip6_mloopback pass the physical
501 		 * interface to looutput.
502 		 */
503 		if (m->m_flags & M_LOOP) /* XXX: grotty flag, but efficient */
504 			break;
505 
506 		if (!IN6_IS_ADDR_MULTICAST(&mldh->mld_addr))
507 			break;
508 
509 		/*
510 		 * If we belong to the group being reported, stop
511 		 * our timer for that group.
512 		 */
513 		rw_enter(&in6_multilock, RW_WRITER);
514 		in6m = in6_lookup_multi(&mld_addr, ifp);
515 		if (in6m) {
516 			in6m_ref(in6m);
517 			mld_stoptimer(in6m); /* transit to idle state */
518 			in6m->in6m_state = MLD_OTHERLISTENER; /* clear flag */
519 			in6m_unref(in6m);
520 			in6m = NULL; /* in6m might be freed */
521 		}
522 		rw_exit(&in6_multilock);
523 		break;
524 	default:		/* this is impossible */
525 #if 0
526 		/*
527 		 * this case should be impossible because of filtering in
528 		 * icmp6_input().  But we explicitly disabled this part
529 		 * just in case.
530 		 */
531 		log(LOG_ERR, "mld_input: illegal type(%d)", mldh->mld_type);
532 #endif
533 		break;
534 	}
535 
536 out:
537 	m_freem(m);
538 out_nodrop:
539 	m_put_rcvif_psref(ifp, &psref);
540 }
541 
542 /*
543  * XXX mld_sendpkt must be called with in6_multilock held and
544  * will release in6_multilock before calling ip6_output and
545  * returning to avoid locking against myself in ip6_output.
546  */
547 static void
548 mld_sendpkt(struct in6_multi *in6m, int type, const struct in6_addr *dst)
549 {
550 	struct mbuf *mh;
551 	struct mld_hdr *mldh;
552 	struct ip6_hdr *ip6 = NULL;
553 	struct ip6_moptions im6o;
554 	struct in6_ifaddr *ia = NULL;
555 	struct ifnet *ifp = in6m->in6m_ifp;
556 	int ignflags;
557 	struct psref psref;
558 	int bound;
559 
560 	KASSERT(rw_write_held(&in6_multilock));
561 
562 	/*
563 	 * At first, find a link local address on the outgoing interface
564 	 * to use as the source address of the MLD packet.
565 	 * We do not reject tentative addresses for MLD report to deal with
566 	 * the case where we first join a link-local address.
567 	 */
568 	ignflags = (IN6_IFF_NOTREADY|IN6_IFF_ANYCAST) & ~IN6_IFF_TENTATIVE;
569 	bound = curlwp_bind();
570 	ia = in6ifa_ifpforlinklocal_psref(ifp, ignflags, &psref);
571 	if (ia == NULL) {
572 		curlwp_bindx(bound);
573 		return;
574 	}
575 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
576 		ia6_release(ia, &psref);
577 		ia = NULL;
578 	}
579 
580 	/* Allocate two mbufs to store IPv6 header and MLD header */
581 	mldh = mld_allocbuf(&mh, in6m, type);
582 	if (mldh == NULL) {
583 		ia6_release(ia, &psref);
584 		curlwp_bindx(bound);
585 		return;
586 	}
587 
588 	/* fill src/dst here */
589 	ip6 = mtod(mh, struct ip6_hdr *);
590 	ip6->ip6_src = ia ? ia->ia_addr.sin6_addr : in6addr_any;
591 	ip6->ip6_dst = dst ? *dst : in6m->in6m_addr;
592 	ia6_release(ia, &psref);
593 	curlwp_bindx(bound);
594 
595 	mldh->mld_addr = in6m->in6m_addr;
596 	in6_clearscope(&mldh->mld_addr); /* XXX */
597 	mldh->mld_cksum = in6_cksum(mh, IPPROTO_ICMPV6, sizeof(struct ip6_hdr),
598 	    sizeof(struct mld_hdr));
599 
600 	/* construct multicast option */
601 	memset(&im6o, 0, sizeof(im6o));
602 	im6o.im6o_multicast_if_index = if_get_index(ifp);
603 	im6o.im6o_multicast_hlim = 1;
604 
605 	/*
606 	 * Request loopback of the report if we are acting as a multicast
607 	 * router, so that the process-level routing daemon can hear it.
608 	 */
609 	im6o.im6o_multicast_loop = (ip6_mrouter != NULL);
610 
611 	/* increment output statistics */
612 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + type);
613 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
614 	switch (type) {
615 	case MLD_LISTENER_QUERY:
616 		icmp6_ifstat_inc(ifp, ifs6_out_mldquery);
617 		break;
618 	case MLD_LISTENER_REPORT:
619 		icmp6_ifstat_inc(ifp, ifs6_out_mldreport);
620 		break;
621 	case MLD_LISTENER_DONE:
622 		icmp6_ifstat_inc(ifp, ifs6_out_mlddone);
623 		break;
624 	}
625 
626 	/* XXX we cannot call ip6_output with holding in6_multilock */
627 	rw_exit(&in6_multilock);
628 
629 	ip6_output(mh, &ip6_opts, NULL, ia ? 0 : IPV6_UNSPECSRC,
630 	    &im6o, NULL, NULL);
631 
632 	rw_enter(&in6_multilock, RW_WRITER);
633 }
634 
635 static struct mld_hdr *
636 mld_allocbuf(struct mbuf **mh, struct in6_multi *in6m, int type)
637 {
638 	struct mbuf *md;
639 	struct mld_hdr *mldh;
640 	struct ip6_hdr *ip6;
641 
642 	/*
643 	 * Allocate mbufs to store ip6 header and MLD header.
644 	 * We allocate 2 mbufs and make chain in advance because
645 	 * it is more convenient when inserting the hop-by-hop option later.
646 	 */
647 	MGETHDR(*mh, M_DONTWAIT, MT_HEADER);
648 	if (*mh == NULL)
649 		return NULL;
650 	MGET(md, M_DONTWAIT, MT_DATA);
651 	if (md == NULL) {
652 		m_free(*mh);
653 		*mh = NULL;
654 		return NULL;
655 	}
656 	(*mh)->m_next = md;
657 	md->m_next = NULL;
658 
659 	m_reset_rcvif((*mh));
660 	(*mh)->m_pkthdr.len = sizeof(struct ip6_hdr) + sizeof(struct mld_hdr);
661 	(*mh)->m_len = sizeof(struct ip6_hdr);
662 	MH_ALIGN(*mh, sizeof(struct ip6_hdr));
663 
664 	/* fill in the ip6 header */
665 	ip6 = mtod(*mh, struct ip6_hdr *);
666 	memset(ip6, 0, sizeof(*ip6));
667 	ip6->ip6_flow = 0;
668 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
669 	ip6->ip6_vfc |= IPV6_VERSION;
670 	/* ip6_plen will be set later */
671 	ip6->ip6_nxt = IPPROTO_ICMPV6;
672 	/* ip6_hlim will be set by im6o.im6o_multicast_hlim */
673 	/* ip6_src/dst will be set by mld_sendpkt() or mld_sendbuf() */
674 
675 	/* fill in the MLD header as much as possible */
676 	md->m_len = sizeof(struct mld_hdr);
677 	mldh = mtod(md, struct mld_hdr *);
678 	memset(mldh, 0, sizeof(struct mld_hdr));
679 	mldh->mld_type = type;
680 	return mldh;
681 }
682 
683 static void
684 in6m_ref(struct in6_multi *in6m)
685 {
686 
687 	KASSERT(rw_write_held(&in6_multilock));
688 	in6m->in6m_refcount++;
689 }
690 
691 static void
692 in6m_unref(struct in6_multi *in6m)
693 {
694 
695 	KASSERT(rw_write_held(&in6_multilock));
696 	if (--in6m->in6m_refcount == 0)
697 		in6m_destroy(in6m);
698 }
699 
700 /*
701  * Add an address to the list of IP6 multicast addresses for a given interface.
702  */
703 struct	in6_multi *
704 in6_addmulti(struct in6_addr *maddr6, struct ifnet *ifp, int *errorp,
705     int timer)
706 {
707 	struct	sockaddr_in6 sin6;
708 	struct	in6_multi *in6m;
709 
710 	*errorp = 0;
711 
712 	rw_enter(&in6_multilock, RW_WRITER);
713 	/*
714 	 * See if address already in list.
715 	 */
716 	in6m = in6_lookup_multi(maddr6, ifp);
717 	if (in6m != NULL) {
718 		/*
719 		 * Found it; just increment the reference count.
720 		 */
721 		in6m->in6m_refcount++;
722 	} else {
723 		/*
724 		 * New address; allocate a new multicast record
725 		 * and link it into the interface's multicast list.
726 		 */
727 		in6m = malloc(sizeof(*in6m), M_IPMADDR, M_NOWAIT|M_ZERO);
728 		if (in6m == NULL) {
729 			*errorp = ENOBUFS;
730 			goto out;
731 		}
732 
733 		in6m->in6m_addr = *maddr6;
734 		in6m->in6m_ifp = ifp;
735 		in6m->in6m_refcount = 1;
736 		in6m->in6m_timer = IN6M_TIMER_UNDEF;
737 		callout_init(&in6m->in6m_timer_ch, CALLOUT_MPSAFE);
738 		callout_setfunc(&in6m->in6m_timer_ch, mld_timeo, in6m);
739 
740 		LIST_INSERT_HEAD(&ifp->if_multiaddrs, in6m, in6m_entry);
741 
742 		/*
743 		 * Ask the network driver to update its multicast reception
744 		 * filter appropriately for the new address.
745 		 */
746 		sockaddr_in6_init(&sin6, maddr6, 0, 0, 0);
747 		*errorp = if_mcast_op(ifp, SIOCADDMULTI, sin6tosa(&sin6));
748 		if (*errorp) {
749 			callout_destroy(&in6m->in6m_timer_ch);
750 			LIST_REMOVE(in6m, in6m_entry);
751 			free(in6m, M_IPMADDR);
752 			in6m = NULL;
753 			goto out;
754 		}
755 
756 		in6m->in6m_timer = timer;
757 		if (in6m->in6m_timer > 0) {
758 			in6m->in6m_state = MLD_REPORTPENDING;
759 			mld_starttimer(in6m);
760 			goto out;
761 		}
762 
763 		/*
764 		 * Let MLD6 know that we have joined a new IP6 multicast
765 		 * group.
766 		 */
767 		mld_start_listening(in6m);
768 	}
769 out:
770 	rw_exit(&in6_multilock);
771 	return in6m;
772 }
773 
774 static void
775 in6m_destroy(struct in6_multi *in6m)
776 {
777 	struct sockaddr_in6 sin6;
778 
779 	KASSERT(rw_write_held(&in6_multilock));
780 	KASSERT(in6m->in6m_refcount == 0);
781 
782 	/*
783 	 * No remaining claims to this record; let MLD6 know
784 	 * that we are leaving the multicast group.
785 	 */
786 	mld_stop_listening(in6m);
787 
788 	/*
789 	 * Unlink from list.
790 	 */
791 	LIST_REMOVE(in6m, in6m_entry);
792 
793 	/*
794 	 * Delete all references of this multicasting group from
795 	 * the membership arrays
796 	 */
797 	in6_purge_mcast_references(in6m);
798 
799 	/*
800 	 * Notify the network driver to update its multicast
801 	 * reception filter.
802 	 */
803 	sockaddr_in6_init(&sin6, &in6m->in6m_addr, 0, 0, 0);
804 	if_mcast_op(in6m->in6m_ifp, SIOCDELMULTI, sin6tosa(&sin6));
805 
806 	/* Tell mld_timeo we're halting the timer */
807 	in6m->in6m_timer = IN6M_TIMER_UNDEF;
808 	if (mutex_owned(softnet_lock))
809 		callout_halt(&in6m->in6m_timer_ch, softnet_lock);
810 	else
811 		callout_halt(&in6m->in6m_timer_ch, NULL);
812 	callout_destroy(&in6m->in6m_timer_ch);
813 
814 	free(in6m, M_IPMADDR);
815 }
816 
817 /*
818  * Delete a multicast address record.
819  */
820 void
821 in6_delmulti(struct in6_multi *in6m)
822 {
823 
824 	KASSERT(in6m->in6m_refcount > 0);
825 
826 	rw_enter(&in6_multilock, RW_WRITER);
827 	/*
828 	 * The caller should have a reference to in6m. So we don't need to care
829 	 * of releasing the lock in mld_stoptimer.
830 	 */
831 	mld_stoptimer(in6m);
832 	if (--in6m->in6m_refcount == 0)
833 		in6m_destroy(in6m);
834 	rw_exit(&in6_multilock);
835 }
836 
837 /*
838  * Look up the in6_multi record for a given IP6 multicast address
839  * on a given interface. If no matching record is found, "in6m"
840  * returns NULL.
841  */
842 struct in6_multi *
843 in6_lookup_multi(const struct in6_addr *addr, const struct ifnet *ifp)
844 {
845 	struct in6_multi *in6m;
846 
847 	KASSERT(rw_lock_held(&in6_multilock));
848 
849 	LIST_FOREACH(in6m, &ifp->if_multiaddrs, in6m_entry) {
850 		if (IN6_ARE_ADDR_EQUAL(&in6m->in6m_addr, addr))
851 			break;
852 	}
853 	return in6m;
854 }
855 
856 bool
857 in6_multi_group(const struct in6_addr *addr, const struct ifnet *ifp)
858 {
859 	bool ingroup;
860 
861 	rw_enter(&in6_multilock, RW_READER);
862 	ingroup = in6_lookup_multi(addr, ifp) != NULL;
863 	rw_exit(&in6_multilock);
864 
865 	return ingroup;
866 }
867 
868 /*
869  * Purge in6_multi records associated to the interface.
870  */
871 void
872 in6_purge_multi(struct ifnet *ifp)
873 {
874 	struct in6_multi *in6m, *next;
875 
876 	rw_enter(&in6_multilock, RW_WRITER);
877 	LIST_FOREACH_SAFE(in6m, &ifp->if_multiaddrs, in6m_entry, next) {
878 		/*
879 		 * Normally multicast addresses are already purged at this
880 		 * point. Remaining references aren't accessible via ifp,
881 		 * so what we can do here is to prevent ifp from being
882 		 * accessed via in6m by removing it from the list of ifp.
883 		 */
884 		mld_stoptimer(in6m);
885 		LIST_REMOVE(in6m, in6m_entry);
886 	}
887 	rw_exit(&in6_multilock);
888 }
889 
890 void
891 in6_multi_lock(int op)
892 {
893 
894 	rw_enter(&in6_multilock, op);
895 }
896 
897 void
898 in6_multi_unlock(void)
899 {
900 
901 	rw_exit(&in6_multilock);
902 }
903 
904 bool
905 in6_multi_locked(int op)
906 {
907 
908 	switch (op) {
909 	case RW_READER:
910 		return rw_read_held(&in6_multilock);
911 	case RW_WRITER:
912 		return rw_write_held(&in6_multilock);
913 	default:
914 		return rw_lock_held(&in6_multilock);
915 	}
916 }
917 
918 struct in6_multi_mship *
919 in6_joingroup(struct ifnet *ifp, struct in6_addr *addr, int *errorp, int timer)
920 {
921 	struct in6_multi_mship *imm;
922 
923 	imm = malloc(sizeof(*imm), M_IPMADDR, M_NOWAIT|M_ZERO);
924 	if (imm == NULL) {
925 		*errorp = ENOBUFS;
926 		return NULL;
927 	}
928 
929 	imm->i6mm_maddr = in6_addmulti(addr, ifp, errorp, timer);
930 	if (!imm->i6mm_maddr) {
931 		/* *errorp is already set */
932 		free(imm, M_IPMADDR);
933 		return NULL;
934 	}
935 	return imm;
936 }
937 
938 int
939 in6_leavegroup(struct in6_multi_mship *imm)
940 {
941 	struct in6_multi *in6m;
942 
943 	rw_enter(&in6_multilock, RW_READER);
944 	in6m = imm->i6mm_maddr;
945 	rw_exit(&in6_multilock);
946 	if (in6m != NULL) {
947 		in6_delmulti(in6m);
948 	}
949 	free(imm, M_IPMADDR);
950 	return 0;
951 }
952 
953 /*
954  * DEPRECATED: keep it just to avoid breaking old sysctl users.
955  */
956 static int
957 in6_mkludge_sysctl(SYSCTLFN_ARGS)
958 {
959 
960 	if (namelen != 1)
961 		return EINVAL;
962 	*oldlenp = 0;
963 	return 0;
964 }
965 
966 static int
967 in6_multicast_sysctl(SYSCTLFN_ARGS)
968 {
969 	struct ifnet *ifp;
970 	struct ifaddr *ifa;
971 	struct in6_ifaddr *ia6;
972 	struct in6_multi *in6m;
973 	uint32_t tmp;
974 	int error;
975 	size_t written;
976 	struct psref psref, psref_ia;
977 	int bound, s;
978 
979 	if (namelen != 1)
980 		return EINVAL;
981 
982 	rw_enter(&in6_multilock, RW_READER);
983 
984 	bound = curlwp_bind();
985 	ifp = if_get_byindex(name[0], &psref);
986 	if (ifp == NULL) {
987 		curlwp_bindx(bound);
988 		rw_exit(&in6_multilock);
989 		return ENODEV;
990 	}
991 
992 	if (oldp == NULL) {
993 		*oldlenp = 0;
994 		s = pserialize_read_enter();
995 		IFADDR_READER_FOREACH(ifa, ifp) {
996 			LIST_FOREACH(in6m, &ifp->if_multiaddrs, in6m_entry) {
997 				*oldlenp += 2 * sizeof(struct in6_addr) +
998 				    sizeof(uint32_t);
999 			}
1000 		}
1001 		pserialize_read_exit(s);
1002 		if_put(ifp, &psref);
1003 		curlwp_bindx(bound);
1004 		rw_exit(&in6_multilock);
1005 		return 0;
1006 	}
1007 
1008 	error = 0;
1009 	written = 0;
1010 	s = pserialize_read_enter();
1011 	IFADDR_READER_FOREACH(ifa, ifp) {
1012 		if (ifa->ifa_addr->sa_family != AF_INET6)
1013 			continue;
1014 
1015 		ifa_acquire(ifa, &psref_ia);
1016 		pserialize_read_exit(s);
1017 
1018 		ia6 = ifatoia6(ifa);
1019 		LIST_FOREACH(in6m, &ifp->if_multiaddrs, in6m_entry) {
1020 			if (written + 2 * sizeof(struct in6_addr) +
1021 			    sizeof(uint32_t) > *oldlenp)
1022 				goto done;
1023 			/*
1024 			 * XXX return the first IPv6 address to keep backward
1025 			 * compatibility, however now multicast addresses
1026 			 * don't belong to any IPv6 addresses so it should be
1027 			 * unnecessary.
1028 			 */
1029 			error = sysctl_copyout(l, &ia6->ia_addr.sin6_addr,
1030 			    oldp, sizeof(struct in6_addr));
1031 			if (error)
1032 				goto done;
1033 			oldp = (char *)oldp + sizeof(struct in6_addr);
1034 			written += sizeof(struct in6_addr);
1035 			error = sysctl_copyout(l, &in6m->in6m_addr,
1036 			    oldp, sizeof(struct in6_addr));
1037 			if (error)
1038 				goto done;
1039 			oldp = (char *)oldp + sizeof(struct in6_addr);
1040 			written += sizeof(struct in6_addr);
1041 			tmp = in6m->in6m_refcount;
1042 			error = sysctl_copyout(l, &tmp, oldp, sizeof(tmp));
1043 			if (error)
1044 				goto done;
1045 			oldp = (char *)oldp + sizeof(tmp);
1046 			written += sizeof(tmp);
1047 		}
1048 
1049 		s = pserialize_read_enter();
1050 
1051 		break;
1052 	}
1053 	pserialize_read_exit(s);
1054 done:
1055 	ifa_release(ifa, &psref_ia);
1056 	if_put(ifp, &psref);
1057 	curlwp_bindx(bound);
1058 	rw_exit(&in6_multilock);
1059 	*oldlenp = written;
1060 	return error;
1061 }
1062 
1063 void
1064 in6_sysctl_multicast_setup(struct sysctllog **clog)
1065 {
1066 
1067 	sysctl_createv(clog, 0, NULL, NULL,
1068 		       CTLFLAG_PERMANENT,
1069 		       CTLTYPE_NODE, "inet6", NULL,
1070 		       NULL, 0, NULL, 0,
1071 		       CTL_NET, PF_INET6, CTL_EOL);
1072 
1073 	sysctl_createv(clog, 0, NULL, NULL,
1074 		       CTLFLAG_PERMANENT,
1075 		       CTLTYPE_NODE, "multicast",
1076 		       SYSCTL_DESCR("Multicast information"),
1077 		       in6_multicast_sysctl, 0, NULL, 0,
1078 		       CTL_NET, PF_INET6, CTL_CREATE, CTL_EOL);
1079 
1080 	sysctl_createv(clog, 0, NULL, NULL,
1081 		       CTLFLAG_PERMANENT,
1082 		       CTLTYPE_NODE, "multicast_kludge",
1083 		       SYSCTL_DESCR("multicast kludge information"),
1084 		       in6_mkludge_sysctl, 0, NULL, 0,
1085 		       CTL_NET, PF_INET6, CTL_CREATE, CTL_EOL);
1086 }
1087