xref: /openbmc/linux/net/ipv6/mcast.c (revision 8ee90c5c)
1 /*
2  *	Multicast support for IPv6
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Pedro Roque		<roque@di.fc.ul.pt>
7  *
8  *	Based on linux/ipv4/igmp.c and linux/ipv4/ip_sockglue.c
9  *
10  *	This program is free software; you can redistribute it and/or
11  *      modify it under the terms of the GNU General Public License
12  *      as published by the Free Software Foundation; either version
13  *      2 of the License, or (at your option) any later version.
14  */
15 
16 /* Changes:
17  *
18  *	yoshfuji	: fix format of router-alert option
19  *	YOSHIFUJI Hideaki @USAGI:
20  *		Fixed source address for MLD message based on
21  *		<draft-ietf-magma-mld-source-05.txt>.
22  *	YOSHIFUJI Hideaki @USAGI:
23  *		- Ignore Queries for invalid addresses.
24  *		- MLD for link-local addresses.
25  *	David L Stevens <dlstevens@us.ibm.com>:
26  *		- MLDv2 support
27  */
28 
29 #include <linux/module.h>
30 #include <linux/errno.h>
31 #include <linux/types.h>
32 #include <linux/string.h>
33 #include <linux/socket.h>
34 #include <linux/sockios.h>
35 #include <linux/jiffies.h>
36 #include <linux/times.h>
37 #include <linux/net.h>
38 #include <linux/in.h>
39 #include <linux/in6.h>
40 #include <linux/netdevice.h>
41 #include <linux/if_arp.h>
42 #include <linux/route.h>
43 #include <linux/init.h>
44 #include <linux/proc_fs.h>
45 #include <linux/seq_file.h>
46 #include <linux/slab.h>
47 #include <linux/pkt_sched.h>
48 #include <net/mld.h>
49 
50 #include <linux/netfilter.h>
51 #include <linux/netfilter_ipv6.h>
52 
53 #include <net/net_namespace.h>
54 #include <net/sock.h>
55 #include <net/snmp.h>
56 
57 #include <net/ipv6.h>
58 #include <net/protocol.h>
59 #include <net/if_inet6.h>
60 #include <net/ndisc.h>
61 #include <net/addrconf.h>
62 #include <net/ip6_route.h>
63 #include <net/inet_common.h>
64 
65 #include <net/ip6_checksum.h>
66 
67 /* Ensure that we have struct in6_addr aligned on 32bit word. */
68 static void *__mld2_query_bugs[] __attribute__((__unused__)) = {
69 	BUILD_BUG_ON_NULL(offsetof(struct mld2_query, mld2q_srcs) % 4),
70 	BUILD_BUG_ON_NULL(offsetof(struct mld2_report, mld2r_grec) % 4),
71 	BUILD_BUG_ON_NULL(offsetof(struct mld2_grec, grec_mca) % 4)
72 };
73 
74 static struct in6_addr mld2_all_mcr = MLD2_ALL_MCR_INIT;
75 
76 static void igmp6_join_group(struct ifmcaddr6 *ma);
77 static void igmp6_leave_group(struct ifmcaddr6 *ma);
78 static void igmp6_timer_handler(unsigned long data);
79 
80 static void mld_gq_timer_expire(unsigned long data);
81 static void mld_ifc_timer_expire(unsigned long data);
82 static void mld_ifc_event(struct inet6_dev *idev);
83 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *pmc);
84 static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *pmc);
85 static void mld_clear_delrec(struct inet6_dev *idev);
86 static bool mld_in_v1_mode(const struct inet6_dev *idev);
87 static int sf_setstate(struct ifmcaddr6 *pmc);
88 static void sf_markstate(struct ifmcaddr6 *pmc);
89 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc);
90 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
91 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
92 			  int delta);
93 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
94 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
95 			  int delta);
96 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
97 			    struct inet6_dev *idev);
98 
99 #define MLD_QRV_DEFAULT		2
100 /* RFC3810, 9.2. Query Interval */
101 #define MLD_QI_DEFAULT		(125 * HZ)
102 /* RFC3810, 9.3. Query Response Interval */
103 #define MLD_QRI_DEFAULT		(10 * HZ)
104 
105 /* RFC3810, 8.1 Query Version Distinctions */
106 #define MLD_V1_QUERY_LEN	24
107 #define MLD_V2_QUERY_LEN_MIN	28
108 
109 #define IPV6_MLD_MAX_MSF	64
110 
111 int sysctl_mld_max_msf __read_mostly = IPV6_MLD_MAX_MSF;
112 int sysctl_mld_qrv __read_mostly = MLD_QRV_DEFAULT;
113 
114 /*
115  *	socket join on multicast group
116  */
117 
118 #define for_each_pmc_rcu(np, pmc)				\
119 	for (pmc = rcu_dereference(np->ipv6_mc_list);		\
120 	     pmc != NULL;					\
121 	     pmc = rcu_dereference(pmc->next))
122 
123 static int unsolicited_report_interval(struct inet6_dev *idev)
124 {
125 	int iv;
126 
127 	if (mld_in_v1_mode(idev))
128 		iv = idev->cnf.mldv1_unsolicited_report_interval;
129 	else
130 		iv = idev->cnf.mldv2_unsolicited_report_interval;
131 
132 	return iv > 0 ? iv : 1;
133 }
134 
135 int ipv6_sock_mc_join(struct sock *sk, int ifindex, const struct in6_addr *addr)
136 {
137 	struct net_device *dev = NULL;
138 	struct ipv6_mc_socklist *mc_lst;
139 	struct ipv6_pinfo *np = inet6_sk(sk);
140 	struct net *net = sock_net(sk);
141 	int err;
142 
143 	ASSERT_RTNL();
144 
145 	if (!ipv6_addr_is_multicast(addr))
146 		return -EINVAL;
147 
148 	rcu_read_lock();
149 	for_each_pmc_rcu(np, mc_lst) {
150 		if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
151 		    ipv6_addr_equal(&mc_lst->addr, addr)) {
152 			rcu_read_unlock();
153 			return -EADDRINUSE;
154 		}
155 	}
156 	rcu_read_unlock();
157 
158 	mc_lst = sock_kmalloc(sk, sizeof(struct ipv6_mc_socklist), GFP_KERNEL);
159 
160 	if (!mc_lst)
161 		return -ENOMEM;
162 
163 	mc_lst->next = NULL;
164 	mc_lst->addr = *addr;
165 
166 	if (ifindex == 0) {
167 		struct rt6_info *rt;
168 		rt = rt6_lookup(net, addr, NULL, 0, 0);
169 		if (rt) {
170 			dev = rt->dst.dev;
171 			ip6_rt_put(rt);
172 		}
173 	} else
174 		dev = __dev_get_by_index(net, ifindex);
175 
176 	if (!dev) {
177 		sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
178 		return -ENODEV;
179 	}
180 
181 	mc_lst->ifindex = dev->ifindex;
182 	mc_lst->sfmode = MCAST_EXCLUDE;
183 	rwlock_init(&mc_lst->sflock);
184 	mc_lst->sflist = NULL;
185 
186 	/*
187 	 *	now add/increase the group membership on the device
188 	 */
189 
190 	err = ipv6_dev_mc_inc(dev, addr);
191 
192 	if (err) {
193 		sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
194 		return err;
195 	}
196 
197 	mc_lst->next = np->ipv6_mc_list;
198 	rcu_assign_pointer(np->ipv6_mc_list, mc_lst);
199 
200 	return 0;
201 }
202 EXPORT_SYMBOL(ipv6_sock_mc_join);
203 
204 /*
205  *	socket leave on multicast group
206  */
207 int ipv6_sock_mc_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
208 {
209 	struct ipv6_pinfo *np = inet6_sk(sk);
210 	struct ipv6_mc_socklist *mc_lst;
211 	struct ipv6_mc_socklist __rcu **lnk;
212 	struct net *net = sock_net(sk);
213 
214 	ASSERT_RTNL();
215 
216 	if (!ipv6_addr_is_multicast(addr))
217 		return -EINVAL;
218 
219 	for (lnk = &np->ipv6_mc_list;
220 	     (mc_lst = rtnl_dereference(*lnk)) != NULL;
221 	      lnk = &mc_lst->next) {
222 		if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
223 		    ipv6_addr_equal(&mc_lst->addr, addr)) {
224 			struct net_device *dev;
225 
226 			*lnk = mc_lst->next;
227 
228 			dev = __dev_get_by_index(net, mc_lst->ifindex);
229 			if (dev) {
230 				struct inet6_dev *idev = __in6_dev_get(dev);
231 
232 				(void) ip6_mc_leave_src(sk, mc_lst, idev);
233 				if (idev)
234 					__ipv6_dev_mc_dec(idev, &mc_lst->addr);
235 			} else
236 				(void) ip6_mc_leave_src(sk, mc_lst, NULL);
237 
238 			atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
239 			kfree_rcu(mc_lst, rcu);
240 			return 0;
241 		}
242 	}
243 
244 	return -EADDRNOTAVAIL;
245 }
246 EXPORT_SYMBOL(ipv6_sock_mc_drop);
247 
248 /* called with rcu_read_lock() */
249 static struct inet6_dev *ip6_mc_find_dev_rcu(struct net *net,
250 					     const struct in6_addr *group,
251 					     int ifindex)
252 {
253 	struct net_device *dev = NULL;
254 	struct inet6_dev *idev = NULL;
255 
256 	if (ifindex == 0) {
257 		struct rt6_info *rt = rt6_lookup(net, group, NULL, 0, 0);
258 
259 		if (rt) {
260 			dev = rt->dst.dev;
261 			ip6_rt_put(rt);
262 		}
263 	} else
264 		dev = dev_get_by_index_rcu(net, ifindex);
265 
266 	if (!dev)
267 		return NULL;
268 	idev = __in6_dev_get(dev);
269 	if (!idev)
270 		return NULL;
271 	read_lock_bh(&idev->lock);
272 	if (idev->dead) {
273 		read_unlock_bh(&idev->lock);
274 		return NULL;
275 	}
276 	return idev;
277 }
278 
279 void __ipv6_sock_mc_close(struct sock *sk)
280 {
281 	struct ipv6_pinfo *np = inet6_sk(sk);
282 	struct ipv6_mc_socklist *mc_lst;
283 	struct net *net = sock_net(sk);
284 
285 	ASSERT_RTNL();
286 
287 	while ((mc_lst = rtnl_dereference(np->ipv6_mc_list)) != NULL) {
288 		struct net_device *dev;
289 
290 		np->ipv6_mc_list = mc_lst->next;
291 
292 		dev = __dev_get_by_index(net, mc_lst->ifindex);
293 		if (dev) {
294 			struct inet6_dev *idev = __in6_dev_get(dev);
295 
296 			(void) ip6_mc_leave_src(sk, mc_lst, idev);
297 			if (idev)
298 				__ipv6_dev_mc_dec(idev, &mc_lst->addr);
299 		} else
300 			(void) ip6_mc_leave_src(sk, mc_lst, NULL);
301 
302 		atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
303 		kfree_rcu(mc_lst, rcu);
304 	}
305 }
306 
307 void ipv6_sock_mc_close(struct sock *sk)
308 {
309 	struct ipv6_pinfo *np = inet6_sk(sk);
310 
311 	if (!rcu_access_pointer(np->ipv6_mc_list))
312 		return;
313 	rtnl_lock();
314 	__ipv6_sock_mc_close(sk);
315 	rtnl_unlock();
316 }
317 
318 int ip6_mc_source(int add, int omode, struct sock *sk,
319 	struct group_source_req *pgsr)
320 {
321 	struct in6_addr *source, *group;
322 	struct ipv6_mc_socklist *pmc;
323 	struct inet6_dev *idev;
324 	struct ipv6_pinfo *inet6 = inet6_sk(sk);
325 	struct ip6_sf_socklist *psl;
326 	struct net *net = sock_net(sk);
327 	int i, j, rv;
328 	int leavegroup = 0;
329 	int pmclocked = 0;
330 	int err;
331 
332 	source = &((struct sockaddr_in6 *)&pgsr->gsr_source)->sin6_addr;
333 	group = &((struct sockaddr_in6 *)&pgsr->gsr_group)->sin6_addr;
334 
335 	if (!ipv6_addr_is_multicast(group))
336 		return -EINVAL;
337 
338 	rcu_read_lock();
339 	idev = ip6_mc_find_dev_rcu(net, group, pgsr->gsr_interface);
340 	if (!idev) {
341 		rcu_read_unlock();
342 		return -ENODEV;
343 	}
344 
345 	err = -EADDRNOTAVAIL;
346 
347 	for_each_pmc_rcu(inet6, pmc) {
348 		if (pgsr->gsr_interface && pmc->ifindex != pgsr->gsr_interface)
349 			continue;
350 		if (ipv6_addr_equal(&pmc->addr, group))
351 			break;
352 	}
353 	if (!pmc) {		/* must have a prior join */
354 		err = -EINVAL;
355 		goto done;
356 	}
357 	/* if a source filter was set, must be the same mode as before */
358 	if (pmc->sflist) {
359 		if (pmc->sfmode != omode) {
360 			err = -EINVAL;
361 			goto done;
362 		}
363 	} else if (pmc->sfmode != omode) {
364 		/* allow mode switches for empty-set filters */
365 		ip6_mc_add_src(idev, group, omode, 0, NULL, 0);
366 		ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
367 		pmc->sfmode = omode;
368 	}
369 
370 	write_lock(&pmc->sflock);
371 	pmclocked = 1;
372 
373 	psl = pmc->sflist;
374 	if (!add) {
375 		if (!psl)
376 			goto done;	/* err = -EADDRNOTAVAIL */
377 		rv = !0;
378 		for (i = 0; i < psl->sl_count; i++) {
379 			rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
380 			if (rv == 0)
381 				break;
382 		}
383 		if (rv)		/* source not found */
384 			goto done;	/* err = -EADDRNOTAVAIL */
385 
386 		/* special case - (INCLUDE, empty) == LEAVE_GROUP */
387 		if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
388 			leavegroup = 1;
389 			goto done;
390 		}
391 
392 		/* update the interface filter */
393 		ip6_mc_del_src(idev, group, omode, 1, source, 1);
394 
395 		for (j = i+1; j < psl->sl_count; j++)
396 			psl->sl_addr[j-1] = psl->sl_addr[j];
397 		psl->sl_count--;
398 		err = 0;
399 		goto done;
400 	}
401 	/* else, add a new source to the filter */
402 
403 	if (psl && psl->sl_count >= sysctl_mld_max_msf) {
404 		err = -ENOBUFS;
405 		goto done;
406 	}
407 	if (!psl || psl->sl_count == psl->sl_max) {
408 		struct ip6_sf_socklist *newpsl;
409 		int count = IP6_SFBLOCK;
410 
411 		if (psl)
412 			count += psl->sl_max;
413 		newpsl = sock_kmalloc(sk, IP6_SFLSIZE(count), GFP_ATOMIC);
414 		if (!newpsl) {
415 			err = -ENOBUFS;
416 			goto done;
417 		}
418 		newpsl->sl_max = count;
419 		newpsl->sl_count = count - IP6_SFBLOCK;
420 		if (psl) {
421 			for (i = 0; i < psl->sl_count; i++)
422 				newpsl->sl_addr[i] = psl->sl_addr[i];
423 			sock_kfree_s(sk, psl, IP6_SFLSIZE(psl->sl_max));
424 		}
425 		pmc->sflist = psl = newpsl;
426 	}
427 	rv = 1;	/* > 0 for insert logic below if sl_count is 0 */
428 	for (i = 0; i < psl->sl_count; i++) {
429 		rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
430 		if (rv == 0) /* There is an error in the address. */
431 			goto done;
432 	}
433 	for (j = psl->sl_count-1; j >= i; j--)
434 		psl->sl_addr[j+1] = psl->sl_addr[j];
435 	psl->sl_addr[i] = *source;
436 	psl->sl_count++;
437 	err = 0;
438 	/* update the interface list */
439 	ip6_mc_add_src(idev, group, omode, 1, source, 1);
440 done:
441 	if (pmclocked)
442 		write_unlock(&pmc->sflock);
443 	read_unlock_bh(&idev->lock);
444 	rcu_read_unlock();
445 	if (leavegroup)
446 		err = ipv6_sock_mc_drop(sk, pgsr->gsr_interface, group);
447 	return err;
448 }
449 
450 int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf)
451 {
452 	const struct in6_addr *group;
453 	struct ipv6_mc_socklist *pmc;
454 	struct inet6_dev *idev;
455 	struct ipv6_pinfo *inet6 = inet6_sk(sk);
456 	struct ip6_sf_socklist *newpsl, *psl;
457 	struct net *net = sock_net(sk);
458 	int leavegroup = 0;
459 	int i, err;
460 
461 	group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
462 
463 	if (!ipv6_addr_is_multicast(group))
464 		return -EINVAL;
465 	if (gsf->gf_fmode != MCAST_INCLUDE &&
466 	    gsf->gf_fmode != MCAST_EXCLUDE)
467 		return -EINVAL;
468 
469 	rcu_read_lock();
470 	idev = ip6_mc_find_dev_rcu(net, group, gsf->gf_interface);
471 
472 	if (!idev) {
473 		rcu_read_unlock();
474 		return -ENODEV;
475 	}
476 
477 	err = 0;
478 
479 	if (gsf->gf_fmode == MCAST_INCLUDE && gsf->gf_numsrc == 0) {
480 		leavegroup = 1;
481 		goto done;
482 	}
483 
484 	for_each_pmc_rcu(inet6, pmc) {
485 		if (pmc->ifindex != gsf->gf_interface)
486 			continue;
487 		if (ipv6_addr_equal(&pmc->addr, group))
488 			break;
489 	}
490 	if (!pmc) {		/* must have a prior join */
491 		err = -EINVAL;
492 		goto done;
493 	}
494 	if (gsf->gf_numsrc) {
495 		newpsl = sock_kmalloc(sk, IP6_SFLSIZE(gsf->gf_numsrc),
496 							  GFP_ATOMIC);
497 		if (!newpsl) {
498 			err = -ENOBUFS;
499 			goto done;
500 		}
501 		newpsl->sl_max = newpsl->sl_count = gsf->gf_numsrc;
502 		for (i = 0; i < newpsl->sl_count; ++i) {
503 			struct sockaddr_in6 *psin6;
504 
505 			psin6 = (struct sockaddr_in6 *)&gsf->gf_slist[i];
506 			newpsl->sl_addr[i] = psin6->sin6_addr;
507 		}
508 		err = ip6_mc_add_src(idev, group, gsf->gf_fmode,
509 			newpsl->sl_count, newpsl->sl_addr, 0);
510 		if (err) {
511 			sock_kfree_s(sk, newpsl, IP6_SFLSIZE(newpsl->sl_max));
512 			goto done;
513 		}
514 	} else {
515 		newpsl = NULL;
516 		(void) ip6_mc_add_src(idev, group, gsf->gf_fmode, 0, NULL, 0);
517 	}
518 
519 	write_lock(&pmc->sflock);
520 	psl = pmc->sflist;
521 	if (psl) {
522 		(void) ip6_mc_del_src(idev, group, pmc->sfmode,
523 			psl->sl_count, psl->sl_addr, 0);
524 		sock_kfree_s(sk, psl, IP6_SFLSIZE(psl->sl_max));
525 	} else
526 		(void) ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
527 	pmc->sflist = newpsl;
528 	pmc->sfmode = gsf->gf_fmode;
529 	write_unlock(&pmc->sflock);
530 	err = 0;
531 done:
532 	read_unlock_bh(&idev->lock);
533 	rcu_read_unlock();
534 	if (leavegroup)
535 		err = ipv6_sock_mc_drop(sk, gsf->gf_interface, group);
536 	return err;
537 }
538 
539 int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
540 	struct group_filter __user *optval, int __user *optlen)
541 {
542 	int err, i, count, copycount;
543 	const struct in6_addr *group;
544 	struct ipv6_mc_socklist *pmc;
545 	struct inet6_dev *idev;
546 	struct ipv6_pinfo *inet6 = inet6_sk(sk);
547 	struct ip6_sf_socklist *psl;
548 	struct net *net = sock_net(sk);
549 
550 	group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
551 
552 	if (!ipv6_addr_is_multicast(group))
553 		return -EINVAL;
554 
555 	rcu_read_lock();
556 	idev = ip6_mc_find_dev_rcu(net, group, gsf->gf_interface);
557 
558 	if (!idev) {
559 		rcu_read_unlock();
560 		return -ENODEV;
561 	}
562 
563 	err = -EADDRNOTAVAIL;
564 	/* changes to the ipv6_mc_list require the socket lock and
565 	 * rtnl lock. We have the socket lock and rcu read lock,
566 	 * so reading the list is safe.
567 	 */
568 
569 	for_each_pmc_rcu(inet6, pmc) {
570 		if (pmc->ifindex != gsf->gf_interface)
571 			continue;
572 		if (ipv6_addr_equal(group, &pmc->addr))
573 			break;
574 	}
575 	if (!pmc)		/* must have a prior join */
576 		goto done;
577 	gsf->gf_fmode = pmc->sfmode;
578 	psl = pmc->sflist;
579 	count = psl ? psl->sl_count : 0;
580 	read_unlock_bh(&idev->lock);
581 	rcu_read_unlock();
582 
583 	copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
584 	gsf->gf_numsrc = count;
585 	if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
586 	    copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
587 		return -EFAULT;
588 	}
589 	/* changes to psl require the socket lock, and a write lock
590 	 * on pmc->sflock. We have the socket lock so reading here is safe.
591 	 */
592 	for (i = 0; i < copycount; i++) {
593 		struct sockaddr_in6 *psin6;
594 		struct sockaddr_storage ss;
595 
596 		psin6 = (struct sockaddr_in6 *)&ss;
597 		memset(&ss, 0, sizeof(ss));
598 		psin6->sin6_family = AF_INET6;
599 		psin6->sin6_addr = psl->sl_addr[i];
600 		if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
601 			return -EFAULT;
602 	}
603 	return 0;
604 done:
605 	read_unlock_bh(&idev->lock);
606 	rcu_read_unlock();
607 	return err;
608 }
609 
610 bool inet6_mc_check(struct sock *sk, const struct in6_addr *mc_addr,
611 		    const struct in6_addr *src_addr)
612 {
613 	struct ipv6_pinfo *np = inet6_sk(sk);
614 	struct ipv6_mc_socklist *mc;
615 	struct ip6_sf_socklist *psl;
616 	bool rv = true;
617 
618 	rcu_read_lock();
619 	for_each_pmc_rcu(np, mc) {
620 		if (ipv6_addr_equal(&mc->addr, mc_addr))
621 			break;
622 	}
623 	if (!mc) {
624 		rcu_read_unlock();
625 		return true;
626 	}
627 	read_lock(&mc->sflock);
628 	psl = mc->sflist;
629 	if (!psl) {
630 		rv = mc->sfmode == MCAST_EXCLUDE;
631 	} else {
632 		int i;
633 
634 		for (i = 0; i < psl->sl_count; i++) {
635 			if (ipv6_addr_equal(&psl->sl_addr[i], src_addr))
636 				break;
637 		}
638 		if (mc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
639 			rv = false;
640 		if (mc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
641 			rv = false;
642 	}
643 	read_unlock(&mc->sflock);
644 	rcu_read_unlock();
645 
646 	return rv;
647 }
648 
649 static void igmp6_group_added(struct ifmcaddr6 *mc)
650 {
651 	struct net_device *dev = mc->idev->dev;
652 	char buf[MAX_ADDR_LEN];
653 
654 	if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
655 	    IPV6_ADDR_SCOPE_LINKLOCAL)
656 		return;
657 
658 	spin_lock_bh(&mc->mca_lock);
659 	if (!(mc->mca_flags&MAF_LOADED)) {
660 		mc->mca_flags |= MAF_LOADED;
661 		if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
662 			dev_mc_add(dev, buf);
663 	}
664 	spin_unlock_bh(&mc->mca_lock);
665 
666 	if (!(dev->flags & IFF_UP) || (mc->mca_flags & MAF_NOREPORT))
667 		return;
668 
669 	if (mld_in_v1_mode(mc->idev)) {
670 		igmp6_join_group(mc);
671 		return;
672 	}
673 	/* else v2 */
674 
675 	mc->mca_crcount = mc->idev->mc_qrv;
676 	mld_ifc_event(mc->idev);
677 }
678 
679 static void igmp6_group_dropped(struct ifmcaddr6 *mc)
680 {
681 	struct net_device *dev = mc->idev->dev;
682 	char buf[MAX_ADDR_LEN];
683 
684 	if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
685 	    IPV6_ADDR_SCOPE_LINKLOCAL)
686 		return;
687 
688 	spin_lock_bh(&mc->mca_lock);
689 	if (mc->mca_flags&MAF_LOADED) {
690 		mc->mca_flags &= ~MAF_LOADED;
691 		if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
692 			dev_mc_del(dev, buf);
693 	}
694 
695 	spin_unlock_bh(&mc->mca_lock);
696 	if (mc->mca_flags & MAF_NOREPORT)
697 		return;
698 
699 	if (!mc->idev->dead)
700 		igmp6_leave_group(mc);
701 
702 	spin_lock_bh(&mc->mca_lock);
703 	if (del_timer(&mc->mca_timer))
704 		refcount_dec(&mc->mca_refcnt);
705 	spin_unlock_bh(&mc->mca_lock);
706 }
707 
708 /*
709  * deleted ifmcaddr6 manipulation
710  */
711 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
712 {
713 	struct ifmcaddr6 *pmc;
714 
715 	/* this is an "ifmcaddr6" for convenience; only the fields below
716 	 * are actually used. In particular, the refcnt and users are not
717 	 * used for management of the delete list. Using the same structure
718 	 * for deleted items allows change reports to use common code with
719 	 * non-deleted or query-response MCA's.
720 	 */
721 	pmc = kzalloc(sizeof(*pmc), GFP_ATOMIC);
722 	if (!pmc)
723 		return;
724 
725 	spin_lock_bh(&im->mca_lock);
726 	spin_lock_init(&pmc->mca_lock);
727 	pmc->idev = im->idev;
728 	in6_dev_hold(idev);
729 	pmc->mca_addr = im->mca_addr;
730 	pmc->mca_crcount = idev->mc_qrv;
731 	pmc->mca_sfmode = im->mca_sfmode;
732 	if (pmc->mca_sfmode == MCAST_INCLUDE) {
733 		struct ip6_sf_list *psf;
734 
735 		pmc->mca_tomb = im->mca_tomb;
736 		pmc->mca_sources = im->mca_sources;
737 		im->mca_tomb = im->mca_sources = NULL;
738 		for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
739 			psf->sf_crcount = pmc->mca_crcount;
740 	}
741 	spin_unlock_bh(&im->mca_lock);
742 
743 	spin_lock_bh(&idev->mc_lock);
744 	pmc->next = idev->mc_tomb;
745 	idev->mc_tomb = pmc;
746 	spin_unlock_bh(&idev->mc_lock);
747 }
748 
749 static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
750 {
751 	struct ifmcaddr6 *pmc, *pmc_prev;
752 	struct ip6_sf_list *psf;
753 	struct in6_addr *pmca = &im->mca_addr;
754 
755 	spin_lock_bh(&idev->mc_lock);
756 	pmc_prev = NULL;
757 	for (pmc = idev->mc_tomb; pmc; pmc = pmc->next) {
758 		if (ipv6_addr_equal(&pmc->mca_addr, pmca))
759 			break;
760 		pmc_prev = pmc;
761 	}
762 	if (pmc) {
763 		if (pmc_prev)
764 			pmc_prev->next = pmc->next;
765 		else
766 			idev->mc_tomb = pmc->next;
767 	}
768 	spin_unlock_bh(&idev->mc_lock);
769 
770 	spin_lock_bh(&im->mca_lock);
771 	if (pmc) {
772 		im->idev = pmc->idev;
773 		im->mca_crcount = idev->mc_qrv;
774 		im->mca_sfmode = pmc->mca_sfmode;
775 		if (pmc->mca_sfmode == MCAST_INCLUDE) {
776 			im->mca_tomb = pmc->mca_tomb;
777 			im->mca_sources = pmc->mca_sources;
778 			for (psf = im->mca_sources; psf; psf = psf->sf_next)
779 				psf->sf_crcount = im->mca_crcount;
780 		}
781 		in6_dev_put(pmc->idev);
782 		kfree(pmc);
783 	}
784 	spin_unlock_bh(&im->mca_lock);
785 }
786 
787 static void mld_clear_delrec(struct inet6_dev *idev)
788 {
789 	struct ifmcaddr6 *pmc, *nextpmc;
790 
791 	spin_lock_bh(&idev->mc_lock);
792 	pmc = idev->mc_tomb;
793 	idev->mc_tomb = NULL;
794 	spin_unlock_bh(&idev->mc_lock);
795 
796 	for (; pmc; pmc = nextpmc) {
797 		nextpmc = pmc->next;
798 		ip6_mc_clear_src(pmc);
799 		in6_dev_put(pmc->idev);
800 		kfree(pmc);
801 	}
802 
803 	/* clear dead sources, too */
804 	read_lock_bh(&idev->lock);
805 	for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
806 		struct ip6_sf_list *psf, *psf_next;
807 
808 		spin_lock_bh(&pmc->mca_lock);
809 		psf = pmc->mca_tomb;
810 		pmc->mca_tomb = NULL;
811 		spin_unlock_bh(&pmc->mca_lock);
812 		for (; psf; psf = psf_next) {
813 			psf_next = psf->sf_next;
814 			kfree(psf);
815 		}
816 	}
817 	read_unlock_bh(&idev->lock);
818 }
819 
820 static void mca_get(struct ifmcaddr6 *mc)
821 {
822 	refcount_inc(&mc->mca_refcnt);
823 }
824 
825 static void ma_put(struct ifmcaddr6 *mc)
826 {
827 	if (refcount_dec_and_test(&mc->mca_refcnt)) {
828 		in6_dev_put(mc->idev);
829 		kfree(mc);
830 	}
831 }
832 
833 static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev,
834 				   const struct in6_addr *addr)
835 {
836 	struct ifmcaddr6 *mc;
837 
838 	mc = kzalloc(sizeof(*mc), GFP_ATOMIC);
839 	if (!mc)
840 		return NULL;
841 
842 	setup_timer(&mc->mca_timer, igmp6_timer_handler, (unsigned long)mc);
843 
844 	mc->mca_addr = *addr;
845 	mc->idev = idev; /* reference taken by caller */
846 	mc->mca_users = 1;
847 	/* mca_stamp should be updated upon changes */
848 	mc->mca_cstamp = mc->mca_tstamp = jiffies;
849 	refcount_set(&mc->mca_refcnt, 1);
850 	spin_lock_init(&mc->mca_lock);
851 
852 	/* initial mode is (EX, empty) */
853 	mc->mca_sfmode = MCAST_EXCLUDE;
854 	mc->mca_sfcount[MCAST_EXCLUDE] = 1;
855 
856 	if (ipv6_addr_is_ll_all_nodes(&mc->mca_addr) ||
857 	    IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
858 		mc->mca_flags |= MAF_NOREPORT;
859 
860 	return mc;
861 }
862 
863 /*
864  *	device multicast group inc (add if not found)
865  */
866 int ipv6_dev_mc_inc(struct net_device *dev, const struct in6_addr *addr)
867 {
868 	struct ifmcaddr6 *mc;
869 	struct inet6_dev *idev;
870 
871 	ASSERT_RTNL();
872 
873 	/* we need to take a reference on idev */
874 	idev = in6_dev_get(dev);
875 
876 	if (!idev)
877 		return -EINVAL;
878 
879 	write_lock_bh(&idev->lock);
880 	if (idev->dead) {
881 		write_unlock_bh(&idev->lock);
882 		in6_dev_put(idev);
883 		return -ENODEV;
884 	}
885 
886 	for (mc = idev->mc_list; mc; mc = mc->next) {
887 		if (ipv6_addr_equal(&mc->mca_addr, addr)) {
888 			mc->mca_users++;
889 			write_unlock_bh(&idev->lock);
890 			ip6_mc_add_src(idev, &mc->mca_addr, MCAST_EXCLUDE, 0,
891 				NULL, 0);
892 			in6_dev_put(idev);
893 			return 0;
894 		}
895 	}
896 
897 	mc = mca_alloc(idev, addr);
898 	if (!mc) {
899 		write_unlock_bh(&idev->lock);
900 		in6_dev_put(idev);
901 		return -ENOMEM;
902 	}
903 
904 	mc->next = idev->mc_list;
905 	idev->mc_list = mc;
906 
907 	/* Hold this for the code below before we unlock,
908 	 * it is already exposed via idev->mc_list.
909 	 */
910 	mca_get(mc);
911 	write_unlock_bh(&idev->lock);
912 
913 	mld_del_delrec(idev, mc);
914 	igmp6_group_added(mc);
915 	ma_put(mc);
916 	return 0;
917 }
918 
919 /*
920  *	device multicast group del
921  */
922 int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
923 {
924 	struct ifmcaddr6 *ma, **map;
925 
926 	ASSERT_RTNL();
927 
928 	write_lock_bh(&idev->lock);
929 	for (map = &idev->mc_list; (ma = *map) != NULL; map = &ma->next) {
930 		if (ipv6_addr_equal(&ma->mca_addr, addr)) {
931 			if (--ma->mca_users == 0) {
932 				*map = ma->next;
933 				write_unlock_bh(&idev->lock);
934 
935 				igmp6_group_dropped(ma);
936 				ip6_mc_clear_src(ma);
937 
938 				ma_put(ma);
939 				return 0;
940 			}
941 			write_unlock_bh(&idev->lock);
942 			return 0;
943 		}
944 	}
945 	write_unlock_bh(&idev->lock);
946 
947 	return -ENOENT;
948 }
949 
950 int ipv6_dev_mc_dec(struct net_device *dev, const struct in6_addr *addr)
951 {
952 	struct inet6_dev *idev;
953 	int err;
954 
955 	ASSERT_RTNL();
956 
957 	idev = __in6_dev_get(dev);
958 	if (!idev)
959 		err = -ENODEV;
960 	else
961 		err = __ipv6_dev_mc_dec(idev, addr);
962 
963 	return err;
964 }
965 
966 /*
967  *	check if the interface/address pair is valid
968  */
969 bool ipv6_chk_mcast_addr(struct net_device *dev, const struct in6_addr *group,
970 			 const struct in6_addr *src_addr)
971 {
972 	struct inet6_dev *idev;
973 	struct ifmcaddr6 *mc;
974 	bool rv = false;
975 
976 	rcu_read_lock();
977 	idev = __in6_dev_get(dev);
978 	if (idev) {
979 		read_lock_bh(&idev->lock);
980 		for (mc = idev->mc_list; mc; mc = mc->next) {
981 			if (ipv6_addr_equal(&mc->mca_addr, group))
982 				break;
983 		}
984 		if (mc) {
985 			if (src_addr && !ipv6_addr_any(src_addr)) {
986 				struct ip6_sf_list *psf;
987 
988 				spin_lock_bh(&mc->mca_lock);
989 				for (psf = mc->mca_sources; psf; psf = psf->sf_next) {
990 					if (ipv6_addr_equal(&psf->sf_addr, src_addr))
991 						break;
992 				}
993 				if (psf)
994 					rv = psf->sf_count[MCAST_INCLUDE] ||
995 						psf->sf_count[MCAST_EXCLUDE] !=
996 						mc->mca_sfcount[MCAST_EXCLUDE];
997 				else
998 					rv = mc->mca_sfcount[MCAST_EXCLUDE] != 0;
999 				spin_unlock_bh(&mc->mca_lock);
1000 			} else
1001 				rv = true; /* don't filter unspecified source */
1002 		}
1003 		read_unlock_bh(&idev->lock);
1004 	}
1005 	rcu_read_unlock();
1006 	return rv;
1007 }
1008 
1009 static void mld_gq_start_timer(struct inet6_dev *idev)
1010 {
1011 	unsigned long tv = prandom_u32() % idev->mc_maxdelay;
1012 
1013 	idev->mc_gq_running = 1;
1014 	if (!mod_timer(&idev->mc_gq_timer, jiffies+tv+2))
1015 		in6_dev_hold(idev);
1016 }
1017 
1018 static void mld_gq_stop_timer(struct inet6_dev *idev)
1019 {
1020 	idev->mc_gq_running = 0;
1021 	if (del_timer(&idev->mc_gq_timer))
1022 		__in6_dev_put(idev);
1023 }
1024 
1025 static void mld_ifc_start_timer(struct inet6_dev *idev, unsigned long delay)
1026 {
1027 	unsigned long tv = prandom_u32() % delay;
1028 
1029 	if (!mod_timer(&idev->mc_ifc_timer, jiffies+tv+2))
1030 		in6_dev_hold(idev);
1031 }
1032 
1033 static void mld_ifc_stop_timer(struct inet6_dev *idev)
1034 {
1035 	idev->mc_ifc_count = 0;
1036 	if (del_timer(&idev->mc_ifc_timer))
1037 		__in6_dev_put(idev);
1038 }
1039 
1040 static void mld_dad_start_timer(struct inet6_dev *idev, unsigned long delay)
1041 {
1042 	unsigned long tv = prandom_u32() % delay;
1043 
1044 	if (!mod_timer(&idev->mc_dad_timer, jiffies+tv+2))
1045 		in6_dev_hold(idev);
1046 }
1047 
1048 static void mld_dad_stop_timer(struct inet6_dev *idev)
1049 {
1050 	if (del_timer(&idev->mc_dad_timer))
1051 		__in6_dev_put(idev);
1052 }
1053 
1054 /*
1055  *	IGMP handling (alias multicast ICMPv6 messages)
1056  */
1057 
1058 static void igmp6_group_queried(struct ifmcaddr6 *ma, unsigned long resptime)
1059 {
1060 	unsigned long delay = resptime;
1061 
1062 	/* Do not start timer for these addresses */
1063 	if (ipv6_addr_is_ll_all_nodes(&ma->mca_addr) ||
1064 	    IPV6_ADDR_MC_SCOPE(&ma->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
1065 		return;
1066 
1067 	if (del_timer(&ma->mca_timer)) {
1068 		refcount_dec(&ma->mca_refcnt);
1069 		delay = ma->mca_timer.expires - jiffies;
1070 	}
1071 
1072 	if (delay >= resptime)
1073 		delay = prandom_u32() % resptime;
1074 
1075 	ma->mca_timer.expires = jiffies + delay;
1076 	if (!mod_timer(&ma->mca_timer, jiffies + delay))
1077 		refcount_inc(&ma->mca_refcnt);
1078 	ma->mca_flags |= MAF_TIMER_RUNNING;
1079 }
1080 
1081 /* mark EXCLUDE-mode sources */
1082 static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs,
1083 			     const struct in6_addr *srcs)
1084 {
1085 	struct ip6_sf_list *psf;
1086 	int i, scount;
1087 
1088 	scount = 0;
1089 	for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1090 		if (scount == nsrcs)
1091 			break;
1092 		for (i = 0; i < nsrcs; i++) {
1093 			/* skip inactive filters */
1094 			if (psf->sf_count[MCAST_INCLUDE] ||
1095 			    pmc->mca_sfcount[MCAST_EXCLUDE] !=
1096 			    psf->sf_count[MCAST_EXCLUDE])
1097 				break;
1098 			if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1099 				scount++;
1100 				break;
1101 			}
1102 		}
1103 	}
1104 	pmc->mca_flags &= ~MAF_GSQUERY;
1105 	if (scount == nsrcs)	/* all sources excluded */
1106 		return false;
1107 	return true;
1108 }
1109 
1110 static bool mld_marksources(struct ifmcaddr6 *pmc, int nsrcs,
1111 			    const struct in6_addr *srcs)
1112 {
1113 	struct ip6_sf_list *psf;
1114 	int i, scount;
1115 
1116 	if (pmc->mca_sfmode == MCAST_EXCLUDE)
1117 		return mld_xmarksources(pmc, nsrcs, srcs);
1118 
1119 	/* mark INCLUDE-mode sources */
1120 
1121 	scount = 0;
1122 	for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1123 		if (scount == nsrcs)
1124 			break;
1125 		for (i = 0; i < nsrcs; i++) {
1126 			if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1127 				psf->sf_gsresp = 1;
1128 				scount++;
1129 				break;
1130 			}
1131 		}
1132 	}
1133 	if (!scount) {
1134 		pmc->mca_flags &= ~MAF_GSQUERY;
1135 		return false;
1136 	}
1137 	pmc->mca_flags |= MAF_GSQUERY;
1138 	return true;
1139 }
1140 
1141 static int mld_force_mld_version(const struct inet6_dev *idev)
1142 {
1143 	/* Normally, both are 0 here. If enforcement to a particular is
1144 	 * being used, individual device enforcement will have a lower
1145 	 * precedence over 'all' device (.../conf/all/force_mld_version).
1146 	 */
1147 
1148 	if (dev_net(idev->dev)->ipv6.devconf_all->force_mld_version != 0)
1149 		return dev_net(idev->dev)->ipv6.devconf_all->force_mld_version;
1150 	else
1151 		return idev->cnf.force_mld_version;
1152 }
1153 
1154 static bool mld_in_v2_mode_only(const struct inet6_dev *idev)
1155 {
1156 	return mld_force_mld_version(idev) == 2;
1157 }
1158 
1159 static bool mld_in_v1_mode_only(const struct inet6_dev *idev)
1160 {
1161 	return mld_force_mld_version(idev) == 1;
1162 }
1163 
1164 static bool mld_in_v1_mode(const struct inet6_dev *idev)
1165 {
1166 	if (mld_in_v2_mode_only(idev))
1167 		return false;
1168 	if (mld_in_v1_mode_only(idev))
1169 		return true;
1170 	if (idev->mc_v1_seen && time_before(jiffies, idev->mc_v1_seen))
1171 		return true;
1172 
1173 	return false;
1174 }
1175 
1176 static void mld_set_v1_mode(struct inet6_dev *idev)
1177 {
1178 	/* RFC3810, relevant sections:
1179 	 *  - 9.1. Robustness Variable
1180 	 *  - 9.2. Query Interval
1181 	 *  - 9.3. Query Response Interval
1182 	 *  - 9.12. Older Version Querier Present Timeout
1183 	 */
1184 	unsigned long switchback;
1185 
1186 	switchback = (idev->mc_qrv * idev->mc_qi) + idev->mc_qri;
1187 
1188 	idev->mc_v1_seen = jiffies + switchback;
1189 }
1190 
1191 static void mld_update_qrv(struct inet6_dev *idev,
1192 			   const struct mld2_query *mlh2)
1193 {
1194 	/* RFC3810, relevant sections:
1195 	 *  - 5.1.8. QRV (Querier's Robustness Variable)
1196 	 *  - 9.1. Robustness Variable
1197 	 */
1198 
1199 	/* The value of the Robustness Variable MUST NOT be zero,
1200 	 * and SHOULD NOT be one. Catch this here if we ever run
1201 	 * into such a case in future.
1202 	 */
1203 	const int min_qrv = min(MLD_QRV_DEFAULT, sysctl_mld_qrv);
1204 	WARN_ON(idev->mc_qrv == 0);
1205 
1206 	if (mlh2->mld2q_qrv > 0)
1207 		idev->mc_qrv = mlh2->mld2q_qrv;
1208 
1209 	if (unlikely(idev->mc_qrv < min_qrv)) {
1210 		net_warn_ratelimited("IPv6: MLD: clamping QRV from %u to %u!\n",
1211 				     idev->mc_qrv, min_qrv);
1212 		idev->mc_qrv = min_qrv;
1213 	}
1214 }
1215 
1216 static void mld_update_qi(struct inet6_dev *idev,
1217 			  const struct mld2_query *mlh2)
1218 {
1219 	/* RFC3810, relevant sections:
1220 	 *  - 5.1.9. QQIC (Querier's Query Interval Code)
1221 	 *  - 9.2. Query Interval
1222 	 *  - 9.12. Older Version Querier Present Timeout
1223 	 *    (the [Query Interval] in the last Query received)
1224 	 */
1225 	unsigned long mc_qqi;
1226 
1227 	if (mlh2->mld2q_qqic < 128) {
1228 		mc_qqi = mlh2->mld2q_qqic;
1229 	} else {
1230 		unsigned long mc_man, mc_exp;
1231 
1232 		mc_exp = MLDV2_QQIC_EXP(mlh2->mld2q_qqic);
1233 		mc_man = MLDV2_QQIC_MAN(mlh2->mld2q_qqic);
1234 
1235 		mc_qqi = (mc_man | 0x10) << (mc_exp + 3);
1236 	}
1237 
1238 	idev->mc_qi = mc_qqi * HZ;
1239 }
1240 
1241 static void mld_update_qri(struct inet6_dev *idev,
1242 			   const struct mld2_query *mlh2)
1243 {
1244 	/* RFC3810, relevant sections:
1245 	 *  - 5.1.3. Maximum Response Code
1246 	 *  - 9.3. Query Response Interval
1247 	 */
1248 	idev->mc_qri = msecs_to_jiffies(mldv2_mrc(mlh2));
1249 }
1250 
1251 static int mld_process_v1(struct inet6_dev *idev, struct mld_msg *mld,
1252 			  unsigned long *max_delay, bool v1_query)
1253 {
1254 	unsigned long mldv1_md;
1255 
1256 	/* Ignore v1 queries */
1257 	if (mld_in_v2_mode_only(idev))
1258 		return -EINVAL;
1259 
1260 	mldv1_md = ntohs(mld->mld_maxdelay);
1261 
1262 	/* When in MLDv1 fallback and a MLDv2 router start-up being
1263 	 * unaware of current MLDv1 operation, the MRC == MRD mapping
1264 	 * only works when the exponential algorithm is not being
1265 	 * used (as MLDv1 is unaware of such things).
1266 	 *
1267 	 * According to the RFC author, the MLDv2 implementations
1268 	 * he's aware of all use a MRC < 32768 on start up queries.
1269 	 *
1270 	 * Thus, should we *ever* encounter something else larger
1271 	 * than that, just assume the maximum possible within our
1272 	 * reach.
1273 	 */
1274 	if (!v1_query)
1275 		mldv1_md = min(mldv1_md, MLDV1_MRD_MAX_COMPAT);
1276 
1277 	*max_delay = max(msecs_to_jiffies(mldv1_md), 1UL);
1278 
1279 	/* MLDv1 router present: we need to go into v1 mode *only*
1280 	 * when an MLDv1 query is received as per section 9.12. of
1281 	 * RFC3810! And we know from RFC2710 section 3.7 that MLDv1
1282 	 * queries MUST be of exactly 24 octets.
1283 	 */
1284 	if (v1_query)
1285 		mld_set_v1_mode(idev);
1286 
1287 	/* cancel MLDv2 report timer */
1288 	mld_gq_stop_timer(idev);
1289 	/* cancel the interface change timer */
1290 	mld_ifc_stop_timer(idev);
1291 	/* clear deleted report items */
1292 	mld_clear_delrec(idev);
1293 
1294 	return 0;
1295 }
1296 
1297 static int mld_process_v2(struct inet6_dev *idev, struct mld2_query *mld,
1298 			  unsigned long *max_delay)
1299 {
1300 	*max_delay = max(msecs_to_jiffies(mldv2_mrc(mld)), 1UL);
1301 
1302 	mld_update_qrv(idev, mld);
1303 	mld_update_qi(idev, mld);
1304 	mld_update_qri(idev, mld);
1305 
1306 	idev->mc_maxdelay = *max_delay;
1307 
1308 	return 0;
1309 }
1310 
1311 /* called with rcu_read_lock() */
1312 int igmp6_event_query(struct sk_buff *skb)
1313 {
1314 	struct mld2_query *mlh2 = NULL;
1315 	struct ifmcaddr6 *ma;
1316 	const struct in6_addr *group;
1317 	unsigned long max_delay;
1318 	struct inet6_dev *idev;
1319 	struct mld_msg *mld;
1320 	int group_type;
1321 	int mark = 0;
1322 	int len, err;
1323 
1324 	if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
1325 		return -EINVAL;
1326 
1327 	/* compute payload length excluding extension headers */
1328 	len = ntohs(ipv6_hdr(skb)->payload_len) + sizeof(struct ipv6hdr);
1329 	len -= skb_network_header_len(skb);
1330 
1331 	/* RFC3810 6.2
1332 	 * Upon reception of an MLD message that contains a Query, the node
1333 	 * checks if the source address of the message is a valid link-local
1334 	 * address, if the Hop Limit is set to 1, and if the Router Alert
1335 	 * option is present in the Hop-By-Hop Options header of the IPv6
1336 	 * packet.  If any of these checks fails, the packet is dropped.
1337 	 */
1338 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL) ||
1339 	    ipv6_hdr(skb)->hop_limit != 1 ||
1340 	    !(IP6CB(skb)->flags & IP6SKB_ROUTERALERT) ||
1341 	    IP6CB(skb)->ra != htons(IPV6_OPT_ROUTERALERT_MLD))
1342 		return -EINVAL;
1343 
1344 	idev = __in6_dev_get(skb->dev);
1345 	if (!idev)
1346 		return 0;
1347 
1348 	mld = (struct mld_msg *)icmp6_hdr(skb);
1349 	group = &mld->mld_mca;
1350 	group_type = ipv6_addr_type(group);
1351 
1352 	if (group_type != IPV6_ADDR_ANY &&
1353 	    !(group_type&IPV6_ADDR_MULTICAST))
1354 		return -EINVAL;
1355 
1356 	if (len < MLD_V1_QUERY_LEN) {
1357 		return -EINVAL;
1358 	} else if (len == MLD_V1_QUERY_LEN || mld_in_v1_mode(idev)) {
1359 		err = mld_process_v1(idev, mld, &max_delay,
1360 				     len == MLD_V1_QUERY_LEN);
1361 		if (err < 0)
1362 			return err;
1363 	} else if (len >= MLD_V2_QUERY_LEN_MIN) {
1364 		int srcs_offset = sizeof(struct mld2_query) -
1365 				  sizeof(struct icmp6hdr);
1366 
1367 		if (!pskb_may_pull(skb, srcs_offset))
1368 			return -EINVAL;
1369 
1370 		mlh2 = (struct mld2_query *)skb_transport_header(skb);
1371 
1372 		err = mld_process_v2(idev, mlh2, &max_delay);
1373 		if (err < 0)
1374 			return err;
1375 
1376 		if (group_type == IPV6_ADDR_ANY) { /* general query */
1377 			if (mlh2->mld2q_nsrcs)
1378 				return -EINVAL; /* no sources allowed */
1379 
1380 			mld_gq_start_timer(idev);
1381 			return 0;
1382 		}
1383 		/* mark sources to include, if group & source-specific */
1384 		if (mlh2->mld2q_nsrcs != 0) {
1385 			if (!pskb_may_pull(skb, srcs_offset +
1386 			    ntohs(mlh2->mld2q_nsrcs) * sizeof(struct in6_addr)))
1387 				return -EINVAL;
1388 
1389 			mlh2 = (struct mld2_query *)skb_transport_header(skb);
1390 			mark = 1;
1391 		}
1392 	} else {
1393 		return -EINVAL;
1394 	}
1395 
1396 	read_lock_bh(&idev->lock);
1397 	if (group_type == IPV6_ADDR_ANY) {
1398 		for (ma = idev->mc_list; ma; ma = ma->next) {
1399 			spin_lock_bh(&ma->mca_lock);
1400 			igmp6_group_queried(ma, max_delay);
1401 			spin_unlock_bh(&ma->mca_lock);
1402 		}
1403 	} else {
1404 		for (ma = idev->mc_list; ma; ma = ma->next) {
1405 			if (!ipv6_addr_equal(group, &ma->mca_addr))
1406 				continue;
1407 			spin_lock_bh(&ma->mca_lock);
1408 			if (ma->mca_flags & MAF_TIMER_RUNNING) {
1409 				/* gsquery <- gsquery && mark */
1410 				if (!mark)
1411 					ma->mca_flags &= ~MAF_GSQUERY;
1412 			} else {
1413 				/* gsquery <- mark */
1414 				if (mark)
1415 					ma->mca_flags |= MAF_GSQUERY;
1416 				else
1417 					ma->mca_flags &= ~MAF_GSQUERY;
1418 			}
1419 			if (!(ma->mca_flags & MAF_GSQUERY) ||
1420 			    mld_marksources(ma, ntohs(mlh2->mld2q_nsrcs), mlh2->mld2q_srcs))
1421 				igmp6_group_queried(ma, max_delay);
1422 			spin_unlock_bh(&ma->mca_lock);
1423 			break;
1424 		}
1425 	}
1426 	read_unlock_bh(&idev->lock);
1427 
1428 	return 0;
1429 }
1430 
1431 /* called with rcu_read_lock() */
1432 int igmp6_event_report(struct sk_buff *skb)
1433 {
1434 	struct ifmcaddr6 *ma;
1435 	struct inet6_dev *idev;
1436 	struct mld_msg *mld;
1437 	int addr_type;
1438 
1439 	/* Our own report looped back. Ignore it. */
1440 	if (skb->pkt_type == PACKET_LOOPBACK)
1441 		return 0;
1442 
1443 	/* send our report if the MC router may not have heard this report */
1444 	if (skb->pkt_type != PACKET_MULTICAST &&
1445 	    skb->pkt_type != PACKET_BROADCAST)
1446 		return 0;
1447 
1448 	if (!pskb_may_pull(skb, sizeof(*mld) - sizeof(struct icmp6hdr)))
1449 		return -EINVAL;
1450 
1451 	mld = (struct mld_msg *)icmp6_hdr(skb);
1452 
1453 	/* Drop reports with not link local source */
1454 	addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr);
1455 	if (addr_type != IPV6_ADDR_ANY &&
1456 	    !(addr_type&IPV6_ADDR_LINKLOCAL))
1457 		return -EINVAL;
1458 
1459 	idev = __in6_dev_get(skb->dev);
1460 	if (!idev)
1461 		return -ENODEV;
1462 
1463 	/*
1464 	 *	Cancel the timer for this group
1465 	 */
1466 
1467 	read_lock_bh(&idev->lock);
1468 	for (ma = idev->mc_list; ma; ma = ma->next) {
1469 		if (ipv6_addr_equal(&ma->mca_addr, &mld->mld_mca)) {
1470 			spin_lock(&ma->mca_lock);
1471 			if (del_timer(&ma->mca_timer))
1472 				refcount_dec(&ma->mca_refcnt);
1473 			ma->mca_flags &= ~(MAF_LAST_REPORTER|MAF_TIMER_RUNNING);
1474 			spin_unlock(&ma->mca_lock);
1475 			break;
1476 		}
1477 	}
1478 	read_unlock_bh(&idev->lock);
1479 	return 0;
1480 }
1481 
1482 static bool is_in(struct ifmcaddr6 *pmc, struct ip6_sf_list *psf, int type,
1483 		  int gdeleted, int sdeleted)
1484 {
1485 	switch (type) {
1486 	case MLD2_MODE_IS_INCLUDE:
1487 	case MLD2_MODE_IS_EXCLUDE:
1488 		if (gdeleted || sdeleted)
1489 			return false;
1490 		if (!((pmc->mca_flags & MAF_GSQUERY) && !psf->sf_gsresp)) {
1491 			if (pmc->mca_sfmode == MCAST_INCLUDE)
1492 				return true;
1493 			/* don't include if this source is excluded
1494 			 * in all filters
1495 			 */
1496 			if (psf->sf_count[MCAST_INCLUDE])
1497 				return type == MLD2_MODE_IS_INCLUDE;
1498 			return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1499 				psf->sf_count[MCAST_EXCLUDE];
1500 		}
1501 		return false;
1502 	case MLD2_CHANGE_TO_INCLUDE:
1503 		if (gdeleted || sdeleted)
1504 			return false;
1505 		return psf->sf_count[MCAST_INCLUDE] != 0;
1506 	case MLD2_CHANGE_TO_EXCLUDE:
1507 		if (gdeleted || sdeleted)
1508 			return false;
1509 		if (pmc->mca_sfcount[MCAST_EXCLUDE] == 0 ||
1510 		    psf->sf_count[MCAST_INCLUDE])
1511 			return false;
1512 		return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1513 			psf->sf_count[MCAST_EXCLUDE];
1514 	case MLD2_ALLOW_NEW_SOURCES:
1515 		if (gdeleted || !psf->sf_crcount)
1516 			return false;
1517 		return (pmc->mca_sfmode == MCAST_INCLUDE) ^ sdeleted;
1518 	case MLD2_BLOCK_OLD_SOURCES:
1519 		if (pmc->mca_sfmode == MCAST_INCLUDE)
1520 			return gdeleted || (psf->sf_crcount && sdeleted);
1521 		return psf->sf_crcount && !gdeleted && !sdeleted;
1522 	}
1523 	return false;
1524 }
1525 
1526 static int
1527 mld_scount(struct ifmcaddr6 *pmc, int type, int gdeleted, int sdeleted)
1528 {
1529 	struct ip6_sf_list *psf;
1530 	int scount = 0;
1531 
1532 	for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1533 		if (!is_in(pmc, psf, type, gdeleted, sdeleted))
1534 			continue;
1535 		scount++;
1536 	}
1537 	return scount;
1538 }
1539 
1540 static void ip6_mc_hdr(struct sock *sk, struct sk_buff *skb,
1541 		       struct net_device *dev,
1542 		       const struct in6_addr *saddr,
1543 		       const struct in6_addr *daddr,
1544 		       int proto, int len)
1545 {
1546 	struct ipv6hdr *hdr;
1547 
1548 	skb->protocol = htons(ETH_P_IPV6);
1549 	skb->dev = dev;
1550 
1551 	skb_reset_network_header(skb);
1552 	skb_put(skb, sizeof(struct ipv6hdr));
1553 	hdr = ipv6_hdr(skb);
1554 
1555 	ip6_flow_hdr(hdr, 0, 0);
1556 
1557 	hdr->payload_len = htons(len);
1558 	hdr->nexthdr = proto;
1559 	hdr->hop_limit = inet6_sk(sk)->hop_limit;
1560 
1561 	hdr->saddr = *saddr;
1562 	hdr->daddr = *daddr;
1563 }
1564 
1565 static struct sk_buff *mld_newpack(struct inet6_dev *idev, unsigned int mtu)
1566 {
1567 	struct net_device *dev = idev->dev;
1568 	struct net *net = dev_net(dev);
1569 	struct sock *sk = net->ipv6.igmp_sk;
1570 	struct sk_buff *skb;
1571 	struct mld2_report *pmr;
1572 	struct in6_addr addr_buf;
1573 	const struct in6_addr *saddr;
1574 	int hlen = LL_RESERVED_SPACE(dev);
1575 	int tlen = dev->needed_tailroom;
1576 	unsigned int size = mtu + hlen + tlen;
1577 	int err;
1578 	u8 ra[8] = { IPPROTO_ICMPV6, 0,
1579 		     IPV6_TLV_ROUTERALERT, 2, 0, 0,
1580 		     IPV6_TLV_PADN, 0 };
1581 
1582 	/* we assume size > sizeof(ra) here */
1583 	/* limit our allocations to order-0 page */
1584 	size = min_t(int, size, SKB_MAX_ORDER(0, 0));
1585 	skb = sock_alloc_send_skb(sk, size, 1, &err);
1586 
1587 	if (!skb)
1588 		return NULL;
1589 
1590 	skb->priority = TC_PRIO_CONTROL;
1591 	skb_reserve(skb, hlen);
1592 	skb_tailroom_reserve(skb, mtu, tlen);
1593 
1594 	if (__ipv6_get_lladdr(idev, &addr_buf, IFA_F_TENTATIVE)) {
1595 		/* <draft-ietf-magma-mld-source-05.txt>:
1596 		 * use unspecified address as the source address
1597 		 * when a valid link-local address is not available.
1598 		 */
1599 		saddr = &in6addr_any;
1600 	} else
1601 		saddr = &addr_buf;
1602 
1603 	ip6_mc_hdr(sk, skb, dev, saddr, &mld2_all_mcr, NEXTHDR_HOP, 0);
1604 
1605 	skb_put_data(skb, ra, sizeof(ra));
1606 
1607 	skb_set_transport_header(skb, skb_tail_pointer(skb) - skb->data);
1608 	skb_put(skb, sizeof(*pmr));
1609 	pmr = (struct mld2_report *)skb_transport_header(skb);
1610 	pmr->mld2r_type = ICMPV6_MLD2_REPORT;
1611 	pmr->mld2r_resv1 = 0;
1612 	pmr->mld2r_cksum = 0;
1613 	pmr->mld2r_resv2 = 0;
1614 	pmr->mld2r_ngrec = 0;
1615 	return skb;
1616 }
1617 
1618 static void mld_sendpack(struct sk_buff *skb)
1619 {
1620 	struct ipv6hdr *pip6 = ipv6_hdr(skb);
1621 	struct mld2_report *pmr =
1622 			      (struct mld2_report *)skb_transport_header(skb);
1623 	int payload_len, mldlen;
1624 	struct inet6_dev *idev;
1625 	struct net *net = dev_net(skb->dev);
1626 	int err;
1627 	struct flowi6 fl6;
1628 	struct dst_entry *dst;
1629 
1630 	rcu_read_lock();
1631 	idev = __in6_dev_get(skb->dev);
1632 	IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
1633 
1634 	payload_len = (skb_tail_pointer(skb) - skb_network_header(skb)) -
1635 		sizeof(*pip6);
1636 	mldlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1637 	pip6->payload_len = htons(payload_len);
1638 
1639 	pmr->mld2r_cksum = csum_ipv6_magic(&pip6->saddr, &pip6->daddr, mldlen,
1640 					   IPPROTO_ICMPV6,
1641 					   csum_partial(skb_transport_header(skb),
1642 							mldlen, 0));
1643 
1644 	icmpv6_flow_init(net->ipv6.igmp_sk, &fl6, ICMPV6_MLD2_REPORT,
1645 			 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
1646 			 skb->dev->ifindex);
1647 	dst = icmp6_dst_alloc(skb->dev, &fl6);
1648 
1649 	err = 0;
1650 	if (IS_ERR(dst)) {
1651 		err = PTR_ERR(dst);
1652 		dst = NULL;
1653 	}
1654 	skb_dst_set(skb, dst);
1655 	if (err)
1656 		goto err_out;
1657 
1658 	payload_len = skb->len;
1659 
1660 	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
1661 		      net, net->ipv6.igmp_sk, skb, NULL, skb->dev,
1662 		      dst_output);
1663 out:
1664 	if (!err) {
1665 		ICMP6MSGOUT_INC_STATS(net, idev, ICMPV6_MLD2_REPORT);
1666 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1667 	} else {
1668 		IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
1669 	}
1670 
1671 	rcu_read_unlock();
1672 	return;
1673 
1674 err_out:
1675 	kfree_skb(skb);
1676 	goto out;
1677 }
1678 
1679 static int grec_size(struct ifmcaddr6 *pmc, int type, int gdel, int sdel)
1680 {
1681 	return sizeof(struct mld2_grec) + 16 * mld_scount(pmc,type,gdel,sdel);
1682 }
1683 
1684 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1685 	int type, struct mld2_grec **ppgr)
1686 {
1687 	struct net_device *dev = pmc->idev->dev;
1688 	struct mld2_report *pmr;
1689 	struct mld2_grec *pgr;
1690 
1691 	if (!skb)
1692 		skb = mld_newpack(pmc->idev, dev->mtu);
1693 	if (!skb)
1694 		return NULL;
1695 	pgr = skb_put(skb, sizeof(struct mld2_grec));
1696 	pgr->grec_type = type;
1697 	pgr->grec_auxwords = 0;
1698 	pgr->grec_nsrcs = 0;
1699 	pgr->grec_mca = pmc->mca_addr;	/* structure copy */
1700 	pmr = (struct mld2_report *)skb_transport_header(skb);
1701 	pmr->mld2r_ngrec = htons(ntohs(pmr->mld2r_ngrec)+1);
1702 	*ppgr = pgr;
1703 	return skb;
1704 }
1705 
1706 #define AVAILABLE(skb)	((skb) ? skb_availroom(skb) : 0)
1707 
1708 static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1709 	int type, int gdeleted, int sdeleted, int crsend)
1710 {
1711 	struct inet6_dev *idev = pmc->idev;
1712 	struct net_device *dev = idev->dev;
1713 	struct mld2_report *pmr;
1714 	struct mld2_grec *pgr = NULL;
1715 	struct ip6_sf_list *psf, *psf_next, *psf_prev, **psf_list;
1716 	int scount, stotal, first, isquery, truncate;
1717 
1718 	if (pmc->mca_flags & MAF_NOREPORT)
1719 		return skb;
1720 
1721 	isquery = type == MLD2_MODE_IS_INCLUDE ||
1722 		  type == MLD2_MODE_IS_EXCLUDE;
1723 	truncate = type == MLD2_MODE_IS_EXCLUDE ||
1724 		    type == MLD2_CHANGE_TO_EXCLUDE;
1725 
1726 	stotal = scount = 0;
1727 
1728 	psf_list = sdeleted ? &pmc->mca_tomb : &pmc->mca_sources;
1729 
1730 	if (!*psf_list)
1731 		goto empty_source;
1732 
1733 	pmr = skb ? (struct mld2_report *)skb_transport_header(skb) : NULL;
1734 
1735 	/* EX and TO_EX get a fresh packet, if needed */
1736 	if (truncate) {
1737 		if (pmr && pmr->mld2r_ngrec &&
1738 		    AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
1739 			if (skb)
1740 				mld_sendpack(skb);
1741 			skb = mld_newpack(idev, dev->mtu);
1742 		}
1743 	}
1744 	first = 1;
1745 	psf_prev = NULL;
1746 	for (psf = *psf_list; psf; psf = psf_next) {
1747 		struct in6_addr *psrc;
1748 
1749 		psf_next = psf->sf_next;
1750 
1751 		if (!is_in(pmc, psf, type, gdeleted, sdeleted)) {
1752 			psf_prev = psf;
1753 			continue;
1754 		}
1755 
1756 		/* Based on RFC3810 6.1. Should not send source-list change
1757 		 * records when there is a filter mode change.
1758 		 */
1759 		if (((gdeleted && pmc->mca_sfmode == MCAST_EXCLUDE) ||
1760 		     (!gdeleted && pmc->mca_crcount)) &&
1761 		    (type == MLD2_ALLOW_NEW_SOURCES ||
1762 		     type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount)
1763 			goto decrease_sf_crcount;
1764 
1765 		/* clear marks on query responses */
1766 		if (isquery)
1767 			psf->sf_gsresp = 0;
1768 
1769 		if (AVAILABLE(skb) < sizeof(*psrc) +
1770 		    first*sizeof(struct mld2_grec)) {
1771 			if (truncate && !first)
1772 				break;	 /* truncate these */
1773 			if (pgr)
1774 				pgr->grec_nsrcs = htons(scount);
1775 			if (skb)
1776 				mld_sendpack(skb);
1777 			skb = mld_newpack(idev, dev->mtu);
1778 			first = 1;
1779 			scount = 0;
1780 		}
1781 		if (first) {
1782 			skb = add_grhead(skb, pmc, type, &pgr);
1783 			first = 0;
1784 		}
1785 		if (!skb)
1786 			return NULL;
1787 		psrc = skb_put(skb, sizeof(*psrc));
1788 		*psrc = psf->sf_addr;
1789 		scount++; stotal++;
1790 		if ((type == MLD2_ALLOW_NEW_SOURCES ||
1791 		     type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
1792 decrease_sf_crcount:
1793 			psf->sf_crcount--;
1794 			if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
1795 				if (psf_prev)
1796 					psf_prev->sf_next = psf->sf_next;
1797 				else
1798 					*psf_list = psf->sf_next;
1799 				kfree(psf);
1800 				continue;
1801 			}
1802 		}
1803 		psf_prev = psf;
1804 	}
1805 
1806 empty_source:
1807 	if (!stotal) {
1808 		if (type == MLD2_ALLOW_NEW_SOURCES ||
1809 		    type == MLD2_BLOCK_OLD_SOURCES)
1810 			return skb;
1811 		if (pmc->mca_crcount || isquery || crsend) {
1812 			/* make sure we have room for group header */
1813 			if (skb && AVAILABLE(skb) < sizeof(struct mld2_grec)) {
1814 				mld_sendpack(skb);
1815 				skb = NULL; /* add_grhead will get a new one */
1816 			}
1817 			skb = add_grhead(skb, pmc, type, &pgr);
1818 		}
1819 	}
1820 	if (pgr)
1821 		pgr->grec_nsrcs = htons(scount);
1822 
1823 	if (isquery)
1824 		pmc->mca_flags &= ~MAF_GSQUERY;	/* clear query state */
1825 	return skb;
1826 }
1827 
1828 static void mld_send_report(struct inet6_dev *idev, struct ifmcaddr6 *pmc)
1829 {
1830 	struct sk_buff *skb = NULL;
1831 	int type;
1832 
1833 	read_lock_bh(&idev->lock);
1834 	if (!pmc) {
1835 		for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
1836 			if (pmc->mca_flags & MAF_NOREPORT)
1837 				continue;
1838 			spin_lock_bh(&pmc->mca_lock);
1839 			if (pmc->mca_sfcount[MCAST_EXCLUDE])
1840 				type = MLD2_MODE_IS_EXCLUDE;
1841 			else
1842 				type = MLD2_MODE_IS_INCLUDE;
1843 			skb = add_grec(skb, pmc, type, 0, 0, 0);
1844 			spin_unlock_bh(&pmc->mca_lock);
1845 		}
1846 	} else {
1847 		spin_lock_bh(&pmc->mca_lock);
1848 		if (pmc->mca_sfcount[MCAST_EXCLUDE])
1849 			type = MLD2_MODE_IS_EXCLUDE;
1850 		else
1851 			type = MLD2_MODE_IS_INCLUDE;
1852 		skb = add_grec(skb, pmc, type, 0, 0, 0);
1853 		spin_unlock_bh(&pmc->mca_lock);
1854 	}
1855 	read_unlock_bh(&idev->lock);
1856 	if (skb)
1857 		mld_sendpack(skb);
1858 }
1859 
1860 /*
1861  * remove zero-count source records from a source filter list
1862  */
1863 static void mld_clear_zeros(struct ip6_sf_list **ppsf)
1864 {
1865 	struct ip6_sf_list *psf_prev, *psf_next, *psf;
1866 
1867 	psf_prev = NULL;
1868 	for (psf = *ppsf; psf; psf = psf_next) {
1869 		psf_next = psf->sf_next;
1870 		if (psf->sf_crcount == 0) {
1871 			if (psf_prev)
1872 				psf_prev->sf_next = psf->sf_next;
1873 			else
1874 				*ppsf = psf->sf_next;
1875 			kfree(psf);
1876 		} else
1877 			psf_prev = psf;
1878 	}
1879 }
1880 
1881 static void mld_send_cr(struct inet6_dev *idev)
1882 {
1883 	struct ifmcaddr6 *pmc, *pmc_prev, *pmc_next;
1884 	struct sk_buff *skb = NULL;
1885 	int type, dtype;
1886 
1887 	read_lock_bh(&idev->lock);
1888 	spin_lock(&idev->mc_lock);
1889 
1890 	/* deleted MCA's */
1891 	pmc_prev = NULL;
1892 	for (pmc = idev->mc_tomb; pmc; pmc = pmc_next) {
1893 		pmc_next = pmc->next;
1894 		if (pmc->mca_sfmode == MCAST_INCLUDE) {
1895 			type = MLD2_BLOCK_OLD_SOURCES;
1896 			dtype = MLD2_BLOCK_OLD_SOURCES;
1897 			skb = add_grec(skb, pmc, type, 1, 0, 0);
1898 			skb = add_grec(skb, pmc, dtype, 1, 1, 0);
1899 		}
1900 		if (pmc->mca_crcount) {
1901 			if (pmc->mca_sfmode == MCAST_EXCLUDE) {
1902 				type = MLD2_CHANGE_TO_INCLUDE;
1903 				skb = add_grec(skb, pmc, type, 1, 0, 0);
1904 			}
1905 			pmc->mca_crcount--;
1906 			if (pmc->mca_crcount == 0) {
1907 				mld_clear_zeros(&pmc->mca_tomb);
1908 				mld_clear_zeros(&pmc->mca_sources);
1909 			}
1910 		}
1911 		if (pmc->mca_crcount == 0 && !pmc->mca_tomb &&
1912 		    !pmc->mca_sources) {
1913 			if (pmc_prev)
1914 				pmc_prev->next = pmc_next;
1915 			else
1916 				idev->mc_tomb = pmc_next;
1917 			in6_dev_put(pmc->idev);
1918 			kfree(pmc);
1919 		} else
1920 			pmc_prev = pmc;
1921 	}
1922 	spin_unlock(&idev->mc_lock);
1923 
1924 	/* change recs */
1925 	for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
1926 		spin_lock_bh(&pmc->mca_lock);
1927 		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
1928 			type = MLD2_BLOCK_OLD_SOURCES;
1929 			dtype = MLD2_ALLOW_NEW_SOURCES;
1930 		} else {
1931 			type = MLD2_ALLOW_NEW_SOURCES;
1932 			dtype = MLD2_BLOCK_OLD_SOURCES;
1933 		}
1934 		skb = add_grec(skb, pmc, type, 0, 0, 0);
1935 		skb = add_grec(skb, pmc, dtype, 0, 1, 0);	/* deleted sources */
1936 
1937 		/* filter mode changes */
1938 		if (pmc->mca_crcount) {
1939 			if (pmc->mca_sfmode == MCAST_EXCLUDE)
1940 				type = MLD2_CHANGE_TO_EXCLUDE;
1941 			else
1942 				type = MLD2_CHANGE_TO_INCLUDE;
1943 			skb = add_grec(skb, pmc, type, 0, 0, 0);
1944 			pmc->mca_crcount--;
1945 		}
1946 		spin_unlock_bh(&pmc->mca_lock);
1947 	}
1948 	read_unlock_bh(&idev->lock);
1949 	if (!skb)
1950 		return;
1951 	(void) mld_sendpack(skb);
1952 }
1953 
1954 static void igmp6_send(struct in6_addr *addr, struct net_device *dev, int type)
1955 {
1956 	struct net *net = dev_net(dev);
1957 	struct sock *sk = net->ipv6.igmp_sk;
1958 	struct inet6_dev *idev;
1959 	struct sk_buff *skb;
1960 	struct mld_msg *hdr;
1961 	const struct in6_addr *snd_addr, *saddr;
1962 	struct in6_addr addr_buf;
1963 	int hlen = LL_RESERVED_SPACE(dev);
1964 	int tlen = dev->needed_tailroom;
1965 	int err, len, payload_len, full_len;
1966 	u8 ra[8] = { IPPROTO_ICMPV6, 0,
1967 		     IPV6_TLV_ROUTERALERT, 2, 0, 0,
1968 		     IPV6_TLV_PADN, 0 };
1969 	struct flowi6 fl6;
1970 	struct dst_entry *dst;
1971 
1972 	if (type == ICMPV6_MGM_REDUCTION)
1973 		snd_addr = &in6addr_linklocal_allrouters;
1974 	else
1975 		snd_addr = addr;
1976 
1977 	len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
1978 	payload_len = len + sizeof(ra);
1979 	full_len = sizeof(struct ipv6hdr) + payload_len;
1980 
1981 	rcu_read_lock();
1982 	IP6_UPD_PO_STATS(net, __in6_dev_get(dev),
1983 		      IPSTATS_MIB_OUT, full_len);
1984 	rcu_read_unlock();
1985 
1986 	skb = sock_alloc_send_skb(sk, hlen + tlen + full_len, 1, &err);
1987 
1988 	if (!skb) {
1989 		rcu_read_lock();
1990 		IP6_INC_STATS(net, __in6_dev_get(dev),
1991 			      IPSTATS_MIB_OUTDISCARDS);
1992 		rcu_read_unlock();
1993 		return;
1994 	}
1995 	skb->priority = TC_PRIO_CONTROL;
1996 	skb_reserve(skb, hlen);
1997 
1998 	if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
1999 		/* <draft-ietf-magma-mld-source-05.txt>:
2000 		 * use unspecified address as the source address
2001 		 * when a valid link-local address is not available.
2002 		 */
2003 		saddr = &in6addr_any;
2004 	} else
2005 		saddr = &addr_buf;
2006 
2007 	ip6_mc_hdr(sk, skb, dev, saddr, snd_addr, NEXTHDR_HOP, payload_len);
2008 
2009 	skb_put_data(skb, ra, sizeof(ra));
2010 
2011 	hdr = skb_put_zero(skb, sizeof(struct mld_msg));
2012 	hdr->mld_type = type;
2013 	hdr->mld_mca = *addr;
2014 
2015 	hdr->mld_cksum = csum_ipv6_magic(saddr, snd_addr, len,
2016 					 IPPROTO_ICMPV6,
2017 					 csum_partial(hdr, len, 0));
2018 
2019 	rcu_read_lock();
2020 	idev = __in6_dev_get(skb->dev);
2021 
2022 	icmpv6_flow_init(sk, &fl6, type,
2023 			 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
2024 			 skb->dev->ifindex);
2025 	dst = icmp6_dst_alloc(skb->dev, &fl6);
2026 	if (IS_ERR(dst)) {
2027 		err = PTR_ERR(dst);
2028 		goto err_out;
2029 	}
2030 
2031 	skb_dst_set(skb, dst);
2032 	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
2033 		      net, sk, skb, NULL, skb->dev,
2034 		      dst_output);
2035 out:
2036 	if (!err) {
2037 		ICMP6MSGOUT_INC_STATS(net, idev, type);
2038 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
2039 	} else
2040 		IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
2041 
2042 	rcu_read_unlock();
2043 	return;
2044 
2045 err_out:
2046 	kfree_skb(skb);
2047 	goto out;
2048 }
2049 
2050 static void mld_send_initial_cr(struct inet6_dev *idev)
2051 {
2052 	struct sk_buff *skb;
2053 	struct ifmcaddr6 *pmc;
2054 	int type;
2055 
2056 	if (mld_in_v1_mode(idev))
2057 		return;
2058 
2059 	skb = NULL;
2060 	read_lock_bh(&idev->lock);
2061 	for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2062 		spin_lock_bh(&pmc->mca_lock);
2063 		if (pmc->mca_sfcount[MCAST_EXCLUDE])
2064 			type = MLD2_CHANGE_TO_EXCLUDE;
2065 		else
2066 			type = MLD2_CHANGE_TO_INCLUDE;
2067 		skb = add_grec(skb, pmc, type, 0, 0, 1);
2068 		spin_unlock_bh(&pmc->mca_lock);
2069 	}
2070 	read_unlock_bh(&idev->lock);
2071 	if (skb)
2072 		mld_sendpack(skb);
2073 }
2074 
2075 void ipv6_mc_dad_complete(struct inet6_dev *idev)
2076 {
2077 	idev->mc_dad_count = idev->mc_qrv;
2078 	if (idev->mc_dad_count) {
2079 		mld_send_initial_cr(idev);
2080 		idev->mc_dad_count--;
2081 		if (idev->mc_dad_count)
2082 			mld_dad_start_timer(idev, idev->mc_maxdelay);
2083 	}
2084 }
2085 
2086 static void mld_dad_timer_expire(unsigned long data)
2087 {
2088 	struct inet6_dev *idev = (struct inet6_dev *)data;
2089 
2090 	mld_send_initial_cr(idev);
2091 	if (idev->mc_dad_count) {
2092 		idev->mc_dad_count--;
2093 		if (idev->mc_dad_count)
2094 			mld_dad_start_timer(idev, idev->mc_maxdelay);
2095 	}
2096 	in6_dev_put(idev);
2097 }
2098 
2099 static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
2100 	const struct in6_addr *psfsrc)
2101 {
2102 	struct ip6_sf_list *psf, *psf_prev;
2103 	int rv = 0;
2104 
2105 	psf_prev = NULL;
2106 	for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2107 		if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2108 			break;
2109 		psf_prev = psf;
2110 	}
2111 	if (!psf || psf->sf_count[sfmode] == 0) {
2112 		/* source filter not found, or count wrong =>  bug */
2113 		return -ESRCH;
2114 	}
2115 	psf->sf_count[sfmode]--;
2116 	if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
2117 		struct inet6_dev *idev = pmc->idev;
2118 
2119 		/* no more filters for this source */
2120 		if (psf_prev)
2121 			psf_prev->sf_next = psf->sf_next;
2122 		else
2123 			pmc->mca_sources = psf->sf_next;
2124 		if (psf->sf_oldin && !(pmc->mca_flags & MAF_NOREPORT) &&
2125 		    !mld_in_v1_mode(idev)) {
2126 			psf->sf_crcount = idev->mc_qrv;
2127 			psf->sf_next = pmc->mca_tomb;
2128 			pmc->mca_tomb = psf;
2129 			rv = 1;
2130 		} else
2131 			kfree(psf);
2132 	}
2133 	return rv;
2134 }
2135 
2136 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2137 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
2138 			  int delta)
2139 {
2140 	struct ifmcaddr6 *pmc;
2141 	int	changerec = 0;
2142 	int	i, err;
2143 
2144 	if (!idev)
2145 		return -ENODEV;
2146 	read_lock_bh(&idev->lock);
2147 	for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2148 		if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2149 			break;
2150 	}
2151 	if (!pmc) {
2152 		/* MCA not found?? bug */
2153 		read_unlock_bh(&idev->lock);
2154 		return -ESRCH;
2155 	}
2156 	spin_lock_bh(&pmc->mca_lock);
2157 	sf_markstate(pmc);
2158 	if (!delta) {
2159 		if (!pmc->mca_sfcount[sfmode]) {
2160 			spin_unlock_bh(&pmc->mca_lock);
2161 			read_unlock_bh(&idev->lock);
2162 			return -EINVAL;
2163 		}
2164 		pmc->mca_sfcount[sfmode]--;
2165 	}
2166 	err = 0;
2167 	for (i = 0; i < sfcount; i++) {
2168 		int rv = ip6_mc_del1_src(pmc, sfmode, &psfsrc[i]);
2169 
2170 		changerec |= rv > 0;
2171 		if (!err && rv < 0)
2172 			err = rv;
2173 	}
2174 	if (pmc->mca_sfmode == MCAST_EXCLUDE &&
2175 	    pmc->mca_sfcount[MCAST_EXCLUDE] == 0 &&
2176 	    pmc->mca_sfcount[MCAST_INCLUDE]) {
2177 		struct ip6_sf_list *psf;
2178 
2179 		/* filter mode change */
2180 		pmc->mca_sfmode = MCAST_INCLUDE;
2181 		pmc->mca_crcount = idev->mc_qrv;
2182 		idev->mc_ifc_count = pmc->mca_crcount;
2183 		for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2184 			psf->sf_crcount = 0;
2185 		mld_ifc_event(pmc->idev);
2186 	} else if (sf_setstate(pmc) || changerec)
2187 		mld_ifc_event(pmc->idev);
2188 	spin_unlock_bh(&pmc->mca_lock);
2189 	read_unlock_bh(&idev->lock);
2190 	return err;
2191 }
2192 
2193 /*
2194  * Add multicast single-source filter to the interface list
2195  */
2196 static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode,
2197 	const struct in6_addr *psfsrc)
2198 {
2199 	struct ip6_sf_list *psf, *psf_prev;
2200 
2201 	psf_prev = NULL;
2202 	for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2203 		if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2204 			break;
2205 		psf_prev = psf;
2206 	}
2207 	if (!psf) {
2208 		psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
2209 		if (!psf)
2210 			return -ENOBUFS;
2211 
2212 		psf->sf_addr = *psfsrc;
2213 		if (psf_prev) {
2214 			psf_prev->sf_next = psf;
2215 		} else
2216 			pmc->mca_sources = psf;
2217 	}
2218 	psf->sf_count[sfmode]++;
2219 	return 0;
2220 }
2221 
2222 static void sf_markstate(struct ifmcaddr6 *pmc)
2223 {
2224 	struct ip6_sf_list *psf;
2225 	int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2226 
2227 	for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2228 		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2229 			psf->sf_oldin = mca_xcount ==
2230 				psf->sf_count[MCAST_EXCLUDE] &&
2231 				!psf->sf_count[MCAST_INCLUDE];
2232 		} else
2233 			psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
2234 }
2235 
2236 static int sf_setstate(struct ifmcaddr6 *pmc)
2237 {
2238 	struct ip6_sf_list *psf, *dpsf;
2239 	int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2240 	int qrv = pmc->idev->mc_qrv;
2241 	int new_in, rv;
2242 
2243 	rv = 0;
2244 	for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2245 		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2246 			new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
2247 				!psf->sf_count[MCAST_INCLUDE];
2248 		} else
2249 			new_in = psf->sf_count[MCAST_INCLUDE] != 0;
2250 		if (new_in) {
2251 			if (!psf->sf_oldin) {
2252 				struct ip6_sf_list *prev = NULL;
2253 
2254 				for (dpsf = pmc->mca_tomb; dpsf;
2255 				     dpsf = dpsf->sf_next) {
2256 					if (ipv6_addr_equal(&dpsf->sf_addr,
2257 					    &psf->sf_addr))
2258 						break;
2259 					prev = dpsf;
2260 				}
2261 				if (dpsf) {
2262 					if (prev)
2263 						prev->sf_next = dpsf->sf_next;
2264 					else
2265 						pmc->mca_tomb = dpsf->sf_next;
2266 					kfree(dpsf);
2267 				}
2268 				psf->sf_crcount = qrv;
2269 				rv++;
2270 			}
2271 		} else if (psf->sf_oldin) {
2272 			psf->sf_crcount = 0;
2273 			/*
2274 			 * add or update "delete" records if an active filter
2275 			 * is now inactive
2276 			 */
2277 			for (dpsf = pmc->mca_tomb; dpsf; dpsf = dpsf->sf_next)
2278 				if (ipv6_addr_equal(&dpsf->sf_addr,
2279 				    &psf->sf_addr))
2280 					break;
2281 			if (!dpsf) {
2282 				dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
2283 				if (!dpsf)
2284 					continue;
2285 				*dpsf = *psf;
2286 				/* pmc->mca_lock held by callers */
2287 				dpsf->sf_next = pmc->mca_tomb;
2288 				pmc->mca_tomb = dpsf;
2289 			}
2290 			dpsf->sf_crcount = qrv;
2291 			rv++;
2292 		}
2293 	}
2294 	return rv;
2295 }
2296 
2297 /*
2298  * Add multicast source filter list to the interface list
2299  */
2300 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2301 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
2302 			  int delta)
2303 {
2304 	struct ifmcaddr6 *pmc;
2305 	int	isexclude;
2306 	int	i, err;
2307 
2308 	if (!idev)
2309 		return -ENODEV;
2310 	read_lock_bh(&idev->lock);
2311 	for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2312 		if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2313 			break;
2314 	}
2315 	if (!pmc) {
2316 		/* MCA not found?? bug */
2317 		read_unlock_bh(&idev->lock);
2318 		return -ESRCH;
2319 	}
2320 	spin_lock_bh(&pmc->mca_lock);
2321 
2322 	sf_markstate(pmc);
2323 	isexclude = pmc->mca_sfmode == MCAST_EXCLUDE;
2324 	if (!delta)
2325 		pmc->mca_sfcount[sfmode]++;
2326 	err = 0;
2327 	for (i = 0; i < sfcount; i++) {
2328 		err = ip6_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2329 		if (err)
2330 			break;
2331 	}
2332 	if (err) {
2333 		int j;
2334 
2335 		if (!delta)
2336 			pmc->mca_sfcount[sfmode]--;
2337 		for (j = 0; j < i; j++)
2338 			ip6_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2339 	} else if (isexclude != (pmc->mca_sfcount[MCAST_EXCLUDE] != 0)) {
2340 		struct ip6_sf_list *psf;
2341 
2342 		/* filter mode change */
2343 		if (pmc->mca_sfcount[MCAST_EXCLUDE])
2344 			pmc->mca_sfmode = MCAST_EXCLUDE;
2345 		else if (pmc->mca_sfcount[MCAST_INCLUDE])
2346 			pmc->mca_sfmode = MCAST_INCLUDE;
2347 		/* else no filters; keep old mode for reports */
2348 
2349 		pmc->mca_crcount = idev->mc_qrv;
2350 		idev->mc_ifc_count = pmc->mca_crcount;
2351 		for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2352 			psf->sf_crcount = 0;
2353 		mld_ifc_event(idev);
2354 	} else if (sf_setstate(pmc))
2355 		mld_ifc_event(idev);
2356 	spin_unlock_bh(&pmc->mca_lock);
2357 	read_unlock_bh(&idev->lock);
2358 	return err;
2359 }
2360 
2361 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc)
2362 {
2363 	struct ip6_sf_list *psf, *nextpsf;
2364 
2365 	for (psf = pmc->mca_tomb; psf; psf = nextpsf) {
2366 		nextpsf = psf->sf_next;
2367 		kfree(psf);
2368 	}
2369 	pmc->mca_tomb = NULL;
2370 	for (psf = pmc->mca_sources; psf; psf = nextpsf) {
2371 		nextpsf = psf->sf_next;
2372 		kfree(psf);
2373 	}
2374 	pmc->mca_sources = NULL;
2375 	pmc->mca_sfmode = MCAST_EXCLUDE;
2376 	pmc->mca_sfcount[MCAST_INCLUDE] = 0;
2377 	pmc->mca_sfcount[MCAST_EXCLUDE] = 1;
2378 }
2379 
2380 
2381 static void igmp6_join_group(struct ifmcaddr6 *ma)
2382 {
2383 	unsigned long delay;
2384 
2385 	if (ma->mca_flags & MAF_NOREPORT)
2386 		return;
2387 
2388 	igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2389 
2390 	delay = prandom_u32() % unsolicited_report_interval(ma->idev);
2391 
2392 	spin_lock_bh(&ma->mca_lock);
2393 	if (del_timer(&ma->mca_timer)) {
2394 		refcount_dec(&ma->mca_refcnt);
2395 		delay = ma->mca_timer.expires - jiffies;
2396 	}
2397 
2398 	if (!mod_timer(&ma->mca_timer, jiffies + delay))
2399 		refcount_inc(&ma->mca_refcnt);
2400 	ma->mca_flags |= MAF_TIMER_RUNNING | MAF_LAST_REPORTER;
2401 	spin_unlock_bh(&ma->mca_lock);
2402 }
2403 
2404 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
2405 			    struct inet6_dev *idev)
2406 {
2407 	int err;
2408 
2409 	/* callers have the socket lock and rtnl lock
2410 	 * so no other readers or writers of iml or its sflist
2411 	 */
2412 	if (!iml->sflist) {
2413 		/* any-source empty exclude case */
2414 		return ip6_mc_del_src(idev, &iml->addr, iml->sfmode, 0, NULL, 0);
2415 	}
2416 	err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode,
2417 		iml->sflist->sl_count, iml->sflist->sl_addr, 0);
2418 	sock_kfree_s(sk, iml->sflist, IP6_SFLSIZE(iml->sflist->sl_max));
2419 	iml->sflist = NULL;
2420 	return err;
2421 }
2422 
2423 static void igmp6_leave_group(struct ifmcaddr6 *ma)
2424 {
2425 	if (mld_in_v1_mode(ma->idev)) {
2426 		if (ma->mca_flags & MAF_LAST_REPORTER)
2427 			igmp6_send(&ma->mca_addr, ma->idev->dev,
2428 				ICMPV6_MGM_REDUCTION);
2429 	} else {
2430 		mld_add_delrec(ma->idev, ma);
2431 		mld_ifc_event(ma->idev);
2432 	}
2433 }
2434 
2435 static void mld_gq_timer_expire(unsigned long data)
2436 {
2437 	struct inet6_dev *idev = (struct inet6_dev *)data;
2438 
2439 	idev->mc_gq_running = 0;
2440 	mld_send_report(idev, NULL);
2441 	in6_dev_put(idev);
2442 }
2443 
2444 static void mld_ifc_timer_expire(unsigned long data)
2445 {
2446 	struct inet6_dev *idev = (struct inet6_dev *)data;
2447 
2448 	mld_send_cr(idev);
2449 	if (idev->mc_ifc_count) {
2450 		idev->mc_ifc_count--;
2451 		if (idev->mc_ifc_count)
2452 			mld_ifc_start_timer(idev, idev->mc_maxdelay);
2453 	}
2454 	in6_dev_put(idev);
2455 }
2456 
2457 static void mld_ifc_event(struct inet6_dev *idev)
2458 {
2459 	if (mld_in_v1_mode(idev))
2460 		return;
2461 	idev->mc_ifc_count = idev->mc_qrv;
2462 	mld_ifc_start_timer(idev, 1);
2463 }
2464 
2465 static void igmp6_timer_handler(unsigned long data)
2466 {
2467 	struct ifmcaddr6 *ma = (struct ifmcaddr6 *) data;
2468 
2469 	if (mld_in_v1_mode(ma->idev))
2470 		igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2471 	else
2472 		mld_send_report(ma->idev, ma);
2473 
2474 	spin_lock(&ma->mca_lock);
2475 	ma->mca_flags |=  MAF_LAST_REPORTER;
2476 	ma->mca_flags &= ~MAF_TIMER_RUNNING;
2477 	spin_unlock(&ma->mca_lock);
2478 	ma_put(ma);
2479 }
2480 
2481 /* Device changing type */
2482 
2483 void ipv6_mc_unmap(struct inet6_dev *idev)
2484 {
2485 	struct ifmcaddr6 *i;
2486 
2487 	/* Install multicast list, except for all-nodes (already installed) */
2488 
2489 	read_lock_bh(&idev->lock);
2490 	for (i = idev->mc_list; i; i = i->next)
2491 		igmp6_group_dropped(i);
2492 	read_unlock_bh(&idev->lock);
2493 }
2494 
2495 void ipv6_mc_remap(struct inet6_dev *idev)
2496 {
2497 	ipv6_mc_up(idev);
2498 }
2499 
2500 /* Device going down */
2501 
2502 void ipv6_mc_down(struct inet6_dev *idev)
2503 {
2504 	struct ifmcaddr6 *i;
2505 
2506 	/* Withdraw multicast list */
2507 
2508 	read_lock_bh(&idev->lock);
2509 
2510 	for (i = idev->mc_list; i; i = i->next)
2511 		igmp6_group_dropped(i);
2512 
2513 	/* Should stop timer after group drop. or we will
2514 	 * start timer again in mld_ifc_event()
2515 	 */
2516 	mld_ifc_stop_timer(idev);
2517 	mld_gq_stop_timer(idev);
2518 	mld_dad_stop_timer(idev);
2519 	read_unlock_bh(&idev->lock);
2520 }
2521 
2522 static void ipv6_mc_reset(struct inet6_dev *idev)
2523 {
2524 	idev->mc_qrv = sysctl_mld_qrv;
2525 	idev->mc_qi = MLD_QI_DEFAULT;
2526 	idev->mc_qri = MLD_QRI_DEFAULT;
2527 	idev->mc_v1_seen = 0;
2528 	idev->mc_maxdelay = unsolicited_report_interval(idev);
2529 }
2530 
2531 /* Device going up */
2532 
2533 void ipv6_mc_up(struct inet6_dev *idev)
2534 {
2535 	struct ifmcaddr6 *i;
2536 
2537 	/* Install multicast list, except for all-nodes (already installed) */
2538 
2539 	read_lock_bh(&idev->lock);
2540 	ipv6_mc_reset(idev);
2541 	for (i = idev->mc_list; i; i = i->next) {
2542 		mld_del_delrec(idev, i);
2543 		igmp6_group_added(i);
2544 	}
2545 	read_unlock_bh(&idev->lock);
2546 }
2547 
2548 /* IPv6 device initialization. */
2549 
2550 void ipv6_mc_init_dev(struct inet6_dev *idev)
2551 {
2552 	write_lock_bh(&idev->lock);
2553 	spin_lock_init(&idev->mc_lock);
2554 	idev->mc_gq_running = 0;
2555 	setup_timer(&idev->mc_gq_timer, mld_gq_timer_expire,
2556 			(unsigned long)idev);
2557 	idev->mc_tomb = NULL;
2558 	idev->mc_ifc_count = 0;
2559 	setup_timer(&idev->mc_ifc_timer, mld_ifc_timer_expire,
2560 			(unsigned long)idev);
2561 	setup_timer(&idev->mc_dad_timer, mld_dad_timer_expire,
2562 		    (unsigned long)idev);
2563 	ipv6_mc_reset(idev);
2564 	write_unlock_bh(&idev->lock);
2565 }
2566 
2567 /*
2568  *	Device is about to be destroyed: clean up.
2569  */
2570 
2571 void ipv6_mc_destroy_dev(struct inet6_dev *idev)
2572 {
2573 	struct ifmcaddr6 *i;
2574 
2575 	/* Deactivate timers */
2576 	ipv6_mc_down(idev);
2577 	mld_clear_delrec(idev);
2578 
2579 	/* Delete all-nodes address. */
2580 	/* We cannot call ipv6_dev_mc_dec() directly, our caller in
2581 	 * addrconf.c has NULL'd out dev->ip6_ptr so in6_dev_get() will
2582 	 * fail.
2583 	 */
2584 	__ipv6_dev_mc_dec(idev, &in6addr_linklocal_allnodes);
2585 
2586 	if (idev->cnf.forwarding)
2587 		__ipv6_dev_mc_dec(idev, &in6addr_linklocal_allrouters);
2588 
2589 	write_lock_bh(&idev->lock);
2590 	while ((i = idev->mc_list) != NULL) {
2591 		idev->mc_list = i->next;
2592 
2593 		write_unlock_bh(&idev->lock);
2594 		ma_put(i);
2595 		write_lock_bh(&idev->lock);
2596 	}
2597 	write_unlock_bh(&idev->lock);
2598 }
2599 
2600 static void ipv6_mc_rejoin_groups(struct inet6_dev *idev)
2601 {
2602 	struct ifmcaddr6 *pmc;
2603 
2604 	ASSERT_RTNL();
2605 
2606 	if (mld_in_v1_mode(idev)) {
2607 		read_lock_bh(&idev->lock);
2608 		for (pmc = idev->mc_list; pmc; pmc = pmc->next)
2609 			igmp6_join_group(pmc);
2610 		read_unlock_bh(&idev->lock);
2611 	} else
2612 		mld_send_report(idev, NULL);
2613 }
2614 
2615 static int ipv6_mc_netdev_event(struct notifier_block *this,
2616 				unsigned long event,
2617 				void *ptr)
2618 {
2619 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2620 	struct inet6_dev *idev = __in6_dev_get(dev);
2621 
2622 	switch (event) {
2623 	case NETDEV_RESEND_IGMP:
2624 		if (idev)
2625 			ipv6_mc_rejoin_groups(idev);
2626 		break;
2627 	default:
2628 		break;
2629 	}
2630 
2631 	return NOTIFY_DONE;
2632 }
2633 
2634 static struct notifier_block igmp6_netdev_notifier = {
2635 	.notifier_call = ipv6_mc_netdev_event,
2636 };
2637 
2638 #ifdef CONFIG_PROC_FS
2639 struct igmp6_mc_iter_state {
2640 	struct seq_net_private p;
2641 	struct net_device *dev;
2642 	struct inet6_dev *idev;
2643 };
2644 
2645 #define igmp6_mc_seq_private(seq)	((struct igmp6_mc_iter_state *)(seq)->private)
2646 
2647 static inline struct ifmcaddr6 *igmp6_mc_get_first(struct seq_file *seq)
2648 {
2649 	struct ifmcaddr6 *im = NULL;
2650 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2651 	struct net *net = seq_file_net(seq);
2652 
2653 	state->idev = NULL;
2654 	for_each_netdev_rcu(net, state->dev) {
2655 		struct inet6_dev *idev;
2656 		idev = __in6_dev_get(state->dev);
2657 		if (!idev)
2658 			continue;
2659 		read_lock_bh(&idev->lock);
2660 		im = idev->mc_list;
2661 		if (im) {
2662 			state->idev = idev;
2663 			break;
2664 		}
2665 		read_unlock_bh(&idev->lock);
2666 	}
2667 	return im;
2668 }
2669 
2670 static struct ifmcaddr6 *igmp6_mc_get_next(struct seq_file *seq, struct ifmcaddr6 *im)
2671 {
2672 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2673 
2674 	im = im->next;
2675 	while (!im) {
2676 		if (likely(state->idev))
2677 			read_unlock_bh(&state->idev->lock);
2678 
2679 		state->dev = next_net_device_rcu(state->dev);
2680 		if (!state->dev) {
2681 			state->idev = NULL;
2682 			break;
2683 		}
2684 		state->idev = __in6_dev_get(state->dev);
2685 		if (!state->idev)
2686 			continue;
2687 		read_lock_bh(&state->idev->lock);
2688 		im = state->idev->mc_list;
2689 	}
2690 	return im;
2691 }
2692 
2693 static struct ifmcaddr6 *igmp6_mc_get_idx(struct seq_file *seq, loff_t pos)
2694 {
2695 	struct ifmcaddr6 *im = igmp6_mc_get_first(seq);
2696 	if (im)
2697 		while (pos && (im = igmp6_mc_get_next(seq, im)) != NULL)
2698 			--pos;
2699 	return pos ? NULL : im;
2700 }
2701 
2702 static void *igmp6_mc_seq_start(struct seq_file *seq, loff_t *pos)
2703 	__acquires(RCU)
2704 {
2705 	rcu_read_lock();
2706 	return igmp6_mc_get_idx(seq, *pos);
2707 }
2708 
2709 static void *igmp6_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2710 {
2711 	struct ifmcaddr6 *im = igmp6_mc_get_next(seq, v);
2712 
2713 	++*pos;
2714 	return im;
2715 }
2716 
2717 static void igmp6_mc_seq_stop(struct seq_file *seq, void *v)
2718 	__releases(RCU)
2719 {
2720 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2721 
2722 	if (likely(state->idev)) {
2723 		read_unlock_bh(&state->idev->lock);
2724 		state->idev = NULL;
2725 	}
2726 	state->dev = NULL;
2727 	rcu_read_unlock();
2728 }
2729 
2730 static int igmp6_mc_seq_show(struct seq_file *seq, void *v)
2731 {
2732 	struct ifmcaddr6 *im = (struct ifmcaddr6 *)v;
2733 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2734 
2735 	seq_printf(seq,
2736 		   "%-4d %-15s %pi6 %5d %08X %ld\n",
2737 		   state->dev->ifindex, state->dev->name,
2738 		   &im->mca_addr,
2739 		   im->mca_users, im->mca_flags,
2740 		   (im->mca_flags&MAF_TIMER_RUNNING) ?
2741 		   jiffies_to_clock_t(im->mca_timer.expires-jiffies) : 0);
2742 	return 0;
2743 }
2744 
2745 static const struct seq_operations igmp6_mc_seq_ops = {
2746 	.start	=	igmp6_mc_seq_start,
2747 	.next	=	igmp6_mc_seq_next,
2748 	.stop	=	igmp6_mc_seq_stop,
2749 	.show	=	igmp6_mc_seq_show,
2750 };
2751 
2752 static int igmp6_mc_seq_open(struct inode *inode, struct file *file)
2753 {
2754 	return seq_open_net(inode, file, &igmp6_mc_seq_ops,
2755 			    sizeof(struct igmp6_mc_iter_state));
2756 }
2757 
2758 static const struct file_operations igmp6_mc_seq_fops = {
2759 	.owner		=	THIS_MODULE,
2760 	.open		=	igmp6_mc_seq_open,
2761 	.read		=	seq_read,
2762 	.llseek		=	seq_lseek,
2763 	.release	=	seq_release_net,
2764 };
2765 
2766 struct igmp6_mcf_iter_state {
2767 	struct seq_net_private p;
2768 	struct net_device *dev;
2769 	struct inet6_dev *idev;
2770 	struct ifmcaddr6 *im;
2771 };
2772 
2773 #define igmp6_mcf_seq_private(seq)	((struct igmp6_mcf_iter_state *)(seq)->private)
2774 
2775 static inline struct ip6_sf_list *igmp6_mcf_get_first(struct seq_file *seq)
2776 {
2777 	struct ip6_sf_list *psf = NULL;
2778 	struct ifmcaddr6 *im = NULL;
2779 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2780 	struct net *net = seq_file_net(seq);
2781 
2782 	state->idev = NULL;
2783 	state->im = NULL;
2784 	for_each_netdev_rcu(net, state->dev) {
2785 		struct inet6_dev *idev;
2786 		idev = __in6_dev_get(state->dev);
2787 		if (unlikely(idev == NULL))
2788 			continue;
2789 		read_lock_bh(&idev->lock);
2790 		im = idev->mc_list;
2791 		if (likely(im)) {
2792 			spin_lock_bh(&im->mca_lock);
2793 			psf = im->mca_sources;
2794 			if (likely(psf)) {
2795 				state->im = im;
2796 				state->idev = idev;
2797 				break;
2798 			}
2799 			spin_unlock_bh(&im->mca_lock);
2800 		}
2801 		read_unlock_bh(&idev->lock);
2802 	}
2803 	return psf;
2804 }
2805 
2806 static struct ip6_sf_list *igmp6_mcf_get_next(struct seq_file *seq, struct ip6_sf_list *psf)
2807 {
2808 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2809 
2810 	psf = psf->sf_next;
2811 	while (!psf) {
2812 		spin_unlock_bh(&state->im->mca_lock);
2813 		state->im = state->im->next;
2814 		while (!state->im) {
2815 			if (likely(state->idev))
2816 				read_unlock_bh(&state->idev->lock);
2817 
2818 			state->dev = next_net_device_rcu(state->dev);
2819 			if (!state->dev) {
2820 				state->idev = NULL;
2821 				goto out;
2822 			}
2823 			state->idev = __in6_dev_get(state->dev);
2824 			if (!state->idev)
2825 				continue;
2826 			read_lock_bh(&state->idev->lock);
2827 			state->im = state->idev->mc_list;
2828 		}
2829 		if (!state->im)
2830 			break;
2831 		spin_lock_bh(&state->im->mca_lock);
2832 		psf = state->im->mca_sources;
2833 	}
2834 out:
2835 	return psf;
2836 }
2837 
2838 static struct ip6_sf_list *igmp6_mcf_get_idx(struct seq_file *seq, loff_t pos)
2839 {
2840 	struct ip6_sf_list *psf = igmp6_mcf_get_first(seq);
2841 	if (psf)
2842 		while (pos && (psf = igmp6_mcf_get_next(seq, psf)) != NULL)
2843 			--pos;
2844 	return pos ? NULL : psf;
2845 }
2846 
2847 static void *igmp6_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2848 	__acquires(RCU)
2849 {
2850 	rcu_read_lock();
2851 	return *pos ? igmp6_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2852 }
2853 
2854 static void *igmp6_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2855 {
2856 	struct ip6_sf_list *psf;
2857 	if (v == SEQ_START_TOKEN)
2858 		psf = igmp6_mcf_get_first(seq);
2859 	else
2860 		psf = igmp6_mcf_get_next(seq, v);
2861 	++*pos;
2862 	return psf;
2863 }
2864 
2865 static void igmp6_mcf_seq_stop(struct seq_file *seq, void *v)
2866 	__releases(RCU)
2867 {
2868 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2869 	if (likely(state->im)) {
2870 		spin_unlock_bh(&state->im->mca_lock);
2871 		state->im = NULL;
2872 	}
2873 	if (likely(state->idev)) {
2874 		read_unlock_bh(&state->idev->lock);
2875 		state->idev = NULL;
2876 	}
2877 	state->dev = NULL;
2878 	rcu_read_unlock();
2879 }
2880 
2881 static int igmp6_mcf_seq_show(struct seq_file *seq, void *v)
2882 {
2883 	struct ip6_sf_list *psf = (struct ip6_sf_list *)v;
2884 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2885 
2886 	if (v == SEQ_START_TOKEN) {
2887 		seq_puts(seq, "Idx Device                Multicast Address                   Source Address    INC    EXC\n");
2888 	} else {
2889 		seq_printf(seq,
2890 			   "%3d %6.6s %pi6 %pi6 %6lu %6lu\n",
2891 			   state->dev->ifindex, state->dev->name,
2892 			   &state->im->mca_addr,
2893 			   &psf->sf_addr,
2894 			   psf->sf_count[MCAST_INCLUDE],
2895 			   psf->sf_count[MCAST_EXCLUDE]);
2896 	}
2897 	return 0;
2898 }
2899 
2900 static const struct seq_operations igmp6_mcf_seq_ops = {
2901 	.start	=	igmp6_mcf_seq_start,
2902 	.next	=	igmp6_mcf_seq_next,
2903 	.stop	=	igmp6_mcf_seq_stop,
2904 	.show	=	igmp6_mcf_seq_show,
2905 };
2906 
2907 static int igmp6_mcf_seq_open(struct inode *inode, struct file *file)
2908 {
2909 	return seq_open_net(inode, file, &igmp6_mcf_seq_ops,
2910 			    sizeof(struct igmp6_mcf_iter_state));
2911 }
2912 
2913 static const struct file_operations igmp6_mcf_seq_fops = {
2914 	.owner		=	THIS_MODULE,
2915 	.open		=	igmp6_mcf_seq_open,
2916 	.read		=	seq_read,
2917 	.llseek		=	seq_lseek,
2918 	.release	=	seq_release_net,
2919 };
2920 
2921 static int __net_init igmp6_proc_init(struct net *net)
2922 {
2923 	int err;
2924 
2925 	err = -ENOMEM;
2926 	if (!proc_create("igmp6", S_IRUGO, net->proc_net, &igmp6_mc_seq_fops))
2927 		goto out;
2928 	if (!proc_create("mcfilter6", S_IRUGO, net->proc_net,
2929 			 &igmp6_mcf_seq_fops))
2930 		goto out_proc_net_igmp6;
2931 
2932 	err = 0;
2933 out:
2934 	return err;
2935 
2936 out_proc_net_igmp6:
2937 	remove_proc_entry("igmp6", net->proc_net);
2938 	goto out;
2939 }
2940 
2941 static void __net_exit igmp6_proc_exit(struct net *net)
2942 {
2943 	remove_proc_entry("mcfilter6", net->proc_net);
2944 	remove_proc_entry("igmp6", net->proc_net);
2945 }
2946 #else
2947 static inline int igmp6_proc_init(struct net *net)
2948 {
2949 	return 0;
2950 }
2951 static inline void igmp6_proc_exit(struct net *net)
2952 {
2953 }
2954 #endif
2955 
2956 static int __net_init igmp6_net_init(struct net *net)
2957 {
2958 	int err;
2959 
2960 	err = inet_ctl_sock_create(&net->ipv6.igmp_sk, PF_INET6,
2961 				   SOCK_RAW, IPPROTO_ICMPV6, net);
2962 	if (err < 0) {
2963 		pr_err("Failed to initialize the IGMP6 control socket (err %d)\n",
2964 		       err);
2965 		goto out;
2966 	}
2967 
2968 	inet6_sk(net->ipv6.igmp_sk)->hop_limit = 1;
2969 
2970 	err = inet_ctl_sock_create(&net->ipv6.mc_autojoin_sk, PF_INET6,
2971 				   SOCK_RAW, IPPROTO_ICMPV6, net);
2972 	if (err < 0) {
2973 		pr_err("Failed to initialize the IGMP6 autojoin socket (err %d)\n",
2974 		       err);
2975 		goto out_sock_create;
2976 	}
2977 
2978 	err = igmp6_proc_init(net);
2979 	if (err)
2980 		goto out_sock_create_autojoin;
2981 
2982 	return 0;
2983 
2984 out_sock_create_autojoin:
2985 	inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
2986 out_sock_create:
2987 	inet_ctl_sock_destroy(net->ipv6.igmp_sk);
2988 out:
2989 	return err;
2990 }
2991 
2992 static void __net_exit igmp6_net_exit(struct net *net)
2993 {
2994 	inet_ctl_sock_destroy(net->ipv6.igmp_sk);
2995 	inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
2996 	igmp6_proc_exit(net);
2997 }
2998 
2999 static struct pernet_operations igmp6_net_ops = {
3000 	.init = igmp6_net_init,
3001 	.exit = igmp6_net_exit,
3002 };
3003 
3004 int __init igmp6_init(void)
3005 {
3006 	return register_pernet_subsys(&igmp6_net_ops);
3007 }
3008 
3009 int __init igmp6_late_init(void)
3010 {
3011 	return register_netdevice_notifier(&igmp6_netdev_notifier);
3012 }
3013 
3014 void igmp6_cleanup(void)
3015 {
3016 	unregister_pernet_subsys(&igmp6_net_ops);
3017 }
3018 
3019 void igmp6_late_cleanup(void)
3020 {
3021 	unregister_netdevice_notifier(&igmp6_netdev_notifier);
3022 }
3023