xref: /openbmc/linux/net/ipv6/mcast.c (revision ae213c44)
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 int __mld2_query_bugs[] __attribute__((__unused__)) = {
69 	BUILD_BUG_ON_ZERO(offsetof(struct mld2_query, mld2q_srcs) % 4),
70 	BUILD_BUG_ON_ZERO(offsetof(struct mld2_report, mld2r_grec) % 4),
71 	BUILD_BUG_ON_ZERO(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(struct timer_list *t);
79 
80 static void mld_gq_timer_expire(struct timer_list *t);
81 static void mld_ifc_timer_expire(struct timer_list *t);
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 static int __ipv6_dev_mc_inc(struct net_device *dev,
99 			     const struct in6_addr *addr, unsigned int mode);
100 
101 #define MLD_QRV_DEFAULT		2
102 /* RFC3810, 9.2. Query Interval */
103 #define MLD_QI_DEFAULT		(125 * HZ)
104 /* RFC3810, 9.3. Query Response Interval */
105 #define MLD_QRI_DEFAULT		(10 * HZ)
106 
107 /* RFC3810, 8.1 Query Version Distinctions */
108 #define MLD_V1_QUERY_LEN	24
109 #define MLD_V2_QUERY_LEN_MIN	28
110 
111 #define IPV6_MLD_MAX_MSF	64
112 
113 int sysctl_mld_max_msf __read_mostly = IPV6_MLD_MAX_MSF;
114 int sysctl_mld_qrv __read_mostly = MLD_QRV_DEFAULT;
115 
116 /*
117  *	socket join on multicast group
118  */
119 
120 #define for_each_pmc_rcu(np, pmc)				\
121 	for (pmc = rcu_dereference(np->ipv6_mc_list);		\
122 	     pmc != NULL;					\
123 	     pmc = rcu_dereference(pmc->next))
124 
125 static int unsolicited_report_interval(struct inet6_dev *idev)
126 {
127 	int iv;
128 
129 	if (mld_in_v1_mode(idev))
130 		iv = idev->cnf.mldv1_unsolicited_report_interval;
131 	else
132 		iv = idev->cnf.mldv2_unsolicited_report_interval;
133 
134 	return iv > 0 ? iv : 1;
135 }
136 
137 static int __ipv6_sock_mc_join(struct sock *sk, int ifindex,
138 			       const struct in6_addr *addr, unsigned int mode)
139 {
140 	struct net_device *dev = NULL;
141 	struct ipv6_mc_socklist *mc_lst;
142 	struct ipv6_pinfo *np = inet6_sk(sk);
143 	struct net *net = sock_net(sk);
144 	int err;
145 
146 	ASSERT_RTNL();
147 
148 	if (!ipv6_addr_is_multicast(addr))
149 		return -EINVAL;
150 
151 	rcu_read_lock();
152 	for_each_pmc_rcu(np, mc_lst) {
153 		if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
154 		    ipv6_addr_equal(&mc_lst->addr, addr)) {
155 			rcu_read_unlock();
156 			return -EADDRINUSE;
157 		}
158 	}
159 	rcu_read_unlock();
160 
161 	mc_lst = sock_kmalloc(sk, sizeof(struct ipv6_mc_socklist), GFP_KERNEL);
162 
163 	if (!mc_lst)
164 		return -ENOMEM;
165 
166 	mc_lst->next = NULL;
167 	mc_lst->addr = *addr;
168 
169 	if (ifindex == 0) {
170 		struct rt6_info *rt;
171 		rt = rt6_lookup(net, addr, NULL, 0, NULL, 0);
172 		if (rt) {
173 			dev = rt->dst.dev;
174 			ip6_rt_put(rt);
175 		}
176 	} else
177 		dev = __dev_get_by_index(net, ifindex);
178 
179 	if (!dev) {
180 		sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
181 		return -ENODEV;
182 	}
183 
184 	mc_lst->ifindex = dev->ifindex;
185 	mc_lst->sfmode = mode;
186 	rwlock_init(&mc_lst->sflock);
187 	mc_lst->sflist = NULL;
188 
189 	/*
190 	 *	now add/increase the group membership on the device
191 	 */
192 
193 	err = __ipv6_dev_mc_inc(dev, addr, mode);
194 
195 	if (err) {
196 		sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
197 		return err;
198 	}
199 
200 	mc_lst->next = np->ipv6_mc_list;
201 	rcu_assign_pointer(np->ipv6_mc_list, mc_lst);
202 
203 	return 0;
204 }
205 
206 int ipv6_sock_mc_join(struct sock *sk, int ifindex, const struct in6_addr *addr)
207 {
208 	return __ipv6_sock_mc_join(sk, ifindex, addr, MCAST_EXCLUDE);
209 }
210 EXPORT_SYMBOL(ipv6_sock_mc_join);
211 
212 int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
213 			  const struct in6_addr *addr, unsigned int mode)
214 {
215 	return __ipv6_sock_mc_join(sk, ifindex, addr, mode);
216 }
217 
218 /*
219  *	socket leave on multicast group
220  */
221 int ipv6_sock_mc_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
222 {
223 	struct ipv6_pinfo *np = inet6_sk(sk);
224 	struct ipv6_mc_socklist *mc_lst;
225 	struct ipv6_mc_socklist __rcu **lnk;
226 	struct net *net = sock_net(sk);
227 
228 	ASSERT_RTNL();
229 
230 	if (!ipv6_addr_is_multicast(addr))
231 		return -EINVAL;
232 
233 	for (lnk = &np->ipv6_mc_list;
234 	     (mc_lst = rtnl_dereference(*lnk)) != NULL;
235 	      lnk = &mc_lst->next) {
236 		if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
237 		    ipv6_addr_equal(&mc_lst->addr, addr)) {
238 			struct net_device *dev;
239 
240 			*lnk = mc_lst->next;
241 
242 			dev = __dev_get_by_index(net, mc_lst->ifindex);
243 			if (dev) {
244 				struct inet6_dev *idev = __in6_dev_get(dev);
245 
246 				(void) ip6_mc_leave_src(sk, mc_lst, idev);
247 				if (idev)
248 					__ipv6_dev_mc_dec(idev, &mc_lst->addr);
249 			} else
250 				(void) ip6_mc_leave_src(sk, mc_lst, NULL);
251 
252 			atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
253 			kfree_rcu(mc_lst, rcu);
254 			return 0;
255 		}
256 	}
257 
258 	return -EADDRNOTAVAIL;
259 }
260 EXPORT_SYMBOL(ipv6_sock_mc_drop);
261 
262 /* called with rcu_read_lock() */
263 static struct inet6_dev *ip6_mc_find_dev_rcu(struct net *net,
264 					     const struct in6_addr *group,
265 					     int ifindex)
266 {
267 	struct net_device *dev = NULL;
268 	struct inet6_dev *idev = NULL;
269 
270 	if (ifindex == 0) {
271 		struct rt6_info *rt = rt6_lookup(net, group, NULL, 0, NULL, 0);
272 
273 		if (rt) {
274 			dev = rt->dst.dev;
275 			ip6_rt_put(rt);
276 		}
277 	} else
278 		dev = dev_get_by_index_rcu(net, ifindex);
279 
280 	if (!dev)
281 		return NULL;
282 	idev = __in6_dev_get(dev);
283 	if (!idev)
284 		return NULL;
285 	read_lock_bh(&idev->lock);
286 	if (idev->dead) {
287 		read_unlock_bh(&idev->lock);
288 		return NULL;
289 	}
290 	return idev;
291 }
292 
293 void __ipv6_sock_mc_close(struct sock *sk)
294 {
295 	struct ipv6_pinfo *np = inet6_sk(sk);
296 	struct ipv6_mc_socklist *mc_lst;
297 	struct net *net = sock_net(sk);
298 
299 	ASSERT_RTNL();
300 
301 	while ((mc_lst = rtnl_dereference(np->ipv6_mc_list)) != NULL) {
302 		struct net_device *dev;
303 
304 		np->ipv6_mc_list = mc_lst->next;
305 
306 		dev = __dev_get_by_index(net, mc_lst->ifindex);
307 		if (dev) {
308 			struct inet6_dev *idev = __in6_dev_get(dev);
309 
310 			(void) ip6_mc_leave_src(sk, mc_lst, idev);
311 			if (idev)
312 				__ipv6_dev_mc_dec(idev, &mc_lst->addr);
313 		} else
314 			(void) ip6_mc_leave_src(sk, mc_lst, NULL);
315 
316 		atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
317 		kfree_rcu(mc_lst, rcu);
318 	}
319 }
320 
321 void ipv6_sock_mc_close(struct sock *sk)
322 {
323 	struct ipv6_pinfo *np = inet6_sk(sk);
324 
325 	if (!rcu_access_pointer(np->ipv6_mc_list))
326 		return;
327 	rtnl_lock();
328 	__ipv6_sock_mc_close(sk);
329 	rtnl_unlock();
330 }
331 
332 int ip6_mc_source(int add, int omode, struct sock *sk,
333 	struct group_source_req *pgsr)
334 {
335 	struct in6_addr *source, *group;
336 	struct ipv6_mc_socklist *pmc;
337 	struct inet6_dev *idev;
338 	struct ipv6_pinfo *inet6 = inet6_sk(sk);
339 	struct ip6_sf_socklist *psl;
340 	struct net *net = sock_net(sk);
341 	int i, j, rv;
342 	int leavegroup = 0;
343 	int pmclocked = 0;
344 	int err;
345 
346 	source = &((struct sockaddr_in6 *)&pgsr->gsr_source)->sin6_addr;
347 	group = &((struct sockaddr_in6 *)&pgsr->gsr_group)->sin6_addr;
348 
349 	if (!ipv6_addr_is_multicast(group))
350 		return -EINVAL;
351 
352 	rcu_read_lock();
353 	idev = ip6_mc_find_dev_rcu(net, group, pgsr->gsr_interface);
354 	if (!idev) {
355 		rcu_read_unlock();
356 		return -ENODEV;
357 	}
358 
359 	err = -EADDRNOTAVAIL;
360 
361 	for_each_pmc_rcu(inet6, pmc) {
362 		if (pgsr->gsr_interface && pmc->ifindex != pgsr->gsr_interface)
363 			continue;
364 		if (ipv6_addr_equal(&pmc->addr, group))
365 			break;
366 	}
367 	if (!pmc) {		/* must have a prior join */
368 		err = -EINVAL;
369 		goto done;
370 	}
371 	/* if a source filter was set, must be the same mode as before */
372 	if (pmc->sflist) {
373 		if (pmc->sfmode != omode) {
374 			err = -EINVAL;
375 			goto done;
376 		}
377 	} else if (pmc->sfmode != omode) {
378 		/* allow mode switches for empty-set filters */
379 		ip6_mc_add_src(idev, group, omode, 0, NULL, 0);
380 		ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
381 		pmc->sfmode = omode;
382 	}
383 
384 	write_lock(&pmc->sflock);
385 	pmclocked = 1;
386 
387 	psl = pmc->sflist;
388 	if (!add) {
389 		if (!psl)
390 			goto done;	/* err = -EADDRNOTAVAIL */
391 		rv = !0;
392 		for (i = 0; i < psl->sl_count; i++) {
393 			rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
394 			if (rv == 0)
395 				break;
396 		}
397 		if (rv)		/* source not found */
398 			goto done;	/* err = -EADDRNOTAVAIL */
399 
400 		/* special case - (INCLUDE, empty) == LEAVE_GROUP */
401 		if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
402 			leavegroup = 1;
403 			goto done;
404 		}
405 
406 		/* update the interface filter */
407 		ip6_mc_del_src(idev, group, omode, 1, source, 1);
408 
409 		for (j = i+1; j < psl->sl_count; j++)
410 			psl->sl_addr[j-1] = psl->sl_addr[j];
411 		psl->sl_count--;
412 		err = 0;
413 		goto done;
414 	}
415 	/* else, add a new source to the filter */
416 
417 	if (psl && psl->sl_count >= sysctl_mld_max_msf) {
418 		err = -ENOBUFS;
419 		goto done;
420 	}
421 	if (!psl || psl->sl_count == psl->sl_max) {
422 		struct ip6_sf_socklist *newpsl;
423 		int count = IP6_SFBLOCK;
424 
425 		if (psl)
426 			count += psl->sl_max;
427 		newpsl = sock_kmalloc(sk, IP6_SFLSIZE(count), GFP_ATOMIC);
428 		if (!newpsl) {
429 			err = -ENOBUFS;
430 			goto done;
431 		}
432 		newpsl->sl_max = count;
433 		newpsl->sl_count = count - IP6_SFBLOCK;
434 		if (psl) {
435 			for (i = 0; i < psl->sl_count; i++)
436 				newpsl->sl_addr[i] = psl->sl_addr[i];
437 			sock_kfree_s(sk, psl, IP6_SFLSIZE(psl->sl_max));
438 		}
439 		pmc->sflist = psl = newpsl;
440 	}
441 	rv = 1;	/* > 0 for insert logic below if sl_count is 0 */
442 	for (i = 0; i < psl->sl_count; i++) {
443 		rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
444 		if (rv == 0) /* There is an error in the address. */
445 			goto done;
446 	}
447 	for (j = psl->sl_count-1; j >= i; j--)
448 		psl->sl_addr[j+1] = psl->sl_addr[j];
449 	psl->sl_addr[i] = *source;
450 	psl->sl_count++;
451 	err = 0;
452 	/* update the interface list */
453 	ip6_mc_add_src(idev, group, omode, 1, source, 1);
454 done:
455 	if (pmclocked)
456 		write_unlock(&pmc->sflock);
457 	read_unlock_bh(&idev->lock);
458 	rcu_read_unlock();
459 	if (leavegroup)
460 		err = ipv6_sock_mc_drop(sk, pgsr->gsr_interface, group);
461 	return err;
462 }
463 
464 int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf)
465 {
466 	const struct in6_addr *group;
467 	struct ipv6_mc_socklist *pmc;
468 	struct inet6_dev *idev;
469 	struct ipv6_pinfo *inet6 = inet6_sk(sk);
470 	struct ip6_sf_socklist *newpsl, *psl;
471 	struct net *net = sock_net(sk);
472 	int leavegroup = 0;
473 	int i, err;
474 
475 	group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
476 
477 	if (!ipv6_addr_is_multicast(group))
478 		return -EINVAL;
479 	if (gsf->gf_fmode != MCAST_INCLUDE &&
480 	    gsf->gf_fmode != MCAST_EXCLUDE)
481 		return -EINVAL;
482 
483 	rcu_read_lock();
484 	idev = ip6_mc_find_dev_rcu(net, group, gsf->gf_interface);
485 
486 	if (!idev) {
487 		rcu_read_unlock();
488 		return -ENODEV;
489 	}
490 
491 	err = 0;
492 
493 	if (gsf->gf_fmode == MCAST_INCLUDE && gsf->gf_numsrc == 0) {
494 		leavegroup = 1;
495 		goto done;
496 	}
497 
498 	for_each_pmc_rcu(inet6, pmc) {
499 		if (pmc->ifindex != gsf->gf_interface)
500 			continue;
501 		if (ipv6_addr_equal(&pmc->addr, group))
502 			break;
503 	}
504 	if (!pmc) {		/* must have a prior join */
505 		err = -EINVAL;
506 		goto done;
507 	}
508 	if (gsf->gf_numsrc) {
509 		newpsl = sock_kmalloc(sk, IP6_SFLSIZE(gsf->gf_numsrc),
510 							  GFP_ATOMIC);
511 		if (!newpsl) {
512 			err = -ENOBUFS;
513 			goto done;
514 		}
515 		newpsl->sl_max = newpsl->sl_count = gsf->gf_numsrc;
516 		for (i = 0; i < newpsl->sl_count; ++i) {
517 			struct sockaddr_in6 *psin6;
518 
519 			psin6 = (struct sockaddr_in6 *)&gsf->gf_slist[i];
520 			newpsl->sl_addr[i] = psin6->sin6_addr;
521 		}
522 		err = ip6_mc_add_src(idev, group, gsf->gf_fmode,
523 			newpsl->sl_count, newpsl->sl_addr, 0);
524 		if (err) {
525 			sock_kfree_s(sk, newpsl, IP6_SFLSIZE(newpsl->sl_max));
526 			goto done;
527 		}
528 	} else {
529 		newpsl = NULL;
530 		(void) ip6_mc_add_src(idev, group, gsf->gf_fmode, 0, NULL, 0);
531 	}
532 
533 	write_lock(&pmc->sflock);
534 	psl = pmc->sflist;
535 	if (psl) {
536 		(void) ip6_mc_del_src(idev, group, pmc->sfmode,
537 			psl->sl_count, psl->sl_addr, 0);
538 		sock_kfree_s(sk, psl, IP6_SFLSIZE(psl->sl_max));
539 	} else
540 		(void) ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
541 	pmc->sflist = newpsl;
542 	pmc->sfmode = gsf->gf_fmode;
543 	write_unlock(&pmc->sflock);
544 	err = 0;
545 done:
546 	read_unlock_bh(&idev->lock);
547 	rcu_read_unlock();
548 	if (leavegroup)
549 		err = ipv6_sock_mc_drop(sk, gsf->gf_interface, group);
550 	return err;
551 }
552 
553 int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
554 	struct group_filter __user *optval, int __user *optlen)
555 {
556 	int err, i, count, copycount;
557 	const struct in6_addr *group;
558 	struct ipv6_mc_socklist *pmc;
559 	struct inet6_dev *idev;
560 	struct ipv6_pinfo *inet6 = inet6_sk(sk);
561 	struct ip6_sf_socklist *psl;
562 	struct net *net = sock_net(sk);
563 
564 	group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
565 
566 	if (!ipv6_addr_is_multicast(group))
567 		return -EINVAL;
568 
569 	rcu_read_lock();
570 	idev = ip6_mc_find_dev_rcu(net, group, gsf->gf_interface);
571 
572 	if (!idev) {
573 		rcu_read_unlock();
574 		return -ENODEV;
575 	}
576 
577 	err = -EADDRNOTAVAIL;
578 	/* changes to the ipv6_mc_list require the socket lock and
579 	 * rtnl lock. We have the socket lock and rcu read lock,
580 	 * so reading the list is safe.
581 	 */
582 
583 	for_each_pmc_rcu(inet6, pmc) {
584 		if (pmc->ifindex != gsf->gf_interface)
585 			continue;
586 		if (ipv6_addr_equal(group, &pmc->addr))
587 			break;
588 	}
589 	if (!pmc)		/* must have a prior join */
590 		goto done;
591 	gsf->gf_fmode = pmc->sfmode;
592 	psl = pmc->sflist;
593 	count = psl ? psl->sl_count : 0;
594 	read_unlock_bh(&idev->lock);
595 	rcu_read_unlock();
596 
597 	copycount = count < gsf->gf_numsrc ? count : gsf->gf_numsrc;
598 	gsf->gf_numsrc = count;
599 	if (put_user(GROUP_FILTER_SIZE(copycount), optlen) ||
600 	    copy_to_user(optval, gsf, GROUP_FILTER_SIZE(0))) {
601 		return -EFAULT;
602 	}
603 	/* changes to psl require the socket lock, and a write lock
604 	 * on pmc->sflock. We have the socket lock so reading here is safe.
605 	 */
606 	for (i = 0; i < copycount; i++) {
607 		struct sockaddr_in6 *psin6;
608 		struct sockaddr_storage ss;
609 
610 		psin6 = (struct sockaddr_in6 *)&ss;
611 		memset(&ss, 0, sizeof(ss));
612 		psin6->sin6_family = AF_INET6;
613 		psin6->sin6_addr = psl->sl_addr[i];
614 		if (copy_to_user(&optval->gf_slist[i], &ss, sizeof(ss)))
615 			return -EFAULT;
616 	}
617 	return 0;
618 done:
619 	read_unlock_bh(&idev->lock);
620 	rcu_read_unlock();
621 	return err;
622 }
623 
624 bool inet6_mc_check(struct sock *sk, const struct in6_addr *mc_addr,
625 		    const struct in6_addr *src_addr)
626 {
627 	struct ipv6_pinfo *np = inet6_sk(sk);
628 	struct ipv6_mc_socklist *mc;
629 	struct ip6_sf_socklist *psl;
630 	bool rv = true;
631 
632 	rcu_read_lock();
633 	for_each_pmc_rcu(np, mc) {
634 		if (ipv6_addr_equal(&mc->addr, mc_addr))
635 			break;
636 	}
637 	if (!mc) {
638 		rcu_read_unlock();
639 		return np->mc_all;
640 	}
641 	read_lock(&mc->sflock);
642 	psl = mc->sflist;
643 	if (!psl) {
644 		rv = mc->sfmode == MCAST_EXCLUDE;
645 	} else {
646 		int i;
647 
648 		for (i = 0; i < psl->sl_count; i++) {
649 			if (ipv6_addr_equal(&psl->sl_addr[i], src_addr))
650 				break;
651 		}
652 		if (mc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
653 			rv = false;
654 		if (mc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
655 			rv = false;
656 	}
657 	read_unlock(&mc->sflock);
658 	rcu_read_unlock();
659 
660 	return rv;
661 }
662 
663 static void igmp6_group_added(struct ifmcaddr6 *mc)
664 {
665 	struct net_device *dev = mc->idev->dev;
666 	char buf[MAX_ADDR_LEN];
667 
668 	if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
669 	    IPV6_ADDR_SCOPE_LINKLOCAL)
670 		return;
671 
672 	spin_lock_bh(&mc->mca_lock);
673 	if (!(mc->mca_flags&MAF_LOADED)) {
674 		mc->mca_flags |= MAF_LOADED;
675 		if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
676 			dev_mc_add(dev, buf);
677 	}
678 	spin_unlock_bh(&mc->mca_lock);
679 
680 	if (!(dev->flags & IFF_UP) || (mc->mca_flags & MAF_NOREPORT))
681 		return;
682 
683 	if (mld_in_v1_mode(mc->idev)) {
684 		igmp6_join_group(mc);
685 		return;
686 	}
687 	/* else v2 */
688 
689 	/* Based on RFC3810 6.1, for newly added INCLUDE SSM, we
690 	 * should not send filter-mode change record as the mode
691 	 * should be from IN() to IN(A).
692 	 */
693 	if (mc->mca_sfmode == MCAST_EXCLUDE)
694 		mc->mca_crcount = mc->idev->mc_qrv;
695 
696 	mld_ifc_event(mc->idev);
697 }
698 
699 static void igmp6_group_dropped(struct ifmcaddr6 *mc)
700 {
701 	struct net_device *dev = mc->idev->dev;
702 	char buf[MAX_ADDR_LEN];
703 
704 	if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
705 	    IPV6_ADDR_SCOPE_LINKLOCAL)
706 		return;
707 
708 	spin_lock_bh(&mc->mca_lock);
709 	if (mc->mca_flags&MAF_LOADED) {
710 		mc->mca_flags &= ~MAF_LOADED;
711 		if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
712 			dev_mc_del(dev, buf);
713 	}
714 
715 	spin_unlock_bh(&mc->mca_lock);
716 	if (mc->mca_flags & MAF_NOREPORT)
717 		return;
718 
719 	if (!mc->idev->dead)
720 		igmp6_leave_group(mc);
721 
722 	spin_lock_bh(&mc->mca_lock);
723 	if (del_timer(&mc->mca_timer))
724 		refcount_dec(&mc->mca_refcnt);
725 	spin_unlock_bh(&mc->mca_lock);
726 }
727 
728 /*
729  * deleted ifmcaddr6 manipulation
730  */
731 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
732 {
733 	struct ifmcaddr6 *pmc;
734 
735 	/* this is an "ifmcaddr6" for convenience; only the fields below
736 	 * are actually used. In particular, the refcnt and users are not
737 	 * used for management of the delete list. Using the same structure
738 	 * for deleted items allows change reports to use common code with
739 	 * non-deleted or query-response MCA's.
740 	 */
741 	pmc = kzalloc(sizeof(*pmc), GFP_ATOMIC);
742 	if (!pmc)
743 		return;
744 
745 	spin_lock_bh(&im->mca_lock);
746 	spin_lock_init(&pmc->mca_lock);
747 	pmc->idev = im->idev;
748 	in6_dev_hold(idev);
749 	pmc->mca_addr = im->mca_addr;
750 	pmc->mca_crcount = idev->mc_qrv;
751 	pmc->mca_sfmode = im->mca_sfmode;
752 	if (pmc->mca_sfmode == MCAST_INCLUDE) {
753 		struct ip6_sf_list *psf;
754 
755 		pmc->mca_tomb = im->mca_tomb;
756 		pmc->mca_sources = im->mca_sources;
757 		im->mca_tomb = im->mca_sources = NULL;
758 		for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
759 			psf->sf_crcount = pmc->mca_crcount;
760 	}
761 	spin_unlock_bh(&im->mca_lock);
762 
763 	spin_lock_bh(&idev->mc_lock);
764 	pmc->next = idev->mc_tomb;
765 	idev->mc_tomb = pmc;
766 	spin_unlock_bh(&idev->mc_lock);
767 }
768 
769 static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
770 {
771 	struct ifmcaddr6 *pmc, *pmc_prev;
772 	struct ip6_sf_list *psf;
773 	struct in6_addr *pmca = &im->mca_addr;
774 
775 	spin_lock_bh(&idev->mc_lock);
776 	pmc_prev = NULL;
777 	for (pmc = idev->mc_tomb; pmc; pmc = pmc->next) {
778 		if (ipv6_addr_equal(&pmc->mca_addr, pmca))
779 			break;
780 		pmc_prev = pmc;
781 	}
782 	if (pmc) {
783 		if (pmc_prev)
784 			pmc_prev->next = pmc->next;
785 		else
786 			idev->mc_tomb = pmc->next;
787 	}
788 	spin_unlock_bh(&idev->mc_lock);
789 
790 	spin_lock_bh(&im->mca_lock);
791 	if (pmc) {
792 		im->idev = pmc->idev;
793 		if (im->mca_sfmode == MCAST_INCLUDE) {
794 			im->mca_tomb = pmc->mca_tomb;
795 			im->mca_sources = pmc->mca_sources;
796 			for (psf = im->mca_sources; psf; psf = psf->sf_next)
797 				psf->sf_crcount = idev->mc_qrv;
798 		} else {
799 			im->mca_crcount = idev->mc_qrv;
800 		}
801 		in6_dev_put(pmc->idev);
802 		kfree(pmc);
803 	}
804 	spin_unlock_bh(&im->mca_lock);
805 }
806 
807 static void mld_clear_delrec(struct inet6_dev *idev)
808 {
809 	struct ifmcaddr6 *pmc, *nextpmc;
810 
811 	spin_lock_bh(&idev->mc_lock);
812 	pmc = idev->mc_tomb;
813 	idev->mc_tomb = NULL;
814 	spin_unlock_bh(&idev->mc_lock);
815 
816 	for (; pmc; pmc = nextpmc) {
817 		nextpmc = pmc->next;
818 		ip6_mc_clear_src(pmc);
819 		in6_dev_put(pmc->idev);
820 		kfree(pmc);
821 	}
822 
823 	/* clear dead sources, too */
824 	read_lock_bh(&idev->lock);
825 	for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
826 		struct ip6_sf_list *psf, *psf_next;
827 
828 		spin_lock_bh(&pmc->mca_lock);
829 		psf = pmc->mca_tomb;
830 		pmc->mca_tomb = NULL;
831 		spin_unlock_bh(&pmc->mca_lock);
832 		for (; psf; psf = psf_next) {
833 			psf_next = psf->sf_next;
834 			kfree(psf);
835 		}
836 	}
837 	read_unlock_bh(&idev->lock);
838 }
839 
840 static void mca_get(struct ifmcaddr6 *mc)
841 {
842 	refcount_inc(&mc->mca_refcnt);
843 }
844 
845 static void ma_put(struct ifmcaddr6 *mc)
846 {
847 	if (refcount_dec_and_test(&mc->mca_refcnt)) {
848 		in6_dev_put(mc->idev);
849 		kfree(mc);
850 	}
851 }
852 
853 static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev,
854 				   const struct in6_addr *addr,
855 				   unsigned int mode)
856 {
857 	struct ifmcaddr6 *mc;
858 
859 	mc = kzalloc(sizeof(*mc), GFP_ATOMIC);
860 	if (!mc)
861 		return NULL;
862 
863 	timer_setup(&mc->mca_timer, igmp6_timer_handler, 0);
864 
865 	mc->mca_addr = *addr;
866 	mc->idev = idev; /* reference taken by caller */
867 	mc->mca_users = 1;
868 	/* mca_stamp should be updated upon changes */
869 	mc->mca_cstamp = mc->mca_tstamp = jiffies;
870 	refcount_set(&mc->mca_refcnt, 1);
871 	spin_lock_init(&mc->mca_lock);
872 
873 	mc->mca_sfmode = mode;
874 	mc->mca_sfcount[mode] = 1;
875 
876 	if (ipv6_addr_is_ll_all_nodes(&mc->mca_addr) ||
877 	    IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
878 		mc->mca_flags |= MAF_NOREPORT;
879 
880 	return mc;
881 }
882 
883 /*
884  *	device multicast group inc (add if not found)
885  */
886 static int __ipv6_dev_mc_inc(struct net_device *dev,
887 			     const struct in6_addr *addr, unsigned int mode)
888 {
889 	struct ifmcaddr6 *mc;
890 	struct inet6_dev *idev;
891 
892 	ASSERT_RTNL();
893 
894 	/* we need to take a reference on idev */
895 	idev = in6_dev_get(dev);
896 
897 	if (!idev)
898 		return -EINVAL;
899 
900 	write_lock_bh(&idev->lock);
901 	if (idev->dead) {
902 		write_unlock_bh(&idev->lock);
903 		in6_dev_put(idev);
904 		return -ENODEV;
905 	}
906 
907 	for (mc = idev->mc_list; mc; mc = mc->next) {
908 		if (ipv6_addr_equal(&mc->mca_addr, addr)) {
909 			mc->mca_users++;
910 			write_unlock_bh(&idev->lock);
911 			ip6_mc_add_src(idev, &mc->mca_addr, mode, 0, NULL, 0);
912 			in6_dev_put(idev);
913 			return 0;
914 		}
915 	}
916 
917 	mc = mca_alloc(idev, addr, mode);
918 	if (!mc) {
919 		write_unlock_bh(&idev->lock);
920 		in6_dev_put(idev);
921 		return -ENOMEM;
922 	}
923 
924 	mc->next = idev->mc_list;
925 	idev->mc_list = mc;
926 
927 	/* Hold this for the code below before we unlock,
928 	 * it is already exposed via idev->mc_list.
929 	 */
930 	mca_get(mc);
931 	write_unlock_bh(&idev->lock);
932 
933 	mld_del_delrec(idev, mc);
934 	igmp6_group_added(mc);
935 	ma_put(mc);
936 	return 0;
937 }
938 
939 int ipv6_dev_mc_inc(struct net_device *dev, const struct in6_addr *addr)
940 {
941 	return __ipv6_dev_mc_inc(dev, addr, MCAST_EXCLUDE);
942 }
943 EXPORT_SYMBOL(ipv6_dev_mc_inc);
944 
945 /*
946  *	device multicast group del
947  */
948 int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
949 {
950 	struct ifmcaddr6 *ma, **map;
951 
952 	ASSERT_RTNL();
953 
954 	write_lock_bh(&idev->lock);
955 	for (map = &idev->mc_list; (ma = *map) != NULL; map = &ma->next) {
956 		if (ipv6_addr_equal(&ma->mca_addr, addr)) {
957 			if (--ma->mca_users == 0) {
958 				*map = ma->next;
959 				write_unlock_bh(&idev->lock);
960 
961 				igmp6_group_dropped(ma);
962 				ip6_mc_clear_src(ma);
963 
964 				ma_put(ma);
965 				return 0;
966 			}
967 			write_unlock_bh(&idev->lock);
968 			return 0;
969 		}
970 	}
971 	write_unlock_bh(&idev->lock);
972 
973 	return -ENOENT;
974 }
975 
976 int ipv6_dev_mc_dec(struct net_device *dev, const struct in6_addr *addr)
977 {
978 	struct inet6_dev *idev;
979 	int err;
980 
981 	ASSERT_RTNL();
982 
983 	idev = __in6_dev_get(dev);
984 	if (!idev)
985 		err = -ENODEV;
986 	else
987 		err = __ipv6_dev_mc_dec(idev, addr);
988 
989 	return err;
990 }
991 EXPORT_SYMBOL(ipv6_dev_mc_dec);
992 
993 /*
994  *	check if the interface/address pair is valid
995  */
996 bool ipv6_chk_mcast_addr(struct net_device *dev, const struct in6_addr *group,
997 			 const struct in6_addr *src_addr)
998 {
999 	struct inet6_dev *idev;
1000 	struct ifmcaddr6 *mc;
1001 	bool rv = false;
1002 
1003 	rcu_read_lock();
1004 	idev = __in6_dev_get(dev);
1005 	if (idev) {
1006 		read_lock_bh(&idev->lock);
1007 		for (mc = idev->mc_list; mc; mc = mc->next) {
1008 			if (ipv6_addr_equal(&mc->mca_addr, group))
1009 				break;
1010 		}
1011 		if (mc) {
1012 			if (src_addr && !ipv6_addr_any(src_addr)) {
1013 				struct ip6_sf_list *psf;
1014 
1015 				spin_lock_bh(&mc->mca_lock);
1016 				for (psf = mc->mca_sources; psf; psf = psf->sf_next) {
1017 					if (ipv6_addr_equal(&psf->sf_addr, src_addr))
1018 						break;
1019 				}
1020 				if (psf)
1021 					rv = psf->sf_count[MCAST_INCLUDE] ||
1022 						psf->sf_count[MCAST_EXCLUDE] !=
1023 						mc->mca_sfcount[MCAST_EXCLUDE];
1024 				else
1025 					rv = mc->mca_sfcount[MCAST_EXCLUDE] != 0;
1026 				spin_unlock_bh(&mc->mca_lock);
1027 			} else
1028 				rv = true; /* don't filter unspecified source */
1029 		}
1030 		read_unlock_bh(&idev->lock);
1031 	}
1032 	rcu_read_unlock();
1033 	return rv;
1034 }
1035 
1036 static void mld_gq_start_timer(struct inet6_dev *idev)
1037 {
1038 	unsigned long tv = prandom_u32() % idev->mc_maxdelay;
1039 
1040 	idev->mc_gq_running = 1;
1041 	if (!mod_timer(&idev->mc_gq_timer, jiffies+tv+2))
1042 		in6_dev_hold(idev);
1043 }
1044 
1045 static void mld_gq_stop_timer(struct inet6_dev *idev)
1046 {
1047 	idev->mc_gq_running = 0;
1048 	if (del_timer(&idev->mc_gq_timer))
1049 		__in6_dev_put(idev);
1050 }
1051 
1052 static void mld_ifc_start_timer(struct inet6_dev *idev, unsigned long delay)
1053 {
1054 	unsigned long tv = prandom_u32() % delay;
1055 
1056 	if (!mod_timer(&idev->mc_ifc_timer, jiffies+tv+2))
1057 		in6_dev_hold(idev);
1058 }
1059 
1060 static void mld_ifc_stop_timer(struct inet6_dev *idev)
1061 {
1062 	idev->mc_ifc_count = 0;
1063 	if (del_timer(&idev->mc_ifc_timer))
1064 		__in6_dev_put(idev);
1065 }
1066 
1067 static void mld_dad_start_timer(struct inet6_dev *idev, unsigned long delay)
1068 {
1069 	unsigned long tv = prandom_u32() % delay;
1070 
1071 	if (!mod_timer(&idev->mc_dad_timer, jiffies+tv+2))
1072 		in6_dev_hold(idev);
1073 }
1074 
1075 static void mld_dad_stop_timer(struct inet6_dev *idev)
1076 {
1077 	if (del_timer(&idev->mc_dad_timer))
1078 		__in6_dev_put(idev);
1079 }
1080 
1081 /*
1082  *	IGMP handling (alias multicast ICMPv6 messages)
1083  */
1084 
1085 static void igmp6_group_queried(struct ifmcaddr6 *ma, unsigned long resptime)
1086 {
1087 	unsigned long delay = resptime;
1088 
1089 	/* Do not start timer for these addresses */
1090 	if (ipv6_addr_is_ll_all_nodes(&ma->mca_addr) ||
1091 	    IPV6_ADDR_MC_SCOPE(&ma->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
1092 		return;
1093 
1094 	if (del_timer(&ma->mca_timer)) {
1095 		refcount_dec(&ma->mca_refcnt);
1096 		delay = ma->mca_timer.expires - jiffies;
1097 	}
1098 
1099 	if (delay >= resptime)
1100 		delay = prandom_u32() % resptime;
1101 
1102 	ma->mca_timer.expires = jiffies + delay;
1103 	if (!mod_timer(&ma->mca_timer, jiffies + delay))
1104 		refcount_inc(&ma->mca_refcnt);
1105 	ma->mca_flags |= MAF_TIMER_RUNNING;
1106 }
1107 
1108 /* mark EXCLUDE-mode sources */
1109 static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs,
1110 			     const struct in6_addr *srcs)
1111 {
1112 	struct ip6_sf_list *psf;
1113 	int i, scount;
1114 
1115 	scount = 0;
1116 	for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1117 		if (scount == nsrcs)
1118 			break;
1119 		for (i = 0; i < nsrcs; i++) {
1120 			/* skip inactive filters */
1121 			if (psf->sf_count[MCAST_INCLUDE] ||
1122 			    pmc->mca_sfcount[MCAST_EXCLUDE] !=
1123 			    psf->sf_count[MCAST_EXCLUDE])
1124 				break;
1125 			if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1126 				scount++;
1127 				break;
1128 			}
1129 		}
1130 	}
1131 	pmc->mca_flags &= ~MAF_GSQUERY;
1132 	if (scount == nsrcs)	/* all sources excluded */
1133 		return false;
1134 	return true;
1135 }
1136 
1137 static bool mld_marksources(struct ifmcaddr6 *pmc, int nsrcs,
1138 			    const struct in6_addr *srcs)
1139 {
1140 	struct ip6_sf_list *psf;
1141 	int i, scount;
1142 
1143 	if (pmc->mca_sfmode == MCAST_EXCLUDE)
1144 		return mld_xmarksources(pmc, nsrcs, srcs);
1145 
1146 	/* mark INCLUDE-mode sources */
1147 
1148 	scount = 0;
1149 	for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1150 		if (scount == nsrcs)
1151 			break;
1152 		for (i = 0; i < nsrcs; i++) {
1153 			if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1154 				psf->sf_gsresp = 1;
1155 				scount++;
1156 				break;
1157 			}
1158 		}
1159 	}
1160 	if (!scount) {
1161 		pmc->mca_flags &= ~MAF_GSQUERY;
1162 		return false;
1163 	}
1164 	pmc->mca_flags |= MAF_GSQUERY;
1165 	return true;
1166 }
1167 
1168 static int mld_force_mld_version(const struct inet6_dev *idev)
1169 {
1170 	/* Normally, both are 0 here. If enforcement to a particular is
1171 	 * being used, individual device enforcement will have a lower
1172 	 * precedence over 'all' device (.../conf/all/force_mld_version).
1173 	 */
1174 
1175 	if (dev_net(idev->dev)->ipv6.devconf_all->force_mld_version != 0)
1176 		return dev_net(idev->dev)->ipv6.devconf_all->force_mld_version;
1177 	else
1178 		return idev->cnf.force_mld_version;
1179 }
1180 
1181 static bool mld_in_v2_mode_only(const struct inet6_dev *idev)
1182 {
1183 	return mld_force_mld_version(idev) == 2;
1184 }
1185 
1186 static bool mld_in_v1_mode_only(const struct inet6_dev *idev)
1187 {
1188 	return mld_force_mld_version(idev) == 1;
1189 }
1190 
1191 static bool mld_in_v1_mode(const struct inet6_dev *idev)
1192 {
1193 	if (mld_in_v2_mode_only(idev))
1194 		return false;
1195 	if (mld_in_v1_mode_only(idev))
1196 		return true;
1197 	if (idev->mc_v1_seen && time_before(jiffies, idev->mc_v1_seen))
1198 		return true;
1199 
1200 	return false;
1201 }
1202 
1203 static void mld_set_v1_mode(struct inet6_dev *idev)
1204 {
1205 	/* RFC3810, relevant sections:
1206 	 *  - 9.1. Robustness Variable
1207 	 *  - 9.2. Query Interval
1208 	 *  - 9.3. Query Response Interval
1209 	 *  - 9.12. Older Version Querier Present Timeout
1210 	 */
1211 	unsigned long switchback;
1212 
1213 	switchback = (idev->mc_qrv * idev->mc_qi) + idev->mc_qri;
1214 
1215 	idev->mc_v1_seen = jiffies + switchback;
1216 }
1217 
1218 static void mld_update_qrv(struct inet6_dev *idev,
1219 			   const struct mld2_query *mlh2)
1220 {
1221 	/* RFC3810, relevant sections:
1222 	 *  - 5.1.8. QRV (Querier's Robustness Variable)
1223 	 *  - 9.1. Robustness Variable
1224 	 */
1225 
1226 	/* The value of the Robustness Variable MUST NOT be zero,
1227 	 * and SHOULD NOT be one. Catch this here if we ever run
1228 	 * into such a case in future.
1229 	 */
1230 	const int min_qrv = min(MLD_QRV_DEFAULT, sysctl_mld_qrv);
1231 	WARN_ON(idev->mc_qrv == 0);
1232 
1233 	if (mlh2->mld2q_qrv > 0)
1234 		idev->mc_qrv = mlh2->mld2q_qrv;
1235 
1236 	if (unlikely(idev->mc_qrv < min_qrv)) {
1237 		net_warn_ratelimited("IPv6: MLD: clamping QRV from %u to %u!\n",
1238 				     idev->mc_qrv, min_qrv);
1239 		idev->mc_qrv = min_qrv;
1240 	}
1241 }
1242 
1243 static void mld_update_qi(struct inet6_dev *idev,
1244 			  const struct mld2_query *mlh2)
1245 {
1246 	/* RFC3810, relevant sections:
1247 	 *  - 5.1.9. QQIC (Querier's Query Interval Code)
1248 	 *  - 9.2. Query Interval
1249 	 *  - 9.12. Older Version Querier Present Timeout
1250 	 *    (the [Query Interval] in the last Query received)
1251 	 */
1252 	unsigned long mc_qqi;
1253 
1254 	if (mlh2->mld2q_qqic < 128) {
1255 		mc_qqi = mlh2->mld2q_qqic;
1256 	} else {
1257 		unsigned long mc_man, mc_exp;
1258 
1259 		mc_exp = MLDV2_QQIC_EXP(mlh2->mld2q_qqic);
1260 		mc_man = MLDV2_QQIC_MAN(mlh2->mld2q_qqic);
1261 
1262 		mc_qqi = (mc_man | 0x10) << (mc_exp + 3);
1263 	}
1264 
1265 	idev->mc_qi = mc_qqi * HZ;
1266 }
1267 
1268 static void mld_update_qri(struct inet6_dev *idev,
1269 			   const struct mld2_query *mlh2)
1270 {
1271 	/* RFC3810, relevant sections:
1272 	 *  - 5.1.3. Maximum Response Code
1273 	 *  - 9.3. Query Response Interval
1274 	 */
1275 	idev->mc_qri = msecs_to_jiffies(mldv2_mrc(mlh2));
1276 }
1277 
1278 static int mld_process_v1(struct inet6_dev *idev, struct mld_msg *mld,
1279 			  unsigned long *max_delay, bool v1_query)
1280 {
1281 	unsigned long mldv1_md;
1282 
1283 	/* Ignore v1 queries */
1284 	if (mld_in_v2_mode_only(idev))
1285 		return -EINVAL;
1286 
1287 	mldv1_md = ntohs(mld->mld_maxdelay);
1288 
1289 	/* When in MLDv1 fallback and a MLDv2 router start-up being
1290 	 * unaware of current MLDv1 operation, the MRC == MRD mapping
1291 	 * only works when the exponential algorithm is not being
1292 	 * used (as MLDv1 is unaware of such things).
1293 	 *
1294 	 * According to the RFC author, the MLDv2 implementations
1295 	 * he's aware of all use a MRC < 32768 on start up queries.
1296 	 *
1297 	 * Thus, should we *ever* encounter something else larger
1298 	 * than that, just assume the maximum possible within our
1299 	 * reach.
1300 	 */
1301 	if (!v1_query)
1302 		mldv1_md = min(mldv1_md, MLDV1_MRD_MAX_COMPAT);
1303 
1304 	*max_delay = max(msecs_to_jiffies(mldv1_md), 1UL);
1305 
1306 	/* MLDv1 router present: we need to go into v1 mode *only*
1307 	 * when an MLDv1 query is received as per section 9.12. of
1308 	 * RFC3810! And we know from RFC2710 section 3.7 that MLDv1
1309 	 * queries MUST be of exactly 24 octets.
1310 	 */
1311 	if (v1_query)
1312 		mld_set_v1_mode(idev);
1313 
1314 	/* cancel MLDv2 report timer */
1315 	mld_gq_stop_timer(idev);
1316 	/* cancel the interface change timer */
1317 	mld_ifc_stop_timer(idev);
1318 	/* clear deleted report items */
1319 	mld_clear_delrec(idev);
1320 
1321 	return 0;
1322 }
1323 
1324 static int mld_process_v2(struct inet6_dev *idev, struct mld2_query *mld,
1325 			  unsigned long *max_delay)
1326 {
1327 	*max_delay = max(msecs_to_jiffies(mldv2_mrc(mld)), 1UL);
1328 
1329 	mld_update_qrv(idev, mld);
1330 	mld_update_qi(idev, mld);
1331 	mld_update_qri(idev, mld);
1332 
1333 	idev->mc_maxdelay = *max_delay;
1334 
1335 	return 0;
1336 }
1337 
1338 /* called with rcu_read_lock() */
1339 int igmp6_event_query(struct sk_buff *skb)
1340 {
1341 	struct mld2_query *mlh2 = NULL;
1342 	struct ifmcaddr6 *ma;
1343 	const struct in6_addr *group;
1344 	unsigned long max_delay;
1345 	struct inet6_dev *idev;
1346 	struct mld_msg *mld;
1347 	int group_type;
1348 	int mark = 0;
1349 	int len, err;
1350 
1351 	if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
1352 		return -EINVAL;
1353 
1354 	/* compute payload length excluding extension headers */
1355 	len = ntohs(ipv6_hdr(skb)->payload_len) + sizeof(struct ipv6hdr);
1356 	len -= skb_network_header_len(skb);
1357 
1358 	/* RFC3810 6.2
1359 	 * Upon reception of an MLD message that contains a Query, the node
1360 	 * checks if the source address of the message is a valid link-local
1361 	 * address, if the Hop Limit is set to 1, and if the Router Alert
1362 	 * option is present in the Hop-By-Hop Options header of the IPv6
1363 	 * packet.  If any of these checks fails, the packet is dropped.
1364 	 */
1365 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL) ||
1366 	    ipv6_hdr(skb)->hop_limit != 1 ||
1367 	    !(IP6CB(skb)->flags & IP6SKB_ROUTERALERT) ||
1368 	    IP6CB(skb)->ra != htons(IPV6_OPT_ROUTERALERT_MLD))
1369 		return -EINVAL;
1370 
1371 	idev = __in6_dev_get(skb->dev);
1372 	if (!idev)
1373 		return 0;
1374 
1375 	mld = (struct mld_msg *)icmp6_hdr(skb);
1376 	group = &mld->mld_mca;
1377 	group_type = ipv6_addr_type(group);
1378 
1379 	if (group_type != IPV6_ADDR_ANY &&
1380 	    !(group_type&IPV6_ADDR_MULTICAST))
1381 		return -EINVAL;
1382 
1383 	if (len < MLD_V1_QUERY_LEN) {
1384 		return -EINVAL;
1385 	} else if (len == MLD_V1_QUERY_LEN || mld_in_v1_mode(idev)) {
1386 		err = mld_process_v1(idev, mld, &max_delay,
1387 				     len == MLD_V1_QUERY_LEN);
1388 		if (err < 0)
1389 			return err;
1390 	} else if (len >= MLD_V2_QUERY_LEN_MIN) {
1391 		int srcs_offset = sizeof(struct mld2_query) -
1392 				  sizeof(struct icmp6hdr);
1393 
1394 		if (!pskb_may_pull(skb, srcs_offset))
1395 			return -EINVAL;
1396 
1397 		mlh2 = (struct mld2_query *)skb_transport_header(skb);
1398 
1399 		err = mld_process_v2(idev, mlh2, &max_delay);
1400 		if (err < 0)
1401 			return err;
1402 
1403 		if (group_type == IPV6_ADDR_ANY) { /* general query */
1404 			if (mlh2->mld2q_nsrcs)
1405 				return -EINVAL; /* no sources allowed */
1406 
1407 			mld_gq_start_timer(idev);
1408 			return 0;
1409 		}
1410 		/* mark sources to include, if group & source-specific */
1411 		if (mlh2->mld2q_nsrcs != 0) {
1412 			if (!pskb_may_pull(skb, srcs_offset +
1413 			    ntohs(mlh2->mld2q_nsrcs) * sizeof(struct in6_addr)))
1414 				return -EINVAL;
1415 
1416 			mlh2 = (struct mld2_query *)skb_transport_header(skb);
1417 			mark = 1;
1418 		}
1419 	} else {
1420 		return -EINVAL;
1421 	}
1422 
1423 	read_lock_bh(&idev->lock);
1424 	if (group_type == IPV6_ADDR_ANY) {
1425 		for (ma = idev->mc_list; ma; ma = ma->next) {
1426 			spin_lock_bh(&ma->mca_lock);
1427 			igmp6_group_queried(ma, max_delay);
1428 			spin_unlock_bh(&ma->mca_lock);
1429 		}
1430 	} else {
1431 		for (ma = idev->mc_list; ma; ma = ma->next) {
1432 			if (!ipv6_addr_equal(group, &ma->mca_addr))
1433 				continue;
1434 			spin_lock_bh(&ma->mca_lock);
1435 			if (ma->mca_flags & MAF_TIMER_RUNNING) {
1436 				/* gsquery <- gsquery && mark */
1437 				if (!mark)
1438 					ma->mca_flags &= ~MAF_GSQUERY;
1439 			} else {
1440 				/* gsquery <- mark */
1441 				if (mark)
1442 					ma->mca_flags |= MAF_GSQUERY;
1443 				else
1444 					ma->mca_flags &= ~MAF_GSQUERY;
1445 			}
1446 			if (!(ma->mca_flags & MAF_GSQUERY) ||
1447 			    mld_marksources(ma, ntohs(mlh2->mld2q_nsrcs), mlh2->mld2q_srcs))
1448 				igmp6_group_queried(ma, max_delay);
1449 			spin_unlock_bh(&ma->mca_lock);
1450 			break;
1451 		}
1452 	}
1453 	read_unlock_bh(&idev->lock);
1454 
1455 	return 0;
1456 }
1457 
1458 /* called with rcu_read_lock() */
1459 int igmp6_event_report(struct sk_buff *skb)
1460 {
1461 	struct ifmcaddr6 *ma;
1462 	struct inet6_dev *idev;
1463 	struct mld_msg *mld;
1464 	int addr_type;
1465 
1466 	/* Our own report looped back. Ignore it. */
1467 	if (skb->pkt_type == PACKET_LOOPBACK)
1468 		return 0;
1469 
1470 	/* send our report if the MC router may not have heard this report */
1471 	if (skb->pkt_type != PACKET_MULTICAST &&
1472 	    skb->pkt_type != PACKET_BROADCAST)
1473 		return 0;
1474 
1475 	if (!pskb_may_pull(skb, sizeof(*mld) - sizeof(struct icmp6hdr)))
1476 		return -EINVAL;
1477 
1478 	mld = (struct mld_msg *)icmp6_hdr(skb);
1479 
1480 	/* Drop reports with not link local source */
1481 	addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr);
1482 	if (addr_type != IPV6_ADDR_ANY &&
1483 	    !(addr_type&IPV6_ADDR_LINKLOCAL))
1484 		return -EINVAL;
1485 
1486 	idev = __in6_dev_get(skb->dev);
1487 	if (!idev)
1488 		return -ENODEV;
1489 
1490 	/*
1491 	 *	Cancel the timer for this group
1492 	 */
1493 
1494 	read_lock_bh(&idev->lock);
1495 	for (ma = idev->mc_list; ma; ma = ma->next) {
1496 		if (ipv6_addr_equal(&ma->mca_addr, &mld->mld_mca)) {
1497 			spin_lock(&ma->mca_lock);
1498 			if (del_timer(&ma->mca_timer))
1499 				refcount_dec(&ma->mca_refcnt);
1500 			ma->mca_flags &= ~(MAF_LAST_REPORTER|MAF_TIMER_RUNNING);
1501 			spin_unlock(&ma->mca_lock);
1502 			break;
1503 		}
1504 	}
1505 	read_unlock_bh(&idev->lock);
1506 	return 0;
1507 }
1508 
1509 static bool is_in(struct ifmcaddr6 *pmc, struct ip6_sf_list *psf, int type,
1510 		  int gdeleted, int sdeleted)
1511 {
1512 	switch (type) {
1513 	case MLD2_MODE_IS_INCLUDE:
1514 	case MLD2_MODE_IS_EXCLUDE:
1515 		if (gdeleted || sdeleted)
1516 			return false;
1517 		if (!((pmc->mca_flags & MAF_GSQUERY) && !psf->sf_gsresp)) {
1518 			if (pmc->mca_sfmode == MCAST_INCLUDE)
1519 				return true;
1520 			/* don't include if this source is excluded
1521 			 * in all filters
1522 			 */
1523 			if (psf->sf_count[MCAST_INCLUDE])
1524 				return type == MLD2_MODE_IS_INCLUDE;
1525 			return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1526 				psf->sf_count[MCAST_EXCLUDE];
1527 		}
1528 		return false;
1529 	case MLD2_CHANGE_TO_INCLUDE:
1530 		if (gdeleted || sdeleted)
1531 			return false;
1532 		return psf->sf_count[MCAST_INCLUDE] != 0;
1533 	case MLD2_CHANGE_TO_EXCLUDE:
1534 		if (gdeleted || sdeleted)
1535 			return false;
1536 		if (pmc->mca_sfcount[MCAST_EXCLUDE] == 0 ||
1537 		    psf->sf_count[MCAST_INCLUDE])
1538 			return false;
1539 		return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1540 			psf->sf_count[MCAST_EXCLUDE];
1541 	case MLD2_ALLOW_NEW_SOURCES:
1542 		if (gdeleted || !psf->sf_crcount)
1543 			return false;
1544 		return (pmc->mca_sfmode == MCAST_INCLUDE) ^ sdeleted;
1545 	case MLD2_BLOCK_OLD_SOURCES:
1546 		if (pmc->mca_sfmode == MCAST_INCLUDE)
1547 			return gdeleted || (psf->sf_crcount && sdeleted);
1548 		return psf->sf_crcount && !gdeleted && !sdeleted;
1549 	}
1550 	return false;
1551 }
1552 
1553 static int
1554 mld_scount(struct ifmcaddr6 *pmc, int type, int gdeleted, int sdeleted)
1555 {
1556 	struct ip6_sf_list *psf;
1557 	int scount = 0;
1558 
1559 	for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
1560 		if (!is_in(pmc, psf, type, gdeleted, sdeleted))
1561 			continue;
1562 		scount++;
1563 	}
1564 	return scount;
1565 }
1566 
1567 static void ip6_mc_hdr(struct sock *sk, struct sk_buff *skb,
1568 		       struct net_device *dev,
1569 		       const struct in6_addr *saddr,
1570 		       const struct in6_addr *daddr,
1571 		       int proto, int len)
1572 {
1573 	struct ipv6hdr *hdr;
1574 
1575 	skb->protocol = htons(ETH_P_IPV6);
1576 	skb->dev = dev;
1577 
1578 	skb_reset_network_header(skb);
1579 	skb_put(skb, sizeof(struct ipv6hdr));
1580 	hdr = ipv6_hdr(skb);
1581 
1582 	ip6_flow_hdr(hdr, 0, 0);
1583 
1584 	hdr->payload_len = htons(len);
1585 	hdr->nexthdr = proto;
1586 	hdr->hop_limit = inet6_sk(sk)->hop_limit;
1587 
1588 	hdr->saddr = *saddr;
1589 	hdr->daddr = *daddr;
1590 }
1591 
1592 static struct sk_buff *mld_newpack(struct inet6_dev *idev, unsigned int mtu)
1593 {
1594 	struct net_device *dev = idev->dev;
1595 	struct net *net = dev_net(dev);
1596 	struct sock *sk = net->ipv6.igmp_sk;
1597 	struct sk_buff *skb;
1598 	struct mld2_report *pmr;
1599 	struct in6_addr addr_buf;
1600 	const struct in6_addr *saddr;
1601 	int hlen = LL_RESERVED_SPACE(dev);
1602 	int tlen = dev->needed_tailroom;
1603 	unsigned int size = mtu + hlen + tlen;
1604 	int err;
1605 	u8 ra[8] = { IPPROTO_ICMPV6, 0,
1606 		     IPV6_TLV_ROUTERALERT, 2, 0, 0,
1607 		     IPV6_TLV_PADN, 0 };
1608 
1609 	/* we assume size > sizeof(ra) here */
1610 	/* limit our allocations to order-0 page */
1611 	size = min_t(int, size, SKB_MAX_ORDER(0, 0));
1612 	skb = sock_alloc_send_skb(sk, size, 1, &err);
1613 
1614 	if (!skb)
1615 		return NULL;
1616 
1617 	skb->priority = TC_PRIO_CONTROL;
1618 	skb_reserve(skb, hlen);
1619 	skb_tailroom_reserve(skb, mtu, tlen);
1620 
1621 	if (__ipv6_get_lladdr(idev, &addr_buf, IFA_F_TENTATIVE)) {
1622 		/* <draft-ietf-magma-mld-source-05.txt>:
1623 		 * use unspecified address as the source address
1624 		 * when a valid link-local address is not available.
1625 		 */
1626 		saddr = &in6addr_any;
1627 	} else
1628 		saddr = &addr_buf;
1629 
1630 	ip6_mc_hdr(sk, skb, dev, saddr, &mld2_all_mcr, NEXTHDR_HOP, 0);
1631 
1632 	skb_put_data(skb, ra, sizeof(ra));
1633 
1634 	skb_set_transport_header(skb, skb_tail_pointer(skb) - skb->data);
1635 	skb_put(skb, sizeof(*pmr));
1636 	pmr = (struct mld2_report *)skb_transport_header(skb);
1637 	pmr->mld2r_type = ICMPV6_MLD2_REPORT;
1638 	pmr->mld2r_resv1 = 0;
1639 	pmr->mld2r_cksum = 0;
1640 	pmr->mld2r_resv2 = 0;
1641 	pmr->mld2r_ngrec = 0;
1642 	return skb;
1643 }
1644 
1645 static void mld_sendpack(struct sk_buff *skb)
1646 {
1647 	struct ipv6hdr *pip6 = ipv6_hdr(skb);
1648 	struct mld2_report *pmr =
1649 			      (struct mld2_report *)skb_transport_header(skb);
1650 	int payload_len, mldlen;
1651 	struct inet6_dev *idev;
1652 	struct net *net = dev_net(skb->dev);
1653 	int err;
1654 	struct flowi6 fl6;
1655 	struct dst_entry *dst;
1656 
1657 	rcu_read_lock();
1658 	idev = __in6_dev_get(skb->dev);
1659 	IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_OUT, skb->len);
1660 
1661 	payload_len = (skb_tail_pointer(skb) - skb_network_header(skb)) -
1662 		sizeof(*pip6);
1663 	mldlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1664 	pip6->payload_len = htons(payload_len);
1665 
1666 	pmr->mld2r_cksum = csum_ipv6_magic(&pip6->saddr, &pip6->daddr, mldlen,
1667 					   IPPROTO_ICMPV6,
1668 					   csum_partial(skb_transport_header(skb),
1669 							mldlen, 0));
1670 
1671 	icmpv6_flow_init(net->ipv6.igmp_sk, &fl6, ICMPV6_MLD2_REPORT,
1672 			 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
1673 			 skb->dev->ifindex);
1674 	dst = icmp6_dst_alloc(skb->dev, &fl6);
1675 
1676 	err = 0;
1677 	if (IS_ERR(dst)) {
1678 		err = PTR_ERR(dst);
1679 		dst = NULL;
1680 	}
1681 	skb_dst_set(skb, dst);
1682 	if (err)
1683 		goto err_out;
1684 
1685 	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
1686 		      net, net->ipv6.igmp_sk, skb, NULL, skb->dev,
1687 		      dst_output);
1688 out:
1689 	if (!err) {
1690 		ICMP6MSGOUT_INC_STATS(net, idev, ICMPV6_MLD2_REPORT);
1691 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1692 	} else {
1693 		IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
1694 	}
1695 
1696 	rcu_read_unlock();
1697 	return;
1698 
1699 err_out:
1700 	kfree_skb(skb);
1701 	goto out;
1702 }
1703 
1704 static int grec_size(struct ifmcaddr6 *pmc, int type, int gdel, int sdel)
1705 {
1706 	return sizeof(struct mld2_grec) + 16 * mld_scount(pmc,type,gdel,sdel);
1707 }
1708 
1709 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1710 	int type, struct mld2_grec **ppgr, unsigned int mtu)
1711 {
1712 	struct mld2_report *pmr;
1713 	struct mld2_grec *pgr;
1714 
1715 	if (!skb) {
1716 		skb = mld_newpack(pmc->idev, mtu);
1717 		if (!skb)
1718 			return NULL;
1719 	}
1720 	pgr = skb_put(skb, sizeof(struct mld2_grec));
1721 	pgr->grec_type = type;
1722 	pgr->grec_auxwords = 0;
1723 	pgr->grec_nsrcs = 0;
1724 	pgr->grec_mca = pmc->mca_addr;	/* structure copy */
1725 	pmr = (struct mld2_report *)skb_transport_header(skb);
1726 	pmr->mld2r_ngrec = htons(ntohs(pmr->mld2r_ngrec)+1);
1727 	*ppgr = pgr;
1728 	return skb;
1729 }
1730 
1731 #define AVAILABLE(skb)	((skb) ? skb_availroom(skb) : 0)
1732 
1733 static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1734 	int type, int gdeleted, int sdeleted, int crsend)
1735 {
1736 	struct inet6_dev *idev = pmc->idev;
1737 	struct net_device *dev = idev->dev;
1738 	struct mld2_report *pmr;
1739 	struct mld2_grec *pgr = NULL;
1740 	struct ip6_sf_list *psf, *psf_next, *psf_prev, **psf_list;
1741 	int scount, stotal, first, isquery, truncate;
1742 	unsigned int mtu;
1743 
1744 	if (pmc->mca_flags & MAF_NOREPORT)
1745 		return skb;
1746 
1747 	mtu = READ_ONCE(dev->mtu);
1748 	if (mtu < IPV6_MIN_MTU)
1749 		return skb;
1750 
1751 	isquery = type == MLD2_MODE_IS_INCLUDE ||
1752 		  type == MLD2_MODE_IS_EXCLUDE;
1753 	truncate = type == MLD2_MODE_IS_EXCLUDE ||
1754 		    type == MLD2_CHANGE_TO_EXCLUDE;
1755 
1756 	stotal = scount = 0;
1757 
1758 	psf_list = sdeleted ? &pmc->mca_tomb : &pmc->mca_sources;
1759 
1760 	if (!*psf_list)
1761 		goto empty_source;
1762 
1763 	pmr = skb ? (struct mld2_report *)skb_transport_header(skb) : NULL;
1764 
1765 	/* EX and TO_EX get a fresh packet, if needed */
1766 	if (truncate) {
1767 		if (pmr && pmr->mld2r_ngrec &&
1768 		    AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
1769 			if (skb)
1770 				mld_sendpack(skb);
1771 			skb = mld_newpack(idev, mtu);
1772 		}
1773 	}
1774 	first = 1;
1775 	psf_prev = NULL;
1776 	for (psf = *psf_list; psf; psf = psf_next) {
1777 		struct in6_addr *psrc;
1778 
1779 		psf_next = psf->sf_next;
1780 
1781 		if (!is_in(pmc, psf, type, gdeleted, sdeleted) && !crsend) {
1782 			psf_prev = psf;
1783 			continue;
1784 		}
1785 
1786 		/* Based on RFC3810 6.1. Should not send source-list change
1787 		 * records when there is a filter mode change.
1788 		 */
1789 		if (((gdeleted && pmc->mca_sfmode == MCAST_EXCLUDE) ||
1790 		     (!gdeleted && pmc->mca_crcount)) &&
1791 		    (type == MLD2_ALLOW_NEW_SOURCES ||
1792 		     type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount)
1793 			goto decrease_sf_crcount;
1794 
1795 		/* clear marks on query responses */
1796 		if (isquery)
1797 			psf->sf_gsresp = 0;
1798 
1799 		if (AVAILABLE(skb) < sizeof(*psrc) +
1800 		    first*sizeof(struct mld2_grec)) {
1801 			if (truncate && !first)
1802 				break;	 /* truncate these */
1803 			if (pgr)
1804 				pgr->grec_nsrcs = htons(scount);
1805 			if (skb)
1806 				mld_sendpack(skb);
1807 			skb = mld_newpack(idev, mtu);
1808 			first = 1;
1809 			scount = 0;
1810 		}
1811 		if (first) {
1812 			skb = add_grhead(skb, pmc, type, &pgr, mtu);
1813 			first = 0;
1814 		}
1815 		if (!skb)
1816 			return NULL;
1817 		psrc = skb_put(skb, sizeof(*psrc));
1818 		*psrc = psf->sf_addr;
1819 		scount++; stotal++;
1820 		if ((type == MLD2_ALLOW_NEW_SOURCES ||
1821 		     type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
1822 decrease_sf_crcount:
1823 			psf->sf_crcount--;
1824 			if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
1825 				if (psf_prev)
1826 					psf_prev->sf_next = psf->sf_next;
1827 				else
1828 					*psf_list = psf->sf_next;
1829 				kfree(psf);
1830 				continue;
1831 			}
1832 		}
1833 		psf_prev = psf;
1834 	}
1835 
1836 empty_source:
1837 	if (!stotal) {
1838 		if (type == MLD2_ALLOW_NEW_SOURCES ||
1839 		    type == MLD2_BLOCK_OLD_SOURCES)
1840 			return skb;
1841 		if (pmc->mca_crcount || isquery || crsend) {
1842 			/* make sure we have room for group header */
1843 			if (skb && AVAILABLE(skb) < sizeof(struct mld2_grec)) {
1844 				mld_sendpack(skb);
1845 				skb = NULL; /* add_grhead will get a new one */
1846 			}
1847 			skb = add_grhead(skb, pmc, type, &pgr, mtu);
1848 		}
1849 	}
1850 	if (pgr)
1851 		pgr->grec_nsrcs = htons(scount);
1852 
1853 	if (isquery)
1854 		pmc->mca_flags &= ~MAF_GSQUERY;	/* clear query state */
1855 	return skb;
1856 }
1857 
1858 static void mld_send_report(struct inet6_dev *idev, struct ifmcaddr6 *pmc)
1859 {
1860 	struct sk_buff *skb = NULL;
1861 	int type;
1862 
1863 	read_lock_bh(&idev->lock);
1864 	if (!pmc) {
1865 		for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
1866 			if (pmc->mca_flags & MAF_NOREPORT)
1867 				continue;
1868 			spin_lock_bh(&pmc->mca_lock);
1869 			if (pmc->mca_sfcount[MCAST_EXCLUDE])
1870 				type = MLD2_MODE_IS_EXCLUDE;
1871 			else
1872 				type = MLD2_MODE_IS_INCLUDE;
1873 			skb = add_grec(skb, pmc, type, 0, 0, 0);
1874 			spin_unlock_bh(&pmc->mca_lock);
1875 		}
1876 	} else {
1877 		spin_lock_bh(&pmc->mca_lock);
1878 		if (pmc->mca_sfcount[MCAST_EXCLUDE])
1879 			type = MLD2_MODE_IS_EXCLUDE;
1880 		else
1881 			type = MLD2_MODE_IS_INCLUDE;
1882 		skb = add_grec(skb, pmc, type, 0, 0, 0);
1883 		spin_unlock_bh(&pmc->mca_lock);
1884 	}
1885 	read_unlock_bh(&idev->lock);
1886 	if (skb)
1887 		mld_sendpack(skb);
1888 }
1889 
1890 /*
1891  * remove zero-count source records from a source filter list
1892  */
1893 static void mld_clear_zeros(struct ip6_sf_list **ppsf)
1894 {
1895 	struct ip6_sf_list *psf_prev, *psf_next, *psf;
1896 
1897 	psf_prev = NULL;
1898 	for (psf = *ppsf; psf; psf = psf_next) {
1899 		psf_next = psf->sf_next;
1900 		if (psf->sf_crcount == 0) {
1901 			if (psf_prev)
1902 				psf_prev->sf_next = psf->sf_next;
1903 			else
1904 				*ppsf = psf->sf_next;
1905 			kfree(psf);
1906 		} else
1907 			psf_prev = psf;
1908 	}
1909 }
1910 
1911 static void mld_send_cr(struct inet6_dev *idev)
1912 {
1913 	struct ifmcaddr6 *pmc, *pmc_prev, *pmc_next;
1914 	struct sk_buff *skb = NULL;
1915 	int type, dtype;
1916 
1917 	read_lock_bh(&idev->lock);
1918 	spin_lock(&idev->mc_lock);
1919 
1920 	/* deleted MCA's */
1921 	pmc_prev = NULL;
1922 	for (pmc = idev->mc_tomb; pmc; pmc = pmc_next) {
1923 		pmc_next = pmc->next;
1924 		if (pmc->mca_sfmode == MCAST_INCLUDE) {
1925 			type = MLD2_BLOCK_OLD_SOURCES;
1926 			dtype = MLD2_BLOCK_OLD_SOURCES;
1927 			skb = add_grec(skb, pmc, type, 1, 0, 0);
1928 			skb = add_grec(skb, pmc, dtype, 1, 1, 0);
1929 		}
1930 		if (pmc->mca_crcount) {
1931 			if (pmc->mca_sfmode == MCAST_EXCLUDE) {
1932 				type = MLD2_CHANGE_TO_INCLUDE;
1933 				skb = add_grec(skb, pmc, type, 1, 0, 0);
1934 			}
1935 			pmc->mca_crcount--;
1936 			if (pmc->mca_crcount == 0) {
1937 				mld_clear_zeros(&pmc->mca_tomb);
1938 				mld_clear_zeros(&pmc->mca_sources);
1939 			}
1940 		}
1941 		if (pmc->mca_crcount == 0 && !pmc->mca_tomb &&
1942 		    !pmc->mca_sources) {
1943 			if (pmc_prev)
1944 				pmc_prev->next = pmc_next;
1945 			else
1946 				idev->mc_tomb = pmc_next;
1947 			in6_dev_put(pmc->idev);
1948 			kfree(pmc);
1949 		} else
1950 			pmc_prev = pmc;
1951 	}
1952 	spin_unlock(&idev->mc_lock);
1953 
1954 	/* change recs */
1955 	for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
1956 		spin_lock_bh(&pmc->mca_lock);
1957 		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
1958 			type = MLD2_BLOCK_OLD_SOURCES;
1959 			dtype = MLD2_ALLOW_NEW_SOURCES;
1960 		} else {
1961 			type = MLD2_ALLOW_NEW_SOURCES;
1962 			dtype = MLD2_BLOCK_OLD_SOURCES;
1963 		}
1964 		skb = add_grec(skb, pmc, type, 0, 0, 0);
1965 		skb = add_grec(skb, pmc, dtype, 0, 1, 0);	/* deleted sources */
1966 
1967 		/* filter mode changes */
1968 		if (pmc->mca_crcount) {
1969 			if (pmc->mca_sfmode == MCAST_EXCLUDE)
1970 				type = MLD2_CHANGE_TO_EXCLUDE;
1971 			else
1972 				type = MLD2_CHANGE_TO_INCLUDE;
1973 			skb = add_grec(skb, pmc, type, 0, 0, 0);
1974 			pmc->mca_crcount--;
1975 		}
1976 		spin_unlock_bh(&pmc->mca_lock);
1977 	}
1978 	read_unlock_bh(&idev->lock);
1979 	if (!skb)
1980 		return;
1981 	(void) mld_sendpack(skb);
1982 }
1983 
1984 static void igmp6_send(struct in6_addr *addr, struct net_device *dev, int type)
1985 {
1986 	struct net *net = dev_net(dev);
1987 	struct sock *sk = net->ipv6.igmp_sk;
1988 	struct inet6_dev *idev;
1989 	struct sk_buff *skb;
1990 	struct mld_msg *hdr;
1991 	const struct in6_addr *snd_addr, *saddr;
1992 	struct in6_addr addr_buf;
1993 	int hlen = LL_RESERVED_SPACE(dev);
1994 	int tlen = dev->needed_tailroom;
1995 	int err, len, payload_len, full_len;
1996 	u8 ra[8] = { IPPROTO_ICMPV6, 0,
1997 		     IPV6_TLV_ROUTERALERT, 2, 0, 0,
1998 		     IPV6_TLV_PADN, 0 };
1999 	struct flowi6 fl6;
2000 	struct dst_entry *dst;
2001 
2002 	if (type == ICMPV6_MGM_REDUCTION)
2003 		snd_addr = &in6addr_linklocal_allrouters;
2004 	else
2005 		snd_addr = addr;
2006 
2007 	len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
2008 	payload_len = len + sizeof(ra);
2009 	full_len = sizeof(struct ipv6hdr) + payload_len;
2010 
2011 	rcu_read_lock();
2012 	IP6_UPD_PO_STATS(net, __in6_dev_get(dev),
2013 		      IPSTATS_MIB_OUT, full_len);
2014 	rcu_read_unlock();
2015 
2016 	skb = sock_alloc_send_skb(sk, hlen + tlen + full_len, 1, &err);
2017 
2018 	if (!skb) {
2019 		rcu_read_lock();
2020 		IP6_INC_STATS(net, __in6_dev_get(dev),
2021 			      IPSTATS_MIB_OUTDISCARDS);
2022 		rcu_read_unlock();
2023 		return;
2024 	}
2025 	skb->priority = TC_PRIO_CONTROL;
2026 	skb_reserve(skb, hlen);
2027 
2028 	if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
2029 		/* <draft-ietf-magma-mld-source-05.txt>:
2030 		 * use unspecified address as the source address
2031 		 * when a valid link-local address is not available.
2032 		 */
2033 		saddr = &in6addr_any;
2034 	} else
2035 		saddr = &addr_buf;
2036 
2037 	ip6_mc_hdr(sk, skb, dev, saddr, snd_addr, NEXTHDR_HOP, payload_len);
2038 
2039 	skb_put_data(skb, ra, sizeof(ra));
2040 
2041 	hdr = skb_put_zero(skb, sizeof(struct mld_msg));
2042 	hdr->mld_type = type;
2043 	hdr->mld_mca = *addr;
2044 
2045 	hdr->mld_cksum = csum_ipv6_magic(saddr, snd_addr, len,
2046 					 IPPROTO_ICMPV6,
2047 					 csum_partial(hdr, len, 0));
2048 
2049 	rcu_read_lock();
2050 	idev = __in6_dev_get(skb->dev);
2051 
2052 	icmpv6_flow_init(sk, &fl6, type,
2053 			 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
2054 			 skb->dev->ifindex);
2055 	dst = icmp6_dst_alloc(skb->dev, &fl6);
2056 	if (IS_ERR(dst)) {
2057 		err = PTR_ERR(dst);
2058 		goto err_out;
2059 	}
2060 
2061 	skb_dst_set(skb, dst);
2062 	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
2063 		      net, sk, skb, NULL, skb->dev,
2064 		      dst_output);
2065 out:
2066 	if (!err) {
2067 		ICMP6MSGOUT_INC_STATS(net, idev, type);
2068 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
2069 	} else
2070 		IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
2071 
2072 	rcu_read_unlock();
2073 	return;
2074 
2075 err_out:
2076 	kfree_skb(skb);
2077 	goto out;
2078 }
2079 
2080 static void mld_send_initial_cr(struct inet6_dev *idev)
2081 {
2082 	struct sk_buff *skb;
2083 	struct ifmcaddr6 *pmc;
2084 	int type;
2085 
2086 	if (mld_in_v1_mode(idev))
2087 		return;
2088 
2089 	skb = NULL;
2090 	read_lock_bh(&idev->lock);
2091 	for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2092 		spin_lock_bh(&pmc->mca_lock);
2093 		if (pmc->mca_sfcount[MCAST_EXCLUDE])
2094 			type = MLD2_CHANGE_TO_EXCLUDE;
2095 		else
2096 			type = MLD2_ALLOW_NEW_SOURCES;
2097 		skb = add_grec(skb, pmc, type, 0, 0, 1);
2098 		spin_unlock_bh(&pmc->mca_lock);
2099 	}
2100 	read_unlock_bh(&idev->lock);
2101 	if (skb)
2102 		mld_sendpack(skb);
2103 }
2104 
2105 void ipv6_mc_dad_complete(struct inet6_dev *idev)
2106 {
2107 	idev->mc_dad_count = idev->mc_qrv;
2108 	if (idev->mc_dad_count) {
2109 		mld_send_initial_cr(idev);
2110 		idev->mc_dad_count--;
2111 		if (idev->mc_dad_count)
2112 			mld_dad_start_timer(idev,
2113 					    unsolicited_report_interval(idev));
2114 	}
2115 }
2116 
2117 static void mld_dad_timer_expire(struct timer_list *t)
2118 {
2119 	struct inet6_dev *idev = from_timer(idev, t, mc_dad_timer);
2120 
2121 	mld_send_initial_cr(idev);
2122 	if (idev->mc_dad_count) {
2123 		idev->mc_dad_count--;
2124 		if (idev->mc_dad_count)
2125 			mld_dad_start_timer(idev,
2126 					    unsolicited_report_interval(idev));
2127 	}
2128 	in6_dev_put(idev);
2129 }
2130 
2131 static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
2132 	const struct in6_addr *psfsrc)
2133 {
2134 	struct ip6_sf_list *psf, *psf_prev;
2135 	int rv = 0;
2136 
2137 	psf_prev = NULL;
2138 	for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2139 		if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2140 			break;
2141 		psf_prev = psf;
2142 	}
2143 	if (!psf || psf->sf_count[sfmode] == 0) {
2144 		/* source filter not found, or count wrong =>  bug */
2145 		return -ESRCH;
2146 	}
2147 	psf->sf_count[sfmode]--;
2148 	if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
2149 		struct inet6_dev *idev = pmc->idev;
2150 
2151 		/* no more filters for this source */
2152 		if (psf_prev)
2153 			psf_prev->sf_next = psf->sf_next;
2154 		else
2155 			pmc->mca_sources = psf->sf_next;
2156 		if (psf->sf_oldin && !(pmc->mca_flags & MAF_NOREPORT) &&
2157 		    !mld_in_v1_mode(idev)) {
2158 			psf->sf_crcount = idev->mc_qrv;
2159 			psf->sf_next = pmc->mca_tomb;
2160 			pmc->mca_tomb = psf;
2161 			rv = 1;
2162 		} else
2163 			kfree(psf);
2164 	}
2165 	return rv;
2166 }
2167 
2168 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2169 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
2170 			  int delta)
2171 {
2172 	struct ifmcaddr6 *pmc;
2173 	int	changerec = 0;
2174 	int	i, err;
2175 
2176 	if (!idev)
2177 		return -ENODEV;
2178 	read_lock_bh(&idev->lock);
2179 	for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2180 		if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2181 			break;
2182 	}
2183 	if (!pmc) {
2184 		/* MCA not found?? bug */
2185 		read_unlock_bh(&idev->lock);
2186 		return -ESRCH;
2187 	}
2188 	spin_lock_bh(&pmc->mca_lock);
2189 	sf_markstate(pmc);
2190 	if (!delta) {
2191 		if (!pmc->mca_sfcount[sfmode]) {
2192 			spin_unlock_bh(&pmc->mca_lock);
2193 			read_unlock_bh(&idev->lock);
2194 			return -EINVAL;
2195 		}
2196 		pmc->mca_sfcount[sfmode]--;
2197 	}
2198 	err = 0;
2199 	for (i = 0; i < sfcount; i++) {
2200 		int rv = ip6_mc_del1_src(pmc, sfmode, &psfsrc[i]);
2201 
2202 		changerec |= rv > 0;
2203 		if (!err && rv < 0)
2204 			err = rv;
2205 	}
2206 	if (pmc->mca_sfmode == MCAST_EXCLUDE &&
2207 	    pmc->mca_sfcount[MCAST_EXCLUDE] == 0 &&
2208 	    pmc->mca_sfcount[MCAST_INCLUDE]) {
2209 		struct ip6_sf_list *psf;
2210 
2211 		/* filter mode change */
2212 		pmc->mca_sfmode = MCAST_INCLUDE;
2213 		pmc->mca_crcount = idev->mc_qrv;
2214 		idev->mc_ifc_count = pmc->mca_crcount;
2215 		for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2216 			psf->sf_crcount = 0;
2217 		mld_ifc_event(pmc->idev);
2218 	} else if (sf_setstate(pmc) || changerec)
2219 		mld_ifc_event(pmc->idev);
2220 	spin_unlock_bh(&pmc->mca_lock);
2221 	read_unlock_bh(&idev->lock);
2222 	return err;
2223 }
2224 
2225 /*
2226  * Add multicast single-source filter to the interface list
2227  */
2228 static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode,
2229 	const struct in6_addr *psfsrc)
2230 {
2231 	struct ip6_sf_list *psf, *psf_prev;
2232 
2233 	psf_prev = NULL;
2234 	for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2235 		if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2236 			break;
2237 		psf_prev = psf;
2238 	}
2239 	if (!psf) {
2240 		psf = kzalloc(sizeof(*psf), GFP_ATOMIC);
2241 		if (!psf)
2242 			return -ENOBUFS;
2243 
2244 		psf->sf_addr = *psfsrc;
2245 		if (psf_prev) {
2246 			psf_prev->sf_next = psf;
2247 		} else
2248 			pmc->mca_sources = psf;
2249 	}
2250 	psf->sf_count[sfmode]++;
2251 	return 0;
2252 }
2253 
2254 static void sf_markstate(struct ifmcaddr6 *pmc)
2255 {
2256 	struct ip6_sf_list *psf;
2257 	int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2258 
2259 	for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2260 		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2261 			psf->sf_oldin = mca_xcount ==
2262 				psf->sf_count[MCAST_EXCLUDE] &&
2263 				!psf->sf_count[MCAST_INCLUDE];
2264 		} else
2265 			psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
2266 }
2267 
2268 static int sf_setstate(struct ifmcaddr6 *pmc)
2269 {
2270 	struct ip6_sf_list *psf, *dpsf;
2271 	int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2272 	int qrv = pmc->idev->mc_qrv;
2273 	int new_in, rv;
2274 
2275 	rv = 0;
2276 	for (psf = pmc->mca_sources; psf; psf = psf->sf_next) {
2277 		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2278 			new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
2279 				!psf->sf_count[MCAST_INCLUDE];
2280 		} else
2281 			new_in = psf->sf_count[MCAST_INCLUDE] != 0;
2282 		if (new_in) {
2283 			if (!psf->sf_oldin) {
2284 				struct ip6_sf_list *prev = NULL;
2285 
2286 				for (dpsf = pmc->mca_tomb; dpsf;
2287 				     dpsf = dpsf->sf_next) {
2288 					if (ipv6_addr_equal(&dpsf->sf_addr,
2289 					    &psf->sf_addr))
2290 						break;
2291 					prev = dpsf;
2292 				}
2293 				if (dpsf) {
2294 					if (prev)
2295 						prev->sf_next = dpsf->sf_next;
2296 					else
2297 						pmc->mca_tomb = dpsf->sf_next;
2298 					kfree(dpsf);
2299 				}
2300 				psf->sf_crcount = qrv;
2301 				rv++;
2302 			}
2303 		} else if (psf->sf_oldin) {
2304 			psf->sf_crcount = 0;
2305 			/*
2306 			 * add or update "delete" records if an active filter
2307 			 * is now inactive
2308 			 */
2309 			for (dpsf = pmc->mca_tomb; dpsf; dpsf = dpsf->sf_next)
2310 				if (ipv6_addr_equal(&dpsf->sf_addr,
2311 				    &psf->sf_addr))
2312 					break;
2313 			if (!dpsf) {
2314 				dpsf = kmalloc(sizeof(*dpsf), GFP_ATOMIC);
2315 				if (!dpsf)
2316 					continue;
2317 				*dpsf = *psf;
2318 				/* pmc->mca_lock held by callers */
2319 				dpsf->sf_next = pmc->mca_tomb;
2320 				pmc->mca_tomb = dpsf;
2321 			}
2322 			dpsf->sf_crcount = qrv;
2323 			rv++;
2324 		}
2325 	}
2326 	return rv;
2327 }
2328 
2329 /*
2330  * Add multicast source filter list to the interface list
2331  */
2332 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2333 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
2334 			  int delta)
2335 {
2336 	struct ifmcaddr6 *pmc;
2337 	int	isexclude;
2338 	int	i, err;
2339 
2340 	if (!idev)
2341 		return -ENODEV;
2342 	read_lock_bh(&idev->lock);
2343 	for (pmc = idev->mc_list; pmc; pmc = pmc->next) {
2344 		if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2345 			break;
2346 	}
2347 	if (!pmc) {
2348 		/* MCA not found?? bug */
2349 		read_unlock_bh(&idev->lock);
2350 		return -ESRCH;
2351 	}
2352 	spin_lock_bh(&pmc->mca_lock);
2353 
2354 	sf_markstate(pmc);
2355 	isexclude = pmc->mca_sfmode == MCAST_EXCLUDE;
2356 	if (!delta)
2357 		pmc->mca_sfcount[sfmode]++;
2358 	err = 0;
2359 	for (i = 0; i < sfcount; i++) {
2360 		err = ip6_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2361 		if (err)
2362 			break;
2363 	}
2364 	if (err) {
2365 		int j;
2366 
2367 		if (!delta)
2368 			pmc->mca_sfcount[sfmode]--;
2369 		for (j = 0; j < i; j++)
2370 			ip6_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2371 	} else if (isexclude != (pmc->mca_sfcount[MCAST_EXCLUDE] != 0)) {
2372 		struct ip6_sf_list *psf;
2373 
2374 		/* filter mode change */
2375 		if (pmc->mca_sfcount[MCAST_EXCLUDE])
2376 			pmc->mca_sfmode = MCAST_EXCLUDE;
2377 		else if (pmc->mca_sfcount[MCAST_INCLUDE])
2378 			pmc->mca_sfmode = MCAST_INCLUDE;
2379 		/* else no filters; keep old mode for reports */
2380 
2381 		pmc->mca_crcount = idev->mc_qrv;
2382 		idev->mc_ifc_count = pmc->mca_crcount;
2383 		for (psf = pmc->mca_sources; psf; psf = psf->sf_next)
2384 			psf->sf_crcount = 0;
2385 		mld_ifc_event(idev);
2386 	} else if (sf_setstate(pmc))
2387 		mld_ifc_event(idev);
2388 	spin_unlock_bh(&pmc->mca_lock);
2389 	read_unlock_bh(&idev->lock);
2390 	return err;
2391 }
2392 
2393 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc)
2394 {
2395 	struct ip6_sf_list *psf, *nextpsf;
2396 
2397 	for (psf = pmc->mca_tomb; psf; psf = nextpsf) {
2398 		nextpsf = psf->sf_next;
2399 		kfree(psf);
2400 	}
2401 	pmc->mca_tomb = NULL;
2402 	for (psf = pmc->mca_sources; psf; psf = nextpsf) {
2403 		nextpsf = psf->sf_next;
2404 		kfree(psf);
2405 	}
2406 	pmc->mca_sources = NULL;
2407 	pmc->mca_sfmode = MCAST_EXCLUDE;
2408 	pmc->mca_sfcount[MCAST_INCLUDE] = 0;
2409 	pmc->mca_sfcount[MCAST_EXCLUDE] = 1;
2410 }
2411 
2412 
2413 static void igmp6_join_group(struct ifmcaddr6 *ma)
2414 {
2415 	unsigned long delay;
2416 
2417 	if (ma->mca_flags & MAF_NOREPORT)
2418 		return;
2419 
2420 	igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2421 
2422 	delay = prandom_u32() % unsolicited_report_interval(ma->idev);
2423 
2424 	spin_lock_bh(&ma->mca_lock);
2425 	if (del_timer(&ma->mca_timer)) {
2426 		refcount_dec(&ma->mca_refcnt);
2427 		delay = ma->mca_timer.expires - jiffies;
2428 	}
2429 
2430 	if (!mod_timer(&ma->mca_timer, jiffies + delay))
2431 		refcount_inc(&ma->mca_refcnt);
2432 	ma->mca_flags |= MAF_TIMER_RUNNING | MAF_LAST_REPORTER;
2433 	spin_unlock_bh(&ma->mca_lock);
2434 }
2435 
2436 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
2437 			    struct inet6_dev *idev)
2438 {
2439 	int err;
2440 
2441 	write_lock_bh(&iml->sflock);
2442 	if (!iml->sflist) {
2443 		/* any-source empty exclude case */
2444 		err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode, 0, NULL, 0);
2445 	} else {
2446 		err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode,
2447 				iml->sflist->sl_count, iml->sflist->sl_addr, 0);
2448 		sock_kfree_s(sk, iml->sflist, IP6_SFLSIZE(iml->sflist->sl_max));
2449 		iml->sflist = NULL;
2450 	}
2451 	write_unlock_bh(&iml->sflock);
2452 	return err;
2453 }
2454 
2455 static void igmp6_leave_group(struct ifmcaddr6 *ma)
2456 {
2457 	if (mld_in_v1_mode(ma->idev)) {
2458 		if (ma->mca_flags & MAF_LAST_REPORTER)
2459 			igmp6_send(&ma->mca_addr, ma->idev->dev,
2460 				ICMPV6_MGM_REDUCTION);
2461 	} else {
2462 		mld_add_delrec(ma->idev, ma);
2463 		mld_ifc_event(ma->idev);
2464 	}
2465 }
2466 
2467 static void mld_gq_timer_expire(struct timer_list *t)
2468 {
2469 	struct inet6_dev *idev = from_timer(idev, t, mc_gq_timer);
2470 
2471 	idev->mc_gq_running = 0;
2472 	mld_send_report(idev, NULL);
2473 	in6_dev_put(idev);
2474 }
2475 
2476 static void mld_ifc_timer_expire(struct timer_list *t)
2477 {
2478 	struct inet6_dev *idev = from_timer(idev, t, mc_ifc_timer);
2479 
2480 	mld_send_cr(idev);
2481 	if (idev->mc_ifc_count) {
2482 		idev->mc_ifc_count--;
2483 		if (idev->mc_ifc_count)
2484 			mld_ifc_start_timer(idev,
2485 					    unsolicited_report_interval(idev));
2486 	}
2487 	in6_dev_put(idev);
2488 }
2489 
2490 static void mld_ifc_event(struct inet6_dev *idev)
2491 {
2492 	if (mld_in_v1_mode(idev))
2493 		return;
2494 	idev->mc_ifc_count = idev->mc_qrv;
2495 	mld_ifc_start_timer(idev, 1);
2496 }
2497 
2498 static void igmp6_timer_handler(struct timer_list *t)
2499 {
2500 	struct ifmcaddr6 *ma = from_timer(ma, t, mca_timer);
2501 
2502 	if (mld_in_v1_mode(ma->idev))
2503 		igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2504 	else
2505 		mld_send_report(ma->idev, ma);
2506 
2507 	spin_lock(&ma->mca_lock);
2508 	ma->mca_flags |=  MAF_LAST_REPORTER;
2509 	ma->mca_flags &= ~MAF_TIMER_RUNNING;
2510 	spin_unlock(&ma->mca_lock);
2511 	ma_put(ma);
2512 }
2513 
2514 /* Device changing type */
2515 
2516 void ipv6_mc_unmap(struct inet6_dev *idev)
2517 {
2518 	struct ifmcaddr6 *i;
2519 
2520 	/* Install multicast list, except for all-nodes (already installed) */
2521 
2522 	read_lock_bh(&idev->lock);
2523 	for (i = idev->mc_list; i; i = i->next)
2524 		igmp6_group_dropped(i);
2525 	read_unlock_bh(&idev->lock);
2526 }
2527 
2528 void ipv6_mc_remap(struct inet6_dev *idev)
2529 {
2530 	ipv6_mc_up(idev);
2531 }
2532 
2533 /* Device going down */
2534 
2535 void ipv6_mc_down(struct inet6_dev *idev)
2536 {
2537 	struct ifmcaddr6 *i;
2538 
2539 	/* Withdraw multicast list */
2540 
2541 	read_lock_bh(&idev->lock);
2542 
2543 	for (i = idev->mc_list; i; i = i->next)
2544 		igmp6_group_dropped(i);
2545 
2546 	/* Should stop timer after group drop. or we will
2547 	 * start timer again in mld_ifc_event()
2548 	 */
2549 	mld_ifc_stop_timer(idev);
2550 	mld_gq_stop_timer(idev);
2551 	mld_dad_stop_timer(idev);
2552 	read_unlock_bh(&idev->lock);
2553 }
2554 
2555 static void ipv6_mc_reset(struct inet6_dev *idev)
2556 {
2557 	idev->mc_qrv = sysctl_mld_qrv;
2558 	idev->mc_qi = MLD_QI_DEFAULT;
2559 	idev->mc_qri = MLD_QRI_DEFAULT;
2560 	idev->mc_v1_seen = 0;
2561 	idev->mc_maxdelay = unsolicited_report_interval(idev);
2562 }
2563 
2564 /* Device going up */
2565 
2566 void ipv6_mc_up(struct inet6_dev *idev)
2567 {
2568 	struct ifmcaddr6 *i;
2569 
2570 	/* Install multicast list, except for all-nodes (already installed) */
2571 
2572 	read_lock_bh(&idev->lock);
2573 	ipv6_mc_reset(idev);
2574 	for (i = idev->mc_list; i; i = i->next) {
2575 		mld_del_delrec(idev, i);
2576 		igmp6_group_added(i);
2577 	}
2578 	read_unlock_bh(&idev->lock);
2579 }
2580 
2581 /* IPv6 device initialization. */
2582 
2583 void ipv6_mc_init_dev(struct inet6_dev *idev)
2584 {
2585 	write_lock_bh(&idev->lock);
2586 	spin_lock_init(&idev->mc_lock);
2587 	idev->mc_gq_running = 0;
2588 	timer_setup(&idev->mc_gq_timer, mld_gq_timer_expire, 0);
2589 	idev->mc_tomb = NULL;
2590 	idev->mc_ifc_count = 0;
2591 	timer_setup(&idev->mc_ifc_timer, mld_ifc_timer_expire, 0);
2592 	timer_setup(&idev->mc_dad_timer, mld_dad_timer_expire, 0);
2593 	ipv6_mc_reset(idev);
2594 	write_unlock_bh(&idev->lock);
2595 }
2596 
2597 /*
2598  *	Device is about to be destroyed: clean up.
2599  */
2600 
2601 void ipv6_mc_destroy_dev(struct inet6_dev *idev)
2602 {
2603 	struct ifmcaddr6 *i;
2604 
2605 	/* Deactivate timers */
2606 	ipv6_mc_down(idev);
2607 	mld_clear_delrec(idev);
2608 
2609 	/* Delete all-nodes address. */
2610 	/* We cannot call ipv6_dev_mc_dec() directly, our caller in
2611 	 * addrconf.c has NULL'd out dev->ip6_ptr so in6_dev_get() will
2612 	 * fail.
2613 	 */
2614 	__ipv6_dev_mc_dec(idev, &in6addr_linklocal_allnodes);
2615 
2616 	if (idev->cnf.forwarding)
2617 		__ipv6_dev_mc_dec(idev, &in6addr_linklocal_allrouters);
2618 
2619 	write_lock_bh(&idev->lock);
2620 	while ((i = idev->mc_list) != NULL) {
2621 		idev->mc_list = i->next;
2622 
2623 		write_unlock_bh(&idev->lock);
2624 		ma_put(i);
2625 		write_lock_bh(&idev->lock);
2626 	}
2627 	write_unlock_bh(&idev->lock);
2628 }
2629 
2630 static void ipv6_mc_rejoin_groups(struct inet6_dev *idev)
2631 {
2632 	struct ifmcaddr6 *pmc;
2633 
2634 	ASSERT_RTNL();
2635 
2636 	if (mld_in_v1_mode(idev)) {
2637 		read_lock_bh(&idev->lock);
2638 		for (pmc = idev->mc_list; pmc; pmc = pmc->next)
2639 			igmp6_join_group(pmc);
2640 		read_unlock_bh(&idev->lock);
2641 	} else
2642 		mld_send_report(idev, NULL);
2643 }
2644 
2645 static int ipv6_mc_netdev_event(struct notifier_block *this,
2646 				unsigned long event,
2647 				void *ptr)
2648 {
2649 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2650 	struct inet6_dev *idev = __in6_dev_get(dev);
2651 
2652 	switch (event) {
2653 	case NETDEV_RESEND_IGMP:
2654 		if (idev)
2655 			ipv6_mc_rejoin_groups(idev);
2656 		break;
2657 	default:
2658 		break;
2659 	}
2660 
2661 	return NOTIFY_DONE;
2662 }
2663 
2664 static struct notifier_block igmp6_netdev_notifier = {
2665 	.notifier_call = ipv6_mc_netdev_event,
2666 };
2667 
2668 #ifdef CONFIG_PROC_FS
2669 struct igmp6_mc_iter_state {
2670 	struct seq_net_private p;
2671 	struct net_device *dev;
2672 	struct inet6_dev *idev;
2673 };
2674 
2675 #define igmp6_mc_seq_private(seq)	((struct igmp6_mc_iter_state *)(seq)->private)
2676 
2677 static inline struct ifmcaddr6 *igmp6_mc_get_first(struct seq_file *seq)
2678 {
2679 	struct ifmcaddr6 *im = NULL;
2680 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2681 	struct net *net = seq_file_net(seq);
2682 
2683 	state->idev = NULL;
2684 	for_each_netdev_rcu(net, state->dev) {
2685 		struct inet6_dev *idev;
2686 		idev = __in6_dev_get(state->dev);
2687 		if (!idev)
2688 			continue;
2689 		read_lock_bh(&idev->lock);
2690 		im = idev->mc_list;
2691 		if (im) {
2692 			state->idev = idev;
2693 			break;
2694 		}
2695 		read_unlock_bh(&idev->lock);
2696 	}
2697 	return im;
2698 }
2699 
2700 static struct ifmcaddr6 *igmp6_mc_get_next(struct seq_file *seq, struct ifmcaddr6 *im)
2701 {
2702 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2703 
2704 	im = im->next;
2705 	while (!im) {
2706 		if (likely(state->idev))
2707 			read_unlock_bh(&state->idev->lock);
2708 
2709 		state->dev = next_net_device_rcu(state->dev);
2710 		if (!state->dev) {
2711 			state->idev = NULL;
2712 			break;
2713 		}
2714 		state->idev = __in6_dev_get(state->dev);
2715 		if (!state->idev)
2716 			continue;
2717 		read_lock_bh(&state->idev->lock);
2718 		im = state->idev->mc_list;
2719 	}
2720 	return im;
2721 }
2722 
2723 static struct ifmcaddr6 *igmp6_mc_get_idx(struct seq_file *seq, loff_t pos)
2724 {
2725 	struct ifmcaddr6 *im = igmp6_mc_get_first(seq);
2726 	if (im)
2727 		while (pos && (im = igmp6_mc_get_next(seq, im)) != NULL)
2728 			--pos;
2729 	return pos ? NULL : im;
2730 }
2731 
2732 static void *igmp6_mc_seq_start(struct seq_file *seq, loff_t *pos)
2733 	__acquires(RCU)
2734 {
2735 	rcu_read_lock();
2736 	return igmp6_mc_get_idx(seq, *pos);
2737 }
2738 
2739 static void *igmp6_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2740 {
2741 	struct ifmcaddr6 *im = igmp6_mc_get_next(seq, v);
2742 
2743 	++*pos;
2744 	return im;
2745 }
2746 
2747 static void igmp6_mc_seq_stop(struct seq_file *seq, void *v)
2748 	__releases(RCU)
2749 {
2750 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2751 
2752 	if (likely(state->idev)) {
2753 		read_unlock_bh(&state->idev->lock);
2754 		state->idev = NULL;
2755 	}
2756 	state->dev = NULL;
2757 	rcu_read_unlock();
2758 }
2759 
2760 static int igmp6_mc_seq_show(struct seq_file *seq, void *v)
2761 {
2762 	struct ifmcaddr6 *im = (struct ifmcaddr6 *)v;
2763 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2764 
2765 	seq_printf(seq,
2766 		   "%-4d %-15s %pi6 %5d %08X %ld\n",
2767 		   state->dev->ifindex, state->dev->name,
2768 		   &im->mca_addr,
2769 		   im->mca_users, im->mca_flags,
2770 		   (im->mca_flags&MAF_TIMER_RUNNING) ?
2771 		   jiffies_to_clock_t(im->mca_timer.expires-jiffies) : 0);
2772 	return 0;
2773 }
2774 
2775 static const struct seq_operations igmp6_mc_seq_ops = {
2776 	.start	=	igmp6_mc_seq_start,
2777 	.next	=	igmp6_mc_seq_next,
2778 	.stop	=	igmp6_mc_seq_stop,
2779 	.show	=	igmp6_mc_seq_show,
2780 };
2781 
2782 struct igmp6_mcf_iter_state {
2783 	struct seq_net_private p;
2784 	struct net_device *dev;
2785 	struct inet6_dev *idev;
2786 	struct ifmcaddr6 *im;
2787 };
2788 
2789 #define igmp6_mcf_seq_private(seq)	((struct igmp6_mcf_iter_state *)(seq)->private)
2790 
2791 static inline struct ip6_sf_list *igmp6_mcf_get_first(struct seq_file *seq)
2792 {
2793 	struct ip6_sf_list *psf = NULL;
2794 	struct ifmcaddr6 *im = NULL;
2795 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2796 	struct net *net = seq_file_net(seq);
2797 
2798 	state->idev = NULL;
2799 	state->im = NULL;
2800 	for_each_netdev_rcu(net, state->dev) {
2801 		struct inet6_dev *idev;
2802 		idev = __in6_dev_get(state->dev);
2803 		if (unlikely(idev == NULL))
2804 			continue;
2805 		read_lock_bh(&idev->lock);
2806 		im = idev->mc_list;
2807 		if (likely(im)) {
2808 			spin_lock_bh(&im->mca_lock);
2809 			psf = im->mca_sources;
2810 			if (likely(psf)) {
2811 				state->im = im;
2812 				state->idev = idev;
2813 				break;
2814 			}
2815 			spin_unlock_bh(&im->mca_lock);
2816 		}
2817 		read_unlock_bh(&idev->lock);
2818 	}
2819 	return psf;
2820 }
2821 
2822 static struct ip6_sf_list *igmp6_mcf_get_next(struct seq_file *seq, struct ip6_sf_list *psf)
2823 {
2824 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2825 
2826 	psf = psf->sf_next;
2827 	while (!psf) {
2828 		spin_unlock_bh(&state->im->mca_lock);
2829 		state->im = state->im->next;
2830 		while (!state->im) {
2831 			if (likely(state->idev))
2832 				read_unlock_bh(&state->idev->lock);
2833 
2834 			state->dev = next_net_device_rcu(state->dev);
2835 			if (!state->dev) {
2836 				state->idev = NULL;
2837 				goto out;
2838 			}
2839 			state->idev = __in6_dev_get(state->dev);
2840 			if (!state->idev)
2841 				continue;
2842 			read_lock_bh(&state->idev->lock);
2843 			state->im = state->idev->mc_list;
2844 		}
2845 		if (!state->im)
2846 			break;
2847 		spin_lock_bh(&state->im->mca_lock);
2848 		psf = state->im->mca_sources;
2849 	}
2850 out:
2851 	return psf;
2852 }
2853 
2854 static struct ip6_sf_list *igmp6_mcf_get_idx(struct seq_file *seq, loff_t pos)
2855 {
2856 	struct ip6_sf_list *psf = igmp6_mcf_get_first(seq);
2857 	if (psf)
2858 		while (pos && (psf = igmp6_mcf_get_next(seq, psf)) != NULL)
2859 			--pos;
2860 	return pos ? NULL : psf;
2861 }
2862 
2863 static void *igmp6_mcf_seq_start(struct seq_file *seq, loff_t *pos)
2864 	__acquires(RCU)
2865 {
2866 	rcu_read_lock();
2867 	return *pos ? igmp6_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
2868 }
2869 
2870 static void *igmp6_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2871 {
2872 	struct ip6_sf_list *psf;
2873 	if (v == SEQ_START_TOKEN)
2874 		psf = igmp6_mcf_get_first(seq);
2875 	else
2876 		psf = igmp6_mcf_get_next(seq, v);
2877 	++*pos;
2878 	return psf;
2879 }
2880 
2881 static void igmp6_mcf_seq_stop(struct seq_file *seq, void *v)
2882 	__releases(RCU)
2883 {
2884 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2885 	if (likely(state->im)) {
2886 		spin_unlock_bh(&state->im->mca_lock);
2887 		state->im = NULL;
2888 	}
2889 	if (likely(state->idev)) {
2890 		read_unlock_bh(&state->idev->lock);
2891 		state->idev = NULL;
2892 	}
2893 	state->dev = NULL;
2894 	rcu_read_unlock();
2895 }
2896 
2897 static int igmp6_mcf_seq_show(struct seq_file *seq, void *v)
2898 {
2899 	struct ip6_sf_list *psf = (struct ip6_sf_list *)v;
2900 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
2901 
2902 	if (v == SEQ_START_TOKEN) {
2903 		seq_puts(seq, "Idx Device                Multicast Address                   Source Address    INC    EXC\n");
2904 	} else {
2905 		seq_printf(seq,
2906 			   "%3d %6.6s %pi6 %pi6 %6lu %6lu\n",
2907 			   state->dev->ifindex, state->dev->name,
2908 			   &state->im->mca_addr,
2909 			   &psf->sf_addr,
2910 			   psf->sf_count[MCAST_INCLUDE],
2911 			   psf->sf_count[MCAST_EXCLUDE]);
2912 	}
2913 	return 0;
2914 }
2915 
2916 static const struct seq_operations igmp6_mcf_seq_ops = {
2917 	.start	=	igmp6_mcf_seq_start,
2918 	.next	=	igmp6_mcf_seq_next,
2919 	.stop	=	igmp6_mcf_seq_stop,
2920 	.show	=	igmp6_mcf_seq_show,
2921 };
2922 
2923 static int __net_init igmp6_proc_init(struct net *net)
2924 {
2925 	int err;
2926 
2927 	err = -ENOMEM;
2928 	if (!proc_create_net("igmp6", 0444, net->proc_net, &igmp6_mc_seq_ops,
2929 			sizeof(struct igmp6_mc_iter_state)))
2930 		goto out;
2931 	if (!proc_create_net("mcfilter6", 0444, net->proc_net,
2932 			&igmp6_mcf_seq_ops,
2933 			sizeof(struct igmp6_mcf_iter_state)))
2934 		goto out_proc_net_igmp6;
2935 
2936 	err = 0;
2937 out:
2938 	return err;
2939 
2940 out_proc_net_igmp6:
2941 	remove_proc_entry("igmp6", net->proc_net);
2942 	goto out;
2943 }
2944 
2945 static void __net_exit igmp6_proc_exit(struct net *net)
2946 {
2947 	remove_proc_entry("mcfilter6", net->proc_net);
2948 	remove_proc_entry("igmp6", net->proc_net);
2949 }
2950 #else
2951 static inline int igmp6_proc_init(struct net *net)
2952 {
2953 	return 0;
2954 }
2955 static inline void igmp6_proc_exit(struct net *net)
2956 {
2957 }
2958 #endif
2959 
2960 static int __net_init igmp6_net_init(struct net *net)
2961 {
2962 	int err;
2963 
2964 	err = inet_ctl_sock_create(&net->ipv6.igmp_sk, PF_INET6,
2965 				   SOCK_RAW, IPPROTO_ICMPV6, net);
2966 	if (err < 0) {
2967 		pr_err("Failed to initialize the IGMP6 control socket (err %d)\n",
2968 		       err);
2969 		goto out;
2970 	}
2971 
2972 	inet6_sk(net->ipv6.igmp_sk)->hop_limit = 1;
2973 
2974 	err = inet_ctl_sock_create(&net->ipv6.mc_autojoin_sk, PF_INET6,
2975 				   SOCK_RAW, IPPROTO_ICMPV6, net);
2976 	if (err < 0) {
2977 		pr_err("Failed to initialize the IGMP6 autojoin socket (err %d)\n",
2978 		       err);
2979 		goto out_sock_create;
2980 	}
2981 
2982 	err = igmp6_proc_init(net);
2983 	if (err)
2984 		goto out_sock_create_autojoin;
2985 
2986 	return 0;
2987 
2988 out_sock_create_autojoin:
2989 	inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
2990 out_sock_create:
2991 	inet_ctl_sock_destroy(net->ipv6.igmp_sk);
2992 out:
2993 	return err;
2994 }
2995 
2996 static void __net_exit igmp6_net_exit(struct net *net)
2997 {
2998 	inet_ctl_sock_destroy(net->ipv6.igmp_sk);
2999 	inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
3000 	igmp6_proc_exit(net);
3001 }
3002 
3003 static struct pernet_operations igmp6_net_ops = {
3004 	.init = igmp6_net_init,
3005 	.exit = igmp6_net_exit,
3006 };
3007 
3008 int __init igmp6_init(void)
3009 {
3010 	return register_pernet_subsys(&igmp6_net_ops);
3011 }
3012 
3013 int __init igmp6_late_init(void)
3014 {
3015 	return register_netdevice_notifier(&igmp6_netdev_notifier);
3016 }
3017 
3018 void igmp6_cleanup(void)
3019 {
3020 	unregister_pernet_subsys(&igmp6_net_ops);
3021 }
3022 
3023 void igmp6_late_cleanup(void)
3024 {
3025 	unregister_netdevice_notifier(&igmp6_netdev_notifier);
3026 }
3027