xref: /linux/net/ipv6/mcast.c (revision f7c595c9d9f4cce9ec335f0d3c5d875bb547f9d5)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *	Multicast support for IPv6
4  *	Linux INET6 implementation
5  *
6  *	Authors:
7  *	Pedro Roque		<roque@di.fc.ul.pt>
8  *
9  *	Based on linux/ipv4/igmp.c and linux/ipv4/ip_sockglue.c
10  */
11 
12 /* Changes:
13  *
14  *	yoshfuji	: fix format of router-alert option
15  *	YOSHIFUJI Hideaki @USAGI:
16  *		Fixed source address for MLD message based on
17  *		<draft-ietf-magma-mld-source-05.txt>.
18  *	YOSHIFUJI Hideaki @USAGI:
19  *		- Ignore Queries for invalid addresses.
20  *		- MLD for link-local addresses.
21  *	David L Stevens <dlstevens@us.ibm.com>:
22  *		- MLDv2 support
23  */
24 
25 #include <linux/module.h>
26 #include <linux/errno.h>
27 #include <linux/types.h>
28 #include <linux/string.h>
29 #include <linux/socket.h>
30 #include <linux/sockios.h>
31 #include <linux/jiffies.h>
32 #include <linux/net.h>
33 #include <linux/in.h>
34 #include <linux/in6.h>
35 #include <linux/netdevice.h>
36 #include <linux/if_addr.h>
37 #include <linux/if_arp.h>
38 #include <linux/route.h>
39 #include <linux/rtnetlink.h>
40 #include <linux/init.h>
41 #include <linux/proc_fs.h>
42 #include <linux/seq_file.h>
43 #include <linux/slab.h>
44 #include <linux/pkt_sched.h>
45 #include <net/mld.h>
46 #include <linux/workqueue.h>
47 
48 #include <linux/netfilter.h>
49 #include <linux/netfilter_ipv6.h>
50 
51 #include <net/net_namespace.h>
52 #include <net/netlink.h>
53 #include <net/sock.h>
54 #include <net/snmp.h>
55 
56 #include <net/ipv6.h>
57 #include <net/protocol.h>
58 #include <net/if_inet6.h>
59 #include <net/ndisc.h>
60 #include <net/addrconf.h>
61 #include <net/ip6_route.h>
62 #include <net/inet_common.h>
63 
64 #include <net/ip6_checksum.h>
65 
66 /* Ensure that we have struct in6_addr aligned on 32bit word. */
67 static int __mld2_query_bugs[] __attribute__((__unused__)) = {
68 	BUILD_BUG_ON_ZERO(offsetof(struct mld2_query, mld2q_srcs) % 4),
69 	BUILD_BUG_ON_ZERO(offsetof(struct mld2_report, mld2r_grec) % 4),
70 	BUILD_BUG_ON_ZERO(offsetof(struct mld2_grec, grec_mca) % 4)
71 };
72 
73 static struct workqueue_struct *mld_wq;
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 mld_mca_work(struct work_struct *work);
79 
80 static void mld_ifc_event(struct inet6_dev *idev);
81 static bool mld_in_v1_mode(const struct inet6_dev *idev);
82 static int sf_setstate(struct ifmcaddr6 *pmc);
83 static void sf_markstate(struct ifmcaddr6 *pmc);
84 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc);
85 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
86 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
87 			  int delta);
88 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
89 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
90 			  int delta);
91 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
92 			    struct inet6_dev *idev);
93 static int __ipv6_dev_mc_inc(struct net_device *dev,
94 			     const struct in6_addr *addr, unsigned int mode);
95 
96 #define MLD_QRV_DEFAULT		2
97 /* RFC3810, 9.2. Query Interval */
98 #define MLD_QI_DEFAULT		(125 * HZ)
99 /* RFC3810, 9.3. Query Response Interval */
100 #define MLD_QRI_DEFAULT		(10 * HZ)
101 
102 /* RFC3810, 8.1 Query Version Distinctions */
103 #define MLD_V1_QUERY_LEN	24
104 #define MLD_V2_QUERY_LEN_MIN	28
105 
106 #define IPV6_MLD_MAX_MSF	64
107 
108 int sysctl_mld_max_msf __read_mostly = IPV6_MLD_MAX_MSF;
109 int sysctl_mld_qrv __read_mostly = MLD_QRV_DEFAULT;
110 
111 #define mc_assert_locked(idev)			\
112 	lockdep_assert_held(&(idev)->mc_lock)
113 
114 #define mc_dereference(e, idev) \
115 	rcu_dereference_protected(e, lockdep_is_held(&(idev)->mc_lock))
116 
117 #define sock_dereference(e, sk) \
118 	rcu_dereference_protected(e, lockdep_sock_is_held(sk))
119 
120 #define for_each_pmc_socklock(np, sk, pmc)			\
121 	for (pmc = sock_dereference((np)->ipv6_mc_list, sk);	\
122 	     pmc;						\
123 	     pmc = sock_dereference(pmc->next, sk))
124 
125 #define for_each_pmc_rcu(np, pmc)				\
126 	for (pmc = rcu_dereference((np)->ipv6_mc_list);		\
127 	     pmc;						\
128 	     pmc = rcu_dereference(pmc->next))
129 
130 #define for_each_psf_mclock(mc, psf)				\
131 	for (psf = mc_dereference((mc)->mca_sources, mc->idev);	\
132 	     psf;						\
133 	     psf = mc_dereference(psf->sf_next, mc->idev))
134 
135 #define for_each_psf_rcu(mc, psf)				\
136 	for (psf = rcu_dereference((mc)->mca_sources);		\
137 	     psf;						\
138 	     psf = rcu_dereference(psf->sf_next))
139 
140 #define for_each_psf_tomb(mc, psf)				\
141 	for (psf = mc_dereference((mc)->mca_tomb, mc->idev);	\
142 	     psf;						\
143 	     psf = mc_dereference(psf->sf_next, mc->idev))
144 
145 #define for_each_mc_mclock(idev, mc)				\
146 	for (mc = mc_dereference((idev)->mc_list, idev);	\
147 	     mc;						\
148 	     mc = mc_dereference(mc->next, idev))
149 
150 #define for_each_mc_rcu(idev, mc)				\
151 	for (mc = rcu_dereference((idev)->mc_list);             \
152 	     mc;                                                \
153 	     mc = rcu_dereference(mc->next))
154 
155 #define for_each_mc_tomb(idev, mc)				\
156 	for (mc = mc_dereference((idev)->mc_tomb, idev);	\
157 	     mc;						\
158 	     mc = mc_dereference(mc->next, idev))
159 
160 static int unsolicited_report_interval(struct inet6_dev *idev)
161 {
162 	int iv;
163 
164 	if (mld_in_v1_mode(idev))
165 		iv = READ_ONCE(idev->cnf.mldv1_unsolicited_report_interval);
166 	else
167 		iv = READ_ONCE(idev->cnf.mldv2_unsolicited_report_interval);
168 
169 	return iv > 0 ? iv : 1;
170 }
171 
172 /*
173  *	socket join on multicast group
174  */
175 static int __ipv6_sock_mc_join(struct sock *sk, int ifindex,
176 			       const struct in6_addr *addr, unsigned int mode)
177 {
178 	struct ipv6_pinfo *np = inet6_sk(sk);
179 	struct ipv6_mc_socklist *mc_lst;
180 	struct net *net = sock_net(sk);
181 	struct net_device *dev = NULL;
182 	int err;
183 
184 	if (!ipv6_addr_is_multicast(addr))
185 		return -EINVAL;
186 
187 	for_each_pmc_socklock(np, sk, mc_lst) {
188 		if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
189 		    ipv6_addr_equal(&mc_lst->addr, addr))
190 			return -EADDRINUSE;
191 	}
192 
193 	mc_lst = sock_kmalloc(sk, sizeof(struct ipv6_mc_socklist), GFP_KERNEL);
194 
195 	if (!mc_lst)
196 		return -ENOMEM;
197 
198 	mc_lst->next = NULL;
199 	mc_lst->addr = *addr;
200 
201 	if (ifindex == 0) {
202 		struct rt6_info *rt;
203 
204 		rcu_read_lock();
205 		rt = rt6_lookup(net, addr, NULL, 0, NULL, 0);
206 		if (rt) {
207 			dev = dst_dev(&rt->dst);
208 			dev_hold(dev);
209 			ip6_rt_put(rt);
210 		}
211 		rcu_read_unlock();
212 	} else {
213 		dev = dev_get_by_index(net, ifindex);
214 	}
215 
216 	if (!dev) {
217 		sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
218 		return -ENODEV;
219 	}
220 
221 	mc_lst->ifindex = dev->ifindex;
222 	mc_lst->sfmode = mode;
223 	RCU_INIT_POINTER(mc_lst->sflist, NULL);
224 
225 	/* now add/increase the group membership on the device */
226 	err = __ipv6_dev_mc_inc(dev, addr, mode);
227 
228 	dev_put(dev);
229 
230 	if (err) {
231 		sock_kfree_s(sk, mc_lst, sizeof(*mc_lst));
232 		return err;
233 	}
234 
235 	mc_lst->next = np->ipv6_mc_list;
236 	rcu_assign_pointer(np->ipv6_mc_list, mc_lst);
237 
238 	return 0;
239 }
240 
241 int ipv6_sock_mc_join(struct sock *sk, int ifindex, const struct in6_addr *addr)
242 {
243 	return __ipv6_sock_mc_join(sk, ifindex, addr, MCAST_EXCLUDE);
244 }
245 EXPORT_SYMBOL(ipv6_sock_mc_join);
246 
247 int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
248 			  const struct in6_addr *addr, unsigned int mode)
249 {
250 	return __ipv6_sock_mc_join(sk, ifindex, addr, mode);
251 }
252 
253 /*
254  *	socket leave on multicast group
255  */
256 static void __ipv6_sock_mc_drop(struct sock *sk, struct ipv6_mc_socklist *mc_lst)
257 {
258 	struct net *net = sock_net(sk);
259 	struct net_device *dev;
260 
261 	dev = dev_get_by_index(net, mc_lst->ifindex);
262 	if (dev) {
263 		struct inet6_dev *idev = in6_dev_get(dev);
264 
265 		ip6_mc_leave_src(sk, mc_lst, idev);
266 
267 		if (idev) {
268 			__ipv6_dev_mc_dec(idev, &mc_lst->addr);
269 			in6_dev_put(idev);
270 		}
271 
272 		dev_put(dev);
273 	} else {
274 		ip6_mc_leave_src(sk, mc_lst, NULL);
275 	}
276 
277 	atomic_sub(sizeof(*mc_lst), &sk->sk_omem_alloc);
278 	kfree_rcu(mc_lst, rcu);
279 }
280 
281 int ipv6_sock_mc_drop(struct sock *sk, int ifindex, const struct in6_addr *addr)
282 {
283 	struct ipv6_pinfo *np = inet6_sk(sk);
284 	struct ipv6_mc_socklist __rcu **lnk;
285 	struct ipv6_mc_socklist *mc_lst;
286 
287 	if (!ipv6_addr_is_multicast(addr))
288 		return -EINVAL;
289 
290 	for (lnk = &np->ipv6_mc_list;
291 	     (mc_lst = sock_dereference(*lnk, sk)) != NULL;
292 	      lnk = &mc_lst->next) {
293 		if ((ifindex == 0 || mc_lst->ifindex == ifindex) &&
294 		    ipv6_addr_equal(&mc_lst->addr, addr)) {
295 			*lnk = mc_lst->next;
296 			__ipv6_sock_mc_drop(sk, mc_lst);
297 			return 0;
298 		}
299 	}
300 
301 	return -EADDRNOTAVAIL;
302 }
303 EXPORT_SYMBOL(ipv6_sock_mc_drop);
304 
305 static struct inet6_dev *ip6_mc_find_dev(struct net *net,
306 					 const struct in6_addr *group,
307 					 int ifindex)
308 {
309 	struct net_device *dev = NULL;
310 	struct inet6_dev *idev;
311 
312 	if (ifindex == 0) {
313 		struct rt6_info *rt;
314 
315 		rcu_read_lock();
316 		rt = rt6_lookup(net, group, NULL, 0, NULL, 0);
317 		if (rt) {
318 			dev = dst_dev(&rt->dst);
319 			dev_hold(dev);
320 			ip6_rt_put(rt);
321 		}
322 		rcu_read_unlock();
323 	} else {
324 		dev = dev_get_by_index(net, ifindex);
325 	}
326 	if (!dev)
327 		return NULL;
328 
329 	idev = in6_dev_get(dev);
330 	dev_put(dev);
331 
332 	if (!idev)
333 		return NULL;
334 
335 	return idev;
336 }
337 
338 void __ipv6_sock_mc_close(struct sock *sk)
339 {
340 	struct ipv6_pinfo *np = inet6_sk(sk);
341 	struct ipv6_mc_socklist *mc_lst;
342 
343 	while ((mc_lst = sock_dereference(np->ipv6_mc_list, sk)) != NULL) {
344 		np->ipv6_mc_list = mc_lst->next;
345 		__ipv6_sock_mc_drop(sk, mc_lst);
346 	}
347 }
348 
349 void ipv6_sock_mc_close(struct sock *sk)
350 {
351 	struct ipv6_pinfo *np = inet6_sk(sk);
352 
353 	if (!rcu_access_pointer(np->ipv6_mc_list))
354 		return;
355 
356 	lock_sock(sk);
357 	__ipv6_sock_mc_close(sk);
358 	release_sock(sk);
359 }
360 
361 int ip6_mc_source(int add, int omode, struct sock *sk,
362 		  struct group_source_req *pgsr)
363 {
364 	struct ipv6_pinfo *inet6 = inet6_sk(sk);
365 	struct in6_addr *source, *group;
366 	struct net *net = sock_net(sk);
367 	struct ipv6_mc_socklist *pmc;
368 	struct ip6_sf_socklist *psl;
369 	struct inet6_dev *idev;
370 	int leavegroup = 0;
371 	int i, j, rv;
372 	int err;
373 
374 	source = &((struct sockaddr_in6 *)&pgsr->gsr_source)->sin6_addr;
375 	group = &((struct sockaddr_in6 *)&pgsr->gsr_group)->sin6_addr;
376 
377 	if (!ipv6_addr_is_multicast(group))
378 		return -EINVAL;
379 
380 	idev = ip6_mc_find_dev(net, group, pgsr->gsr_interface);
381 	if (!idev)
382 		return -ENODEV;
383 
384 	mutex_lock(&idev->mc_lock);
385 
386 	if (idev->dead) {
387 		err = -ENODEV;
388 		goto done;
389 	}
390 
391 	err = -EADDRNOTAVAIL;
392 
393 	for_each_pmc_socklock(inet6, sk, pmc) {
394 		if (pgsr->gsr_interface && pmc->ifindex != pgsr->gsr_interface)
395 			continue;
396 		if (ipv6_addr_equal(&pmc->addr, group))
397 			break;
398 	}
399 	if (!pmc) {		/* must have a prior join */
400 		err = -EINVAL;
401 		goto done;
402 	}
403 	/* if a source filter was set, must be the same mode as before */
404 	if (rcu_access_pointer(pmc->sflist)) {
405 		if (pmc->sfmode != omode) {
406 			err = -EINVAL;
407 			goto done;
408 		}
409 	} else if (pmc->sfmode != omode) {
410 		/* allow mode switches for empty-set filters */
411 		ip6_mc_add_src(idev, group, omode, 0, NULL, 0);
412 		ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
413 		pmc->sfmode = omode;
414 	}
415 
416 	psl = sock_dereference(pmc->sflist, sk);
417 	if (!add) {
418 		if (!psl)
419 			goto done;	/* err = -EADDRNOTAVAIL */
420 		rv = !0;
421 		for (i = 0; i < psl->sl_count; i++) {
422 			rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
423 			if (rv == 0)
424 				break;
425 		}
426 		if (rv)		/* source not found */
427 			goto done;	/* err = -EADDRNOTAVAIL */
428 
429 		/* special case - (INCLUDE, empty) == LEAVE_GROUP */
430 		if (psl->sl_count == 1 && omode == MCAST_INCLUDE) {
431 			leavegroup = 1;
432 			goto done;
433 		}
434 
435 		/* update the interface filter */
436 		ip6_mc_del_src(idev, group, omode, 1, source, 1);
437 
438 		for (j = i+1; j < psl->sl_count; j++)
439 			psl->sl_addr[j-1] = psl->sl_addr[j];
440 		psl->sl_count--;
441 		err = 0;
442 		goto done;
443 	}
444 	/* else, add a new source to the filter */
445 
446 	if (psl && psl->sl_count >= sysctl_mld_max_msf) {
447 		err = -ENOBUFS;
448 		goto done;
449 	}
450 	if (!psl || psl->sl_count == psl->sl_max) {
451 		struct ip6_sf_socklist *newpsl;
452 		int count = IP6_SFBLOCK;
453 
454 		if (psl)
455 			count += psl->sl_max;
456 		newpsl = sock_kmalloc(sk, struct_size(newpsl, sl_addr, count),
457 				      GFP_KERNEL);
458 		if (!newpsl) {
459 			err = -ENOBUFS;
460 			goto done;
461 		}
462 		newpsl->sl_max = count;
463 		newpsl->sl_count = count - IP6_SFBLOCK;
464 		if (psl) {
465 			for (i = 0; i < psl->sl_count; i++)
466 				newpsl->sl_addr[i] = psl->sl_addr[i];
467 			atomic_sub(struct_size(psl, sl_addr, psl->sl_max),
468 				   &sk->sk_omem_alloc);
469 		}
470 		rcu_assign_pointer(pmc->sflist, newpsl);
471 		kfree_rcu(psl, rcu);
472 		psl = newpsl;
473 	}
474 	rv = 1;	/* > 0 for insert logic below if sl_count is 0 */
475 	for (i = 0; i < psl->sl_count; i++) {
476 		rv = !ipv6_addr_equal(&psl->sl_addr[i], source);
477 		if (rv == 0) /* There is an error in the address. */
478 			goto done;
479 	}
480 	for (j = psl->sl_count-1; j >= i; j--)
481 		psl->sl_addr[j+1] = psl->sl_addr[j];
482 	psl->sl_addr[i] = *source;
483 	psl->sl_count++;
484 	err = 0;
485 	/* update the interface list */
486 	ip6_mc_add_src(idev, group, omode, 1, source, 1);
487 done:
488 	mutex_unlock(&idev->mc_lock);
489 	in6_dev_put(idev);
490 	if (leavegroup)
491 		err = ipv6_sock_mc_drop(sk, pgsr->gsr_interface, group);
492 	return err;
493 }
494 
495 int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf,
496 		    struct sockaddr_storage *list)
497 {
498 	struct ipv6_pinfo *inet6 = inet6_sk(sk);
499 	struct ip6_sf_socklist *newpsl, *psl;
500 	struct net *net = sock_net(sk);
501 	const struct in6_addr *group;
502 	struct ipv6_mc_socklist *pmc;
503 	struct inet6_dev *idev;
504 	int leavegroup = 0;
505 	int i, err;
506 
507 	group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
508 
509 	if (!ipv6_addr_is_multicast(group))
510 		return -EINVAL;
511 	if (gsf->gf_fmode != MCAST_INCLUDE &&
512 	    gsf->gf_fmode != MCAST_EXCLUDE)
513 		return -EINVAL;
514 
515 	idev = ip6_mc_find_dev(net, group, gsf->gf_interface);
516 	if (!idev)
517 		return -ENODEV;
518 
519 	mutex_lock(&idev->mc_lock);
520 
521 	if (idev->dead) {
522 		err = -ENODEV;
523 		goto done;
524 	}
525 
526 	err = 0;
527 
528 	if (gsf->gf_fmode == MCAST_INCLUDE && gsf->gf_numsrc == 0) {
529 		leavegroup = 1;
530 		goto done;
531 	}
532 
533 	for_each_pmc_socklock(inet6, sk, pmc) {
534 		if (pmc->ifindex != gsf->gf_interface)
535 			continue;
536 		if (ipv6_addr_equal(&pmc->addr, group))
537 			break;
538 	}
539 	if (!pmc) {		/* must have a prior join */
540 		err = -EINVAL;
541 		goto done;
542 	}
543 	if (gsf->gf_numsrc) {
544 		newpsl = sock_kmalloc(sk, struct_size(newpsl, sl_addr,
545 						      gsf->gf_numsrc),
546 				      GFP_KERNEL);
547 		if (!newpsl) {
548 			err = -ENOBUFS;
549 			goto done;
550 		}
551 		newpsl->sl_max = newpsl->sl_count = gsf->gf_numsrc;
552 		for (i = 0; i < newpsl->sl_count; ++i, ++list) {
553 			struct sockaddr_in6 *psin6;
554 
555 			psin6 = (struct sockaddr_in6 *)list;
556 			newpsl->sl_addr[i] = psin6->sin6_addr;
557 		}
558 
559 		err = ip6_mc_add_src(idev, group, gsf->gf_fmode,
560 				     newpsl->sl_count, newpsl->sl_addr, 0);
561 		if (err) {
562 			sock_kfree_s(sk, newpsl, struct_size(newpsl, sl_addr,
563 							     newpsl->sl_max));
564 			goto done;
565 		}
566 	} else {
567 		newpsl = NULL;
568 		ip6_mc_add_src(idev, group, gsf->gf_fmode, 0, NULL, 0);
569 	}
570 
571 	psl = sock_dereference(pmc->sflist, sk);
572 	if (psl) {
573 		ip6_mc_del_src(idev, group, pmc->sfmode,
574 			       psl->sl_count, psl->sl_addr, 0);
575 		atomic_sub(struct_size(psl, sl_addr, psl->sl_max),
576 			   &sk->sk_omem_alloc);
577 	} else {
578 		ip6_mc_del_src(idev, group, pmc->sfmode, 0, NULL, 0);
579 	}
580 
581 	rcu_assign_pointer(pmc->sflist, newpsl);
582 	kfree_rcu(psl, rcu);
583 	pmc->sfmode = gsf->gf_fmode;
584 	err = 0;
585 done:
586 	mutex_unlock(&idev->mc_lock);
587 	in6_dev_put(idev);
588 	if (leavegroup)
589 		err = ipv6_sock_mc_drop(sk, gsf->gf_interface, group);
590 	return err;
591 }
592 
593 int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
594 		  sockptr_t optval, size_t ss_offset)
595 {
596 	struct ipv6_pinfo *inet6 = inet6_sk(sk);
597 	const struct in6_addr *group;
598 	struct ipv6_mc_socklist *pmc;
599 	struct ip6_sf_socklist *psl;
600 	unsigned int count;
601 	int i, copycount;
602 
603 	group = &((struct sockaddr_in6 *)&gsf->gf_group)->sin6_addr;
604 
605 	if (!ipv6_addr_is_multicast(group))
606 		return -EINVAL;
607 
608 	for_each_pmc_socklock(inet6, sk, pmc) {
609 		if (pmc->ifindex != gsf->gf_interface)
610 			continue;
611 		if (ipv6_addr_equal(group, &pmc->addr))
612 			break;
613 	}
614 	if (!pmc)		/* must have a prior join */
615 		return -EADDRNOTAVAIL;
616 
617 	gsf->gf_fmode = pmc->sfmode;
618 	psl = sock_dereference(pmc->sflist, sk);
619 	count = psl ? psl->sl_count : 0;
620 
621 	copycount = min(count, gsf->gf_numsrc);
622 	gsf->gf_numsrc = count;
623 	for (i = 0; i < copycount; i++) {
624 		struct sockaddr_in6 *psin6;
625 		struct sockaddr_storage ss;
626 
627 		psin6 = (struct sockaddr_in6 *)&ss;
628 		memset(&ss, 0, sizeof(ss));
629 		psin6->sin6_family = AF_INET6;
630 		psin6->sin6_addr = psl->sl_addr[i];
631 		if (copy_to_sockptr_offset(optval, ss_offset, &ss, sizeof(ss)))
632 			return -EFAULT;
633 		ss_offset += sizeof(ss);
634 	}
635 	return 0;
636 }
637 
638 bool inet6_mc_check(const struct sock *sk, const struct in6_addr *mc_addr,
639 		    const struct in6_addr *src_addr)
640 {
641 	const struct ipv6_pinfo *np = inet6_sk(sk);
642 	const struct ipv6_mc_socklist *mc;
643 	const struct ip6_sf_socklist *psl;
644 	bool rv = true;
645 
646 	rcu_read_lock();
647 	for_each_pmc_rcu(np, mc) {
648 		if (ipv6_addr_equal(&mc->addr, mc_addr))
649 			break;
650 	}
651 	if (!mc) {
652 		rcu_read_unlock();
653 		return inet6_test_bit(MC6_ALL, sk);
654 	}
655 	psl = rcu_dereference(mc->sflist);
656 	if (!psl) {
657 		rv = mc->sfmode == MCAST_EXCLUDE;
658 	} else {
659 		int i;
660 
661 		for (i = 0; i < psl->sl_count; i++) {
662 			if (ipv6_addr_equal(&psl->sl_addr[i], src_addr))
663 				break;
664 		}
665 		if (mc->sfmode == MCAST_INCLUDE && i >= psl->sl_count)
666 			rv = false;
667 		if (mc->sfmode == MCAST_EXCLUDE && i < psl->sl_count)
668 			rv = false;
669 	}
670 	rcu_read_unlock();
671 
672 	return rv;
673 }
674 
675 static void igmp6_group_added(struct ifmcaddr6 *mc)
676 {
677 	struct net_device *dev = mc->idev->dev;
678 	char buf[MAX_ADDR_LEN];
679 
680 	mc_assert_locked(mc->idev);
681 
682 	if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
683 	    IPV6_ADDR_SCOPE_LINKLOCAL)
684 		return;
685 
686 	if (!(mc->mca_flags&MAF_LOADED)) {
687 		mc->mca_flags |= MAF_LOADED;
688 		if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
689 			dev_mc_add(dev, buf);
690 	}
691 
692 	if (!(dev->flags & IFF_UP) || (mc->mca_flags & MAF_NOREPORT))
693 		return;
694 
695 	if (mld_in_v1_mode(mc->idev)) {
696 		igmp6_join_group(mc);
697 		return;
698 	}
699 	/* else v2 */
700 
701 	/* Based on RFC3810 6.1, for newly added INCLUDE SSM, we
702 	 * should not send filter-mode change record as the mode
703 	 * should be from IN() to IN(A).
704 	 */
705 	if (mc->mca_sfmode == MCAST_EXCLUDE)
706 		mc->mca_crcount = mc->idev->mc_qrv;
707 
708 	mld_ifc_event(mc->idev);
709 }
710 
711 static void igmp6_group_dropped(struct ifmcaddr6 *mc)
712 {
713 	struct net_device *dev = mc->idev->dev;
714 	char buf[MAX_ADDR_LEN];
715 
716 	mc_assert_locked(mc->idev);
717 
718 	if (IPV6_ADDR_MC_SCOPE(&mc->mca_addr) <
719 	    IPV6_ADDR_SCOPE_LINKLOCAL)
720 		return;
721 
722 	if (mc->mca_flags&MAF_LOADED) {
723 		mc->mca_flags &= ~MAF_LOADED;
724 		if (ndisc_mc_map(&mc->mca_addr, buf, dev, 0) == 0)
725 			dev_mc_del(dev, buf);
726 	}
727 
728 	if (mc->mca_flags & MAF_NOREPORT)
729 		return;
730 
731 	if (!mc->idev->dead)
732 		igmp6_leave_group(mc);
733 
734 	if (cancel_delayed_work(&mc->mca_work))
735 		refcount_dec(&mc->mca_refcnt);
736 }
737 
738 /* deleted ifmcaddr6 manipulation */
739 static void mld_add_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
740 {
741 	struct ifmcaddr6 *pmc;
742 
743 	mc_assert_locked(idev);
744 
745 	/* this is an "ifmcaddr6" for convenience; only the fields below
746 	 * are actually used. In particular, the refcnt and users are not
747 	 * used for management of the delete list. Using the same structure
748 	 * for deleted items allows change reports to use common code with
749 	 * non-deleted or query-response MCA's.
750 	 */
751 	pmc = kzalloc(sizeof(*pmc), GFP_KERNEL);
752 	if (!pmc)
753 		return;
754 
755 	pmc->idev = im->idev;
756 	in6_dev_hold(idev);
757 	pmc->mca_addr = im->mca_addr;
758 	pmc->mca_crcount = idev->mc_qrv;
759 	pmc->mca_sfmode = im->mca_sfmode;
760 	if (pmc->mca_sfmode == MCAST_INCLUDE) {
761 		struct ip6_sf_list *psf;
762 
763 		rcu_assign_pointer(pmc->mca_tomb,
764 				   mc_dereference(im->mca_tomb, idev));
765 		rcu_assign_pointer(pmc->mca_sources,
766 				   mc_dereference(im->mca_sources, idev));
767 		RCU_INIT_POINTER(im->mca_tomb, NULL);
768 		RCU_INIT_POINTER(im->mca_sources, NULL);
769 
770 		for_each_psf_mclock(pmc, psf)
771 			psf->sf_crcount = pmc->mca_crcount;
772 	}
773 
774 	rcu_assign_pointer(pmc->next, idev->mc_tomb);
775 	rcu_assign_pointer(idev->mc_tomb, pmc);
776 }
777 
778 static void mld_del_delrec(struct inet6_dev *idev, struct ifmcaddr6 *im)
779 {
780 	struct ip6_sf_list *psf, *sources, *tomb;
781 	struct in6_addr *pmca = &im->mca_addr;
782 	struct ifmcaddr6 *pmc, *pmc_prev;
783 
784 	mc_assert_locked(idev);
785 
786 	pmc_prev = NULL;
787 	for_each_mc_tomb(idev, pmc) {
788 		if (ipv6_addr_equal(&pmc->mca_addr, pmca))
789 			break;
790 		pmc_prev = pmc;
791 	}
792 	if (pmc) {
793 		if (pmc_prev)
794 			rcu_assign_pointer(pmc_prev->next, pmc->next);
795 		else
796 			rcu_assign_pointer(idev->mc_tomb, pmc->next);
797 	}
798 
799 	if (pmc) {
800 		im->idev = pmc->idev;
801 		if (im->mca_sfmode == MCAST_INCLUDE) {
802 			tomb = rcu_replace_pointer(im->mca_tomb,
803 						   mc_dereference(pmc->mca_tomb, pmc->idev),
804 						   lockdep_is_held(&im->idev->mc_lock));
805 			rcu_assign_pointer(pmc->mca_tomb, tomb);
806 
807 			sources = rcu_replace_pointer(im->mca_sources,
808 						      mc_dereference(pmc->mca_sources, pmc->idev),
809 						      lockdep_is_held(&im->idev->mc_lock));
810 			rcu_assign_pointer(pmc->mca_sources, sources);
811 			for_each_psf_mclock(im, psf)
812 				psf->sf_crcount = idev->mc_qrv;
813 		} else {
814 			im->mca_crcount = idev->mc_qrv;
815 		}
816 		in6_dev_put(pmc->idev);
817 		ip6_mc_clear_src(pmc);
818 		kfree_rcu(pmc, rcu);
819 	}
820 }
821 
822 static void mld_clear_delrec(struct inet6_dev *idev)
823 {
824 	struct ifmcaddr6 *pmc, *nextpmc;
825 
826 	mc_assert_locked(idev);
827 
828 	pmc = mc_dereference(idev->mc_tomb, idev);
829 	RCU_INIT_POINTER(idev->mc_tomb, NULL);
830 
831 	for (; pmc; pmc = nextpmc) {
832 		nextpmc = mc_dereference(pmc->next, idev);
833 		ip6_mc_clear_src(pmc);
834 		in6_dev_put(pmc->idev);
835 		kfree_rcu(pmc, rcu);
836 	}
837 
838 	/* clear dead sources, too */
839 	for_each_mc_mclock(idev, pmc) {
840 		struct ip6_sf_list *psf, *psf_next;
841 
842 		psf = mc_dereference(pmc->mca_tomb, idev);
843 		RCU_INIT_POINTER(pmc->mca_tomb, NULL);
844 		for (; psf; psf = psf_next) {
845 			psf_next = mc_dereference(psf->sf_next, idev);
846 			kfree_rcu(psf, rcu);
847 		}
848 	}
849 }
850 
851 static void mld_clear_query(struct inet6_dev *idev)
852 {
853 	struct sk_buff *skb;
854 
855 	spin_lock_bh(&idev->mc_query_lock);
856 	while ((skb = __skb_dequeue(&idev->mc_query_queue)))
857 		kfree_skb(skb);
858 	spin_unlock_bh(&idev->mc_query_lock);
859 }
860 
861 static void mld_clear_report(struct inet6_dev *idev)
862 {
863 	struct sk_buff *skb;
864 
865 	spin_lock_bh(&idev->mc_report_lock);
866 	while ((skb = __skb_dequeue(&idev->mc_report_queue)))
867 		kfree_skb(skb);
868 	spin_unlock_bh(&idev->mc_report_lock);
869 }
870 
871 static void ma_put(struct ifmcaddr6 *mc)
872 {
873 	if (refcount_dec_and_test(&mc->mca_refcnt)) {
874 		in6_dev_put(mc->idev);
875 		kfree_rcu(mc, rcu);
876 	}
877 }
878 
879 static struct ifmcaddr6 *mca_alloc(struct inet6_dev *idev,
880 				   const struct in6_addr *addr,
881 				   unsigned int mode)
882 {
883 	struct ifmcaddr6 *mc;
884 
885 	mc_assert_locked(idev);
886 
887 	mc = kzalloc(sizeof(*mc), GFP_KERNEL);
888 	if (!mc)
889 		return NULL;
890 
891 	INIT_DELAYED_WORK(&mc->mca_work, mld_mca_work);
892 
893 	mc->mca_addr = *addr;
894 	mc->idev = idev; /* reference taken by caller */
895 	mc->mca_users = 1;
896 	/* mca_stamp should be updated upon changes */
897 	mc->mca_cstamp = mc->mca_tstamp = jiffies;
898 	refcount_set(&mc->mca_refcnt, 1);
899 
900 	mc->mca_sfmode = mode;
901 	mc->mca_sfcount[mode] = 1;
902 
903 	if (ipv6_addr_is_ll_all_nodes(&mc->mca_addr) ||
904 	    IPV6_ADDR_MC_SCOPE(&mc->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
905 		mc->mca_flags |= MAF_NOREPORT;
906 
907 	return mc;
908 }
909 
910 static void inet6_ifmcaddr_notify(struct net_device *dev,
911 				  const struct ifmcaddr6 *ifmca, int event)
912 {
913 	struct inet6_fill_args fillargs = {
914 		.portid = 0,
915 		.seq = 0,
916 		.event = event,
917 		.flags = 0,
918 		.netnsid = -1,
919 		.force_rt_scope_universe = true,
920 	};
921 	struct net *net = dev_net(dev);
922 	struct sk_buff *skb;
923 	int err = -ENOMEM;
924 
925 	skb = nlmsg_new(NLMSG_ALIGN(sizeof(struct ifaddrmsg)) +
926 			nla_total_size(sizeof(struct in6_addr)) +
927 			nla_total_size(sizeof(struct ifa_cacheinfo)),
928 			GFP_KERNEL);
929 	if (!skb)
930 		goto error;
931 
932 	err = inet6_fill_ifmcaddr(skb, ifmca, &fillargs);
933 	if (err < 0) {
934 		WARN_ON_ONCE(err == -EMSGSIZE);
935 		nlmsg_free(skb);
936 		goto error;
937 	}
938 
939 	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_MCADDR, NULL, GFP_KERNEL);
940 	return;
941 error:
942 	rtnl_set_sk_err(net, RTNLGRP_IPV6_MCADDR, err);
943 }
944 
945 /*
946  *	device multicast group inc (add if not found)
947  */
948 static int __ipv6_dev_mc_inc(struct net_device *dev,
949 			     const struct in6_addr *addr, unsigned int mode)
950 {
951 	struct inet6_dev *idev;
952 	struct ifmcaddr6 *mc;
953 
954 	/* we need to take a reference on idev */
955 	idev = in6_dev_get(dev);
956 	if (!idev)
957 		return -EINVAL;
958 
959 	mutex_lock(&idev->mc_lock);
960 
961 	if (READ_ONCE(idev->dead)) {
962 		mutex_unlock(&idev->mc_lock);
963 		in6_dev_put(idev);
964 		return -ENODEV;
965 	}
966 
967 	for_each_mc_mclock(idev, mc) {
968 		if (ipv6_addr_equal(&mc->mca_addr, addr)) {
969 			mc->mca_users++;
970 			ip6_mc_add_src(idev, &mc->mca_addr, mode, 0, NULL, 0);
971 			mutex_unlock(&idev->mc_lock);
972 			in6_dev_put(idev);
973 			return 0;
974 		}
975 	}
976 
977 	mc = mca_alloc(idev, addr, mode);
978 	if (!mc) {
979 		mutex_unlock(&idev->mc_lock);
980 		in6_dev_put(idev);
981 		return -ENOMEM;
982 	}
983 
984 	rcu_assign_pointer(mc->next, idev->mc_list);
985 	rcu_assign_pointer(idev->mc_list, mc);
986 
987 	mld_del_delrec(idev, mc);
988 	igmp6_group_added(mc);
989 	inet6_ifmcaddr_notify(dev, mc, RTM_NEWMULTICAST);
990 	mutex_unlock(&idev->mc_lock);
991 
992 	return 0;
993 }
994 
995 int ipv6_dev_mc_inc(struct net_device *dev, const struct in6_addr *addr)
996 {
997 	return __ipv6_dev_mc_inc(dev, addr, MCAST_EXCLUDE);
998 }
999 EXPORT_SYMBOL(ipv6_dev_mc_inc);
1000 
1001 /*
1002  * device multicast group del
1003  */
1004 int __ipv6_dev_mc_dec(struct inet6_dev *idev, const struct in6_addr *addr)
1005 {
1006 	struct ifmcaddr6 *ma, __rcu **map;
1007 
1008 	mutex_lock(&idev->mc_lock);
1009 
1010 	for (map = &idev->mc_list;
1011 	     (ma = mc_dereference(*map, idev));
1012 	     map = &ma->next) {
1013 		if (ipv6_addr_equal(&ma->mca_addr, addr)) {
1014 			if (--ma->mca_users == 0) {
1015 				*map = ma->next;
1016 
1017 				igmp6_group_dropped(ma);
1018 				inet6_ifmcaddr_notify(idev->dev, ma,
1019 						      RTM_DELMULTICAST);
1020 				ip6_mc_clear_src(ma);
1021 				mutex_unlock(&idev->mc_lock);
1022 
1023 				ma_put(ma);
1024 				return 0;
1025 			}
1026 			mutex_unlock(&idev->mc_lock);
1027 			return 0;
1028 		}
1029 	}
1030 
1031 	mutex_unlock(&idev->mc_lock);
1032 	return -ENOENT;
1033 }
1034 
1035 int ipv6_dev_mc_dec(struct net_device *dev, const struct in6_addr *addr)
1036 {
1037 	struct inet6_dev *idev;
1038 	int err;
1039 
1040 	idev = in6_dev_get(dev);
1041 	if (!idev)
1042 		return -ENODEV;
1043 
1044 	err = __ipv6_dev_mc_dec(idev, addr);
1045 	in6_dev_put(idev);
1046 
1047 	return err;
1048 }
1049 EXPORT_SYMBOL(ipv6_dev_mc_dec);
1050 
1051 /*
1052  *	check if the interface/address pair is valid
1053  */
1054 bool ipv6_chk_mcast_addr(struct net_device *dev, const struct in6_addr *group,
1055 			 const struct in6_addr *src_addr)
1056 {
1057 	struct inet6_dev *idev;
1058 	struct ifmcaddr6 *mc;
1059 	bool rv = false;
1060 
1061 	rcu_read_lock();
1062 	idev = __in6_dev_get(dev);
1063 	if (!idev)
1064 		goto unlock;
1065 	for_each_mc_rcu(idev, mc) {
1066 		if (ipv6_addr_equal(&mc->mca_addr, group))
1067 			break;
1068 	}
1069 	if (!mc)
1070 		goto unlock;
1071 	if (src_addr && !ipv6_addr_any(src_addr)) {
1072 		struct ip6_sf_list *psf;
1073 
1074 		for_each_psf_rcu(mc, psf) {
1075 			if (ipv6_addr_equal(&psf->sf_addr, src_addr))
1076 				break;
1077 		}
1078 		if (psf)
1079 			rv = READ_ONCE(psf->sf_count[MCAST_INCLUDE]) ||
1080 				READ_ONCE(psf->sf_count[MCAST_EXCLUDE]) !=
1081 				READ_ONCE(mc->mca_sfcount[MCAST_EXCLUDE]);
1082 		else
1083 			rv = READ_ONCE(mc->mca_sfcount[MCAST_EXCLUDE]) != 0;
1084 	} else {
1085 		rv = true; /* don't filter unspecified source */
1086 	}
1087 unlock:
1088 	rcu_read_unlock();
1089 	return rv;
1090 }
1091 
1092 static void mld_gq_start_work(struct inet6_dev *idev)
1093 {
1094 	unsigned long tv = get_random_u32_below(idev->mc_maxdelay);
1095 
1096 	mc_assert_locked(idev);
1097 
1098 	idev->mc_gq_running = 1;
1099 	if (!mod_delayed_work(mld_wq, &idev->mc_gq_work, tv + 2))
1100 		in6_dev_hold(idev);
1101 }
1102 
1103 static void mld_gq_stop_work(struct inet6_dev *idev)
1104 {
1105 	mc_assert_locked(idev);
1106 
1107 	idev->mc_gq_running = 0;
1108 	if (cancel_delayed_work(&idev->mc_gq_work))
1109 		__in6_dev_put(idev);
1110 }
1111 
1112 static void mld_ifc_start_work(struct inet6_dev *idev, unsigned long delay)
1113 {
1114 	unsigned long tv = get_random_u32_below(delay);
1115 
1116 	mc_assert_locked(idev);
1117 
1118 	if (!mod_delayed_work(mld_wq, &idev->mc_ifc_work, tv + 2))
1119 		in6_dev_hold(idev);
1120 }
1121 
1122 static void mld_ifc_stop_work(struct inet6_dev *idev)
1123 {
1124 	mc_assert_locked(idev);
1125 
1126 	idev->mc_ifc_count = 0;
1127 	if (cancel_delayed_work(&idev->mc_ifc_work))
1128 		__in6_dev_put(idev);
1129 }
1130 
1131 static void mld_dad_start_work(struct inet6_dev *idev, unsigned long delay)
1132 {
1133 	unsigned long tv = get_random_u32_below(delay);
1134 
1135 	mc_assert_locked(idev);
1136 
1137 	if (!mod_delayed_work(mld_wq, &idev->mc_dad_work, tv + 2))
1138 		in6_dev_hold(idev);
1139 }
1140 
1141 static void mld_dad_stop_work(struct inet6_dev *idev)
1142 {
1143 	if (cancel_delayed_work(&idev->mc_dad_work))
1144 		__in6_dev_put(idev);
1145 }
1146 
1147 static void mld_query_stop_work(struct inet6_dev *idev)
1148 {
1149 	spin_lock_bh(&idev->mc_query_lock);
1150 	if (cancel_delayed_work(&idev->mc_query_work))
1151 		__in6_dev_put(idev);
1152 	spin_unlock_bh(&idev->mc_query_lock);
1153 }
1154 
1155 static void mld_report_stop_work(struct inet6_dev *idev)
1156 {
1157 	if (cancel_delayed_work_sync(&idev->mc_report_work))
1158 		__in6_dev_put(idev);
1159 }
1160 
1161 /* IGMP handling (alias multicast ICMPv6 messages) */
1162 static void igmp6_group_queried(struct ifmcaddr6 *ma, unsigned long resptime)
1163 {
1164 	unsigned long delay = resptime;
1165 
1166 	mc_assert_locked(ma->idev);
1167 
1168 	/* Do not start work for these addresses */
1169 	if (ipv6_addr_is_ll_all_nodes(&ma->mca_addr) ||
1170 	    IPV6_ADDR_MC_SCOPE(&ma->mca_addr) < IPV6_ADDR_SCOPE_LINKLOCAL)
1171 		return;
1172 
1173 	if (cancel_delayed_work(&ma->mca_work)) {
1174 		refcount_dec(&ma->mca_refcnt);
1175 		delay = ma->mca_work.timer.expires - jiffies;
1176 	}
1177 
1178 	if (delay >= resptime)
1179 		delay = get_random_u32_below(resptime);
1180 
1181 	if (!mod_delayed_work(mld_wq, &ma->mca_work, delay))
1182 		refcount_inc(&ma->mca_refcnt);
1183 	ma->mca_flags |= MAF_TIMER_RUNNING;
1184 }
1185 
1186 /* mark EXCLUDE-mode sources */
1187 static bool mld_xmarksources(struct ifmcaddr6 *pmc, int nsrcs,
1188 			     const struct in6_addr *srcs)
1189 {
1190 	struct ip6_sf_list *psf;
1191 	int i, scount;
1192 
1193 	mc_assert_locked(pmc->idev);
1194 
1195 	scount = 0;
1196 	for_each_psf_mclock(pmc, psf) {
1197 		if (scount == nsrcs)
1198 			break;
1199 		for (i = 0; i < nsrcs; i++) {
1200 			/* skip inactive filters */
1201 			if (psf->sf_count[MCAST_INCLUDE] ||
1202 			    pmc->mca_sfcount[MCAST_EXCLUDE] !=
1203 			    psf->sf_count[MCAST_EXCLUDE])
1204 				break;
1205 			if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1206 				scount++;
1207 				break;
1208 			}
1209 		}
1210 	}
1211 	pmc->mca_flags &= ~MAF_GSQUERY;
1212 	if (scount == nsrcs)	/* all sources excluded */
1213 		return false;
1214 	return true;
1215 }
1216 
1217 static bool mld_marksources(struct ifmcaddr6 *pmc, int nsrcs,
1218 			    const struct in6_addr *srcs)
1219 {
1220 	struct ip6_sf_list *psf;
1221 	int i, scount;
1222 
1223 	mc_assert_locked(pmc->idev);
1224 
1225 	if (pmc->mca_sfmode == MCAST_EXCLUDE)
1226 		return mld_xmarksources(pmc, nsrcs, srcs);
1227 
1228 	/* mark INCLUDE-mode sources */
1229 
1230 	scount = 0;
1231 	for_each_psf_mclock(pmc, psf) {
1232 		if (scount == nsrcs)
1233 			break;
1234 		for (i = 0; i < nsrcs; i++) {
1235 			if (ipv6_addr_equal(&srcs[i], &psf->sf_addr)) {
1236 				psf->sf_gsresp = 1;
1237 				scount++;
1238 				break;
1239 			}
1240 		}
1241 	}
1242 	if (!scount) {
1243 		pmc->mca_flags &= ~MAF_GSQUERY;
1244 		return false;
1245 	}
1246 	pmc->mca_flags |= MAF_GSQUERY;
1247 	return true;
1248 }
1249 
1250 static int mld_force_mld_version(const struct inet6_dev *idev)
1251 {
1252 	const struct net *net = dev_net(idev->dev);
1253 	int all_force;
1254 
1255 	all_force = READ_ONCE(net->ipv6.devconf_all->force_mld_version);
1256 	/* Normally, both are 0 here. If enforcement to a particular is
1257 	 * being used, individual device enforcement will have a lower
1258 	 * precedence over 'all' device (.../conf/all/force_mld_version).
1259 	 */
1260 	return all_force ?: READ_ONCE(idev->cnf.force_mld_version);
1261 }
1262 
1263 static bool mld_in_v2_mode_only(const struct inet6_dev *idev)
1264 {
1265 	return mld_force_mld_version(idev) == 2;
1266 }
1267 
1268 static bool mld_in_v1_mode_only(const struct inet6_dev *idev)
1269 {
1270 	return mld_force_mld_version(idev) == 1;
1271 }
1272 
1273 static bool mld_in_v1_mode(const struct inet6_dev *idev)
1274 {
1275 	if (mld_in_v2_mode_only(idev))
1276 		return false;
1277 	if (mld_in_v1_mode_only(idev))
1278 		return true;
1279 	if (idev->mc_v1_seen && time_before(jiffies, idev->mc_v1_seen))
1280 		return true;
1281 
1282 	return false;
1283 }
1284 
1285 static void mld_set_v1_mode(struct inet6_dev *idev)
1286 {
1287 	/* RFC3810, relevant sections:
1288 	 *  - 9.1. Robustness Variable
1289 	 *  - 9.2. Query Interval
1290 	 *  - 9.3. Query Response Interval
1291 	 *  - 9.12. Older Version Querier Present Timeout
1292 	 */
1293 	unsigned long switchback;
1294 
1295 	switchback = (idev->mc_qrv * idev->mc_qi) + idev->mc_qri;
1296 
1297 	idev->mc_v1_seen = jiffies + switchback;
1298 }
1299 
1300 static void mld_update_qrv(struct inet6_dev *idev,
1301 			   const struct mld2_query *mlh2)
1302 {
1303 	/* RFC3810, relevant sections:
1304 	 *  - 5.1.8. QRV (Querier's Robustness Variable)
1305 	 *  - 9.1. Robustness Variable
1306 	 */
1307 
1308 	/* The value of the Robustness Variable MUST NOT be zero,
1309 	 * and SHOULD NOT be one. Catch this here if we ever run
1310 	 * into such a case in future.
1311 	 */
1312 	const int min_qrv = min(MLD_QRV_DEFAULT, sysctl_mld_qrv);
1313 	WARN_ON(idev->mc_qrv == 0);
1314 
1315 	if (mlh2->mld2q_qrv > 0)
1316 		idev->mc_qrv = mlh2->mld2q_qrv;
1317 
1318 	if (unlikely(idev->mc_qrv < min_qrv)) {
1319 		net_warn_ratelimited("IPv6: MLD: clamping QRV from %u to %u!\n",
1320 				     idev->mc_qrv, min_qrv);
1321 		idev->mc_qrv = min_qrv;
1322 	}
1323 }
1324 
1325 static void mld_update_qi(struct inet6_dev *idev,
1326 			  const struct mld2_query *mlh2)
1327 {
1328 	/* RFC3810, relevant sections:
1329 	 *  - 5.1.9. QQIC (Querier's Query Interval Code)
1330 	 *  - 9.2. Query Interval
1331 	 *  - 9.12. Older Version Querier Present Timeout
1332 	 *    (the [Query Interval] in the last Query received)
1333 	 */
1334 	unsigned long mc_qqi;
1335 
1336 	if (mlh2->mld2q_qqic < 128) {
1337 		mc_qqi = mlh2->mld2q_qqic;
1338 	} else {
1339 		unsigned long mc_man, mc_exp;
1340 
1341 		mc_exp = MLDV2_QQIC_EXP(mlh2->mld2q_qqic);
1342 		mc_man = MLDV2_QQIC_MAN(mlh2->mld2q_qqic);
1343 
1344 		mc_qqi = (mc_man | 0x10) << (mc_exp + 3);
1345 	}
1346 
1347 	idev->mc_qi = mc_qqi * HZ;
1348 }
1349 
1350 static void mld_update_qri(struct inet6_dev *idev,
1351 			   const struct mld2_query *mlh2)
1352 {
1353 	/* RFC3810, relevant sections:
1354 	 *  - 5.1.3. Maximum Response Code
1355 	 *  - 9.3. Query Response Interval
1356 	 */
1357 	idev->mc_qri = msecs_to_jiffies(mldv2_mrc(mlh2));
1358 }
1359 
1360 static int mld_process_v1(struct inet6_dev *idev, struct mld_msg *mld,
1361 			  unsigned long *max_delay, bool v1_query)
1362 {
1363 	unsigned long mldv1_md;
1364 
1365 	/* Ignore v1 queries */
1366 	if (mld_in_v2_mode_only(idev))
1367 		return -EINVAL;
1368 
1369 	mldv1_md = ntohs(mld->mld_maxdelay);
1370 
1371 	/* When in MLDv1 fallback and a MLDv2 router start-up being
1372 	 * unaware of current MLDv1 operation, the MRC == MRD mapping
1373 	 * only works when the exponential algorithm is not being
1374 	 * used (as MLDv1 is unaware of such things).
1375 	 *
1376 	 * According to the RFC author, the MLDv2 implementations
1377 	 * he's aware of all use a MRC < 32768 on start up queries.
1378 	 *
1379 	 * Thus, should we *ever* encounter something else larger
1380 	 * than that, just assume the maximum possible within our
1381 	 * reach.
1382 	 */
1383 	if (!v1_query)
1384 		mldv1_md = min(mldv1_md, MLDV1_MRD_MAX_COMPAT);
1385 
1386 	*max_delay = max(msecs_to_jiffies(mldv1_md), 1UL);
1387 
1388 	/* MLDv1 router present: we need to go into v1 mode *only*
1389 	 * when an MLDv1 query is received as per section 9.12. of
1390 	 * RFC3810! And we know from RFC2710 section 3.7 that MLDv1
1391 	 * queries MUST be of exactly 24 octets.
1392 	 */
1393 	if (v1_query)
1394 		mld_set_v1_mode(idev);
1395 
1396 	/* cancel MLDv2 report work */
1397 	mld_gq_stop_work(idev);
1398 	/* cancel the interface change work */
1399 	mld_ifc_stop_work(idev);
1400 	/* clear deleted report items */
1401 	mld_clear_delrec(idev);
1402 
1403 	return 0;
1404 }
1405 
1406 static void mld_process_v2(struct inet6_dev *idev, struct mld2_query *mld,
1407 			   unsigned long *max_delay)
1408 {
1409 	*max_delay = max(msecs_to_jiffies(mldv2_mrc(mld)), 1UL);
1410 
1411 	mld_update_qrv(idev, mld);
1412 	mld_update_qi(idev, mld);
1413 	mld_update_qri(idev, mld);
1414 
1415 	idev->mc_maxdelay = *max_delay;
1416 
1417 	return;
1418 }
1419 
1420 /* called with rcu_read_lock() */
1421 void igmp6_event_query(struct sk_buff *skb)
1422 {
1423 	struct inet6_dev *idev = __in6_dev_get(skb->dev);
1424 
1425 	if (!idev || idev->dead)
1426 		goto out;
1427 
1428 	spin_lock_bh(&idev->mc_query_lock);
1429 	if (skb_queue_len(&idev->mc_query_queue) < MLD_MAX_SKBS) {
1430 		__skb_queue_tail(&idev->mc_query_queue, skb);
1431 		if (!mod_delayed_work(mld_wq, &idev->mc_query_work, 0))
1432 			in6_dev_hold(idev);
1433 		skb = NULL;
1434 	}
1435 	spin_unlock_bh(&idev->mc_query_lock);
1436 out:
1437 	kfree_skb(skb);
1438 }
1439 
1440 static void __mld_query_work(struct sk_buff *skb)
1441 {
1442 	struct mld2_query *mlh2 = NULL;
1443 	const struct in6_addr *group;
1444 	unsigned long max_delay;
1445 	struct inet6_dev *idev;
1446 	struct ifmcaddr6 *ma;
1447 	struct mld_msg *mld;
1448 	int group_type;
1449 	int mark = 0;
1450 	int len, err;
1451 
1452 	if (!pskb_may_pull(skb, sizeof(struct in6_addr)))
1453 		goto kfree_skb;
1454 
1455 	/* compute payload length excluding extension headers */
1456 	len = ntohs(ipv6_hdr(skb)->payload_len) + sizeof(struct ipv6hdr);
1457 	len -= skb_network_header_len(skb);
1458 
1459 	/* RFC3810 6.2
1460 	 * Upon reception of an MLD message that contains a Query, the node
1461 	 * checks if the source address of the message is a valid link-local
1462 	 * address, if the Hop Limit is set to 1, and if the Router Alert
1463 	 * option is present in the Hop-By-Hop Options header of the IPv6
1464 	 * packet.  If any of these checks fails, the packet is dropped.
1465 	 */
1466 	if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL) ||
1467 	    ipv6_hdr(skb)->hop_limit != 1 ||
1468 	    !(IP6CB(skb)->flags & IP6SKB_ROUTERALERT) ||
1469 	    IP6CB(skb)->ra != htons(IPV6_OPT_ROUTERALERT_MLD))
1470 		goto kfree_skb;
1471 
1472 	idev = in6_dev_get(skb->dev);
1473 	if (!idev)
1474 		goto kfree_skb;
1475 
1476 	mld = (struct mld_msg *)icmp6_hdr(skb);
1477 	group = &mld->mld_mca;
1478 	group_type = ipv6_addr_type(group);
1479 
1480 	if (group_type != IPV6_ADDR_ANY &&
1481 	    !(group_type&IPV6_ADDR_MULTICAST))
1482 		goto out;
1483 
1484 	if (len < MLD_V1_QUERY_LEN) {
1485 		goto out;
1486 	} else if (len == MLD_V1_QUERY_LEN || mld_in_v1_mode(idev)) {
1487 		err = mld_process_v1(idev, mld, &max_delay,
1488 				     len == MLD_V1_QUERY_LEN);
1489 		if (err < 0)
1490 			goto out;
1491 	} else if (len >= MLD_V2_QUERY_LEN_MIN) {
1492 		int srcs_offset = sizeof(struct mld2_query) -
1493 				  sizeof(struct icmp6hdr);
1494 
1495 		if (!pskb_may_pull(skb, srcs_offset))
1496 			goto out;
1497 
1498 		mlh2 = (struct mld2_query *)skb_transport_header(skb);
1499 
1500 		mld_process_v2(idev, mlh2, &max_delay);
1501 
1502 		if (group_type == IPV6_ADDR_ANY) { /* general query */
1503 			if (mlh2->mld2q_nsrcs)
1504 				goto out; /* no sources allowed */
1505 
1506 			mld_gq_start_work(idev);
1507 			goto out;
1508 		}
1509 		/* mark sources to include, if group & source-specific */
1510 		if (mlh2->mld2q_nsrcs != 0) {
1511 			if (!pskb_may_pull(skb, srcs_offset +
1512 			    ntohs(mlh2->mld2q_nsrcs) * sizeof(struct in6_addr)))
1513 				goto out;
1514 
1515 			mlh2 = (struct mld2_query *)skb_transport_header(skb);
1516 			mark = 1;
1517 		}
1518 	} else {
1519 		goto out;
1520 	}
1521 
1522 	if (group_type == IPV6_ADDR_ANY) {
1523 		for_each_mc_mclock(idev, ma) {
1524 			igmp6_group_queried(ma, max_delay);
1525 		}
1526 	} else {
1527 		for_each_mc_mclock(idev, ma) {
1528 			if (!ipv6_addr_equal(group, &ma->mca_addr))
1529 				continue;
1530 			if (ma->mca_flags & MAF_TIMER_RUNNING) {
1531 				/* gsquery <- gsquery && mark */
1532 				if (!mark)
1533 					ma->mca_flags &= ~MAF_GSQUERY;
1534 			} else {
1535 				/* gsquery <- mark */
1536 				if (mark)
1537 					ma->mca_flags |= MAF_GSQUERY;
1538 				else
1539 					ma->mca_flags &= ~MAF_GSQUERY;
1540 			}
1541 			if (!(ma->mca_flags & MAF_GSQUERY) ||
1542 			    mld_marksources(ma, ntohs(mlh2->mld2q_nsrcs), mlh2->mld2q_srcs))
1543 				igmp6_group_queried(ma, max_delay);
1544 			break;
1545 		}
1546 	}
1547 
1548 out:
1549 	in6_dev_put(idev);
1550 kfree_skb:
1551 	consume_skb(skb);
1552 }
1553 
1554 static void mld_query_work(struct work_struct *work)
1555 {
1556 	struct inet6_dev *idev = container_of(to_delayed_work(work),
1557 					      struct inet6_dev,
1558 					      mc_query_work);
1559 	struct sk_buff_head q;
1560 	struct sk_buff *skb;
1561 	bool rework = false;
1562 	int cnt = 0;
1563 
1564 	skb_queue_head_init(&q);
1565 
1566 	spin_lock_bh(&idev->mc_query_lock);
1567 	while ((skb = __skb_dequeue(&idev->mc_query_queue))) {
1568 		__skb_queue_tail(&q, skb);
1569 
1570 		if (++cnt >= MLD_MAX_QUEUE) {
1571 			rework = true;
1572 			break;
1573 		}
1574 	}
1575 	spin_unlock_bh(&idev->mc_query_lock);
1576 
1577 	mutex_lock(&idev->mc_lock);
1578 	while ((skb = __skb_dequeue(&q)))
1579 		__mld_query_work(skb);
1580 	mutex_unlock(&idev->mc_lock);
1581 
1582 	if (rework && queue_delayed_work(mld_wq, &idev->mc_query_work, 0))
1583 		return;
1584 
1585 	in6_dev_put(idev);
1586 }
1587 
1588 /* called with rcu_read_lock() */
1589 void igmp6_event_report(struct sk_buff *skb)
1590 {
1591 	struct inet6_dev *idev = __in6_dev_get(skb->dev);
1592 
1593 	if (!idev || idev->dead)
1594 		goto out;
1595 
1596 	spin_lock_bh(&idev->mc_report_lock);
1597 	if (skb_queue_len(&idev->mc_report_queue) < MLD_MAX_SKBS) {
1598 		__skb_queue_tail(&idev->mc_report_queue, skb);
1599 		if (!mod_delayed_work(mld_wq, &idev->mc_report_work, 0))
1600 			in6_dev_hold(idev);
1601 		skb = NULL;
1602 	}
1603 	spin_unlock_bh(&idev->mc_report_lock);
1604 out:
1605 	kfree_skb(skb);
1606 }
1607 
1608 static void __mld_report_work(struct sk_buff *skb)
1609 {
1610 	struct inet6_dev *idev;
1611 	struct ifmcaddr6 *ma;
1612 	struct mld_msg *mld;
1613 	int addr_type;
1614 
1615 	/* Our own report looped back. Ignore it. */
1616 	if (skb->pkt_type == PACKET_LOOPBACK)
1617 		goto kfree_skb;
1618 
1619 	/* send our report if the MC router may not have heard this report */
1620 	if (skb->pkt_type != PACKET_MULTICAST &&
1621 	    skb->pkt_type != PACKET_BROADCAST)
1622 		goto kfree_skb;
1623 
1624 	if (!pskb_may_pull(skb, sizeof(*mld) - sizeof(struct icmp6hdr)))
1625 		goto kfree_skb;
1626 
1627 	mld = (struct mld_msg *)icmp6_hdr(skb);
1628 
1629 	/* Drop reports with not link local source */
1630 	addr_type = ipv6_addr_type(&ipv6_hdr(skb)->saddr);
1631 	if (addr_type != IPV6_ADDR_ANY &&
1632 	    !(addr_type&IPV6_ADDR_LINKLOCAL))
1633 		goto kfree_skb;
1634 
1635 	idev = in6_dev_get(skb->dev);
1636 	if (!idev)
1637 		goto kfree_skb;
1638 
1639 	/*
1640 	 *	Cancel the work for this group
1641 	 */
1642 
1643 	for_each_mc_mclock(idev, ma) {
1644 		if (ipv6_addr_equal(&ma->mca_addr, &mld->mld_mca)) {
1645 			if (cancel_delayed_work(&ma->mca_work))
1646 				refcount_dec(&ma->mca_refcnt);
1647 			ma->mca_flags &= ~(MAF_LAST_REPORTER |
1648 					   MAF_TIMER_RUNNING);
1649 			break;
1650 		}
1651 	}
1652 
1653 	in6_dev_put(idev);
1654 kfree_skb:
1655 	consume_skb(skb);
1656 }
1657 
1658 static void mld_report_work(struct work_struct *work)
1659 {
1660 	struct inet6_dev *idev = container_of(to_delayed_work(work),
1661 					      struct inet6_dev,
1662 					      mc_report_work);
1663 	struct sk_buff_head q;
1664 	struct sk_buff *skb;
1665 	bool rework = false;
1666 	int cnt = 0;
1667 
1668 	skb_queue_head_init(&q);
1669 	spin_lock_bh(&idev->mc_report_lock);
1670 	while ((skb = __skb_dequeue(&idev->mc_report_queue))) {
1671 		__skb_queue_tail(&q, skb);
1672 
1673 		if (++cnt >= MLD_MAX_QUEUE) {
1674 			rework = true;
1675 			break;
1676 		}
1677 	}
1678 	spin_unlock_bh(&idev->mc_report_lock);
1679 
1680 	mutex_lock(&idev->mc_lock);
1681 	while ((skb = __skb_dequeue(&q)))
1682 		__mld_report_work(skb);
1683 	mutex_unlock(&idev->mc_lock);
1684 
1685 	if (rework && queue_delayed_work(mld_wq, &idev->mc_report_work, 0))
1686 		return;
1687 
1688 	in6_dev_put(idev);
1689 }
1690 
1691 static bool is_in(struct ifmcaddr6 *pmc, struct ip6_sf_list *psf, int type,
1692 		  int gdeleted, int sdeleted)
1693 {
1694 	switch (type) {
1695 	case MLD2_MODE_IS_INCLUDE:
1696 	case MLD2_MODE_IS_EXCLUDE:
1697 		if (gdeleted || sdeleted)
1698 			return false;
1699 		if (!((pmc->mca_flags & MAF_GSQUERY) && !psf->sf_gsresp)) {
1700 			if (pmc->mca_sfmode == MCAST_INCLUDE)
1701 				return true;
1702 			/* don't include if this source is excluded
1703 			 * in all filters
1704 			 */
1705 			if (psf->sf_count[MCAST_INCLUDE])
1706 				return type == MLD2_MODE_IS_INCLUDE;
1707 			return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1708 				psf->sf_count[MCAST_EXCLUDE];
1709 		}
1710 		return false;
1711 	case MLD2_CHANGE_TO_INCLUDE:
1712 		if (gdeleted || sdeleted)
1713 			return false;
1714 		return psf->sf_count[MCAST_INCLUDE] != 0;
1715 	case MLD2_CHANGE_TO_EXCLUDE:
1716 		if (gdeleted || sdeleted)
1717 			return false;
1718 		if (pmc->mca_sfcount[MCAST_EXCLUDE] == 0 ||
1719 		    psf->sf_count[MCAST_INCLUDE])
1720 			return false;
1721 		return pmc->mca_sfcount[MCAST_EXCLUDE] ==
1722 			psf->sf_count[MCAST_EXCLUDE];
1723 	case MLD2_ALLOW_NEW_SOURCES:
1724 		if (gdeleted || !psf->sf_crcount)
1725 			return false;
1726 		return (pmc->mca_sfmode == MCAST_INCLUDE) ^ sdeleted;
1727 	case MLD2_BLOCK_OLD_SOURCES:
1728 		if (pmc->mca_sfmode == MCAST_INCLUDE)
1729 			return gdeleted || (psf->sf_crcount && sdeleted);
1730 		return psf->sf_crcount && !gdeleted && !sdeleted;
1731 	}
1732 	return false;
1733 }
1734 
1735 static int
1736 mld_scount(struct ifmcaddr6 *pmc, int type, int gdeleted, int sdeleted)
1737 {
1738 	struct ip6_sf_list *psf;
1739 	int scount = 0;
1740 
1741 	for_each_psf_mclock(pmc, psf) {
1742 		if (!is_in(pmc, psf, type, gdeleted, sdeleted))
1743 			continue;
1744 		scount++;
1745 	}
1746 	return scount;
1747 }
1748 
1749 static void ip6_mc_hdr(const struct sock *sk, struct sk_buff *skb,
1750 		       struct net_device *dev, const struct in6_addr *saddr,
1751 		       const struct in6_addr *daddr, int proto, int len)
1752 {
1753 	struct ipv6hdr *hdr;
1754 
1755 	skb->protocol = htons(ETH_P_IPV6);
1756 	skb->dev = dev;
1757 
1758 	skb_reset_network_header(skb);
1759 	skb_put(skb, sizeof(struct ipv6hdr));
1760 	hdr = ipv6_hdr(skb);
1761 
1762 	ip6_flow_hdr(hdr, 0, 0);
1763 
1764 	hdr->payload_len = htons(len);
1765 	hdr->nexthdr = proto;
1766 	hdr->hop_limit = READ_ONCE(inet6_sk(sk)->hop_limit);
1767 
1768 	hdr->saddr = *saddr;
1769 	hdr->daddr = *daddr;
1770 }
1771 
1772 static struct sk_buff *mld_newpack(struct inet6_dev *idev, unsigned int mtu)
1773 {
1774 	u8 ra[8] = { IPPROTO_ICMPV6, 0, IPV6_TLV_ROUTERALERT,
1775 		     2, 0, 0, IPV6_TLV_PADN, 0 };
1776 	struct net_device *dev = idev->dev;
1777 	int hlen = LL_RESERVED_SPACE(dev);
1778 	int tlen = dev->needed_tailroom;
1779 	const struct in6_addr *saddr;
1780 	struct in6_addr addr_buf;
1781 	struct mld2_report *pmr;
1782 	struct sk_buff *skb;
1783 	unsigned int size;
1784 	struct sock *sk;
1785 	struct net *net;
1786 
1787 	/* we assume size > sizeof(ra) here
1788 	 * Also try to not allocate high-order pages for big MTU
1789 	 */
1790 	size = min_t(int, mtu, PAGE_SIZE / 2) + hlen + tlen;
1791 	skb = alloc_skb(size, GFP_KERNEL);
1792 	if (!skb)
1793 		return NULL;
1794 
1795 	skb->priority = TC_PRIO_CONTROL;
1796 	skb_reserve(skb, hlen);
1797 	skb_tailroom_reserve(skb, mtu, tlen);
1798 
1799 	rcu_read_lock();
1800 
1801 	net = dev_net_rcu(dev);
1802 	sk = net->ipv6.igmp_sk;
1803 	skb_set_owner_w(skb, sk);
1804 
1805 	if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
1806 		/* <draft-ietf-magma-mld-source-05.txt>:
1807 		 * use unspecified address as the source address
1808 		 * when a valid link-local address is not available.
1809 		 */
1810 		saddr = &in6addr_any;
1811 	} else
1812 		saddr = &addr_buf;
1813 
1814 	ip6_mc_hdr(sk, skb, dev, saddr, &mld2_all_mcr, NEXTHDR_HOP, 0);
1815 
1816 	rcu_read_unlock();
1817 
1818 	skb_put_data(skb, ra, sizeof(ra));
1819 
1820 	skb_set_transport_header(skb, skb_tail_pointer(skb) - skb->data);
1821 	skb_put(skb, sizeof(*pmr));
1822 	pmr = (struct mld2_report *)skb_transport_header(skb);
1823 	pmr->mld2r_type = ICMPV6_MLD2_REPORT;
1824 	pmr->mld2r_resv1 = 0;
1825 	pmr->mld2r_cksum = 0;
1826 	pmr->mld2r_resv2 = 0;
1827 	pmr->mld2r_ngrec = 0;
1828 	return skb;
1829 }
1830 
1831 static void mld_sendpack(struct sk_buff *skb)
1832 {
1833 	struct ipv6hdr *pip6 = ipv6_hdr(skb);
1834 	struct mld2_report *pmr =
1835 			      (struct mld2_report *)skb_transport_header(skb);
1836 	int payload_len, mldlen;
1837 	struct inet6_dev *idev;
1838 	struct net *net = dev_net(skb->dev);
1839 	int err;
1840 	struct flowi6 fl6;
1841 	struct dst_entry *dst;
1842 
1843 	rcu_read_lock();
1844 	idev = __in6_dev_get(skb->dev);
1845 	IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTREQUESTS);
1846 
1847 	payload_len = (skb_tail_pointer(skb) - skb_network_header(skb)) -
1848 		sizeof(*pip6);
1849 	mldlen = skb_tail_pointer(skb) - skb_transport_header(skb);
1850 	pip6->payload_len = htons(payload_len);
1851 
1852 	pmr->mld2r_cksum = csum_ipv6_magic(&pip6->saddr, &pip6->daddr, mldlen,
1853 					   IPPROTO_ICMPV6,
1854 					   csum_partial(skb_transport_header(skb),
1855 							mldlen, 0));
1856 
1857 	icmpv6_flow_init(net->ipv6.igmp_sk, &fl6, ICMPV6_MLD2_REPORT,
1858 			 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
1859 			 skb->dev->ifindex);
1860 	dst = icmp6_dst_alloc(skb->dev, &fl6);
1861 
1862 	err = 0;
1863 	if (IS_ERR(dst)) {
1864 		err = PTR_ERR(dst);
1865 		dst = NULL;
1866 	}
1867 	skb_dst_set(skb, dst);
1868 	if (err)
1869 		goto err_out;
1870 
1871 	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
1872 		      net, net->ipv6.igmp_sk, skb, NULL, skb->dev,
1873 		      dst_output);
1874 out:
1875 	if (!err) {
1876 		ICMP6MSGOUT_INC_STATS(net, idev, ICMPV6_MLD2_REPORT);
1877 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
1878 	} else {
1879 		IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
1880 	}
1881 
1882 	rcu_read_unlock();
1883 	return;
1884 
1885 err_out:
1886 	kfree_skb(skb);
1887 	goto out;
1888 }
1889 
1890 static int grec_size(struct ifmcaddr6 *pmc, int type, int gdel, int sdel)
1891 {
1892 	return sizeof(struct mld2_grec) + 16 * mld_scount(pmc,type,gdel,sdel);
1893 }
1894 
1895 static struct sk_buff *add_grhead(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1896 	int type, struct mld2_grec **ppgr, unsigned int mtu)
1897 {
1898 	struct mld2_report *pmr;
1899 	struct mld2_grec *pgr;
1900 
1901 	if (!skb) {
1902 		skb = mld_newpack(pmc->idev, mtu);
1903 		if (!skb)
1904 			return NULL;
1905 	}
1906 	pgr = skb_put(skb, sizeof(struct mld2_grec));
1907 	pgr->grec_type = type;
1908 	pgr->grec_auxwords = 0;
1909 	pgr->grec_nsrcs = 0;
1910 	pgr->grec_mca = pmc->mca_addr;	/* structure copy */
1911 	pmr = (struct mld2_report *)skb_transport_header(skb);
1912 	pmr->mld2r_ngrec = htons(ntohs(pmr->mld2r_ngrec)+1);
1913 	*ppgr = pgr;
1914 	return skb;
1915 }
1916 
1917 #define AVAILABLE(skb)	((skb) ? skb_availroom(skb) : 0)
1918 
1919 static struct sk_buff *add_grec(struct sk_buff *skb, struct ifmcaddr6 *pmc,
1920 				int type, int gdeleted, int sdeleted,
1921 				int crsend)
1922 {
1923 	struct ip6_sf_list *psf, *psf_prev, *psf_next;
1924 	int scount, stotal, first, isquery, truncate;
1925 	struct ip6_sf_list __rcu **psf_list;
1926 	struct inet6_dev *idev = pmc->idev;
1927 	struct net_device *dev = idev->dev;
1928 	struct mld2_grec *pgr = NULL;
1929 	struct mld2_report *pmr;
1930 	unsigned int mtu;
1931 
1932 	mc_assert_locked(idev);
1933 
1934 	if (pmc->mca_flags & MAF_NOREPORT)
1935 		return skb;
1936 
1937 	mtu = READ_ONCE(dev->mtu);
1938 	if (mtu < IPV6_MIN_MTU)
1939 		return skb;
1940 
1941 	isquery = type == MLD2_MODE_IS_INCLUDE ||
1942 		  type == MLD2_MODE_IS_EXCLUDE;
1943 	truncate = type == MLD2_MODE_IS_EXCLUDE ||
1944 		    type == MLD2_CHANGE_TO_EXCLUDE;
1945 
1946 	stotal = scount = 0;
1947 
1948 	psf_list = sdeleted ? &pmc->mca_tomb : &pmc->mca_sources;
1949 
1950 	if (!rcu_access_pointer(*psf_list))
1951 		goto empty_source;
1952 
1953 	pmr = skb ? (struct mld2_report *)skb_transport_header(skb) : NULL;
1954 
1955 	/* EX and TO_EX get a fresh packet, if needed */
1956 	if (truncate) {
1957 		if (pmr && pmr->mld2r_ngrec &&
1958 		    AVAILABLE(skb) < grec_size(pmc, type, gdeleted, sdeleted)) {
1959 			if (skb)
1960 				mld_sendpack(skb);
1961 			skb = mld_newpack(idev, mtu);
1962 		}
1963 	}
1964 	first = 1;
1965 	psf_prev = NULL;
1966 	for (psf = mc_dereference(*psf_list, idev);
1967 	     psf;
1968 	     psf = psf_next) {
1969 		struct in6_addr *psrc;
1970 
1971 		psf_next = mc_dereference(psf->sf_next, idev);
1972 
1973 		if (!is_in(pmc, psf, type, gdeleted, sdeleted) && !crsend) {
1974 			psf_prev = psf;
1975 			continue;
1976 		}
1977 
1978 		/* Based on RFC3810 6.1. Should not send source-list change
1979 		 * records when there is a filter mode change.
1980 		 */
1981 		if (((gdeleted && pmc->mca_sfmode == MCAST_EXCLUDE) ||
1982 		     (!gdeleted && pmc->mca_crcount)) &&
1983 		    (type == MLD2_ALLOW_NEW_SOURCES ||
1984 		     type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount)
1985 			goto decrease_sf_crcount;
1986 
1987 		/* clear marks on query responses */
1988 		if (isquery)
1989 			psf->sf_gsresp = 0;
1990 
1991 		if (AVAILABLE(skb) < sizeof(*psrc) +
1992 		    first*sizeof(struct mld2_grec)) {
1993 			if (truncate && !first)
1994 				break;	 /* truncate these */
1995 			if (pgr)
1996 				pgr->grec_nsrcs = htons(scount);
1997 			if (skb)
1998 				mld_sendpack(skb);
1999 			skb = mld_newpack(idev, mtu);
2000 			first = 1;
2001 			scount = 0;
2002 		}
2003 		if (first) {
2004 			skb = add_grhead(skb, pmc, type, &pgr, mtu);
2005 			first = 0;
2006 		}
2007 		if (!skb)
2008 			return NULL;
2009 		psrc = skb_put(skb, sizeof(*psrc));
2010 		*psrc = psf->sf_addr;
2011 		scount++; stotal++;
2012 		if ((type == MLD2_ALLOW_NEW_SOURCES ||
2013 		     type == MLD2_BLOCK_OLD_SOURCES) && psf->sf_crcount) {
2014 decrease_sf_crcount:
2015 			psf->sf_crcount--;
2016 			if ((sdeleted || gdeleted) && psf->sf_crcount == 0) {
2017 				if (psf_prev)
2018 					rcu_assign_pointer(psf_prev->sf_next,
2019 							   mc_dereference(psf->sf_next, idev));
2020 				else
2021 					rcu_assign_pointer(*psf_list,
2022 							   mc_dereference(psf->sf_next, idev));
2023 				kfree_rcu(psf, rcu);
2024 				continue;
2025 			}
2026 		}
2027 		psf_prev = psf;
2028 	}
2029 
2030 empty_source:
2031 	if (!stotal) {
2032 		if (type == MLD2_ALLOW_NEW_SOURCES ||
2033 		    type == MLD2_BLOCK_OLD_SOURCES)
2034 			return skb;
2035 		if (pmc->mca_crcount || isquery || crsend) {
2036 			/* make sure we have room for group header */
2037 			if (skb && AVAILABLE(skb) < sizeof(struct mld2_grec)) {
2038 				mld_sendpack(skb);
2039 				skb = NULL; /* add_grhead will get a new one */
2040 			}
2041 			skb = add_grhead(skb, pmc, type, &pgr, mtu);
2042 		}
2043 	}
2044 	if (pgr)
2045 		pgr->grec_nsrcs = htons(scount);
2046 
2047 	if (isquery)
2048 		pmc->mca_flags &= ~MAF_GSQUERY;	/* clear query state */
2049 	return skb;
2050 }
2051 
2052 static void mld_send_report(struct inet6_dev *idev, struct ifmcaddr6 *pmc)
2053 {
2054 	struct sk_buff *skb = NULL;
2055 	int type;
2056 
2057 	mc_assert_locked(idev);
2058 
2059 	if (!pmc) {
2060 		for_each_mc_mclock(idev, pmc) {
2061 			if (pmc->mca_flags & MAF_NOREPORT)
2062 				continue;
2063 			if (pmc->mca_sfcount[MCAST_EXCLUDE])
2064 				type = MLD2_MODE_IS_EXCLUDE;
2065 			else
2066 				type = MLD2_MODE_IS_INCLUDE;
2067 			skb = add_grec(skb, pmc, type, 0, 0, 0);
2068 		}
2069 	} else {
2070 		if (pmc->mca_sfcount[MCAST_EXCLUDE])
2071 			type = MLD2_MODE_IS_EXCLUDE;
2072 		else
2073 			type = MLD2_MODE_IS_INCLUDE;
2074 		skb = add_grec(skb, pmc, type, 0, 0, 0);
2075 	}
2076 	if (skb)
2077 		mld_sendpack(skb);
2078 }
2079 
2080 /* remove zero-count source records from a source filter list */
2081 static void mld_clear_zeros(struct ip6_sf_list __rcu **ppsf, struct inet6_dev *idev)
2082 {
2083 	struct ip6_sf_list *psf_prev, *psf_next, *psf;
2084 
2085 	psf_prev = NULL;
2086 	for (psf = mc_dereference(*ppsf, idev);
2087 	     psf;
2088 	     psf = psf_next) {
2089 		psf_next = mc_dereference(psf->sf_next, idev);
2090 		if (psf->sf_crcount == 0) {
2091 			if (psf_prev)
2092 				rcu_assign_pointer(psf_prev->sf_next,
2093 						   mc_dereference(psf->sf_next, idev));
2094 			else
2095 				rcu_assign_pointer(*ppsf,
2096 						   mc_dereference(psf->sf_next, idev));
2097 			kfree_rcu(psf, rcu);
2098 		} else {
2099 			psf_prev = psf;
2100 		}
2101 	}
2102 }
2103 
2104 static void mld_send_cr(struct inet6_dev *idev)
2105 {
2106 	struct ifmcaddr6 *pmc, *pmc_prev, *pmc_next;
2107 	struct sk_buff *skb = NULL;
2108 	int type, dtype;
2109 
2110 	/* deleted MCA's */
2111 	pmc_prev = NULL;
2112 	for (pmc = mc_dereference(idev->mc_tomb, idev);
2113 	     pmc;
2114 	     pmc = pmc_next) {
2115 		pmc_next = mc_dereference(pmc->next, idev);
2116 		if (pmc->mca_sfmode == MCAST_INCLUDE) {
2117 			type = MLD2_BLOCK_OLD_SOURCES;
2118 			dtype = MLD2_BLOCK_OLD_SOURCES;
2119 			skb = add_grec(skb, pmc, type, 1, 0, 0);
2120 			skb = add_grec(skb, pmc, dtype, 1, 1, 0);
2121 		}
2122 		if (pmc->mca_crcount) {
2123 			if (pmc->mca_sfmode == MCAST_EXCLUDE) {
2124 				type = MLD2_CHANGE_TO_INCLUDE;
2125 				skb = add_grec(skb, pmc, type, 1, 0, 0);
2126 			}
2127 			pmc->mca_crcount--;
2128 			if (pmc->mca_crcount == 0) {
2129 				mld_clear_zeros(&pmc->mca_tomb, idev);
2130 				mld_clear_zeros(&pmc->mca_sources, idev);
2131 			}
2132 		}
2133 		if (pmc->mca_crcount == 0 &&
2134 		    !rcu_access_pointer(pmc->mca_tomb) &&
2135 		    !rcu_access_pointer(pmc->mca_sources)) {
2136 			if (pmc_prev)
2137 				rcu_assign_pointer(pmc_prev->next, pmc_next);
2138 			else
2139 				rcu_assign_pointer(idev->mc_tomb, pmc_next);
2140 			in6_dev_put(pmc->idev);
2141 			kfree_rcu(pmc, rcu);
2142 		} else
2143 			pmc_prev = pmc;
2144 	}
2145 
2146 	/* change recs */
2147 	for_each_mc_mclock(idev, pmc) {
2148 		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2149 			type = MLD2_BLOCK_OLD_SOURCES;
2150 			dtype = MLD2_ALLOW_NEW_SOURCES;
2151 		} else {
2152 			type = MLD2_ALLOW_NEW_SOURCES;
2153 			dtype = MLD2_BLOCK_OLD_SOURCES;
2154 		}
2155 		skb = add_grec(skb, pmc, type, 0, 0, 0);
2156 		skb = add_grec(skb, pmc, dtype, 0, 1, 0);	/* deleted sources */
2157 
2158 		/* filter mode changes */
2159 		if (pmc->mca_crcount) {
2160 			if (pmc->mca_sfmode == MCAST_EXCLUDE)
2161 				type = MLD2_CHANGE_TO_EXCLUDE;
2162 			else
2163 				type = MLD2_CHANGE_TO_INCLUDE;
2164 			skb = add_grec(skb, pmc, type, 0, 0, 0);
2165 			pmc->mca_crcount--;
2166 		}
2167 	}
2168 	if (!skb)
2169 		return;
2170 	(void) mld_sendpack(skb);
2171 }
2172 
2173 static void igmp6_send(struct in6_addr *addr, struct net_device *dev, int type)
2174 {
2175 	const struct in6_addr *snd_addr, *saddr;
2176 	int err, len, payload_len, full_len;
2177 	struct in6_addr addr_buf;
2178 	struct inet6_dev *idev;
2179 	struct sk_buff *skb;
2180 	struct mld_msg *hdr;
2181 	int hlen = LL_RESERVED_SPACE(dev);
2182 	int tlen = dev->needed_tailroom;
2183 	u8 ra[8] = { IPPROTO_ICMPV6, 0,
2184 		     IPV6_TLV_ROUTERALERT, 2, 0, 0,
2185 		     IPV6_TLV_PADN, 0 };
2186 	struct dst_entry *dst;
2187 	struct flowi6 fl6;
2188 	struct net *net;
2189 	struct sock *sk;
2190 
2191 	if (type == ICMPV6_MGM_REDUCTION)
2192 		snd_addr = &in6addr_linklocal_allrouters;
2193 	else
2194 		snd_addr = addr;
2195 
2196 	len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
2197 	payload_len = len + sizeof(ra);
2198 	full_len = sizeof(struct ipv6hdr) + payload_len;
2199 
2200 	skb = alloc_skb(hlen + tlen + full_len, GFP_KERNEL);
2201 
2202 	rcu_read_lock();
2203 
2204 	net = dev_net_rcu(dev);
2205 	idev = __in6_dev_get(dev);
2206 	IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTREQUESTS);
2207 	if (!skb) {
2208 		IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
2209 		rcu_read_unlock();
2210 		return;
2211 	}
2212 	sk = net->ipv6.igmp_sk;
2213 	skb_set_owner_w(skb, sk);
2214 
2215 	skb->priority = TC_PRIO_CONTROL;
2216 	skb_reserve(skb, hlen);
2217 
2218 	if (ipv6_get_lladdr(dev, &addr_buf, IFA_F_TENTATIVE)) {
2219 		/* <draft-ietf-magma-mld-source-05.txt>:
2220 		 * use unspecified address as the source address
2221 		 * when a valid link-local address is not available.
2222 		 */
2223 		saddr = &in6addr_any;
2224 	} else
2225 		saddr = &addr_buf;
2226 
2227 	ip6_mc_hdr(sk, skb, dev, saddr, snd_addr, NEXTHDR_HOP, payload_len);
2228 
2229 	skb_put_data(skb, ra, sizeof(ra));
2230 
2231 	hdr = skb_put_zero(skb, sizeof(struct mld_msg));
2232 	hdr->mld_type = type;
2233 	hdr->mld_mca = *addr;
2234 
2235 	hdr->mld_cksum = csum_ipv6_magic(saddr, snd_addr, len,
2236 					 IPPROTO_ICMPV6,
2237 					 csum_partial(hdr, len, 0));
2238 
2239 	icmpv6_flow_init(sk, &fl6, type,
2240 			 &ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
2241 			 skb->dev->ifindex);
2242 	dst = icmp6_dst_alloc(skb->dev, &fl6);
2243 	if (IS_ERR(dst)) {
2244 		err = PTR_ERR(dst);
2245 		goto err_out;
2246 	}
2247 
2248 	skb_dst_set(skb, dst);
2249 	err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT,
2250 		      net, sk, skb, NULL, skb->dev,
2251 		      dst_output);
2252 out:
2253 	if (!err) {
2254 		ICMP6MSGOUT_INC_STATS(net, idev, type);
2255 		ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS);
2256 	} else
2257 		IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTDISCARDS);
2258 
2259 	rcu_read_unlock();
2260 	return;
2261 
2262 err_out:
2263 	kfree_skb(skb);
2264 	goto out;
2265 }
2266 
2267 static void mld_send_initial_cr(struct inet6_dev *idev)
2268 {
2269 	struct ifmcaddr6 *pmc;
2270 	struct sk_buff *skb;
2271 	int type;
2272 
2273 	mc_assert_locked(idev);
2274 
2275 	if (mld_in_v1_mode(idev))
2276 		return;
2277 
2278 	skb = NULL;
2279 	for_each_mc_mclock(idev, pmc) {
2280 		if (pmc->mca_sfcount[MCAST_EXCLUDE])
2281 			type = MLD2_CHANGE_TO_EXCLUDE;
2282 		else
2283 			type = MLD2_ALLOW_NEW_SOURCES;
2284 		skb = add_grec(skb, pmc, type, 0, 0, 1);
2285 	}
2286 	if (skb)
2287 		mld_sendpack(skb);
2288 }
2289 
2290 void ipv6_mc_dad_complete(struct inet6_dev *idev)
2291 {
2292 	mutex_lock(&idev->mc_lock);
2293 	idev->mc_dad_count = idev->mc_qrv;
2294 	if (idev->mc_dad_count) {
2295 		mld_send_initial_cr(idev);
2296 		idev->mc_dad_count--;
2297 		if (idev->mc_dad_count)
2298 			mld_dad_start_work(idev,
2299 					   unsolicited_report_interval(idev));
2300 	}
2301 	mutex_unlock(&idev->mc_lock);
2302 }
2303 
2304 static void mld_dad_work(struct work_struct *work)
2305 {
2306 	struct inet6_dev *idev = container_of(to_delayed_work(work),
2307 					      struct inet6_dev,
2308 					      mc_dad_work);
2309 	mutex_lock(&idev->mc_lock);
2310 	mld_send_initial_cr(idev);
2311 	if (idev->mc_dad_count) {
2312 		idev->mc_dad_count--;
2313 		if (idev->mc_dad_count)
2314 			mld_dad_start_work(idev,
2315 					   unsolicited_report_interval(idev));
2316 	}
2317 	mutex_unlock(&idev->mc_lock);
2318 	in6_dev_put(idev);
2319 }
2320 
2321 static int ip6_mc_del1_src(struct ifmcaddr6 *pmc, int sfmode,
2322 			   const struct in6_addr *psfsrc)
2323 {
2324 	struct ip6_sf_list *psf, *psf_prev;
2325 	int rv = 0;
2326 
2327 	mc_assert_locked(pmc->idev);
2328 
2329 	psf_prev = NULL;
2330 	for_each_psf_mclock(pmc, psf) {
2331 		if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2332 			break;
2333 		psf_prev = psf;
2334 	}
2335 	if (!psf || psf->sf_count[sfmode] == 0) {
2336 		/* source filter not found, or count wrong =>  bug */
2337 		return -ESRCH;
2338 	}
2339 	WRITE_ONCE(psf->sf_count[sfmode], psf->sf_count[sfmode] - 1);
2340 	if (!psf->sf_count[MCAST_INCLUDE] && !psf->sf_count[MCAST_EXCLUDE]) {
2341 		struct inet6_dev *idev = pmc->idev;
2342 
2343 		/* no more filters for this source */
2344 		if (psf_prev)
2345 			rcu_assign_pointer(psf_prev->sf_next,
2346 					   mc_dereference(psf->sf_next, idev));
2347 		else
2348 			rcu_assign_pointer(pmc->mca_sources,
2349 					   mc_dereference(psf->sf_next, idev));
2350 
2351 		if (psf->sf_oldin && !(pmc->mca_flags & MAF_NOREPORT) &&
2352 		    !mld_in_v1_mode(idev)) {
2353 			psf->sf_crcount = idev->mc_qrv;
2354 			rcu_assign_pointer(psf->sf_next,
2355 					   mc_dereference(pmc->mca_tomb, idev));
2356 			rcu_assign_pointer(pmc->mca_tomb, psf);
2357 			rv = 1;
2358 		} else {
2359 			kfree_rcu(psf, rcu);
2360 		}
2361 	}
2362 	return rv;
2363 }
2364 
2365 static int ip6_mc_del_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2366 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
2367 			  int delta)
2368 {
2369 	struct ifmcaddr6 *pmc;
2370 	int	changerec = 0;
2371 	int	i, err;
2372 
2373 	if (!idev)
2374 		return -ENODEV;
2375 
2376 	mc_assert_locked(idev);
2377 
2378 	for_each_mc_mclock(idev, pmc) {
2379 		if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2380 			break;
2381 	}
2382 	if (!pmc)
2383 		return -ESRCH;
2384 
2385 	sf_markstate(pmc);
2386 	if (!delta) {
2387 		if (!pmc->mca_sfcount[sfmode])
2388 			return -EINVAL;
2389 
2390 		pmc->mca_sfcount[sfmode]--;
2391 	}
2392 	err = 0;
2393 	for (i = 0; i < sfcount; i++) {
2394 		int rv = ip6_mc_del1_src(pmc, sfmode, &psfsrc[i]);
2395 
2396 		changerec |= rv > 0;
2397 		if (!err && rv < 0)
2398 			err = rv;
2399 	}
2400 	if (pmc->mca_sfmode == MCAST_EXCLUDE &&
2401 	    pmc->mca_sfcount[MCAST_EXCLUDE] == 0 &&
2402 	    pmc->mca_sfcount[MCAST_INCLUDE]) {
2403 		struct ip6_sf_list *psf;
2404 
2405 		/* filter mode change */
2406 		pmc->mca_sfmode = MCAST_INCLUDE;
2407 		pmc->mca_crcount = idev->mc_qrv;
2408 		idev->mc_ifc_count = pmc->mca_crcount;
2409 		for_each_psf_mclock(pmc, psf)
2410 			psf->sf_crcount = 0;
2411 		mld_ifc_event(pmc->idev);
2412 	} else if (sf_setstate(pmc) || changerec) {
2413 		mld_ifc_event(pmc->idev);
2414 	}
2415 
2416 	return err;
2417 }
2418 
2419 /* Add multicast single-source filter to the interface list */
2420 static int ip6_mc_add1_src(struct ifmcaddr6 *pmc, int sfmode,
2421 			   const struct in6_addr *psfsrc)
2422 {
2423 	struct ip6_sf_list *psf, *psf_prev;
2424 
2425 	mc_assert_locked(pmc->idev);
2426 
2427 	psf_prev = NULL;
2428 	for_each_psf_mclock(pmc, psf) {
2429 		if (ipv6_addr_equal(&psf->sf_addr, psfsrc))
2430 			break;
2431 		psf_prev = psf;
2432 	}
2433 	if (!psf) {
2434 		psf = kzalloc(sizeof(*psf), GFP_KERNEL);
2435 		if (!psf)
2436 			return -ENOBUFS;
2437 
2438 		psf->sf_addr = *psfsrc;
2439 		if (psf_prev) {
2440 			rcu_assign_pointer(psf_prev->sf_next, psf);
2441 		} else {
2442 			rcu_assign_pointer(pmc->mca_sources, psf);
2443 		}
2444 	}
2445 	WRITE_ONCE(psf->sf_count[sfmode], psf->sf_count[sfmode] + 1);
2446 	return 0;
2447 }
2448 
2449 static void sf_markstate(struct ifmcaddr6 *pmc)
2450 {
2451 	int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2452 	struct ip6_sf_list *psf;
2453 
2454 	mc_assert_locked(pmc->idev);
2455 
2456 	for_each_psf_mclock(pmc, psf) {
2457 		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2458 			psf->sf_oldin = mca_xcount ==
2459 				psf->sf_count[MCAST_EXCLUDE] &&
2460 				!psf->sf_count[MCAST_INCLUDE];
2461 		} else {
2462 			psf->sf_oldin = psf->sf_count[MCAST_INCLUDE] != 0;
2463 		}
2464 	}
2465 }
2466 
2467 static int sf_setstate(struct ifmcaddr6 *pmc)
2468 {
2469 	int mca_xcount = pmc->mca_sfcount[MCAST_EXCLUDE];
2470 	struct ip6_sf_list *psf, *dpsf;
2471 	int qrv = pmc->idev->mc_qrv;
2472 	int new_in, rv;
2473 
2474 	mc_assert_locked(pmc->idev);
2475 
2476 	rv = 0;
2477 	for_each_psf_mclock(pmc, psf) {
2478 		if (pmc->mca_sfcount[MCAST_EXCLUDE]) {
2479 			new_in = mca_xcount == psf->sf_count[MCAST_EXCLUDE] &&
2480 				!psf->sf_count[MCAST_INCLUDE];
2481 		} else
2482 			new_in = psf->sf_count[MCAST_INCLUDE] != 0;
2483 		if (new_in) {
2484 			if (!psf->sf_oldin) {
2485 				struct ip6_sf_list *prev = NULL;
2486 
2487 				for_each_psf_tomb(pmc, dpsf) {
2488 					if (ipv6_addr_equal(&dpsf->sf_addr,
2489 					    &psf->sf_addr))
2490 						break;
2491 					prev = dpsf;
2492 				}
2493 				if (dpsf) {
2494 					if (prev)
2495 						rcu_assign_pointer(prev->sf_next,
2496 								   mc_dereference(dpsf->sf_next,
2497 										  pmc->idev));
2498 					else
2499 						rcu_assign_pointer(pmc->mca_tomb,
2500 								   mc_dereference(dpsf->sf_next,
2501 										  pmc->idev));
2502 					kfree_rcu(dpsf, rcu);
2503 				}
2504 				psf->sf_crcount = qrv;
2505 				rv++;
2506 			}
2507 		} else if (psf->sf_oldin) {
2508 			psf->sf_crcount = 0;
2509 			/*
2510 			 * add or update "delete" records if an active filter
2511 			 * is now inactive
2512 			 */
2513 
2514 			for_each_psf_tomb(pmc, dpsf)
2515 				if (ipv6_addr_equal(&dpsf->sf_addr,
2516 				    &psf->sf_addr))
2517 					break;
2518 			if (!dpsf) {
2519 				dpsf = kmalloc(sizeof(*dpsf), GFP_KERNEL);
2520 				if (!dpsf)
2521 					continue;
2522 				*dpsf = *psf;
2523 				rcu_assign_pointer(dpsf->sf_next,
2524 						   mc_dereference(pmc->mca_tomb, pmc->idev));
2525 				rcu_assign_pointer(pmc->mca_tomb, dpsf);
2526 			}
2527 			dpsf->sf_crcount = qrv;
2528 			rv++;
2529 		}
2530 	}
2531 	return rv;
2532 }
2533 
2534 /* Add multicast source filter list to the interface list */
2535 static int ip6_mc_add_src(struct inet6_dev *idev, const struct in6_addr *pmca,
2536 			  int sfmode, int sfcount, const struct in6_addr *psfsrc,
2537 			  int delta)
2538 {
2539 	struct ifmcaddr6 *pmc;
2540 	int	isexclude;
2541 	int	i, err;
2542 
2543 	if (!idev)
2544 		return -ENODEV;
2545 
2546 	mc_assert_locked(idev);
2547 
2548 	for_each_mc_mclock(idev, pmc) {
2549 		if (ipv6_addr_equal(pmca, &pmc->mca_addr))
2550 			break;
2551 	}
2552 	if (!pmc)
2553 		return -ESRCH;
2554 
2555 	sf_markstate(pmc);
2556 	isexclude = pmc->mca_sfmode == MCAST_EXCLUDE;
2557 	if (!delta)
2558 		WRITE_ONCE(pmc->mca_sfcount[sfmode],
2559 			   pmc->mca_sfcount[sfmode] + 1);
2560 	err = 0;
2561 	for (i = 0; i < sfcount; i++) {
2562 		err = ip6_mc_add1_src(pmc, sfmode, &psfsrc[i]);
2563 		if (err)
2564 			break;
2565 	}
2566 	if (err) {
2567 		int j;
2568 
2569 		if (!delta)
2570 			WRITE_ONCE(pmc->mca_sfcount[sfmode],
2571 				   pmc->mca_sfcount[sfmode] - 1);
2572 		for (j = 0; j < i; j++)
2573 			ip6_mc_del1_src(pmc, sfmode, &psfsrc[j]);
2574 	} else if (isexclude != (pmc->mca_sfcount[MCAST_EXCLUDE] != 0)) {
2575 		struct ip6_sf_list *psf;
2576 
2577 		/* filter mode change */
2578 		if (pmc->mca_sfcount[MCAST_EXCLUDE])
2579 			pmc->mca_sfmode = MCAST_EXCLUDE;
2580 		else if (pmc->mca_sfcount[MCAST_INCLUDE])
2581 			pmc->mca_sfmode = MCAST_INCLUDE;
2582 		/* else no filters; keep old mode for reports */
2583 
2584 		pmc->mca_crcount = idev->mc_qrv;
2585 		idev->mc_ifc_count = pmc->mca_crcount;
2586 		for_each_psf_mclock(pmc, psf)
2587 			psf->sf_crcount = 0;
2588 		mld_ifc_event(idev);
2589 	} else if (sf_setstate(pmc)) {
2590 		mld_ifc_event(idev);
2591 	}
2592 	return err;
2593 }
2594 
2595 static void ip6_mc_clear_src(struct ifmcaddr6 *pmc)
2596 {
2597 	struct ip6_sf_list *psf, *nextpsf;
2598 
2599 	mc_assert_locked(pmc->idev);
2600 
2601 	for (psf = mc_dereference(pmc->mca_tomb, pmc->idev);
2602 	     psf;
2603 	     psf = nextpsf) {
2604 		nextpsf = mc_dereference(psf->sf_next, pmc->idev);
2605 		kfree_rcu(psf, rcu);
2606 	}
2607 	RCU_INIT_POINTER(pmc->mca_tomb, NULL);
2608 	for (psf = mc_dereference(pmc->mca_sources, pmc->idev);
2609 	     psf;
2610 	     psf = nextpsf) {
2611 		nextpsf = mc_dereference(psf->sf_next, pmc->idev);
2612 		kfree_rcu(psf, rcu);
2613 	}
2614 	RCU_INIT_POINTER(pmc->mca_sources, NULL);
2615 	pmc->mca_sfmode = MCAST_EXCLUDE;
2616 	pmc->mca_sfcount[MCAST_INCLUDE] = 0;
2617 	/* Paired with the READ_ONCE() from ipv6_chk_mcast_addr() */
2618 	WRITE_ONCE(pmc->mca_sfcount[MCAST_EXCLUDE], 1);
2619 }
2620 
2621 static void igmp6_join_group(struct ifmcaddr6 *ma)
2622 {
2623 	unsigned long delay;
2624 
2625 	mc_assert_locked(ma->idev);
2626 
2627 	if (ma->mca_flags & MAF_NOREPORT)
2628 		return;
2629 
2630 	igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2631 
2632 	delay = get_random_u32_below(unsolicited_report_interval(ma->idev));
2633 
2634 	if (cancel_delayed_work(&ma->mca_work)) {
2635 		refcount_dec(&ma->mca_refcnt);
2636 		delay = ma->mca_work.timer.expires - jiffies;
2637 	}
2638 
2639 	if (!mod_delayed_work(mld_wq, &ma->mca_work, delay))
2640 		refcount_inc(&ma->mca_refcnt);
2641 	ma->mca_flags |= MAF_TIMER_RUNNING | MAF_LAST_REPORTER;
2642 }
2643 
2644 static int ip6_mc_leave_src(struct sock *sk, struct ipv6_mc_socklist *iml,
2645 			    struct inet6_dev *idev)
2646 {
2647 	struct ip6_sf_socklist *psl;
2648 	int err;
2649 
2650 	psl = sock_dereference(iml->sflist, sk);
2651 
2652 	if (idev)
2653 		mutex_lock(&idev->mc_lock);
2654 
2655 	if (!psl) {
2656 		/* any-source empty exclude case */
2657 		err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode, 0, NULL, 0);
2658 	} else {
2659 		err = ip6_mc_del_src(idev, &iml->addr, iml->sfmode,
2660 				     psl->sl_count, psl->sl_addr, 0);
2661 		RCU_INIT_POINTER(iml->sflist, NULL);
2662 		atomic_sub(struct_size(psl, sl_addr, psl->sl_max),
2663 			   &sk->sk_omem_alloc);
2664 		kfree_rcu(psl, rcu);
2665 	}
2666 
2667 	if (idev)
2668 		mutex_unlock(&idev->mc_lock);
2669 
2670 	return err;
2671 }
2672 
2673 static void igmp6_leave_group(struct ifmcaddr6 *ma)
2674 {
2675 	mc_assert_locked(ma->idev);
2676 
2677 	if (mld_in_v1_mode(ma->idev)) {
2678 		if (ma->mca_flags & MAF_LAST_REPORTER) {
2679 			igmp6_send(&ma->mca_addr, ma->idev->dev,
2680 				ICMPV6_MGM_REDUCTION);
2681 		}
2682 	} else {
2683 		mld_add_delrec(ma->idev, ma);
2684 		mld_ifc_event(ma->idev);
2685 	}
2686 }
2687 
2688 static void mld_gq_work(struct work_struct *work)
2689 {
2690 	struct inet6_dev *idev = container_of(to_delayed_work(work),
2691 					      struct inet6_dev,
2692 					      mc_gq_work);
2693 
2694 	mutex_lock(&idev->mc_lock);
2695 	mld_send_report(idev, NULL);
2696 	idev->mc_gq_running = 0;
2697 	mutex_unlock(&idev->mc_lock);
2698 
2699 	in6_dev_put(idev);
2700 }
2701 
2702 static void mld_ifc_work(struct work_struct *work)
2703 {
2704 	struct inet6_dev *idev = container_of(to_delayed_work(work),
2705 					      struct inet6_dev,
2706 					      mc_ifc_work);
2707 
2708 	mutex_lock(&idev->mc_lock);
2709 	mld_send_cr(idev);
2710 
2711 	if (idev->mc_ifc_count) {
2712 		idev->mc_ifc_count--;
2713 		if (idev->mc_ifc_count)
2714 			mld_ifc_start_work(idev,
2715 					   unsolicited_report_interval(idev));
2716 	}
2717 	mutex_unlock(&idev->mc_lock);
2718 	in6_dev_put(idev);
2719 }
2720 
2721 static void mld_ifc_event(struct inet6_dev *idev)
2722 {
2723 	mc_assert_locked(idev);
2724 
2725 	if (mld_in_v1_mode(idev))
2726 		return;
2727 
2728 	idev->mc_ifc_count = idev->mc_qrv;
2729 	mld_ifc_start_work(idev, 1);
2730 }
2731 
2732 static void mld_mca_work(struct work_struct *work)
2733 {
2734 	struct ifmcaddr6 *ma = container_of(to_delayed_work(work),
2735 					    struct ifmcaddr6, mca_work);
2736 
2737 	mutex_lock(&ma->idev->mc_lock);
2738 	if (mld_in_v1_mode(ma->idev))
2739 		igmp6_send(&ma->mca_addr, ma->idev->dev, ICMPV6_MGM_REPORT);
2740 	else
2741 		mld_send_report(ma->idev, ma);
2742 	ma->mca_flags |=  MAF_LAST_REPORTER;
2743 	ma->mca_flags &= ~MAF_TIMER_RUNNING;
2744 	mutex_unlock(&ma->idev->mc_lock);
2745 
2746 	ma_put(ma);
2747 }
2748 
2749 /* Device changing type */
2750 
2751 void ipv6_mc_unmap(struct inet6_dev *idev)
2752 {
2753 	struct ifmcaddr6 *i;
2754 
2755 	/* Install multicast list, except for all-nodes (already installed) */
2756 
2757 	mutex_lock(&idev->mc_lock);
2758 	for_each_mc_mclock(idev, i)
2759 		igmp6_group_dropped(i);
2760 	mutex_unlock(&idev->mc_lock);
2761 }
2762 
2763 void ipv6_mc_remap(struct inet6_dev *idev)
2764 {
2765 	ipv6_mc_up(idev);
2766 }
2767 
2768 /* Device going down */
2769 void ipv6_mc_down(struct inet6_dev *idev)
2770 {
2771 	struct ifmcaddr6 *i;
2772 
2773 	mutex_lock(&idev->mc_lock);
2774 	/* Withdraw multicast list */
2775 	for_each_mc_mclock(idev, i)
2776 		igmp6_group_dropped(i);
2777 	mutex_unlock(&idev->mc_lock);
2778 
2779 	/* Should stop work after group drop. or we will
2780 	 * start work again in mld_ifc_event()
2781 	 */
2782 	mld_query_stop_work(idev);
2783 	mld_report_stop_work(idev);
2784 
2785 	mutex_lock(&idev->mc_lock);
2786 	mld_ifc_stop_work(idev);
2787 	mld_gq_stop_work(idev);
2788 	mutex_unlock(&idev->mc_lock);
2789 
2790 	mld_dad_stop_work(idev);
2791 }
2792 
2793 static void ipv6_mc_reset(struct inet6_dev *idev)
2794 {
2795 	idev->mc_qrv = sysctl_mld_qrv;
2796 	idev->mc_qi = MLD_QI_DEFAULT;
2797 	idev->mc_qri = MLD_QRI_DEFAULT;
2798 	idev->mc_v1_seen = 0;
2799 	idev->mc_maxdelay = unsolicited_report_interval(idev);
2800 }
2801 
2802 /* Device going up */
2803 
2804 void ipv6_mc_up(struct inet6_dev *idev)
2805 {
2806 	struct ifmcaddr6 *i;
2807 
2808 	/* Install multicast list, except for all-nodes (already installed) */
2809 
2810 	ipv6_mc_reset(idev);
2811 	mutex_lock(&idev->mc_lock);
2812 	for_each_mc_mclock(idev, i) {
2813 		mld_del_delrec(idev, i);
2814 		igmp6_group_added(i);
2815 	}
2816 	mutex_unlock(&idev->mc_lock);
2817 }
2818 
2819 /* IPv6 device initialization. */
2820 
2821 void ipv6_mc_init_dev(struct inet6_dev *idev)
2822 {
2823 	idev->mc_gq_running = 0;
2824 	INIT_DELAYED_WORK(&idev->mc_gq_work, mld_gq_work);
2825 	RCU_INIT_POINTER(idev->mc_tomb, NULL);
2826 	idev->mc_ifc_count = 0;
2827 	INIT_DELAYED_WORK(&idev->mc_ifc_work, mld_ifc_work);
2828 	INIT_DELAYED_WORK(&idev->mc_dad_work, mld_dad_work);
2829 	INIT_DELAYED_WORK(&idev->mc_query_work, mld_query_work);
2830 	INIT_DELAYED_WORK(&idev->mc_report_work, mld_report_work);
2831 	skb_queue_head_init(&idev->mc_query_queue);
2832 	skb_queue_head_init(&idev->mc_report_queue);
2833 	spin_lock_init(&idev->mc_query_lock);
2834 	spin_lock_init(&idev->mc_report_lock);
2835 	mutex_init(&idev->mc_lock);
2836 	ipv6_mc_reset(idev);
2837 }
2838 
2839 /*
2840  *	Device is about to be destroyed: clean up.
2841  */
2842 
2843 void ipv6_mc_destroy_dev(struct inet6_dev *idev)
2844 {
2845 	struct ifmcaddr6 *i;
2846 
2847 	/* Deactivate works */
2848 	ipv6_mc_down(idev);
2849 	mutex_lock(&idev->mc_lock);
2850 	mld_clear_delrec(idev);
2851 	mutex_unlock(&idev->mc_lock);
2852 	mld_clear_query(idev);
2853 	mld_clear_report(idev);
2854 
2855 	/* Delete all-nodes address. */
2856 	/* We cannot call ipv6_dev_mc_dec() directly, our caller in
2857 	 * addrconf.c has NULL'd out dev->ip6_ptr so in6_dev_get() will
2858 	 * fail.
2859 	 */
2860 	__ipv6_dev_mc_dec(idev, &in6addr_linklocal_allnodes);
2861 
2862 	if (idev->cnf.forwarding)
2863 		__ipv6_dev_mc_dec(idev, &in6addr_linklocal_allrouters);
2864 
2865 	mutex_lock(&idev->mc_lock);
2866 	while ((i = mc_dereference(idev->mc_list, idev))) {
2867 		rcu_assign_pointer(idev->mc_list, mc_dereference(i->next, idev));
2868 
2869 		ip6_mc_clear_src(i);
2870 		ma_put(i);
2871 	}
2872 	mutex_unlock(&idev->mc_lock);
2873 }
2874 
2875 static void ipv6_mc_rejoin_groups(struct inet6_dev *idev)
2876 {
2877 	struct ifmcaddr6 *pmc;
2878 
2879 	mutex_lock(&idev->mc_lock);
2880 	if (mld_in_v1_mode(idev)) {
2881 		for_each_mc_mclock(idev, pmc)
2882 			igmp6_join_group(pmc);
2883 	} else {
2884 		mld_send_report(idev, NULL);
2885 	}
2886 	mutex_unlock(&idev->mc_lock);
2887 }
2888 
2889 static int ipv6_mc_netdev_event(struct notifier_block *this,
2890 				unsigned long event,
2891 				void *ptr)
2892 {
2893 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2894 	struct inet6_dev *idev = __in6_dev_get(dev);
2895 
2896 	switch (event) {
2897 	case NETDEV_RESEND_IGMP:
2898 		if (idev)
2899 			ipv6_mc_rejoin_groups(idev);
2900 		break;
2901 	default:
2902 		break;
2903 	}
2904 
2905 	return NOTIFY_DONE;
2906 }
2907 
2908 static struct notifier_block igmp6_netdev_notifier = {
2909 	.notifier_call = ipv6_mc_netdev_event,
2910 };
2911 
2912 #ifdef CONFIG_PROC_FS
2913 struct igmp6_mc_iter_state {
2914 	struct seq_net_private p;
2915 	struct net_device *dev;
2916 	struct inet6_dev *idev;
2917 };
2918 
2919 #define igmp6_mc_seq_private(seq)	((struct igmp6_mc_iter_state *)(seq)->private)
2920 
2921 static inline struct ifmcaddr6 *igmp6_mc_get_first(struct seq_file *seq)
2922 {
2923 	struct ifmcaddr6 *im = NULL;
2924 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2925 	struct net *net = seq_file_net(seq);
2926 
2927 	state->idev = NULL;
2928 	for_each_netdev_rcu(net, state->dev) {
2929 		struct inet6_dev *idev;
2930 		idev = __in6_dev_get(state->dev);
2931 		if (!idev)
2932 			continue;
2933 
2934 		im = rcu_dereference(idev->mc_list);
2935 		if (im) {
2936 			state->idev = idev;
2937 			break;
2938 		}
2939 	}
2940 	return im;
2941 }
2942 
2943 static struct ifmcaddr6 *igmp6_mc_get_next(struct seq_file *seq, struct ifmcaddr6 *im)
2944 {
2945 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2946 
2947 	im = rcu_dereference(im->next);
2948 	while (!im) {
2949 		state->dev = next_net_device_rcu(state->dev);
2950 		if (!state->dev) {
2951 			state->idev = NULL;
2952 			break;
2953 		}
2954 		state->idev = __in6_dev_get(state->dev);
2955 		if (!state->idev)
2956 			continue;
2957 		im = rcu_dereference(state->idev->mc_list);
2958 	}
2959 	return im;
2960 }
2961 
2962 static struct ifmcaddr6 *igmp6_mc_get_idx(struct seq_file *seq, loff_t pos)
2963 {
2964 	struct ifmcaddr6 *im = igmp6_mc_get_first(seq);
2965 	if (im)
2966 		while (pos && (im = igmp6_mc_get_next(seq, im)) != NULL)
2967 			--pos;
2968 	return pos ? NULL : im;
2969 }
2970 
2971 static void *igmp6_mc_seq_start(struct seq_file *seq, loff_t *pos)
2972 	__acquires(RCU)
2973 {
2974 	rcu_read_lock();
2975 	return igmp6_mc_get_idx(seq, *pos);
2976 }
2977 
2978 static void *igmp6_mc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2979 {
2980 	struct ifmcaddr6 *im = igmp6_mc_get_next(seq, v);
2981 
2982 	++*pos;
2983 	return im;
2984 }
2985 
2986 static void igmp6_mc_seq_stop(struct seq_file *seq, void *v)
2987 	__releases(RCU)
2988 {
2989 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
2990 
2991 	if (likely(state->idev))
2992 		state->idev = NULL;
2993 	state->dev = NULL;
2994 	rcu_read_unlock();
2995 }
2996 
2997 static int igmp6_mc_seq_show(struct seq_file *seq, void *v)
2998 {
2999 	struct ifmcaddr6 *im = (struct ifmcaddr6 *)v;
3000 	struct igmp6_mc_iter_state *state = igmp6_mc_seq_private(seq);
3001 
3002 	seq_printf(seq,
3003 		   "%-4d %-15s %pi6 %5d %08X %ld\n",
3004 		   state->dev->ifindex, state->dev->name,
3005 		   &im->mca_addr,
3006 		   im->mca_users, im->mca_flags,
3007 		   (im->mca_flags & MAF_TIMER_RUNNING) ?
3008 		   jiffies_to_clock_t(im->mca_work.timer.expires - jiffies) : 0);
3009 	return 0;
3010 }
3011 
3012 static const struct seq_operations igmp6_mc_seq_ops = {
3013 	.start	=	igmp6_mc_seq_start,
3014 	.next	=	igmp6_mc_seq_next,
3015 	.stop	=	igmp6_mc_seq_stop,
3016 	.show	=	igmp6_mc_seq_show,
3017 };
3018 
3019 struct igmp6_mcf_iter_state {
3020 	struct seq_net_private p;
3021 	struct net_device *dev;
3022 	struct inet6_dev *idev;
3023 	struct ifmcaddr6 *im;
3024 };
3025 
3026 #define igmp6_mcf_seq_private(seq)	((struct igmp6_mcf_iter_state *)(seq)->private)
3027 
3028 static inline struct ip6_sf_list *igmp6_mcf_get_first(struct seq_file *seq)
3029 {
3030 	struct ip6_sf_list *psf = NULL;
3031 	struct ifmcaddr6 *im = NULL;
3032 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3033 	struct net *net = seq_file_net(seq);
3034 
3035 	state->idev = NULL;
3036 	state->im = NULL;
3037 	for_each_netdev_rcu(net, state->dev) {
3038 		struct inet6_dev *idev;
3039 		idev = __in6_dev_get(state->dev);
3040 		if (unlikely(idev == NULL))
3041 			continue;
3042 
3043 		im = rcu_dereference(idev->mc_list);
3044 		if (likely(im)) {
3045 			psf = rcu_dereference(im->mca_sources);
3046 			if (likely(psf)) {
3047 				state->im = im;
3048 				state->idev = idev;
3049 				break;
3050 			}
3051 		}
3052 	}
3053 	return psf;
3054 }
3055 
3056 static struct ip6_sf_list *igmp6_mcf_get_next(struct seq_file *seq, struct ip6_sf_list *psf)
3057 {
3058 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3059 
3060 	psf = rcu_dereference(psf->sf_next);
3061 	while (!psf) {
3062 		state->im = rcu_dereference(state->im->next);
3063 		while (!state->im) {
3064 			state->dev = next_net_device_rcu(state->dev);
3065 			if (!state->dev) {
3066 				state->idev = NULL;
3067 				goto out;
3068 			}
3069 			state->idev = __in6_dev_get(state->dev);
3070 			if (!state->idev)
3071 				continue;
3072 			state->im = rcu_dereference(state->idev->mc_list);
3073 		}
3074 		psf = rcu_dereference(state->im->mca_sources);
3075 	}
3076 out:
3077 	return psf;
3078 }
3079 
3080 static struct ip6_sf_list *igmp6_mcf_get_idx(struct seq_file *seq, loff_t pos)
3081 {
3082 	struct ip6_sf_list *psf = igmp6_mcf_get_first(seq);
3083 	if (psf)
3084 		while (pos && (psf = igmp6_mcf_get_next(seq, psf)) != NULL)
3085 			--pos;
3086 	return pos ? NULL : psf;
3087 }
3088 
3089 static void *igmp6_mcf_seq_start(struct seq_file *seq, loff_t *pos)
3090 	__acquires(RCU)
3091 {
3092 	rcu_read_lock();
3093 	return *pos ? igmp6_mcf_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
3094 }
3095 
3096 static void *igmp6_mcf_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3097 {
3098 	struct ip6_sf_list *psf;
3099 	if (v == SEQ_START_TOKEN)
3100 		psf = igmp6_mcf_get_first(seq);
3101 	else
3102 		psf = igmp6_mcf_get_next(seq, v);
3103 	++*pos;
3104 	return psf;
3105 }
3106 
3107 static void igmp6_mcf_seq_stop(struct seq_file *seq, void *v)
3108 	__releases(RCU)
3109 {
3110 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3111 
3112 	if (likely(state->im))
3113 		state->im = NULL;
3114 	if (likely(state->idev))
3115 		state->idev = NULL;
3116 
3117 	state->dev = NULL;
3118 	rcu_read_unlock();
3119 }
3120 
3121 static int igmp6_mcf_seq_show(struct seq_file *seq, void *v)
3122 {
3123 	struct ip6_sf_list *psf = (struct ip6_sf_list *)v;
3124 	struct igmp6_mcf_iter_state *state = igmp6_mcf_seq_private(seq);
3125 
3126 	if (v == SEQ_START_TOKEN) {
3127 		seq_puts(seq, "Idx Device                Multicast Address                   Source Address    INC    EXC\n");
3128 	} else {
3129 		seq_printf(seq,
3130 			   "%3d %6.6s %pi6 %pi6 %6lu %6lu\n",
3131 			   state->dev->ifindex, state->dev->name,
3132 			   &state->im->mca_addr,
3133 			   &psf->sf_addr,
3134 			   READ_ONCE(psf->sf_count[MCAST_INCLUDE]),
3135 			   READ_ONCE(psf->sf_count[MCAST_EXCLUDE]));
3136 	}
3137 	return 0;
3138 }
3139 
3140 static const struct seq_operations igmp6_mcf_seq_ops = {
3141 	.start	=	igmp6_mcf_seq_start,
3142 	.next	=	igmp6_mcf_seq_next,
3143 	.stop	=	igmp6_mcf_seq_stop,
3144 	.show	=	igmp6_mcf_seq_show,
3145 };
3146 
3147 static int __net_init igmp6_proc_init(struct net *net)
3148 {
3149 	int err;
3150 
3151 	err = -ENOMEM;
3152 	if (!proc_create_net("igmp6", 0444, net->proc_net, &igmp6_mc_seq_ops,
3153 			sizeof(struct igmp6_mc_iter_state)))
3154 		goto out;
3155 	if (!proc_create_net("mcfilter6", 0444, net->proc_net,
3156 			&igmp6_mcf_seq_ops,
3157 			sizeof(struct igmp6_mcf_iter_state)))
3158 		goto out_proc_net_igmp6;
3159 
3160 	err = 0;
3161 out:
3162 	return err;
3163 
3164 out_proc_net_igmp6:
3165 	remove_proc_entry("igmp6", net->proc_net);
3166 	goto out;
3167 }
3168 
3169 static void __net_exit igmp6_proc_exit(struct net *net)
3170 {
3171 	remove_proc_entry("mcfilter6", net->proc_net);
3172 	remove_proc_entry("igmp6", net->proc_net);
3173 }
3174 #else
3175 static inline int igmp6_proc_init(struct net *net)
3176 {
3177 	return 0;
3178 }
3179 static inline void igmp6_proc_exit(struct net *net)
3180 {
3181 }
3182 #endif
3183 
3184 static int __net_init igmp6_net_init(struct net *net)
3185 {
3186 	int err;
3187 
3188 	err = inet_ctl_sock_create(&net->ipv6.igmp_sk, PF_INET6,
3189 				   SOCK_RAW, IPPROTO_ICMPV6, net);
3190 	if (err < 0) {
3191 		pr_err("Failed to initialize the IGMP6 control socket (err %d)\n",
3192 		       err);
3193 		goto out;
3194 	}
3195 
3196 	inet6_sk(net->ipv6.igmp_sk)->hop_limit = 1;
3197 	net->ipv6.igmp_sk->sk_allocation = GFP_KERNEL;
3198 
3199 	err = inet_ctl_sock_create(&net->ipv6.mc_autojoin_sk, PF_INET6,
3200 				   SOCK_RAW, IPPROTO_ICMPV6, net);
3201 	if (err < 0) {
3202 		pr_err("Failed to initialize the IGMP6 autojoin socket (err %d)\n",
3203 		       err);
3204 		goto out_sock_create;
3205 	}
3206 
3207 	err = igmp6_proc_init(net);
3208 	if (err)
3209 		goto out_sock_create_autojoin;
3210 
3211 	return 0;
3212 
3213 out_sock_create_autojoin:
3214 	inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
3215 out_sock_create:
3216 	inet_ctl_sock_destroy(net->ipv6.igmp_sk);
3217 out:
3218 	return err;
3219 }
3220 
3221 static void __net_exit igmp6_net_exit(struct net *net)
3222 {
3223 	inet_ctl_sock_destroy(net->ipv6.igmp_sk);
3224 	inet_ctl_sock_destroy(net->ipv6.mc_autojoin_sk);
3225 	igmp6_proc_exit(net);
3226 }
3227 
3228 static struct pernet_operations igmp6_net_ops = {
3229 	.init = igmp6_net_init,
3230 	.exit = igmp6_net_exit,
3231 };
3232 
3233 int __init igmp6_init(void)
3234 {
3235 	int err;
3236 
3237 	err = register_pernet_subsys(&igmp6_net_ops);
3238 	if (err)
3239 		return err;
3240 
3241 	mld_wq = create_workqueue("mld");
3242 	if (!mld_wq) {
3243 		unregister_pernet_subsys(&igmp6_net_ops);
3244 		return -ENOMEM;
3245 	}
3246 
3247 	return err;
3248 }
3249 
3250 int __init igmp6_late_init(void)
3251 {
3252 	return register_netdevice_notifier(&igmp6_netdev_notifier);
3253 }
3254 
3255 void igmp6_cleanup(void)
3256 {
3257 	unregister_pernet_subsys(&igmp6_net_ops);
3258 	destroy_workqueue(mld_wq);
3259 }
3260 
3261 void igmp6_late_cleanup(void)
3262 {
3263 	unregister_netdevice_notifier(&igmp6_netdev_notifier);
3264 }
3265